From 75bbe9b98c06b4d769c027bc275ac7e19e88d22d Mon Sep 17 00:00:00 2001 From: dlyakhov Date: Wed, 23 Oct 2024 14:54:41 +0200 Subject: [PATCH 01/15] WIP experimental quantization --- .../algorithms/post_training/__init__.py | 10 + .../algorithms/post_training/algorithm.py | 103 ++++ .../algorithms/post_training/pipeline.py | 139 +++++ .../algorithms/quantizer/fx_quantizer.py | 110 ++++ .../algorithms/quantizer/quantizer.py | 26 + .../algorithms/range_estimator/backend.py | 154 ++++++ .../range_estimator/range_estimator.py | 495 ++++++++++++++++++ .../range_estimator/torch_fx_backend.py | 221 ++++++++ .../torch/fx/quantization/quantize_pt2e.py | 91 ++++ torch_fx_experimental_q.py | 107 ++++ 10 files changed, 1456 insertions(+) create mode 100644 nncf/experimental/common/quantization/algorithms/post_training/__init__.py create mode 100644 nncf/experimental/common/quantization/algorithms/post_training/algorithm.py create mode 100644 nncf/experimental/common/quantization/algorithms/post_training/pipeline.py create mode 100644 nncf/experimental/common/quantization/algorithms/quantizer/fx_quantizer.py create mode 100644 nncf/experimental/common/quantization/algorithms/quantizer/quantizer.py create mode 100644 nncf/experimental/common/quantization/algorithms/range_estimator/backend.py create mode 100644 nncf/experimental/common/quantization/algorithms/range_estimator/range_estimator.py create mode 100644 nncf/experimental/common/quantization/algorithms/range_estimator/torch_fx_backend.py create mode 100644 nncf/experimental/torch/fx/quantization/quantize_pt2e.py create mode 100644 torch_fx_experimental_q.py diff --git a/nncf/experimental/common/quantization/algorithms/post_training/__init__.py b/nncf/experimental/common/quantization/algorithms/post_training/__init__.py new file mode 100644 index 00000000000..2e49d63977d --- /dev/null +++ b/nncf/experimental/common/quantization/algorithms/post_training/__init__.py @@ -0,0 +1,10 @@ +# Copyright (c) 2024 Intel Corporation +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# http://www.apache.org/licenses/LICENSE-2.0 +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. diff --git a/nncf/experimental/common/quantization/algorithms/post_training/algorithm.py b/nncf/experimental/common/quantization/algorithms/post_training/algorithm.py new file mode 100644 index 00000000000..0f7a90d3d67 --- /dev/null +++ b/nncf/experimental/common/quantization/algorithms/post_training/algorithm.py @@ -0,0 +1,103 @@ +# Copyright (c) 2024 Intel Corporation +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# http://www.apache.org/licenses/LICENSE-2.0 +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import itertools +from typing import Callable, List, Optional, TypeVar + +from nncf import Dataset +from nncf.common.graph.graph import NNCFGraph +from nncf.common.tensor_statistics.statistic_point import StatisticPointsContainer +from nncf.common.utils.backend import BackendType +from nncf.experimental.common.quantization.algorithms.post_training.pipeline import create_ptq_pipeline +from nncf.experimental.common.quantization.algorithms.quantizer.quantizer import NNCFQuantizer +from nncf.parameters import ModelType +from nncf.quantization.advanced_parameters import AdvancedQuantizationParameters +from nncf.quantization.algorithms.algorithm import Algorithm + +TModel = TypeVar("TModel") +TPass = Callable[[TModel], TModel] + + +class PostTrainingQuantization(Algorithm): + """ + Implements Post-Training Quantization algorithm, which basically includes: + 1) ChannelAlignment + 2) MinMaxQuantization + 3) FastBiasCorrection or BiasCorrection + """ + + def __init__( + self, + quantizer: NNCFQuantizer, + subset_size: int = 300, + fast_bias_correction: bool = True, + model_type: Optional[ModelType] = None, + advanced_parameters: Optional[AdvancedQuantizationParameters] = None, + ): + """ + :param mode: Special quantization mode that specify different ways of the optimization. + :param preset: A preset controls the quantization mode (symmetric and asymmetric). + It can take the following values: + - `performance`: Symmetric quantization of weights and activations. + - `mixed`: Symmetric quantization of weights and asymmetric quantization of activations. + Default value is None. In this case, `mixed` preset is used for `transformer` + model type otherwise `performace`. + :param target_device: A target device the specificity of which will be taken + into account while compressing in order to obtain the best performance + for this type of device. + :param subset_size: Size of a subset to calculate activations + statistics used for quantization. + :param fast_bias_correction: Setting this option to `False` enables a different + bias correction method which is more accurate, in general, and takes + more time but requires less memory. + :param model_type: Model type is needed to specify additional patterns + in the model. Supported only `transformer` now. + :param ignored_scope: An ignored scope that defined the list of model control + flow graph nodes to be ignored during quantization. + :param advanced_parameters: Advanced quantization parameters for + fine-tuning the quantization algorithm + """ + self._pipeline = create_ptq_pipeline( + quantizer=quantizer, + subset_size=subset_size, + fast_bias_correction=fast_bias_correction, + model_type=model_type, + advanced_parameters=advanced_parameters, + ) + + @property + def available_backends(self) -> List[BackendType]: + backends = set(BackendType) + for algorithm in itertools.chain.from_iterable(self._pipeline.pipeline_steps): + backends = backends.intersection(algorithm.available_backends) + return list(backends) + + def get_statistic_points(self, model: TModel, graph: NNCFGraph) -> StatisticPointsContainer: + return self._pipeline.get_statistic_points_for_step(0, model, graph) + + def apply( + self, + model: TModel, + graph: NNCFGraph, + statistic_points: Optional[StatisticPointsContainer] = None, + dataset: Optional[Dataset] = None, + ) -> TModel: + if dataset is None and len(self._pipeline.pipeline_steps) > 1: + raise ValueError( + "A dataset is required for the post-training quantization " + "algorithm to collect statistics for intermediate models." + ) + + step_index_to_statistics = None + if statistic_points: + step_index_to_statistics = {0: statistic_points} + + return self._pipeline.run_from_step(model, dataset, graph, 0, step_index_to_statistics) diff --git a/nncf/experimental/common/quantization/algorithms/post_training/pipeline.py b/nncf/experimental/common/quantization/algorithms/post_training/pipeline.py new file mode 100644 index 00000000000..bb1c0ec0bba --- /dev/null +++ b/nncf/experimental/common/quantization/algorithms/post_training/pipeline.py @@ -0,0 +1,139 @@ +# Copyright (c) 2024 Intel Corporation +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# http://www.apache.org/licenses/LICENSE-2.0 +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Optional, TypeVar + +from nncf.common.deprecation import warning_deprecated +from nncf.experimental.common.quantization.algorithms.quantizer.quantizer import NNCFQuantizer +from nncf.experimental.common.quantization.algorithms.range_estimator.range_estimator import MinMaxRangeEstimator +from nncf.parameters import ModelType +from nncf.quantization.advanced_parameters import AdvancedQuantizationParameters +from nncf.quantization.algorithms.bias_correction.algorithm import BIAS_CORRECTION_THRESHOLD +from nncf.quantization.algorithms.bias_correction.algorithm import BiasCorrection +from nncf.quantization.algorithms.channel_alignment.algorithm import ChannelAlignment +from nncf.quantization.algorithms.fast_bias_correction.algorithm import FAST_BIAS_CORRECTION_THRESHOLD +from nncf.quantization.algorithms.fast_bias_correction.algorithm import FastBiasCorrection +from nncf.quantization.algorithms.pipeline import Pipeline +from nncf.quantization.algorithms.smooth_quant.algorithm import SmoothQuant + +TModel = TypeVar("TModel") + + +def create_ptq_pipeline( + quantizer: NNCFQuantizer, + subset_size: int = 300, + fast_bias_correction: bool = True, + model_type: Optional[ModelType] = None, + advanced_parameters: Optional[AdvancedQuantizationParameters] = None, +) -> Pipeline: + """ + Creates a post-training quantization pipeline. + + The post-training quantization pipeline includes the following steps: + 1) SmoothQuant + 2) ChannelAlignment + 3) MinMaxQuantization + 4) FastBiasCorrection or BiasCorrection + + :param mode: Special quantization mode that specify different ways of the optimization. + :param preset: A preset controls the quantization mode (symmetric and asymmetric). + It can take the following values: + - `performance`: Symmetric quantization of weights and activations. + - `mixed`: Symmetric quantization of weights and asymmetric quantization of activations. + Default value is None. In this case, `mixed` preset is used for `transformer` + model type otherwise `performace`. + :param target_device: A target device the specificity of which will be taken + into account while compressing in order to obtain the best performance + for this type of device. + :param subset_size: Size of a subset to calculate activations + statistics used for quantization. + :param fast_bias_correction: Setting this option to `False` enables a different + bias correction method which is more accurate, in general, and takes + more time but requires less memory. + :param model_type: Model type is needed to specify additional patterns + in the model. Supported only `transformer` now. + :param advanced_parameters: Advanced quantization parameters for + fine-tuning the quantization algorithm + :return: A post-training quantization pipeline. + """ + + if advanced_parameters is None: + advanced_parameters = AdvancedQuantizationParameters() + + # Build the post-training quantization pipeline. + pipeline_steps = [] + + # Add the `SmoothQuant` algorithm as the first step of the pipeline. + # It is added only for `ModelType.TRANSFORMER`. + sq_params = advanced_parameters.smooth_quant_alphas + sq_alpha = advanced_parameters.smooth_quant_alpha + if sq_alpha is not None: + warning_deprecated( + "`AdvancedQuantizationParameters(smooth_quant_alpha=..)` is deprecated." + "Please, use `AdvancedQuantizationParameters(smooth_quant_alphas)` option " + "with AdvancedSmoothQuantParameters(convolution=.., matmul=..) as value instead." + ) + if sq_alpha < 0: + sq_params.convolution = -1 + sq_params.matmul = -1 + else: + sq_params.matmul = sq_alpha + + if model_type == ModelType.TRANSFORMER and (sq_params.convolution >= 0 or sq_params.matmul >= 0): + alpha_map = {"convolution": sq_params.convolution, "matmul": sq_params.matmul} + pipeline_steps.append([SmoothQuant(subset_size, advanced_parameters.inplace_statistics, alpha_map=alpha_map)]) + + # Add the `ChannelAlignment` algorithm as the second step of the pipeline. + if not advanced_parameters.disable_channel_alignment: + pipeline_steps.append([ChannelAlignment(subset_size, advanced_parameters.inplace_statistics)]) + + # Add the `MinMaxQuantization` algorithm as the third step of the pipeline. + pipeline_steps.append( + [ + MinMaxRangeEstimator( + quantizer=quantizer, + subset_size=subset_size, + inplace_statistics=advanced_parameters.inplace_statistics, + batchwise_statistics=advanced_parameters.batchwise_statistics, + activations_range_estimator_params=advanced_parameters.activations_range_estimator_params, + weights_range_estimator_params=advanced_parameters.weights_range_estimator_params, + ) + ] + ) + + if not advanced_parameters.disable_bias_correction: + # Add the `FastBiasCorrection` or `BiasCorrection` as additional algorithm + # inside the third step of the pipeline. It is added after `MinMaxQuantization` + # algorithm. + bias_correction_params = advanced_parameters.bias_correction_params + if fast_bias_correction: + threshold = FAST_BIAS_CORRECTION_THRESHOLD + bias_correction_subset_size = subset_size + bias_correction_cls = FastBiasCorrection + else: + threshold = BIAS_CORRECTION_THRESHOLD + bias_correction_subset_size = max(int(subset_size * 0.2), 1) + bias_correction_cls = BiasCorrection + + if bias_correction_params.threshold is not None: + threshold = bias_correction_params.threshold + + pipeline_steps[-1].append( + bias_correction_cls( + bias_correction_subset_size, + threshold, + bias_correction_params.apply_for_all_nodes, + advanced_parameters.inplace_statistics, + advanced_parameters.backend_params, + ) + ) + + return Pipeline(pipeline_steps) diff --git a/nncf/experimental/common/quantization/algorithms/quantizer/fx_quantizer.py b/nncf/experimental/common/quantization/algorithms/quantizer/fx_quantizer.py new file mode 100644 index 00000000000..e7d80fbaff4 --- /dev/null +++ b/nncf/experimental/common/quantization/algorithms/quantizer/fx_quantizer.py @@ -0,0 +1,110 @@ +# Copyright (c) 2024 Intel Corporation +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# http://www.apache.org/licenses/LICENSE-2.0 +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + + +from collections import defaultdict +from copy import deepcopy + +import torch +import torch.fx +from torch.ao.quantization.pt2e.prepare import _get_edge_or_node_to_group_id +from torch.ao.quantization.pt2e.prepare import _get_edge_or_node_to_qspec +from torch.ao.quantization.pt2e.prepare import _get_obs_or_fq_map +from torch.ao.quantization.quantizer import Quantizer +from torch.ao.quantization.quantizer.quantizer import QuantizationSpec +from torch.ao.quantization.quantizer.quantizer import SharedQuantizationSpec + +import nncf +from nncf.common.graph.graph import NNCFGraph +from nncf.common.quantization.quantizer_setup import ActivationQuantizationInsertionPoint +from nncf.common.quantization.quantizer_setup import SingleConfigQuantizationPoint +from nncf.common.quantization.quantizer_setup import SingleConfigQuantizerSetup +from nncf.common.quantization.quantizer_setup import WeightQuantizationInsertionPoint +from nncf.common.quantization.structs import QuantizationScheme as QuantizationMode +from nncf.common.quantization.structs import QuantizerConfig +from nncf.experimental.common.quantization.algorithms.quantizer.quantizer import NNCFQuantizer + + +class NNCFFXQuantizer(NNCFQuantizer): + def __init__(self, quantizer: Quantizer): + self._quantizer = quantizer + + def get_quantization_setup(self, model: torch.fx.GraphModule, nncf_graph: NNCFGraph) -> SingleConfigQuantizerSetup: + anotated_model = deepcopy(model) + + self._quantizer.transform_for_annotation(anotated_model) + self._quantizer.annotate(anotated_model) + self._quantizer.validate(anotated_model) + return self.get_quantizer_config_from_anotated_model(anotated_model) + + @staticmethod + def get_quantizer_config_from_anotated_model(anotated_model: torch.fx.GraphModule) -> SingleConfigQuantizerSetup: + is_qat = False + edge_or_node_to_qspec = _get_edge_or_node_to_qspec(anotated_model) + edge_or_node_to_group_id = _get_edge_or_node_to_group_id(edge_or_node_to_qspec) + obs_or_fq_map = _get_obs_or_fq_map(edge_or_node_to_group_id, edge_or_node_to_qspec, is_qat) + if obs_or_fq_map: + pass + + q_map = defaultdict(list) + for edge, qspec in edge_or_node_to_qspec.items(): + if not isinstance(edge, tuple): + continue + from_n, to_n = edge + q_map[from_n].append(to_n) + + q_setup = SingleConfigQuantizerSetup() + for from_n, to_nodes in q_map.items(): + to_n = to_nodes[0] + qspec = edge_or_node_to_qspec[(from_n, to_n)] + if qspec is None: + continue + if isinstance(qspec, QuantizationSpec): + if qspec.qscheme in [torch.per_channel_affine, torch.per_channel_symmetric]: + per_channel = True + elif qspec.qscheme in [torch.per_tensor_affine, torch.per_tensor_symmetric]: + per_channel = False + else: + raise nncf.InternalError(f"Unknown qscheme: {qspec.qscheme}") + signed = qspec.dtype is torch.uint8 + mode = ( + QuantizationMode.SYMMETRIC + if qspec.qscheme in [torch.per_channel_symmetric, torch.per_tensor_symmetric] + else QuantizationMode.ASYMMETRIC + ) + qconfig = QuantizerConfig(mode=mode, signedness_to_force=signed, per_channel=per_channel) + qps = [] + # If input node is a constant and placed not at activations port (0) + if from_n.op == "get_attr" and to_n.args.index(from_n) != 0: + qip = WeightQuantizationInsertionPoint(to_n.name) + qp = SingleConfigQuantizationPoint(qip, qconfig, [x.name for x in to_nodes]) + qps.append(qp) + else: + if len(from_n.users) == len(to_nodes): + qip = ActivationQuantizationInsertionPoint(from_n.name) + qp = SingleConfigQuantizationPoint(qip, qconfig, [x.name for x in to_nodes]) + qps.append(qp) + else: + for to_n_ in to_nodes: + input_port_id = to_n_.args.index(from_n) + qip = ActivationQuantizationInsertionPoint(to_n_.name, input_port_id) + qp = SingleConfigQuantizationPoint(qip, qconfig, [to_n_.name]) + qps.append(qp) + + for qp in qps: + q_setup.add_independent_quantization_point(qp) + + elif isinstance(qspec, SharedQuantizationSpec): + pass + else: + raise nncf.InternalError(f"Unknown torch.ao quantization spec: {qspec}") + + return q_setup diff --git a/nncf/experimental/common/quantization/algorithms/quantizer/quantizer.py b/nncf/experimental/common/quantization/algorithms/quantizer/quantizer.py new file mode 100644 index 00000000000..b0d40234210 --- /dev/null +++ b/nncf/experimental/common/quantization/algorithms/quantizer/quantizer.py @@ -0,0 +1,26 @@ +# Copyright (c) 2024 Intel Corporation +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# http://www.apache.org/licenses/LICENSE-2.0 +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from abc import abstractmethod +from typing import TypeVar + +from nncf.common.graph.graph import NNCFGraph +from nncf.common.quantization.quantizer_setup import SingleConfigQuantizerSetup + +TModel = TypeVar("TModel") + + +class NNCFQuantizer: + @abstractmethod + def get_quantization_setup(self, model: TModel, nncf_graph: NNCFGraph) -> SingleConfigQuantizerSetup: + """ + Return quantization setup. + """ diff --git a/nncf/experimental/common/quantization/algorithms/range_estimator/backend.py b/nncf/experimental/common/quantization/algorithms/range_estimator/backend.py new file mode 100644 index 00000000000..dbd11f3f6b7 --- /dev/null +++ b/nncf/experimental/common/quantization/algorithms/range_estimator/backend.py @@ -0,0 +1,154 @@ +# Copyright (c) 2024 Intel Corporation +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# http://www.apache.org/licenses/LICENSE-2.0 +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from abc import ABC +from abc import abstractmethod +from typing import List, Optional, Set, Tuple, TypeVar + +from nncf.common.graph.graph import NNCFGraph +from nncf.common.graph.graph import NNCFNode +from nncf.common.graph.transformations.commands import TargetPoint +from nncf.common.graph.transformations.commands import TargetType +from nncf.common.graph.transformations.commands import TransformationCommand +from nncf.common.quantization.structs import QuantizerConfig +from nncf.common.tensor_statistics.collectors import TensorStatisticCollectorBase +from nncf.quantization.fake_quantize import FakeQuantizeParameters +from nncf.quantization.range_estimator import RangeEstimatorParameters + +TModel = TypeVar("TModel") + + +class RangeEstimatorAlgoBackend(ABC): + @staticmethod + @abstractmethod + def target_point(target_type: TargetType, target_node_name: str, port_id: int) -> TargetPoint: + """ + Returns backend-specific target point. + + :param target_type: Type of the location that should be modified. + :param target_node_name: Name of the located node. + :param port_id: Port ID of the tensor for the statistics distribution. + :return: Backend-specific TargetPoint. + """ + + @staticmethod + @abstractmethod + def create_quantizer_insertion_command( + nncf_graph: NNCFGraph, + target_point: TargetPoint, + quantizer_config: QuantizerConfig, + parameters: FakeQuantizeParameters, + ) -> TransformationCommand: + """ + Returns backend-specific quantizer insertion command. + + :param nncf_graph: NNCFGraph to get input/output shapes for the target point. + :param target_point: Target location for the quantizer insertion. + :param quantizer_config: QuantizerConfig instance for the current layer. + :param parameters: FakeQuantizeParameters to calculate activation quantization parameters. + :return: Backend-specific TransformationCommand for the quantizer insertion operation. + """ + + @staticmethod + @abstractmethod + def create_unified_scales_quantizers_insertion_commands( + nncf_graph: NNCFGraph, + target_points: List[TargetPoint], + quantizer_config: QuantizerConfig, + parameters: FakeQuantizeParameters, + ) -> List[TransformationCommand]: + """ + Returns backend-specific unified scales quantizers insertion commands. + + :param nncf_graph: NNCFGraph to get input/output shapes for the target point. + :param target_points: List of target locations for the quantizers insertion. + :param quantizer_config: QuantizerConfig instance for the current layer. + :param parameters: FakeQuantizeParameters to calculate activation quantization parameters. + :return: List of backend-specific TransformationCommands + for the quantizers with unified scales insertion operations. + """ + + @staticmethod + @abstractmethod + def get_target_point_shape(nncf_graph: NNCFGraph, node: NNCFNode, target_point: TargetPoint) -> Tuple[int, ...]: + """ + Returns shape of a target point tensor. + + :param nncf_graph: NNCFGraph instance. + :param node: NNCFNode. + :param target_point: Target point of which tensor shape is seeked. + :return: Shape of target point tensor. + """ + + @staticmethod + @abstractmethod + def get_weight_quantization_axes(node: NNCFNode, target_point: TargetPoint, ndims: int) -> Tuple[int, ...]: + """ + Returns axes for per-channel quantization of weights of the node placed on a input port_id. + + :param node: Quantized node with the weight. + :param target_point: Corresponding target point. + :param ndims: Number of dimensions of weight. + :return: Axes for per-channel quantization of weights. + """ + + @staticmethod + @abstractmethod + def get_statistic_collector( + range_estimator_params: RangeEstimatorParameters, + use_abs_max: bool, + reduction_axes: Optional[Tuple[int, ...]], + aggregation_axes: Optional[Tuple[int, ...]], + inplace: bool, + num_samples: Optional[int] = None, + ) -> TensorStatisticCollectorBase: + """ + Returns backend-specific statistic collector. + + :param range_estimator_params: Parameters that specify estimators types. + :param use_abs_max: Wheather reduce absolute values of input tensors or not. + :param reduction_axes: Axes for reducer. + :param aggregation_axes: Axes for aggregator. + :param inplace: Whether to calculate statistic inplace or not. + :param num_samples: Maximum number of samples to collect. + :return: Backend-specific TensorStatisticCollectorBase for the statistics calculation. + """ + + @staticmethod + @abstractmethod + def get_weight_tensor_port_ids(node: NNCFNode, graph: NNCFGraph) -> List[Optional[int]]: + """ + Returns node's input port indices with weight tensors. + + :param node: NNCFNode to find its weight input port indices. + :param graph: NNCFGraph instance. + :return: Weights input port indices. + """ + + @staticmethod + def get_weight_name(nncf_graph: NNCFGraph, target_point: TargetPoint) -> str: + """ + Returns node's weight name corresponding to port ID. + + :param nncf_graph: NNCFGraph instance. + :param target_point: The TargetPoint instance that contains layer's information. + :return: Weight name. + """ + + @staticmethod + def should_quantize_weight(weight_name: str, quantized_weight_names: Set[str]) -> bool: + """ + Return True if weight should be quantized. + + :param weight_name: Weight name. + :param quantized_weight_names: Set containing already quantized weight names. + :return: A boolean value specifying whether a weight should be quantized. + """ diff --git a/nncf/experimental/common/quantization/algorithms/range_estimator/range_estimator.py b/nncf/experimental/common/quantization/algorithms/range_estimator/range_estimator.py new file mode 100644 index 00000000000..5431703cdb1 --- /dev/null +++ b/nncf/experimental/common/quantization/algorithms/range_estimator/range_estimator.py @@ -0,0 +1,495 @@ +# Copyright (c) 2024 Intel Corporation +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# http://www.apache.org/licenses/LICENSE-2.0 +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import collections +import dataclasses +from copy import deepcopy +from typing import List, Optional, OrderedDict, Tuple, TypeVar + +import nncf +import nncf.tensor.functions as fns +from nncf import Dataset +from nncf.common.factory import ModelTransformerFactory +from nncf.common.graph.graph import NNCFGraph +from nncf.common.graph.transformations.commands import TargetPoint +from nncf.common.graph.transformations.commands import TargetType +from nncf.common.graph.transformations.layout import TransformationLayout +from nncf.common.logging import nncf_logger +from nncf.common.quantization.initialization.range import RangeInitCollectorParams +from nncf.common.quantization.quantizer_setup import SingleConfigQuantizationPoint +from nncf.common.quantization.quantizer_setup import SingleConfigQuantizerSetup +from nncf.common.quantization.structs import QuantizerConfig +from nncf.common.quantization.structs import QuantizerGroup +from nncf.common.tensor_statistics.collectors import TensorStatisticCollectorBase +from nncf.common.tensor_statistics.statistic_point import StatisticPoint +from nncf.common.tensor_statistics.statistic_point import StatisticPointsContainer +from nncf.common.utils.backend import BackendType +from nncf.common.utils.backend import get_backend +from nncf.experimental.common.quantization.algorithms.quantizer.quantizer import NNCFQuantizer +from nncf.experimental.common.tensor_statistics.statistics import MinMaxTensorStatistic +from nncf.quantization.advanced_parameters import changes_asdict +from nncf.quantization.algorithms.algorithm import Algorithm +from nncf.quantization.fake_quantize import calculate_quantizer_parameters +from nncf.quantization.fake_quantize import get_quantizer_narrow_range +from nncf.quantization.range_estimator import RangeEstimatorParameters +from nncf.quantization.range_estimator import RangeEstimatorParametersSet + +TModel = TypeVar("TModel") + + +class MinMaxRangeEstimator(Algorithm): + def __init__( + self, + quantizer: NNCFQuantizer, + subset_size: int = 300, + inplace_statistics: bool = True, + batchwise_statistics: bool = False, + activations_range_estimator_params: Optional[RangeEstimatorParameters] = None, + weights_range_estimator_params: Optional[RangeEstimatorParameters] = None, + ): + """ + :param subset_size: Size of a subset to calculate activations statistics used + for quantization, defaults to 300. + :param inplace_statistics: Defines wheather to calculate quantizers statistics + by backend graph operations or by default Python implementation, defaults + to True. + :param batchwise_statistics: Determines whether quantizer statistics should be calculated + for each item of the batch or for the entire batch, default is False. + :param activations_range_estimator_params: Quantization range estimation + parameters for activation. + :param weights_range_estimator_params: Quantization range estimation parameters + for weights. + """ + self._quantizer = quantizer + self._subset_size = subset_size + self._inplace_statistics = inplace_statistics + self._batchwise_statistics = batchwise_statistics + self._activations_range_estimator_params = activations_range_estimator_params + self._weights_range_estimator_params = weights_range_estimator_params + + self._range_estimator_params = { + QuantizerGroup.WEIGHTS: self._weights_range_estimator_params, + QuantizerGroup.ACTIVATIONS: self._activations_range_estimator_params, + } + # Calculates global quantizer constraints + self._reset_cache() + self._algorithm_key = f"MMQ_{hash(self)}" + + def _reset_cache(self) -> None: + """ + Marks cache by noninitialized values. Needs to be called when the new quantizer setup is needed. + """ + self._quantization_target_points_to_qconfig: OrderedDict[TargetPoint, QuantizerConfig] = None + self._unified_scale_groups = None + + def _init_cache(self) -> None: + """ + Initializes cache. + """ + self._quantization_target_points_to_qconfig: OrderedDict[TargetPoint, QuantizerConfig] = ( + collections.OrderedDict() + ) + self._unified_scale_groups = [] + + @property + def available_backends(self) -> List[BackendType]: + return [BackendType.TORCH_FX] + + def _set_backend_entity(self, model: TModel) -> None: + """ + Creates a helper class with a backed-specific logic of the algorithm + + :param model: backend-specific input model + """ + model_backend = get_backend(model) + if model_backend == BackendType.TORCH_FX: + from nncf.experimental.common.quantization.algorithms.range_estimator.torch_fx_backend import ( + FXRangeEstimatorAlgoBackend, + ) + + self._backend_entity = FXRangeEstimatorAlgoBackend() + else: + raise nncf.UnsupportedBackendError( + "Cannot return backend-specific entity because {} is not supported!".format(model_backend.value) + ) + + def _get_range_estimator_parameters( + self, target_point: TargetPoint, quantizer_config: QuantizerConfig + ) -> RangeEstimatorParameters: + """ + Returns range estimator parameters. + + :param target_point: Quantizer target point. + :param quantizer_config: Quantizer config. + :return: Range estimator parameters. + """ + quantizer_group = QuantizerGroup.ACTIVATIONS + if target_point.is_weight_target_point(): + quantizer_group = QuantizerGroup.WEIGHTS + + if quantizer_group == QuantizerGroup.WEIGHTS or ( + quantizer_group == QuantizerGroup.ACTIVATIONS and quantizer_config.per_channel + ): + params = RangeEstimatorParametersSet.MINMAX + else: + params = RangeEstimatorParametersSet.MEAN_MINMAX + + user_params = self._range_estimator_params[quantizer_group] + if user_params is None: + return deepcopy(params) + + min_changes = changes_asdict(user_params.min) + min_statistic_collector = dataclasses.replace(params.min, **min_changes) + + max_changes = changes_asdict(user_params.max) + max_statistic_collector = dataclasses.replace(params.max, **max_changes) + + return RangeEstimatorParameters(min_statistic_collector, max_statistic_collector) + + def _get_stat_collector( + self, + graph: NNCFGraph, + target_point: TargetPoint, + qconfig: QuantizerConfig, + batchwise_statistics: bool, + ) -> TensorStatisticCollectorBase: + """ + Creates and returns a statistic collector based on the quantizer's configuration. + + :param graph: NNCFGraph instance. + :param target_point: Target point indicates where statistics should be collected. + :param qconfig: Configuration of a quantizer layer, + defining the configuration of created statistic collector. + :param batchwise_statistics: Determines whether quantizer statistics should be calculated + for each item of the batch or for the entire batch. + :return: Statistic Collector. + """ + is_weight = target_point.is_weight_target_point() + node = graph.get_node_by_name(target_point.target_node_name) + shape = self._backend_entity.get_target_point_shape(graph, node, target_point) + range_estimator_params = self._get_range_estimator_parameters(target_point, qconfig) + + channel_axes = () + if qconfig.per_channel: + channel_axes = ( + self._backend_entity.get_weight_quantization_axes(node, target_point, len(shape)) if is_weight else (1,) + ) + + # Weight statistics is constant, so only one collection is enough. + num_samples = self._subset_size if not is_weight else 1 + + batchwise_statistics = batchwise_statistics and not is_weight + + collector_params = RangeInitCollectorParams( + is_weights=is_weight, scheme=qconfig.mode, per_channel=qconfig.per_channel + ) + reduction_axes, aggregation_axes = None, None + if shape is not None: + reduction_axes, aggregation_axes = collector_params.get_reduction_aggregation_axes( + shape, channel_axes, batchwise_statistics + ) + + return self._backend_entity.get_statistic_collector( + range_estimator_params, + collector_params.use_abs_max, + reduction_axes, + aggregation_axes, + self._inplace_statistics, + num_samples=num_samples, + ) + + def _add_weight_quantization_target_point( + self, quantization_point: SingleConfigQuantizationPoint, nncf_graph: NNCFGraph + ) -> None: + """ + Adds weight quantization target point to the set of existing points. + + :param quantization_point: SingleConfigQuantizationPoint for the needed layer. + :param nncf_graph: The built NNCFGraph of the model. + """ + weight_quantization_target_points = self._get_weight_quantization_target_points(quantization_point, nncf_graph) + for weight_quantization_target_point in weight_quantization_target_points: + self._quantization_target_points_to_qconfig[weight_quantization_target_point] = quantization_point.qconfig + + def _add_activation_quantization_target_point( + self, quantization_point: SingleConfigQuantizationPoint, nncf_graph: NNCFGraph + ) -> None: + """ + Adds activation quantization target point to the set of existing points. + + :param quantization_point: SingleConfigQuantizationPoint for the needed layer. + :param nncf_graph: NNCFGraph instance for working with the graph and nodes. + """ + activation_quantization_target_point = self._get_activation_quantization_target_point( + quantization_point, nncf_graph + ) + self._quantization_target_points_to_qconfig[activation_quantization_target_point] = quantization_point.qconfig + + def _get_weight_quantization_target_points( + self, quantization_point: SingleConfigQuantizationPoint, nncf_graph: NNCFGraph + ) -> List[SingleConfigQuantizationPoint]: + """ + Returns weight quantization target points to the set of existing points. + + :param quantization_point: SingleConfigQuantizationPoint for the needed layer. + :param nncf_graph: NNCFGraph instance for working with the graph and nodes. + :return: List of SingleConfigQuantizationPoints for the needed layer. + """ + weight_quantization_target_points = [] + node_name = quantization_point.insertion_point.target_node_name + node = nncf_graph.get_node_by_name(node_name) + weights_port_ids = self._backend_entity.get_weight_tensor_port_ids(node, nncf_graph) + for port_id in weights_port_ids: + weight_quantization_target_points.append( + self._backend_entity.target_point(TargetType.OPERATION_WITH_WEIGHTS, node_name, port_id) + ) + return weight_quantization_target_points + + def _get_activation_quantization_target_point( + self, quantization_point: SingleConfigQuantizationPoint, nncf_graph: NNCFGraph + ) -> SingleConfigQuantizationPoint: + """ + Returns activation quantization target point to the set of existing points. + + :param quantization_point: SingleConfigQuantizationPoint for the needed layer. + :param nncf_graph: NNCFGraph instance for working with the graph and nodes. + :return: SingleConfigQuantizationPoint for the needed layer. + """ + node_name = quantization_point.insertion_point.target_node_name + # If Quantization of node's input + if quantization_point.insertion_point.input_port_id is not None: + input_port_id = quantization_point.insertion_point.input_port_id + activation_quantization_target_point = self._backend_entity.target_point( + TargetType.PRE_LAYER_OPERATION, node_name, input_port_id + ) + # If quantization of node's output or Model Input node + else: + # NOTE: Assumes that the operation has output edges only from one output port because + # we haven't encountered a model with operations that have multiple output edges with different + # output port IDs. Currently, such models are not supported. Usually, `output_port_id = 0` is used. + # However, there are operations, such as LSTMSequence, where the `output_port_id` changes from case + # to case. Therefore, the code below is required to dynamically determine the `output_port_id` where + # the quantize operation should be inserted." + node = nncf_graph.get_node_by_name(node_name) + unique_output_port_ids = set(e.output_port_id for e in nncf_graph.get_output_edges(node)) + if len(unique_output_port_ids) > 1: + nncf_logger.warning( + f"Cannot determine the output_port_id for the operation: {node_name}, " + "output_port_id = 0 will be used." + ) + output_port_id = 0 + else: + output_port_id = next(iter(unique_output_port_ids)) + + activation_quantization_target_point = self._backend_entity.target_point( + TargetType.POST_LAYER_OPERATION, node_name, output_port_id + ) + return activation_quantization_target_point + + def _find_quantization_target_points( + self, model: TModel, nncf_graph: NNCFGraph + ) -> Tuple[OrderedDict[TargetPoint, QuantizerConfig], List[List[TargetPoint]]]: + """ + Initializes a cache, finds quantization target points and them puts in the cache. + + :param model: Backend-specific model, for which Quantization Target Points are being seek. + :param nncf_graph: NNCFGraph instance. + :return: Mapping of quantization target points with associated quantization configuration, + along with target points for scale unification. + """ + quantizer_setup = self._quantizer.get_quantization_setup(model, nncf_graph) + self._unified_scale_groups = self._collect_unified_groups(quantizer_setup, nncf_graph) + quantization_points = list(quantizer_setup.quantization_points.values()) + quantization_points = self._topological_sort_quantization_points(quantization_points, nncf_graph) + for quantization_point in quantization_points: + if quantization_point.is_weight_quantization_point(): + self._add_weight_quantization_target_point(quantization_point, nncf_graph) + elif quantization_point.is_activation_quantization_point(): + self._add_activation_quantization_target_point(quantization_point, nncf_graph) + else: + raise nncf.InternalError("Incorrect quantization point") + return self._quantization_target_points_to_qconfig, self._unified_scale_groups + + def _get_quantization_target_points( + self, model: TModel, nncf_graph: NNCFGraph + ) -> Tuple[OrderedDict[TargetPoint, QuantizerConfig], List[List[TargetPoint]]]: + """ + Returns Quantization Target Points. + Returns a cache with target points if exists. Otherwise, initiates a procedure of finding them. + + :param model: Backend-specific model, for which Quantization Target Points are being seek. + :param nncf_graph: NNCFGraph instance. + :return: Mapping of quantization target points with associated quantization configuration, + along with target points for scale unification. + """ + if self._quantization_target_points_to_qconfig is not None: + return self._quantization_target_points_to_qconfig, self._unified_scale_groups + self._init_cache() + return self._find_quantization_target_points(model, nncf_graph) + + def _collect_unified_groups( + self, quantizer_setup: SingleConfigQuantizerSetup, nncf_graph: NNCFGraph + ) -> List[List[TargetPoint]]: + """ + Collects the group of quantizers for unification. + + :param quantizer_setup: SingleConfigQuantizerSetup instance. + :param nncf_graph: NNCFGraph instance. + :return: List with the groups of the TargetPoints. + """ + unified_scale_groups = [] + for quantizer_ids in quantizer_setup.unified_scale_groups.values(): + unified_scale_group = [] + for quantizer_id in quantizer_ids: + quantization_point = quantizer_setup.quantization_points[quantizer_id] + + # Only activation quantizers can be unified + if quantization_point.is_activation_quantization_point(): + activation_target_point = self._get_activation_quantization_target_point( + quantization_point, nncf_graph + ) + unified_scale_group.append(activation_target_point) + else: + weight_target_points = self._get_weight_quantization_target_points(quantization_point, nncf_graph) + for weight_target_point in weight_target_points: + unified_scale_group.append(weight_target_point) + unified_scale_groups.append(unified_scale_group) + return unified_scale_groups + + def _topological_sort_quantization_points( + self, quantization_points: List[SingleConfigQuantizationPoint], nncf_graph: NNCFGraph + ) -> List[SingleConfigQuantizationPoint]: + """ + Sorts quantization_points based on the topological order of nodes obtained form nncf_graph. + + :param quantization_points: Quantization points. + :param nncf_graph: Instance of NNCFgraph used to get topological sort. + :return: Sorted quantization_points. + """ + node_names_to_pos = {node.node_name: i for i, node in enumerate(nncf_graph.topological_sort())} + quantization_points.sort(key=lambda point: node_names_to_pos[point.insertion_point.target_node_name]) + return quantization_points + + def apply( + self, + model: TModel, + graph: NNCFGraph, + statistic_points: Optional[StatisticPointsContainer] = None, + dataset: Optional[Dataset] = None, + ) -> TModel: + transformation_layout = TransformationLayout() + model_transformer = ModelTransformerFactory.create(model) + quantization_target_points, unified_scale_groups = self._get_quantization_target_points(model, graph) + weight_layer_names = set() + + def filter_func(point: StatisticPoint) -> bool: + return ( + self._algorithm_key in point.algorithm_to_tensor_collectors + and point.target_point == quantization_target_point + ) + + unified_ops_list = set() + for unified_scale_group in unified_scale_groups: + group_statistics = [] + for quantization_target_point in unified_scale_group: + target_node_name = quantization_target_point.target_node_name + for tensor_collector in statistic_points.get_algo_statistics_for_node( + target_node_name, filter_func, self._algorithm_key + ): + statistics = tensor_collector.get_statistics() + if statistics.min_values is None or statistics.max_values is None: + raise nncf.InternalError(f"Statistics were not collected for the node {target_node_name}") + group_statistics.append(statistics) + + unified_values = self._unify_statistics(group_statistics) + qconfigs = [quantization_target_points[qtp] for qtp in unified_scale_group] + if any(qconfigs[0] != qconfig for qconfig in qconfigs[1:]): + raise nncf.InternalError(f"QConfigs for unified scale group {unified_scale_group} are not equal") + qconfig = qconfigs[0] + q_group = QuantizerGroup.ACTIVATIONS + narrow_range = get_quantizer_narrow_range(qconfig, q_group) + parameters = calculate_quantizer_parameters(unified_values, qconfig, q_group, narrow_range) + commands = self._backend_entity.create_unified_scales_quantizers_insertion_commands( + graph, unified_scale_group, qconfig, parameters + ) + for command in commands: + transformation_layout.register(command) + unified_ops_list.update(unified_scale_group) + + for quantization_target_point, qconfig in quantization_target_points.items(): + if quantization_target_point in unified_ops_list: + continue + target_node_name = quantization_target_point.target_node_name + for tensor_collector in statistic_points.get_algo_statistics_for_node( + target_node_name, filter_func, self._algorithm_key + ): + if quantization_target_point.is_weight_target_point(): + weights_name = self._backend_entity.get_weight_name(graph, quantization_target_point) + if not self._backend_entity.should_quantize_weight(weights_name, weight_layer_names): + continue + weight_layer_names.add(weights_name) + quant_group = QuantizerGroup.WEIGHTS + else: + quant_group = QuantizerGroup.ACTIVATIONS + + half_range = False + narrow_range = get_quantizer_narrow_range(qconfig, quant_group) + statistics = tensor_collector.get_statistics() + if statistics.min_values is None or statistics.max_values is None: + raise nncf.InternalError(f"Statistics were not collected for the node {target_node_name}") + parameters = calculate_quantizer_parameters(statistics, qconfig, quant_group, narrow_range, half_range) + command = self._backend_entity.create_quantizer_insertion_command( + graph, quantization_target_point, qconfig, parameters + ) + transformation_layout.register(command) + if not transformation_layout.transformations: + nncf_logger.info("The model has no operations to apply quantization.") + quantized_model = model_transformer.transform(transformation_layout) + return quantized_model + + def get_statistic_points(self, model: TModel, graph: NNCFGraph) -> StatisticPointsContainer: + self._set_backend_entity(model) + self._reset_cache() + quantization_target_points, _ = self._get_quantization_target_points(model, graph) + output = StatisticPointsContainer() + for quantization_target_point, qconfig in quantization_target_points.items(): + nncf_logger.debug( + f"Adding target point {quantization_target_point.target_node_name}" + f" with type {quantization_target_point.type} for statistics collection" + ) + stat_collector = self._get_stat_collector( + graph, quantization_target_point, qconfig, self._batchwise_statistics + ) + output.add_statistic_point( + StatisticPoint( + target_point=quantization_target_point, + tensor_collector=stat_collector, + algorithm=self._algorithm_key, + ) + ) + return output + + @staticmethod + def _unify_statistics(statistics: List[MinMaxTensorStatistic]) -> MinMaxTensorStatistic: + """ + Returns backend-specific unified statistics. + + :param statistics: List of MinMaxTensorStatistic instances. + :return: Unified MinMaxTensorStatistic value. + """ + + max_values, min_values = [], [] + for statistic in statistics: + max_values.append(statistic.max_values.flatten()) + min_values.append(statistic.min_values.flatten()) + max_values = fns.max(fns.stack(max_values), axis=0) + min_values = fns.min(fns.stack(min_values), axis=0) + return MinMaxTensorStatistic(min_values=min_values, max_values=max_values) diff --git a/nncf/experimental/common/quantization/algorithms/range_estimator/torch_fx_backend.py b/nncf/experimental/common/quantization/algorithms/range_estimator/torch_fx_backend.py new file mode 100644 index 00000000000..0e30e70ae57 --- /dev/null +++ b/nncf/experimental/common/quantization/algorithms/range_estimator/torch_fx_backend.py @@ -0,0 +1,221 @@ +# Copyright (c) 2024 Intel Corporation +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# http://www.apache.org/licenses/LICENSE-2.0 +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import List, Optional, Set, Tuple + +import torch +from torch.quantization.fake_quantize import FakeQuantize + +import nncf +from nncf.common.graph.graph import NNCFGraph +from nncf.common.graph.graph import NNCFNode +from nncf.common.graph.transformations.commands import TargetType +from nncf.common.quantization.structs import QuantizationScheme as QuantizationMode +from nncf.common.quantization.structs import QuantizerConfig +from nncf.experimental.common.quantization.algorithms.range_estimator.backend import RangeEstimatorAlgoBackend +from nncf.experimental.common.tensor_statistics.collectors import AGGREGATORS_MAP +from nncf.experimental.common.tensor_statistics.collectors import TensorCollector +from nncf.experimental.common.tensor_statistics.statistics import MinMaxTensorStatistic +from nncf.experimental.torch.fx.commands import FXApplyTransformationCommand +from nncf.experimental.torch.fx.model_utils import get_target_point +from nncf.experimental.torch.fx.transformations import qdq_insertion_transformation_builder +from nncf.quantization.advanced_parameters import StatisticsType +from nncf.quantization.fake_quantize import FakeQuantizeParameters +from nncf.quantization.range_estimator import AggregatorType +from nncf.quantization.range_estimator import RangeEstimatorParameters +from nncf.torch.graph.graph import PTNNCFGraph +from nncf.torch.graph.graph import PTTargetPoint +from nncf.torch.graph.transformations.commands import PTSharedFnInsertionCommand +from nncf.torch.model_graph_manager import get_weight_tensor_port_ids +from nncf.torch.quantization.layers import QUANTIZATION_MODULES +from nncf.torch.quantization.layers import AsymmetricQuantizer +from nncf.torch.quantization.layers import BaseQuantizer +from nncf.torch.quantization.layers import PTQuantizerSpec +from nncf.torch.quantization.layers import get_scale_shape +from nncf.torch.quantization.strip import convert_to_torch_fakequantizer +from nncf.torch.tensor_statistics.collectors import PT_REDUCERS_MAP + + +class FXRangeEstimatorAlgoBackend(RangeEstimatorAlgoBackend): + @staticmethod + def target_point(target_type: TargetType, target_node_name: str, port_id: int) -> PTTargetPoint: + return get_target_point(target_type, target_node_name, port_id) + + @staticmethod + def get_target_point_shape(nncf_graph: PTNNCFGraph, node: NNCFNode, target_point: PTTargetPoint) -> Tuple[int, ...]: + return nncf_graph.get_input_shape_for_insertion_point(target_point) + + @staticmethod + def get_weight_quantization_axes(node: NNCFNode, target_point: PTTargetPoint, ndims: int) -> Tuple[int]: + # TODO(dlyakhov): support transpose conv and other cases + return (0,) + + @staticmethod + def get_statistic_collector( + range_estimator_params: RangeEstimatorParameters, + use_abs_max: bool, + reduction_axes: Optional[Tuple[int, ...]], + aggregation_axes: Optional[Tuple[int, ...]], + inplace: bool, + num_samples: Optional[int] = None, + ) -> TensorCollector: + collector = TensorCollector(MinMaxTensorStatistic) + for params, container_key in zip( + [range_estimator_params.min, range_estimator_params.max], + [MinMaxTensorStatistic.MIN_STAT, MinMaxTensorStatistic.MAX_STAT], + ): + if params.statistics_type not in PT_REDUCERS_MAP: + raise nncf.InternalError( + f"Statistic type: {params.statistics_type} is not supported for Torch PTQ backend yet." + ) + + if params.aggregator_type not in AGGREGATORS_MAP: + raise nncf.InternalError( + f"Aggregator type: {params.aggregator_type} is not supported for Torch PTQ backend yet." + ) + + statistic_type = params.statistics_type + if statistic_type in [StatisticsType.QUANTILE, StatisticsType.ABS_QUANTILE]: + # TODO(dlyakhov): merge two quantile aggregators in one + if container_key == MinMaxTensorStatistic.MIN_STAT: + quantile = params.quantile_outlier_prob + else: + quantile = 1 - params.quantile_outlier_prob + reducer = PT_REDUCERS_MAP[statistic_type](reduction_axes=reduction_axes, quantile=[quantile]) + else: + if use_abs_max and statistic_type == StatisticsType.MAX: + statistic_type = StatisticsType.ABS_MAX + reducer = PT_REDUCERS_MAP[statistic_type](reduction_axes=reduction_axes) + + kwargs = { + "num_samples": num_samples, + "aggregation_axes": aggregation_axes, + } + if params.aggregator_type in [AggregatorType.MEAN_NO_OUTLIERS, AggregatorType.MEDIAN_NO_OUTLIERS]: + kwargs.update({"quantile": params.quantile_outlier_prob}) + aggregator = AGGREGATORS_MAP[params.aggregator_type](**kwargs) + + collector.register_statistic_branch(container_key, reducer, aggregator) + return collector + + @staticmethod + def get_weight_tensor_port_ids(node: NNCFNode, graph: NNCFGraph) -> List[Optional[int]]: + return get_weight_tensor_port_ids(node, graph) + + @staticmethod + def get_weight_name(nncf_graph: NNCFGraph, target_point: PTTargetPoint) -> str: + weighted_node = nncf_graph.get_node_by_name(target_point.target_node_name) + weight_edge = nncf_graph.get_input_edge_by_port_id(weighted_node, target_point.input_port_id) + weight = weight_edge.from_node + return weight.node_name + + @staticmethod + def should_quantize_weight(weight_name: str, quantized_weight_names: Set[str]) -> bool: + # If the nodes share one weight tensor, we should have only one quantizer on that + return weight_name not in quantized_weight_names + + @staticmethod + def _get_input_scale_shape( + nncf_graph: NNCFGraph, target_point: PTTargetPoint, per_channel: bool + ) -> Tuple[Tuple[int, ...], Tuple[int, ...], int]: + is_weights = target_point.is_weight_target_point() + if is_weights: + # TODO(dlyakhov): support transpose conv/ make channel_idx common + channel_idx = 0 + else: + channel_idx = 1 # channel dim for activations + + input_shape = nncf_graph.get_input_shape_for_insertion_point(target_point) + scale_shape = tuple( + get_scale_shape(input_shape, is_weights=is_weights, per_channel=per_channel, channel_idx=channel_idx) + ) + + return input_shape, scale_shape, channel_idx + + @staticmethod + def _create_quantizer( + quantizer_config: QuantizerConfig, + scale_shape: Tuple, + parameters: FakeQuantizeParameters, + target_type: TargetType, + ) -> FakeQuantize: + mode = quantizer_config.mode + quantizer_cls = QUANTIZATION_MODULES.get(mode) + narrow_range = target_type == TargetType.OPERATION_WITH_WEIGHTS and mode == QuantizationMode.SYMMETRIC + quantizer_spec = PTQuantizerSpec.from_config( + quantizer_config, + narrow_range=narrow_range, + scale_shape=scale_shape, + half_range=False, + logarithm_scale=False, + is_quantized_on_export=False, + compression_lr_multiplier=None, + ) + quantizer = quantizer_cls(quantizer_spec) + + # Fill it with minmax + # TODO(dlyakhov) Prevent creation of intermediate objects like nncf quantizer. + FXRangeEstimatorAlgoBackend._fill_quantizer_parameters(quantizer, parameters, quantizer_spec.scale_shape) + # Convert to the torch fake quantizer + torch_fq = convert_to_torch_fakequantizer(quantizer) + return torch_fq + + @staticmethod + def _fill_quantizer_parameters(quantizer: BaseQuantizer, parameters: FakeQuantizeParameters, scale_shape) -> None: + if isinstance(quantizer, AsymmetricQuantizer): + quantizer.input_low = torch.nn.Parameter(parameters.input_low.data.reshape(scale_shape)) + input_range = parameters.input_high - parameters.input_low + # Subtract eps from the input_range to make quantizer parameters equal to + # original parameters on the forward call. + quantizer.input_range = torch.nn.Parameter((input_range.data - quantizer.eps).reshape(scale_shape)) + else: + quantizer.signed = bool(torch.any(parameters.input_low.data < 0)) + # Subtract eps from the scale to make quantizer parameters equal to + # original parameters on the forward call. + quantizer.scale = torch.nn.Parameter((parameters.input_high.data - quantizer.eps).reshape(scale_shape)) + + @staticmethod + def create_quantizer_insertion_command( + nncf_graph: NNCFGraph, + target_point: PTTargetPoint, + quantizer_config: QuantizerConfig, + parameters: FakeQuantizeParameters, + ) -> FXApplyTransformationCommand: + _, scale_shape, _ = FXRangeEstimatorAlgoBackend._get_input_scale_shape( + nncf_graph, target_point, quantizer_config.per_channel + ) + + quantizer = FXRangeEstimatorAlgoBackend._create_quantizer( + quantizer_config, scale_shape, parameters, target_point.target_type + ) + transformation = qdq_insertion_transformation_builder(quantizer, [target_point]) + return FXApplyTransformationCommand(transformation) + + @staticmethod + def create_unified_scales_quantizers_insertion_commands( + nncf_graph: NNCFGraph, + target_points: List[PTTargetPoint], + quantizer_config: QuantizerConfig, + parameters: FakeQuantizeParameters, + ) -> List[PTSharedFnInsertionCommand]: + _, scale_shape, _ = FXRangeEstimatorAlgoBackend._get_input_scale_shape( + nncf_graph, target_points[0], quantizer_config.per_channel + ) + + quantizer = FXRangeEstimatorAlgoBackend._create_quantizer( + quantizer_config, scale_shape, parameters, target_points[0].target_type + ) + + transformations = [] + for tp in target_points: + transformation = qdq_insertion_transformation_builder(quantizer, [tp]) + transformations.append(FXApplyTransformationCommand(transformation)) + return transformations diff --git a/nncf/experimental/torch/fx/quantization/quantize_pt2e.py b/nncf/experimental/torch/fx/quantization/quantize_pt2e.py new file mode 100644 index 00000000000..efa32af48d6 --- /dev/null +++ b/nncf/experimental/torch/fx/quantization/quantize_pt2e.py @@ -0,0 +1,91 @@ +# Copyright (c) 2024 Intel Corporation +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# http://www.apache.org/licenses/LICENSE-2.0 +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from copy import deepcopy +from typing import Optional + +import torch +import torch.fx +from torch.ao.quantization.pt2e.duplicate_dq_pass import DuplicateDQPass +from torch.ao.quantization.pt2e.port_metadata_pass import PortNodeMetaForQDQ +from torch.ao.quantization.pt2e.qat_utils import _fold_conv_bn_qat +from torch.ao.quantization.pt2e.utils import _disallow_eval_train +from torch.ao.quantization.quantizer import Quantizer +from torch.fx import GraphModule +from torch.fx.passes.infra.pass_manager import PassManager + +from nncf.common.factory import NNCFGraphFactory +from nncf.common.logging import nncf_logger +from nncf.data import Dataset +from nncf.experimental.common.quantization.algorithms.post_training.algorithm import PostTrainingQuantization +from nncf.experimental.common.quantization.algorithms.quantizer.fx_quantizer import NNCFFXQuantizer +from nncf.experimental.torch.fx.transformations import fuse_conv_bn +from nncf.parameters import ModelType +from nncf.quantization.advanced_parameters import AdvancedQuantizationParameters + +DEFAULT_RANGE_TYPE = "mean_min_max" + + +def quantize_pt2e( + model: torch.fx.GraphModule, + quantizer: Quantizer, + calibration_dataset: Dataset, + subset_size: int = 300, + fast_bias_correction: bool = True, + model_type: Optional[ModelType] = None, + advanced_parameters: Optional[AdvancedQuantizationParameters] = None, +) -> torch.fx.GraphModule: + """ + Implementation of the `quantize()` method for the Torch FX backend. + """ + nncf_logger.warning( + "Experimental Torch FX quantization backend is being used for the given torch.fx.GraphModule model." + " Torch FX PTQ is an experimental feature, consider using Torch or OpenVino PTQ backends" + " in case of errors or a poor model performance." + ) + + original_graph_meta = model.meta + + copied_model = deepcopy(model) + + quantization_algorithm = PostTrainingQuantization( + quantizer=NNCFFXQuantizer(quantizer), + subset_size=subset_size, + fast_bias_correction=fast_bias_correction, + model_type=model_type, + advanced_parameters=advanced_parameters, + ) + + # To make it easier for bias correction algorithms, + # biases are being separated by the followng calls. + fuse_conv_bn(copied_model) + + nncf_graph = NNCFGraphFactory.create(copied_model) + quantized_model = quantization_algorithm.apply(copied_model, nncf_graph, dataset=calibration_dataset) + + # Magic. Without this call compiled model + # is not preformant + quantized_model = GraphModule(quantized_model, quantized_model.graph) + + quantized_model = _fold_conv_bn_qat(quantized_model) + pm = PassManager([DuplicateDQPass()]) + + quantized_model = pm(quantized_model).graph_module + pm = PassManager([PortNodeMetaForQDQ()]) + quantized_model = pm(quantized_model).graph_module + + quantized_model.meta.update(original_graph_meta) + quantized_model = _disallow_eval_train(quantized_model) + # Each transformation adds a duplicate tensor value to the model buffer. + # This step removes the duplicates tensor values from the buffer. + quantized_model = GraphModule(quantized_model, quantized_model.graph) + + return quantized_model diff --git a/torch_fx_experimental_q.py b/torch_fx_experimental_q.py new file mode 100644 index 00000000000..77ca8859442 --- /dev/null +++ b/torch_fx_experimental_q.py @@ -0,0 +1,107 @@ +# Copyright (c) 2024 Intel Corporation +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# http://www.apache.org/licenses/LICENSE-2.0 +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import os + +os.environ["TORCHINDUCTOR_FREEZING"] = "1" + +from time import time + +import torch +import torch.fx +from torch._export import capture_pre_autograd_graph +from torch.ao.quantization.quantize_pt2e import convert_pt2e +from torch.ao.quantization.quantize_pt2e import prepare_pt2e +from torch.ao.quantization.quantizer.x86_inductor_quantizer import X86InductorQuantizer +from torch.ao.quantization.quantizer.x86_inductor_quantizer import get_default_x86_inductor_quantization_config +from torchvision import models + +import nncf +import nncf.torch +from nncf.experimental.torch.fx.quantization.quantize_pt2e import quantize_pt2e +from tests.torch.fx.helpers import visualize_fx_model + + +def measure_time(model, example_inputs, num_iters=3000): + with torch.no_grad(): + model(*example_inputs) + total_time = 0 + for _ in range(num_iters): + start_time = time() + model(*example_inputs) + total_time += time() - start_time + average_time = (total_time / num_iters) * 1000 + return average_time + + +def main(model_cls): + model = model_cls() + example_inputs = torch.ones((1, 3, 224, 224)) + exported_model = capture_pre_autograd_graph(model.eval(), (example_inputs,)) + + quantizer = X86InductorQuantizer() + quantizer.set_global(get_default_x86_inductor_quantization_config()) + + nncf_quantizer_model = quantize_pt2e(exported_model, quantizer, calibration_dataset=nncf.Dataset([example_inputs])) + + visualize_fx_model(nncf_quantizer_model, "nncf_quantizer_before_fold_resnet.svg") + return nncf_quantizer_model + + # exported_model = capture_pre_autograd_graph(model.eval(), (example_inputs,)) + # nncf_int8 = nncf.quantize(exported_model, nncf.Dataset([example_inputs])) + # visualize_fx_model(nncf_int8, "nncf_resnet.svg") + + +def main_native(model_cls): + model = model_cls() + example_inputs = torch.ones((1, 3, 224, 224)) + exported_model = capture_pre_autograd_graph(model.eval(), (example_inputs,)) + + quantizer = X86InductorQuantizer() + quantizer.set_global(get_default_x86_inductor_quantization_config()) + + prepared_model = prepare_pt2e(exported_model, quantizer) + prepared_model(example_inputs) + converted_model = convert_pt2e(prepared_model) + visualize_fx_model(converted_model, "x86int8_resnet.svg") + return converted_model + + +def constant_fold(m): + pass + + +if __name__ == "__main__": + with nncf.torch.disable_patching(): + for model_cls in (models.resnet18, models.mobilenet_v3_small, models.vit_b_16, models.swin_v2_s): + # for model_cls in (models.mobilenet_v3_small,): + print(f"{model_cls} check!") + nncf_q_model = main(model_cls) + + constant_fold(nncf_q_model) + visualize_fx_model(nncf_q_model, "nncf_quantizer_after_constant_fold_resnet.svg") + + pt_q_model = main_native(model_cls) + print("benchmarking...") + pt_compiled = torch.compile(model_cls()) + pt_int8_compiled = torch.compile(pt_q_model) + nncf_comipled = torch.compile(nncf_q_model) + + example_inputs = (torch.ones((1, 3, 224, 224)),) + + pt_time = measure_time(pt_compiled, example_inputs) + print(f"PT fp32 performance measured: {pt_time}") + + pt_int8_time = measure_time(pt_int8_compiled, example_inputs) + print(f"PT int8 performance measured: {pt_int8_time}") + + nncf_int8_time = measure_time(nncf_comipled, example_inputs) + print(f"NNCF int8 performance measured: {nncf_int8_time}") From 0c406cc1318202295cdb1a962a17ca6d7d6a897a Mon Sep 17 00:00:00 2001 From: dlyakhov Date: Thu, 28 Nov 2024 17:36:53 +0100 Subject: [PATCH 02/15] Experimental quantization --- .../algorithms/post_training/algorithm.py | 57 +- .../algorithms/post_training/pipeline.py | 100 +- .../algorithms/quantizer/fx_quantizer.py | 35 +- .../torch/fx/quantization/quantize_pt2e.py | 44 +- nncf/experimental/torch/fx/transformations.py | 2 + .../mobilenet_v3_small.dot | 1082 ++++ .../X86InductorQuantizer/resnet18.dot | 435 ++ .../X86InductorQuantizer/swin_v2_s.dot | 5734 +++++++++++++++++ .../synthetic_transformer.dot | 45 + .../X86InductorQuantizer/unet.dot | 437 ++ .../X86InductorQuantizer/vit_b_16.dot | 1577 +++++ .../yolo11n_sdpa_block.dot | 53 + tests/torch/fx/test_quantizer.py | 162 + torch_fx_experimental_q.py | 107 - 14 files changed, 9653 insertions(+), 217 deletions(-) create mode 100644 tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/mobilenet_v3_small.dot create mode 100644 tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/resnet18.dot create mode 100644 tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/swin_v2_s.dot create mode 100644 tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/synthetic_transformer.dot create mode 100644 tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/unet.dot create mode 100644 tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/vit_b_16.dot create mode 100644 tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/yolo11n_sdpa_block.dot create mode 100644 tests/torch/fx/test_quantizer.py delete mode 100644 torch_fx_experimental_q.py diff --git a/nncf/experimental/common/quantization/algorithms/post_training/algorithm.py b/nncf/experimental/common/quantization/algorithms/post_training/algorithm.py index 0f7a90d3d67..17558fbb1db 100644 --- a/nncf/experimental/common/quantization/algorithms/post_training/algorithm.py +++ b/nncf/experimental/common/quantization/algorithms/post_training/algorithm.py @@ -16,21 +16,22 @@ from nncf.common.graph.graph import NNCFGraph from nncf.common.tensor_statistics.statistic_point import StatisticPointsContainer from nncf.common.utils.backend import BackendType -from nncf.experimental.common.quantization.algorithms.post_training.pipeline import create_ptq_pipeline +from nncf.experimental.common.quantization.algorithms.post_training.pipeline import experimental_create_ptq_pipeline from nncf.experimental.common.quantization.algorithms.quantizer.quantizer import NNCFQuantizer -from nncf.parameters import ModelType -from nncf.quantization.advanced_parameters import AdvancedQuantizationParameters +from nncf.quantization.advanced_parameters import AdvancedBiasCorrectionParameters +from nncf.quantization.advanced_parameters import AdvancedSmoothQuantParameters +from nncf.quantization.advanced_parameters import RangeEstimatorParameters from nncf.quantization.algorithms.algorithm import Algorithm TModel = TypeVar("TModel") TPass = Callable[[TModel], TModel] -class PostTrainingQuantization(Algorithm): +class ExperimentalPostTrainingQuantization(Algorithm): """ - Implements Post-Training Quantization algorithm, which basically includes: + Implements Experimental Post-Training Quantization algorithm, which basically includes: 1) ChannelAlignment - 2) MinMaxQuantization + 2) MinMaxRangeInit 3) FastBiasCorrection or BiasCorrection """ @@ -38,39 +39,37 @@ def __init__( self, quantizer: NNCFQuantizer, subset_size: int = 300, - fast_bias_correction: bool = True, - model_type: Optional[ModelType] = None, - advanced_parameters: Optional[AdvancedQuantizationParameters] = None, + fast_bias_correction: Optional[bool] = True, + smooth_quant: bool = False, + bias_correction_params: Optional[AdvancedBiasCorrectionParameters] = None, + smooth_quant_params: Optional[AdvancedSmoothQuantParameters] = None, + activations_range_estimator_params: Optional[RangeEstimatorParameters] = None, + weights_range_estimator_params: Optional[RangeEstimatorParameters] = None, ): """ - :param mode: Special quantization mode that specify different ways of the optimization. - :param preset: A preset controls the quantization mode (symmetric and asymmetric). - It can take the following values: - - `performance`: Symmetric quantization of weights and activations. - - `mixed`: Symmetric quantization of weights and asymmetric quantization of activations. - Default value is None. In this case, `mixed` preset is used for `transformer` - model type otherwise `performace`. - :param target_device: A target device the specificity of which will be taken - into account while compressing in order to obtain the best performance - for this type of device. + :param quantizer: NNCFQuantizer to use in MiMaxRageInit algorithm. :param subset_size: Size of a subset to calculate activations statistics used for quantization. :param fast_bias_correction: Setting this option to `False` enables a different bias correction method which is more accurate, in general, and takes - more time but requires less memory. - :param model_type: Model type is needed to specify additional patterns - in the model. Supported only `transformer` now. - :param ignored_scope: An ignored scope that defined the list of model control - flow graph nodes to be ignored during quantization. - :param advanced_parameters: Advanced quantization parameters for - fine-tuning the quantization algorithm + more time but requires less memory. None disables the bias correction algorithm. + :param smooth_quant: Setting this option to `True` enables the SmoothQuant algorithm. + :param bias_correction_params: Contains advanced parameters for fine-tuning bias correction algorithm. + :param smooth_quant_params: Contains advanced alpha parameters for SmoothQuant algorithm. + :param activations_range_estimator_params: Contains parameters for estimating the range + of activations of the model. + :param weights_range_estimator_params: Contains parameters for estimating the range + of weights of the model. """ - self._pipeline = create_ptq_pipeline( + self._pipeline = experimental_create_ptq_pipeline( quantizer=quantizer, subset_size=subset_size, fast_bias_correction=fast_bias_correction, - model_type=model_type, - advanced_parameters=advanced_parameters, + smooth_quant=smooth_quant, + bias_correction_params=bias_correction_params, + smooth_quant_params=smooth_quant_params, + activations_range_estimator_params=activations_range_estimator_params, + weights_range_estimator_params=weights_range_estimator_params, ) @property diff --git a/nncf/experimental/common/quantization/algorithms/post_training/pipeline.py b/nncf/experimental/common/quantization/algorithms/post_training/pipeline.py index bb1c0ec0bba..de496eba2b3 100644 --- a/nncf/experimental/common/quantization/algorithms/post_training/pipeline.py +++ b/nncf/experimental/common/quantization/algorithms/post_training/pipeline.py @@ -11,14 +11,13 @@ from typing import Optional, TypeVar -from nncf.common.deprecation import warning_deprecated from nncf.experimental.common.quantization.algorithms.quantizer.quantizer import NNCFQuantizer from nncf.experimental.common.quantization.algorithms.range_estimator.range_estimator import MinMaxRangeEstimator -from nncf.parameters import ModelType -from nncf.quantization.advanced_parameters import AdvancedQuantizationParameters +from nncf.quantization.advanced_parameters import AdvancedBiasCorrectionParameters +from nncf.quantization.advanced_parameters import AdvancedSmoothQuantParameters +from nncf.quantization.advanced_parameters import RangeEstimatorParameters from nncf.quantization.algorithms.bias_correction.algorithm import BIAS_CORRECTION_THRESHOLD from nncf.quantization.algorithms.bias_correction.algorithm import BiasCorrection -from nncf.quantization.algorithms.channel_alignment.algorithm import ChannelAlignment from nncf.quantization.algorithms.fast_bias_correction.algorithm import FAST_BIAS_CORRECTION_THRESHOLD from nncf.quantization.algorithms.fast_bias_correction.algorithm import FastBiasCorrection from nncf.quantization.algorithms.pipeline import Pipeline @@ -27,73 +26,49 @@ TModel = TypeVar("TModel") -def create_ptq_pipeline( +def experimental_create_ptq_pipeline( quantizer: NNCFQuantizer, subset_size: int = 300, - fast_bias_correction: bool = True, - model_type: Optional[ModelType] = None, - advanced_parameters: Optional[AdvancedQuantizationParameters] = None, + fast_bias_correction: Optional[bool] = True, + smooth_quant: bool = False, + bias_correction_params: Optional[AdvancedBiasCorrectionParameters] = None, + smooth_quant_params: Optional[AdvancedSmoothQuantParameters] = None, + activations_range_estimator_params: Optional[RangeEstimatorParameters] = None, + weights_range_estimator_params: Optional[RangeEstimatorParameters] = None, ) -> Pipeline: """ - Creates a post-training quantization pipeline. + Creates an experimental post-training quantization pipeline. - The post-training quantization pipeline includes the following steps: + The experimental post-training quantization pipeline includes the following steps: 1) SmoothQuant - 2) ChannelAlignment - 3) MinMaxQuantization - 4) FastBiasCorrection or BiasCorrection - - :param mode: Special quantization mode that specify different ways of the optimization. - :param preset: A preset controls the quantization mode (symmetric and asymmetric). - It can take the following values: - - `performance`: Symmetric quantization of weights and activations. - - `mixed`: Symmetric quantization of weights and asymmetric quantization of activations. - Default value is None. In this case, `mixed` preset is used for `transformer` - model type otherwise `performace`. - :param target_device: A target device the specificity of which will be taken - into account while compressing in order to obtain the best performance - for this type of device. + 2) MinMaxRangeInit + 3) FastBiasCorrection or BiasCorrection + + :param quantizer: NNCFQuantizer to use in MiMaxRageInit algorithm. :param subset_size: Size of a subset to calculate activations statistics used for quantization. :param fast_bias_correction: Setting this option to `False` enables a different bias correction method which is more accurate, in general, and takes - more time but requires less memory. - :param model_type: Model type is needed to specify additional patterns - in the model. Supported only `transformer` now. - :param advanced_parameters: Advanced quantization parameters for - fine-tuning the quantization algorithm - :return: A post-training quantization pipeline. + more time but requires less memory. None disables the bias correction algorithm. + :param smooth_quant: Setting this option to `True` enables the SmoothQuant algorithm. + :param bias_correction_params: Contains advanced parameters for fine-tuning bias correction algorithm. + :param smooth_quant_params: Contains advanced alpha parameters for SmoothQuant algorithm. + :param activations_range_estimator_params: Contains parameters for estimating the range + of activations of the model. + :param weights_range_estimator_params: Contains parameters for estimating the range + of weights of the model. + :return: An experimental post-training quantization pipeline. """ - if advanced_parameters is None: - advanced_parameters = AdvancedQuantizationParameters() - # Build the post-training quantization pipeline. pipeline_steps = [] - # Add the `SmoothQuant` algorithm as the first step of the pipeline. - # It is added only for `ModelType.TRANSFORMER`. - sq_params = advanced_parameters.smooth_quant_alphas - sq_alpha = advanced_parameters.smooth_quant_alpha - if sq_alpha is not None: - warning_deprecated( - "`AdvancedQuantizationParameters(smooth_quant_alpha=..)` is deprecated." - "Please, use `AdvancedQuantizationParameters(smooth_quant_alphas)` option " - "with AdvancedSmoothQuantParameters(convolution=.., matmul=..) as value instead." - ) - if sq_alpha < 0: - sq_params.convolution = -1 - sq_params.matmul = -1 - else: - sq_params.matmul = sq_alpha - - if model_type == ModelType.TRANSFORMER and (sq_params.convolution >= 0 or sq_params.matmul >= 0): - alpha_map = {"convolution": sq_params.convolution, "matmul": sq_params.matmul} - pipeline_steps.append([SmoothQuant(subset_size, advanced_parameters.inplace_statistics, alpha_map=alpha_map)]) + if smooth_quant_params is None: + smooth_quant_params = AdvancedSmoothQuantParameters() - # Add the `ChannelAlignment` algorithm as the second step of the pipeline. - if not advanced_parameters.disable_channel_alignment: - pipeline_steps.append([ChannelAlignment(subset_size, advanced_parameters.inplace_statistics)]) + if smooth_quant and smooth_quant_params.convolution >= 0 or smooth_quant_params.matmul >= 0: + alpha_map = {"convolution": smooth_quant_params.convolution, "matmul": smooth_quant_params.matmul} + pipeline_steps.append([SmoothQuant(subset_size, False, alpha_map=alpha_map)]) # Add the `MinMaxQuantization` algorithm as the third step of the pipeline. pipeline_steps.append( @@ -101,19 +76,17 @@ def create_ptq_pipeline( MinMaxRangeEstimator( quantizer=quantizer, subset_size=subset_size, - inplace_statistics=advanced_parameters.inplace_statistics, - batchwise_statistics=advanced_parameters.batchwise_statistics, - activations_range_estimator_params=advanced_parameters.activations_range_estimator_params, - weights_range_estimator_params=advanced_parameters.weights_range_estimator_params, + inplace_statistics=False, + activations_range_estimator_params=activations_range_estimator_params, + weights_range_estimator_params=weights_range_estimator_params, ) ] ) - if not advanced_parameters.disable_bias_correction: + if fast_bias_correction is not None: # Add the `FastBiasCorrection` or `BiasCorrection` as additional algorithm # inside the third step of the pipeline. It is added after `MinMaxQuantization` # algorithm. - bias_correction_params = advanced_parameters.bias_correction_params if fast_bias_correction: threshold = FAST_BIAS_CORRECTION_THRESHOLD bias_correction_subset_size = subset_size @@ -123,6 +96,9 @@ def create_ptq_pipeline( bias_correction_subset_size = max(int(subset_size * 0.2), 1) bias_correction_cls = BiasCorrection + if bias_correction_params is None: + bias_correction_params = AdvancedBiasCorrectionParameters() + if bias_correction_params.threshold is not None: threshold = bias_correction_params.threshold @@ -131,8 +107,6 @@ def create_ptq_pipeline( bias_correction_subset_size, threshold, bias_correction_params.apply_for_all_nodes, - advanced_parameters.inplace_statistics, - advanced_parameters.backend_params, ) ) diff --git a/nncf/experimental/common/quantization/algorithms/quantizer/fx_quantizer.py b/nncf/experimental/common/quantization/algorithms/quantizer/fx_quantizer.py index e7d80fbaff4..7842e7475f0 100644 --- a/nncf/experimental/common/quantization/algorithms/quantizer/fx_quantizer.py +++ b/nncf/experimental/common/quantization/algorithms/quantizer/fx_quantizer.py @@ -12,14 +12,13 @@ from collections import defaultdict from copy import deepcopy +from typing import Dict, Tuple, Union import torch import torch.fx -from torch.ao.quantization.pt2e.prepare import _get_edge_or_node_to_group_id -from torch.ao.quantization.pt2e.prepare import _get_edge_or_node_to_qspec -from torch.ao.quantization.pt2e.prepare import _get_obs_or_fq_map from torch.ao.quantization.quantizer import Quantizer from torch.ao.quantization.quantizer.quantizer import QuantizationSpec +from torch.ao.quantization.quantizer.quantizer import QuantizationSpecBase from torch.ao.quantization.quantizer.quantizer import SharedQuantizationSpec import nncf @@ -32,6 +31,8 @@ from nncf.common.quantization.structs import QuantizerConfig from nncf.experimental.common.quantization.algorithms.quantizer.quantizer import NNCFQuantizer +EdgeOrNode = Union[Tuple[torch.fx.Node, torch.fx.Node]] + class NNCFFXQuantizer(NNCFQuantizer): def __init__(self, quantizer: Quantizer): @@ -47,12 +48,7 @@ def get_quantization_setup(self, model: torch.fx.GraphModule, nncf_graph: NNCFGr @staticmethod def get_quantizer_config_from_anotated_model(anotated_model: torch.fx.GraphModule) -> SingleConfigQuantizerSetup: - is_qat = False edge_or_node_to_qspec = _get_edge_or_node_to_qspec(anotated_model) - edge_or_node_to_group_id = _get_edge_or_node_to_group_id(edge_or_node_to_qspec) - obs_or_fq_map = _get_obs_or_fq_map(edge_or_node_to_group_id, edge_or_node_to_qspec, is_qat) - if obs_or_fq_map: - pass q_map = defaultdict(list) for edge, qspec in edge_or_node_to_qspec.items(): @@ -108,3 +104,26 @@ def get_quantizer_config_from_anotated_model(anotated_model: torch.fx.GraphModul raise nncf.InternalError(f"Unknown torch.ao quantization spec: {qspec}") return q_setup + + +def _get_edge_or_node_to_qspec( + model: torch.fx.GraphModule, +) -> Dict[EdgeOrNode, QuantizationSpecBase]: + """ + Get a map from EdgeOrNode to quantization spec based on annotations on the nodes. + + :param model: torch.fx.GraphModule instance. + :return: A map from EdgeOrNode to quantization spec based on annotations on the nodes. + """ + edge_or_node_to_qspec: Dict[EdgeOrNode, QuantizationSpecBase] = {} + for n in model.graph.nodes: + if hasattr(n, "meta") and "quantization_annotation" in n.meta: + qa = n.meta["quantization_annotation"] + for input_to_n, qspec in qa.input_qspec_map.items(): + input_edge = (input_to_n, n) + edge_or_node_to_qspec[input_edge] = qspec + if qa.output_qspec is not None: + output_node = n + qspec = qa.output_qspec + edge_or_node_to_qspec[output_node] = qspec + return edge_or_node_to_qspec diff --git a/nncf/experimental/torch/fx/quantization/quantize_pt2e.py b/nncf/experimental/torch/fx/quantization/quantize_pt2e.py index efa32af48d6..78ba278826e 100644 --- a/nncf/experimental/torch/fx/quantization/quantize_pt2e.py +++ b/nncf/experimental/torch/fx/quantization/quantize_pt2e.py @@ -25,13 +25,16 @@ from nncf.common.factory import NNCFGraphFactory from nncf.common.logging import nncf_logger from nncf.data import Dataset -from nncf.experimental.common.quantization.algorithms.post_training.algorithm import PostTrainingQuantization +from nncf.experimental.common.quantization.algorithms.post_training.algorithm import ( + ExperimentalPostTrainingQuantization, +) from nncf.experimental.common.quantization.algorithms.quantizer.fx_quantizer import NNCFFXQuantizer +from nncf.experimental.torch.fx.constant_folding import constant_fold +from nncf.experimental.torch.fx.transformations import QUANTIZE_NODE_TARGETS from nncf.experimental.torch.fx.transformations import fuse_conv_bn -from nncf.parameters import ModelType -from nncf.quantization.advanced_parameters import AdvancedQuantizationParameters - -DEFAULT_RANGE_TYPE = "mean_min_max" +from nncf.quantization.advanced_parameters import AdvancedBiasCorrectionParameters +from nncf.quantization.advanced_parameters import AdvancedSmoothQuantParameters +from nncf.quantization.advanced_parameters import RangeEstimatorParameters def quantize_pt2e( @@ -40,8 +43,12 @@ def quantize_pt2e( calibration_dataset: Dataset, subset_size: int = 300, fast_bias_correction: bool = True, - model_type: Optional[ModelType] = None, - advanced_parameters: Optional[AdvancedQuantizationParameters] = None, + smooth_quant: bool = False, + bias_correction_params: Optional[AdvancedBiasCorrectionParameters] = None, + smooth_quant_params: Optional[AdvancedSmoothQuantParameters] = None, + activations_range_estimator_params: Optional[RangeEstimatorParameters] = None, + weights_range_estimator_params: Optional[RangeEstimatorParameters] = None, + fold_quantize: Optional[bool] = False, ) -> torch.fx.GraphModule: """ Implementation of the `quantize()` method for the Torch FX backend. @@ -56,12 +63,15 @@ def quantize_pt2e( copied_model = deepcopy(model) - quantization_algorithm = PostTrainingQuantization( + quantization_algorithm = ExperimentalPostTrainingQuantization( quantizer=NNCFFXQuantizer(quantizer), subset_size=subset_size, fast_bias_correction=fast_bias_correction, - model_type=model_type, - advanced_parameters=advanced_parameters, + smooth_quant=smooth_quant, + bias_correction_params=bias_correction_params, + smooth_quant_params=smooth_quant_params, + activations_range_estimator_params=activations_range_estimator_params, + weights_range_estimator_params=weights_range_estimator_params, ) # To make it easier for bias correction algorithms, @@ -76,6 +86,9 @@ def quantize_pt2e( quantized_model = GraphModule(quantized_model, quantized_model.graph) quantized_model = _fold_conv_bn_qat(quantized_model) + if fold_quantize: + constant_fold(quantized_model, _quant_node_constraint) + pm = PassManager([DuplicateDQPass()]) quantized_model = pm(quantized_model).graph_module @@ -89,3 +102,14 @@ def quantize_pt2e( quantized_model = GraphModule(quantized_model, quantized_model.graph) return quantized_model + + +def _quant_node_constraint(n: torch.fx.Node) -> bool: + """If there is any pure ops between get_attr and quantize op they will be const propagated + e.g. get_attr(weight) -> transpose -> quantize -> dequantize* + (Note: dequantize op is not going to be constant propagated) + + This filter is added because we don't want to constant fold the things that are not + related to quantization + """ + return n.op == "call_function" and n.target in QUANTIZE_NODE_TARGETS diff --git a/nncf/experimental/torch/fx/transformations.py b/nncf/experimental/torch/fx/transformations.py index 6d4ecaf72a8..38541a1ae5f 100644 --- a/nncf/experimental/torch/fx/transformations.py +++ b/nncf/experimental/torch/fx/transformations.py @@ -47,10 +47,12 @@ QUANTIZE_NODE_TARGETS = [ torch.ops.quantized_decomposed.quantize_per_tensor.default, + torch.ops.quantized_decomposed.quantize_per_tensor.tensor, torch.ops.quantized_decomposed.quantize_per_channel.default, ] DEQUANTIZE_NODE_TARGETS = [ torch.ops.quantized_decomposed.dequantize_per_tensor.default, + torch.ops.quantized_decomposed.dequantize_per_tensor.tensor, torch.ops.quantized_decomposed.dequantize_per_channel.default, ] # Map quantize_per_tensor to dequantize_per_tensor, the same for per_channel and vice-versa diff --git a/tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/mobilenet_v3_small.dot b/tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/mobilenet_v3_small.dot new file mode 100644 index 00000000000..6afb1789a12 --- /dev/null +++ b/tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/mobilenet_v3_small.dot @@ -0,0 +1,1082 @@ +strict digraph { +"0 features_1_block_1_fc1_bias" [id=0, type=get_attr]; +"1 features_1_block_1_fc2_bias" [id=1, type=get_attr]; +"2 features_4_block_2_fc1_bias" [id=2, type=get_attr]; +"3 features_4_block_2_fc2_bias" [id=3, type=get_attr]; +"4 features_5_block_2_fc1_bias" [id=4, type=get_attr]; +"5 features_5_block_2_fc2_bias" [id=5, type=get_attr]; +"6 features_6_block_2_fc1_bias" [id=6, type=get_attr]; +"7 features_6_block_2_fc2_bias" [id=7, type=get_attr]; +"8 features_7_block_2_fc1_bias" [id=8, type=get_attr]; +"9 features_7_block_2_fc2_bias" [id=9, type=get_attr]; +"10 features_8_block_2_fc1_bias" [id=10, type=get_attr]; +"11 features_8_block_2_fc2_bias" [id=11, type=get_attr]; +"12 features_9_block_2_fc1_bias" [id=12, type=get_attr]; +"13 features_9_block_2_fc2_bias" [id=13, type=get_attr]; +"14 features_10_block_2_fc1_bias" [id=14, type=get_attr]; +"15 features_10_block_2_fc2_bias" [id=15, type=get_attr]; +"16 features_11_block_2_fc1_bias" [id=16, type=get_attr]; +"17 features_11_block_2_fc2_bias" [id=17, type=get_attr]; +"18 classifier_0_bias" [id=18, type=get_attr]; +"19 classifier_3_bias" [id=19, type=get_attr]; +"20 x" [id=20, type=input]; +"21 quantize_per_tensor_default" [id=21, type=quantize_per_tensor]; +"22 dequantize_per_tensor_default" [id=22, type=dequantize_per_tensor]; +"23 features_0_0_weight_bias" [id=23, type=get_attr]; +"24 conv2d_scale_0" [id=24, type=get_attr]; +"25 conv2d_zero_point_0" [id=25, type=get_attr]; +"26 _frozen_param0" [id=26, type=get_attr]; +"27 dequantize_per_channel_default" [id=27, type=dequantize_per_channel]; +"28 conv2d" [id=28, type=conv2d]; +"29 hardswish_" [id=29, type=hardswish_]; +"30 quantize_per_tensor_default_1" [id=30, type=quantize_per_tensor]; +"31 dequantize_per_tensor_default_1" [id=31, type=dequantize_per_tensor]; +"32 features_1_block_0_0_weight_bias" [id=32, type=get_attr]; +"33 conv2d_1_scale_0" [id=33, type=get_attr]; +"34 conv2d_1_zero_point_0" [id=34, type=get_attr]; +"35 _frozen_param1" [id=35, type=get_attr]; +"36 dequantize_per_channel_default_1" [id=36, type=dequantize_per_channel]; +"37 conv2d_1" [id=37, type=conv2d]; +"38 relu_" [id=38, type=relu_]; +"39 quantize_per_tensor_default_2" [id=39, type=quantize_per_tensor]; +"40 dequantize_per_tensor_default_2" [id=40, type=dequantize_per_tensor]; +"41 adaptive_avg_pool2d" [id=41, type=adaptive_avg_pool2d]; +"42 quantize_per_tensor_default_3" [id=42, type=quantize_per_tensor]; +"43 dequantize_per_tensor_default_3" [id=43, type=dequantize_per_tensor]; +"44 conv2d_2_scale_0" [id=44, type=get_attr]; +"45 conv2d_2_zero_point_0" [id=45, type=get_attr]; +"46 _frozen_param2" [id=46, type=get_attr]; +"47 dequantize_per_channel_default_2" [id=47, type=dequantize_per_channel]; +"48 conv2d_2" [id=48, type=conv2d]; +"49 relu" [id=49, type=relu]; +"50 quantize_per_tensor_default_4" [id=50, type=quantize_per_tensor]; +"51 dequantize_per_tensor_default_4" [id=51, type=dequantize_per_tensor]; +"52 conv2d_3_scale_0" [id=52, type=get_attr]; +"53 conv2d_3_zero_point_0" [id=53, type=get_attr]; +"54 _frozen_param3" [id=54, type=get_attr]; +"55 dequantize_per_channel_default_3" [id=55, type=dequantize_per_channel]; +"56 conv2d_3" [id=56, type=conv2d]; +"57 hardsigmoid" [id=57, type=hardsigmoid]; +"58 mul" [id=58, type=mul]; +"59 quantize_per_tensor_default_5" [id=59, type=quantize_per_tensor]; +"60 dequantize_per_tensor_default_5" [id=60, type=dequantize_per_tensor]; +"61 features_1_block_2_0_weight_bias" [id=61, type=get_attr]; +"62 conv2d_4_scale_0" [id=62, type=get_attr]; +"63 conv2d_4_zero_point_0" [id=63, type=get_attr]; +"64 _frozen_param4" [id=64, type=get_attr]; +"65 dequantize_per_channel_default_4" [id=65, type=dequantize_per_channel]; +"66 conv2d_4" [id=66, type=conv2d]; +"67 quantize_per_tensor_default_6" [id=67, type=quantize_per_tensor]; +"68 dequantize_per_tensor_default_6" [id=68, type=dequantize_per_tensor]; +"69 features_2_block_0_0_weight_bias" [id=69, type=get_attr]; +"70 conv2d_5_scale_0" [id=70, type=get_attr]; +"71 conv2d_5_zero_point_0" [id=71, type=get_attr]; +"72 _frozen_param5" [id=72, type=get_attr]; +"73 dequantize_per_channel_default_5" [id=73, type=dequantize_per_channel]; +"74 conv2d_5" [id=74, type=conv2d]; +"75 relu__1" [id=75, type=relu_]; +"76 quantize_per_tensor_default_7" [id=76, type=quantize_per_tensor]; +"77 dequantize_per_tensor_default_7" [id=77, type=dequantize_per_tensor]; +"78 features_2_block_1_0_weight_bias" [id=78, type=get_attr]; +"79 conv2d_6_scale_0" [id=79, type=get_attr]; +"80 conv2d_6_zero_point_0" [id=80, type=get_attr]; +"81 _frozen_param6" [id=81, type=get_attr]; +"82 dequantize_per_channel_default_6" [id=82, type=dequantize_per_channel]; +"83 conv2d_6" [id=83, type=conv2d]; +"84 relu__2" [id=84, type=relu_]; +"85 quantize_per_tensor_default_8" [id=85, type=quantize_per_tensor]; +"86 dequantize_per_tensor_default_8" [id=86, type=dequantize_per_tensor]; +"87 features_2_block_2_0_weight_bias" [id=87, type=get_attr]; +"88 conv2d_7_scale_0" [id=88, type=get_attr]; +"89 conv2d_7_zero_point_0" [id=89, type=get_attr]; +"90 _frozen_param7" [id=90, type=get_attr]; +"91 dequantize_per_channel_default_7" [id=91, type=dequantize_per_channel]; +"92 conv2d_7" [id=92, type=conv2d]; +"93 quantize_per_tensor_default_9" [id=93, type=quantize_per_tensor]; +"94 dequantize_per_tensor_default_10" [id=94, type=dequantize_per_tensor]; +"95 dequantize_per_tensor_default_9" [id=95, type=dequantize_per_tensor]; +"96 features_3_block_0_0_weight_bias" [id=96, type=get_attr]; +"97 conv2d_8_scale_0" [id=97, type=get_attr]; +"98 conv2d_8_zero_point_0" [id=98, type=get_attr]; +"99 _frozen_param8" [id=99, type=get_attr]; +"100 dequantize_per_channel_default_8" [id=100, type=dequantize_per_channel]; +"101 conv2d_8" [id=101, type=conv2d]; +"102 relu__3" [id=102, type=relu_]; +"103 quantize_per_tensor_default_10" [id=103, type=quantize_per_tensor]; +"104 dequantize_per_tensor_default_11" [id=104, type=dequantize_per_tensor]; +"105 features_3_block_1_0_weight_bias" [id=105, type=get_attr]; +"106 conv2d_9_scale_0" [id=106, type=get_attr]; +"107 conv2d_9_zero_point_0" [id=107, type=get_attr]; +"108 _frozen_param9" [id=108, type=get_attr]; +"109 dequantize_per_channel_default_9" [id=109, type=dequantize_per_channel]; +"110 conv2d_9" [id=110, type=conv2d]; +"111 relu__4" [id=111, type=relu_]; +"112 quantize_per_tensor_default_11" [id=112, type=quantize_per_tensor]; +"113 dequantize_per_tensor_default_12" [id=113, type=dequantize_per_tensor]; +"114 features_3_block_2_0_weight_bias" [id=114, type=get_attr]; +"115 conv2d_10_scale_0" [id=115, type=get_attr]; +"116 conv2d_10_zero_point_0" [id=116, type=get_attr]; +"117 _frozen_param10" [id=117, type=get_attr]; +"118 dequantize_per_channel_default_10" [id=118, type=dequantize_per_channel]; +"119 conv2d_10" [id=119, type=conv2d]; +"120 add_" [id=120, type=add_]; +"121 quantize_per_tensor_default_12" [id=121, type=quantize_per_tensor]; +"122 dequantize_per_tensor_default_13" [id=122, type=dequantize_per_tensor]; +"123 features_4_block_0_0_weight_bias" [id=123, type=get_attr]; +"124 conv2d_11_scale_0" [id=124, type=get_attr]; +"125 conv2d_11_zero_point_0" [id=125, type=get_attr]; +"126 _frozen_param11" [id=126, type=get_attr]; +"127 dequantize_per_channel_default_11" [id=127, type=dequantize_per_channel]; +"128 conv2d_11" [id=128, type=conv2d]; +"129 hardswish__1" [id=129, type=hardswish_]; +"130 quantize_per_tensor_default_13" [id=130, type=quantize_per_tensor]; +"131 dequantize_per_tensor_default_14" [id=131, type=dequantize_per_tensor]; +"132 features_4_block_1_0_weight_bias" [id=132, type=get_attr]; +"133 conv2d_12_scale_0" [id=133, type=get_attr]; +"134 conv2d_12_zero_point_0" [id=134, type=get_attr]; +"135 _frozen_param12" [id=135, type=get_attr]; +"136 dequantize_per_channel_default_12" [id=136, type=dequantize_per_channel]; +"137 conv2d_12" [id=137, type=conv2d]; +"138 hardswish__2" [id=138, type=hardswish_]; +"139 quantize_per_tensor_default_14" [id=139, type=quantize_per_tensor]; +"140 dequantize_per_tensor_default_15" [id=140, type=dequantize_per_tensor]; +"141 adaptive_avg_pool2d_1" [id=141, type=adaptive_avg_pool2d]; +"142 quantize_per_tensor_default_15" [id=142, type=quantize_per_tensor]; +"143 dequantize_per_tensor_default_16" [id=143, type=dequantize_per_tensor]; +"144 conv2d_13_scale_0" [id=144, type=get_attr]; +"145 conv2d_13_zero_point_0" [id=145, type=get_attr]; +"146 _frozen_param13" [id=146, type=get_attr]; +"147 dequantize_per_channel_default_13" [id=147, type=dequantize_per_channel]; +"148 conv2d_13" [id=148, type=conv2d]; +"149 relu_1" [id=149, type=relu]; +"150 quantize_per_tensor_default_16" [id=150, type=quantize_per_tensor]; +"151 dequantize_per_tensor_default_17" [id=151, type=dequantize_per_tensor]; +"152 conv2d_14_scale_0" [id=152, type=get_attr]; +"153 conv2d_14_zero_point_0" [id=153, type=get_attr]; +"154 _frozen_param14" [id=154, type=get_attr]; +"155 dequantize_per_channel_default_14" [id=155, type=dequantize_per_channel]; +"156 conv2d_14" [id=156, type=conv2d]; +"157 hardsigmoid_1" [id=157, type=hardsigmoid]; +"158 mul_1" [id=158, type=mul]; +"159 quantize_per_tensor_default_17" [id=159, type=quantize_per_tensor]; +"160 dequantize_per_tensor_default_18" [id=160, type=dequantize_per_tensor]; +"161 features_4_block_3_0_weight_bias" [id=161, type=get_attr]; +"162 conv2d_15_scale_0" [id=162, type=get_attr]; +"163 conv2d_15_zero_point_0" [id=163, type=get_attr]; +"164 _frozen_param15" [id=164, type=get_attr]; +"165 dequantize_per_channel_default_15" [id=165, type=dequantize_per_channel]; +"166 conv2d_15" [id=166, type=conv2d]; +"167 quantize_per_tensor_default_18" [id=167, type=quantize_per_tensor]; +"168 dequantize_per_tensor_default_20" [id=168, type=dequantize_per_tensor]; +"169 dequantize_per_tensor_default_19" [id=169, type=dequantize_per_tensor]; +"170 features_5_block_0_0_weight_bias" [id=170, type=get_attr]; +"171 conv2d_16_scale_0" [id=171, type=get_attr]; +"172 conv2d_16_zero_point_0" [id=172, type=get_attr]; +"173 _frozen_param16" [id=173, type=get_attr]; +"174 dequantize_per_channel_default_16" [id=174, type=dequantize_per_channel]; +"175 conv2d_16" [id=175, type=conv2d]; +"176 hardswish__3" [id=176, type=hardswish_]; +"177 quantize_per_tensor_default_19" [id=177, type=quantize_per_tensor]; +"178 dequantize_per_tensor_default_21" [id=178, type=dequantize_per_tensor]; +"179 features_5_block_1_0_weight_bias" [id=179, type=get_attr]; +"180 conv2d_17_scale_0" [id=180, type=get_attr]; +"181 conv2d_17_zero_point_0" [id=181, type=get_attr]; +"182 _frozen_param17" [id=182, type=get_attr]; +"183 dequantize_per_channel_default_17" [id=183, type=dequantize_per_channel]; +"184 conv2d_17" [id=184, type=conv2d]; +"185 hardswish__4" [id=185, type=hardswish_]; +"186 quantize_per_tensor_default_20" [id=186, type=quantize_per_tensor]; +"187 dequantize_per_tensor_default_22" [id=187, type=dequantize_per_tensor]; +"188 adaptive_avg_pool2d_2" [id=188, type=adaptive_avg_pool2d]; +"189 quantize_per_tensor_default_21" [id=189, type=quantize_per_tensor]; +"190 dequantize_per_tensor_default_23" [id=190, type=dequantize_per_tensor]; +"191 conv2d_18_scale_0" [id=191, type=get_attr]; +"192 conv2d_18_zero_point_0" [id=192, type=get_attr]; +"193 _frozen_param18" [id=193, type=get_attr]; +"194 dequantize_per_channel_default_18" [id=194, type=dequantize_per_channel]; +"195 conv2d_18" [id=195, type=conv2d]; +"196 relu_2" [id=196, type=relu]; +"197 quantize_per_tensor_default_22" [id=197, type=quantize_per_tensor]; +"198 dequantize_per_tensor_default_24" [id=198, type=dequantize_per_tensor]; +"199 conv2d_19_scale_0" [id=199, type=get_attr]; +"200 conv2d_19_zero_point_0" [id=200, type=get_attr]; +"201 _frozen_param19" [id=201, type=get_attr]; +"202 dequantize_per_channel_default_19" [id=202, type=dequantize_per_channel]; +"203 conv2d_19" [id=203, type=conv2d]; +"204 hardsigmoid_2" [id=204, type=hardsigmoid]; +"205 mul_2" [id=205, type=mul]; +"206 quantize_per_tensor_default_23" [id=206, type=quantize_per_tensor]; +"207 dequantize_per_tensor_default_25" [id=207, type=dequantize_per_tensor]; +"208 features_5_block_3_0_weight_bias" [id=208, type=get_attr]; +"209 conv2d_20_scale_0" [id=209, type=get_attr]; +"210 conv2d_20_zero_point_0" [id=210, type=get_attr]; +"211 _frozen_param20" [id=211, type=get_attr]; +"212 dequantize_per_channel_default_20" [id=212, type=dequantize_per_channel]; +"213 conv2d_20" [id=213, type=conv2d]; +"214 add__1" [id=214, type=add_]; +"215 quantize_per_tensor_default_24" [id=215, type=quantize_per_tensor]; +"216 dequantize_per_tensor_default_27" [id=216, type=dequantize_per_tensor]; +"217 dequantize_per_tensor_default_26" [id=217, type=dequantize_per_tensor]; +"218 features_6_block_0_0_weight_bias" [id=218, type=get_attr]; +"219 conv2d_21_scale_0" [id=219, type=get_attr]; +"220 conv2d_21_zero_point_0" [id=220, type=get_attr]; +"221 _frozen_param21" [id=221, type=get_attr]; +"222 dequantize_per_channel_default_21" [id=222, type=dequantize_per_channel]; +"223 conv2d_21" [id=223, type=conv2d]; +"224 hardswish__5" [id=224, type=hardswish_]; +"225 quantize_per_tensor_default_25" [id=225, type=quantize_per_tensor]; +"226 dequantize_per_tensor_default_28" [id=226, type=dequantize_per_tensor]; +"227 features_6_block_1_0_weight_bias" [id=227, type=get_attr]; +"228 conv2d_22_scale_0" [id=228, type=get_attr]; +"229 conv2d_22_zero_point_0" [id=229, type=get_attr]; +"230 _frozen_param22" [id=230, type=get_attr]; +"231 dequantize_per_channel_default_22" [id=231, type=dequantize_per_channel]; +"232 conv2d_22" [id=232, type=conv2d]; +"233 hardswish__6" [id=233, type=hardswish_]; +"234 quantize_per_tensor_default_26" [id=234, type=quantize_per_tensor]; +"235 dequantize_per_tensor_default_29" [id=235, type=dequantize_per_tensor]; +"236 adaptive_avg_pool2d_3" [id=236, type=adaptive_avg_pool2d]; +"237 quantize_per_tensor_default_27" [id=237, type=quantize_per_tensor]; +"238 dequantize_per_tensor_default_30" [id=238, type=dequantize_per_tensor]; +"239 conv2d_23_scale_0" [id=239, type=get_attr]; +"240 conv2d_23_zero_point_0" [id=240, type=get_attr]; +"241 _frozen_param23" [id=241, type=get_attr]; +"242 dequantize_per_channel_default_23" [id=242, type=dequantize_per_channel]; +"243 conv2d_23" [id=243, type=conv2d]; +"244 relu_3" [id=244, type=relu]; +"245 quantize_per_tensor_default_28" [id=245, type=quantize_per_tensor]; +"246 dequantize_per_tensor_default_31" [id=246, type=dequantize_per_tensor]; +"247 conv2d_24_scale_0" [id=247, type=get_attr]; +"248 conv2d_24_zero_point_0" [id=248, type=get_attr]; +"249 _frozen_param24" [id=249, type=get_attr]; +"250 dequantize_per_channel_default_24" [id=250, type=dequantize_per_channel]; +"251 conv2d_24" [id=251, type=conv2d]; +"252 hardsigmoid_3" [id=252, type=hardsigmoid]; +"253 mul_3" [id=253, type=mul]; +"254 quantize_per_tensor_default_29" [id=254, type=quantize_per_tensor]; +"255 dequantize_per_tensor_default_32" [id=255, type=dequantize_per_tensor]; +"256 features_6_block_3_0_weight_bias" [id=256, type=get_attr]; +"257 conv2d_25_scale_0" [id=257, type=get_attr]; +"258 conv2d_25_zero_point_0" [id=258, type=get_attr]; +"259 _frozen_param25" [id=259, type=get_attr]; +"260 dequantize_per_channel_default_25" [id=260, type=dequantize_per_channel]; +"261 conv2d_25" [id=261, type=conv2d]; +"262 add__2" [id=262, type=add_]; +"263 quantize_per_tensor_default_30" [id=263, type=quantize_per_tensor]; +"264 dequantize_per_tensor_default_33" [id=264, type=dequantize_per_tensor]; +"265 features_7_block_0_0_weight_bias" [id=265, type=get_attr]; +"266 conv2d_26_scale_0" [id=266, type=get_attr]; +"267 conv2d_26_zero_point_0" [id=267, type=get_attr]; +"268 _frozen_param26" [id=268, type=get_attr]; +"269 dequantize_per_channel_default_26" [id=269, type=dequantize_per_channel]; +"270 conv2d_26" [id=270, type=conv2d]; +"271 hardswish__7" [id=271, type=hardswish_]; +"272 quantize_per_tensor_default_31" [id=272, type=quantize_per_tensor]; +"273 dequantize_per_tensor_default_34" [id=273, type=dequantize_per_tensor]; +"274 features_7_block_1_0_weight_bias" [id=274, type=get_attr]; +"275 conv2d_27_scale_0" [id=275, type=get_attr]; +"276 conv2d_27_zero_point_0" [id=276, type=get_attr]; +"277 _frozen_param27" [id=277, type=get_attr]; +"278 dequantize_per_channel_default_27" [id=278, type=dequantize_per_channel]; +"279 conv2d_27" [id=279, type=conv2d]; +"280 hardswish__8" [id=280, type=hardswish_]; +"281 quantize_per_tensor_default_32" [id=281, type=quantize_per_tensor]; +"282 dequantize_per_tensor_default_35" [id=282, type=dequantize_per_tensor]; +"283 adaptive_avg_pool2d_4" [id=283, type=adaptive_avg_pool2d]; +"284 quantize_per_tensor_default_33" [id=284, type=quantize_per_tensor]; +"285 dequantize_per_tensor_default_36" [id=285, type=dequantize_per_tensor]; +"286 conv2d_28_scale_0" [id=286, type=get_attr]; +"287 conv2d_28_zero_point_0" [id=287, type=get_attr]; +"288 _frozen_param28" [id=288, type=get_attr]; +"289 dequantize_per_channel_default_28" [id=289, type=dequantize_per_channel]; +"290 conv2d_28" [id=290, type=conv2d]; +"291 relu_4" [id=291, type=relu]; +"292 quantize_per_tensor_default_34" [id=292, type=quantize_per_tensor]; +"293 dequantize_per_tensor_default_37" [id=293, type=dequantize_per_tensor]; +"294 conv2d_29_scale_0" [id=294, type=get_attr]; +"295 conv2d_29_zero_point_0" [id=295, type=get_attr]; +"296 _frozen_param29" [id=296, type=get_attr]; +"297 dequantize_per_channel_default_29" [id=297, type=dequantize_per_channel]; +"298 conv2d_29" [id=298, type=conv2d]; +"299 hardsigmoid_4" [id=299, type=hardsigmoid]; +"300 mul_4" [id=300, type=mul]; +"301 quantize_per_tensor_default_35" [id=301, type=quantize_per_tensor]; +"302 dequantize_per_tensor_default_38" [id=302, type=dequantize_per_tensor]; +"303 features_7_block_3_0_weight_bias" [id=303, type=get_attr]; +"304 conv2d_30_scale_0" [id=304, type=get_attr]; +"305 conv2d_30_zero_point_0" [id=305, type=get_attr]; +"306 _frozen_param30" [id=306, type=get_attr]; +"307 dequantize_per_channel_default_30" [id=307, type=dequantize_per_channel]; +"308 conv2d_30" [id=308, type=conv2d]; +"309 quantize_per_tensor_default_36" [id=309, type=quantize_per_tensor]; +"310 dequantize_per_tensor_default_40" [id=310, type=dequantize_per_tensor]; +"311 dequantize_per_tensor_default_39" [id=311, type=dequantize_per_tensor]; +"312 features_8_block_0_0_weight_bias" [id=312, type=get_attr]; +"313 conv2d_31_scale_0" [id=313, type=get_attr]; +"314 conv2d_31_zero_point_0" [id=314, type=get_attr]; +"315 _frozen_param31" [id=315, type=get_attr]; +"316 dequantize_per_channel_default_31" [id=316, type=dequantize_per_channel]; +"317 conv2d_31" [id=317, type=conv2d]; +"318 hardswish__9" [id=318, type=hardswish_]; +"319 quantize_per_tensor_default_37" [id=319, type=quantize_per_tensor]; +"320 dequantize_per_tensor_default_41" [id=320, type=dequantize_per_tensor]; +"321 features_8_block_1_0_weight_bias" [id=321, type=get_attr]; +"322 conv2d_32_scale_0" [id=322, type=get_attr]; +"323 conv2d_32_zero_point_0" [id=323, type=get_attr]; +"324 _frozen_param32" [id=324, type=get_attr]; +"325 dequantize_per_channel_default_32" [id=325, type=dequantize_per_channel]; +"326 conv2d_32" [id=326, type=conv2d]; +"327 hardswish__10" [id=327, type=hardswish_]; +"328 quantize_per_tensor_default_38" [id=328, type=quantize_per_tensor]; +"329 dequantize_per_tensor_default_42" [id=329, type=dequantize_per_tensor]; +"330 adaptive_avg_pool2d_5" [id=330, type=adaptive_avg_pool2d]; +"331 quantize_per_tensor_default_39" [id=331, type=quantize_per_tensor]; +"332 dequantize_per_tensor_default_43" [id=332, type=dequantize_per_tensor]; +"333 conv2d_33_scale_0" [id=333, type=get_attr]; +"334 conv2d_33_zero_point_0" [id=334, type=get_attr]; +"335 _frozen_param33" [id=335, type=get_attr]; +"336 dequantize_per_channel_default_33" [id=336, type=dequantize_per_channel]; +"337 conv2d_33" [id=337, type=conv2d]; +"338 relu_5" [id=338, type=relu]; +"339 quantize_per_tensor_default_40" [id=339, type=quantize_per_tensor]; +"340 dequantize_per_tensor_default_44" [id=340, type=dequantize_per_tensor]; +"341 conv2d_34_scale_0" [id=341, type=get_attr]; +"342 conv2d_34_zero_point_0" [id=342, type=get_attr]; +"343 _frozen_param34" [id=343, type=get_attr]; +"344 dequantize_per_channel_default_34" [id=344, type=dequantize_per_channel]; +"345 conv2d_34" [id=345, type=conv2d]; +"346 hardsigmoid_5" [id=346, type=hardsigmoid]; +"347 mul_5" [id=347, type=mul]; +"348 quantize_per_tensor_default_41" [id=348, type=quantize_per_tensor]; +"349 dequantize_per_tensor_default_45" [id=349, type=dequantize_per_tensor]; +"350 features_8_block_3_0_weight_bias" [id=350, type=get_attr]; +"351 conv2d_35_scale_0" [id=351, type=get_attr]; +"352 conv2d_35_zero_point_0" [id=352, type=get_attr]; +"353 _frozen_param35" [id=353, type=get_attr]; +"354 dequantize_per_channel_default_35" [id=354, type=dequantize_per_channel]; +"355 conv2d_35" [id=355, type=conv2d]; +"356 add__3" [id=356, type=add_]; +"357 quantize_per_tensor_default_42" [id=357, type=quantize_per_tensor]; +"358 dequantize_per_tensor_default_46" [id=358, type=dequantize_per_tensor]; +"359 features_9_block_0_0_weight_bias" [id=359, type=get_attr]; +"360 conv2d_36_scale_0" [id=360, type=get_attr]; +"361 conv2d_36_zero_point_0" [id=361, type=get_attr]; +"362 _frozen_param36" [id=362, type=get_attr]; +"363 dequantize_per_channel_default_36" [id=363, type=dequantize_per_channel]; +"364 conv2d_36" [id=364, type=conv2d]; +"365 hardswish__11" [id=365, type=hardswish_]; +"366 quantize_per_tensor_default_43" [id=366, type=quantize_per_tensor]; +"367 dequantize_per_tensor_default_47" [id=367, type=dequantize_per_tensor]; +"368 features_9_block_1_0_weight_bias" [id=368, type=get_attr]; +"369 conv2d_37_scale_0" [id=369, type=get_attr]; +"370 conv2d_37_zero_point_0" [id=370, type=get_attr]; +"371 _frozen_param37" [id=371, type=get_attr]; +"372 dequantize_per_channel_default_37" [id=372, type=dequantize_per_channel]; +"373 conv2d_37" [id=373, type=conv2d]; +"374 hardswish__12" [id=374, type=hardswish_]; +"375 quantize_per_tensor_default_44" [id=375, type=quantize_per_tensor]; +"376 dequantize_per_tensor_default_48" [id=376, type=dequantize_per_tensor]; +"377 adaptive_avg_pool2d_6" [id=377, type=adaptive_avg_pool2d]; +"378 quantize_per_tensor_default_45" [id=378, type=quantize_per_tensor]; +"379 dequantize_per_tensor_default_49" [id=379, type=dequantize_per_tensor]; +"380 conv2d_38_scale_0" [id=380, type=get_attr]; +"381 conv2d_38_zero_point_0" [id=381, type=get_attr]; +"382 _frozen_param38" [id=382, type=get_attr]; +"383 dequantize_per_channel_default_38" [id=383, type=dequantize_per_channel]; +"384 conv2d_38" [id=384, type=conv2d]; +"385 relu_6" [id=385, type=relu]; +"386 quantize_per_tensor_default_46" [id=386, type=quantize_per_tensor]; +"387 dequantize_per_tensor_default_50" [id=387, type=dequantize_per_tensor]; +"388 conv2d_39_scale_0" [id=388, type=get_attr]; +"389 conv2d_39_zero_point_0" [id=389, type=get_attr]; +"390 _frozen_param39" [id=390, type=get_attr]; +"391 dequantize_per_channel_default_39" [id=391, type=dequantize_per_channel]; +"392 conv2d_39" [id=392, type=conv2d]; +"393 hardsigmoid_6" [id=393, type=hardsigmoid]; +"394 mul_6" [id=394, type=mul]; +"395 quantize_per_tensor_default_47" [id=395, type=quantize_per_tensor]; +"396 dequantize_per_tensor_default_51" [id=396, type=dequantize_per_tensor]; +"397 features_9_block_3_0_weight_bias" [id=397, type=get_attr]; +"398 conv2d_40_scale_0" [id=398, type=get_attr]; +"399 conv2d_40_zero_point_0" [id=399, type=get_attr]; +"400 _frozen_param40" [id=400, type=get_attr]; +"401 dequantize_per_channel_default_40" [id=401, type=dequantize_per_channel]; +"402 conv2d_40" [id=402, type=conv2d]; +"403 quantize_per_tensor_default_48" [id=403, type=quantize_per_tensor]; +"404 dequantize_per_tensor_default_53" [id=404, type=dequantize_per_tensor]; +"405 dequantize_per_tensor_default_52" [id=405, type=dequantize_per_tensor]; +"406 features_10_block_0_0_weight_bias" [id=406, type=get_attr]; +"407 conv2d_41_scale_0" [id=407, type=get_attr]; +"408 conv2d_41_zero_point_0" [id=408, type=get_attr]; +"409 _frozen_param41" [id=409, type=get_attr]; +"410 dequantize_per_channel_default_41" [id=410, type=dequantize_per_channel]; +"411 conv2d_41" [id=411, type=conv2d]; +"412 hardswish__13" [id=412, type=hardswish_]; +"413 quantize_per_tensor_default_49" [id=413, type=quantize_per_tensor]; +"414 dequantize_per_tensor_default_54" [id=414, type=dequantize_per_tensor]; +"415 features_10_block_1_0_weight_bias" [id=415, type=get_attr]; +"416 conv2d_42_scale_0" [id=416, type=get_attr]; +"417 conv2d_42_zero_point_0" [id=417, type=get_attr]; +"418 _frozen_param42" [id=418, type=get_attr]; +"419 dequantize_per_channel_default_42" [id=419, type=dequantize_per_channel]; +"420 conv2d_42" [id=420, type=conv2d]; +"421 hardswish__14" [id=421, type=hardswish_]; +"422 quantize_per_tensor_default_50" [id=422, type=quantize_per_tensor]; +"423 dequantize_per_tensor_default_55" [id=423, type=dequantize_per_tensor]; +"424 adaptive_avg_pool2d_7" [id=424, type=adaptive_avg_pool2d]; +"425 quantize_per_tensor_default_51" [id=425, type=quantize_per_tensor]; +"426 dequantize_per_tensor_default_56" [id=426, type=dequantize_per_tensor]; +"427 conv2d_43_scale_0" [id=427, type=get_attr]; +"428 conv2d_43_zero_point_0" [id=428, type=get_attr]; +"429 _frozen_param43" [id=429, type=get_attr]; +"430 dequantize_per_channel_default_43" [id=430, type=dequantize_per_channel]; +"431 conv2d_43" [id=431, type=conv2d]; +"432 relu_7" [id=432, type=relu]; +"433 quantize_per_tensor_default_52" [id=433, type=quantize_per_tensor]; +"434 dequantize_per_tensor_default_57" [id=434, type=dequantize_per_tensor]; +"435 conv2d_44_scale_0" [id=435, type=get_attr]; +"436 conv2d_44_zero_point_0" [id=436, type=get_attr]; +"437 _frozen_param44" [id=437, type=get_attr]; +"438 dequantize_per_channel_default_44" [id=438, type=dequantize_per_channel]; +"439 conv2d_44" [id=439, type=conv2d]; +"440 hardsigmoid_7" [id=440, type=hardsigmoid]; +"441 mul_7" [id=441, type=mul]; +"442 quantize_per_tensor_default_53" [id=442, type=quantize_per_tensor]; +"443 dequantize_per_tensor_default_58" [id=443, type=dequantize_per_tensor]; +"444 features_10_block_3_0_weight_bias" [id=444, type=get_attr]; +"445 conv2d_45_scale_0" [id=445, type=get_attr]; +"446 conv2d_45_zero_point_0" [id=446, type=get_attr]; +"447 _frozen_param45" [id=447, type=get_attr]; +"448 dequantize_per_channel_default_45" [id=448, type=dequantize_per_channel]; +"449 conv2d_45" [id=449, type=conv2d]; +"450 add__4" [id=450, type=add_]; +"451 quantize_per_tensor_default_54" [id=451, type=quantize_per_tensor]; +"452 dequantize_per_tensor_default_60" [id=452, type=dequantize_per_tensor]; +"453 dequantize_per_tensor_default_59" [id=453, type=dequantize_per_tensor]; +"454 features_11_block_0_0_weight_bias" [id=454, type=get_attr]; +"455 conv2d_46_scale_0" [id=455, type=get_attr]; +"456 conv2d_46_zero_point_0" [id=456, type=get_attr]; +"457 _frozen_param46" [id=457, type=get_attr]; +"458 dequantize_per_channel_default_46" [id=458, type=dequantize_per_channel]; +"459 conv2d_46" [id=459, type=conv2d]; +"460 hardswish__15" [id=460, type=hardswish_]; +"461 quantize_per_tensor_default_55" [id=461, type=quantize_per_tensor]; +"462 dequantize_per_tensor_default_61" [id=462, type=dequantize_per_tensor]; +"463 features_11_block_1_0_weight_bias" [id=463, type=get_attr]; +"464 conv2d_47_scale_0" [id=464, type=get_attr]; +"465 conv2d_47_zero_point_0" [id=465, type=get_attr]; +"466 _frozen_param47" [id=466, type=get_attr]; +"467 dequantize_per_channel_default_47" [id=467, type=dequantize_per_channel]; +"468 conv2d_47" [id=468, type=conv2d]; +"469 hardswish__16" [id=469, type=hardswish_]; +"470 quantize_per_tensor_default_56" [id=470, type=quantize_per_tensor]; +"471 dequantize_per_tensor_default_62" [id=471, type=dequantize_per_tensor]; +"472 adaptive_avg_pool2d_8" [id=472, type=adaptive_avg_pool2d]; +"473 quantize_per_tensor_default_57" [id=473, type=quantize_per_tensor]; +"474 dequantize_per_tensor_default_63" [id=474, type=dequantize_per_tensor]; +"475 conv2d_48_scale_0" [id=475, type=get_attr]; +"476 conv2d_48_zero_point_0" [id=476, type=get_attr]; +"477 _frozen_param48" [id=477, type=get_attr]; +"478 dequantize_per_channel_default_48" [id=478, type=dequantize_per_channel]; +"479 conv2d_48" [id=479, type=conv2d]; +"480 relu_8" [id=480, type=relu]; +"481 quantize_per_tensor_default_58" [id=481, type=quantize_per_tensor]; +"482 dequantize_per_tensor_default_64" [id=482, type=dequantize_per_tensor]; +"483 conv2d_49_scale_0" [id=483, type=get_attr]; +"484 conv2d_49_zero_point_0" [id=484, type=get_attr]; +"485 _frozen_param49" [id=485, type=get_attr]; +"486 dequantize_per_channel_default_49" [id=486, type=dequantize_per_channel]; +"487 conv2d_49" [id=487, type=conv2d]; +"488 hardsigmoid_8" [id=488, type=hardsigmoid]; +"489 mul_8" [id=489, type=mul]; +"490 quantize_per_tensor_default_59" [id=490, type=quantize_per_tensor]; +"491 dequantize_per_tensor_default_65" [id=491, type=dequantize_per_tensor]; +"492 features_11_block_3_0_weight_bias" [id=492, type=get_attr]; +"493 conv2d_50_scale_0" [id=493, type=get_attr]; +"494 conv2d_50_zero_point_0" [id=494, type=get_attr]; +"495 _frozen_param50" [id=495, type=get_attr]; +"496 dequantize_per_channel_default_50" [id=496, type=dequantize_per_channel]; +"497 conv2d_50" [id=497, type=conv2d]; +"498 add__5" [id=498, type=add_]; +"499 quantize_per_tensor_default_60" [id=499, type=quantize_per_tensor]; +"500 dequantize_per_tensor_default_66" [id=500, type=dequantize_per_tensor]; +"501 features_12_0_weight_bias" [id=501, type=get_attr]; +"502 conv2d_51_scale_0" [id=502, type=get_attr]; +"503 conv2d_51_zero_point_0" [id=503, type=get_attr]; +"504 _frozen_param51" [id=504, type=get_attr]; +"505 dequantize_per_channel_default_51" [id=505, type=dequantize_per_channel]; +"506 conv2d_51" [id=506, type=conv2d]; +"507 hardswish__17" [id=507, type=hardswish_]; +"508 quantize_per_tensor_default_61" [id=508, type=quantize_per_tensor]; +"509 dequantize_per_tensor_default_67" [id=509, type=dequantize_per_tensor]; +"510 adaptive_avg_pool2d_9" [id=510, type=adaptive_avg_pool2d]; +"511 quantize_per_tensor_default_62" [id=511, type=quantize_per_tensor]; +"512 dequantize_per_tensor_default_68" [id=512, type=dequantize_per_tensor]; +"513 flatten" [id=513, type=flatten]; +"514 flatten_0_0_nncf_smooth_quant_0" [id=514, type=call_module]; +"515 quantize_per_tensor_default_63" [id=515, type=quantize_per_tensor]; +"516 dequantize_per_tensor_default_69" [id=516, type=dequantize_per_tensor]; +"517 linear_scale_0" [id=517, type=get_attr]; +"518 linear_zero_point_0" [id=518, type=get_attr]; +"519 _frozen_param52" [id=519, type=get_attr]; +"520 dequantize_per_channel_default_52" [id=520, type=dequantize_per_channel]; +"521 linear" [id=521, type=linear]; +"522 hardswish__18" [id=522, type=hardswish_]; +"523 dropout_" [id=523, type=dropout_]; +"524 dropout__0_0_nncf_smooth_quant_0" [id=524, type=call_module]; +"525 quantize_per_tensor_default_64" [id=525, type=quantize_per_tensor]; +"526 dequantize_per_tensor_default_70" [id=526, type=dequantize_per_tensor]; +"527 linear_1_scale_0" [id=527, type=get_attr]; +"528 linear_1_zero_point_0" [id=528, type=get_attr]; +"529 _frozen_param53" [id=529, type=get_attr]; +"530 dequantize_per_channel_default_53" [id=530, type=dequantize_per_channel]; +"531 linear_1" [id=531, type=linear]; +"532 output" [id=532, type=output]; +"0 features_1_block_1_fc1_bias" -> "48 conv2d_2" [label="(8,)", style=solid]; +"1 features_1_block_1_fc2_bias" -> "56 conv2d_3" [label="(16,)", style=solid]; +"2 features_4_block_2_fc1_bias" -> "148 conv2d_13" [label="(24,)", style=solid]; +"3 features_4_block_2_fc2_bias" -> "156 conv2d_14" [label="(96,)", style=solid]; +"4 features_5_block_2_fc1_bias" -> "195 conv2d_18" [label="(64,)", style=solid]; +"5 features_5_block_2_fc2_bias" -> "203 conv2d_19" [label="(240,)", style=solid]; +"6 features_6_block_2_fc1_bias" -> "243 conv2d_23" [label="(64,)", style=solid]; +"7 features_6_block_2_fc2_bias" -> "251 conv2d_24" [label="(240,)", style=solid]; +"8 features_7_block_2_fc1_bias" -> "290 conv2d_28" [label="(32,)", style=solid]; +"9 features_7_block_2_fc2_bias" -> "298 conv2d_29" [label="(120,)", style=solid]; +"10 features_8_block_2_fc1_bias" -> "337 conv2d_33" [label="(40,)", style=solid]; +"11 features_8_block_2_fc2_bias" -> "345 conv2d_34" [label="(144,)", style=solid]; +"12 features_9_block_2_fc1_bias" -> "384 conv2d_38" [label="(72,)", style=solid]; +"13 features_9_block_2_fc2_bias" -> "392 conv2d_39" [label="(288,)", style=solid]; +"14 features_10_block_2_fc1_bias" -> "431 conv2d_43" [label="(144,)", style=solid]; +"15 features_10_block_2_fc2_bias" -> "439 conv2d_44" [label="(576,)", style=solid]; +"16 features_11_block_2_fc1_bias" -> "479 conv2d_48" [label="(144,)", style=solid]; +"17 features_11_block_2_fc2_bias" -> "487 conv2d_49" [label="(576,)", style=solid]; +"18 classifier_0_bias" -> "521 linear" [label="(1024,)", style=solid]; +"19 classifier_3_bias" -> "531 linear_1" [label="(1000,)", style=solid]; +"20 x" -> "21 quantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; +"21 quantize_per_tensor_default" -> "22 dequantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; +"22 dequantize_per_tensor_default" -> "28 conv2d" [label="(1, 3, 224, 224)", style=solid]; +"23 features_0_0_weight_bias" -> "28 conv2d" [label="(16,)", style=solid]; +"24 conv2d_scale_0" -> "27 dequantize_per_channel_default" [label="(16,)", style=solid]; +"25 conv2d_zero_point_0" -> "27 dequantize_per_channel_default" [label="(16,)", style=solid]; +"26 _frozen_param0" -> "27 dequantize_per_channel_default" [label="(16, 3, 3, 3)", style=solid]; +"27 dequantize_per_channel_default" -> "28 conv2d" [label="(16, 3, 3, 3)", style=solid]; +"28 conv2d" -> "29 hardswish_" [label="(1, 16, 112, 112)", style=solid]; +"29 hardswish_" -> "30 quantize_per_tensor_default_1" [label="(1, 16, 112, 112)", style=solid]; +"30 quantize_per_tensor_default_1" -> "31 dequantize_per_tensor_default_1" [label="(1, 16, 112, 112)", style=solid]; +"31 dequantize_per_tensor_default_1" -> "37 conv2d_1" [label="(1, 16, 112, 112)", style=solid]; +"32 features_1_block_0_0_weight_bias" -> "37 conv2d_1" [label="(16,)", style=solid]; +"33 conv2d_1_scale_0" -> "36 dequantize_per_channel_default_1" [label="(16,)", style=solid]; +"34 conv2d_1_zero_point_0" -> "36 dequantize_per_channel_default_1" [label="(16,)", style=solid]; +"35 _frozen_param1" -> "36 dequantize_per_channel_default_1" [label="(16, 1, 3, 3)", style=solid]; +"36 dequantize_per_channel_default_1" -> "37 conv2d_1" [label="(16, 1, 3, 3)", style=solid]; +"37 conv2d_1" -> "38 relu_" [label="(1, 16, 56, 56)", style=solid]; +"38 relu_" -> "39 quantize_per_tensor_default_2" [label="(1, 16, 56, 56)", style=solid]; +"38 relu_" -> "58 mul" [label="(1, 16, 56, 56)", style=solid]; +"39 quantize_per_tensor_default_2" -> "40 dequantize_per_tensor_default_2" [label="(1, 16, 56, 56)", style=solid]; +"40 dequantize_per_tensor_default_2" -> "41 adaptive_avg_pool2d" [label="(1, 16, 56, 56)", style=solid]; +"41 adaptive_avg_pool2d" -> "42 quantize_per_tensor_default_3" [label="(1, 16, 1, 1)", style=solid]; +"42 quantize_per_tensor_default_3" -> "43 dequantize_per_tensor_default_3" [label="(1, 16, 1, 1)", style=solid]; +"43 dequantize_per_tensor_default_3" -> "48 conv2d_2" [label="(1, 16, 1, 1)", style=solid]; +"44 conv2d_2_scale_0" -> "47 dequantize_per_channel_default_2" [label="(8,)", style=solid]; +"45 conv2d_2_zero_point_0" -> "47 dequantize_per_channel_default_2" [label="(8,)", style=solid]; +"46 _frozen_param2" -> "47 dequantize_per_channel_default_2" [label="(8, 16, 1, 1)", style=solid]; +"47 dequantize_per_channel_default_2" -> "48 conv2d_2" [label="(8, 16, 1, 1)", style=solid]; +"48 conv2d_2" -> "49 relu" [label="(1, 8, 1, 1)", style=solid]; +"49 relu" -> "50 quantize_per_tensor_default_4" [label="(1, 8, 1, 1)", style=solid]; +"50 quantize_per_tensor_default_4" -> "51 dequantize_per_tensor_default_4" [label="(1, 8, 1, 1)", style=solid]; +"51 dequantize_per_tensor_default_4" -> "56 conv2d_3" [label="(1, 8, 1, 1)", style=solid]; +"52 conv2d_3_scale_0" -> "55 dequantize_per_channel_default_3" [label="(16,)", style=solid]; +"53 conv2d_3_zero_point_0" -> "55 dequantize_per_channel_default_3" [label="(16,)", style=solid]; +"54 _frozen_param3" -> "55 dequantize_per_channel_default_3" [label="(16, 8, 1, 1)", style=solid]; +"55 dequantize_per_channel_default_3" -> "56 conv2d_3" [label="(16, 8, 1, 1)", style=solid]; +"56 conv2d_3" -> "57 hardsigmoid" [label="(1, 16, 1, 1)", style=solid]; +"57 hardsigmoid" -> "58 mul" [label="(1, 16, 1, 1)", style=solid]; +"58 mul" -> "59 quantize_per_tensor_default_5" [label="(1, 16, 56, 56)", style=solid]; +"59 quantize_per_tensor_default_5" -> "60 dequantize_per_tensor_default_5" [label="(1, 16, 56, 56)", style=solid]; +"60 dequantize_per_tensor_default_5" -> "66 conv2d_4" [label="(1, 16, 56, 56)", style=solid]; +"61 features_1_block_2_0_weight_bias" -> "66 conv2d_4" [label="(16,)", style=solid]; +"62 conv2d_4_scale_0" -> "65 dequantize_per_channel_default_4" [label="(16,)", style=solid]; +"63 conv2d_4_zero_point_0" -> "65 dequantize_per_channel_default_4" [label="(16,)", style=solid]; +"64 _frozen_param4" -> "65 dequantize_per_channel_default_4" [label="(16, 16, 1, 1)", style=solid]; +"65 dequantize_per_channel_default_4" -> "66 conv2d_4" [label="(16, 16, 1, 1)", style=solid]; +"66 conv2d_4" -> "67 quantize_per_tensor_default_6" [label="(1, 16, 56, 56)", style=solid]; +"67 quantize_per_tensor_default_6" -> "68 dequantize_per_tensor_default_6" [label="(1, 16, 56, 56)", style=solid]; +"68 dequantize_per_tensor_default_6" -> "74 conv2d_5" [label="(1, 16, 56, 56)", style=solid]; +"69 features_2_block_0_0_weight_bias" -> "74 conv2d_5" [label="(72,)", style=solid]; +"70 conv2d_5_scale_0" -> "73 dequantize_per_channel_default_5" [label="(72,)", style=solid]; +"71 conv2d_5_zero_point_0" -> "73 dequantize_per_channel_default_5" [label="(72,)", style=solid]; +"72 _frozen_param5" -> "73 dequantize_per_channel_default_5" [label="(72, 16, 1, 1)", style=solid]; +"73 dequantize_per_channel_default_5" -> "74 conv2d_5" [label="(72, 16, 1, 1)", style=solid]; +"74 conv2d_5" -> "75 relu__1" [label="(1, 72, 56, 56)", style=solid]; +"75 relu__1" -> "76 quantize_per_tensor_default_7" [label="(1, 72, 56, 56)", style=solid]; +"76 quantize_per_tensor_default_7" -> "77 dequantize_per_tensor_default_7" [label="(1, 72, 56, 56)", style=solid]; +"77 dequantize_per_tensor_default_7" -> "83 conv2d_6" [label="(1, 72, 56, 56)", style=solid]; +"78 features_2_block_1_0_weight_bias" -> "83 conv2d_6" [label="(72,)", style=solid]; +"79 conv2d_6_scale_0" -> "82 dequantize_per_channel_default_6" [label="(72,)", style=solid]; +"80 conv2d_6_zero_point_0" -> "82 dequantize_per_channel_default_6" [label="(72,)", style=solid]; +"81 _frozen_param6" -> "82 dequantize_per_channel_default_6" [label="(72, 1, 3, 3)", style=solid]; +"82 dequantize_per_channel_default_6" -> "83 conv2d_6" [label="(72, 1, 3, 3)", style=solid]; +"83 conv2d_6" -> "84 relu__2" [label="(1, 72, 28, 28)", style=solid]; +"84 relu__2" -> "85 quantize_per_tensor_default_8" [label="(1, 72, 28, 28)", style=solid]; +"85 quantize_per_tensor_default_8" -> "86 dequantize_per_tensor_default_8" [label="(1, 72, 28, 28)", style=solid]; +"86 dequantize_per_tensor_default_8" -> "92 conv2d_7" [label="(1, 72, 28, 28)", style=solid]; +"87 features_2_block_2_0_weight_bias" -> "92 conv2d_7" [label="(24,)", style=solid]; +"88 conv2d_7_scale_0" -> "91 dequantize_per_channel_default_7" [label="(24,)", style=solid]; +"89 conv2d_7_zero_point_0" -> "91 dequantize_per_channel_default_7" [label="(24,)", style=solid]; +"90 _frozen_param7" -> "91 dequantize_per_channel_default_7" [label="(24, 72, 1, 1)", style=solid]; +"91 dequantize_per_channel_default_7" -> "92 conv2d_7" [label="(24, 72, 1, 1)", style=solid]; +"92 conv2d_7" -> "93 quantize_per_tensor_default_9" [label="(1, 24, 28, 28)", style=solid]; +"93 quantize_per_tensor_default_9" -> "94 dequantize_per_tensor_default_10" [label="(1, 24, 28, 28)", style=solid]; +"93 quantize_per_tensor_default_9" -> "95 dequantize_per_tensor_default_9" [label="(1, 24, 28, 28)", style=solid]; +"94 dequantize_per_tensor_default_10" -> "120 add_" [label="(1, 24, 28, 28)", style=solid]; +"95 dequantize_per_tensor_default_9" -> "101 conv2d_8" [label="(1, 24, 28, 28)", style=solid]; +"96 features_3_block_0_0_weight_bias" -> "101 conv2d_8" [label="(88,)", style=solid]; +"97 conv2d_8_scale_0" -> "100 dequantize_per_channel_default_8" [label="(88,)", style=solid]; +"98 conv2d_8_zero_point_0" -> "100 dequantize_per_channel_default_8" [label="(88,)", style=solid]; +"99 _frozen_param8" -> "100 dequantize_per_channel_default_8" [label="(88, 24, 1, 1)", style=solid]; +"100 dequantize_per_channel_default_8" -> "101 conv2d_8" [label="(88, 24, 1, 1)", style=solid]; +"101 conv2d_8" -> "102 relu__3" [label="(1, 88, 28, 28)", style=solid]; +"102 relu__3" -> "103 quantize_per_tensor_default_10" [label="(1, 88, 28, 28)", style=solid]; +"103 quantize_per_tensor_default_10" -> "104 dequantize_per_tensor_default_11" [label="(1, 88, 28, 28)", style=solid]; +"104 dequantize_per_tensor_default_11" -> "110 conv2d_9" [label="(1, 88, 28, 28)", style=solid]; +"105 features_3_block_1_0_weight_bias" -> "110 conv2d_9" [label="(88,)", style=solid]; +"106 conv2d_9_scale_0" -> "109 dequantize_per_channel_default_9" [label="(88,)", style=solid]; +"107 conv2d_9_zero_point_0" -> "109 dequantize_per_channel_default_9" [label="(88,)", style=solid]; +"108 _frozen_param9" -> "109 dequantize_per_channel_default_9" [label="(88, 1, 3, 3)", style=solid]; +"109 dequantize_per_channel_default_9" -> "110 conv2d_9" [label="(88, 1, 3, 3)", style=solid]; +"110 conv2d_9" -> "111 relu__4" [label="(1, 88, 28, 28)", style=solid]; +"111 relu__4" -> "112 quantize_per_tensor_default_11" [label="(1, 88, 28, 28)", style=solid]; +"112 quantize_per_tensor_default_11" -> "113 dequantize_per_tensor_default_12" [label="(1, 88, 28, 28)", style=solid]; +"113 dequantize_per_tensor_default_12" -> "119 conv2d_10" [label="(1, 88, 28, 28)", style=solid]; +"114 features_3_block_2_0_weight_bias" -> "119 conv2d_10" [label="(24,)", style=solid]; +"115 conv2d_10_scale_0" -> "118 dequantize_per_channel_default_10" [label="(24,)", style=solid]; +"116 conv2d_10_zero_point_0" -> "118 dequantize_per_channel_default_10" [label="(24,)", style=solid]; +"117 _frozen_param10" -> "118 dequantize_per_channel_default_10" [label="(24, 88, 1, 1)", style=solid]; +"118 dequantize_per_channel_default_10" -> "119 conv2d_10" [label="(24, 88, 1, 1)", style=solid]; +"119 conv2d_10" -> "120 add_" [label="(1, 24, 28, 28)", style=solid]; +"120 add_" -> "121 quantize_per_tensor_default_12" [label="(1, 24, 28, 28)", style=solid]; +"121 quantize_per_tensor_default_12" -> "122 dequantize_per_tensor_default_13" [label="(1, 24, 28, 28)", style=solid]; +"122 dequantize_per_tensor_default_13" -> "128 conv2d_11" [label="(1, 24, 28, 28)", style=solid]; +"123 features_4_block_0_0_weight_bias" -> "128 conv2d_11" [label="(96,)", style=solid]; +"124 conv2d_11_scale_0" -> "127 dequantize_per_channel_default_11" [label="(96,)", style=solid]; +"125 conv2d_11_zero_point_0" -> "127 dequantize_per_channel_default_11" [label="(96,)", style=solid]; +"126 _frozen_param11" -> "127 dequantize_per_channel_default_11" [label="(96, 24, 1, 1)", style=solid]; +"127 dequantize_per_channel_default_11" -> "128 conv2d_11" [label="(96, 24, 1, 1)", style=solid]; +"128 conv2d_11" -> "129 hardswish__1" [label="(1, 96, 28, 28)", style=solid]; +"129 hardswish__1" -> "130 quantize_per_tensor_default_13" [label="(1, 96, 28, 28)", style=solid]; +"130 quantize_per_tensor_default_13" -> "131 dequantize_per_tensor_default_14" [label="(1, 96, 28, 28)", style=solid]; +"131 dequantize_per_tensor_default_14" -> "137 conv2d_12" [label="(1, 96, 28, 28)", style=solid]; +"132 features_4_block_1_0_weight_bias" -> "137 conv2d_12" [label="(96,)", style=solid]; +"133 conv2d_12_scale_0" -> "136 dequantize_per_channel_default_12" [label="(96,)", style=solid]; +"134 conv2d_12_zero_point_0" -> "136 dequantize_per_channel_default_12" [label="(96,)", style=solid]; +"135 _frozen_param12" -> "136 dequantize_per_channel_default_12" [label="(96, 1, 5, 5)", style=solid]; +"136 dequantize_per_channel_default_12" -> "137 conv2d_12" [label="(96, 1, 5, 5)", style=solid]; +"137 conv2d_12" -> "138 hardswish__2" [label="(1, 96, 14, 14)", style=solid]; +"138 hardswish__2" -> "139 quantize_per_tensor_default_14" [label="(1, 96, 14, 14)", style=solid]; +"138 hardswish__2" -> "158 mul_1" [label="(1, 96, 14, 14)", style=solid]; +"139 quantize_per_tensor_default_14" -> "140 dequantize_per_tensor_default_15" [label="(1, 96, 14, 14)", style=solid]; +"140 dequantize_per_tensor_default_15" -> "141 adaptive_avg_pool2d_1" [label="(1, 96, 14, 14)", style=solid]; +"141 adaptive_avg_pool2d_1" -> "142 quantize_per_tensor_default_15" [label="(1, 96, 1, 1)", style=solid]; +"142 quantize_per_tensor_default_15" -> "143 dequantize_per_tensor_default_16" [label="(1, 96, 1, 1)", style=solid]; +"143 dequantize_per_tensor_default_16" -> "148 conv2d_13" [label="(1, 96, 1, 1)", style=solid]; +"144 conv2d_13_scale_0" -> "147 dequantize_per_channel_default_13" [label="(24,)", style=solid]; +"145 conv2d_13_zero_point_0" -> "147 dequantize_per_channel_default_13" [label="(24,)", style=solid]; +"146 _frozen_param13" -> "147 dequantize_per_channel_default_13" [label="(24, 96, 1, 1)", style=solid]; +"147 dequantize_per_channel_default_13" -> "148 conv2d_13" [label="(24, 96, 1, 1)", style=solid]; +"148 conv2d_13" -> "149 relu_1" [label="(1, 24, 1, 1)", style=solid]; +"149 relu_1" -> "150 quantize_per_tensor_default_16" [label="(1, 24, 1, 1)", style=solid]; +"150 quantize_per_tensor_default_16" -> "151 dequantize_per_tensor_default_17" [label="(1, 24, 1, 1)", style=solid]; +"151 dequantize_per_tensor_default_17" -> "156 conv2d_14" [label="(1, 24, 1, 1)", style=solid]; +"152 conv2d_14_scale_0" -> "155 dequantize_per_channel_default_14" [label="(96,)", style=solid]; +"153 conv2d_14_zero_point_0" -> "155 dequantize_per_channel_default_14" [label="(96,)", style=solid]; +"154 _frozen_param14" -> "155 dequantize_per_channel_default_14" [label="(96, 24, 1, 1)", style=solid]; +"155 dequantize_per_channel_default_14" -> "156 conv2d_14" [label="(96, 24, 1, 1)", style=solid]; +"156 conv2d_14" -> "157 hardsigmoid_1" [label="(1, 96, 1, 1)", style=solid]; +"157 hardsigmoid_1" -> "158 mul_1" [label="(1, 96, 1, 1)", style=solid]; +"158 mul_1" -> "159 quantize_per_tensor_default_17" [label="(1, 96, 14, 14)", style=solid]; +"159 quantize_per_tensor_default_17" -> "160 dequantize_per_tensor_default_18" [label="(1, 96, 14, 14)", style=solid]; +"160 dequantize_per_tensor_default_18" -> "166 conv2d_15" [label="(1, 96, 14, 14)", style=solid]; +"161 features_4_block_3_0_weight_bias" -> "166 conv2d_15" [label="(40,)", style=solid]; +"162 conv2d_15_scale_0" -> "165 dequantize_per_channel_default_15" [label="(40,)", style=solid]; +"163 conv2d_15_zero_point_0" -> "165 dequantize_per_channel_default_15" [label="(40,)", style=solid]; +"164 _frozen_param15" -> "165 dequantize_per_channel_default_15" [label="(40, 96, 1, 1)", style=solid]; +"165 dequantize_per_channel_default_15" -> "166 conv2d_15" [label="(40, 96, 1, 1)", style=solid]; +"166 conv2d_15" -> "167 quantize_per_tensor_default_18" [label="(1, 40, 14, 14)", style=solid]; +"167 quantize_per_tensor_default_18" -> "168 dequantize_per_tensor_default_20" [label="(1, 40, 14, 14)", style=solid]; +"167 quantize_per_tensor_default_18" -> "169 dequantize_per_tensor_default_19" [label="(1, 40, 14, 14)", style=solid]; +"168 dequantize_per_tensor_default_20" -> "214 add__1" [label="(1, 40, 14, 14)", style=solid]; +"169 dequantize_per_tensor_default_19" -> "175 conv2d_16" [label="(1, 40, 14, 14)", style=solid]; +"170 features_5_block_0_0_weight_bias" -> "175 conv2d_16" [label="(240,)", style=solid]; +"171 conv2d_16_scale_0" -> "174 dequantize_per_channel_default_16" [label="(240,)", style=solid]; +"172 conv2d_16_zero_point_0" -> "174 dequantize_per_channel_default_16" [label="(240,)", style=solid]; +"173 _frozen_param16" -> "174 dequantize_per_channel_default_16" [label="(240, 40, 1, 1)", style=solid]; +"174 dequantize_per_channel_default_16" -> "175 conv2d_16" [label="(240, 40, 1, 1)", style=solid]; +"175 conv2d_16" -> "176 hardswish__3" [label="(1, 240, 14, 14)", style=solid]; +"176 hardswish__3" -> "177 quantize_per_tensor_default_19" [label="(1, 240, 14, 14)", style=solid]; +"177 quantize_per_tensor_default_19" -> "178 dequantize_per_tensor_default_21" [label="(1, 240, 14, 14)", style=solid]; +"178 dequantize_per_tensor_default_21" -> "184 conv2d_17" [label="(1, 240, 14, 14)", style=solid]; +"179 features_5_block_1_0_weight_bias" -> "184 conv2d_17" [label="(240,)", style=solid]; +"180 conv2d_17_scale_0" -> "183 dequantize_per_channel_default_17" [label="(240,)", style=solid]; +"181 conv2d_17_zero_point_0" -> "183 dequantize_per_channel_default_17" [label="(240,)", style=solid]; +"182 _frozen_param17" -> "183 dequantize_per_channel_default_17" [label="(240, 1, 5, 5)", style=solid]; +"183 dequantize_per_channel_default_17" -> "184 conv2d_17" [label="(240, 1, 5, 5)", style=solid]; +"184 conv2d_17" -> "185 hardswish__4" [label="(1, 240, 14, 14)", style=solid]; +"185 hardswish__4" -> "186 quantize_per_tensor_default_20" [label="(1, 240, 14, 14)", style=solid]; +"185 hardswish__4" -> "205 mul_2" [label="(1, 240, 14, 14)", style=solid]; +"186 quantize_per_tensor_default_20" -> "187 dequantize_per_tensor_default_22" [label="(1, 240, 14, 14)", style=solid]; +"187 dequantize_per_tensor_default_22" -> "188 adaptive_avg_pool2d_2" [label="(1, 240, 14, 14)", style=solid]; +"188 adaptive_avg_pool2d_2" -> "189 quantize_per_tensor_default_21" [label="(1, 240, 1, 1)", style=solid]; +"189 quantize_per_tensor_default_21" -> "190 dequantize_per_tensor_default_23" [label="(1, 240, 1, 1)", style=solid]; +"190 dequantize_per_tensor_default_23" -> "195 conv2d_18" [label="(1, 240, 1, 1)", style=solid]; +"191 conv2d_18_scale_0" -> "194 dequantize_per_channel_default_18" [label="(64,)", style=solid]; +"192 conv2d_18_zero_point_0" -> "194 dequantize_per_channel_default_18" [label="(64,)", style=solid]; +"193 _frozen_param18" -> "194 dequantize_per_channel_default_18" [label="(64, 240, 1, 1)", style=solid]; +"194 dequantize_per_channel_default_18" -> "195 conv2d_18" [label="(64, 240, 1, 1)", style=solid]; +"195 conv2d_18" -> "196 relu_2" [label="(1, 64, 1, 1)", style=solid]; +"196 relu_2" -> "197 quantize_per_tensor_default_22" [label="(1, 64, 1, 1)", style=solid]; +"197 quantize_per_tensor_default_22" -> "198 dequantize_per_tensor_default_24" [label="(1, 64, 1, 1)", style=solid]; +"198 dequantize_per_tensor_default_24" -> "203 conv2d_19" [label="(1, 64, 1, 1)", style=solid]; +"199 conv2d_19_scale_0" -> "202 dequantize_per_channel_default_19" [label="(240,)", style=solid]; +"200 conv2d_19_zero_point_0" -> "202 dequantize_per_channel_default_19" [label="(240,)", style=solid]; +"201 _frozen_param19" -> "202 dequantize_per_channel_default_19" [label="(240, 64, 1, 1)", style=solid]; +"202 dequantize_per_channel_default_19" -> "203 conv2d_19" [label="(240, 64, 1, 1)", style=solid]; +"203 conv2d_19" -> "204 hardsigmoid_2" [label="(1, 240, 1, 1)", style=solid]; +"204 hardsigmoid_2" -> "205 mul_2" [label="(1, 240, 1, 1)", style=solid]; +"205 mul_2" -> "206 quantize_per_tensor_default_23" [label="(1, 240, 14, 14)", style=solid]; +"206 quantize_per_tensor_default_23" -> "207 dequantize_per_tensor_default_25" [label="(1, 240, 14, 14)", style=solid]; +"207 dequantize_per_tensor_default_25" -> "213 conv2d_20" [label="(1, 240, 14, 14)", style=solid]; +"208 features_5_block_3_0_weight_bias" -> "213 conv2d_20" [label="(40,)", style=solid]; +"209 conv2d_20_scale_0" -> "212 dequantize_per_channel_default_20" [label="(40,)", style=solid]; +"210 conv2d_20_zero_point_0" -> "212 dequantize_per_channel_default_20" [label="(40,)", style=solid]; +"211 _frozen_param20" -> "212 dequantize_per_channel_default_20" [label="(40, 240, 1, 1)", style=solid]; +"212 dequantize_per_channel_default_20" -> "213 conv2d_20" [label="(40, 240, 1, 1)", style=solid]; +"213 conv2d_20" -> "214 add__1" [label="(1, 40, 14, 14)", style=solid]; +"214 add__1" -> "215 quantize_per_tensor_default_24" [label="(1, 40, 14, 14)", style=solid]; +"215 quantize_per_tensor_default_24" -> "216 dequantize_per_tensor_default_27" [label="(1, 40, 14, 14)", style=solid]; +"215 quantize_per_tensor_default_24" -> "217 dequantize_per_tensor_default_26" [label="(1, 40, 14, 14)", style=solid]; +"216 dequantize_per_tensor_default_27" -> "262 add__2" [label="(1, 40, 14, 14)", style=solid]; +"217 dequantize_per_tensor_default_26" -> "223 conv2d_21" [label="(1, 40, 14, 14)", style=solid]; +"218 features_6_block_0_0_weight_bias" -> "223 conv2d_21" [label="(240,)", style=solid]; +"219 conv2d_21_scale_0" -> "222 dequantize_per_channel_default_21" [label="(240,)", style=solid]; +"220 conv2d_21_zero_point_0" -> "222 dequantize_per_channel_default_21" [label="(240,)", style=solid]; +"221 _frozen_param21" -> "222 dequantize_per_channel_default_21" [label="(240, 40, 1, 1)", style=solid]; +"222 dequantize_per_channel_default_21" -> "223 conv2d_21" [label="(240, 40, 1, 1)", style=solid]; +"223 conv2d_21" -> "224 hardswish__5" [label="(1, 240, 14, 14)", style=solid]; +"224 hardswish__5" -> "225 quantize_per_tensor_default_25" [label="(1, 240, 14, 14)", style=solid]; +"225 quantize_per_tensor_default_25" -> "226 dequantize_per_tensor_default_28" [label="(1, 240, 14, 14)", style=solid]; +"226 dequantize_per_tensor_default_28" -> "232 conv2d_22" [label="(1, 240, 14, 14)", style=solid]; +"227 features_6_block_1_0_weight_bias" -> "232 conv2d_22" [label="(240,)", style=solid]; +"228 conv2d_22_scale_0" -> "231 dequantize_per_channel_default_22" [label="(240,)", style=solid]; +"229 conv2d_22_zero_point_0" -> "231 dequantize_per_channel_default_22" [label="(240,)", style=solid]; +"230 _frozen_param22" -> "231 dequantize_per_channel_default_22" [label="(240, 1, 5, 5)", style=solid]; +"231 dequantize_per_channel_default_22" -> "232 conv2d_22" [label="(240, 1, 5, 5)", style=solid]; +"232 conv2d_22" -> "233 hardswish__6" [label="(1, 240, 14, 14)", style=solid]; +"233 hardswish__6" -> "234 quantize_per_tensor_default_26" [label="(1, 240, 14, 14)", style=solid]; +"233 hardswish__6" -> "253 mul_3" [label="(1, 240, 14, 14)", style=solid]; +"234 quantize_per_tensor_default_26" -> "235 dequantize_per_tensor_default_29" [label="(1, 240, 14, 14)", style=solid]; +"235 dequantize_per_tensor_default_29" -> "236 adaptive_avg_pool2d_3" [label="(1, 240, 14, 14)", style=solid]; +"236 adaptive_avg_pool2d_3" -> "237 quantize_per_tensor_default_27" [label="(1, 240, 1, 1)", style=solid]; +"237 quantize_per_tensor_default_27" -> "238 dequantize_per_tensor_default_30" [label="(1, 240, 1, 1)", style=solid]; +"238 dequantize_per_tensor_default_30" -> "243 conv2d_23" [label="(1, 240, 1, 1)", style=solid]; +"239 conv2d_23_scale_0" -> "242 dequantize_per_channel_default_23" [label="(64,)", style=solid]; +"240 conv2d_23_zero_point_0" -> "242 dequantize_per_channel_default_23" [label="(64,)", style=solid]; +"241 _frozen_param23" -> "242 dequantize_per_channel_default_23" [label="(64, 240, 1, 1)", style=solid]; +"242 dequantize_per_channel_default_23" -> "243 conv2d_23" [label="(64, 240, 1, 1)", style=solid]; +"243 conv2d_23" -> "244 relu_3" [label="(1, 64, 1, 1)", style=solid]; +"244 relu_3" -> "245 quantize_per_tensor_default_28" [label="(1, 64, 1, 1)", style=solid]; +"245 quantize_per_tensor_default_28" -> "246 dequantize_per_tensor_default_31" [label="(1, 64, 1, 1)", style=solid]; +"246 dequantize_per_tensor_default_31" -> "251 conv2d_24" [label="(1, 64, 1, 1)", style=solid]; +"247 conv2d_24_scale_0" -> "250 dequantize_per_channel_default_24" [label="(240,)", style=solid]; +"248 conv2d_24_zero_point_0" -> "250 dequantize_per_channel_default_24" [label="(240,)", style=solid]; +"249 _frozen_param24" -> "250 dequantize_per_channel_default_24" [label="(240, 64, 1, 1)", style=solid]; +"250 dequantize_per_channel_default_24" -> "251 conv2d_24" [label="(240, 64, 1, 1)", style=solid]; +"251 conv2d_24" -> "252 hardsigmoid_3" [label="(1, 240, 1, 1)", style=solid]; +"252 hardsigmoid_3" -> "253 mul_3" [label="(1, 240, 1, 1)", style=solid]; +"253 mul_3" -> "254 quantize_per_tensor_default_29" [label="(1, 240, 14, 14)", style=solid]; +"254 quantize_per_tensor_default_29" -> "255 dequantize_per_tensor_default_32" [label="(1, 240, 14, 14)", style=solid]; +"255 dequantize_per_tensor_default_32" -> "261 conv2d_25" [label="(1, 240, 14, 14)", style=solid]; +"256 features_6_block_3_0_weight_bias" -> "261 conv2d_25" [label="(40,)", style=solid]; +"257 conv2d_25_scale_0" -> "260 dequantize_per_channel_default_25" [label="(40,)", style=solid]; +"258 conv2d_25_zero_point_0" -> "260 dequantize_per_channel_default_25" [label="(40,)", style=solid]; +"259 _frozen_param25" -> "260 dequantize_per_channel_default_25" [label="(40, 240, 1, 1)", style=solid]; +"260 dequantize_per_channel_default_25" -> "261 conv2d_25" [label="(40, 240, 1, 1)", style=solid]; +"261 conv2d_25" -> "262 add__2" [label="(1, 40, 14, 14)", style=solid]; +"262 add__2" -> "263 quantize_per_tensor_default_30" [label="(1, 40, 14, 14)", style=solid]; +"263 quantize_per_tensor_default_30" -> "264 dequantize_per_tensor_default_33" [label="(1, 40, 14, 14)", style=solid]; +"264 dequantize_per_tensor_default_33" -> "270 conv2d_26" [label="(1, 40, 14, 14)", style=solid]; +"265 features_7_block_0_0_weight_bias" -> "270 conv2d_26" [label="(120,)", style=solid]; +"266 conv2d_26_scale_0" -> "269 dequantize_per_channel_default_26" [label="(120,)", style=solid]; +"267 conv2d_26_zero_point_0" -> "269 dequantize_per_channel_default_26" [label="(120,)", style=solid]; +"268 _frozen_param26" -> "269 dequantize_per_channel_default_26" [label="(120, 40, 1, 1)", style=solid]; +"269 dequantize_per_channel_default_26" -> "270 conv2d_26" [label="(120, 40, 1, 1)", style=solid]; +"270 conv2d_26" -> "271 hardswish__7" [label="(1, 120, 14, 14)", style=solid]; +"271 hardswish__7" -> "272 quantize_per_tensor_default_31" [label="(1, 120, 14, 14)", style=solid]; +"272 quantize_per_tensor_default_31" -> "273 dequantize_per_tensor_default_34" [label="(1, 120, 14, 14)", style=solid]; +"273 dequantize_per_tensor_default_34" -> "279 conv2d_27" [label="(1, 120, 14, 14)", style=solid]; +"274 features_7_block_1_0_weight_bias" -> "279 conv2d_27" [label="(120,)", style=solid]; +"275 conv2d_27_scale_0" -> "278 dequantize_per_channel_default_27" [label="(120,)", style=solid]; +"276 conv2d_27_zero_point_0" -> "278 dequantize_per_channel_default_27" [label="(120,)", style=solid]; +"277 _frozen_param27" -> "278 dequantize_per_channel_default_27" [label="(120, 1, 5, 5)", style=solid]; +"278 dequantize_per_channel_default_27" -> "279 conv2d_27" [label="(120, 1, 5, 5)", style=solid]; +"279 conv2d_27" -> "280 hardswish__8" [label="(1, 120, 14, 14)", style=solid]; +"280 hardswish__8" -> "281 quantize_per_tensor_default_32" [label="(1, 120, 14, 14)", style=solid]; +"280 hardswish__8" -> "300 mul_4" [label="(1, 120, 14, 14)", style=solid]; +"281 quantize_per_tensor_default_32" -> "282 dequantize_per_tensor_default_35" [label="(1, 120, 14, 14)", style=solid]; +"282 dequantize_per_tensor_default_35" -> "283 adaptive_avg_pool2d_4" [label="(1, 120, 14, 14)", style=solid]; +"283 adaptive_avg_pool2d_4" -> "284 quantize_per_tensor_default_33" [label="(1, 120, 1, 1)", style=solid]; +"284 quantize_per_tensor_default_33" -> "285 dequantize_per_tensor_default_36" [label="(1, 120, 1, 1)", style=solid]; +"285 dequantize_per_tensor_default_36" -> "290 conv2d_28" [label="(1, 120, 1, 1)", style=solid]; +"286 conv2d_28_scale_0" -> "289 dequantize_per_channel_default_28" [label="(32,)", style=solid]; +"287 conv2d_28_zero_point_0" -> "289 dequantize_per_channel_default_28" [label="(32,)", style=solid]; +"288 _frozen_param28" -> "289 dequantize_per_channel_default_28" [label="(32, 120, 1, 1)", style=solid]; +"289 dequantize_per_channel_default_28" -> "290 conv2d_28" [label="(32, 120, 1, 1)", style=solid]; +"290 conv2d_28" -> "291 relu_4" [label="(1, 32, 1, 1)", style=solid]; +"291 relu_4" -> "292 quantize_per_tensor_default_34" [label="(1, 32, 1, 1)", style=solid]; +"292 quantize_per_tensor_default_34" -> "293 dequantize_per_tensor_default_37" [label="(1, 32, 1, 1)", style=solid]; +"293 dequantize_per_tensor_default_37" -> "298 conv2d_29" [label="(1, 32, 1, 1)", style=solid]; +"294 conv2d_29_scale_0" -> "297 dequantize_per_channel_default_29" [label="(120,)", style=solid]; +"295 conv2d_29_zero_point_0" -> "297 dequantize_per_channel_default_29" [label="(120,)", style=solid]; +"296 _frozen_param29" -> "297 dequantize_per_channel_default_29" [label="(120, 32, 1, 1)", style=solid]; +"297 dequantize_per_channel_default_29" -> "298 conv2d_29" [label="(120, 32, 1, 1)", style=solid]; +"298 conv2d_29" -> "299 hardsigmoid_4" [label="(1, 120, 1, 1)", style=solid]; +"299 hardsigmoid_4" -> "300 mul_4" [label="(1, 120, 1, 1)", style=solid]; +"300 mul_4" -> "301 quantize_per_tensor_default_35" [label="(1, 120, 14, 14)", style=solid]; +"301 quantize_per_tensor_default_35" -> "302 dequantize_per_tensor_default_38" [label="(1, 120, 14, 14)", style=solid]; +"302 dequantize_per_tensor_default_38" -> "308 conv2d_30" [label="(1, 120, 14, 14)", style=solid]; +"303 features_7_block_3_0_weight_bias" -> "308 conv2d_30" [label="(48,)", style=solid]; +"304 conv2d_30_scale_0" -> "307 dequantize_per_channel_default_30" [label="(48,)", style=solid]; +"305 conv2d_30_zero_point_0" -> "307 dequantize_per_channel_default_30" [label="(48,)", style=solid]; +"306 _frozen_param30" -> "307 dequantize_per_channel_default_30" [label="(48, 120, 1, 1)", style=solid]; +"307 dequantize_per_channel_default_30" -> "308 conv2d_30" [label="(48, 120, 1, 1)", style=solid]; +"308 conv2d_30" -> "309 quantize_per_tensor_default_36" [label="(1, 48, 14, 14)", style=solid]; +"309 quantize_per_tensor_default_36" -> "310 dequantize_per_tensor_default_40" [label="(1, 48, 14, 14)", style=solid]; +"309 quantize_per_tensor_default_36" -> "311 dequantize_per_tensor_default_39" [label="(1, 48, 14, 14)", style=solid]; +"310 dequantize_per_tensor_default_40" -> "356 add__3" [label="(1, 48, 14, 14)", style=solid]; +"311 dequantize_per_tensor_default_39" -> "317 conv2d_31" [label="(1, 48, 14, 14)", style=solid]; +"312 features_8_block_0_0_weight_bias" -> "317 conv2d_31" [label="(144,)", style=solid]; +"313 conv2d_31_scale_0" -> "316 dequantize_per_channel_default_31" [label="(144,)", style=solid]; +"314 conv2d_31_zero_point_0" -> "316 dequantize_per_channel_default_31" [label="(144,)", style=solid]; +"315 _frozen_param31" -> "316 dequantize_per_channel_default_31" [label="(144, 48, 1, 1)", style=solid]; +"316 dequantize_per_channel_default_31" -> "317 conv2d_31" [label="(144, 48, 1, 1)", style=solid]; +"317 conv2d_31" -> "318 hardswish__9" [label="(1, 144, 14, 14)", style=solid]; +"318 hardswish__9" -> "319 quantize_per_tensor_default_37" [label="(1, 144, 14, 14)", style=solid]; +"319 quantize_per_tensor_default_37" -> "320 dequantize_per_tensor_default_41" [label="(1, 144, 14, 14)", style=solid]; +"320 dequantize_per_tensor_default_41" -> "326 conv2d_32" [label="(1, 144, 14, 14)", style=solid]; +"321 features_8_block_1_0_weight_bias" -> "326 conv2d_32" [label="(144,)", style=solid]; +"322 conv2d_32_scale_0" -> "325 dequantize_per_channel_default_32" [label="(144,)", style=solid]; +"323 conv2d_32_zero_point_0" -> "325 dequantize_per_channel_default_32" [label="(144,)", style=solid]; +"324 _frozen_param32" -> "325 dequantize_per_channel_default_32" [label="(144, 1, 5, 5)", style=solid]; +"325 dequantize_per_channel_default_32" -> "326 conv2d_32" [label="(144, 1, 5, 5)", style=solid]; +"326 conv2d_32" -> "327 hardswish__10" [label="(1, 144, 14, 14)", style=solid]; +"327 hardswish__10" -> "328 quantize_per_tensor_default_38" [label="(1, 144, 14, 14)", style=solid]; +"327 hardswish__10" -> "347 mul_5" [label="(1, 144, 14, 14)", style=solid]; +"328 quantize_per_tensor_default_38" -> "329 dequantize_per_tensor_default_42" [label="(1, 144, 14, 14)", style=solid]; +"329 dequantize_per_tensor_default_42" -> "330 adaptive_avg_pool2d_5" [label="(1, 144, 14, 14)", style=solid]; +"330 adaptive_avg_pool2d_5" -> "331 quantize_per_tensor_default_39" [label="(1, 144, 1, 1)", style=solid]; +"331 quantize_per_tensor_default_39" -> "332 dequantize_per_tensor_default_43" [label="(1, 144, 1, 1)", style=solid]; +"332 dequantize_per_tensor_default_43" -> "337 conv2d_33" [label="(1, 144, 1, 1)", style=solid]; +"333 conv2d_33_scale_0" -> "336 dequantize_per_channel_default_33" [label="(40,)", style=solid]; +"334 conv2d_33_zero_point_0" -> "336 dequantize_per_channel_default_33" [label="(40,)", style=solid]; +"335 _frozen_param33" -> "336 dequantize_per_channel_default_33" [label="(40, 144, 1, 1)", style=solid]; +"336 dequantize_per_channel_default_33" -> "337 conv2d_33" [label="(40, 144, 1, 1)", style=solid]; +"337 conv2d_33" -> "338 relu_5" [label="(1, 40, 1, 1)", style=solid]; +"338 relu_5" -> "339 quantize_per_tensor_default_40" [label="(1, 40, 1, 1)", style=solid]; +"339 quantize_per_tensor_default_40" -> "340 dequantize_per_tensor_default_44" [label="(1, 40, 1, 1)", style=solid]; +"340 dequantize_per_tensor_default_44" -> "345 conv2d_34" [label="(1, 40, 1, 1)", style=solid]; +"341 conv2d_34_scale_0" -> "344 dequantize_per_channel_default_34" [label="(144,)", style=solid]; +"342 conv2d_34_zero_point_0" -> "344 dequantize_per_channel_default_34" [label="(144,)", style=solid]; +"343 _frozen_param34" -> "344 dequantize_per_channel_default_34" [label="(144, 40, 1, 1)", style=solid]; +"344 dequantize_per_channel_default_34" -> "345 conv2d_34" [label="(144, 40, 1, 1)", style=solid]; +"345 conv2d_34" -> "346 hardsigmoid_5" [label="(1, 144, 1, 1)", style=solid]; +"346 hardsigmoid_5" -> "347 mul_5" [label="(1, 144, 1, 1)", style=solid]; +"347 mul_5" -> "348 quantize_per_tensor_default_41" [label="(1, 144, 14, 14)", style=solid]; +"348 quantize_per_tensor_default_41" -> "349 dequantize_per_tensor_default_45" [label="(1, 144, 14, 14)", style=solid]; +"349 dequantize_per_tensor_default_45" -> "355 conv2d_35" [label="(1, 144, 14, 14)", style=solid]; +"350 features_8_block_3_0_weight_bias" -> "355 conv2d_35" [label="(48,)", style=solid]; +"351 conv2d_35_scale_0" -> "354 dequantize_per_channel_default_35" [label="(48,)", style=solid]; +"352 conv2d_35_zero_point_0" -> "354 dequantize_per_channel_default_35" [label="(48,)", style=solid]; +"353 _frozen_param35" -> "354 dequantize_per_channel_default_35" [label="(48, 144, 1, 1)", style=solid]; +"354 dequantize_per_channel_default_35" -> "355 conv2d_35" [label="(48, 144, 1, 1)", style=solid]; +"355 conv2d_35" -> "356 add__3" [label="(1, 48, 14, 14)", style=solid]; +"356 add__3" -> "357 quantize_per_tensor_default_42" [label="(1, 48, 14, 14)", style=solid]; +"357 quantize_per_tensor_default_42" -> "358 dequantize_per_tensor_default_46" [label="(1, 48, 14, 14)", style=solid]; +"358 dequantize_per_tensor_default_46" -> "364 conv2d_36" [label="(1, 48, 14, 14)", style=solid]; +"359 features_9_block_0_0_weight_bias" -> "364 conv2d_36" [label="(288,)", style=solid]; +"360 conv2d_36_scale_0" -> "363 dequantize_per_channel_default_36" [label="(288,)", style=solid]; +"361 conv2d_36_zero_point_0" -> "363 dequantize_per_channel_default_36" [label="(288,)", style=solid]; +"362 _frozen_param36" -> "363 dequantize_per_channel_default_36" [label="(288, 48, 1, 1)", style=solid]; +"363 dequantize_per_channel_default_36" -> "364 conv2d_36" [label="(288, 48, 1, 1)", style=solid]; +"364 conv2d_36" -> "365 hardswish__11" [label="(1, 288, 14, 14)", style=solid]; +"365 hardswish__11" -> "366 quantize_per_tensor_default_43" [label="(1, 288, 14, 14)", style=solid]; +"366 quantize_per_tensor_default_43" -> "367 dequantize_per_tensor_default_47" [label="(1, 288, 14, 14)", style=solid]; +"367 dequantize_per_tensor_default_47" -> "373 conv2d_37" [label="(1, 288, 14, 14)", style=solid]; +"368 features_9_block_1_0_weight_bias" -> "373 conv2d_37" [label="(288,)", style=solid]; +"369 conv2d_37_scale_0" -> "372 dequantize_per_channel_default_37" [label="(288,)", style=solid]; +"370 conv2d_37_zero_point_0" -> "372 dequantize_per_channel_default_37" [label="(288,)", style=solid]; +"371 _frozen_param37" -> "372 dequantize_per_channel_default_37" [label="(288, 1, 5, 5)", style=solid]; +"372 dequantize_per_channel_default_37" -> "373 conv2d_37" [label="(288, 1, 5, 5)", style=solid]; +"373 conv2d_37" -> "374 hardswish__12" [label="(1, 288, 7, 7)", style=solid]; +"374 hardswish__12" -> "375 quantize_per_tensor_default_44" [label="(1, 288, 7, 7)", style=solid]; +"374 hardswish__12" -> "394 mul_6" [label="(1, 288, 7, 7)", style=solid]; +"375 quantize_per_tensor_default_44" -> "376 dequantize_per_tensor_default_48" [label="(1, 288, 7, 7)", style=solid]; +"376 dequantize_per_tensor_default_48" -> "377 adaptive_avg_pool2d_6" [label="(1, 288, 7, 7)", style=solid]; +"377 adaptive_avg_pool2d_6" -> "378 quantize_per_tensor_default_45" [label="(1, 288, 1, 1)", style=solid]; +"378 quantize_per_tensor_default_45" -> "379 dequantize_per_tensor_default_49" [label="(1, 288, 1, 1)", style=solid]; +"379 dequantize_per_tensor_default_49" -> "384 conv2d_38" [label="(1, 288, 1, 1)", style=solid]; +"380 conv2d_38_scale_0" -> "383 dequantize_per_channel_default_38" [label="(72,)", style=solid]; +"381 conv2d_38_zero_point_0" -> "383 dequantize_per_channel_default_38" [label="(72,)", style=solid]; +"382 _frozen_param38" -> "383 dequantize_per_channel_default_38" [label="(72, 288, 1, 1)", style=solid]; +"383 dequantize_per_channel_default_38" -> "384 conv2d_38" [label="(72, 288, 1, 1)", style=solid]; +"384 conv2d_38" -> "385 relu_6" [label="(1, 72, 1, 1)", style=solid]; +"385 relu_6" -> "386 quantize_per_tensor_default_46" [label="(1, 72, 1, 1)", style=solid]; +"386 quantize_per_tensor_default_46" -> "387 dequantize_per_tensor_default_50" [label="(1, 72, 1, 1)", style=solid]; +"387 dequantize_per_tensor_default_50" -> "392 conv2d_39" [label="(1, 72, 1, 1)", style=solid]; +"388 conv2d_39_scale_0" -> "391 dequantize_per_channel_default_39" [label="(288,)", style=solid]; +"389 conv2d_39_zero_point_0" -> "391 dequantize_per_channel_default_39" [label="(288,)", style=solid]; +"390 _frozen_param39" -> "391 dequantize_per_channel_default_39" [label="(288, 72, 1, 1)", style=solid]; +"391 dequantize_per_channel_default_39" -> "392 conv2d_39" [label="(288, 72, 1, 1)", style=solid]; +"392 conv2d_39" -> "393 hardsigmoid_6" [label="(1, 288, 1, 1)", style=solid]; +"393 hardsigmoid_6" -> "394 mul_6" [label="(1, 288, 1, 1)", style=solid]; +"394 mul_6" -> "395 quantize_per_tensor_default_47" [label="(1, 288, 7, 7)", style=solid]; +"395 quantize_per_tensor_default_47" -> "396 dequantize_per_tensor_default_51" [label="(1, 288, 7, 7)", style=solid]; +"396 dequantize_per_tensor_default_51" -> "402 conv2d_40" [label="(1, 288, 7, 7)", style=solid]; +"397 features_9_block_3_0_weight_bias" -> "402 conv2d_40" [label="(96,)", style=solid]; +"398 conv2d_40_scale_0" -> "401 dequantize_per_channel_default_40" [label="(96,)", style=solid]; +"399 conv2d_40_zero_point_0" -> "401 dequantize_per_channel_default_40" [label="(96,)", style=solid]; +"400 _frozen_param40" -> "401 dequantize_per_channel_default_40" [label="(96, 288, 1, 1)", style=solid]; +"401 dequantize_per_channel_default_40" -> "402 conv2d_40" [label="(96, 288, 1, 1)", style=solid]; +"402 conv2d_40" -> "403 quantize_per_tensor_default_48" [label="(1, 96, 7, 7)", style=solid]; +"403 quantize_per_tensor_default_48" -> "404 dequantize_per_tensor_default_53" [label="(1, 96, 7, 7)", style=solid]; +"403 quantize_per_tensor_default_48" -> "405 dequantize_per_tensor_default_52" [label="(1, 96, 7, 7)", style=solid]; +"404 dequantize_per_tensor_default_53" -> "450 add__4" [label="(1, 96, 7, 7)", style=solid]; +"405 dequantize_per_tensor_default_52" -> "411 conv2d_41" [label="(1, 96, 7, 7)", style=solid]; +"406 features_10_block_0_0_weight_bias" -> "411 conv2d_41" [label="(576,)", style=solid]; +"407 conv2d_41_scale_0" -> "410 dequantize_per_channel_default_41" [label="(576,)", style=solid]; +"408 conv2d_41_zero_point_0" -> "410 dequantize_per_channel_default_41" [label="(576,)", style=solid]; +"409 _frozen_param41" -> "410 dequantize_per_channel_default_41" [label="(576, 96, 1, 1)", style=solid]; +"410 dequantize_per_channel_default_41" -> "411 conv2d_41" [label="(576, 96, 1, 1)", style=solid]; +"411 conv2d_41" -> "412 hardswish__13" [label="(1, 576, 7, 7)", style=solid]; +"412 hardswish__13" -> "413 quantize_per_tensor_default_49" [label="(1, 576, 7, 7)", style=solid]; +"413 quantize_per_tensor_default_49" -> "414 dequantize_per_tensor_default_54" [label="(1, 576, 7, 7)", style=solid]; +"414 dequantize_per_tensor_default_54" -> "420 conv2d_42" [label="(1, 576, 7, 7)", style=solid]; +"415 features_10_block_1_0_weight_bias" -> "420 conv2d_42" [label="(576,)", style=solid]; +"416 conv2d_42_scale_0" -> "419 dequantize_per_channel_default_42" [label="(576,)", style=solid]; +"417 conv2d_42_zero_point_0" -> "419 dequantize_per_channel_default_42" [label="(576,)", style=solid]; +"418 _frozen_param42" -> "419 dequantize_per_channel_default_42" [label="(576, 1, 5, 5)", style=solid]; +"419 dequantize_per_channel_default_42" -> "420 conv2d_42" [label="(576, 1, 5, 5)", style=solid]; +"420 conv2d_42" -> "421 hardswish__14" [label="(1, 576, 7, 7)", style=solid]; +"421 hardswish__14" -> "422 quantize_per_tensor_default_50" [label="(1, 576, 7, 7)", style=solid]; +"421 hardswish__14" -> "441 mul_7" [label="(1, 576, 7, 7)", style=solid]; +"422 quantize_per_tensor_default_50" -> "423 dequantize_per_tensor_default_55" [label="(1, 576, 7, 7)", style=solid]; +"423 dequantize_per_tensor_default_55" -> "424 adaptive_avg_pool2d_7" [label="(1, 576, 7, 7)", style=solid]; +"424 adaptive_avg_pool2d_7" -> "425 quantize_per_tensor_default_51" [label="(1, 576, 1, 1)", style=solid]; +"425 quantize_per_tensor_default_51" -> "426 dequantize_per_tensor_default_56" [label="(1, 576, 1, 1)", style=solid]; +"426 dequantize_per_tensor_default_56" -> "431 conv2d_43" [label="(1, 576, 1, 1)", style=solid]; +"427 conv2d_43_scale_0" -> "430 dequantize_per_channel_default_43" [label="(144,)", style=solid]; +"428 conv2d_43_zero_point_0" -> "430 dequantize_per_channel_default_43" [label="(144,)", style=solid]; +"429 _frozen_param43" -> "430 dequantize_per_channel_default_43" [label="(144, 576, 1, 1)", style=solid]; +"430 dequantize_per_channel_default_43" -> "431 conv2d_43" [label="(144, 576, 1, 1)", style=solid]; +"431 conv2d_43" -> "432 relu_7" [label="(1, 144, 1, 1)", style=solid]; +"432 relu_7" -> "433 quantize_per_tensor_default_52" [label="(1, 144, 1, 1)", style=solid]; +"433 quantize_per_tensor_default_52" -> "434 dequantize_per_tensor_default_57" [label="(1, 144, 1, 1)", style=solid]; +"434 dequantize_per_tensor_default_57" -> "439 conv2d_44" [label="(1, 144, 1, 1)", style=solid]; +"435 conv2d_44_scale_0" -> "438 dequantize_per_channel_default_44" [label="(576,)", style=solid]; +"436 conv2d_44_zero_point_0" -> "438 dequantize_per_channel_default_44" [label="(576,)", style=solid]; +"437 _frozen_param44" -> "438 dequantize_per_channel_default_44" [label="(576, 144, 1, 1)", style=solid]; +"438 dequantize_per_channel_default_44" -> "439 conv2d_44" [label="(576, 144, 1, 1)", style=solid]; +"439 conv2d_44" -> "440 hardsigmoid_7" [label="(1, 576, 1, 1)", style=solid]; +"440 hardsigmoid_7" -> "441 mul_7" [label="(1, 576, 1, 1)", style=solid]; +"441 mul_7" -> "442 quantize_per_tensor_default_53" [label="(1, 576, 7, 7)", style=solid]; +"442 quantize_per_tensor_default_53" -> "443 dequantize_per_tensor_default_58" [label="(1, 576, 7, 7)", style=solid]; +"443 dequantize_per_tensor_default_58" -> "449 conv2d_45" [label="(1, 576, 7, 7)", style=solid]; +"444 features_10_block_3_0_weight_bias" -> "449 conv2d_45" [label="(96,)", style=solid]; +"445 conv2d_45_scale_0" -> "448 dequantize_per_channel_default_45" [label="(96,)", style=solid]; +"446 conv2d_45_zero_point_0" -> "448 dequantize_per_channel_default_45" [label="(96,)", style=solid]; +"447 _frozen_param45" -> "448 dequantize_per_channel_default_45" [label="(96, 576, 1, 1)", style=solid]; +"448 dequantize_per_channel_default_45" -> "449 conv2d_45" [label="(96, 576, 1, 1)", style=solid]; +"449 conv2d_45" -> "450 add__4" [label="(1, 96, 7, 7)", style=solid]; +"450 add__4" -> "451 quantize_per_tensor_default_54" [label="(1, 96, 7, 7)", style=solid]; +"451 quantize_per_tensor_default_54" -> "452 dequantize_per_tensor_default_60" [label="(1, 96, 7, 7)", style=solid]; +"451 quantize_per_tensor_default_54" -> "453 dequantize_per_tensor_default_59" [label="(1, 96, 7, 7)", style=solid]; +"452 dequantize_per_tensor_default_60" -> "498 add__5" [label="(1, 96, 7, 7)", style=solid]; +"453 dequantize_per_tensor_default_59" -> "459 conv2d_46" [label="(1, 96, 7, 7)", style=solid]; +"454 features_11_block_0_0_weight_bias" -> "459 conv2d_46" [label="(576,)", style=solid]; +"455 conv2d_46_scale_0" -> "458 dequantize_per_channel_default_46" [label="(576,)", style=solid]; +"456 conv2d_46_zero_point_0" -> "458 dequantize_per_channel_default_46" [label="(576,)", style=solid]; +"457 _frozen_param46" -> "458 dequantize_per_channel_default_46" [label="(576, 96, 1, 1)", style=solid]; +"458 dequantize_per_channel_default_46" -> "459 conv2d_46" [label="(576, 96, 1, 1)", style=solid]; +"459 conv2d_46" -> "460 hardswish__15" [label="(1, 576, 7, 7)", style=solid]; +"460 hardswish__15" -> "461 quantize_per_tensor_default_55" [label="(1, 576, 7, 7)", style=solid]; +"461 quantize_per_tensor_default_55" -> "462 dequantize_per_tensor_default_61" [label="(1, 576, 7, 7)", style=solid]; +"462 dequantize_per_tensor_default_61" -> "468 conv2d_47" [label="(1, 576, 7, 7)", style=solid]; +"463 features_11_block_1_0_weight_bias" -> "468 conv2d_47" [label="(576,)", style=solid]; +"464 conv2d_47_scale_0" -> "467 dequantize_per_channel_default_47" [label="(576,)", style=solid]; +"465 conv2d_47_zero_point_0" -> "467 dequantize_per_channel_default_47" [label="(576,)", style=solid]; +"466 _frozen_param47" -> "467 dequantize_per_channel_default_47" [label="(576, 1, 5, 5)", style=solid]; +"467 dequantize_per_channel_default_47" -> "468 conv2d_47" [label="(576, 1, 5, 5)", style=solid]; +"468 conv2d_47" -> "469 hardswish__16" [label="(1, 576, 7, 7)", style=solid]; +"469 hardswish__16" -> "470 quantize_per_tensor_default_56" [label="(1, 576, 7, 7)", style=solid]; +"469 hardswish__16" -> "489 mul_8" [label="(1, 576, 7, 7)", style=solid]; +"470 quantize_per_tensor_default_56" -> "471 dequantize_per_tensor_default_62" [label="(1, 576, 7, 7)", style=solid]; +"471 dequantize_per_tensor_default_62" -> "472 adaptive_avg_pool2d_8" [label="(1, 576, 7, 7)", style=solid]; +"472 adaptive_avg_pool2d_8" -> "473 quantize_per_tensor_default_57" [label="(1, 576, 1, 1)", style=solid]; +"473 quantize_per_tensor_default_57" -> "474 dequantize_per_tensor_default_63" [label="(1, 576, 1, 1)", style=solid]; +"474 dequantize_per_tensor_default_63" -> "479 conv2d_48" [label="(1, 576, 1, 1)", style=solid]; +"475 conv2d_48_scale_0" -> "478 dequantize_per_channel_default_48" [label="(144,)", style=solid]; +"476 conv2d_48_zero_point_0" -> "478 dequantize_per_channel_default_48" [label="(144,)", style=solid]; +"477 _frozen_param48" -> "478 dequantize_per_channel_default_48" [label="(144, 576, 1, 1)", style=solid]; +"478 dequantize_per_channel_default_48" -> "479 conv2d_48" [label="(144, 576, 1, 1)", style=solid]; +"479 conv2d_48" -> "480 relu_8" [label="(1, 144, 1, 1)", style=solid]; +"480 relu_8" -> "481 quantize_per_tensor_default_58" [label="(1, 144, 1, 1)", style=solid]; +"481 quantize_per_tensor_default_58" -> "482 dequantize_per_tensor_default_64" [label="(1, 144, 1, 1)", style=solid]; +"482 dequantize_per_tensor_default_64" -> "487 conv2d_49" [label="(1, 144, 1, 1)", style=solid]; +"483 conv2d_49_scale_0" -> "486 dequantize_per_channel_default_49" [label="(576,)", style=solid]; +"484 conv2d_49_zero_point_0" -> "486 dequantize_per_channel_default_49" [label="(576,)", style=solid]; +"485 _frozen_param49" -> "486 dequantize_per_channel_default_49" [label="(576, 144, 1, 1)", style=solid]; +"486 dequantize_per_channel_default_49" -> "487 conv2d_49" [label="(576, 144, 1, 1)", style=solid]; +"487 conv2d_49" -> "488 hardsigmoid_8" [label="(1, 576, 1, 1)", style=solid]; +"488 hardsigmoid_8" -> "489 mul_8" [label="(1, 576, 1, 1)", style=solid]; +"489 mul_8" -> "490 quantize_per_tensor_default_59" [label="(1, 576, 7, 7)", style=solid]; +"490 quantize_per_tensor_default_59" -> "491 dequantize_per_tensor_default_65" [label="(1, 576, 7, 7)", style=solid]; +"491 dequantize_per_tensor_default_65" -> "497 conv2d_50" [label="(1, 576, 7, 7)", style=solid]; +"492 features_11_block_3_0_weight_bias" -> "497 conv2d_50" [label="(96,)", style=solid]; +"493 conv2d_50_scale_0" -> "496 dequantize_per_channel_default_50" [label="(96,)", style=solid]; +"494 conv2d_50_zero_point_0" -> "496 dequantize_per_channel_default_50" [label="(96,)", style=solid]; +"495 _frozen_param50" -> "496 dequantize_per_channel_default_50" [label="(96, 576, 1, 1)", style=solid]; +"496 dequantize_per_channel_default_50" -> "497 conv2d_50" [label="(96, 576, 1, 1)", style=solid]; +"497 conv2d_50" -> "498 add__5" [label="(1, 96, 7, 7)", style=solid]; +"498 add__5" -> "499 quantize_per_tensor_default_60" [label="(1, 96, 7, 7)", style=solid]; +"499 quantize_per_tensor_default_60" -> "500 dequantize_per_tensor_default_66" [label="(1, 96, 7, 7)", style=solid]; +"500 dequantize_per_tensor_default_66" -> "506 conv2d_51" [label="(1, 96, 7, 7)", style=solid]; +"501 features_12_0_weight_bias" -> "506 conv2d_51" [label="(576,)", style=solid]; +"502 conv2d_51_scale_0" -> "505 dequantize_per_channel_default_51" [label="(576,)", style=solid]; +"503 conv2d_51_zero_point_0" -> "505 dequantize_per_channel_default_51" [label="(576,)", style=solid]; +"504 _frozen_param51" -> "505 dequantize_per_channel_default_51" [label="(576, 96, 1, 1)", style=solid]; +"505 dequantize_per_channel_default_51" -> "506 conv2d_51" [label="(576, 96, 1, 1)", style=solid]; +"506 conv2d_51" -> "507 hardswish__17" [label="(1, 576, 7, 7)", style=solid]; +"507 hardswish__17" -> "508 quantize_per_tensor_default_61" [label="(1, 576, 7, 7)", style=solid]; +"508 quantize_per_tensor_default_61" -> "509 dequantize_per_tensor_default_67" [label="(1, 576, 7, 7)", style=solid]; +"509 dequantize_per_tensor_default_67" -> "510 adaptive_avg_pool2d_9" [label="(1, 576, 7, 7)", style=solid]; +"510 adaptive_avg_pool2d_9" -> "511 quantize_per_tensor_default_62" [label="(1, 576, 1, 1)", style=solid]; +"511 quantize_per_tensor_default_62" -> "512 dequantize_per_tensor_default_68" [label="(1, 576, 1, 1)", style=solid]; +"512 dequantize_per_tensor_default_68" -> "513 flatten" [label="(1, 576, 1, 1)", style=solid]; +"513 flatten" -> "514 flatten_0_0_nncf_smooth_quant_0" [label="(1, 576)", style=solid]; +"514 flatten_0_0_nncf_smooth_quant_0" -> "515 quantize_per_tensor_default_63" [label="(1, 576)", style=solid]; +"515 quantize_per_tensor_default_63" -> "516 dequantize_per_tensor_default_69" [label="(1, 576)", style=solid]; +"516 dequantize_per_tensor_default_69" -> "521 linear" [label="(1, 576)", style=solid]; +"517 linear_scale_0" -> "520 dequantize_per_channel_default_52" [label="(1024,)", style=solid]; +"518 linear_zero_point_0" -> "520 dequantize_per_channel_default_52" [label="(1024,)", style=solid]; +"519 _frozen_param52" -> "520 dequantize_per_channel_default_52" [label="(1024, 576)", style=solid]; +"520 dequantize_per_channel_default_52" -> "521 linear" [label="(1024, 576)", style=solid]; +"521 linear" -> "522 hardswish__18" [label="(1, 1024)", style=solid]; +"522 hardswish__18" -> "523 dropout_" [label="(1, 1024)", style=solid]; +"523 dropout_" -> "524 dropout__0_0_nncf_smooth_quant_0" [label="(1, 1024)", style=solid]; +"524 dropout__0_0_nncf_smooth_quant_0" -> "525 quantize_per_tensor_default_64" [label="(1, 1024)", style=solid]; +"525 quantize_per_tensor_default_64" -> "526 dequantize_per_tensor_default_70" [label="(1, 1024)", style=solid]; +"526 dequantize_per_tensor_default_70" -> "531 linear_1" [label="(1, 1024)", style=solid]; +"527 linear_1_scale_0" -> "530 dequantize_per_channel_default_53" [label="(1000,)", style=solid]; +"528 linear_1_zero_point_0" -> "530 dequantize_per_channel_default_53" [label="(1000,)", style=solid]; +"529 _frozen_param53" -> "530 dequantize_per_channel_default_53" [label="(1000, 1024)", style=solid]; +"530 dequantize_per_channel_default_53" -> "531 linear_1" [label="(1000, 1024)", style=solid]; +"531 linear_1" -> "532 output" [label="(1, 1000)", style=solid]; +} diff --git a/tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/resnet18.dot b/tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/resnet18.dot new file mode 100644 index 00000000000..e9db48ad8f3 --- /dev/null +++ b/tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/resnet18.dot @@ -0,0 +1,435 @@ +strict digraph { +"0 fc_bias" [id=0, type=get_attr]; +"1 x" [id=1, type=input]; +"2 quantize_per_tensor_default" [id=2, type=quantize_per_tensor]; +"3 dequantize_per_tensor_default" [id=3, type=dequantize_per_tensor]; +"4 conv1_weight_bias" [id=4, type=get_attr]; +"5 conv2d_scale_0" [id=5, type=get_attr]; +"6 conv2d_zero_point_0" [id=6, type=get_attr]; +"7 _frozen_param0" [id=7, type=get_attr]; +"8 dequantize_per_channel_default" [id=8, type=dequantize_per_channel]; +"9 conv2d" [id=9, type=conv2d]; +"10 relu_" [id=10, type=relu_]; +"11 quantize_per_tensor_default_1" [id=11, type=quantize_per_tensor]; +"12 dequantize_per_tensor_default_1" [id=12, type=dequantize_per_tensor]; +"13 max_pool2d" [id=13, type=max_pool2d]; +"14 quantize_per_tensor_default_2" [id=14, type=quantize_per_tensor]; +"15 dequantize_per_tensor_default_3" [id=15, type=dequantize_per_tensor]; +"16 dequantize_per_tensor_default_2" [id=16, type=dequantize_per_tensor]; +"17 layer1_0_conv1_weight_bias" [id=17, type=get_attr]; +"18 conv2d_1_scale_0" [id=18, type=get_attr]; +"19 conv2d_1_zero_point_0" [id=19, type=get_attr]; +"20 _frozen_param1" [id=20, type=get_attr]; +"21 dequantize_per_channel_default_1" [id=21, type=dequantize_per_channel]; +"22 conv2d_1" [id=22, type=conv2d]; +"23 relu__1" [id=23, type=relu_]; +"24 quantize_per_tensor_default_3" [id=24, type=quantize_per_tensor]; +"25 dequantize_per_tensor_default_4" [id=25, type=dequantize_per_tensor]; +"26 layer1_0_conv2_weight_bias" [id=26, type=get_attr]; +"27 conv2d_2_scale_0" [id=27, type=get_attr]; +"28 conv2d_2_zero_point_0" [id=28, type=get_attr]; +"29 _frozen_param2" [id=29, type=get_attr]; +"30 dequantize_per_channel_default_2" [id=30, type=dequantize_per_channel]; +"31 conv2d_2" [id=31, type=conv2d]; +"32 add_" [id=32, type=add_]; +"33 relu__2" [id=33, type=relu_]; +"34 quantize_per_tensor_default_4" [id=34, type=quantize_per_tensor]; +"35 dequantize_per_tensor_default_6" [id=35, type=dequantize_per_tensor]; +"36 dequantize_per_tensor_default_5" [id=36, type=dequantize_per_tensor]; +"37 layer1_1_conv1_weight_bias" [id=37, type=get_attr]; +"38 conv2d_3_scale_0" [id=38, type=get_attr]; +"39 conv2d_3_zero_point_0" [id=39, type=get_attr]; +"40 _frozen_param3" [id=40, type=get_attr]; +"41 dequantize_per_channel_default_3" [id=41, type=dequantize_per_channel]; +"42 conv2d_3" [id=42, type=conv2d]; +"43 relu__3" [id=43, type=relu_]; +"44 quantize_per_tensor_default_5" [id=44, type=quantize_per_tensor]; +"45 dequantize_per_tensor_default_7" [id=45, type=dequantize_per_tensor]; +"46 layer1_1_conv2_weight_bias" [id=46, type=get_attr]; +"47 conv2d_4_scale_0" [id=47, type=get_attr]; +"48 conv2d_4_zero_point_0" [id=48, type=get_attr]; +"49 _frozen_param4" [id=49, type=get_attr]; +"50 dequantize_per_channel_default_4" [id=50, type=dequantize_per_channel]; +"51 conv2d_4" [id=51, type=conv2d]; +"52 add__1" [id=52, type=add_]; +"53 relu__4" [id=53, type=relu_]; +"54 quantize_per_tensor_default_6" [id=54, type=quantize_per_tensor]; +"55 dequantize_per_tensor_default_9" [id=55, type=dequantize_per_tensor]; +"56 dequantize_per_tensor_default_8" [id=56, type=dequantize_per_tensor]; +"57 layer2_0_conv1_weight_bias" [id=57, type=get_attr]; +"58 conv2d_5_scale_0" [id=58, type=get_attr]; +"59 conv2d_5_zero_point_0" [id=59, type=get_attr]; +"60 _frozen_param5" [id=60, type=get_attr]; +"61 dequantize_per_channel_default_5" [id=61, type=dequantize_per_channel]; +"62 conv2d_5" [id=62, type=conv2d]; +"63 relu__5" [id=63, type=relu_]; +"64 quantize_per_tensor_default_7" [id=64, type=quantize_per_tensor]; +"65 dequantize_per_tensor_default_10" [id=65, type=dequantize_per_tensor]; +"66 layer2_0_conv2_weight_bias" [id=66, type=get_attr]; +"67 conv2d_6_scale_0" [id=67, type=get_attr]; +"68 conv2d_6_zero_point_0" [id=68, type=get_attr]; +"69 _frozen_param6" [id=69, type=get_attr]; +"70 dequantize_per_channel_default_6" [id=70, type=dequantize_per_channel]; +"71 conv2d_6" [id=71, type=conv2d]; +"72 layer2_0_downsample_0_weight_bias" [id=72, type=get_attr]; +"73 conv2d_7_scale_0" [id=73, type=get_attr]; +"74 conv2d_7_zero_point_0" [id=74, type=get_attr]; +"75 _frozen_param7" [id=75, type=get_attr]; +"76 dequantize_per_channel_default_7" [id=76, type=dequantize_per_channel]; +"77 conv2d_7" [id=77, type=conv2d]; +"78 quantize_per_tensor_default_8" [id=78, type=quantize_per_tensor]; +"79 dequantize_per_tensor_default_11" [id=79, type=dequantize_per_tensor]; +"80 add__2" [id=80, type=add_]; +"81 relu__6" [id=81, type=relu_]; +"82 quantize_per_tensor_default_9" [id=82, type=quantize_per_tensor]; +"83 dequantize_per_tensor_default_13" [id=83, type=dequantize_per_tensor]; +"84 dequantize_per_tensor_default_12" [id=84, type=dequantize_per_tensor]; +"85 layer2_1_conv1_weight_bias" [id=85, type=get_attr]; +"86 conv2d_8_scale_0" [id=86, type=get_attr]; +"87 conv2d_8_zero_point_0" [id=87, type=get_attr]; +"88 _frozen_param8" [id=88, type=get_attr]; +"89 dequantize_per_channel_default_8" [id=89, type=dequantize_per_channel]; +"90 conv2d_8" [id=90, type=conv2d]; +"91 relu__7" [id=91, type=relu_]; +"92 quantize_per_tensor_default_10" [id=92, type=quantize_per_tensor]; +"93 dequantize_per_tensor_default_14" [id=93, type=dequantize_per_tensor]; +"94 layer2_1_conv2_weight_bias" [id=94, type=get_attr]; +"95 conv2d_9_scale_0" [id=95, type=get_attr]; +"96 conv2d_9_zero_point_0" [id=96, type=get_attr]; +"97 _frozen_param9" [id=97, type=get_attr]; +"98 dequantize_per_channel_default_9" [id=98, type=dequantize_per_channel]; +"99 conv2d_9" [id=99, type=conv2d]; +"100 add__3" [id=100, type=add_]; +"101 relu__8" [id=101, type=relu_]; +"102 quantize_per_tensor_default_11" [id=102, type=quantize_per_tensor]; +"103 dequantize_per_tensor_default_16" [id=103, type=dequantize_per_tensor]; +"104 dequantize_per_tensor_default_15" [id=104, type=dequantize_per_tensor]; +"105 layer3_0_conv1_weight_bias" [id=105, type=get_attr]; +"106 conv2d_10_scale_0" [id=106, type=get_attr]; +"107 conv2d_10_zero_point_0" [id=107, type=get_attr]; +"108 _frozen_param10" [id=108, type=get_attr]; +"109 dequantize_per_channel_default_10" [id=109, type=dequantize_per_channel]; +"110 conv2d_10" [id=110, type=conv2d]; +"111 relu__9" [id=111, type=relu_]; +"112 quantize_per_tensor_default_12" [id=112, type=quantize_per_tensor]; +"113 dequantize_per_tensor_default_17" [id=113, type=dequantize_per_tensor]; +"114 layer3_0_conv2_weight_bias" [id=114, type=get_attr]; +"115 conv2d_11_scale_0" [id=115, type=get_attr]; +"116 conv2d_11_zero_point_0" [id=116, type=get_attr]; +"117 _frozen_param11" [id=117, type=get_attr]; +"118 dequantize_per_channel_default_11" [id=118, type=dequantize_per_channel]; +"119 conv2d_11" [id=119, type=conv2d]; +"120 layer3_0_downsample_0_weight_bias" [id=120, type=get_attr]; +"121 conv2d_12_scale_0" [id=121, type=get_attr]; +"122 conv2d_12_zero_point_0" [id=122, type=get_attr]; +"123 _frozen_param12" [id=123, type=get_attr]; +"124 dequantize_per_channel_default_12" [id=124, type=dequantize_per_channel]; +"125 conv2d_12" [id=125, type=conv2d]; +"126 quantize_per_tensor_default_13" [id=126, type=quantize_per_tensor]; +"127 dequantize_per_tensor_default_18" [id=127, type=dequantize_per_tensor]; +"128 add__4" [id=128, type=add_]; +"129 relu__10" [id=129, type=relu_]; +"130 quantize_per_tensor_default_14" [id=130, type=quantize_per_tensor]; +"131 dequantize_per_tensor_default_20" [id=131, type=dequantize_per_tensor]; +"132 dequantize_per_tensor_default_19" [id=132, type=dequantize_per_tensor]; +"133 layer3_1_conv1_weight_bias" [id=133, type=get_attr]; +"134 conv2d_13_scale_0" [id=134, type=get_attr]; +"135 conv2d_13_zero_point_0" [id=135, type=get_attr]; +"136 _frozen_param13" [id=136, type=get_attr]; +"137 dequantize_per_channel_default_13" [id=137, type=dequantize_per_channel]; +"138 conv2d_13" [id=138, type=conv2d]; +"139 relu__11" [id=139, type=relu_]; +"140 quantize_per_tensor_default_15" [id=140, type=quantize_per_tensor]; +"141 dequantize_per_tensor_default_21" [id=141, type=dequantize_per_tensor]; +"142 layer3_1_conv2_weight_bias" [id=142, type=get_attr]; +"143 conv2d_14_scale_0" [id=143, type=get_attr]; +"144 conv2d_14_zero_point_0" [id=144, type=get_attr]; +"145 _frozen_param14" [id=145, type=get_attr]; +"146 dequantize_per_channel_default_14" [id=146, type=dequantize_per_channel]; +"147 conv2d_14" [id=147, type=conv2d]; +"148 add__5" [id=148, type=add_]; +"149 relu__12" [id=149, type=relu_]; +"150 quantize_per_tensor_default_16" [id=150, type=quantize_per_tensor]; +"151 dequantize_per_tensor_default_23" [id=151, type=dequantize_per_tensor]; +"152 dequantize_per_tensor_default_22" [id=152, type=dequantize_per_tensor]; +"153 layer4_0_conv1_weight_bias" [id=153, type=get_attr]; +"154 conv2d_15_scale_0" [id=154, type=get_attr]; +"155 conv2d_15_zero_point_0" [id=155, type=get_attr]; +"156 _frozen_param15" [id=156, type=get_attr]; +"157 dequantize_per_channel_default_15" [id=157, type=dequantize_per_channel]; +"158 conv2d_15" [id=158, type=conv2d]; +"159 relu__13" [id=159, type=relu_]; +"160 quantize_per_tensor_default_17" [id=160, type=quantize_per_tensor]; +"161 dequantize_per_tensor_default_24" [id=161, type=dequantize_per_tensor]; +"162 layer4_0_conv2_weight_bias" [id=162, type=get_attr]; +"163 conv2d_16_scale_0" [id=163, type=get_attr]; +"164 conv2d_16_zero_point_0" [id=164, type=get_attr]; +"165 _frozen_param16" [id=165, type=get_attr]; +"166 dequantize_per_channel_default_16" [id=166, type=dequantize_per_channel]; +"167 conv2d_16" [id=167, type=conv2d]; +"168 layer4_0_downsample_0_weight_bias" [id=168, type=get_attr]; +"169 conv2d_17_scale_0" [id=169, type=get_attr]; +"170 conv2d_17_zero_point_0" [id=170, type=get_attr]; +"171 _frozen_param17" [id=171, type=get_attr]; +"172 dequantize_per_channel_default_17" [id=172, type=dequantize_per_channel]; +"173 conv2d_17" [id=173, type=conv2d]; +"174 quantize_per_tensor_default_18" [id=174, type=quantize_per_tensor]; +"175 dequantize_per_tensor_default_25" [id=175, type=dequantize_per_tensor]; +"176 add__6" [id=176, type=add_]; +"177 relu__14" [id=177, type=relu_]; +"178 quantize_per_tensor_default_19" [id=178, type=quantize_per_tensor]; +"179 dequantize_per_tensor_default_27" [id=179, type=dequantize_per_tensor]; +"180 dequantize_per_tensor_default_26" [id=180, type=dequantize_per_tensor]; +"181 layer4_1_conv1_weight_bias" [id=181, type=get_attr]; +"182 conv2d_18_scale_0" [id=182, type=get_attr]; +"183 conv2d_18_zero_point_0" [id=183, type=get_attr]; +"184 _frozen_param18" [id=184, type=get_attr]; +"185 dequantize_per_channel_default_18" [id=185, type=dequantize_per_channel]; +"186 conv2d_18" [id=186, type=conv2d]; +"187 relu__15" [id=187, type=relu_]; +"188 quantize_per_tensor_default_20" [id=188, type=quantize_per_tensor]; +"189 dequantize_per_tensor_default_28" [id=189, type=dequantize_per_tensor]; +"190 layer4_1_conv2_weight_bias" [id=190, type=get_attr]; +"191 conv2d_19_scale_0" [id=191, type=get_attr]; +"192 conv2d_19_zero_point_0" [id=192, type=get_attr]; +"193 _frozen_param19" [id=193, type=get_attr]; +"194 dequantize_per_channel_default_19" [id=194, type=dequantize_per_channel]; +"195 conv2d_19" [id=195, type=conv2d]; +"196 add__7" [id=196, type=add_]; +"197 relu__16" [id=197, type=relu_]; +"198 quantize_per_tensor_default_21" [id=198, type=quantize_per_tensor]; +"199 dequantize_per_tensor_default_29" [id=199, type=dequantize_per_tensor]; +"200 adaptive_avg_pool2d" [id=200, type=adaptive_avg_pool2d]; +"201 quantize_per_tensor_default_22" [id=201, type=quantize_per_tensor]; +"202 dequantize_per_tensor_default_30" [id=202, type=dequantize_per_tensor]; +"203 flatten" [id=203, type=flatten]; +"204 flatten_0_0_nncf_smooth_quant_0" [id=204, type=call_module]; +"205 quantize_per_tensor_default_23" [id=205, type=quantize_per_tensor]; +"206 dequantize_per_tensor_default_31" [id=206, type=dequantize_per_tensor]; +"207 linear_scale_0" [id=207, type=get_attr]; +"208 linear_zero_point_0" [id=208, type=get_attr]; +"209 _frozen_param20" [id=209, type=get_attr]; +"210 dequantize_per_channel_default_20" [id=210, type=dequantize_per_channel]; +"211 linear" [id=211, type=linear]; +"212 output" [id=212, type=output]; +"0 fc_bias" -> "211 linear" [label="(1000,)", style=solid]; +"1 x" -> "2 quantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; +"2 quantize_per_tensor_default" -> "3 dequantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; +"3 dequantize_per_tensor_default" -> "9 conv2d" [label="(1, 3, 224, 224)", style=solid]; +"4 conv1_weight_bias" -> "9 conv2d" [label="(64,)", style=solid]; +"5 conv2d_scale_0" -> "8 dequantize_per_channel_default" [label="(64,)", style=solid]; +"6 conv2d_zero_point_0" -> "8 dequantize_per_channel_default" [label="(64,)", style=solid]; +"7 _frozen_param0" -> "8 dequantize_per_channel_default" [label="(64, 3, 7, 7)", style=solid]; +"8 dequantize_per_channel_default" -> "9 conv2d" [label="(64, 3, 7, 7)", style=solid]; +"9 conv2d" -> "10 relu_" [label="(1, 64, 112, 112)", style=solid]; +"10 relu_" -> "11 quantize_per_tensor_default_1" [label="(1, 64, 112, 112)", style=solid]; +"11 quantize_per_tensor_default_1" -> "12 dequantize_per_tensor_default_1" [label="(1, 64, 112, 112)", style=solid]; +"12 dequantize_per_tensor_default_1" -> "13 max_pool2d" [label="(1, 64, 112, 112)", style=solid]; +"13 max_pool2d" -> "14 quantize_per_tensor_default_2" [label="(1, 64, 56, 56)", style=solid]; +"14 quantize_per_tensor_default_2" -> "15 dequantize_per_tensor_default_3" [label="(1, 64, 56, 56)", style=solid]; +"14 quantize_per_tensor_default_2" -> "16 dequantize_per_tensor_default_2" [label="(1, 64, 56, 56)", style=solid]; +"15 dequantize_per_tensor_default_3" -> "32 add_" [label="(1, 64, 56, 56)", style=solid]; +"16 dequantize_per_tensor_default_2" -> "22 conv2d_1" [label="(1, 64, 56, 56)", style=solid]; +"17 layer1_0_conv1_weight_bias" -> "22 conv2d_1" [label="(64,)", style=solid]; +"18 conv2d_1_scale_0" -> "21 dequantize_per_channel_default_1" [label="(64,)", style=solid]; +"19 conv2d_1_zero_point_0" -> "21 dequantize_per_channel_default_1" [label="(64,)", style=solid]; +"20 _frozen_param1" -> "21 dequantize_per_channel_default_1" [label="(64, 64, 3, 3)", style=solid]; +"21 dequantize_per_channel_default_1" -> "22 conv2d_1" [label="(64, 64, 3, 3)", style=solid]; +"22 conv2d_1" -> "23 relu__1" [label="(1, 64, 56, 56)", style=solid]; +"23 relu__1" -> "24 quantize_per_tensor_default_3" [label="(1, 64, 56, 56)", style=solid]; +"24 quantize_per_tensor_default_3" -> "25 dequantize_per_tensor_default_4" [label="(1, 64, 56, 56)", style=solid]; +"25 dequantize_per_tensor_default_4" -> "31 conv2d_2" [label="(1, 64, 56, 56)", style=solid]; +"26 layer1_0_conv2_weight_bias" -> "31 conv2d_2" [label="(64,)", style=solid]; +"27 conv2d_2_scale_0" -> "30 dequantize_per_channel_default_2" [label="(64,)", style=solid]; +"28 conv2d_2_zero_point_0" -> "30 dequantize_per_channel_default_2" [label="(64,)", style=solid]; +"29 _frozen_param2" -> "30 dequantize_per_channel_default_2" [label="(64, 64, 3, 3)", style=solid]; +"30 dequantize_per_channel_default_2" -> "31 conv2d_2" [label="(64, 64, 3, 3)", style=solid]; +"31 conv2d_2" -> "32 add_" [label="(1, 64, 56, 56)", style=solid]; +"32 add_" -> "33 relu__2" [label="(1, 64, 56, 56)", style=solid]; +"33 relu__2" -> "34 quantize_per_tensor_default_4" [label="(1, 64, 56, 56)", style=solid]; +"34 quantize_per_tensor_default_4" -> "35 dequantize_per_tensor_default_6" [label="(1, 64, 56, 56)", style=solid]; +"34 quantize_per_tensor_default_4" -> "36 dequantize_per_tensor_default_5" [label="(1, 64, 56, 56)", style=solid]; +"35 dequantize_per_tensor_default_6" -> "52 add__1" [label="(1, 64, 56, 56)", style=solid]; +"36 dequantize_per_tensor_default_5" -> "42 conv2d_3" [label="(1, 64, 56, 56)", style=solid]; +"37 layer1_1_conv1_weight_bias" -> "42 conv2d_3" [label="(64,)", style=solid]; +"38 conv2d_3_scale_0" -> "41 dequantize_per_channel_default_3" [label="(64,)", style=solid]; +"39 conv2d_3_zero_point_0" -> "41 dequantize_per_channel_default_3" [label="(64,)", style=solid]; +"40 _frozen_param3" -> "41 dequantize_per_channel_default_3" [label="(64, 64, 3, 3)", style=solid]; +"41 dequantize_per_channel_default_3" -> "42 conv2d_3" [label="(64, 64, 3, 3)", style=solid]; +"42 conv2d_3" -> "43 relu__3" [label="(1, 64, 56, 56)", style=solid]; +"43 relu__3" -> "44 quantize_per_tensor_default_5" [label="(1, 64, 56, 56)", style=solid]; +"44 quantize_per_tensor_default_5" -> "45 dequantize_per_tensor_default_7" [label="(1, 64, 56, 56)", style=solid]; +"45 dequantize_per_tensor_default_7" -> "51 conv2d_4" [label="(1, 64, 56, 56)", style=solid]; +"46 layer1_1_conv2_weight_bias" -> "51 conv2d_4" [label="(64,)", style=solid]; +"47 conv2d_4_scale_0" -> "50 dequantize_per_channel_default_4" [label="(64,)", style=solid]; +"48 conv2d_4_zero_point_0" -> "50 dequantize_per_channel_default_4" [label="(64,)", style=solid]; +"49 _frozen_param4" -> "50 dequantize_per_channel_default_4" [label="(64, 64, 3, 3)", style=solid]; +"50 dequantize_per_channel_default_4" -> "51 conv2d_4" [label="(64, 64, 3, 3)", style=solid]; +"51 conv2d_4" -> "52 add__1" [label="(1, 64, 56, 56)", style=solid]; +"52 add__1" -> "53 relu__4" [label="(1, 64, 56, 56)", style=solid]; +"53 relu__4" -> "54 quantize_per_tensor_default_6" [label="(1, 64, 56, 56)", style=solid]; +"54 quantize_per_tensor_default_6" -> "55 dequantize_per_tensor_default_9" [label="(1, 64, 56, 56)", style=solid]; +"54 quantize_per_tensor_default_6" -> "56 dequantize_per_tensor_default_8" [label="(1, 64, 56, 56)", style=solid]; +"55 dequantize_per_tensor_default_9" -> "77 conv2d_7" [label="(1, 64, 56, 56)", style=solid]; +"56 dequantize_per_tensor_default_8" -> "62 conv2d_5" [label="(1, 64, 56, 56)", style=solid]; +"57 layer2_0_conv1_weight_bias" -> "62 conv2d_5" [label="(128,)", style=solid]; +"58 conv2d_5_scale_0" -> "61 dequantize_per_channel_default_5" [label="(128,)", style=solid]; +"59 conv2d_5_zero_point_0" -> "61 dequantize_per_channel_default_5" [label="(128,)", style=solid]; +"60 _frozen_param5" -> "61 dequantize_per_channel_default_5" [label="(128, 64, 3, 3)", style=solid]; +"61 dequantize_per_channel_default_5" -> "62 conv2d_5" [label="(128, 64, 3, 3)", style=solid]; +"62 conv2d_5" -> "63 relu__5" [label="(1, 128, 28, 28)", style=solid]; +"63 relu__5" -> "64 quantize_per_tensor_default_7" [label="(1, 128, 28, 28)", style=solid]; +"64 quantize_per_tensor_default_7" -> "65 dequantize_per_tensor_default_10" [label="(1, 128, 28, 28)", style=solid]; +"65 dequantize_per_tensor_default_10" -> "71 conv2d_6" [label="(1, 128, 28, 28)", style=solid]; +"66 layer2_0_conv2_weight_bias" -> "71 conv2d_6" [label="(128,)", style=solid]; +"67 conv2d_6_scale_0" -> "70 dequantize_per_channel_default_6" [label="(128,)", style=solid]; +"68 conv2d_6_zero_point_0" -> "70 dequantize_per_channel_default_6" [label="(128,)", style=solid]; +"69 _frozen_param6" -> "70 dequantize_per_channel_default_6" [label="(128, 128, 3, 3)", style=solid]; +"70 dequantize_per_channel_default_6" -> "71 conv2d_6" [label="(128, 128, 3, 3)", style=solid]; +"71 conv2d_6" -> "80 add__2" [label="(1, 128, 28, 28)", style=solid]; +"72 layer2_0_downsample_0_weight_bias" -> "77 conv2d_7" [label="(128,)", style=solid]; +"73 conv2d_7_scale_0" -> "76 dequantize_per_channel_default_7" [label="(128,)", style=solid]; +"74 conv2d_7_zero_point_0" -> "76 dequantize_per_channel_default_7" [label="(128,)", style=solid]; +"75 _frozen_param7" -> "76 dequantize_per_channel_default_7" [label="(128, 64, 1, 1)", style=solid]; +"76 dequantize_per_channel_default_7" -> "77 conv2d_7" [label="(128, 64, 1, 1)", style=solid]; +"77 conv2d_7" -> "78 quantize_per_tensor_default_8" [label="(1, 128, 28, 28)", style=solid]; +"78 quantize_per_tensor_default_8" -> "79 dequantize_per_tensor_default_11" [label="(1, 128, 28, 28)", style=solid]; +"79 dequantize_per_tensor_default_11" -> "80 add__2" [label="(1, 128, 28, 28)", style=solid]; +"80 add__2" -> "81 relu__6" [label="(1, 128, 28, 28)", style=solid]; +"81 relu__6" -> "82 quantize_per_tensor_default_9" [label="(1, 128, 28, 28)", style=solid]; +"82 quantize_per_tensor_default_9" -> "83 dequantize_per_tensor_default_13" [label="(1, 128, 28, 28)", style=solid]; +"82 quantize_per_tensor_default_9" -> "84 dequantize_per_tensor_default_12" [label="(1, 128, 28, 28)", style=solid]; +"83 dequantize_per_tensor_default_13" -> "100 add__3" [label="(1, 128, 28, 28)", style=solid]; +"84 dequantize_per_tensor_default_12" -> "90 conv2d_8" [label="(1, 128, 28, 28)", style=solid]; +"85 layer2_1_conv1_weight_bias" -> "90 conv2d_8" [label="(128,)", style=solid]; +"86 conv2d_8_scale_0" -> "89 dequantize_per_channel_default_8" [label="(128,)", style=solid]; +"87 conv2d_8_zero_point_0" -> "89 dequantize_per_channel_default_8" [label="(128,)", style=solid]; +"88 _frozen_param8" -> "89 dequantize_per_channel_default_8" [label="(128, 128, 3, 3)", style=solid]; +"89 dequantize_per_channel_default_8" -> "90 conv2d_8" [label="(128, 128, 3, 3)", style=solid]; +"90 conv2d_8" -> "91 relu__7" [label="(1, 128, 28, 28)", style=solid]; +"91 relu__7" -> "92 quantize_per_tensor_default_10" [label="(1, 128, 28, 28)", style=solid]; +"92 quantize_per_tensor_default_10" -> "93 dequantize_per_tensor_default_14" [label="(1, 128, 28, 28)", style=solid]; +"93 dequantize_per_tensor_default_14" -> "99 conv2d_9" [label="(1, 128, 28, 28)", style=solid]; +"94 layer2_1_conv2_weight_bias" -> "99 conv2d_9" [label="(128,)", style=solid]; +"95 conv2d_9_scale_0" -> "98 dequantize_per_channel_default_9" [label="(128,)", style=solid]; +"96 conv2d_9_zero_point_0" -> "98 dequantize_per_channel_default_9" [label="(128,)", style=solid]; +"97 _frozen_param9" -> "98 dequantize_per_channel_default_9" [label="(128, 128, 3, 3)", style=solid]; +"98 dequantize_per_channel_default_9" -> "99 conv2d_9" [label="(128, 128, 3, 3)", style=solid]; +"99 conv2d_9" -> "100 add__3" [label="(1, 128, 28, 28)", style=solid]; +"100 add__3" -> "101 relu__8" [label="(1, 128, 28, 28)", style=solid]; +"101 relu__8" -> "102 quantize_per_tensor_default_11" [label="(1, 128, 28, 28)", style=solid]; +"102 quantize_per_tensor_default_11" -> "103 dequantize_per_tensor_default_16" [label="(1, 128, 28, 28)", style=solid]; +"102 quantize_per_tensor_default_11" -> "104 dequantize_per_tensor_default_15" [label="(1, 128, 28, 28)", style=solid]; +"103 dequantize_per_tensor_default_16" -> "125 conv2d_12" [label="(1, 128, 28, 28)", style=solid]; +"104 dequantize_per_tensor_default_15" -> "110 conv2d_10" [label="(1, 128, 28, 28)", style=solid]; +"105 layer3_0_conv1_weight_bias" -> "110 conv2d_10" [label="(256,)", style=solid]; +"106 conv2d_10_scale_0" -> "109 dequantize_per_channel_default_10" [label="(256,)", style=solid]; +"107 conv2d_10_zero_point_0" -> "109 dequantize_per_channel_default_10" [label="(256,)", style=solid]; +"108 _frozen_param10" -> "109 dequantize_per_channel_default_10" [label="(256, 128, 3, 3)", style=solid]; +"109 dequantize_per_channel_default_10" -> "110 conv2d_10" [label="(256, 128, 3, 3)", style=solid]; +"110 conv2d_10" -> "111 relu__9" [label="(1, 256, 14, 14)", style=solid]; +"111 relu__9" -> "112 quantize_per_tensor_default_12" [label="(1, 256, 14, 14)", style=solid]; +"112 quantize_per_tensor_default_12" -> "113 dequantize_per_tensor_default_17" [label="(1, 256, 14, 14)", style=solid]; +"113 dequantize_per_tensor_default_17" -> "119 conv2d_11" [label="(1, 256, 14, 14)", style=solid]; +"114 layer3_0_conv2_weight_bias" -> "119 conv2d_11" [label="(256,)", style=solid]; +"115 conv2d_11_scale_0" -> "118 dequantize_per_channel_default_11" [label="(256,)", style=solid]; +"116 conv2d_11_zero_point_0" -> "118 dequantize_per_channel_default_11" [label="(256,)", style=solid]; +"117 _frozen_param11" -> "118 dequantize_per_channel_default_11" [label="(256, 256, 3, 3)", style=solid]; +"118 dequantize_per_channel_default_11" -> "119 conv2d_11" [label="(256, 256, 3, 3)", style=solid]; +"119 conv2d_11" -> "128 add__4" [label="(1, 256, 14, 14)", style=solid]; +"120 layer3_0_downsample_0_weight_bias" -> "125 conv2d_12" [label="(256,)", style=solid]; +"121 conv2d_12_scale_0" -> "124 dequantize_per_channel_default_12" [label="(256,)", style=solid]; +"122 conv2d_12_zero_point_0" -> "124 dequantize_per_channel_default_12" [label="(256,)", style=solid]; +"123 _frozen_param12" -> "124 dequantize_per_channel_default_12" [label="(256, 128, 1, 1)", style=solid]; +"124 dequantize_per_channel_default_12" -> "125 conv2d_12" [label="(256, 128, 1, 1)", style=solid]; +"125 conv2d_12" -> "126 quantize_per_tensor_default_13" [label="(1, 256, 14, 14)", style=solid]; +"126 quantize_per_tensor_default_13" -> "127 dequantize_per_tensor_default_18" [label="(1, 256, 14, 14)", style=solid]; +"127 dequantize_per_tensor_default_18" -> "128 add__4" [label="(1, 256, 14, 14)", style=solid]; +"128 add__4" -> "129 relu__10" [label="(1, 256, 14, 14)", style=solid]; +"129 relu__10" -> "130 quantize_per_tensor_default_14" [label="(1, 256, 14, 14)", style=solid]; +"130 quantize_per_tensor_default_14" -> "131 dequantize_per_tensor_default_20" [label="(1, 256, 14, 14)", style=solid]; +"130 quantize_per_tensor_default_14" -> "132 dequantize_per_tensor_default_19" [label="(1, 256, 14, 14)", style=solid]; +"131 dequantize_per_tensor_default_20" -> "148 add__5" [label="(1, 256, 14, 14)", style=solid]; +"132 dequantize_per_tensor_default_19" -> "138 conv2d_13" [label="(1, 256, 14, 14)", style=solid]; +"133 layer3_1_conv1_weight_bias" -> "138 conv2d_13" [label="(256,)", style=solid]; +"134 conv2d_13_scale_0" -> "137 dequantize_per_channel_default_13" [label="(256,)", style=solid]; +"135 conv2d_13_zero_point_0" -> "137 dequantize_per_channel_default_13" [label="(256,)", style=solid]; +"136 _frozen_param13" -> "137 dequantize_per_channel_default_13" [label="(256, 256, 3, 3)", style=solid]; +"137 dequantize_per_channel_default_13" -> "138 conv2d_13" [label="(256, 256, 3, 3)", style=solid]; +"138 conv2d_13" -> "139 relu__11" [label="(1, 256, 14, 14)", style=solid]; +"139 relu__11" -> "140 quantize_per_tensor_default_15" [label="(1, 256, 14, 14)", style=solid]; +"140 quantize_per_tensor_default_15" -> "141 dequantize_per_tensor_default_21" [label="(1, 256, 14, 14)", style=solid]; +"141 dequantize_per_tensor_default_21" -> "147 conv2d_14" [label="(1, 256, 14, 14)", style=solid]; +"142 layer3_1_conv2_weight_bias" -> "147 conv2d_14" [label="(256,)", style=solid]; +"143 conv2d_14_scale_0" -> "146 dequantize_per_channel_default_14" [label="(256,)", style=solid]; +"144 conv2d_14_zero_point_0" -> "146 dequantize_per_channel_default_14" [label="(256,)", style=solid]; +"145 _frozen_param14" -> "146 dequantize_per_channel_default_14" [label="(256, 256, 3, 3)", style=solid]; +"146 dequantize_per_channel_default_14" -> "147 conv2d_14" [label="(256, 256, 3, 3)", style=solid]; +"147 conv2d_14" -> "148 add__5" [label="(1, 256, 14, 14)", style=solid]; +"148 add__5" -> "149 relu__12" [label="(1, 256, 14, 14)", style=solid]; +"149 relu__12" -> "150 quantize_per_tensor_default_16" [label="(1, 256, 14, 14)", style=solid]; +"150 quantize_per_tensor_default_16" -> "151 dequantize_per_tensor_default_23" [label="(1, 256, 14, 14)", style=solid]; +"150 quantize_per_tensor_default_16" -> "152 dequantize_per_tensor_default_22" [label="(1, 256, 14, 14)", style=solid]; +"151 dequantize_per_tensor_default_23" -> "173 conv2d_17" [label="(1, 256, 14, 14)", style=solid]; +"152 dequantize_per_tensor_default_22" -> "158 conv2d_15" [label="(1, 256, 14, 14)", style=solid]; +"153 layer4_0_conv1_weight_bias" -> "158 conv2d_15" [label="(512,)", style=solid]; +"154 conv2d_15_scale_0" -> "157 dequantize_per_channel_default_15" [label="(512,)", style=solid]; +"155 conv2d_15_zero_point_0" -> "157 dequantize_per_channel_default_15" [label="(512,)", style=solid]; +"156 _frozen_param15" -> "157 dequantize_per_channel_default_15" [label="(512, 256, 3, 3)", style=solid]; +"157 dequantize_per_channel_default_15" -> "158 conv2d_15" [label="(512, 256, 3, 3)", style=solid]; +"158 conv2d_15" -> "159 relu__13" [label="(1, 512, 7, 7)", style=solid]; +"159 relu__13" -> "160 quantize_per_tensor_default_17" [label="(1, 512, 7, 7)", style=solid]; +"160 quantize_per_tensor_default_17" -> "161 dequantize_per_tensor_default_24" [label="(1, 512, 7, 7)", style=solid]; +"161 dequantize_per_tensor_default_24" -> "167 conv2d_16" [label="(1, 512, 7, 7)", style=solid]; +"162 layer4_0_conv2_weight_bias" -> "167 conv2d_16" [label="(512,)", style=solid]; +"163 conv2d_16_scale_0" -> "166 dequantize_per_channel_default_16" [label="(512,)", style=solid]; +"164 conv2d_16_zero_point_0" -> "166 dequantize_per_channel_default_16" [label="(512,)", style=solid]; +"165 _frozen_param16" -> "166 dequantize_per_channel_default_16" [label="(512, 512, 3, 3)", style=solid]; +"166 dequantize_per_channel_default_16" -> "167 conv2d_16" [label="(512, 512, 3, 3)", style=solid]; +"167 conv2d_16" -> "176 add__6" [label="(1, 512, 7, 7)", style=solid]; +"168 layer4_0_downsample_0_weight_bias" -> "173 conv2d_17" [label="(512,)", style=solid]; +"169 conv2d_17_scale_0" -> "172 dequantize_per_channel_default_17" [label="(512,)", style=solid]; +"170 conv2d_17_zero_point_0" -> "172 dequantize_per_channel_default_17" [label="(512,)", style=solid]; +"171 _frozen_param17" -> "172 dequantize_per_channel_default_17" [label="(512, 256, 1, 1)", style=solid]; +"172 dequantize_per_channel_default_17" -> "173 conv2d_17" [label="(512, 256, 1, 1)", style=solid]; +"173 conv2d_17" -> "174 quantize_per_tensor_default_18" [label="(1, 512, 7, 7)", style=solid]; +"174 quantize_per_tensor_default_18" -> "175 dequantize_per_tensor_default_25" [label="(1, 512, 7, 7)", style=solid]; +"175 dequantize_per_tensor_default_25" -> "176 add__6" [label="(1, 512, 7, 7)", style=solid]; +"176 add__6" -> "177 relu__14" [label="(1, 512, 7, 7)", style=solid]; +"177 relu__14" -> "178 quantize_per_tensor_default_19" [label="(1, 512, 7, 7)", style=solid]; +"178 quantize_per_tensor_default_19" -> "179 dequantize_per_tensor_default_27" [label="(1, 512, 7, 7)", style=solid]; +"178 quantize_per_tensor_default_19" -> "180 dequantize_per_tensor_default_26" [label="(1, 512, 7, 7)", style=solid]; +"179 dequantize_per_tensor_default_27" -> "196 add__7" [label="(1, 512, 7, 7)", style=solid]; +"180 dequantize_per_tensor_default_26" -> "186 conv2d_18" [label="(1, 512, 7, 7)", style=solid]; +"181 layer4_1_conv1_weight_bias" -> "186 conv2d_18" [label="(512,)", style=solid]; +"182 conv2d_18_scale_0" -> "185 dequantize_per_channel_default_18" [label="(512,)", style=solid]; +"183 conv2d_18_zero_point_0" -> "185 dequantize_per_channel_default_18" [label="(512,)", style=solid]; +"184 _frozen_param18" -> "185 dequantize_per_channel_default_18" [label="(512, 512, 3, 3)", style=solid]; +"185 dequantize_per_channel_default_18" -> "186 conv2d_18" [label="(512, 512, 3, 3)", style=solid]; +"186 conv2d_18" -> "187 relu__15" [label="(1, 512, 7, 7)", style=solid]; +"187 relu__15" -> "188 quantize_per_tensor_default_20" [label="(1, 512, 7, 7)", style=solid]; +"188 quantize_per_tensor_default_20" -> "189 dequantize_per_tensor_default_28" [label="(1, 512, 7, 7)", style=solid]; +"189 dequantize_per_tensor_default_28" -> "195 conv2d_19" [label="(1, 512, 7, 7)", style=solid]; +"190 layer4_1_conv2_weight_bias" -> "195 conv2d_19" [label="(512,)", style=solid]; +"191 conv2d_19_scale_0" -> "194 dequantize_per_channel_default_19" [label="(512,)", style=solid]; +"192 conv2d_19_zero_point_0" -> "194 dequantize_per_channel_default_19" [label="(512,)", style=solid]; +"193 _frozen_param19" -> "194 dequantize_per_channel_default_19" [label="(512, 512, 3, 3)", style=solid]; +"194 dequantize_per_channel_default_19" -> "195 conv2d_19" [label="(512, 512, 3, 3)", style=solid]; +"195 conv2d_19" -> "196 add__7" [label="(1, 512, 7, 7)", style=solid]; +"196 add__7" -> "197 relu__16" [label="(1, 512, 7, 7)", style=solid]; +"197 relu__16" -> "198 quantize_per_tensor_default_21" [label="(1, 512, 7, 7)", style=solid]; +"198 quantize_per_tensor_default_21" -> "199 dequantize_per_tensor_default_29" [label="(1, 512, 7, 7)", style=solid]; +"199 dequantize_per_tensor_default_29" -> "200 adaptive_avg_pool2d" [label="(1, 512, 7, 7)", style=solid]; +"200 adaptive_avg_pool2d" -> "201 quantize_per_tensor_default_22" [label="(1, 512, 1, 1)", style=solid]; +"201 quantize_per_tensor_default_22" -> "202 dequantize_per_tensor_default_30" [label="(1, 512, 1, 1)", style=solid]; +"202 dequantize_per_tensor_default_30" -> "203 flatten" [label="(1, 512, 1, 1)", style=solid]; +"203 flatten" -> "204 flatten_0_0_nncf_smooth_quant_0" [label="(1, 512)", style=solid]; +"204 flatten_0_0_nncf_smooth_quant_0" -> "205 quantize_per_tensor_default_23" [label="(1, 512)", style=solid]; +"205 quantize_per_tensor_default_23" -> "206 dequantize_per_tensor_default_31" [label="(1, 512)", style=solid]; +"206 dequantize_per_tensor_default_31" -> "211 linear" [label="(1, 512)", style=solid]; +"207 linear_scale_0" -> "210 dequantize_per_channel_default_20" [label="(1000,)", style=solid]; +"208 linear_zero_point_0" -> "210 dequantize_per_channel_default_20" [label="(1000,)", style=solid]; +"209 _frozen_param20" -> "210 dequantize_per_channel_default_20" [label="(1000, 512)", style=solid]; +"210 dequantize_per_channel_default_20" -> "211 linear" [label="(1000, 512)", style=solid]; +"211 linear" -> "212 output" [label="(1, 1000)", style=solid]; +} diff --git a/tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/swin_v2_s.dot b/tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/swin_v2_s.dot new file mode 100644 index 00000000000..15bede0a9ae --- /dev/null +++ b/tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/swin_v2_s.dot @@ -0,0 +1,5734 @@ +strict digraph { +"0 features_1_0_attn_qkv_bias" [id=0, type=get_attr]; +"1 features_1_0_attn_proj_bias" [id=1, type=get_attr]; +"2 features_1_0_attn_logit_scale" [id=2, type=get_attr]; +"3 features_1_1_attn_qkv_bias" [id=3, type=get_attr]; +"4 features_1_1_attn_proj_bias" [id=4, type=get_attr]; +"5 features_1_1_attn_logit_scale" [id=5, type=get_attr]; +"6 features_3_0_attn_qkv_bias" [id=6, type=get_attr]; +"7 features_3_0_attn_proj_bias" [id=7, type=get_attr]; +"8 features_3_0_attn_logit_scale" [id=8, type=get_attr]; +"9 features_3_1_attn_qkv_bias" [id=9, type=get_attr]; +"10 features_3_1_attn_proj_bias" [id=10, type=get_attr]; +"11 features_3_1_attn_logit_scale" [id=11, type=get_attr]; +"12 features_5_0_attn_qkv_bias" [id=12, type=get_attr]; +"13 features_5_0_attn_proj_bias" [id=13, type=get_attr]; +"14 features_5_0_attn_logit_scale" [id=14, type=get_attr]; +"15 features_5_1_attn_qkv_bias" [id=15, type=get_attr]; +"16 features_5_1_attn_proj_bias" [id=16, type=get_attr]; +"17 features_5_1_attn_logit_scale" [id=17, type=get_attr]; +"18 features_5_2_attn_qkv_bias" [id=18, type=get_attr]; +"19 features_5_2_attn_proj_bias" [id=19, type=get_attr]; +"20 features_5_2_attn_logit_scale" [id=20, type=get_attr]; +"21 features_5_3_attn_qkv_bias" [id=21, type=get_attr]; +"22 features_5_3_attn_proj_bias" [id=22, type=get_attr]; +"23 features_5_3_attn_logit_scale" [id=23, type=get_attr]; +"24 features_5_4_attn_qkv_bias" [id=24, type=get_attr]; +"25 features_5_4_attn_proj_bias" [id=25, type=get_attr]; +"26 features_5_4_attn_logit_scale" [id=26, type=get_attr]; +"27 features_5_5_attn_qkv_bias" [id=27, type=get_attr]; +"28 features_5_5_attn_proj_bias" [id=28, type=get_attr]; +"29 features_5_5_attn_logit_scale" [id=29, type=get_attr]; +"30 features_5_6_attn_qkv_bias" [id=30, type=get_attr]; +"31 features_5_6_attn_proj_bias" [id=31, type=get_attr]; +"32 features_5_6_attn_logit_scale" [id=32, type=get_attr]; +"33 features_5_7_attn_qkv_bias" [id=33, type=get_attr]; +"34 features_5_7_attn_proj_bias" [id=34, type=get_attr]; +"35 features_5_7_attn_logit_scale" [id=35, type=get_attr]; +"36 features_5_8_attn_qkv_bias" [id=36, type=get_attr]; +"37 features_5_8_attn_proj_bias" [id=37, type=get_attr]; +"38 features_5_8_attn_logit_scale" [id=38, type=get_attr]; +"39 features_5_9_attn_qkv_bias" [id=39, type=get_attr]; +"40 features_5_9_attn_proj_bias" [id=40, type=get_attr]; +"41 features_5_9_attn_logit_scale" [id=41, type=get_attr]; +"42 features_5_10_attn_qkv_bias" [id=42, type=get_attr]; +"43 features_5_10_attn_proj_bias" [id=43, type=get_attr]; +"44 features_5_10_attn_logit_scale" [id=44, type=get_attr]; +"45 features_5_11_attn_qkv_bias" [id=45, type=get_attr]; +"46 features_5_11_attn_proj_bias" [id=46, type=get_attr]; +"47 features_5_11_attn_logit_scale" [id=47, type=get_attr]; +"48 features_5_12_attn_qkv_bias" [id=48, type=get_attr]; +"49 features_5_12_attn_proj_bias" [id=49, type=get_attr]; +"50 features_5_12_attn_logit_scale" [id=50, type=get_attr]; +"51 features_5_13_attn_qkv_bias" [id=51, type=get_attr]; +"52 features_5_13_attn_proj_bias" [id=52, type=get_attr]; +"53 features_5_13_attn_logit_scale" [id=53, type=get_attr]; +"54 features_5_14_attn_qkv_bias" [id=54, type=get_attr]; +"55 features_5_14_attn_proj_bias" [id=55, type=get_attr]; +"56 features_5_14_attn_logit_scale" [id=56, type=get_attr]; +"57 features_5_15_attn_qkv_bias" [id=57, type=get_attr]; +"58 features_5_15_attn_proj_bias" [id=58, type=get_attr]; +"59 features_5_15_attn_logit_scale" [id=59, type=get_attr]; +"60 features_5_16_attn_qkv_bias" [id=60, type=get_attr]; +"61 features_5_16_attn_proj_bias" [id=61, type=get_attr]; +"62 features_5_16_attn_logit_scale" [id=62, type=get_attr]; +"63 features_5_17_attn_qkv_bias" [id=63, type=get_attr]; +"64 features_5_17_attn_proj_bias" [id=64, type=get_attr]; +"65 features_5_17_attn_logit_scale" [id=65, type=get_attr]; +"66 features_7_0_attn_qkv_bias" [id=66, type=get_attr]; +"67 features_7_0_attn_proj_bias" [id=67, type=get_attr]; +"68 features_7_0_attn_logit_scale" [id=68, type=get_attr]; +"69 features_7_1_attn_qkv_bias" [id=69, type=get_attr]; +"70 features_7_1_attn_proj_bias" [id=70, type=get_attr]; +"71 features_7_1_attn_logit_scale" [id=71, type=get_attr]; +"72 features_0_0_bias" [id=72, type=get_attr]; +"73 features_0_2_weight" [id=73, type=get_attr]; +"74 features_0_2_bias" [id=74, type=get_attr]; +"75 features_1_0_attn_cpb_mlp_0_bias" [id=75, type=get_attr]; +"76 features_1_0_norm1_weight" [id=76, type=get_attr]; +"77 features_1_0_norm1_bias" [id=77, type=get_attr]; +"78 features_1_0_mlp_0_bias" [id=78, type=get_attr]; +"79 features_1_0_mlp_3_bias" [id=79, type=get_attr]; +"80 features_1_0_norm2_weight" [id=80, type=get_attr]; +"81 features_1_0_norm2_bias" [id=81, type=get_attr]; +"82 features_1_1_attn_cpb_mlp_0_bias" [id=82, type=get_attr]; +"83 features_1_1_norm1_weight" [id=83, type=get_attr]; +"84 features_1_1_norm1_bias" [id=84, type=get_attr]; +"85 features_1_1_mlp_0_bias" [id=85, type=get_attr]; +"86 features_1_1_mlp_3_bias" [id=86, type=get_attr]; +"87 features_1_1_norm2_weight" [id=87, type=get_attr]; +"88 features_1_1_norm2_bias" [id=88, type=get_attr]; +"89 features_2_norm_weight" [id=89, type=get_attr]; +"90 features_2_norm_bias" [id=90, type=get_attr]; +"91 features_3_0_attn_cpb_mlp_0_bias" [id=91, type=get_attr]; +"92 features_3_0_norm1_weight" [id=92, type=get_attr]; +"93 features_3_0_norm1_bias" [id=93, type=get_attr]; +"94 features_3_0_mlp_0_bias" [id=94, type=get_attr]; +"95 features_3_0_mlp_3_bias" [id=95, type=get_attr]; +"96 features_3_0_norm2_weight" [id=96, type=get_attr]; +"97 features_3_0_norm2_bias" [id=97, type=get_attr]; +"98 features_3_1_attn_cpb_mlp_0_bias" [id=98, type=get_attr]; +"99 features_3_1_norm1_weight" [id=99, type=get_attr]; +"100 features_3_1_norm1_bias" [id=100, type=get_attr]; +"101 features_3_1_mlp_0_bias" [id=101, type=get_attr]; +"102 features_3_1_mlp_3_bias" [id=102, type=get_attr]; +"103 features_3_1_norm2_weight" [id=103, type=get_attr]; +"104 features_3_1_norm2_bias" [id=104, type=get_attr]; +"105 features_4_norm_weight" [id=105, type=get_attr]; +"106 features_4_norm_bias" [id=106, type=get_attr]; +"107 features_5_0_attn_cpb_mlp_0_bias" [id=107, type=get_attr]; +"108 features_5_0_norm1_weight" [id=108, type=get_attr]; +"109 features_5_0_norm1_bias" [id=109, type=get_attr]; +"110 features_5_0_mlp_0_bias" [id=110, type=get_attr]; +"111 features_5_0_mlp_3_bias" [id=111, type=get_attr]; +"112 features_5_0_norm2_weight" [id=112, type=get_attr]; +"113 features_5_0_norm2_bias" [id=113, type=get_attr]; +"114 features_5_1_attn_cpb_mlp_0_bias" [id=114, type=get_attr]; +"115 features_5_1_norm1_weight" [id=115, type=get_attr]; +"116 features_5_1_norm1_bias" [id=116, type=get_attr]; +"117 features_5_1_mlp_0_bias" [id=117, type=get_attr]; +"118 features_5_1_mlp_3_bias" [id=118, type=get_attr]; +"119 features_5_1_norm2_weight" [id=119, type=get_attr]; +"120 features_5_1_norm2_bias" [id=120, type=get_attr]; +"121 features_5_2_attn_cpb_mlp_0_bias" [id=121, type=get_attr]; +"122 features_5_2_norm1_weight" [id=122, type=get_attr]; +"123 features_5_2_norm1_bias" [id=123, type=get_attr]; +"124 features_5_2_mlp_0_bias" [id=124, type=get_attr]; +"125 features_5_2_mlp_3_bias" [id=125, type=get_attr]; +"126 features_5_2_norm2_weight" [id=126, type=get_attr]; +"127 features_5_2_norm2_bias" [id=127, type=get_attr]; +"128 features_5_3_attn_cpb_mlp_0_bias" [id=128, type=get_attr]; +"129 features_5_3_norm1_weight" [id=129, type=get_attr]; +"130 features_5_3_norm1_bias" [id=130, type=get_attr]; +"131 features_5_3_mlp_0_bias" [id=131, type=get_attr]; +"132 features_5_3_mlp_3_bias" [id=132, type=get_attr]; +"133 features_5_3_norm2_weight" [id=133, type=get_attr]; +"134 features_5_3_norm2_bias" [id=134, type=get_attr]; +"135 features_5_4_attn_cpb_mlp_0_bias" [id=135, type=get_attr]; +"136 features_5_4_norm1_weight" [id=136, type=get_attr]; +"137 features_5_4_norm1_bias" [id=137, type=get_attr]; +"138 features_5_4_mlp_0_bias" [id=138, type=get_attr]; +"139 features_5_4_mlp_3_bias" [id=139, type=get_attr]; +"140 features_5_4_norm2_weight" [id=140, type=get_attr]; +"141 features_5_4_norm2_bias" [id=141, type=get_attr]; +"142 features_5_5_attn_cpb_mlp_0_bias" [id=142, type=get_attr]; +"143 features_5_5_norm1_weight" [id=143, type=get_attr]; +"144 features_5_5_norm1_bias" [id=144, type=get_attr]; +"145 features_5_5_mlp_0_bias" [id=145, type=get_attr]; +"146 features_5_5_mlp_3_bias" [id=146, type=get_attr]; +"147 features_5_5_norm2_weight" [id=147, type=get_attr]; +"148 features_5_5_norm2_bias" [id=148, type=get_attr]; +"149 features_5_6_attn_cpb_mlp_0_bias" [id=149, type=get_attr]; +"150 features_5_6_norm1_weight" [id=150, type=get_attr]; +"151 features_5_6_norm1_bias" [id=151, type=get_attr]; +"152 features_5_6_mlp_0_bias" [id=152, type=get_attr]; +"153 features_5_6_mlp_3_bias" [id=153, type=get_attr]; +"154 features_5_6_norm2_weight" [id=154, type=get_attr]; +"155 features_5_6_norm2_bias" [id=155, type=get_attr]; +"156 features_5_7_attn_cpb_mlp_0_bias" [id=156, type=get_attr]; +"157 features_5_7_norm1_weight" [id=157, type=get_attr]; +"158 features_5_7_norm1_bias" [id=158, type=get_attr]; +"159 features_5_7_mlp_0_bias" [id=159, type=get_attr]; +"160 features_5_7_mlp_3_bias" [id=160, type=get_attr]; +"161 features_5_7_norm2_weight" [id=161, type=get_attr]; +"162 features_5_7_norm2_bias" [id=162, type=get_attr]; +"163 features_5_8_attn_cpb_mlp_0_bias" [id=163, type=get_attr]; +"164 features_5_8_norm1_weight" [id=164, type=get_attr]; +"165 features_5_8_norm1_bias" [id=165, type=get_attr]; +"166 features_5_8_mlp_0_bias" [id=166, type=get_attr]; +"167 features_5_8_mlp_3_bias" [id=167, type=get_attr]; +"168 features_5_8_norm2_weight" [id=168, type=get_attr]; +"169 features_5_8_norm2_bias" [id=169, type=get_attr]; +"170 features_5_9_attn_cpb_mlp_0_bias" [id=170, type=get_attr]; +"171 features_5_9_norm1_weight" [id=171, type=get_attr]; +"172 features_5_9_norm1_bias" [id=172, type=get_attr]; +"173 features_5_9_mlp_0_bias" [id=173, type=get_attr]; +"174 features_5_9_mlp_3_bias" [id=174, type=get_attr]; +"175 features_5_9_norm2_weight" [id=175, type=get_attr]; +"176 features_5_9_norm2_bias" [id=176, type=get_attr]; +"177 features_5_10_attn_cpb_mlp_0_bias" [id=177, type=get_attr]; +"178 features_5_10_norm1_weight" [id=178, type=get_attr]; +"179 features_5_10_norm1_bias" [id=179, type=get_attr]; +"180 features_5_10_mlp_0_bias" [id=180, type=get_attr]; +"181 features_5_10_mlp_3_bias" [id=181, type=get_attr]; +"182 features_5_10_norm2_weight" [id=182, type=get_attr]; +"183 features_5_10_norm2_bias" [id=183, type=get_attr]; +"184 features_5_11_attn_cpb_mlp_0_bias" [id=184, type=get_attr]; +"185 features_5_11_norm1_weight" [id=185, type=get_attr]; +"186 features_5_11_norm1_bias" [id=186, type=get_attr]; +"187 features_5_11_mlp_0_bias" [id=187, type=get_attr]; +"188 features_5_11_mlp_3_bias" [id=188, type=get_attr]; +"189 features_5_11_norm2_weight" [id=189, type=get_attr]; +"190 features_5_11_norm2_bias" [id=190, type=get_attr]; +"191 features_5_12_attn_cpb_mlp_0_bias" [id=191, type=get_attr]; +"192 features_5_12_norm1_weight" [id=192, type=get_attr]; +"193 features_5_12_norm1_bias" [id=193, type=get_attr]; +"194 features_5_12_mlp_0_bias" [id=194, type=get_attr]; +"195 features_5_12_mlp_3_bias" [id=195, type=get_attr]; +"196 features_5_12_norm2_weight" [id=196, type=get_attr]; +"197 features_5_12_norm2_bias" [id=197, type=get_attr]; +"198 features_5_13_attn_cpb_mlp_0_bias" [id=198, type=get_attr]; +"199 features_5_13_norm1_weight" [id=199, type=get_attr]; +"200 features_5_13_norm1_bias" [id=200, type=get_attr]; +"201 features_5_13_mlp_0_bias" [id=201, type=get_attr]; +"202 features_5_13_mlp_3_bias" [id=202, type=get_attr]; +"203 features_5_13_norm2_weight" [id=203, type=get_attr]; +"204 features_5_13_norm2_bias" [id=204, type=get_attr]; +"205 features_5_14_attn_cpb_mlp_0_bias" [id=205, type=get_attr]; +"206 features_5_14_norm1_weight" [id=206, type=get_attr]; +"207 features_5_14_norm1_bias" [id=207, type=get_attr]; +"208 features_5_14_mlp_0_bias" [id=208, type=get_attr]; +"209 features_5_14_mlp_3_bias" [id=209, type=get_attr]; +"210 features_5_14_norm2_weight" [id=210, type=get_attr]; +"211 features_5_14_norm2_bias" [id=211, type=get_attr]; +"212 features_5_15_attn_cpb_mlp_0_bias" [id=212, type=get_attr]; +"213 features_5_15_norm1_weight" [id=213, type=get_attr]; +"214 features_5_15_norm1_bias" [id=214, type=get_attr]; +"215 features_5_15_mlp_0_bias" [id=215, type=get_attr]; +"216 features_5_15_mlp_3_bias" [id=216, type=get_attr]; +"217 features_5_15_norm2_weight" [id=217, type=get_attr]; +"218 features_5_15_norm2_bias" [id=218, type=get_attr]; +"219 features_5_16_attn_cpb_mlp_0_bias" [id=219, type=get_attr]; +"220 features_5_16_norm1_weight" [id=220, type=get_attr]; +"221 features_5_16_norm1_bias" [id=221, type=get_attr]; +"222 features_5_16_mlp_0_bias" [id=222, type=get_attr]; +"223 features_5_16_mlp_3_bias" [id=223, type=get_attr]; +"224 features_5_16_norm2_weight" [id=224, type=get_attr]; +"225 features_5_16_norm2_bias" [id=225, type=get_attr]; +"226 features_5_17_attn_cpb_mlp_0_bias" [id=226, type=get_attr]; +"227 features_5_17_norm1_weight" [id=227, type=get_attr]; +"228 features_5_17_norm1_bias" [id=228, type=get_attr]; +"229 features_5_17_mlp_0_bias" [id=229, type=get_attr]; +"230 features_5_17_mlp_3_bias" [id=230, type=get_attr]; +"231 features_5_17_norm2_weight" [id=231, type=get_attr]; +"232 features_5_17_norm2_bias" [id=232, type=get_attr]; +"233 features_6_norm_weight" [id=233, type=get_attr]; +"234 features_6_norm_bias" [id=234, type=get_attr]; +"235 features_7_0_attn_cpb_mlp_0_bias" [id=235, type=get_attr]; +"236 features_7_0_norm1_weight" [id=236, type=get_attr]; +"237 features_7_0_norm1_bias" [id=237, type=get_attr]; +"238 features_7_0_mlp_0_bias" [id=238, type=get_attr]; +"239 features_7_0_mlp_3_bias" [id=239, type=get_attr]; +"240 features_7_0_norm2_weight" [id=240, type=get_attr]; +"241 features_7_0_norm2_bias" [id=241, type=get_attr]; +"242 features_7_1_attn_cpb_mlp_0_bias" [id=242, type=get_attr]; +"243 features_7_1_norm1_weight" [id=243, type=get_attr]; +"244 features_7_1_norm1_bias" [id=244, type=get_attr]; +"245 features_7_1_mlp_0_bias" [id=245, type=get_attr]; +"246 features_7_1_mlp_3_bias" [id=246, type=get_attr]; +"247 features_7_1_norm2_weight" [id=247, type=get_attr]; +"248 features_7_1_norm2_bias" [id=248, type=get_attr]; +"249 norm_weight" [id=249, type=get_attr]; +"250 norm_bias" [id=250, type=get_attr]; +"251 head_bias" [id=251, type=get_attr]; +"252 features_1_0_attn_relative_position_index" [id=252, type=get_attr]; +"253 features_1_1_attn_relative_position_index" [id=253, type=get_attr]; +"254 features_3_0_attn_relative_position_index" [id=254, type=get_attr]; +"255 features_3_1_attn_relative_position_index" [id=255, type=get_attr]; +"256 features_5_0_attn_relative_position_index" [id=256, type=get_attr]; +"257 features_5_1_attn_relative_position_index" [id=257, type=get_attr]; +"258 features_5_2_attn_relative_position_index" [id=258, type=get_attr]; +"259 features_5_3_attn_relative_position_index" [id=259, type=get_attr]; +"260 features_5_4_attn_relative_position_index" [id=260, type=get_attr]; +"261 features_5_5_attn_relative_position_index" [id=261, type=get_attr]; +"262 features_5_6_attn_relative_position_index" [id=262, type=get_attr]; +"263 features_5_7_attn_relative_position_index" [id=263, type=get_attr]; +"264 features_5_8_attn_relative_position_index" [id=264, type=get_attr]; +"265 features_5_9_attn_relative_position_index" [id=265, type=get_attr]; +"266 features_5_10_attn_relative_position_index" [id=266, type=get_attr]; +"267 features_5_11_attn_relative_position_index" [id=267, type=get_attr]; +"268 features_5_12_attn_relative_position_index" [id=268, type=get_attr]; +"269 features_5_13_attn_relative_position_index" [id=269, type=get_attr]; +"270 features_5_14_attn_relative_position_index" [id=270, type=get_attr]; +"271 features_5_15_attn_relative_position_index" [id=271, type=get_attr]; +"272 features_5_16_attn_relative_position_index" [id=272, type=get_attr]; +"273 features_5_17_attn_relative_position_index" [id=273, type=get_attr]; +"274 features_7_0_attn_relative_position_index" [id=274, type=get_attr]; +"275 features_7_1_attn_relative_position_index" [id=275, type=get_attr]; +"276 x" [id=276, type=input]; +"277 quantize_per_tensor_default" [id=277, type=quantize_per_tensor]; +"278 dequantize_per_tensor_default" [id=278, type=dequantize_per_tensor]; +"279 conv2d_scale_0" [id=279, type=get_attr]; +"280 conv2d_zero_point_0" [id=280, type=get_attr]; +"281 _frozen_param0" [id=281, type=get_attr]; +"282 dequantize_per_channel_default" [id=282, type=dequantize_per_channel]; +"283 conv2d" [id=283, type=conv2d]; +"284 permute" [id=284, type=permute]; +"285 layer_norm" [id=285, type=layer_norm]; +"286 _frozen_param1" [id=286, type=get_attr]; +"287 dequantize_per_tensor_default_1" [id=287, type=dequantize_per_tensor]; +"288 linear_scale_0" [id=288, type=get_attr]; +"289 linear_zero_point_0" [id=289, type=get_attr]; +"290 _frozen_param2" [id=290, type=get_attr]; +"291 dequantize_per_channel_default_1" [id=291, type=dequantize_per_channel]; +"292 linear" [id=292, type=linear]; +"293 relu_" [id=293, type=relu_]; +"294 relu__0_0_nncf_smooth_quant_0" [id=294, type=call_module]; +"295 quantize_per_tensor_default_2" [id=295, type=quantize_per_tensor]; +"296 dequantize_per_tensor_default_2" [id=296, type=dequantize_per_tensor]; +"297 linear_1_scale_0" [id=297, type=get_attr]; +"298 linear_1_zero_point_0" [id=298, type=get_attr]; +"299 _frozen_param3" [id=299, type=get_attr]; +"300 dequantize_per_channel_default_2" [id=300, type=dequantize_per_channel]; +"301 linear_1" [id=301, type=linear]; +"302 view" [id=302, type=view]; +"303 index" [id=303, type=index]; +"304 view_1" [id=304, type=view]; +"305 permute_1" [id=305, type=permute]; +"306 contiguous" [id=306, type=contiguous]; +"307 unsqueeze" [id=307, type=unsqueeze]; +"308 sigmoid" [id=308, type=sigmoid]; +"309 mul" [id=309, type=mul]; +"310 pad" [id=310, type=pad]; +"311 view_2" [id=311, type=view]; +"312 permute_2" [id=312, type=permute]; +"313 reshape" [id=313, type=reshape]; +"314 clone" [id=314, type=clone]; +"315 features_1_0_attn_qkv_weight_updated_constant0" [id=315, type=get_attr]; +"316 reshape_0_0_nncf_smooth_quant_0" [id=316, type=call_module]; +"317 linear_2" [id=317, type=linear]; +"318 reshape_1" [id=318, type=reshape]; +"319 permute_3" [id=319, type=permute]; +"320 select" [id=320, type=select]; +"321 select_1" [id=321, type=select]; +"322 select_2" [id=322, type=select]; +"323 linalg_vector_norm" [id=323, type=linalg_vector_norm]; +"324 clamp_min" [id=324, type=clamp_min]; +"325 expand_as" [id=325, type=expand_as]; +"326 div" [id=326, type=div]; +"327 linalg_vector_norm_1" [id=327, type=linalg_vector_norm]; +"328 clamp_min_1" [id=328, type=clamp_min]; +"329 expand_as_1" [id=329, type=expand_as]; +"330 div_1" [id=330, type=div]; +"331 transpose" [id=331, type=transpose]; +"332 matmul" [id=332, type=matmul]; +"333 clamp" [id=333, type=clamp]; +"334 exp" [id=334, type=exp]; +"335 mul_1" [id=335, type=mul]; +"336 add" [id=336, type=add]; +"337 softmax" [id=337, type=softmax]; +"338 dropout" [id=338, type=dropout]; +"339 matmul_1" [id=339, type=matmul]; +"340 transpose_1" [id=340, type=transpose]; +"341 reshape_2" [id=341, type=reshape]; +"342 features_1_0_attn_proj_weight_updated_constant0" [id=342, type=get_attr]; +"343 reshape_2_0_0_nncf_smooth_quant_0" [id=343, type=call_module]; +"344 linear_3" [id=344, type=linear]; +"345 dropout_1" [id=345, type=dropout]; +"346 view_3" [id=346, type=view]; +"347 permute_4" [id=347, type=permute]; +"348 reshape_3" [id=348, type=reshape]; +"349 slice_2" [id=349, type=slice]; +"350 slice_3" [id=350, type=slice]; +"351 layer_norm_1" [id=351, type=layer_norm]; +"352 add_1" [id=352, type=add]; +"353 add_1_0_0_nncf_smooth_quant_0" [id=353, type=call_module]; +"354 quantize_per_tensor_default_3" [id=354, type=quantize_per_tensor]; +"355 dequantize_per_tensor_default_3" [id=355, type=dequantize_per_tensor]; +"356 linear_4_scale_0" [id=356, type=get_attr]; +"357 linear_4_zero_point_0" [id=357, type=get_attr]; +"358 _frozen_param4" [id=358, type=get_attr]; +"359 dequantize_per_channel_default_3" [id=359, type=dequantize_per_channel]; +"360 linear_4" [id=360, type=linear]; +"361 gelu" [id=361, type=gelu]; +"362 dropout_2" [id=362, type=dropout]; +"363 dropout_2_0_0_nncf_smooth_quant_0" [id=363, type=call_module]; +"364 quantize_per_tensor_default_4" [id=364, type=quantize_per_tensor]; +"365 dequantize_per_tensor_default_4" [id=365, type=dequantize_per_tensor]; +"366 linear_5_scale_0" [id=366, type=get_attr]; +"367 linear_5_zero_point_0" [id=367, type=get_attr]; +"368 _frozen_param5" [id=368, type=get_attr]; +"369 dequantize_per_channel_default_4" [id=369, type=dequantize_per_channel]; +"370 linear_5" [id=370, type=linear]; +"371 dropout_3" [id=371, type=dropout]; +"372 layer_norm_2" [id=372, type=layer_norm]; +"373 add_2" [id=373, type=add]; +"374 _frozen_param6" [id=374, type=get_attr]; +"375 dequantize_per_tensor_default_5" [id=375, type=dequantize_per_tensor]; +"376 linear_6_scale_0" [id=376, type=get_attr]; +"377 linear_6_zero_point_0" [id=377, type=get_attr]; +"378 _frozen_param7" [id=378, type=get_attr]; +"379 dequantize_per_channel_default_5" [id=379, type=dequantize_per_channel]; +"380 linear_6" [id=380, type=linear]; +"381 relu__1" [id=381, type=relu_]; +"382 relu__1_0_0_nncf_smooth_quant_0" [id=382, type=call_module]; +"383 quantize_per_tensor_default_6" [id=383, type=quantize_per_tensor]; +"384 dequantize_per_tensor_default_6" [id=384, type=dequantize_per_tensor]; +"385 linear_7_scale_0" [id=385, type=get_attr]; +"386 linear_7_zero_point_0" [id=386, type=get_attr]; +"387 _frozen_param8" [id=387, type=get_attr]; +"388 dequantize_per_channel_default_6" [id=388, type=dequantize_per_channel]; +"389 linear_7" [id=389, type=linear]; +"390 view_4" [id=390, type=view]; +"391 index_1" [id=391, type=index]; +"392 view_5" [id=392, type=view]; +"393 permute_5" [id=393, type=permute]; +"394 contiguous_1" [id=394, type=contiguous]; +"395 unsqueeze_1" [id=395, type=unsqueeze]; +"396 sigmoid_1" [id=396, type=sigmoid]; +"397 mul_2" [id=397, type=mul]; +"398 pad_1" [id=398, type=pad]; +"399 roll" [id=399, type=roll]; +"400 view_6" [id=400, type=view]; +"401 permute_6" [id=401, type=permute]; +"402 reshape_4" [id=402, type=reshape]; +"403 clone_1" [id=403, type=clone]; +"404 features_1_1_attn_qkv_weight_updated_constant0" [id=404, type=get_attr]; +"405 reshape_4_0_0_nncf_smooth_quant_0" [id=405, type=call_module]; +"406 linear_8" [id=406, type=linear]; +"407 reshape_5" [id=407, type=reshape]; +"408 permute_7" [id=408, type=permute]; +"409 select_3" [id=409, type=select]; +"410 select_4" [id=410, type=select]; +"411 select_5" [id=411, type=select]; +"412 linalg_vector_norm_2" [id=412, type=linalg_vector_norm]; +"413 clamp_min_2" [id=413, type=clamp_min]; +"414 expand_as_2" [id=414, type=expand_as]; +"415 div_2" [id=415, type=div]; +"416 linalg_vector_norm_3" [id=416, type=linalg_vector_norm]; +"417 clamp_min_3" [id=417, type=clamp_min]; +"418 expand_as_3" [id=418, type=expand_as]; +"419 div_3" [id=419, type=div]; +"420 transpose_2" [id=420, type=transpose]; +"421 matmul_2" [id=421, type=matmul]; +"422 clamp_1" [id=422, type=clamp]; +"423 exp_1" [id=423, type=exp]; +"424 mul_3" [id=424, type=mul]; +"425 add_3" [id=425, type=add]; +"426 new_zeros" [id=426, type=new_zeros]; +"427 view_7" [id=427, type=view]; +"428 permute_8" [id=428, type=permute]; +"429 reshape_6" [id=429, type=reshape]; +"430 unsqueeze_2" [id=430, type=unsqueeze]; +"431 unsqueeze_3" [id=431, type=unsqueeze]; +"432 sub" [id=432, type=sub]; +"433 ne" [id=433, type=ne]; +"434 masked_fill" [id=434, type=masked_fill]; +"435 eq" [id=435, type=eq]; +"436 masked_fill_1" [id=436, type=masked_fill]; +"437 view_8" [id=437, type=view]; +"438 unsqueeze_4" [id=438, type=unsqueeze]; +"439 unsqueeze_5" [id=439, type=unsqueeze]; +"440 add_4" [id=440, type=add]; +"441 view_9" [id=441, type=view]; +"442 softmax_1" [id=442, type=softmax]; +"443 dropout_4" [id=443, type=dropout]; +"444 matmul_3" [id=444, type=matmul]; +"445 transpose_3" [id=445, type=transpose]; +"446 reshape_7" [id=446, type=reshape]; +"447 features_1_1_attn_proj_weight_updated_constant0" [id=447, type=get_attr]; +"448 reshape_7_0_0_nncf_smooth_quant_0" [id=448, type=call_module]; +"449 linear_9" [id=449, type=linear]; +"450 dropout_5" [id=450, type=dropout]; +"451 view_10" [id=451, type=view]; +"452 permute_9" [id=452, type=permute]; +"453 reshape_8" [id=453, type=reshape]; +"454 roll_1" [id=454, type=roll]; +"455 slice_23" [id=455, type=slice]; +"456 slice_24" [id=456, type=slice]; +"457 layer_norm_3" [id=457, type=layer_norm]; +"458 add_5" [id=458, type=add]; +"459 add_5_0_0_nncf_smooth_quant_0" [id=459, type=call_module]; +"460 quantize_per_tensor_default_7" [id=460, type=quantize_per_tensor]; +"461 dequantize_per_tensor_default_7" [id=461, type=dequantize_per_tensor]; +"462 linear_10_scale_0" [id=462, type=get_attr]; +"463 linear_10_zero_point_0" [id=463, type=get_attr]; +"464 _frozen_param9" [id=464, type=get_attr]; +"465 dequantize_per_channel_default_7" [id=465, type=dequantize_per_channel]; +"466 linear_10" [id=466, type=linear]; +"467 gelu_1" [id=467, type=gelu]; +"468 dropout_6" [id=468, type=dropout]; +"469 dropout_6_0_0_nncf_smooth_quant_0" [id=469, type=call_module]; +"470 quantize_per_tensor_default_8" [id=470, type=quantize_per_tensor]; +"471 dequantize_per_tensor_default_8" [id=471, type=dequantize_per_tensor]; +"472 linear_11_scale_0" [id=472, type=get_attr]; +"473 linear_11_zero_point_0" [id=473, type=get_attr]; +"474 _frozen_param10" [id=474, type=get_attr]; +"475 dequantize_per_channel_default_8" [id=475, type=dequantize_per_channel]; +"476 linear_11" [id=476, type=linear]; +"477 dropout_7" [id=477, type=dropout]; +"478 layer_norm_4" [id=478, type=layer_norm]; +"479 add_6" [id=479, type=add]; +"480 pad_2" [id=480, type=pad]; +"481 slice_25" [id=481, type=slice]; +"482 slice_26" [id=482, type=slice]; +"483 slice_27" [id=483, type=slice]; +"484 slice_28" [id=484, type=slice]; +"485 slice_29" [id=485, type=slice]; +"486 slice_30" [id=486, type=slice]; +"487 slice_31" [id=487, type=slice]; +"488 slice_32" [id=488, type=slice]; +"489 slice_33" [id=489, type=slice]; +"490 slice_34" [id=490, type=slice]; +"491 slice_35" [id=491, type=slice]; +"492 slice_36" [id=492, type=slice]; +"493 cat" [id=493, type=cat]; +"494 cat_0_0_nncf_smooth_quant_0" [id=494, type=call_module]; +"495 quantize_per_tensor_default_9" [id=495, type=quantize_per_tensor]; +"496 dequantize_per_tensor_default_9" [id=496, type=dequantize_per_tensor]; +"497 linear_12_scale_0" [id=497, type=get_attr]; +"498 linear_12_zero_point_0" [id=498, type=get_attr]; +"499 _frozen_param11" [id=499, type=get_attr]; +"500 dequantize_per_channel_default_9" [id=500, type=dequantize_per_channel]; +"501 linear_12" [id=501, type=linear]; +"502 layer_norm_5" [id=502, type=layer_norm]; +"503 _frozen_param12" [id=503, type=get_attr]; +"504 dequantize_per_tensor_default_10" [id=504, type=dequantize_per_tensor]; +"505 linear_13_scale_0" [id=505, type=get_attr]; +"506 linear_13_zero_point_0" [id=506, type=get_attr]; +"507 _frozen_param13" [id=507, type=get_attr]; +"508 dequantize_per_channel_default_10" [id=508, type=dequantize_per_channel]; +"509 linear_13" [id=509, type=linear]; +"510 relu__2" [id=510, type=relu_]; +"511 relu__2_0_0_nncf_smooth_quant_0" [id=511, type=call_module]; +"512 quantize_per_tensor_default_11" [id=512, type=quantize_per_tensor]; +"513 dequantize_per_tensor_default_11" [id=513, type=dequantize_per_tensor]; +"514 linear_14_scale_0" [id=514, type=get_attr]; +"515 linear_14_zero_point_0" [id=515, type=get_attr]; +"516 _frozen_param14" [id=516, type=get_attr]; +"517 dequantize_per_channel_default_11" [id=517, type=dequantize_per_channel]; +"518 linear_14" [id=518, type=linear]; +"519 view_11" [id=519, type=view]; +"520 index_2" [id=520, type=index]; +"521 view_12" [id=521, type=view]; +"522 permute_10" [id=522, type=permute]; +"523 contiguous_2" [id=523, type=contiguous]; +"524 unsqueeze_6" [id=524, type=unsqueeze]; +"525 sigmoid_2" [id=525, type=sigmoid]; +"526 mul_4" [id=526, type=mul]; +"527 pad_3" [id=527, type=pad]; +"528 view_13" [id=528, type=view]; +"529 permute_11" [id=529, type=permute]; +"530 reshape_9" [id=530, type=reshape]; +"531 clone_2" [id=531, type=clone]; +"532 features_3_0_attn_qkv_weight_updated_constant0" [id=532, type=get_attr]; +"533 reshape_9_0_0_nncf_smooth_quant_0" [id=533, type=call_module]; +"534 linear_15" [id=534, type=linear]; +"535 reshape_10" [id=535, type=reshape]; +"536 permute_12" [id=536, type=permute]; +"537 select_6" [id=537, type=select]; +"538 select_7" [id=538, type=select]; +"539 select_8" [id=539, type=select]; +"540 linalg_vector_norm_4" [id=540, type=linalg_vector_norm]; +"541 clamp_min_4" [id=541, type=clamp_min]; +"542 expand_as_4" [id=542, type=expand_as]; +"543 div_4" [id=543, type=div]; +"544 linalg_vector_norm_5" [id=544, type=linalg_vector_norm]; +"545 clamp_min_5" [id=545, type=clamp_min]; +"546 expand_as_5" [id=546, type=expand_as]; +"547 div_5" [id=547, type=div]; +"548 transpose_4" [id=548, type=transpose]; +"549 matmul_4" [id=549, type=matmul]; +"550 clamp_2" [id=550, type=clamp]; +"551 exp_2" [id=551, type=exp]; +"552 mul_5" [id=552, type=mul]; +"553 add_7" [id=553, type=add]; +"554 softmax_2" [id=554, type=softmax]; +"555 dropout_8" [id=555, type=dropout]; +"556 matmul_5" [id=556, type=matmul]; +"557 transpose_5" [id=557, type=transpose]; +"558 reshape_11" [id=558, type=reshape]; +"559 features_3_0_attn_proj_weight_updated_constant0" [id=559, type=get_attr]; +"560 reshape_11_0_0_nncf_smooth_quant_0" [id=560, type=call_module]; +"561 linear_16" [id=561, type=linear]; +"562 dropout_9" [id=562, type=dropout]; +"563 view_14" [id=563, type=view]; +"564 permute_13" [id=564, type=permute]; +"565 reshape_12" [id=565, type=reshape]; +"566 slice_38" [id=566, type=slice]; +"567 slice_39" [id=567, type=slice]; +"568 slice_40" [id=568, type=slice]; +"569 slice_41" [id=569, type=slice]; +"570 contiguous_3" [id=570, type=contiguous]; +"571 layer_norm_6" [id=571, type=layer_norm]; +"572 add_8" [id=572, type=add]; +"573 add_8_0_0_nncf_smooth_quant_0" [id=573, type=call_module]; +"574 quantize_per_tensor_default_12" [id=574, type=quantize_per_tensor]; +"575 dequantize_per_tensor_default_12" [id=575, type=dequantize_per_tensor]; +"576 linear_17_scale_0" [id=576, type=get_attr]; +"577 linear_17_zero_point_0" [id=577, type=get_attr]; +"578 _frozen_param15" [id=578, type=get_attr]; +"579 dequantize_per_channel_default_12" [id=579, type=dequantize_per_channel]; +"580 linear_17" [id=580, type=linear]; +"581 gelu_2" [id=581, type=gelu]; +"582 dropout_10" [id=582, type=dropout]; +"583 dropout_10_0_0_nncf_smooth_quant_0" [id=583, type=call_module]; +"584 quantize_per_tensor_default_13" [id=584, type=quantize_per_tensor]; +"585 dequantize_per_tensor_default_13" [id=585, type=dequantize_per_tensor]; +"586 linear_18_scale_0" [id=586, type=get_attr]; +"587 linear_18_zero_point_0" [id=587, type=get_attr]; +"588 _frozen_param16" [id=588, type=get_attr]; +"589 dequantize_per_channel_default_13" [id=589, type=dequantize_per_channel]; +"590 linear_18" [id=590, type=linear]; +"591 dropout_11" [id=591, type=dropout]; +"592 layer_norm_7" [id=592, type=layer_norm]; +"593 add_9" [id=593, type=add]; +"594 _frozen_param17" [id=594, type=get_attr]; +"595 dequantize_per_tensor_default_14" [id=595, type=dequantize_per_tensor]; +"596 linear_19_scale_0" [id=596, type=get_attr]; +"597 linear_19_zero_point_0" [id=597, type=get_attr]; +"598 _frozen_param18" [id=598, type=get_attr]; +"599 dequantize_per_channel_default_14" [id=599, type=dequantize_per_channel]; +"600 linear_19" [id=600, type=linear]; +"601 relu__3" [id=601, type=relu_]; +"602 relu__3_0_0_nncf_smooth_quant_0" [id=602, type=call_module]; +"603 quantize_per_tensor_default_15" [id=603, type=quantize_per_tensor]; +"604 dequantize_per_tensor_default_15" [id=604, type=dequantize_per_tensor]; +"605 linear_20_scale_0" [id=605, type=get_attr]; +"606 linear_20_zero_point_0" [id=606, type=get_attr]; +"607 _frozen_param19" [id=607, type=get_attr]; +"608 dequantize_per_channel_default_15" [id=608, type=dequantize_per_channel]; +"609 linear_20" [id=609, type=linear]; +"610 view_15" [id=610, type=view]; +"611 index_3" [id=611, type=index]; +"612 view_16" [id=612, type=view]; +"613 permute_14" [id=613, type=permute]; +"614 contiguous_4" [id=614, type=contiguous]; +"615 unsqueeze_7" [id=615, type=unsqueeze]; +"616 sigmoid_3" [id=616, type=sigmoid]; +"617 mul_6" [id=617, type=mul]; +"618 pad_4" [id=618, type=pad]; +"619 roll_2" [id=619, type=roll]; +"620 view_17" [id=620, type=view]; +"621 permute_15" [id=621, type=permute]; +"622 reshape_13" [id=622, type=reshape]; +"623 clone_3" [id=623, type=clone]; +"624 features_3_1_attn_qkv_weight_updated_constant0" [id=624, type=get_attr]; +"625 reshape_13_0_0_nncf_smooth_quant_0" [id=625, type=call_module]; +"626 linear_21" [id=626, type=linear]; +"627 reshape_14" [id=627, type=reshape]; +"628 permute_16" [id=628, type=permute]; +"629 select_9" [id=629, type=select]; +"630 select_10" [id=630, type=select]; +"631 select_11" [id=631, type=select]; +"632 linalg_vector_norm_6" [id=632, type=linalg_vector_norm]; +"633 clamp_min_6" [id=633, type=clamp_min]; +"634 expand_as_6" [id=634, type=expand_as]; +"635 div_6" [id=635, type=div]; +"636 linalg_vector_norm_7" [id=636, type=linalg_vector_norm]; +"637 clamp_min_7" [id=637, type=clamp_min]; +"638 expand_as_7" [id=638, type=expand_as]; +"639 div_7" [id=639, type=div]; +"640 transpose_6" [id=640, type=transpose]; +"641 matmul_6" [id=641, type=matmul]; +"642 clamp_3" [id=642, type=clamp]; +"643 exp_3" [id=643, type=exp]; +"644 mul_7" [id=644, type=mul]; +"645 add_10" [id=645, type=add]; +"646 new_zeros_1" [id=646, type=new_zeros]; +"647 view_18" [id=647, type=view]; +"648 permute_17" [id=648, type=permute]; +"649 reshape_15" [id=649, type=reshape]; +"650 unsqueeze_8" [id=650, type=unsqueeze]; +"651 unsqueeze_9" [id=651, type=unsqueeze]; +"652 sub_1" [id=652, type=sub]; +"653 ne_1" [id=653, type=ne]; +"654 masked_fill_2" [id=654, type=masked_fill]; +"655 eq_1" [id=655, type=eq]; +"656 masked_fill_3" [id=656, type=masked_fill]; +"657 view_19" [id=657, type=view]; +"658 unsqueeze_10" [id=658, type=unsqueeze]; +"659 unsqueeze_11" [id=659, type=unsqueeze]; +"660 add_11" [id=660, type=add]; +"661 view_20" [id=661, type=view]; +"662 softmax_3" [id=662, type=softmax]; +"663 dropout_12" [id=663, type=dropout]; +"664 matmul_7" [id=664, type=matmul]; +"665 transpose_7" [id=665, type=transpose]; +"666 reshape_16" [id=666, type=reshape]; +"667 features_3_1_attn_proj_weight_updated_constant0" [id=667, type=get_attr]; +"668 reshape_16_0_0_nncf_smooth_quant_0" [id=668, type=call_module]; +"669 linear_22" [id=669, type=linear]; +"670 dropout_13" [id=670, type=dropout]; +"671 view_21" [id=671, type=view]; +"672 permute_18" [id=672, type=permute]; +"673 reshape_17" [id=673, type=reshape]; +"674 roll_3" [id=674, type=roll]; +"675 slice_61" [id=675, type=slice]; +"676 slice_62" [id=676, type=slice]; +"677 slice_63" [id=677, type=slice]; +"678 slice_64" [id=678, type=slice]; +"679 contiguous_5" [id=679, type=contiguous]; +"680 layer_norm_8" [id=680, type=layer_norm]; +"681 add_12" [id=681, type=add]; +"682 add_12_0_0_nncf_smooth_quant_0" [id=682, type=call_module]; +"683 quantize_per_tensor_default_16" [id=683, type=quantize_per_tensor]; +"684 dequantize_per_tensor_default_16" [id=684, type=dequantize_per_tensor]; +"685 linear_23_scale_0" [id=685, type=get_attr]; +"686 linear_23_zero_point_0" [id=686, type=get_attr]; +"687 _frozen_param20" [id=687, type=get_attr]; +"688 dequantize_per_channel_default_16" [id=688, type=dequantize_per_channel]; +"689 linear_23" [id=689, type=linear]; +"690 gelu_3" [id=690, type=gelu]; +"691 dropout_14" [id=691, type=dropout]; +"692 dropout_14_0_0_nncf_smooth_quant_0" [id=692, type=call_module]; +"693 quantize_per_tensor_default_17" [id=693, type=quantize_per_tensor]; +"694 dequantize_per_tensor_default_17" [id=694, type=dequantize_per_tensor]; +"695 linear_24_scale_0" [id=695, type=get_attr]; +"696 linear_24_zero_point_0" [id=696, type=get_attr]; +"697 _frozen_param21" [id=697, type=get_attr]; +"698 dequantize_per_channel_default_17" [id=698, type=dequantize_per_channel]; +"699 linear_24" [id=699, type=linear]; +"700 dropout_15" [id=700, type=dropout]; +"701 layer_norm_9" [id=701, type=layer_norm]; +"702 add_13" [id=702, type=add]; +"703 pad_5" [id=703, type=pad]; +"704 slice_65" [id=704, type=slice]; +"705 slice_66" [id=705, type=slice]; +"706 slice_67" [id=706, type=slice]; +"707 slice_68" [id=707, type=slice]; +"708 slice_69" [id=708, type=slice]; +"709 slice_70" [id=709, type=slice]; +"710 slice_71" [id=710, type=slice]; +"711 slice_72" [id=711, type=slice]; +"712 slice_73" [id=712, type=slice]; +"713 slice_74" [id=713, type=slice]; +"714 slice_75" [id=714, type=slice]; +"715 slice_76" [id=715, type=slice]; +"716 cat_1" [id=716, type=cat]; +"717 cat_1_0_0_nncf_smooth_quant_0" [id=717, type=call_module]; +"718 quantize_per_tensor_default_18" [id=718, type=quantize_per_tensor]; +"719 dequantize_per_tensor_default_18" [id=719, type=dequantize_per_tensor]; +"720 linear_25_scale_0" [id=720, type=get_attr]; +"721 linear_25_zero_point_0" [id=721, type=get_attr]; +"722 _frozen_param22" [id=722, type=get_attr]; +"723 dequantize_per_channel_default_18" [id=723, type=dequantize_per_channel]; +"724 linear_25" [id=724, type=linear]; +"725 layer_norm_10" [id=725, type=layer_norm]; +"726 _frozen_param23" [id=726, type=get_attr]; +"727 dequantize_per_tensor_default_19" [id=727, type=dequantize_per_tensor]; +"728 linear_26_scale_0" [id=728, type=get_attr]; +"729 linear_26_zero_point_0" [id=729, type=get_attr]; +"730 _frozen_param24" [id=730, type=get_attr]; +"731 dequantize_per_channel_default_19" [id=731, type=dequantize_per_channel]; +"732 linear_26" [id=732, type=linear]; +"733 relu__4" [id=733, type=relu_]; +"734 relu__4_0_0_nncf_smooth_quant_0" [id=734, type=call_module]; +"735 quantize_per_tensor_default_20" [id=735, type=quantize_per_tensor]; +"736 dequantize_per_tensor_default_20" [id=736, type=dequantize_per_tensor]; +"737 linear_27_scale_0" [id=737, type=get_attr]; +"738 linear_27_zero_point_0" [id=738, type=get_attr]; +"739 _frozen_param25" [id=739, type=get_attr]; +"740 dequantize_per_channel_default_20" [id=740, type=dequantize_per_channel]; +"741 linear_27" [id=741, type=linear]; +"742 view_22" [id=742, type=view]; +"743 index_4" [id=743, type=index]; +"744 view_23" [id=744, type=view]; +"745 permute_19" [id=745, type=permute]; +"746 contiguous_6" [id=746, type=contiguous]; +"747 unsqueeze_12" [id=747, type=unsqueeze]; +"748 sigmoid_4" [id=748, type=sigmoid]; +"749 mul_8" [id=749, type=mul]; +"750 pad_6" [id=750, type=pad]; +"751 view_24" [id=751, type=view]; +"752 permute_20" [id=752, type=permute]; +"753 reshape_18" [id=753, type=reshape]; +"754 clone_4" [id=754, type=clone]; +"755 features_5_0_attn_qkv_weight_updated_constant0" [id=755, type=get_attr]; +"756 reshape_18_0_0_nncf_smooth_quant_0" [id=756, type=call_module]; +"757 linear_28" [id=757, type=linear]; +"758 reshape_19" [id=758, type=reshape]; +"759 permute_21" [id=759, type=permute]; +"760 select_12" [id=760, type=select]; +"761 select_13" [id=761, type=select]; +"762 select_14" [id=762, type=select]; +"763 linalg_vector_norm_8" [id=763, type=linalg_vector_norm]; +"764 clamp_min_8" [id=764, type=clamp_min]; +"765 expand_as_8" [id=765, type=expand_as]; +"766 div_8" [id=766, type=div]; +"767 linalg_vector_norm_9" [id=767, type=linalg_vector_norm]; +"768 clamp_min_9" [id=768, type=clamp_min]; +"769 expand_as_9" [id=769, type=expand_as]; +"770 div_9" [id=770, type=div]; +"771 transpose_8" [id=771, type=transpose]; +"772 matmul_8" [id=772, type=matmul]; +"773 clamp_4" [id=773, type=clamp]; +"774 exp_4" [id=774, type=exp]; +"775 mul_9" [id=775, type=mul]; +"776 add_14" [id=776, type=add]; +"777 softmax_4" [id=777, type=softmax]; +"778 dropout_16" [id=778, type=dropout]; +"779 matmul_9" [id=779, type=matmul]; +"780 transpose_9" [id=780, type=transpose]; +"781 reshape_20" [id=781, type=reshape]; +"782 features_5_0_attn_proj_weight_updated_constant0" [id=782, type=get_attr]; +"783 reshape_20_0_0_nncf_smooth_quant_0" [id=783, type=call_module]; +"784 linear_29" [id=784, type=linear]; +"785 dropout_17" [id=785, type=dropout]; +"786 view_25" [id=786, type=view]; +"787 permute_22" [id=787, type=permute]; +"788 reshape_21" [id=788, type=reshape]; +"789 slice_78" [id=789, type=slice]; +"790 slice_79" [id=790, type=slice]; +"791 slice_80" [id=791, type=slice]; +"792 slice_81" [id=792, type=slice]; +"793 contiguous_7" [id=793, type=contiguous]; +"794 layer_norm_11" [id=794, type=layer_norm]; +"795 add_15" [id=795, type=add]; +"796 add_15_0_0_nncf_smooth_quant_0" [id=796, type=call_module]; +"797 quantize_per_tensor_default_21" [id=797, type=quantize_per_tensor]; +"798 dequantize_per_tensor_default_21" [id=798, type=dequantize_per_tensor]; +"799 linear_30_scale_0" [id=799, type=get_attr]; +"800 linear_30_zero_point_0" [id=800, type=get_attr]; +"801 _frozen_param26" [id=801, type=get_attr]; +"802 dequantize_per_channel_default_21" [id=802, type=dequantize_per_channel]; +"803 linear_30" [id=803, type=linear]; +"804 gelu_4" [id=804, type=gelu]; +"805 dropout_18" [id=805, type=dropout]; +"806 dropout_18_0_0_nncf_smooth_quant_0" [id=806, type=call_module]; +"807 quantize_per_tensor_default_22" [id=807, type=quantize_per_tensor]; +"808 dequantize_per_tensor_default_22" [id=808, type=dequantize_per_tensor]; +"809 linear_31_scale_0" [id=809, type=get_attr]; +"810 linear_31_zero_point_0" [id=810, type=get_attr]; +"811 _frozen_param27" [id=811, type=get_attr]; +"812 dequantize_per_channel_default_22" [id=812, type=dequantize_per_channel]; +"813 linear_31" [id=813, type=linear]; +"814 dropout_19" [id=814, type=dropout]; +"815 layer_norm_12" [id=815, type=layer_norm]; +"816 add_16" [id=816, type=add]; +"817 _frozen_param28" [id=817, type=get_attr]; +"818 dequantize_per_tensor_default_23" [id=818, type=dequantize_per_tensor]; +"819 linear_32_scale_0" [id=819, type=get_attr]; +"820 linear_32_zero_point_0" [id=820, type=get_attr]; +"821 _frozen_param29" [id=821, type=get_attr]; +"822 dequantize_per_channel_default_23" [id=822, type=dequantize_per_channel]; +"823 linear_32" [id=823, type=linear]; +"824 relu__5" [id=824, type=relu_]; +"825 relu__5_0_0_nncf_smooth_quant_0" [id=825, type=call_module]; +"826 quantize_per_tensor_default_24" [id=826, type=quantize_per_tensor]; +"827 dequantize_per_tensor_default_24" [id=827, type=dequantize_per_tensor]; +"828 linear_33_scale_0" [id=828, type=get_attr]; +"829 linear_33_zero_point_0" [id=829, type=get_attr]; +"830 _frozen_param30" [id=830, type=get_attr]; +"831 dequantize_per_channel_default_24" [id=831, type=dequantize_per_channel]; +"832 linear_33" [id=832, type=linear]; +"833 view_26" [id=833, type=view]; +"834 index_5" [id=834, type=index]; +"835 view_27" [id=835, type=view]; +"836 permute_23" [id=836, type=permute]; +"837 contiguous_8" [id=837, type=contiguous]; +"838 unsqueeze_13" [id=838, type=unsqueeze]; +"839 sigmoid_5" [id=839, type=sigmoid]; +"840 mul_10" [id=840, type=mul]; +"841 pad_7" [id=841, type=pad]; +"842 roll_4" [id=842, type=roll]; +"843 view_28" [id=843, type=view]; +"844 permute_24" [id=844, type=permute]; +"845 reshape_22" [id=845, type=reshape]; +"846 clone_5" [id=846, type=clone]; +"847 features_5_1_attn_qkv_weight_updated_constant0" [id=847, type=get_attr]; +"848 reshape_22_0_0_nncf_smooth_quant_0" [id=848, type=call_module]; +"849 linear_34" [id=849, type=linear]; +"850 reshape_23" [id=850, type=reshape]; +"851 permute_25" [id=851, type=permute]; +"852 select_15" [id=852, type=select]; +"853 select_16" [id=853, type=select]; +"854 select_17" [id=854, type=select]; +"855 linalg_vector_norm_10" [id=855, type=linalg_vector_norm]; +"856 clamp_min_10" [id=856, type=clamp_min]; +"857 expand_as_10" [id=857, type=expand_as]; +"858 div_10" [id=858, type=div]; +"859 linalg_vector_norm_11" [id=859, type=linalg_vector_norm]; +"860 clamp_min_11" [id=860, type=clamp_min]; +"861 expand_as_11" [id=861, type=expand_as]; +"862 div_11" [id=862, type=div]; +"863 transpose_10" [id=863, type=transpose]; +"864 matmul_10" [id=864, type=matmul]; +"865 clamp_5" [id=865, type=clamp]; +"866 exp_5" [id=866, type=exp]; +"867 mul_11" [id=867, type=mul]; +"868 add_17" [id=868, type=add]; +"869 new_zeros_2" [id=869, type=new_zeros]; +"870 view_29" [id=870, type=view]; +"871 permute_26" [id=871, type=permute]; +"872 reshape_24" [id=872, type=reshape]; +"873 unsqueeze_14" [id=873, type=unsqueeze]; +"874 unsqueeze_15" [id=874, type=unsqueeze]; +"875 sub_2" [id=875, type=sub]; +"876 ne_2" [id=876, type=ne]; +"877 masked_fill_4" [id=877, type=masked_fill]; +"878 eq_2" [id=878, type=eq]; +"879 masked_fill_5" [id=879, type=masked_fill]; +"880 view_30" [id=880, type=view]; +"881 unsqueeze_16" [id=881, type=unsqueeze]; +"882 unsqueeze_17" [id=882, type=unsqueeze]; +"883 add_18" [id=883, type=add]; +"884 view_31" [id=884, type=view]; +"885 softmax_5" [id=885, type=softmax]; +"886 dropout_20" [id=886, type=dropout]; +"887 matmul_11" [id=887, type=matmul]; +"888 transpose_11" [id=888, type=transpose]; +"889 reshape_25" [id=889, type=reshape]; +"890 features_5_1_attn_proj_weight_updated_constant0" [id=890, type=get_attr]; +"891 reshape_25_0_0_nncf_smooth_quant_0" [id=891, type=call_module]; +"892 linear_35" [id=892, type=linear]; +"893 dropout_21" [id=893, type=dropout]; +"894 view_32" [id=894, type=view]; +"895 permute_27" [id=895, type=permute]; +"896 reshape_26" [id=896, type=reshape]; +"897 roll_5" [id=897, type=roll]; +"898 slice_101" [id=898, type=slice]; +"899 slice_102" [id=899, type=slice]; +"900 slice_103" [id=900, type=slice]; +"901 slice_104" [id=901, type=slice]; +"902 contiguous_9" [id=902, type=contiguous]; +"903 layer_norm_13" [id=903, type=layer_norm]; +"904 add_19" [id=904, type=add]; +"905 add_19_0_0_nncf_smooth_quant_0" [id=905, type=call_module]; +"906 quantize_per_tensor_default_25" [id=906, type=quantize_per_tensor]; +"907 dequantize_per_tensor_default_25" [id=907, type=dequantize_per_tensor]; +"908 linear_36_scale_0" [id=908, type=get_attr]; +"909 linear_36_zero_point_0" [id=909, type=get_attr]; +"910 _frozen_param31" [id=910, type=get_attr]; +"911 dequantize_per_channel_default_25" [id=911, type=dequantize_per_channel]; +"912 linear_36" [id=912, type=linear]; +"913 gelu_5" [id=913, type=gelu]; +"914 dropout_22" [id=914, type=dropout]; +"915 dropout_22_0_0_nncf_smooth_quant_0" [id=915, type=call_module]; +"916 quantize_per_tensor_default_26" [id=916, type=quantize_per_tensor]; +"917 dequantize_per_tensor_default_26" [id=917, type=dequantize_per_tensor]; +"918 linear_37_scale_0" [id=918, type=get_attr]; +"919 linear_37_zero_point_0" [id=919, type=get_attr]; +"920 _frozen_param32" [id=920, type=get_attr]; +"921 dequantize_per_channel_default_26" [id=921, type=dequantize_per_channel]; +"922 linear_37" [id=922, type=linear]; +"923 dropout_23" [id=923, type=dropout]; +"924 layer_norm_14" [id=924, type=layer_norm]; +"925 add_20" [id=925, type=add]; +"926 _frozen_param33" [id=926, type=get_attr]; +"927 dequantize_per_tensor_default_27" [id=927, type=dequantize_per_tensor]; +"928 linear_38_scale_0" [id=928, type=get_attr]; +"929 linear_38_zero_point_0" [id=929, type=get_attr]; +"930 _frozen_param34" [id=930, type=get_attr]; +"931 dequantize_per_channel_default_27" [id=931, type=dequantize_per_channel]; +"932 linear_38" [id=932, type=linear]; +"933 relu__6" [id=933, type=relu_]; +"934 relu__6_0_0_nncf_smooth_quant_0" [id=934, type=call_module]; +"935 quantize_per_tensor_default_28" [id=935, type=quantize_per_tensor]; +"936 dequantize_per_tensor_default_28" [id=936, type=dequantize_per_tensor]; +"937 linear_39_scale_0" [id=937, type=get_attr]; +"938 linear_39_zero_point_0" [id=938, type=get_attr]; +"939 _frozen_param35" [id=939, type=get_attr]; +"940 dequantize_per_channel_default_28" [id=940, type=dequantize_per_channel]; +"941 linear_39" [id=941, type=linear]; +"942 view_33" [id=942, type=view]; +"943 index_6" [id=943, type=index]; +"944 view_34" [id=944, type=view]; +"945 permute_28" [id=945, type=permute]; +"946 contiguous_10" [id=946, type=contiguous]; +"947 unsqueeze_18" [id=947, type=unsqueeze]; +"948 sigmoid_6" [id=948, type=sigmoid]; +"949 mul_12" [id=949, type=mul]; +"950 pad_8" [id=950, type=pad]; +"951 view_35" [id=951, type=view]; +"952 permute_29" [id=952, type=permute]; +"953 reshape_27" [id=953, type=reshape]; +"954 clone_6" [id=954, type=clone]; +"955 features_5_2_attn_qkv_weight_updated_constant0" [id=955, type=get_attr]; +"956 reshape_27_0_0_nncf_smooth_quant_0" [id=956, type=call_module]; +"957 linear_40" [id=957, type=linear]; +"958 reshape_28" [id=958, type=reshape]; +"959 permute_30" [id=959, type=permute]; +"960 select_18" [id=960, type=select]; +"961 select_19" [id=961, type=select]; +"962 select_20" [id=962, type=select]; +"963 linalg_vector_norm_12" [id=963, type=linalg_vector_norm]; +"964 clamp_min_12" [id=964, type=clamp_min]; +"965 expand_as_12" [id=965, type=expand_as]; +"966 div_12" [id=966, type=div]; +"967 linalg_vector_norm_13" [id=967, type=linalg_vector_norm]; +"968 clamp_min_13" [id=968, type=clamp_min]; +"969 expand_as_13" [id=969, type=expand_as]; +"970 div_13" [id=970, type=div]; +"971 transpose_12" [id=971, type=transpose]; +"972 matmul_12" [id=972, type=matmul]; +"973 clamp_6" [id=973, type=clamp]; +"974 exp_6" [id=974, type=exp]; +"975 mul_13" [id=975, type=mul]; +"976 add_21" [id=976, type=add]; +"977 softmax_6" [id=977, type=softmax]; +"978 dropout_24" [id=978, type=dropout]; +"979 matmul_13" [id=979, type=matmul]; +"980 transpose_13" [id=980, type=transpose]; +"981 reshape_29" [id=981, type=reshape]; +"982 features_5_2_attn_proj_weight_updated_constant0" [id=982, type=get_attr]; +"983 reshape_29_0_0_nncf_smooth_quant_0" [id=983, type=call_module]; +"984 linear_41" [id=984, type=linear]; +"985 dropout_25" [id=985, type=dropout]; +"986 view_36" [id=986, type=view]; +"987 permute_31" [id=987, type=permute]; +"988 reshape_30" [id=988, type=reshape]; +"989 slice_106" [id=989, type=slice]; +"990 slice_107" [id=990, type=slice]; +"991 slice_108" [id=991, type=slice]; +"992 slice_109" [id=992, type=slice]; +"993 contiguous_11" [id=993, type=contiguous]; +"994 layer_norm_15" [id=994, type=layer_norm]; +"995 add_22" [id=995, type=add]; +"996 add_22_0_0_nncf_smooth_quant_0" [id=996, type=call_module]; +"997 quantize_per_tensor_default_29" [id=997, type=quantize_per_tensor]; +"998 dequantize_per_tensor_default_29" [id=998, type=dequantize_per_tensor]; +"999 linear_42_scale_0" [id=999, type=get_attr]; +"1000 linear_42_zero_point_0" [id=1000, type=get_attr]; +"1001 _frozen_param36" [id=1001, type=get_attr]; +"1002 dequantize_per_channel_default_29" [id=1002, type=dequantize_per_channel]; +"1003 linear_42" [id=1003, type=linear]; +"1004 gelu_6" [id=1004, type=gelu]; +"1005 dropout_26" [id=1005, type=dropout]; +"1006 dropout_26_0_0_nncf_smooth_quant_0" [id=1006, type=call_module]; +"1007 quantize_per_tensor_default_30" [id=1007, type=quantize_per_tensor]; +"1008 dequantize_per_tensor_default_30" [id=1008, type=dequantize_per_tensor]; +"1009 linear_43_scale_0" [id=1009, type=get_attr]; +"1010 linear_43_zero_point_0" [id=1010, type=get_attr]; +"1011 _frozen_param37" [id=1011, type=get_attr]; +"1012 dequantize_per_channel_default_30" [id=1012, type=dequantize_per_channel]; +"1013 linear_43" [id=1013, type=linear]; +"1014 dropout_27" [id=1014, type=dropout]; +"1015 layer_norm_16" [id=1015, type=layer_norm]; +"1016 add_23" [id=1016, type=add]; +"1017 _frozen_param38" [id=1017, type=get_attr]; +"1018 dequantize_per_tensor_default_31" [id=1018, type=dequantize_per_tensor]; +"1019 linear_44_scale_0" [id=1019, type=get_attr]; +"1020 linear_44_zero_point_0" [id=1020, type=get_attr]; +"1021 _frozen_param39" [id=1021, type=get_attr]; +"1022 dequantize_per_channel_default_31" [id=1022, type=dequantize_per_channel]; +"1023 linear_44" [id=1023, type=linear]; +"1024 relu__7" [id=1024, type=relu_]; +"1025 relu__7_0_0_nncf_smooth_quant_0" [id=1025, type=call_module]; +"1026 quantize_per_tensor_default_32" [id=1026, type=quantize_per_tensor]; +"1027 dequantize_per_tensor_default_32" [id=1027, type=dequantize_per_tensor]; +"1028 linear_45_scale_0" [id=1028, type=get_attr]; +"1029 linear_45_zero_point_0" [id=1029, type=get_attr]; +"1030 _frozen_param40" [id=1030, type=get_attr]; +"1031 dequantize_per_channel_default_32" [id=1031, type=dequantize_per_channel]; +"1032 linear_45" [id=1032, type=linear]; +"1033 view_37" [id=1033, type=view]; +"1034 index_7" [id=1034, type=index]; +"1035 view_38" [id=1035, type=view]; +"1036 permute_32" [id=1036, type=permute]; +"1037 contiguous_12" [id=1037, type=contiguous]; +"1038 unsqueeze_19" [id=1038, type=unsqueeze]; +"1039 sigmoid_7" [id=1039, type=sigmoid]; +"1040 mul_14" [id=1040, type=mul]; +"1041 pad_9" [id=1041, type=pad]; +"1042 roll_6" [id=1042, type=roll]; +"1043 view_39" [id=1043, type=view]; +"1044 permute_33" [id=1044, type=permute]; +"1045 reshape_31" [id=1045, type=reshape]; +"1046 clone_7" [id=1046, type=clone]; +"1047 features_5_3_attn_qkv_weight_updated_constant0" [id=1047, type=get_attr]; +"1048 reshape_31_0_0_nncf_smooth_quant_0" [id=1048, type=call_module]; +"1049 linear_46" [id=1049, type=linear]; +"1050 reshape_32" [id=1050, type=reshape]; +"1051 permute_34" [id=1051, type=permute]; +"1052 select_21" [id=1052, type=select]; +"1053 select_22" [id=1053, type=select]; +"1054 select_23" [id=1054, type=select]; +"1055 linalg_vector_norm_14" [id=1055, type=linalg_vector_norm]; +"1056 clamp_min_14" [id=1056, type=clamp_min]; +"1057 expand_as_14" [id=1057, type=expand_as]; +"1058 div_14" [id=1058, type=div]; +"1059 linalg_vector_norm_15" [id=1059, type=linalg_vector_norm]; +"1060 clamp_min_15" [id=1060, type=clamp_min]; +"1061 expand_as_15" [id=1061, type=expand_as]; +"1062 div_15" [id=1062, type=div]; +"1063 transpose_14" [id=1063, type=transpose]; +"1064 matmul_14" [id=1064, type=matmul]; +"1065 clamp_7" [id=1065, type=clamp]; +"1066 exp_7" [id=1066, type=exp]; +"1067 mul_15" [id=1067, type=mul]; +"1068 add_24" [id=1068, type=add]; +"1069 new_zeros_3" [id=1069, type=new_zeros]; +"1070 view_40" [id=1070, type=view]; +"1071 permute_35" [id=1071, type=permute]; +"1072 reshape_33" [id=1072, type=reshape]; +"1073 unsqueeze_20" [id=1073, type=unsqueeze]; +"1074 unsqueeze_21" [id=1074, type=unsqueeze]; +"1075 sub_3" [id=1075, type=sub]; +"1076 ne_3" [id=1076, type=ne]; +"1077 masked_fill_6" [id=1077, type=masked_fill]; +"1078 eq_3" [id=1078, type=eq]; +"1079 masked_fill_7" [id=1079, type=masked_fill]; +"1080 view_41" [id=1080, type=view]; +"1081 unsqueeze_22" [id=1081, type=unsqueeze]; +"1082 unsqueeze_23" [id=1082, type=unsqueeze]; +"1083 add_25" [id=1083, type=add]; +"1084 view_42" [id=1084, type=view]; +"1085 softmax_7" [id=1085, type=softmax]; +"1086 dropout_28" [id=1086, type=dropout]; +"1087 matmul_15" [id=1087, type=matmul]; +"1088 transpose_15" [id=1088, type=transpose]; +"1089 reshape_34" [id=1089, type=reshape]; +"1090 features_5_3_attn_proj_weight_updated_constant0" [id=1090, type=get_attr]; +"1091 reshape_34_0_0_nncf_smooth_quant_0" [id=1091, type=call_module]; +"1092 linear_47" [id=1092, type=linear]; +"1093 dropout_29" [id=1093, type=dropout]; +"1094 view_43" [id=1094, type=view]; +"1095 permute_36" [id=1095, type=permute]; +"1096 reshape_35" [id=1096, type=reshape]; +"1097 roll_7" [id=1097, type=roll]; +"1098 slice_129" [id=1098, type=slice]; +"1099 slice_130" [id=1099, type=slice]; +"1100 slice_131" [id=1100, type=slice]; +"1101 slice_132" [id=1101, type=slice]; +"1102 contiguous_13" [id=1102, type=contiguous]; +"1103 layer_norm_17" [id=1103, type=layer_norm]; +"1104 add_26" [id=1104, type=add]; +"1105 add_26_0_0_nncf_smooth_quant_0" [id=1105, type=call_module]; +"1106 quantize_per_tensor_default_33" [id=1106, type=quantize_per_tensor]; +"1107 dequantize_per_tensor_default_33" [id=1107, type=dequantize_per_tensor]; +"1108 linear_48_scale_0" [id=1108, type=get_attr]; +"1109 linear_48_zero_point_0" [id=1109, type=get_attr]; +"1110 _frozen_param41" [id=1110, type=get_attr]; +"1111 dequantize_per_channel_default_33" [id=1111, type=dequantize_per_channel]; +"1112 linear_48" [id=1112, type=linear]; +"1113 gelu_7" [id=1113, type=gelu]; +"1114 dropout_30" [id=1114, type=dropout]; +"1115 dropout_30_0_0_nncf_smooth_quant_0" [id=1115, type=call_module]; +"1116 quantize_per_tensor_default_34" [id=1116, type=quantize_per_tensor]; +"1117 dequantize_per_tensor_default_34" [id=1117, type=dequantize_per_tensor]; +"1118 linear_49_scale_0" [id=1118, type=get_attr]; +"1119 linear_49_zero_point_0" [id=1119, type=get_attr]; +"1120 _frozen_param42" [id=1120, type=get_attr]; +"1121 dequantize_per_channel_default_34" [id=1121, type=dequantize_per_channel]; +"1122 linear_49" [id=1122, type=linear]; +"1123 dropout_31" [id=1123, type=dropout]; +"1124 layer_norm_18" [id=1124, type=layer_norm]; +"1125 add_27" [id=1125, type=add]; +"1126 _frozen_param43" [id=1126, type=get_attr]; +"1127 dequantize_per_tensor_default_35" [id=1127, type=dequantize_per_tensor]; +"1128 linear_50_scale_0" [id=1128, type=get_attr]; +"1129 linear_50_zero_point_0" [id=1129, type=get_attr]; +"1130 _frozen_param44" [id=1130, type=get_attr]; +"1131 dequantize_per_channel_default_35" [id=1131, type=dequantize_per_channel]; +"1132 linear_50" [id=1132, type=linear]; +"1133 relu__8" [id=1133, type=relu_]; +"1134 relu__8_0_0_nncf_smooth_quant_0" [id=1134, type=call_module]; +"1135 quantize_per_tensor_default_36" [id=1135, type=quantize_per_tensor]; +"1136 dequantize_per_tensor_default_36" [id=1136, type=dequantize_per_tensor]; +"1137 linear_51_scale_0" [id=1137, type=get_attr]; +"1138 linear_51_zero_point_0" [id=1138, type=get_attr]; +"1139 _frozen_param45" [id=1139, type=get_attr]; +"1140 dequantize_per_channel_default_36" [id=1140, type=dequantize_per_channel]; +"1141 linear_51" [id=1141, type=linear]; +"1142 view_44" [id=1142, type=view]; +"1143 index_8" [id=1143, type=index]; +"1144 view_45" [id=1144, type=view]; +"1145 permute_37" [id=1145, type=permute]; +"1146 contiguous_14" [id=1146, type=contiguous]; +"1147 unsqueeze_24" [id=1147, type=unsqueeze]; +"1148 sigmoid_8" [id=1148, type=sigmoid]; +"1149 mul_16" [id=1149, type=mul]; +"1150 pad_10" [id=1150, type=pad]; +"1151 view_46" [id=1151, type=view]; +"1152 permute_38" [id=1152, type=permute]; +"1153 reshape_36" [id=1153, type=reshape]; +"1154 clone_8" [id=1154, type=clone]; +"1155 features_5_4_attn_qkv_weight_updated_constant0" [id=1155, type=get_attr]; +"1156 reshape_36_0_0_nncf_smooth_quant_0" [id=1156, type=call_module]; +"1157 linear_52" [id=1157, type=linear]; +"1158 reshape_37" [id=1158, type=reshape]; +"1159 permute_39" [id=1159, type=permute]; +"1160 select_24" [id=1160, type=select]; +"1161 select_25" [id=1161, type=select]; +"1162 select_26" [id=1162, type=select]; +"1163 linalg_vector_norm_16" [id=1163, type=linalg_vector_norm]; +"1164 clamp_min_16" [id=1164, type=clamp_min]; +"1165 expand_as_16" [id=1165, type=expand_as]; +"1166 div_16" [id=1166, type=div]; +"1167 linalg_vector_norm_17" [id=1167, type=linalg_vector_norm]; +"1168 clamp_min_17" [id=1168, type=clamp_min]; +"1169 expand_as_17" [id=1169, type=expand_as]; +"1170 div_17" [id=1170, type=div]; +"1171 transpose_16" [id=1171, type=transpose]; +"1172 matmul_16" [id=1172, type=matmul]; +"1173 clamp_8" [id=1173, type=clamp]; +"1174 exp_8" [id=1174, type=exp]; +"1175 mul_17" [id=1175, type=mul]; +"1176 add_28" [id=1176, type=add]; +"1177 softmax_8" [id=1177, type=softmax]; +"1178 dropout_32" [id=1178, type=dropout]; +"1179 matmul_17" [id=1179, type=matmul]; +"1180 transpose_17" [id=1180, type=transpose]; +"1181 reshape_38" [id=1181, type=reshape]; +"1182 features_5_4_attn_proj_weight_updated_constant0" [id=1182, type=get_attr]; +"1183 reshape_38_0_0_nncf_smooth_quant_0" [id=1183, type=call_module]; +"1184 linear_53" [id=1184, type=linear]; +"1185 dropout_33" [id=1185, type=dropout]; +"1186 view_47" [id=1186, type=view]; +"1187 permute_40" [id=1187, type=permute]; +"1188 reshape_39" [id=1188, type=reshape]; +"1189 slice_134" [id=1189, type=slice]; +"1190 slice_135" [id=1190, type=slice]; +"1191 slice_136" [id=1191, type=slice]; +"1192 slice_137" [id=1192, type=slice]; +"1193 contiguous_15" [id=1193, type=contiguous]; +"1194 layer_norm_19" [id=1194, type=layer_norm]; +"1195 add_29" [id=1195, type=add]; +"1196 add_29_0_0_nncf_smooth_quant_0" [id=1196, type=call_module]; +"1197 quantize_per_tensor_default_37" [id=1197, type=quantize_per_tensor]; +"1198 dequantize_per_tensor_default_37" [id=1198, type=dequantize_per_tensor]; +"1199 linear_54_scale_0" [id=1199, type=get_attr]; +"1200 linear_54_zero_point_0" [id=1200, type=get_attr]; +"1201 _frozen_param46" [id=1201, type=get_attr]; +"1202 dequantize_per_channel_default_37" [id=1202, type=dequantize_per_channel]; +"1203 linear_54" [id=1203, type=linear]; +"1204 gelu_8" [id=1204, type=gelu]; +"1205 dropout_34" [id=1205, type=dropout]; +"1206 dropout_34_0_0_nncf_smooth_quant_0" [id=1206, type=call_module]; +"1207 quantize_per_tensor_default_38" [id=1207, type=quantize_per_tensor]; +"1208 dequantize_per_tensor_default_38" [id=1208, type=dequantize_per_tensor]; +"1209 linear_55_scale_0" [id=1209, type=get_attr]; +"1210 linear_55_zero_point_0" [id=1210, type=get_attr]; +"1211 _frozen_param47" [id=1211, type=get_attr]; +"1212 dequantize_per_channel_default_38" [id=1212, type=dequantize_per_channel]; +"1213 linear_55" [id=1213, type=linear]; +"1214 dropout_35" [id=1214, type=dropout]; +"1215 layer_norm_20" [id=1215, type=layer_norm]; +"1216 add_30" [id=1216, type=add]; +"1217 _frozen_param48" [id=1217, type=get_attr]; +"1218 dequantize_per_tensor_default_39" [id=1218, type=dequantize_per_tensor]; +"1219 linear_56_scale_0" [id=1219, type=get_attr]; +"1220 linear_56_zero_point_0" [id=1220, type=get_attr]; +"1221 _frozen_param49" [id=1221, type=get_attr]; +"1222 dequantize_per_channel_default_39" [id=1222, type=dequantize_per_channel]; +"1223 linear_56" [id=1223, type=linear]; +"1224 relu__9" [id=1224, type=relu_]; +"1225 relu__9_0_0_nncf_smooth_quant_0" [id=1225, type=call_module]; +"1226 quantize_per_tensor_default_40" [id=1226, type=quantize_per_tensor]; +"1227 dequantize_per_tensor_default_40" [id=1227, type=dequantize_per_tensor]; +"1228 linear_57_scale_0" [id=1228, type=get_attr]; +"1229 linear_57_zero_point_0" [id=1229, type=get_attr]; +"1230 _frozen_param50" [id=1230, type=get_attr]; +"1231 dequantize_per_channel_default_40" [id=1231, type=dequantize_per_channel]; +"1232 linear_57" [id=1232, type=linear]; +"1233 view_48" [id=1233, type=view]; +"1234 index_9" [id=1234, type=index]; +"1235 view_49" [id=1235, type=view]; +"1236 permute_41" [id=1236, type=permute]; +"1237 contiguous_16" [id=1237, type=contiguous]; +"1238 unsqueeze_25" [id=1238, type=unsqueeze]; +"1239 sigmoid_9" [id=1239, type=sigmoid]; +"1240 mul_18" [id=1240, type=mul]; +"1241 pad_11" [id=1241, type=pad]; +"1242 roll_8" [id=1242, type=roll]; +"1243 view_50" [id=1243, type=view]; +"1244 permute_42" [id=1244, type=permute]; +"1245 reshape_40" [id=1245, type=reshape]; +"1246 clone_9" [id=1246, type=clone]; +"1247 features_5_5_attn_qkv_weight_updated_constant0" [id=1247, type=get_attr]; +"1248 reshape_40_0_0_nncf_smooth_quant_0" [id=1248, type=call_module]; +"1249 linear_58" [id=1249, type=linear]; +"1250 reshape_41" [id=1250, type=reshape]; +"1251 permute_43" [id=1251, type=permute]; +"1252 select_27" [id=1252, type=select]; +"1253 select_28" [id=1253, type=select]; +"1254 select_29" [id=1254, type=select]; +"1255 linalg_vector_norm_18" [id=1255, type=linalg_vector_norm]; +"1256 clamp_min_18" [id=1256, type=clamp_min]; +"1257 expand_as_18" [id=1257, type=expand_as]; +"1258 div_18" [id=1258, type=div]; +"1259 linalg_vector_norm_19" [id=1259, type=linalg_vector_norm]; +"1260 clamp_min_19" [id=1260, type=clamp_min]; +"1261 expand_as_19" [id=1261, type=expand_as]; +"1262 div_19" [id=1262, type=div]; +"1263 transpose_18" [id=1263, type=transpose]; +"1264 matmul_18" [id=1264, type=matmul]; +"1265 clamp_9" [id=1265, type=clamp]; +"1266 exp_9" [id=1266, type=exp]; +"1267 mul_19" [id=1267, type=mul]; +"1268 add_31" [id=1268, type=add]; +"1269 new_zeros_4" [id=1269, type=new_zeros]; +"1270 view_51" [id=1270, type=view]; +"1271 permute_44" [id=1271, type=permute]; +"1272 reshape_42" [id=1272, type=reshape]; +"1273 unsqueeze_26" [id=1273, type=unsqueeze]; +"1274 unsqueeze_27" [id=1274, type=unsqueeze]; +"1275 sub_4" [id=1275, type=sub]; +"1276 ne_4" [id=1276, type=ne]; +"1277 masked_fill_8" [id=1277, type=masked_fill]; +"1278 eq_4" [id=1278, type=eq]; +"1279 masked_fill_9" [id=1279, type=masked_fill]; +"1280 view_52" [id=1280, type=view]; +"1281 unsqueeze_28" [id=1281, type=unsqueeze]; +"1282 unsqueeze_29" [id=1282, type=unsqueeze]; +"1283 add_32" [id=1283, type=add]; +"1284 view_53" [id=1284, type=view]; +"1285 softmax_9" [id=1285, type=softmax]; +"1286 dropout_36" [id=1286, type=dropout]; +"1287 matmul_19" [id=1287, type=matmul]; +"1288 transpose_19" [id=1288, type=transpose]; +"1289 reshape_43" [id=1289, type=reshape]; +"1290 features_5_5_attn_proj_weight_updated_constant0" [id=1290, type=get_attr]; +"1291 reshape_43_0_0_nncf_smooth_quant_0" [id=1291, type=call_module]; +"1292 linear_59" [id=1292, type=linear]; +"1293 dropout_37" [id=1293, type=dropout]; +"1294 view_54" [id=1294, type=view]; +"1295 permute_45" [id=1295, type=permute]; +"1296 reshape_44" [id=1296, type=reshape]; +"1297 roll_9" [id=1297, type=roll]; +"1298 slice_157" [id=1298, type=slice]; +"1299 slice_158" [id=1299, type=slice]; +"1300 slice_159" [id=1300, type=slice]; +"1301 slice_160" [id=1301, type=slice]; +"1302 contiguous_17" [id=1302, type=contiguous]; +"1303 layer_norm_21" [id=1303, type=layer_norm]; +"1304 add_33" [id=1304, type=add]; +"1305 add_33_0_0_nncf_smooth_quant_0" [id=1305, type=call_module]; +"1306 quantize_per_tensor_default_41" [id=1306, type=quantize_per_tensor]; +"1307 dequantize_per_tensor_default_41" [id=1307, type=dequantize_per_tensor]; +"1308 linear_60_scale_0" [id=1308, type=get_attr]; +"1309 linear_60_zero_point_0" [id=1309, type=get_attr]; +"1310 _frozen_param51" [id=1310, type=get_attr]; +"1311 dequantize_per_channel_default_41" [id=1311, type=dequantize_per_channel]; +"1312 linear_60" [id=1312, type=linear]; +"1313 gelu_9" [id=1313, type=gelu]; +"1314 dropout_38" [id=1314, type=dropout]; +"1315 dropout_38_0_0_nncf_smooth_quant_0" [id=1315, type=call_module]; +"1316 quantize_per_tensor_default_42" [id=1316, type=quantize_per_tensor]; +"1317 dequantize_per_tensor_default_42" [id=1317, type=dequantize_per_tensor]; +"1318 linear_61_scale_0" [id=1318, type=get_attr]; +"1319 linear_61_zero_point_0" [id=1319, type=get_attr]; +"1320 _frozen_param52" [id=1320, type=get_attr]; +"1321 dequantize_per_channel_default_42" [id=1321, type=dequantize_per_channel]; +"1322 linear_61" [id=1322, type=linear]; +"1323 dropout_39" [id=1323, type=dropout]; +"1324 layer_norm_22" [id=1324, type=layer_norm]; +"1325 add_34" [id=1325, type=add]; +"1326 _frozen_param53" [id=1326, type=get_attr]; +"1327 dequantize_per_tensor_default_43" [id=1327, type=dequantize_per_tensor]; +"1328 linear_62_scale_0" [id=1328, type=get_attr]; +"1329 linear_62_zero_point_0" [id=1329, type=get_attr]; +"1330 _frozen_param54" [id=1330, type=get_attr]; +"1331 dequantize_per_channel_default_43" [id=1331, type=dequantize_per_channel]; +"1332 linear_62" [id=1332, type=linear]; +"1333 relu__10" [id=1333, type=relu_]; +"1334 relu__10_0_0_nncf_smooth_quant_0" [id=1334, type=call_module]; +"1335 quantize_per_tensor_default_44" [id=1335, type=quantize_per_tensor]; +"1336 dequantize_per_tensor_default_44" [id=1336, type=dequantize_per_tensor]; +"1337 linear_63_scale_0" [id=1337, type=get_attr]; +"1338 linear_63_zero_point_0" [id=1338, type=get_attr]; +"1339 _frozen_param55" [id=1339, type=get_attr]; +"1340 dequantize_per_channel_default_44" [id=1340, type=dequantize_per_channel]; +"1341 linear_63" [id=1341, type=linear]; +"1342 view_55" [id=1342, type=view]; +"1343 index_10" [id=1343, type=index]; +"1344 view_56" [id=1344, type=view]; +"1345 permute_46" [id=1345, type=permute]; +"1346 contiguous_18" [id=1346, type=contiguous]; +"1347 unsqueeze_30" [id=1347, type=unsqueeze]; +"1348 sigmoid_10" [id=1348, type=sigmoid]; +"1349 mul_20" [id=1349, type=mul]; +"1350 pad_12" [id=1350, type=pad]; +"1351 view_57" [id=1351, type=view]; +"1352 permute_47" [id=1352, type=permute]; +"1353 reshape_45" [id=1353, type=reshape]; +"1354 clone_10" [id=1354, type=clone]; +"1355 features_5_6_attn_qkv_weight_updated_constant0" [id=1355, type=get_attr]; +"1356 reshape_45_0_0_nncf_smooth_quant_0" [id=1356, type=call_module]; +"1357 linear_64" [id=1357, type=linear]; +"1358 reshape_46" [id=1358, type=reshape]; +"1359 permute_48" [id=1359, type=permute]; +"1360 select_30" [id=1360, type=select]; +"1361 select_31" [id=1361, type=select]; +"1362 select_32" [id=1362, type=select]; +"1363 linalg_vector_norm_20" [id=1363, type=linalg_vector_norm]; +"1364 clamp_min_20" [id=1364, type=clamp_min]; +"1365 expand_as_20" [id=1365, type=expand_as]; +"1366 div_20" [id=1366, type=div]; +"1367 linalg_vector_norm_21" [id=1367, type=linalg_vector_norm]; +"1368 clamp_min_21" [id=1368, type=clamp_min]; +"1369 expand_as_21" [id=1369, type=expand_as]; +"1370 div_21" [id=1370, type=div]; +"1371 transpose_20" [id=1371, type=transpose]; +"1372 matmul_20" [id=1372, type=matmul]; +"1373 clamp_10" [id=1373, type=clamp]; +"1374 exp_10" [id=1374, type=exp]; +"1375 mul_21" [id=1375, type=mul]; +"1376 add_35" [id=1376, type=add]; +"1377 softmax_10" [id=1377, type=softmax]; +"1378 dropout_40" [id=1378, type=dropout]; +"1379 matmul_21" [id=1379, type=matmul]; +"1380 transpose_21" [id=1380, type=transpose]; +"1381 reshape_47" [id=1381, type=reshape]; +"1382 features_5_6_attn_proj_weight_updated_constant0" [id=1382, type=get_attr]; +"1383 reshape_47_0_0_nncf_smooth_quant_0" [id=1383, type=call_module]; +"1384 linear_65" [id=1384, type=linear]; +"1385 dropout_41" [id=1385, type=dropout]; +"1386 view_58" [id=1386, type=view]; +"1387 permute_49" [id=1387, type=permute]; +"1388 reshape_48" [id=1388, type=reshape]; +"1389 slice_162" [id=1389, type=slice]; +"1390 slice_163" [id=1390, type=slice]; +"1391 slice_164" [id=1391, type=slice]; +"1392 slice_165" [id=1392, type=slice]; +"1393 contiguous_19" [id=1393, type=contiguous]; +"1394 layer_norm_23" [id=1394, type=layer_norm]; +"1395 add_36" [id=1395, type=add]; +"1396 add_36_0_0_nncf_smooth_quant_0" [id=1396, type=call_module]; +"1397 quantize_per_tensor_default_45" [id=1397, type=quantize_per_tensor]; +"1398 dequantize_per_tensor_default_45" [id=1398, type=dequantize_per_tensor]; +"1399 linear_66_scale_0" [id=1399, type=get_attr]; +"1400 linear_66_zero_point_0" [id=1400, type=get_attr]; +"1401 _frozen_param56" [id=1401, type=get_attr]; +"1402 dequantize_per_channel_default_45" [id=1402, type=dequantize_per_channel]; +"1403 linear_66" [id=1403, type=linear]; +"1404 gelu_10" [id=1404, type=gelu]; +"1405 dropout_42" [id=1405, type=dropout]; +"1406 dropout_42_0_0_nncf_smooth_quant_0" [id=1406, type=call_module]; +"1407 quantize_per_tensor_default_46" [id=1407, type=quantize_per_tensor]; +"1408 dequantize_per_tensor_default_46" [id=1408, type=dequantize_per_tensor]; +"1409 linear_67_scale_0" [id=1409, type=get_attr]; +"1410 linear_67_zero_point_0" [id=1410, type=get_attr]; +"1411 _frozen_param57" [id=1411, type=get_attr]; +"1412 dequantize_per_channel_default_46" [id=1412, type=dequantize_per_channel]; +"1413 linear_67" [id=1413, type=linear]; +"1414 dropout_43" [id=1414, type=dropout]; +"1415 layer_norm_24" [id=1415, type=layer_norm]; +"1416 add_37" [id=1416, type=add]; +"1417 _frozen_param58" [id=1417, type=get_attr]; +"1418 dequantize_per_tensor_default_47" [id=1418, type=dequantize_per_tensor]; +"1419 linear_68_scale_0" [id=1419, type=get_attr]; +"1420 linear_68_zero_point_0" [id=1420, type=get_attr]; +"1421 _frozen_param59" [id=1421, type=get_attr]; +"1422 dequantize_per_channel_default_47" [id=1422, type=dequantize_per_channel]; +"1423 linear_68" [id=1423, type=linear]; +"1424 relu__11" [id=1424, type=relu_]; +"1425 relu__11_0_0_nncf_smooth_quant_0" [id=1425, type=call_module]; +"1426 quantize_per_tensor_default_48" [id=1426, type=quantize_per_tensor]; +"1427 dequantize_per_tensor_default_48" [id=1427, type=dequantize_per_tensor]; +"1428 linear_69_scale_0" [id=1428, type=get_attr]; +"1429 linear_69_zero_point_0" [id=1429, type=get_attr]; +"1430 _frozen_param60" [id=1430, type=get_attr]; +"1431 dequantize_per_channel_default_48" [id=1431, type=dequantize_per_channel]; +"1432 linear_69" [id=1432, type=linear]; +"1433 view_59" [id=1433, type=view]; +"1434 index_11" [id=1434, type=index]; +"1435 view_60" [id=1435, type=view]; +"1436 permute_50" [id=1436, type=permute]; +"1437 contiguous_20" [id=1437, type=contiguous]; +"1438 unsqueeze_31" [id=1438, type=unsqueeze]; +"1439 sigmoid_11" [id=1439, type=sigmoid]; +"1440 mul_22" [id=1440, type=mul]; +"1441 pad_13" [id=1441, type=pad]; +"1442 roll_10" [id=1442, type=roll]; +"1443 view_61" [id=1443, type=view]; +"1444 permute_51" [id=1444, type=permute]; +"1445 reshape_49" [id=1445, type=reshape]; +"1446 clone_11" [id=1446, type=clone]; +"1447 features_5_7_attn_qkv_weight_updated_constant0" [id=1447, type=get_attr]; +"1448 reshape_49_0_0_nncf_smooth_quant_0" [id=1448, type=call_module]; +"1449 linear_70" [id=1449, type=linear]; +"1450 reshape_50" [id=1450, type=reshape]; +"1451 permute_52" [id=1451, type=permute]; +"1452 select_33" [id=1452, type=select]; +"1453 select_34" [id=1453, type=select]; +"1454 select_35" [id=1454, type=select]; +"1455 linalg_vector_norm_22" [id=1455, type=linalg_vector_norm]; +"1456 clamp_min_22" [id=1456, type=clamp_min]; +"1457 expand_as_22" [id=1457, type=expand_as]; +"1458 div_22" [id=1458, type=div]; +"1459 linalg_vector_norm_23" [id=1459, type=linalg_vector_norm]; +"1460 clamp_min_23" [id=1460, type=clamp_min]; +"1461 expand_as_23" [id=1461, type=expand_as]; +"1462 div_23" [id=1462, type=div]; +"1463 transpose_22" [id=1463, type=transpose]; +"1464 matmul_22" [id=1464, type=matmul]; +"1465 clamp_11" [id=1465, type=clamp]; +"1466 exp_11" [id=1466, type=exp]; +"1467 mul_23" [id=1467, type=mul]; +"1468 add_38" [id=1468, type=add]; +"1469 new_zeros_5" [id=1469, type=new_zeros]; +"1470 view_62" [id=1470, type=view]; +"1471 permute_53" [id=1471, type=permute]; +"1472 reshape_51" [id=1472, type=reshape]; +"1473 unsqueeze_32" [id=1473, type=unsqueeze]; +"1474 unsqueeze_33" [id=1474, type=unsqueeze]; +"1475 sub_5" [id=1475, type=sub]; +"1476 ne_5" [id=1476, type=ne]; +"1477 masked_fill_10" [id=1477, type=masked_fill]; +"1478 eq_5" [id=1478, type=eq]; +"1479 masked_fill_11" [id=1479, type=masked_fill]; +"1480 view_63" [id=1480, type=view]; +"1481 unsqueeze_34" [id=1481, type=unsqueeze]; +"1482 unsqueeze_35" [id=1482, type=unsqueeze]; +"1483 add_39" [id=1483, type=add]; +"1484 view_64" [id=1484, type=view]; +"1485 softmax_11" [id=1485, type=softmax]; +"1486 dropout_44" [id=1486, type=dropout]; +"1487 matmul_23" [id=1487, type=matmul]; +"1488 transpose_23" [id=1488, type=transpose]; +"1489 reshape_52" [id=1489, type=reshape]; +"1490 features_5_7_attn_proj_weight_updated_constant0" [id=1490, type=get_attr]; +"1491 reshape_52_0_0_nncf_smooth_quant_0" [id=1491, type=call_module]; +"1492 linear_71" [id=1492, type=linear]; +"1493 dropout_45" [id=1493, type=dropout]; +"1494 view_65" [id=1494, type=view]; +"1495 permute_54" [id=1495, type=permute]; +"1496 reshape_53" [id=1496, type=reshape]; +"1497 roll_11" [id=1497, type=roll]; +"1498 slice_185" [id=1498, type=slice]; +"1499 slice_186" [id=1499, type=slice]; +"1500 slice_187" [id=1500, type=slice]; +"1501 slice_188" [id=1501, type=slice]; +"1502 contiguous_21" [id=1502, type=contiguous]; +"1503 layer_norm_25" [id=1503, type=layer_norm]; +"1504 add_40" [id=1504, type=add]; +"1505 add_40_0_0_nncf_smooth_quant_0" [id=1505, type=call_module]; +"1506 quantize_per_tensor_default_49" [id=1506, type=quantize_per_tensor]; +"1507 dequantize_per_tensor_default_49" [id=1507, type=dequantize_per_tensor]; +"1508 linear_72_scale_0" [id=1508, type=get_attr]; +"1509 linear_72_zero_point_0" [id=1509, type=get_attr]; +"1510 _frozen_param61" [id=1510, type=get_attr]; +"1511 dequantize_per_channel_default_49" [id=1511, type=dequantize_per_channel]; +"1512 linear_72" [id=1512, type=linear]; +"1513 gelu_11" [id=1513, type=gelu]; +"1514 dropout_46" [id=1514, type=dropout]; +"1515 dropout_46_0_0_nncf_smooth_quant_0" [id=1515, type=call_module]; +"1516 quantize_per_tensor_default_50" [id=1516, type=quantize_per_tensor]; +"1517 dequantize_per_tensor_default_50" [id=1517, type=dequantize_per_tensor]; +"1518 linear_73_scale_0" [id=1518, type=get_attr]; +"1519 linear_73_zero_point_0" [id=1519, type=get_attr]; +"1520 _frozen_param62" [id=1520, type=get_attr]; +"1521 dequantize_per_channel_default_50" [id=1521, type=dequantize_per_channel]; +"1522 linear_73" [id=1522, type=linear]; +"1523 dropout_47" [id=1523, type=dropout]; +"1524 layer_norm_26" [id=1524, type=layer_norm]; +"1525 add_41" [id=1525, type=add]; +"1526 _frozen_param63" [id=1526, type=get_attr]; +"1527 dequantize_per_tensor_default_51" [id=1527, type=dequantize_per_tensor]; +"1528 linear_74_scale_0" [id=1528, type=get_attr]; +"1529 linear_74_zero_point_0" [id=1529, type=get_attr]; +"1530 _frozen_param64" [id=1530, type=get_attr]; +"1531 dequantize_per_channel_default_51" [id=1531, type=dequantize_per_channel]; +"1532 linear_74" [id=1532, type=linear]; +"1533 relu__12" [id=1533, type=relu_]; +"1534 relu__12_0_0_nncf_smooth_quant_0" [id=1534, type=call_module]; +"1535 quantize_per_tensor_default_52" [id=1535, type=quantize_per_tensor]; +"1536 dequantize_per_tensor_default_52" [id=1536, type=dequantize_per_tensor]; +"1537 linear_75_scale_0" [id=1537, type=get_attr]; +"1538 linear_75_zero_point_0" [id=1538, type=get_attr]; +"1539 _frozen_param65" [id=1539, type=get_attr]; +"1540 dequantize_per_channel_default_52" [id=1540, type=dequantize_per_channel]; +"1541 linear_75" [id=1541, type=linear]; +"1542 view_66" [id=1542, type=view]; +"1543 index_12" [id=1543, type=index]; +"1544 view_67" [id=1544, type=view]; +"1545 permute_55" [id=1545, type=permute]; +"1546 contiguous_22" [id=1546, type=contiguous]; +"1547 unsqueeze_36" [id=1547, type=unsqueeze]; +"1548 sigmoid_12" [id=1548, type=sigmoid]; +"1549 mul_24" [id=1549, type=mul]; +"1550 pad_14" [id=1550, type=pad]; +"1551 view_68" [id=1551, type=view]; +"1552 permute_56" [id=1552, type=permute]; +"1553 reshape_54" [id=1553, type=reshape]; +"1554 clone_12" [id=1554, type=clone]; +"1555 features_5_8_attn_qkv_weight_updated_constant0" [id=1555, type=get_attr]; +"1556 reshape_54_0_0_nncf_smooth_quant_0" [id=1556, type=call_module]; +"1557 linear_76" [id=1557, type=linear]; +"1558 reshape_55" [id=1558, type=reshape]; +"1559 permute_57" [id=1559, type=permute]; +"1560 select_36" [id=1560, type=select]; +"1561 select_37" [id=1561, type=select]; +"1562 select_38" [id=1562, type=select]; +"1563 linalg_vector_norm_24" [id=1563, type=linalg_vector_norm]; +"1564 clamp_min_24" [id=1564, type=clamp_min]; +"1565 expand_as_24" [id=1565, type=expand_as]; +"1566 div_24" [id=1566, type=div]; +"1567 linalg_vector_norm_25" [id=1567, type=linalg_vector_norm]; +"1568 clamp_min_25" [id=1568, type=clamp_min]; +"1569 expand_as_25" [id=1569, type=expand_as]; +"1570 div_25" [id=1570, type=div]; +"1571 transpose_24" [id=1571, type=transpose]; +"1572 matmul_24" [id=1572, type=matmul]; +"1573 clamp_12" [id=1573, type=clamp]; +"1574 exp_12" [id=1574, type=exp]; +"1575 mul_25" [id=1575, type=mul]; +"1576 add_42" [id=1576, type=add]; +"1577 softmax_12" [id=1577, type=softmax]; +"1578 dropout_48" [id=1578, type=dropout]; +"1579 matmul_25" [id=1579, type=matmul]; +"1580 transpose_25" [id=1580, type=transpose]; +"1581 reshape_56" [id=1581, type=reshape]; +"1582 features_5_8_attn_proj_weight_updated_constant0" [id=1582, type=get_attr]; +"1583 reshape_56_0_0_nncf_smooth_quant_0" [id=1583, type=call_module]; +"1584 linear_77" [id=1584, type=linear]; +"1585 dropout_49" [id=1585, type=dropout]; +"1586 view_69" [id=1586, type=view]; +"1587 permute_58" [id=1587, type=permute]; +"1588 reshape_57" [id=1588, type=reshape]; +"1589 slice_190" [id=1589, type=slice]; +"1590 slice_191" [id=1590, type=slice]; +"1591 slice_192" [id=1591, type=slice]; +"1592 slice_193" [id=1592, type=slice]; +"1593 contiguous_23" [id=1593, type=contiguous]; +"1594 layer_norm_27" [id=1594, type=layer_norm]; +"1595 add_43" [id=1595, type=add]; +"1596 add_43_0_0_nncf_smooth_quant_0" [id=1596, type=call_module]; +"1597 quantize_per_tensor_default_53" [id=1597, type=quantize_per_tensor]; +"1598 dequantize_per_tensor_default_53" [id=1598, type=dequantize_per_tensor]; +"1599 linear_78_scale_0" [id=1599, type=get_attr]; +"1600 linear_78_zero_point_0" [id=1600, type=get_attr]; +"1601 _frozen_param66" [id=1601, type=get_attr]; +"1602 dequantize_per_channel_default_53" [id=1602, type=dequantize_per_channel]; +"1603 linear_78" [id=1603, type=linear]; +"1604 gelu_12" [id=1604, type=gelu]; +"1605 dropout_50" [id=1605, type=dropout]; +"1606 dropout_50_0_0_nncf_smooth_quant_0" [id=1606, type=call_module]; +"1607 quantize_per_tensor_default_54" [id=1607, type=quantize_per_tensor]; +"1608 dequantize_per_tensor_default_54" [id=1608, type=dequantize_per_tensor]; +"1609 linear_79_scale_0" [id=1609, type=get_attr]; +"1610 linear_79_zero_point_0" [id=1610, type=get_attr]; +"1611 _frozen_param67" [id=1611, type=get_attr]; +"1612 dequantize_per_channel_default_54" [id=1612, type=dequantize_per_channel]; +"1613 linear_79" [id=1613, type=linear]; +"1614 dropout_51" [id=1614, type=dropout]; +"1615 layer_norm_28" [id=1615, type=layer_norm]; +"1616 add_44" [id=1616, type=add]; +"1617 _frozen_param68" [id=1617, type=get_attr]; +"1618 dequantize_per_tensor_default_55" [id=1618, type=dequantize_per_tensor]; +"1619 linear_80_scale_0" [id=1619, type=get_attr]; +"1620 linear_80_zero_point_0" [id=1620, type=get_attr]; +"1621 _frozen_param69" [id=1621, type=get_attr]; +"1622 dequantize_per_channel_default_55" [id=1622, type=dequantize_per_channel]; +"1623 linear_80" [id=1623, type=linear]; +"1624 relu__13" [id=1624, type=relu_]; +"1625 relu__13_0_0_nncf_smooth_quant_0" [id=1625, type=call_module]; +"1626 quantize_per_tensor_default_56" [id=1626, type=quantize_per_tensor]; +"1627 dequantize_per_tensor_default_56" [id=1627, type=dequantize_per_tensor]; +"1628 linear_81_scale_0" [id=1628, type=get_attr]; +"1629 linear_81_zero_point_0" [id=1629, type=get_attr]; +"1630 _frozen_param70" [id=1630, type=get_attr]; +"1631 dequantize_per_channel_default_56" [id=1631, type=dequantize_per_channel]; +"1632 linear_81" [id=1632, type=linear]; +"1633 view_70" [id=1633, type=view]; +"1634 index_13" [id=1634, type=index]; +"1635 view_71" [id=1635, type=view]; +"1636 permute_59" [id=1636, type=permute]; +"1637 contiguous_24" [id=1637, type=contiguous]; +"1638 unsqueeze_37" [id=1638, type=unsqueeze]; +"1639 sigmoid_13" [id=1639, type=sigmoid]; +"1640 mul_26" [id=1640, type=mul]; +"1641 pad_15" [id=1641, type=pad]; +"1642 roll_12" [id=1642, type=roll]; +"1643 view_72" [id=1643, type=view]; +"1644 permute_60" [id=1644, type=permute]; +"1645 reshape_58" [id=1645, type=reshape]; +"1646 clone_13" [id=1646, type=clone]; +"1647 features_5_9_attn_qkv_weight_updated_constant0" [id=1647, type=get_attr]; +"1648 reshape_58_0_0_nncf_smooth_quant_0" [id=1648, type=call_module]; +"1649 linear_82" [id=1649, type=linear]; +"1650 reshape_59" [id=1650, type=reshape]; +"1651 permute_61" [id=1651, type=permute]; +"1652 select_39" [id=1652, type=select]; +"1653 select_40" [id=1653, type=select]; +"1654 select_41" [id=1654, type=select]; +"1655 linalg_vector_norm_26" [id=1655, type=linalg_vector_norm]; +"1656 clamp_min_26" [id=1656, type=clamp_min]; +"1657 expand_as_26" [id=1657, type=expand_as]; +"1658 div_26" [id=1658, type=div]; +"1659 linalg_vector_norm_27" [id=1659, type=linalg_vector_norm]; +"1660 clamp_min_27" [id=1660, type=clamp_min]; +"1661 expand_as_27" [id=1661, type=expand_as]; +"1662 div_27" [id=1662, type=div]; +"1663 transpose_26" [id=1663, type=transpose]; +"1664 matmul_26" [id=1664, type=matmul]; +"1665 clamp_13" [id=1665, type=clamp]; +"1666 exp_13" [id=1666, type=exp]; +"1667 mul_27" [id=1667, type=mul]; +"1668 add_45" [id=1668, type=add]; +"1669 new_zeros_6" [id=1669, type=new_zeros]; +"1670 view_73" [id=1670, type=view]; +"1671 permute_62" [id=1671, type=permute]; +"1672 reshape_60" [id=1672, type=reshape]; +"1673 unsqueeze_38" [id=1673, type=unsqueeze]; +"1674 unsqueeze_39" [id=1674, type=unsqueeze]; +"1675 sub_6" [id=1675, type=sub]; +"1676 ne_6" [id=1676, type=ne]; +"1677 masked_fill_12" [id=1677, type=masked_fill]; +"1678 eq_6" [id=1678, type=eq]; +"1679 masked_fill_13" [id=1679, type=masked_fill]; +"1680 view_74" [id=1680, type=view]; +"1681 unsqueeze_40" [id=1681, type=unsqueeze]; +"1682 unsqueeze_41" [id=1682, type=unsqueeze]; +"1683 add_46" [id=1683, type=add]; +"1684 view_75" [id=1684, type=view]; +"1685 softmax_13" [id=1685, type=softmax]; +"1686 dropout_52" [id=1686, type=dropout]; +"1687 matmul_27" [id=1687, type=matmul]; +"1688 transpose_27" [id=1688, type=transpose]; +"1689 reshape_61" [id=1689, type=reshape]; +"1690 features_5_9_attn_proj_weight_updated_constant0" [id=1690, type=get_attr]; +"1691 reshape_61_0_0_nncf_smooth_quant_0" [id=1691, type=call_module]; +"1692 linear_83" [id=1692, type=linear]; +"1693 dropout_53" [id=1693, type=dropout]; +"1694 view_76" [id=1694, type=view]; +"1695 permute_63" [id=1695, type=permute]; +"1696 reshape_62" [id=1696, type=reshape]; +"1697 roll_13" [id=1697, type=roll]; +"1698 slice_213" [id=1698, type=slice]; +"1699 slice_214" [id=1699, type=slice]; +"1700 slice_215" [id=1700, type=slice]; +"1701 slice_216" [id=1701, type=slice]; +"1702 contiguous_25" [id=1702, type=contiguous]; +"1703 layer_norm_29" [id=1703, type=layer_norm]; +"1704 add_47" [id=1704, type=add]; +"1705 add_47_0_0_nncf_smooth_quant_0" [id=1705, type=call_module]; +"1706 quantize_per_tensor_default_57" [id=1706, type=quantize_per_tensor]; +"1707 dequantize_per_tensor_default_57" [id=1707, type=dequantize_per_tensor]; +"1708 linear_84_scale_0" [id=1708, type=get_attr]; +"1709 linear_84_zero_point_0" [id=1709, type=get_attr]; +"1710 _frozen_param71" [id=1710, type=get_attr]; +"1711 dequantize_per_channel_default_57" [id=1711, type=dequantize_per_channel]; +"1712 linear_84" [id=1712, type=linear]; +"1713 gelu_13" [id=1713, type=gelu]; +"1714 dropout_54" [id=1714, type=dropout]; +"1715 dropout_54_0_0_nncf_smooth_quant_0" [id=1715, type=call_module]; +"1716 quantize_per_tensor_default_58" [id=1716, type=quantize_per_tensor]; +"1717 dequantize_per_tensor_default_58" [id=1717, type=dequantize_per_tensor]; +"1718 linear_85_scale_0" [id=1718, type=get_attr]; +"1719 linear_85_zero_point_0" [id=1719, type=get_attr]; +"1720 _frozen_param72" [id=1720, type=get_attr]; +"1721 dequantize_per_channel_default_58" [id=1721, type=dequantize_per_channel]; +"1722 linear_85" [id=1722, type=linear]; +"1723 dropout_55" [id=1723, type=dropout]; +"1724 layer_norm_30" [id=1724, type=layer_norm]; +"1725 add_48" [id=1725, type=add]; +"1726 _frozen_param73" [id=1726, type=get_attr]; +"1727 dequantize_per_tensor_default_59" [id=1727, type=dequantize_per_tensor]; +"1728 linear_86_scale_0" [id=1728, type=get_attr]; +"1729 linear_86_zero_point_0" [id=1729, type=get_attr]; +"1730 _frozen_param74" [id=1730, type=get_attr]; +"1731 dequantize_per_channel_default_59" [id=1731, type=dequantize_per_channel]; +"1732 linear_86" [id=1732, type=linear]; +"1733 relu__14" [id=1733, type=relu_]; +"1734 relu__14_0_0_nncf_smooth_quant_0" [id=1734, type=call_module]; +"1735 quantize_per_tensor_default_60" [id=1735, type=quantize_per_tensor]; +"1736 dequantize_per_tensor_default_60" [id=1736, type=dequantize_per_tensor]; +"1737 linear_87_scale_0" [id=1737, type=get_attr]; +"1738 linear_87_zero_point_0" [id=1738, type=get_attr]; +"1739 _frozen_param75" [id=1739, type=get_attr]; +"1740 dequantize_per_channel_default_60" [id=1740, type=dequantize_per_channel]; +"1741 linear_87" [id=1741, type=linear]; +"1742 view_77" [id=1742, type=view]; +"1743 index_14" [id=1743, type=index]; +"1744 view_78" [id=1744, type=view]; +"1745 permute_64" [id=1745, type=permute]; +"1746 contiguous_26" [id=1746, type=contiguous]; +"1747 unsqueeze_42" [id=1747, type=unsqueeze]; +"1748 sigmoid_14" [id=1748, type=sigmoid]; +"1749 mul_28" [id=1749, type=mul]; +"1750 pad_16" [id=1750, type=pad]; +"1751 view_79" [id=1751, type=view]; +"1752 permute_65" [id=1752, type=permute]; +"1753 reshape_63" [id=1753, type=reshape]; +"1754 clone_14" [id=1754, type=clone]; +"1755 features_5_10_attn_qkv_weight_updated_constant0" [id=1755, type=get_attr]; +"1756 reshape_63_0_0_nncf_smooth_quant_0" [id=1756, type=call_module]; +"1757 linear_88" [id=1757, type=linear]; +"1758 reshape_64" [id=1758, type=reshape]; +"1759 permute_66" [id=1759, type=permute]; +"1760 select_42" [id=1760, type=select]; +"1761 select_43" [id=1761, type=select]; +"1762 select_44" [id=1762, type=select]; +"1763 linalg_vector_norm_28" [id=1763, type=linalg_vector_norm]; +"1764 clamp_min_28" [id=1764, type=clamp_min]; +"1765 expand_as_28" [id=1765, type=expand_as]; +"1766 div_28" [id=1766, type=div]; +"1767 linalg_vector_norm_29" [id=1767, type=linalg_vector_norm]; +"1768 clamp_min_29" [id=1768, type=clamp_min]; +"1769 expand_as_29" [id=1769, type=expand_as]; +"1770 div_29" [id=1770, type=div]; +"1771 transpose_28" [id=1771, type=transpose]; +"1772 matmul_28" [id=1772, type=matmul]; +"1773 clamp_14" [id=1773, type=clamp]; +"1774 exp_14" [id=1774, type=exp]; +"1775 mul_29" [id=1775, type=mul]; +"1776 add_49" [id=1776, type=add]; +"1777 softmax_14" [id=1777, type=softmax]; +"1778 dropout_56" [id=1778, type=dropout]; +"1779 matmul_29" [id=1779, type=matmul]; +"1780 transpose_29" [id=1780, type=transpose]; +"1781 reshape_65" [id=1781, type=reshape]; +"1782 features_5_10_attn_proj_weight_updated_constant0" [id=1782, type=get_attr]; +"1783 reshape_65_0_0_nncf_smooth_quant_0" [id=1783, type=call_module]; +"1784 linear_89" [id=1784, type=linear]; +"1785 dropout_57" [id=1785, type=dropout]; +"1786 view_80" [id=1786, type=view]; +"1787 permute_67" [id=1787, type=permute]; +"1788 reshape_66" [id=1788, type=reshape]; +"1789 slice_218" [id=1789, type=slice]; +"1790 slice_219" [id=1790, type=slice]; +"1791 slice_220" [id=1791, type=slice]; +"1792 slice_221" [id=1792, type=slice]; +"1793 contiguous_27" [id=1793, type=contiguous]; +"1794 layer_norm_31" [id=1794, type=layer_norm]; +"1795 add_50" [id=1795, type=add]; +"1796 add_50_0_0_nncf_smooth_quant_0" [id=1796, type=call_module]; +"1797 quantize_per_tensor_default_61" [id=1797, type=quantize_per_tensor]; +"1798 dequantize_per_tensor_default_61" [id=1798, type=dequantize_per_tensor]; +"1799 linear_90_scale_0" [id=1799, type=get_attr]; +"1800 linear_90_zero_point_0" [id=1800, type=get_attr]; +"1801 _frozen_param76" [id=1801, type=get_attr]; +"1802 dequantize_per_channel_default_61" [id=1802, type=dequantize_per_channel]; +"1803 linear_90" [id=1803, type=linear]; +"1804 gelu_14" [id=1804, type=gelu]; +"1805 dropout_58" [id=1805, type=dropout]; +"1806 dropout_58_0_0_nncf_smooth_quant_0" [id=1806, type=call_module]; +"1807 quantize_per_tensor_default_62" [id=1807, type=quantize_per_tensor]; +"1808 dequantize_per_tensor_default_62" [id=1808, type=dequantize_per_tensor]; +"1809 linear_91_scale_0" [id=1809, type=get_attr]; +"1810 linear_91_zero_point_0" [id=1810, type=get_attr]; +"1811 _frozen_param77" [id=1811, type=get_attr]; +"1812 dequantize_per_channel_default_62" [id=1812, type=dequantize_per_channel]; +"1813 linear_91" [id=1813, type=linear]; +"1814 dropout_59" [id=1814, type=dropout]; +"1815 layer_norm_32" [id=1815, type=layer_norm]; +"1816 add_51" [id=1816, type=add]; +"1817 _frozen_param78" [id=1817, type=get_attr]; +"1818 dequantize_per_tensor_default_63" [id=1818, type=dequantize_per_tensor]; +"1819 linear_92_scale_0" [id=1819, type=get_attr]; +"1820 linear_92_zero_point_0" [id=1820, type=get_attr]; +"1821 _frozen_param79" [id=1821, type=get_attr]; +"1822 dequantize_per_channel_default_63" [id=1822, type=dequantize_per_channel]; +"1823 linear_92" [id=1823, type=linear]; +"1824 relu__15" [id=1824, type=relu_]; +"1825 relu__15_0_0_nncf_smooth_quant_0" [id=1825, type=call_module]; +"1826 quantize_per_tensor_default_64" [id=1826, type=quantize_per_tensor]; +"1827 dequantize_per_tensor_default_64" [id=1827, type=dequantize_per_tensor]; +"1828 linear_93_scale_0" [id=1828, type=get_attr]; +"1829 linear_93_zero_point_0" [id=1829, type=get_attr]; +"1830 _frozen_param80" [id=1830, type=get_attr]; +"1831 dequantize_per_channel_default_64" [id=1831, type=dequantize_per_channel]; +"1832 linear_93" [id=1832, type=linear]; +"1833 view_81" [id=1833, type=view]; +"1834 index_15" [id=1834, type=index]; +"1835 view_82" [id=1835, type=view]; +"1836 permute_68" [id=1836, type=permute]; +"1837 contiguous_28" [id=1837, type=contiguous]; +"1838 unsqueeze_43" [id=1838, type=unsqueeze]; +"1839 sigmoid_15" [id=1839, type=sigmoid]; +"1840 mul_30" [id=1840, type=mul]; +"1841 pad_17" [id=1841, type=pad]; +"1842 roll_14" [id=1842, type=roll]; +"1843 view_83" [id=1843, type=view]; +"1844 permute_69" [id=1844, type=permute]; +"1845 reshape_67" [id=1845, type=reshape]; +"1846 clone_15" [id=1846, type=clone]; +"1847 features_5_11_attn_qkv_weight_updated_constant0" [id=1847, type=get_attr]; +"1848 reshape_67_0_0_nncf_smooth_quant_0" [id=1848, type=call_module]; +"1849 linear_94" [id=1849, type=linear]; +"1850 reshape_68" [id=1850, type=reshape]; +"1851 permute_70" [id=1851, type=permute]; +"1852 select_45" [id=1852, type=select]; +"1853 select_46" [id=1853, type=select]; +"1854 select_47" [id=1854, type=select]; +"1855 linalg_vector_norm_30" [id=1855, type=linalg_vector_norm]; +"1856 clamp_min_30" [id=1856, type=clamp_min]; +"1857 expand_as_30" [id=1857, type=expand_as]; +"1858 div_30" [id=1858, type=div]; +"1859 linalg_vector_norm_31" [id=1859, type=linalg_vector_norm]; +"1860 clamp_min_31" [id=1860, type=clamp_min]; +"1861 expand_as_31" [id=1861, type=expand_as]; +"1862 div_31" [id=1862, type=div]; +"1863 transpose_30" [id=1863, type=transpose]; +"1864 matmul_30" [id=1864, type=matmul]; +"1865 clamp_15" [id=1865, type=clamp]; +"1866 exp_15" [id=1866, type=exp]; +"1867 mul_31" [id=1867, type=mul]; +"1868 add_52" [id=1868, type=add]; +"1869 new_zeros_7" [id=1869, type=new_zeros]; +"1870 view_84" [id=1870, type=view]; +"1871 permute_71" [id=1871, type=permute]; +"1872 reshape_69" [id=1872, type=reshape]; +"1873 unsqueeze_44" [id=1873, type=unsqueeze]; +"1874 unsqueeze_45" [id=1874, type=unsqueeze]; +"1875 sub_7" [id=1875, type=sub]; +"1876 ne_7" [id=1876, type=ne]; +"1877 masked_fill_14" [id=1877, type=masked_fill]; +"1878 eq_7" [id=1878, type=eq]; +"1879 masked_fill_15" [id=1879, type=masked_fill]; +"1880 view_85" [id=1880, type=view]; +"1881 unsqueeze_46" [id=1881, type=unsqueeze]; +"1882 unsqueeze_47" [id=1882, type=unsqueeze]; +"1883 add_53" [id=1883, type=add]; +"1884 view_86" [id=1884, type=view]; +"1885 softmax_15" [id=1885, type=softmax]; +"1886 dropout_60" [id=1886, type=dropout]; +"1887 matmul_31" [id=1887, type=matmul]; +"1888 transpose_31" [id=1888, type=transpose]; +"1889 reshape_70" [id=1889, type=reshape]; +"1890 features_5_11_attn_proj_weight_updated_constant0" [id=1890, type=get_attr]; +"1891 reshape_70_0_0_nncf_smooth_quant_0" [id=1891, type=call_module]; +"1892 linear_95" [id=1892, type=linear]; +"1893 dropout_61" [id=1893, type=dropout]; +"1894 view_87" [id=1894, type=view]; +"1895 permute_72" [id=1895, type=permute]; +"1896 reshape_71" [id=1896, type=reshape]; +"1897 roll_15" [id=1897, type=roll]; +"1898 slice_241" [id=1898, type=slice]; +"1899 slice_242" [id=1899, type=slice]; +"1900 slice_243" [id=1900, type=slice]; +"1901 slice_244" [id=1901, type=slice]; +"1902 contiguous_29" [id=1902, type=contiguous]; +"1903 layer_norm_33" [id=1903, type=layer_norm]; +"1904 add_54" [id=1904, type=add]; +"1905 add_54_0_0_nncf_smooth_quant_0" [id=1905, type=call_module]; +"1906 quantize_per_tensor_default_65" [id=1906, type=quantize_per_tensor]; +"1907 dequantize_per_tensor_default_65" [id=1907, type=dequantize_per_tensor]; +"1908 linear_96_scale_0" [id=1908, type=get_attr]; +"1909 linear_96_zero_point_0" [id=1909, type=get_attr]; +"1910 _frozen_param81" [id=1910, type=get_attr]; +"1911 dequantize_per_channel_default_65" [id=1911, type=dequantize_per_channel]; +"1912 linear_96" [id=1912, type=linear]; +"1913 gelu_15" [id=1913, type=gelu]; +"1914 dropout_62" [id=1914, type=dropout]; +"1915 dropout_62_0_0_nncf_smooth_quant_0" [id=1915, type=call_module]; +"1916 quantize_per_tensor_default_66" [id=1916, type=quantize_per_tensor]; +"1917 dequantize_per_tensor_default_66" [id=1917, type=dequantize_per_tensor]; +"1918 linear_97_scale_0" [id=1918, type=get_attr]; +"1919 linear_97_zero_point_0" [id=1919, type=get_attr]; +"1920 _frozen_param82" [id=1920, type=get_attr]; +"1921 dequantize_per_channel_default_66" [id=1921, type=dequantize_per_channel]; +"1922 linear_97" [id=1922, type=linear]; +"1923 dropout_63" [id=1923, type=dropout]; +"1924 layer_norm_34" [id=1924, type=layer_norm]; +"1925 add_55" [id=1925, type=add]; +"1926 _frozen_param83" [id=1926, type=get_attr]; +"1927 dequantize_per_tensor_default_67" [id=1927, type=dequantize_per_tensor]; +"1928 linear_98_scale_0" [id=1928, type=get_attr]; +"1929 linear_98_zero_point_0" [id=1929, type=get_attr]; +"1930 _frozen_param84" [id=1930, type=get_attr]; +"1931 dequantize_per_channel_default_67" [id=1931, type=dequantize_per_channel]; +"1932 linear_98" [id=1932, type=linear]; +"1933 relu__16" [id=1933, type=relu_]; +"1934 relu__16_0_0_nncf_smooth_quant_0" [id=1934, type=call_module]; +"1935 quantize_per_tensor_default_68" [id=1935, type=quantize_per_tensor]; +"1936 dequantize_per_tensor_default_68" [id=1936, type=dequantize_per_tensor]; +"1937 linear_99_scale_0" [id=1937, type=get_attr]; +"1938 linear_99_zero_point_0" [id=1938, type=get_attr]; +"1939 _frozen_param85" [id=1939, type=get_attr]; +"1940 dequantize_per_channel_default_68" [id=1940, type=dequantize_per_channel]; +"1941 linear_99" [id=1941, type=linear]; +"1942 view_88" [id=1942, type=view]; +"1943 index_16" [id=1943, type=index]; +"1944 view_89" [id=1944, type=view]; +"1945 permute_73" [id=1945, type=permute]; +"1946 contiguous_30" [id=1946, type=contiguous]; +"1947 unsqueeze_48" [id=1947, type=unsqueeze]; +"1948 sigmoid_16" [id=1948, type=sigmoid]; +"1949 mul_32" [id=1949, type=mul]; +"1950 pad_18" [id=1950, type=pad]; +"1951 view_90" [id=1951, type=view]; +"1952 permute_74" [id=1952, type=permute]; +"1953 reshape_72" [id=1953, type=reshape]; +"1954 clone_16" [id=1954, type=clone]; +"1955 features_5_12_attn_qkv_weight_updated_constant0" [id=1955, type=get_attr]; +"1956 reshape_72_0_0_nncf_smooth_quant_0" [id=1956, type=call_module]; +"1957 linear_100" [id=1957, type=linear]; +"1958 reshape_73" [id=1958, type=reshape]; +"1959 permute_75" [id=1959, type=permute]; +"1960 select_48" [id=1960, type=select]; +"1961 select_49" [id=1961, type=select]; +"1962 select_50" [id=1962, type=select]; +"1963 linalg_vector_norm_32" [id=1963, type=linalg_vector_norm]; +"1964 clamp_min_32" [id=1964, type=clamp_min]; +"1965 expand_as_32" [id=1965, type=expand_as]; +"1966 div_32" [id=1966, type=div]; +"1967 linalg_vector_norm_33" [id=1967, type=linalg_vector_norm]; +"1968 clamp_min_33" [id=1968, type=clamp_min]; +"1969 expand_as_33" [id=1969, type=expand_as]; +"1970 div_33" [id=1970, type=div]; +"1971 transpose_32" [id=1971, type=transpose]; +"1972 matmul_32" [id=1972, type=matmul]; +"1973 clamp_16" [id=1973, type=clamp]; +"1974 exp_16" [id=1974, type=exp]; +"1975 mul_33" [id=1975, type=mul]; +"1976 add_56" [id=1976, type=add]; +"1977 softmax_16" [id=1977, type=softmax]; +"1978 dropout_64" [id=1978, type=dropout]; +"1979 matmul_33" [id=1979, type=matmul]; +"1980 transpose_33" [id=1980, type=transpose]; +"1981 reshape_74" [id=1981, type=reshape]; +"1982 features_5_12_attn_proj_weight_updated_constant0" [id=1982, type=get_attr]; +"1983 reshape_74_0_0_nncf_smooth_quant_0" [id=1983, type=call_module]; +"1984 linear_101" [id=1984, type=linear]; +"1985 dropout_65" [id=1985, type=dropout]; +"1986 view_91" [id=1986, type=view]; +"1987 permute_76" [id=1987, type=permute]; +"1988 reshape_75" [id=1988, type=reshape]; +"1989 slice_246" [id=1989, type=slice]; +"1990 slice_247" [id=1990, type=slice]; +"1991 slice_248" [id=1991, type=slice]; +"1992 slice_249" [id=1992, type=slice]; +"1993 contiguous_31" [id=1993, type=contiguous]; +"1994 layer_norm_35" [id=1994, type=layer_norm]; +"1995 add_57" [id=1995, type=add]; +"1996 add_57_0_0_nncf_smooth_quant_0" [id=1996, type=call_module]; +"1997 quantize_per_tensor_default_69" [id=1997, type=quantize_per_tensor]; +"1998 dequantize_per_tensor_default_69" [id=1998, type=dequantize_per_tensor]; +"1999 linear_102_scale_0" [id=1999, type=get_attr]; +"2000 linear_102_zero_point_0" [id=2000, type=get_attr]; +"2001 _frozen_param86" [id=2001, type=get_attr]; +"2002 dequantize_per_channel_default_69" [id=2002, type=dequantize_per_channel]; +"2003 linear_102" [id=2003, type=linear]; +"2004 gelu_16" [id=2004, type=gelu]; +"2005 dropout_66" [id=2005, type=dropout]; +"2006 dropout_66_0_0_nncf_smooth_quant_0" [id=2006, type=call_module]; +"2007 quantize_per_tensor_default_70" [id=2007, type=quantize_per_tensor]; +"2008 dequantize_per_tensor_default_70" [id=2008, type=dequantize_per_tensor]; +"2009 linear_103_scale_0" [id=2009, type=get_attr]; +"2010 linear_103_zero_point_0" [id=2010, type=get_attr]; +"2011 _frozen_param87" [id=2011, type=get_attr]; +"2012 dequantize_per_channel_default_70" [id=2012, type=dequantize_per_channel]; +"2013 linear_103" [id=2013, type=linear]; +"2014 dropout_67" [id=2014, type=dropout]; +"2015 layer_norm_36" [id=2015, type=layer_norm]; +"2016 add_58" [id=2016, type=add]; +"2017 _frozen_param88" [id=2017, type=get_attr]; +"2018 dequantize_per_tensor_default_71" [id=2018, type=dequantize_per_tensor]; +"2019 linear_104_scale_0" [id=2019, type=get_attr]; +"2020 linear_104_zero_point_0" [id=2020, type=get_attr]; +"2021 _frozen_param89" [id=2021, type=get_attr]; +"2022 dequantize_per_channel_default_71" [id=2022, type=dequantize_per_channel]; +"2023 linear_104" [id=2023, type=linear]; +"2024 relu__17" [id=2024, type=relu_]; +"2025 relu__17_0_0_nncf_smooth_quant_0" [id=2025, type=call_module]; +"2026 quantize_per_tensor_default_72" [id=2026, type=quantize_per_tensor]; +"2027 dequantize_per_tensor_default_72" [id=2027, type=dequantize_per_tensor]; +"2028 linear_105_scale_0" [id=2028, type=get_attr]; +"2029 linear_105_zero_point_0" [id=2029, type=get_attr]; +"2030 _frozen_param90" [id=2030, type=get_attr]; +"2031 dequantize_per_channel_default_72" [id=2031, type=dequantize_per_channel]; +"2032 linear_105" [id=2032, type=linear]; +"2033 view_92" [id=2033, type=view]; +"2034 index_17" [id=2034, type=index]; +"2035 view_93" [id=2035, type=view]; +"2036 permute_77" [id=2036, type=permute]; +"2037 contiguous_32" [id=2037, type=contiguous]; +"2038 unsqueeze_49" [id=2038, type=unsqueeze]; +"2039 sigmoid_17" [id=2039, type=sigmoid]; +"2040 mul_34" [id=2040, type=mul]; +"2041 pad_19" [id=2041, type=pad]; +"2042 roll_16" [id=2042, type=roll]; +"2043 view_94" [id=2043, type=view]; +"2044 permute_78" [id=2044, type=permute]; +"2045 reshape_76" [id=2045, type=reshape]; +"2046 clone_17" [id=2046, type=clone]; +"2047 features_5_13_attn_qkv_weight_updated_constant0" [id=2047, type=get_attr]; +"2048 reshape_76_0_0_nncf_smooth_quant_0" [id=2048, type=call_module]; +"2049 linear_106" [id=2049, type=linear]; +"2050 reshape_77" [id=2050, type=reshape]; +"2051 permute_79" [id=2051, type=permute]; +"2052 select_51" [id=2052, type=select]; +"2053 select_52" [id=2053, type=select]; +"2054 select_53" [id=2054, type=select]; +"2055 linalg_vector_norm_34" [id=2055, type=linalg_vector_norm]; +"2056 clamp_min_34" [id=2056, type=clamp_min]; +"2057 expand_as_34" [id=2057, type=expand_as]; +"2058 div_34" [id=2058, type=div]; +"2059 linalg_vector_norm_35" [id=2059, type=linalg_vector_norm]; +"2060 clamp_min_35" [id=2060, type=clamp_min]; +"2061 expand_as_35" [id=2061, type=expand_as]; +"2062 div_35" [id=2062, type=div]; +"2063 transpose_34" [id=2063, type=transpose]; +"2064 matmul_34" [id=2064, type=matmul]; +"2065 clamp_17" [id=2065, type=clamp]; +"2066 exp_17" [id=2066, type=exp]; +"2067 mul_35" [id=2067, type=mul]; +"2068 add_59" [id=2068, type=add]; +"2069 new_zeros_8" [id=2069, type=new_zeros]; +"2070 view_95" [id=2070, type=view]; +"2071 permute_80" [id=2071, type=permute]; +"2072 reshape_78" [id=2072, type=reshape]; +"2073 unsqueeze_50" [id=2073, type=unsqueeze]; +"2074 unsqueeze_51" [id=2074, type=unsqueeze]; +"2075 sub_8" [id=2075, type=sub]; +"2076 ne_8" [id=2076, type=ne]; +"2077 masked_fill_16" [id=2077, type=masked_fill]; +"2078 eq_8" [id=2078, type=eq]; +"2079 masked_fill_17" [id=2079, type=masked_fill]; +"2080 view_96" [id=2080, type=view]; +"2081 unsqueeze_52" [id=2081, type=unsqueeze]; +"2082 unsqueeze_53" [id=2082, type=unsqueeze]; +"2083 add_60" [id=2083, type=add]; +"2084 view_97" [id=2084, type=view]; +"2085 softmax_17" [id=2085, type=softmax]; +"2086 dropout_68" [id=2086, type=dropout]; +"2087 matmul_35" [id=2087, type=matmul]; +"2088 transpose_35" [id=2088, type=transpose]; +"2089 reshape_79" [id=2089, type=reshape]; +"2090 features_5_13_attn_proj_weight_updated_constant0" [id=2090, type=get_attr]; +"2091 reshape_79_0_0_nncf_smooth_quant_0" [id=2091, type=call_module]; +"2092 linear_107" [id=2092, type=linear]; +"2093 dropout_69" [id=2093, type=dropout]; +"2094 view_98" [id=2094, type=view]; +"2095 permute_81" [id=2095, type=permute]; +"2096 reshape_80" [id=2096, type=reshape]; +"2097 roll_17" [id=2097, type=roll]; +"2098 slice_269" [id=2098, type=slice]; +"2099 slice_270" [id=2099, type=slice]; +"2100 slice_271" [id=2100, type=slice]; +"2101 slice_272" [id=2101, type=slice]; +"2102 contiguous_33" [id=2102, type=contiguous]; +"2103 layer_norm_37" [id=2103, type=layer_norm]; +"2104 add_61" [id=2104, type=add]; +"2105 add_61_0_0_nncf_smooth_quant_0" [id=2105, type=call_module]; +"2106 quantize_per_tensor_default_73" [id=2106, type=quantize_per_tensor]; +"2107 dequantize_per_tensor_default_73" [id=2107, type=dequantize_per_tensor]; +"2108 linear_108_scale_0" [id=2108, type=get_attr]; +"2109 linear_108_zero_point_0" [id=2109, type=get_attr]; +"2110 _frozen_param91" [id=2110, type=get_attr]; +"2111 dequantize_per_channel_default_73" [id=2111, type=dequantize_per_channel]; +"2112 linear_108" [id=2112, type=linear]; +"2113 gelu_17" [id=2113, type=gelu]; +"2114 dropout_70" [id=2114, type=dropout]; +"2115 dropout_70_0_0_nncf_smooth_quant_0" [id=2115, type=call_module]; +"2116 quantize_per_tensor_default_74" [id=2116, type=quantize_per_tensor]; +"2117 dequantize_per_tensor_default_74" [id=2117, type=dequantize_per_tensor]; +"2118 linear_109_scale_0" [id=2118, type=get_attr]; +"2119 linear_109_zero_point_0" [id=2119, type=get_attr]; +"2120 _frozen_param92" [id=2120, type=get_attr]; +"2121 dequantize_per_channel_default_74" [id=2121, type=dequantize_per_channel]; +"2122 linear_109" [id=2122, type=linear]; +"2123 dropout_71" [id=2123, type=dropout]; +"2124 layer_norm_38" [id=2124, type=layer_norm]; +"2125 add_62" [id=2125, type=add]; +"2126 _frozen_param93" [id=2126, type=get_attr]; +"2127 dequantize_per_tensor_default_75" [id=2127, type=dequantize_per_tensor]; +"2128 linear_110_scale_0" [id=2128, type=get_attr]; +"2129 linear_110_zero_point_0" [id=2129, type=get_attr]; +"2130 _frozen_param94" [id=2130, type=get_attr]; +"2131 dequantize_per_channel_default_75" [id=2131, type=dequantize_per_channel]; +"2132 linear_110" [id=2132, type=linear]; +"2133 relu__18" [id=2133, type=relu_]; +"2134 relu__18_0_0_nncf_smooth_quant_0" [id=2134, type=call_module]; +"2135 quantize_per_tensor_default_76" [id=2135, type=quantize_per_tensor]; +"2136 dequantize_per_tensor_default_76" [id=2136, type=dequantize_per_tensor]; +"2137 linear_111_scale_0" [id=2137, type=get_attr]; +"2138 linear_111_zero_point_0" [id=2138, type=get_attr]; +"2139 _frozen_param95" [id=2139, type=get_attr]; +"2140 dequantize_per_channel_default_76" [id=2140, type=dequantize_per_channel]; +"2141 linear_111" [id=2141, type=linear]; +"2142 view_99" [id=2142, type=view]; +"2143 index_18" [id=2143, type=index]; +"2144 view_100" [id=2144, type=view]; +"2145 permute_82" [id=2145, type=permute]; +"2146 contiguous_34" [id=2146, type=contiguous]; +"2147 unsqueeze_54" [id=2147, type=unsqueeze]; +"2148 sigmoid_18" [id=2148, type=sigmoid]; +"2149 mul_36" [id=2149, type=mul]; +"2150 pad_20" [id=2150, type=pad]; +"2151 view_101" [id=2151, type=view]; +"2152 permute_83" [id=2152, type=permute]; +"2153 reshape_81" [id=2153, type=reshape]; +"2154 clone_18" [id=2154, type=clone]; +"2155 features_5_14_attn_qkv_weight_updated_constant0" [id=2155, type=get_attr]; +"2156 reshape_81_0_0_nncf_smooth_quant_0" [id=2156, type=call_module]; +"2157 linear_112" [id=2157, type=linear]; +"2158 reshape_82" [id=2158, type=reshape]; +"2159 permute_84" [id=2159, type=permute]; +"2160 select_54" [id=2160, type=select]; +"2161 select_55" [id=2161, type=select]; +"2162 select_56" [id=2162, type=select]; +"2163 linalg_vector_norm_36" [id=2163, type=linalg_vector_norm]; +"2164 clamp_min_36" [id=2164, type=clamp_min]; +"2165 expand_as_36" [id=2165, type=expand_as]; +"2166 div_36" [id=2166, type=div]; +"2167 linalg_vector_norm_37" [id=2167, type=linalg_vector_norm]; +"2168 clamp_min_37" [id=2168, type=clamp_min]; +"2169 expand_as_37" [id=2169, type=expand_as]; +"2170 div_37" [id=2170, type=div]; +"2171 transpose_36" [id=2171, type=transpose]; +"2172 matmul_36" [id=2172, type=matmul]; +"2173 clamp_18" [id=2173, type=clamp]; +"2174 exp_18" [id=2174, type=exp]; +"2175 mul_37" [id=2175, type=mul]; +"2176 add_63" [id=2176, type=add]; +"2177 softmax_18" [id=2177, type=softmax]; +"2178 dropout_72" [id=2178, type=dropout]; +"2179 matmul_37" [id=2179, type=matmul]; +"2180 transpose_37" [id=2180, type=transpose]; +"2181 reshape_83" [id=2181, type=reshape]; +"2182 features_5_14_attn_proj_weight_updated_constant0" [id=2182, type=get_attr]; +"2183 reshape_83_0_0_nncf_smooth_quant_0" [id=2183, type=call_module]; +"2184 linear_113" [id=2184, type=linear]; +"2185 dropout_73" [id=2185, type=dropout]; +"2186 view_102" [id=2186, type=view]; +"2187 permute_85" [id=2187, type=permute]; +"2188 reshape_84" [id=2188, type=reshape]; +"2189 slice_274" [id=2189, type=slice]; +"2190 slice_275" [id=2190, type=slice]; +"2191 slice_276" [id=2191, type=slice]; +"2192 slice_277" [id=2192, type=slice]; +"2193 contiguous_35" [id=2193, type=contiguous]; +"2194 layer_norm_39" [id=2194, type=layer_norm]; +"2195 add_64" [id=2195, type=add]; +"2196 add_64_0_0_nncf_smooth_quant_0" [id=2196, type=call_module]; +"2197 quantize_per_tensor_default_77" [id=2197, type=quantize_per_tensor]; +"2198 dequantize_per_tensor_default_77" [id=2198, type=dequantize_per_tensor]; +"2199 linear_114_scale_0" [id=2199, type=get_attr]; +"2200 linear_114_zero_point_0" [id=2200, type=get_attr]; +"2201 _frozen_param96" [id=2201, type=get_attr]; +"2202 dequantize_per_channel_default_77" [id=2202, type=dequantize_per_channel]; +"2203 linear_114" [id=2203, type=linear]; +"2204 gelu_18" [id=2204, type=gelu]; +"2205 dropout_74" [id=2205, type=dropout]; +"2206 dropout_74_0_0_nncf_smooth_quant_0" [id=2206, type=call_module]; +"2207 quantize_per_tensor_default_78" [id=2207, type=quantize_per_tensor]; +"2208 dequantize_per_tensor_default_78" [id=2208, type=dequantize_per_tensor]; +"2209 linear_115_scale_0" [id=2209, type=get_attr]; +"2210 linear_115_zero_point_0" [id=2210, type=get_attr]; +"2211 _frozen_param97" [id=2211, type=get_attr]; +"2212 dequantize_per_channel_default_78" [id=2212, type=dequantize_per_channel]; +"2213 linear_115" [id=2213, type=linear]; +"2214 dropout_75" [id=2214, type=dropout]; +"2215 layer_norm_40" [id=2215, type=layer_norm]; +"2216 add_65" [id=2216, type=add]; +"2217 _frozen_param98" [id=2217, type=get_attr]; +"2218 dequantize_per_tensor_default_79" [id=2218, type=dequantize_per_tensor]; +"2219 linear_116_scale_0" [id=2219, type=get_attr]; +"2220 linear_116_zero_point_0" [id=2220, type=get_attr]; +"2221 _frozen_param99" [id=2221, type=get_attr]; +"2222 dequantize_per_channel_default_79" [id=2222, type=dequantize_per_channel]; +"2223 linear_116" [id=2223, type=linear]; +"2224 relu__19" [id=2224, type=relu_]; +"2225 relu__19_0_0_nncf_smooth_quant_0" [id=2225, type=call_module]; +"2226 quantize_per_tensor_default_80" [id=2226, type=quantize_per_tensor]; +"2227 dequantize_per_tensor_default_80" [id=2227, type=dequantize_per_tensor]; +"2228 linear_117_scale_0" [id=2228, type=get_attr]; +"2229 linear_117_zero_point_0" [id=2229, type=get_attr]; +"2230 _frozen_param100" [id=2230, type=get_attr]; +"2231 dequantize_per_channel_default_80" [id=2231, type=dequantize_per_channel]; +"2232 linear_117" [id=2232, type=linear]; +"2233 view_103" [id=2233, type=view]; +"2234 index_19" [id=2234, type=index]; +"2235 view_104" [id=2235, type=view]; +"2236 permute_86" [id=2236, type=permute]; +"2237 contiguous_36" [id=2237, type=contiguous]; +"2238 unsqueeze_55" [id=2238, type=unsqueeze]; +"2239 sigmoid_19" [id=2239, type=sigmoid]; +"2240 mul_38" [id=2240, type=mul]; +"2241 pad_21" [id=2241, type=pad]; +"2242 roll_18" [id=2242, type=roll]; +"2243 view_105" [id=2243, type=view]; +"2244 permute_87" [id=2244, type=permute]; +"2245 reshape_85" [id=2245, type=reshape]; +"2246 clone_19" [id=2246, type=clone]; +"2247 features_5_15_attn_qkv_weight_updated_constant0" [id=2247, type=get_attr]; +"2248 reshape_85_0_0_nncf_smooth_quant_0" [id=2248, type=call_module]; +"2249 linear_118" [id=2249, type=linear]; +"2250 reshape_86" [id=2250, type=reshape]; +"2251 permute_88" [id=2251, type=permute]; +"2252 select_57" [id=2252, type=select]; +"2253 select_58" [id=2253, type=select]; +"2254 select_59" [id=2254, type=select]; +"2255 linalg_vector_norm_38" [id=2255, type=linalg_vector_norm]; +"2256 clamp_min_38" [id=2256, type=clamp_min]; +"2257 expand_as_38" [id=2257, type=expand_as]; +"2258 div_38" [id=2258, type=div]; +"2259 linalg_vector_norm_39" [id=2259, type=linalg_vector_norm]; +"2260 clamp_min_39" [id=2260, type=clamp_min]; +"2261 expand_as_39" [id=2261, type=expand_as]; +"2262 div_39" [id=2262, type=div]; +"2263 transpose_38" [id=2263, type=transpose]; +"2264 matmul_38" [id=2264, type=matmul]; +"2265 clamp_19" [id=2265, type=clamp]; +"2266 exp_19" [id=2266, type=exp]; +"2267 mul_39" [id=2267, type=mul]; +"2268 add_66" [id=2268, type=add]; +"2269 new_zeros_9" [id=2269, type=new_zeros]; +"2270 view_106" [id=2270, type=view]; +"2271 permute_89" [id=2271, type=permute]; +"2272 reshape_87" [id=2272, type=reshape]; +"2273 unsqueeze_56" [id=2273, type=unsqueeze]; +"2274 unsqueeze_57" [id=2274, type=unsqueeze]; +"2275 sub_9" [id=2275, type=sub]; +"2276 ne_9" [id=2276, type=ne]; +"2277 masked_fill_18" [id=2277, type=masked_fill]; +"2278 eq_9" [id=2278, type=eq]; +"2279 masked_fill_19" [id=2279, type=masked_fill]; +"2280 view_107" [id=2280, type=view]; +"2281 unsqueeze_58" [id=2281, type=unsqueeze]; +"2282 unsqueeze_59" [id=2282, type=unsqueeze]; +"2283 add_67" [id=2283, type=add]; +"2284 view_108" [id=2284, type=view]; +"2285 softmax_19" [id=2285, type=softmax]; +"2286 dropout_76" [id=2286, type=dropout]; +"2287 matmul_39" [id=2287, type=matmul]; +"2288 transpose_39" [id=2288, type=transpose]; +"2289 reshape_88" [id=2289, type=reshape]; +"2290 features_5_15_attn_proj_weight_updated_constant0" [id=2290, type=get_attr]; +"2291 reshape_88_0_0_nncf_smooth_quant_0" [id=2291, type=call_module]; +"2292 linear_119" [id=2292, type=linear]; +"2293 dropout_77" [id=2293, type=dropout]; +"2294 view_109" [id=2294, type=view]; +"2295 permute_90" [id=2295, type=permute]; +"2296 reshape_89" [id=2296, type=reshape]; +"2297 roll_19" [id=2297, type=roll]; +"2298 slice_297" [id=2298, type=slice]; +"2299 slice_298" [id=2299, type=slice]; +"2300 slice_299" [id=2300, type=slice]; +"2301 slice_300" [id=2301, type=slice]; +"2302 contiguous_37" [id=2302, type=contiguous]; +"2303 layer_norm_41" [id=2303, type=layer_norm]; +"2304 add_68" [id=2304, type=add]; +"2305 add_68_0_0_nncf_smooth_quant_0" [id=2305, type=call_module]; +"2306 quantize_per_tensor_default_81" [id=2306, type=quantize_per_tensor]; +"2307 dequantize_per_tensor_default_81" [id=2307, type=dequantize_per_tensor]; +"2308 linear_120_scale_0" [id=2308, type=get_attr]; +"2309 linear_120_zero_point_0" [id=2309, type=get_attr]; +"2310 _frozen_param101" [id=2310, type=get_attr]; +"2311 dequantize_per_channel_default_81" [id=2311, type=dequantize_per_channel]; +"2312 linear_120" [id=2312, type=linear]; +"2313 gelu_19" [id=2313, type=gelu]; +"2314 dropout_78" [id=2314, type=dropout]; +"2315 dropout_78_0_0_nncf_smooth_quant_0" [id=2315, type=call_module]; +"2316 quantize_per_tensor_default_82" [id=2316, type=quantize_per_tensor]; +"2317 dequantize_per_tensor_default_82" [id=2317, type=dequantize_per_tensor]; +"2318 linear_121_scale_0" [id=2318, type=get_attr]; +"2319 linear_121_zero_point_0" [id=2319, type=get_attr]; +"2320 _frozen_param102" [id=2320, type=get_attr]; +"2321 dequantize_per_channel_default_82" [id=2321, type=dequantize_per_channel]; +"2322 linear_121" [id=2322, type=linear]; +"2323 dropout_79" [id=2323, type=dropout]; +"2324 layer_norm_42" [id=2324, type=layer_norm]; +"2325 add_69" [id=2325, type=add]; +"2326 _frozen_param103" [id=2326, type=get_attr]; +"2327 dequantize_per_tensor_default_83" [id=2327, type=dequantize_per_tensor]; +"2328 linear_122_scale_0" [id=2328, type=get_attr]; +"2329 linear_122_zero_point_0" [id=2329, type=get_attr]; +"2330 _frozen_param104" [id=2330, type=get_attr]; +"2331 dequantize_per_channel_default_83" [id=2331, type=dequantize_per_channel]; +"2332 linear_122" [id=2332, type=linear]; +"2333 relu__20" [id=2333, type=relu_]; +"2334 relu__20_0_0_nncf_smooth_quant_0" [id=2334, type=call_module]; +"2335 quantize_per_tensor_default_84" [id=2335, type=quantize_per_tensor]; +"2336 dequantize_per_tensor_default_84" [id=2336, type=dequantize_per_tensor]; +"2337 linear_123_scale_0" [id=2337, type=get_attr]; +"2338 linear_123_zero_point_0" [id=2338, type=get_attr]; +"2339 _frozen_param105" [id=2339, type=get_attr]; +"2340 dequantize_per_channel_default_84" [id=2340, type=dequantize_per_channel]; +"2341 linear_123" [id=2341, type=linear]; +"2342 view_110" [id=2342, type=view]; +"2343 index_20" [id=2343, type=index]; +"2344 view_111" [id=2344, type=view]; +"2345 permute_91" [id=2345, type=permute]; +"2346 contiguous_38" [id=2346, type=contiguous]; +"2347 unsqueeze_60" [id=2347, type=unsqueeze]; +"2348 sigmoid_20" [id=2348, type=sigmoid]; +"2349 mul_40" [id=2349, type=mul]; +"2350 pad_22" [id=2350, type=pad]; +"2351 view_112" [id=2351, type=view]; +"2352 permute_92" [id=2352, type=permute]; +"2353 reshape_90" [id=2353, type=reshape]; +"2354 clone_20" [id=2354, type=clone]; +"2355 features_5_16_attn_qkv_weight_updated_constant0" [id=2355, type=get_attr]; +"2356 reshape_90_0_0_nncf_smooth_quant_0" [id=2356, type=call_module]; +"2357 linear_124" [id=2357, type=linear]; +"2358 reshape_91" [id=2358, type=reshape]; +"2359 permute_93" [id=2359, type=permute]; +"2360 select_60" [id=2360, type=select]; +"2361 select_61" [id=2361, type=select]; +"2362 select_62" [id=2362, type=select]; +"2363 linalg_vector_norm_40" [id=2363, type=linalg_vector_norm]; +"2364 clamp_min_40" [id=2364, type=clamp_min]; +"2365 expand_as_40" [id=2365, type=expand_as]; +"2366 div_40" [id=2366, type=div]; +"2367 linalg_vector_norm_41" [id=2367, type=linalg_vector_norm]; +"2368 clamp_min_41" [id=2368, type=clamp_min]; +"2369 expand_as_41" [id=2369, type=expand_as]; +"2370 div_41" [id=2370, type=div]; +"2371 transpose_40" [id=2371, type=transpose]; +"2372 matmul_40" [id=2372, type=matmul]; +"2373 clamp_20" [id=2373, type=clamp]; +"2374 exp_20" [id=2374, type=exp]; +"2375 mul_41" [id=2375, type=mul]; +"2376 add_70" [id=2376, type=add]; +"2377 softmax_20" [id=2377, type=softmax]; +"2378 dropout_80" [id=2378, type=dropout]; +"2379 matmul_41" [id=2379, type=matmul]; +"2380 transpose_41" [id=2380, type=transpose]; +"2381 reshape_92" [id=2381, type=reshape]; +"2382 features_5_16_attn_proj_weight_updated_constant0" [id=2382, type=get_attr]; +"2383 reshape_92_0_0_nncf_smooth_quant_0" [id=2383, type=call_module]; +"2384 linear_125" [id=2384, type=linear]; +"2385 dropout_81" [id=2385, type=dropout]; +"2386 view_113" [id=2386, type=view]; +"2387 permute_94" [id=2387, type=permute]; +"2388 reshape_93" [id=2388, type=reshape]; +"2389 slice_302" [id=2389, type=slice]; +"2390 slice_303" [id=2390, type=slice]; +"2391 slice_304" [id=2391, type=slice]; +"2392 slice_305" [id=2392, type=slice]; +"2393 contiguous_39" [id=2393, type=contiguous]; +"2394 layer_norm_43" [id=2394, type=layer_norm]; +"2395 add_71" [id=2395, type=add]; +"2396 add_71_0_0_nncf_smooth_quant_0" [id=2396, type=call_module]; +"2397 quantize_per_tensor_default_85" [id=2397, type=quantize_per_tensor]; +"2398 dequantize_per_tensor_default_85" [id=2398, type=dequantize_per_tensor]; +"2399 linear_126_scale_0" [id=2399, type=get_attr]; +"2400 linear_126_zero_point_0" [id=2400, type=get_attr]; +"2401 _frozen_param106" [id=2401, type=get_attr]; +"2402 dequantize_per_channel_default_85" [id=2402, type=dequantize_per_channel]; +"2403 linear_126" [id=2403, type=linear]; +"2404 gelu_20" [id=2404, type=gelu]; +"2405 dropout_82" [id=2405, type=dropout]; +"2406 dropout_82_0_0_nncf_smooth_quant_0" [id=2406, type=call_module]; +"2407 quantize_per_tensor_default_86" [id=2407, type=quantize_per_tensor]; +"2408 dequantize_per_tensor_default_86" [id=2408, type=dequantize_per_tensor]; +"2409 linear_127_scale_0" [id=2409, type=get_attr]; +"2410 linear_127_zero_point_0" [id=2410, type=get_attr]; +"2411 _frozen_param107" [id=2411, type=get_attr]; +"2412 dequantize_per_channel_default_86" [id=2412, type=dequantize_per_channel]; +"2413 linear_127" [id=2413, type=linear]; +"2414 dropout_83" [id=2414, type=dropout]; +"2415 layer_norm_44" [id=2415, type=layer_norm]; +"2416 add_72" [id=2416, type=add]; +"2417 _frozen_param108" [id=2417, type=get_attr]; +"2418 dequantize_per_tensor_default_87" [id=2418, type=dequantize_per_tensor]; +"2419 linear_128_scale_0" [id=2419, type=get_attr]; +"2420 linear_128_zero_point_0" [id=2420, type=get_attr]; +"2421 _frozen_param109" [id=2421, type=get_attr]; +"2422 dequantize_per_channel_default_87" [id=2422, type=dequantize_per_channel]; +"2423 linear_128" [id=2423, type=linear]; +"2424 relu__21" [id=2424, type=relu_]; +"2425 relu__21_0_0_nncf_smooth_quant_0" [id=2425, type=call_module]; +"2426 quantize_per_tensor_default_88" [id=2426, type=quantize_per_tensor]; +"2427 dequantize_per_tensor_default_88" [id=2427, type=dequantize_per_tensor]; +"2428 linear_129_scale_0" [id=2428, type=get_attr]; +"2429 linear_129_zero_point_0" [id=2429, type=get_attr]; +"2430 _frozen_param110" [id=2430, type=get_attr]; +"2431 dequantize_per_channel_default_88" [id=2431, type=dequantize_per_channel]; +"2432 linear_129" [id=2432, type=linear]; +"2433 view_114" [id=2433, type=view]; +"2434 index_21" [id=2434, type=index]; +"2435 view_115" [id=2435, type=view]; +"2436 permute_95" [id=2436, type=permute]; +"2437 contiguous_40" [id=2437, type=contiguous]; +"2438 unsqueeze_61" [id=2438, type=unsqueeze]; +"2439 sigmoid_21" [id=2439, type=sigmoid]; +"2440 mul_42" [id=2440, type=mul]; +"2441 pad_23" [id=2441, type=pad]; +"2442 roll_20" [id=2442, type=roll]; +"2443 view_116" [id=2443, type=view]; +"2444 permute_96" [id=2444, type=permute]; +"2445 reshape_94" [id=2445, type=reshape]; +"2446 clone_21" [id=2446, type=clone]; +"2447 features_5_17_attn_qkv_weight_updated_constant0" [id=2447, type=get_attr]; +"2448 reshape_94_0_0_nncf_smooth_quant_0" [id=2448, type=call_module]; +"2449 linear_130" [id=2449, type=linear]; +"2450 reshape_95" [id=2450, type=reshape]; +"2451 permute_97" [id=2451, type=permute]; +"2452 select_63" [id=2452, type=select]; +"2453 select_64" [id=2453, type=select]; +"2454 select_65" [id=2454, type=select]; +"2455 linalg_vector_norm_42" [id=2455, type=linalg_vector_norm]; +"2456 clamp_min_42" [id=2456, type=clamp_min]; +"2457 expand_as_42" [id=2457, type=expand_as]; +"2458 div_42" [id=2458, type=div]; +"2459 linalg_vector_norm_43" [id=2459, type=linalg_vector_norm]; +"2460 clamp_min_43" [id=2460, type=clamp_min]; +"2461 expand_as_43" [id=2461, type=expand_as]; +"2462 div_43" [id=2462, type=div]; +"2463 transpose_42" [id=2463, type=transpose]; +"2464 matmul_42" [id=2464, type=matmul]; +"2465 clamp_21" [id=2465, type=clamp]; +"2466 exp_21" [id=2466, type=exp]; +"2467 mul_43" [id=2467, type=mul]; +"2468 add_73" [id=2468, type=add]; +"2469 new_zeros_10" [id=2469, type=new_zeros]; +"2470 view_117" [id=2470, type=view]; +"2471 permute_98" [id=2471, type=permute]; +"2472 reshape_96" [id=2472, type=reshape]; +"2473 unsqueeze_62" [id=2473, type=unsqueeze]; +"2474 unsqueeze_63" [id=2474, type=unsqueeze]; +"2475 sub_10" [id=2475, type=sub]; +"2476 ne_10" [id=2476, type=ne]; +"2477 masked_fill_20" [id=2477, type=masked_fill]; +"2478 eq_10" [id=2478, type=eq]; +"2479 masked_fill_21" [id=2479, type=masked_fill]; +"2480 view_118" [id=2480, type=view]; +"2481 unsqueeze_64" [id=2481, type=unsqueeze]; +"2482 unsqueeze_65" [id=2482, type=unsqueeze]; +"2483 add_74" [id=2483, type=add]; +"2484 view_119" [id=2484, type=view]; +"2485 softmax_21" [id=2485, type=softmax]; +"2486 dropout_84" [id=2486, type=dropout]; +"2487 matmul_43" [id=2487, type=matmul]; +"2488 transpose_43" [id=2488, type=transpose]; +"2489 reshape_97" [id=2489, type=reshape]; +"2490 features_5_17_attn_proj_weight_updated_constant0" [id=2490, type=get_attr]; +"2491 reshape_97_0_0_nncf_smooth_quant_0" [id=2491, type=call_module]; +"2492 linear_131" [id=2492, type=linear]; +"2493 dropout_85" [id=2493, type=dropout]; +"2494 view_120" [id=2494, type=view]; +"2495 permute_99" [id=2495, type=permute]; +"2496 reshape_98" [id=2496, type=reshape]; +"2497 roll_21" [id=2497, type=roll]; +"2498 slice_325" [id=2498, type=slice]; +"2499 slice_326" [id=2499, type=slice]; +"2500 slice_327" [id=2500, type=slice]; +"2501 slice_328" [id=2501, type=slice]; +"2502 contiguous_41" [id=2502, type=contiguous]; +"2503 layer_norm_45" [id=2503, type=layer_norm]; +"2504 add_75" [id=2504, type=add]; +"2505 add_75_0_0_nncf_smooth_quant_0" [id=2505, type=call_module]; +"2506 quantize_per_tensor_default_89" [id=2506, type=quantize_per_tensor]; +"2507 dequantize_per_tensor_default_89" [id=2507, type=dequantize_per_tensor]; +"2508 linear_132_scale_0" [id=2508, type=get_attr]; +"2509 linear_132_zero_point_0" [id=2509, type=get_attr]; +"2510 _frozen_param111" [id=2510, type=get_attr]; +"2511 dequantize_per_channel_default_89" [id=2511, type=dequantize_per_channel]; +"2512 linear_132" [id=2512, type=linear]; +"2513 gelu_21" [id=2513, type=gelu]; +"2514 dropout_86" [id=2514, type=dropout]; +"2515 dropout_86_0_0_nncf_smooth_quant_0" [id=2515, type=call_module]; +"2516 quantize_per_tensor_default_90" [id=2516, type=quantize_per_tensor]; +"2517 dequantize_per_tensor_default_90" [id=2517, type=dequantize_per_tensor]; +"2518 linear_133_scale_0" [id=2518, type=get_attr]; +"2519 linear_133_zero_point_0" [id=2519, type=get_attr]; +"2520 _frozen_param112" [id=2520, type=get_attr]; +"2521 dequantize_per_channel_default_90" [id=2521, type=dequantize_per_channel]; +"2522 linear_133" [id=2522, type=linear]; +"2523 dropout_87" [id=2523, type=dropout]; +"2524 layer_norm_46" [id=2524, type=layer_norm]; +"2525 add_76" [id=2525, type=add]; +"2526 pad_24" [id=2526, type=pad]; +"2527 slice_329" [id=2527, type=slice]; +"2528 slice_330" [id=2528, type=slice]; +"2529 slice_331" [id=2529, type=slice]; +"2530 slice_332" [id=2530, type=slice]; +"2531 slice_333" [id=2531, type=slice]; +"2532 slice_334" [id=2532, type=slice]; +"2533 slice_335" [id=2533, type=slice]; +"2534 slice_336" [id=2534, type=slice]; +"2535 slice_337" [id=2535, type=slice]; +"2536 slice_338" [id=2536, type=slice]; +"2537 slice_339" [id=2537, type=slice]; +"2538 slice_340" [id=2538, type=slice]; +"2539 cat_2" [id=2539, type=cat]; +"2540 cat_2_0_0_nncf_smooth_quant_0" [id=2540, type=call_module]; +"2541 quantize_per_tensor_default_91" [id=2541, type=quantize_per_tensor]; +"2542 dequantize_per_tensor_default_91" [id=2542, type=dequantize_per_tensor]; +"2543 linear_134_scale_0" [id=2543, type=get_attr]; +"2544 linear_134_zero_point_0" [id=2544, type=get_attr]; +"2545 _frozen_param113" [id=2545, type=get_attr]; +"2546 dequantize_per_channel_default_91" [id=2546, type=dequantize_per_channel]; +"2547 linear_134" [id=2547, type=linear]; +"2548 layer_norm_47" [id=2548, type=layer_norm]; +"2549 _frozen_param114" [id=2549, type=get_attr]; +"2550 dequantize_per_tensor_default_92" [id=2550, type=dequantize_per_tensor]; +"2551 linear_135_scale_0" [id=2551, type=get_attr]; +"2552 linear_135_zero_point_0" [id=2552, type=get_attr]; +"2553 _frozen_param115" [id=2553, type=get_attr]; +"2554 dequantize_per_channel_default_92" [id=2554, type=dequantize_per_channel]; +"2555 linear_135" [id=2555, type=linear]; +"2556 relu__22" [id=2556, type=relu_]; +"2557 relu__22_0_0_nncf_smooth_quant_0" [id=2557, type=call_module]; +"2558 quantize_per_tensor_default_93" [id=2558, type=quantize_per_tensor]; +"2559 dequantize_per_tensor_default_93" [id=2559, type=dequantize_per_tensor]; +"2560 linear_136_scale_0" [id=2560, type=get_attr]; +"2561 linear_136_zero_point_0" [id=2561, type=get_attr]; +"2562 _frozen_param116" [id=2562, type=get_attr]; +"2563 dequantize_per_channel_default_93" [id=2563, type=dequantize_per_channel]; +"2564 linear_136" [id=2564, type=linear]; +"2565 view_121" [id=2565, type=view]; +"2566 index_22" [id=2566, type=index]; +"2567 view_122" [id=2567, type=view]; +"2568 permute_100" [id=2568, type=permute]; +"2569 contiguous_42" [id=2569, type=contiguous]; +"2570 unsqueeze_66" [id=2570, type=unsqueeze]; +"2571 sigmoid_22" [id=2571, type=sigmoid]; +"2572 mul_44" [id=2572, type=mul]; +"2573 pad_25" [id=2573, type=pad]; +"2574 view_123" [id=2574, type=view]; +"2575 permute_101" [id=2575, type=permute]; +"2576 reshape_99" [id=2576, type=reshape]; +"2577 clone_22" [id=2577, type=clone]; +"2578 features_7_0_attn_qkv_weight_updated_constant0" [id=2578, type=get_attr]; +"2579 reshape_99_0_0_nncf_smooth_quant_0" [id=2579, type=call_module]; +"2580 linear_137" [id=2580, type=linear]; +"2581 reshape_100" [id=2581, type=reshape]; +"2582 permute_102" [id=2582, type=permute]; +"2583 select_66" [id=2583, type=select]; +"2584 select_67" [id=2584, type=select]; +"2585 select_68" [id=2585, type=select]; +"2586 linalg_vector_norm_44" [id=2586, type=linalg_vector_norm]; +"2587 clamp_min_44" [id=2587, type=clamp_min]; +"2588 expand_as_44" [id=2588, type=expand_as]; +"2589 div_44" [id=2589, type=div]; +"2590 linalg_vector_norm_45" [id=2590, type=linalg_vector_norm]; +"2591 clamp_min_45" [id=2591, type=clamp_min]; +"2592 expand_as_45" [id=2592, type=expand_as]; +"2593 div_45" [id=2593, type=div]; +"2594 transpose_44" [id=2594, type=transpose]; +"2595 matmul_44" [id=2595, type=matmul]; +"2596 clamp_22" [id=2596, type=clamp]; +"2597 exp_22" [id=2597, type=exp]; +"2598 mul_45" [id=2598, type=mul]; +"2599 add_77" [id=2599, type=add]; +"2600 softmax_22" [id=2600, type=softmax]; +"2601 dropout_88" [id=2601, type=dropout]; +"2602 matmul_45" [id=2602, type=matmul]; +"2603 transpose_45" [id=2603, type=transpose]; +"2604 reshape_101" [id=2604, type=reshape]; +"2605 features_7_0_attn_proj_weight_updated_constant0" [id=2605, type=get_attr]; +"2606 reshape_101_0_0_nncf_smooth_quant_0" [id=2606, type=call_module]; +"2607 linear_138" [id=2607, type=linear]; +"2608 dropout_89" [id=2608, type=dropout]; +"2609 view_124" [id=2609, type=view]; +"2610 permute_103" [id=2610, type=permute]; +"2611 reshape_102" [id=2611, type=reshape]; +"2612 slice_342" [id=2612, type=slice]; +"2613 slice_343" [id=2613, type=slice]; +"2614 slice_344" [id=2614, type=slice]; +"2615 slice_345" [id=2615, type=slice]; +"2616 contiguous_43" [id=2616, type=contiguous]; +"2617 layer_norm_48" [id=2617, type=layer_norm]; +"2618 add_78" [id=2618, type=add]; +"2619 add_78_0_0_nncf_smooth_quant_0" [id=2619, type=call_module]; +"2620 quantize_per_tensor_default_94" [id=2620, type=quantize_per_tensor]; +"2621 dequantize_per_tensor_default_94" [id=2621, type=dequantize_per_tensor]; +"2622 linear_139_scale_0" [id=2622, type=get_attr]; +"2623 linear_139_zero_point_0" [id=2623, type=get_attr]; +"2624 _frozen_param117" [id=2624, type=get_attr]; +"2625 dequantize_per_channel_default_94" [id=2625, type=dequantize_per_channel]; +"2626 linear_139" [id=2626, type=linear]; +"2627 gelu_22" [id=2627, type=gelu]; +"2628 dropout_90" [id=2628, type=dropout]; +"2629 dropout_90_0_0_nncf_smooth_quant_0" [id=2629, type=call_module]; +"2630 quantize_per_tensor_default_95" [id=2630, type=quantize_per_tensor]; +"2631 dequantize_per_tensor_default_95" [id=2631, type=dequantize_per_tensor]; +"2632 linear_140_scale_0" [id=2632, type=get_attr]; +"2633 linear_140_zero_point_0" [id=2633, type=get_attr]; +"2634 _frozen_param118" [id=2634, type=get_attr]; +"2635 dequantize_per_channel_default_95" [id=2635, type=dequantize_per_channel]; +"2636 linear_140" [id=2636, type=linear]; +"2637 dropout_91" [id=2637, type=dropout]; +"2638 layer_norm_49" [id=2638, type=layer_norm]; +"2639 add_79" [id=2639, type=add]; +"2640 _frozen_param119" [id=2640, type=get_attr]; +"2641 dequantize_per_tensor_default_96" [id=2641, type=dequantize_per_tensor]; +"2642 linear_141_scale_0" [id=2642, type=get_attr]; +"2643 linear_141_zero_point_0" [id=2643, type=get_attr]; +"2644 _frozen_param120" [id=2644, type=get_attr]; +"2645 dequantize_per_channel_default_96" [id=2645, type=dequantize_per_channel]; +"2646 linear_141" [id=2646, type=linear]; +"2647 relu__23" [id=2647, type=relu_]; +"2648 relu__23_0_0_nncf_smooth_quant_0" [id=2648, type=call_module]; +"2649 quantize_per_tensor_default_97" [id=2649, type=quantize_per_tensor]; +"2650 dequantize_per_tensor_default_97" [id=2650, type=dequantize_per_tensor]; +"2651 linear_142_scale_0" [id=2651, type=get_attr]; +"2652 linear_142_zero_point_0" [id=2652, type=get_attr]; +"2653 _frozen_param121" [id=2653, type=get_attr]; +"2654 dequantize_per_channel_default_97" [id=2654, type=dequantize_per_channel]; +"2655 linear_142" [id=2655, type=linear]; +"2656 view_125" [id=2656, type=view]; +"2657 index_23" [id=2657, type=index]; +"2658 view_126" [id=2658, type=view]; +"2659 permute_104" [id=2659, type=permute]; +"2660 contiguous_44" [id=2660, type=contiguous]; +"2661 unsqueeze_67" [id=2661, type=unsqueeze]; +"2662 sigmoid_23" [id=2662, type=sigmoid]; +"2663 mul_46" [id=2663, type=mul]; +"2664 pad_26" [id=2664, type=pad]; +"2665 view_127" [id=2665, type=view]; +"2666 permute_105" [id=2666, type=permute]; +"2667 reshape_103" [id=2667, type=reshape]; +"2668 clone_23" [id=2668, type=clone]; +"2669 features_7_1_attn_qkv_weight_updated_constant0" [id=2669, type=get_attr]; +"2670 reshape_103_0_0_nncf_smooth_quant_0" [id=2670, type=call_module]; +"2671 linear_143" [id=2671, type=linear]; +"2672 reshape_104" [id=2672, type=reshape]; +"2673 permute_106" [id=2673, type=permute]; +"2674 select_69" [id=2674, type=select]; +"2675 select_70" [id=2675, type=select]; +"2676 select_71" [id=2676, type=select]; +"2677 linalg_vector_norm_46" [id=2677, type=linalg_vector_norm]; +"2678 clamp_min_46" [id=2678, type=clamp_min]; +"2679 expand_as_46" [id=2679, type=expand_as]; +"2680 div_46" [id=2680, type=div]; +"2681 linalg_vector_norm_47" [id=2681, type=linalg_vector_norm]; +"2682 clamp_min_47" [id=2682, type=clamp_min]; +"2683 expand_as_47" [id=2683, type=expand_as]; +"2684 div_47" [id=2684, type=div]; +"2685 transpose_46" [id=2685, type=transpose]; +"2686 matmul_46" [id=2686, type=matmul]; +"2687 clamp_23" [id=2687, type=clamp]; +"2688 exp_23" [id=2688, type=exp]; +"2689 mul_47" [id=2689, type=mul]; +"2690 add_80" [id=2690, type=add]; +"2691 softmax_23" [id=2691, type=softmax]; +"2692 dropout_92" [id=2692, type=dropout]; +"2693 matmul_47" [id=2693, type=matmul]; +"2694 transpose_47" [id=2694, type=transpose]; +"2695 reshape_105" [id=2695, type=reshape]; +"2696 features_7_1_attn_proj_weight_updated_constant0" [id=2696, type=get_attr]; +"2697 reshape_105_0_0_nncf_smooth_quant_0" [id=2697, type=call_module]; +"2698 linear_144" [id=2698, type=linear]; +"2699 dropout_93" [id=2699, type=dropout]; +"2700 view_128" [id=2700, type=view]; +"2701 permute_107" [id=2701, type=permute]; +"2702 reshape_106" [id=2702, type=reshape]; +"2703 slice_347" [id=2703, type=slice]; +"2704 slice_348" [id=2704, type=slice]; +"2705 slice_349" [id=2705, type=slice]; +"2706 slice_350" [id=2706, type=slice]; +"2707 contiguous_45" [id=2707, type=contiguous]; +"2708 layer_norm_50" [id=2708, type=layer_norm]; +"2709 add_81" [id=2709, type=add]; +"2710 add_81_0_0_nncf_smooth_quant_0" [id=2710, type=call_module]; +"2711 quantize_per_tensor_default_98" [id=2711, type=quantize_per_tensor]; +"2712 dequantize_per_tensor_default_98" [id=2712, type=dequantize_per_tensor]; +"2713 linear_145_scale_0" [id=2713, type=get_attr]; +"2714 linear_145_zero_point_0" [id=2714, type=get_attr]; +"2715 _frozen_param122" [id=2715, type=get_attr]; +"2716 dequantize_per_channel_default_98" [id=2716, type=dequantize_per_channel]; +"2717 linear_145" [id=2717, type=linear]; +"2718 gelu_23" [id=2718, type=gelu]; +"2719 dropout_94" [id=2719, type=dropout]; +"2720 dropout_94_0_0_nncf_smooth_quant_0" [id=2720, type=call_module]; +"2721 quantize_per_tensor_default_99" [id=2721, type=quantize_per_tensor]; +"2722 dequantize_per_tensor_default_99" [id=2722, type=dequantize_per_tensor]; +"2723 linear_146_scale_0" [id=2723, type=get_attr]; +"2724 linear_146_zero_point_0" [id=2724, type=get_attr]; +"2725 _frozen_param123" [id=2725, type=get_attr]; +"2726 dequantize_per_channel_default_99" [id=2726, type=dequantize_per_channel]; +"2727 linear_146" [id=2727, type=linear]; +"2728 dropout_95" [id=2728, type=dropout]; +"2729 layer_norm_51" [id=2729, type=layer_norm]; +"2730 add_82" [id=2730, type=add]; +"2731 layer_norm_52" [id=2731, type=layer_norm]; +"2732 permute_108" [id=2732, type=permute]; +"2733 adaptive_avg_pool2d" [id=2733, type=adaptive_avg_pool2d]; +"2734 flatten" [id=2734, type=flatten]; +"2735 flatten_0_0_nncf_smooth_quant_0" [id=2735, type=call_module]; +"2736 quantize_per_tensor_default_100" [id=2736, type=quantize_per_tensor]; +"2737 dequantize_per_tensor_default_100" [id=2737, type=dequantize_per_tensor]; +"2738 linear_147_scale_0" [id=2738, type=get_attr]; +"2739 linear_147_zero_point_0" [id=2739, type=get_attr]; +"2740 _frozen_param124" [id=2740, type=get_attr]; +"2741 dequantize_per_channel_default_100" [id=2741, type=dequantize_per_channel]; +"2742 linear_147" [id=2742, type=linear]; +"2743 output" [id=2743, type=output]; +"0 features_1_0_attn_qkv_bias" -> "314 clone" [label="(288,)", style=solid]; +"1 features_1_0_attn_proj_bias" -> "344 linear_3" [label="(96,)", style=solid]; +"2 features_1_0_attn_logit_scale" -> "333 clamp" [label="(3, 1, 1)", style=solid]; +"3 features_1_1_attn_qkv_bias" -> "403 clone_1" [label="(288,)", style=solid]; +"4 features_1_1_attn_proj_bias" -> "449 linear_9" [label="(96,)", style=solid]; +"5 features_1_1_attn_logit_scale" -> "422 clamp_1" [label="(3, 1, 1)", style=solid]; +"6 features_3_0_attn_qkv_bias" -> "531 clone_2" [label="(576,)", style=solid]; +"7 features_3_0_attn_proj_bias" -> "561 linear_16" [label="(192,)", style=solid]; +"8 features_3_0_attn_logit_scale" -> "550 clamp_2" [label="(6, 1, 1)", style=solid]; +"9 features_3_1_attn_qkv_bias" -> "623 clone_3" [label="(576,)", style=solid]; +"10 features_3_1_attn_proj_bias" -> "669 linear_22" [label="(192,)", style=solid]; +"11 features_3_1_attn_logit_scale" -> "642 clamp_3" [label="(6, 1, 1)", style=solid]; +"12 features_5_0_attn_qkv_bias" -> "754 clone_4" [label="(1152,)", style=solid]; +"13 features_5_0_attn_proj_bias" -> "784 linear_29" [label="(384,)", style=solid]; +"14 features_5_0_attn_logit_scale" -> "773 clamp_4" [label="(12, 1, 1)", style=solid]; +"15 features_5_1_attn_qkv_bias" -> "846 clone_5" [label="(1152,)", style=solid]; +"16 features_5_1_attn_proj_bias" -> "892 linear_35" [label="(384,)", style=solid]; +"17 features_5_1_attn_logit_scale" -> "865 clamp_5" [label="(12, 1, 1)", style=solid]; +"18 features_5_2_attn_qkv_bias" -> "954 clone_6" [label="(1152,)", style=solid]; +"19 features_5_2_attn_proj_bias" -> "984 linear_41" [label="(384,)", style=solid]; +"20 features_5_2_attn_logit_scale" -> "973 clamp_6" [label="(12, 1, 1)", style=solid]; +"21 features_5_3_attn_qkv_bias" -> "1046 clone_7" [label="(1152,)", style=solid]; +"22 features_5_3_attn_proj_bias" -> "1092 linear_47" [label="(384,)", style=solid]; +"23 features_5_3_attn_logit_scale" -> "1065 clamp_7" [label="(12, 1, 1)", style=solid]; +"24 features_5_4_attn_qkv_bias" -> "1154 clone_8" [label="(1152,)", style=solid]; +"25 features_5_4_attn_proj_bias" -> "1184 linear_53" [label="(384,)", style=solid]; +"26 features_5_4_attn_logit_scale" -> "1173 clamp_8" [label="(12, 1, 1)", style=solid]; +"27 features_5_5_attn_qkv_bias" -> "1246 clone_9" [label="(1152,)", style=solid]; +"28 features_5_5_attn_proj_bias" -> "1292 linear_59" [label="(384,)", style=solid]; +"29 features_5_5_attn_logit_scale" -> "1265 clamp_9" [label="(12, 1, 1)", style=solid]; +"30 features_5_6_attn_qkv_bias" -> "1354 clone_10" [label="(1152,)", style=solid]; +"31 features_5_6_attn_proj_bias" -> "1384 linear_65" [label="(384,)", style=solid]; +"32 features_5_6_attn_logit_scale" -> "1373 clamp_10" [label="(12, 1, 1)", style=solid]; +"33 features_5_7_attn_qkv_bias" -> "1446 clone_11" [label="(1152,)", style=solid]; +"34 features_5_7_attn_proj_bias" -> "1492 linear_71" [label="(384,)", style=solid]; +"35 features_5_7_attn_logit_scale" -> "1465 clamp_11" [label="(12, 1, 1)", style=solid]; +"36 features_5_8_attn_qkv_bias" -> "1554 clone_12" [label="(1152,)", style=solid]; +"37 features_5_8_attn_proj_bias" -> "1584 linear_77" [label="(384,)", style=solid]; +"38 features_5_8_attn_logit_scale" -> "1573 clamp_12" [label="(12, 1, 1)", style=solid]; +"39 features_5_9_attn_qkv_bias" -> "1646 clone_13" [label="(1152,)", style=solid]; +"40 features_5_9_attn_proj_bias" -> "1692 linear_83" [label="(384,)", style=solid]; +"41 features_5_9_attn_logit_scale" -> "1665 clamp_13" [label="(12, 1, 1)", style=solid]; +"42 features_5_10_attn_qkv_bias" -> "1754 clone_14" [label="(1152,)", style=solid]; +"43 features_5_10_attn_proj_bias" -> "1784 linear_89" [label="(384,)", style=solid]; +"44 features_5_10_attn_logit_scale" -> "1773 clamp_14" [label="(12, 1, 1)", style=solid]; +"45 features_5_11_attn_qkv_bias" -> "1846 clone_15" [label="(1152,)", style=solid]; +"46 features_5_11_attn_proj_bias" -> "1892 linear_95" [label="(384,)", style=solid]; +"47 features_5_11_attn_logit_scale" -> "1865 clamp_15" [label="(12, 1, 1)", style=solid]; +"48 features_5_12_attn_qkv_bias" -> "1954 clone_16" [label="(1152,)", style=solid]; +"49 features_5_12_attn_proj_bias" -> "1984 linear_101" [label="(384,)", style=solid]; +"50 features_5_12_attn_logit_scale" -> "1973 clamp_16" [label="(12, 1, 1)", style=solid]; +"51 features_5_13_attn_qkv_bias" -> "2046 clone_17" [label="(1152,)", style=solid]; +"52 features_5_13_attn_proj_bias" -> "2092 linear_107" [label="(384,)", style=solid]; +"53 features_5_13_attn_logit_scale" -> "2065 clamp_17" [label="(12, 1, 1)", style=solid]; +"54 features_5_14_attn_qkv_bias" -> "2154 clone_18" [label="(1152,)", style=solid]; +"55 features_5_14_attn_proj_bias" -> "2184 linear_113" [label="(384,)", style=solid]; +"56 features_5_14_attn_logit_scale" -> "2173 clamp_18" [label="(12, 1, 1)", style=solid]; +"57 features_5_15_attn_qkv_bias" -> "2246 clone_19" [label="(1152,)", style=solid]; +"58 features_5_15_attn_proj_bias" -> "2292 linear_119" [label="(384,)", style=solid]; +"59 features_5_15_attn_logit_scale" -> "2265 clamp_19" [label="(12, 1, 1)", style=solid]; +"60 features_5_16_attn_qkv_bias" -> "2354 clone_20" [label="(1152,)", style=solid]; +"61 features_5_16_attn_proj_bias" -> "2384 linear_125" [label="(384,)", style=solid]; +"62 features_5_16_attn_logit_scale" -> "2373 clamp_20" [label="(12, 1, 1)", style=solid]; +"63 features_5_17_attn_qkv_bias" -> "2446 clone_21" [label="(1152,)", style=solid]; +"64 features_5_17_attn_proj_bias" -> "2492 linear_131" [label="(384,)", style=solid]; +"65 features_5_17_attn_logit_scale" -> "2465 clamp_21" [label="(12, 1, 1)", style=solid]; +"66 features_7_0_attn_qkv_bias" -> "2577 clone_22" [label="(2304,)", style=solid]; +"67 features_7_0_attn_proj_bias" -> "2607 linear_138" [label="(768,)", style=solid]; +"68 features_7_0_attn_logit_scale" -> "2596 clamp_22" [label="(24, 1, 1)", style=solid]; +"69 features_7_1_attn_qkv_bias" -> "2668 clone_23" [label="(2304,)", style=solid]; +"70 features_7_1_attn_proj_bias" -> "2698 linear_144" [label="(768,)", style=solid]; +"71 features_7_1_attn_logit_scale" -> "2687 clamp_23" [label="(24, 1, 1)", style=solid]; +"72 features_0_0_bias" -> "283 conv2d" [label="(96,)", style=solid]; +"73 features_0_2_weight" -> "285 layer_norm" [label="(96,)", style=solid]; +"74 features_0_2_bias" -> "285 layer_norm" [label="(96,)", style=solid]; +"75 features_1_0_attn_cpb_mlp_0_bias" -> "292 linear" [label="(512,)", style=solid]; +"76 features_1_0_norm1_weight" -> "351 layer_norm_1" [label="(96,)", style=solid]; +"77 features_1_0_norm1_bias" -> "351 layer_norm_1" [label="(96,)", style=solid]; +"78 features_1_0_mlp_0_bias" -> "360 linear_4" [label="(384,)", style=solid]; +"79 features_1_0_mlp_3_bias" -> "370 linear_5" [label="(96,)", style=solid]; +"80 features_1_0_norm2_weight" -> "372 layer_norm_2" [label="(96,)", style=solid]; +"81 features_1_0_norm2_bias" -> "372 layer_norm_2" [label="(96,)", style=solid]; +"82 features_1_1_attn_cpb_mlp_0_bias" -> "380 linear_6" [label="(512,)", style=solid]; +"83 features_1_1_norm1_weight" -> "457 layer_norm_3" [label="(96,)", style=solid]; +"84 features_1_1_norm1_bias" -> "457 layer_norm_3" [label="(96,)", style=solid]; +"85 features_1_1_mlp_0_bias" -> "466 linear_10" [label="(384,)", style=solid]; +"86 features_1_1_mlp_3_bias" -> "476 linear_11" [label="(96,)", style=solid]; +"87 features_1_1_norm2_weight" -> "478 layer_norm_4" [label="(96,)", style=solid]; +"88 features_1_1_norm2_bias" -> "478 layer_norm_4" [label="(96,)", style=solid]; +"89 features_2_norm_weight" -> "502 layer_norm_5" [label="(192,)", style=solid]; +"90 features_2_norm_bias" -> "502 layer_norm_5" [label="(192,)", style=solid]; +"91 features_3_0_attn_cpb_mlp_0_bias" -> "509 linear_13" [label="(512,)", style=solid]; +"92 features_3_0_norm1_weight" -> "571 layer_norm_6" [label="(192,)", style=solid]; +"93 features_3_0_norm1_bias" -> "571 layer_norm_6" [label="(192,)", style=solid]; +"94 features_3_0_mlp_0_bias" -> "580 linear_17" [label="(768,)", style=solid]; +"95 features_3_0_mlp_3_bias" -> "590 linear_18" [label="(192,)", style=solid]; +"96 features_3_0_norm2_weight" -> "592 layer_norm_7" [label="(192,)", style=solid]; +"97 features_3_0_norm2_bias" -> "592 layer_norm_7" [label="(192,)", style=solid]; +"98 features_3_1_attn_cpb_mlp_0_bias" -> "600 linear_19" [label="(512,)", style=solid]; +"99 features_3_1_norm1_weight" -> "680 layer_norm_8" [label="(192,)", style=solid]; +"100 features_3_1_norm1_bias" -> "680 layer_norm_8" [label="(192,)", style=solid]; +"101 features_3_1_mlp_0_bias" -> "689 linear_23" [label="(768,)", style=solid]; +"102 features_3_1_mlp_3_bias" -> "699 linear_24" [label="(192,)", style=solid]; +"103 features_3_1_norm2_weight" -> "701 layer_norm_9" [label="(192,)", style=solid]; +"104 features_3_1_norm2_bias" -> "701 layer_norm_9" [label="(192,)", style=solid]; +"105 features_4_norm_weight" -> "725 layer_norm_10" [label="(384,)", style=solid]; +"106 features_4_norm_bias" -> "725 layer_norm_10" [label="(384,)", style=solid]; +"107 features_5_0_attn_cpb_mlp_0_bias" -> "732 linear_26" [label="(512,)", style=solid]; +"108 features_5_0_norm1_weight" -> "794 layer_norm_11" [label="(384,)", style=solid]; +"109 features_5_0_norm1_bias" -> "794 layer_norm_11" [label="(384,)", style=solid]; +"110 features_5_0_mlp_0_bias" -> "803 linear_30" [label="(1536,)", style=solid]; +"111 features_5_0_mlp_3_bias" -> "813 linear_31" [label="(384,)", style=solid]; +"112 features_5_0_norm2_weight" -> "815 layer_norm_12" [label="(384,)", style=solid]; +"113 features_5_0_norm2_bias" -> "815 layer_norm_12" [label="(384,)", style=solid]; +"114 features_5_1_attn_cpb_mlp_0_bias" -> "823 linear_32" [label="(512,)", style=solid]; +"115 features_5_1_norm1_weight" -> "903 layer_norm_13" [label="(384,)", style=solid]; +"116 features_5_1_norm1_bias" -> "903 layer_norm_13" [label="(384,)", style=solid]; +"117 features_5_1_mlp_0_bias" -> "912 linear_36" [label="(1536,)", style=solid]; +"118 features_5_1_mlp_3_bias" -> "922 linear_37" [label="(384,)", style=solid]; +"119 features_5_1_norm2_weight" -> "924 layer_norm_14" [label="(384,)", style=solid]; +"120 features_5_1_norm2_bias" -> "924 layer_norm_14" [label="(384,)", style=solid]; +"121 features_5_2_attn_cpb_mlp_0_bias" -> "932 linear_38" [label="(512,)", style=solid]; +"122 features_5_2_norm1_weight" -> "994 layer_norm_15" [label="(384,)", style=solid]; +"123 features_5_2_norm1_bias" -> "994 layer_norm_15" [label="(384,)", style=solid]; +"124 features_5_2_mlp_0_bias" -> "1003 linear_42" [label="(1536,)", style=solid]; +"125 features_5_2_mlp_3_bias" -> "1013 linear_43" [label="(384,)", style=solid]; +"126 features_5_2_norm2_weight" -> "1015 layer_norm_16" [label="(384,)", style=solid]; +"127 features_5_2_norm2_bias" -> "1015 layer_norm_16" [label="(384,)", style=solid]; +"128 features_5_3_attn_cpb_mlp_0_bias" -> "1023 linear_44" [label="(512,)", style=solid]; +"129 features_5_3_norm1_weight" -> "1103 layer_norm_17" [label="(384,)", style=solid]; +"130 features_5_3_norm1_bias" -> "1103 layer_norm_17" [label="(384,)", style=solid]; +"131 features_5_3_mlp_0_bias" -> "1112 linear_48" [label="(1536,)", style=solid]; +"132 features_5_3_mlp_3_bias" -> "1122 linear_49" [label="(384,)", style=solid]; +"133 features_5_3_norm2_weight" -> "1124 layer_norm_18" [label="(384,)", style=solid]; +"134 features_5_3_norm2_bias" -> "1124 layer_norm_18" [label="(384,)", style=solid]; +"135 features_5_4_attn_cpb_mlp_0_bias" -> "1132 linear_50" [label="(512,)", style=solid]; +"136 features_5_4_norm1_weight" -> "1194 layer_norm_19" [label="(384,)", style=solid]; +"137 features_5_4_norm1_bias" -> "1194 layer_norm_19" [label="(384,)", style=solid]; +"138 features_5_4_mlp_0_bias" -> "1203 linear_54" [label="(1536,)", style=solid]; +"139 features_5_4_mlp_3_bias" -> "1213 linear_55" [label="(384,)", style=solid]; +"140 features_5_4_norm2_weight" -> "1215 layer_norm_20" [label="(384,)", style=solid]; +"141 features_5_4_norm2_bias" -> "1215 layer_norm_20" [label="(384,)", style=solid]; +"142 features_5_5_attn_cpb_mlp_0_bias" -> "1223 linear_56" [label="(512,)", style=solid]; +"143 features_5_5_norm1_weight" -> "1303 layer_norm_21" [label="(384,)", style=solid]; +"144 features_5_5_norm1_bias" -> "1303 layer_norm_21" [label="(384,)", style=solid]; +"145 features_5_5_mlp_0_bias" -> "1312 linear_60" [label="(1536,)", style=solid]; +"146 features_5_5_mlp_3_bias" -> "1322 linear_61" [label="(384,)", style=solid]; +"147 features_5_5_norm2_weight" -> "1324 layer_norm_22" [label="(384,)", style=solid]; +"148 features_5_5_norm2_bias" -> "1324 layer_norm_22" [label="(384,)", style=solid]; +"149 features_5_6_attn_cpb_mlp_0_bias" -> "1332 linear_62" [label="(512,)", style=solid]; +"150 features_5_6_norm1_weight" -> "1394 layer_norm_23" [label="(384,)", style=solid]; +"151 features_5_6_norm1_bias" -> "1394 layer_norm_23" [label="(384,)", style=solid]; +"152 features_5_6_mlp_0_bias" -> "1403 linear_66" [label="(1536,)", style=solid]; +"153 features_5_6_mlp_3_bias" -> "1413 linear_67" [label="(384,)", style=solid]; +"154 features_5_6_norm2_weight" -> "1415 layer_norm_24" [label="(384,)", style=solid]; +"155 features_5_6_norm2_bias" -> "1415 layer_norm_24" [label="(384,)", style=solid]; +"156 features_5_7_attn_cpb_mlp_0_bias" -> "1423 linear_68" [label="(512,)", style=solid]; +"157 features_5_7_norm1_weight" -> "1503 layer_norm_25" [label="(384,)", style=solid]; +"158 features_5_7_norm1_bias" -> "1503 layer_norm_25" [label="(384,)", style=solid]; +"159 features_5_7_mlp_0_bias" -> "1512 linear_72" [label="(1536,)", style=solid]; +"160 features_5_7_mlp_3_bias" -> "1522 linear_73" [label="(384,)", style=solid]; +"161 features_5_7_norm2_weight" -> "1524 layer_norm_26" [label="(384,)", style=solid]; +"162 features_5_7_norm2_bias" -> "1524 layer_norm_26" [label="(384,)", style=solid]; +"163 features_5_8_attn_cpb_mlp_0_bias" -> "1532 linear_74" [label="(512,)", style=solid]; +"164 features_5_8_norm1_weight" -> "1594 layer_norm_27" [label="(384,)", style=solid]; +"165 features_5_8_norm1_bias" -> "1594 layer_norm_27" [label="(384,)", style=solid]; +"166 features_5_8_mlp_0_bias" -> "1603 linear_78" [label="(1536,)", style=solid]; +"167 features_5_8_mlp_3_bias" -> "1613 linear_79" [label="(384,)", style=solid]; +"168 features_5_8_norm2_weight" -> "1615 layer_norm_28" [label="(384,)", style=solid]; +"169 features_5_8_norm2_bias" -> "1615 layer_norm_28" [label="(384,)", style=solid]; +"170 features_5_9_attn_cpb_mlp_0_bias" -> "1623 linear_80" [label="(512,)", style=solid]; +"171 features_5_9_norm1_weight" -> "1703 layer_norm_29" [label="(384,)", style=solid]; +"172 features_5_9_norm1_bias" -> "1703 layer_norm_29" [label="(384,)", style=solid]; +"173 features_5_9_mlp_0_bias" -> "1712 linear_84" [label="(1536,)", style=solid]; +"174 features_5_9_mlp_3_bias" -> "1722 linear_85" [label="(384,)", style=solid]; +"175 features_5_9_norm2_weight" -> "1724 layer_norm_30" [label="(384,)", style=solid]; +"176 features_5_9_norm2_bias" -> "1724 layer_norm_30" [label="(384,)", style=solid]; +"177 features_5_10_attn_cpb_mlp_0_bias" -> "1732 linear_86" [label="(512,)", style=solid]; +"178 features_5_10_norm1_weight" -> "1794 layer_norm_31" [label="(384,)", style=solid]; +"179 features_5_10_norm1_bias" -> "1794 layer_norm_31" [label="(384,)", style=solid]; +"180 features_5_10_mlp_0_bias" -> "1803 linear_90" [label="(1536,)", style=solid]; +"181 features_5_10_mlp_3_bias" -> "1813 linear_91" [label="(384,)", style=solid]; +"182 features_5_10_norm2_weight" -> "1815 layer_norm_32" [label="(384,)", style=solid]; +"183 features_5_10_norm2_bias" -> "1815 layer_norm_32" [label="(384,)", style=solid]; +"184 features_5_11_attn_cpb_mlp_0_bias" -> "1823 linear_92" [label="(512,)", style=solid]; +"185 features_5_11_norm1_weight" -> "1903 layer_norm_33" [label="(384,)", style=solid]; +"186 features_5_11_norm1_bias" -> "1903 layer_norm_33" [label="(384,)", style=solid]; +"187 features_5_11_mlp_0_bias" -> "1912 linear_96" [label="(1536,)", style=solid]; +"188 features_5_11_mlp_3_bias" -> "1922 linear_97" [label="(384,)", style=solid]; +"189 features_5_11_norm2_weight" -> "1924 layer_norm_34" [label="(384,)", style=solid]; +"190 features_5_11_norm2_bias" -> "1924 layer_norm_34" [label="(384,)", style=solid]; +"191 features_5_12_attn_cpb_mlp_0_bias" -> "1932 linear_98" [label="(512,)", style=solid]; +"192 features_5_12_norm1_weight" -> "1994 layer_norm_35" [label="(384,)", style=solid]; +"193 features_5_12_norm1_bias" -> "1994 layer_norm_35" [label="(384,)", style=solid]; +"194 features_5_12_mlp_0_bias" -> "2003 linear_102" [label="(1536,)", style=solid]; +"195 features_5_12_mlp_3_bias" -> "2013 linear_103" [label="(384,)", style=solid]; +"196 features_5_12_norm2_weight" -> "2015 layer_norm_36" [label="(384,)", style=solid]; +"197 features_5_12_norm2_bias" -> "2015 layer_norm_36" [label="(384,)", style=solid]; +"198 features_5_13_attn_cpb_mlp_0_bias" -> "2023 linear_104" [label="(512,)", style=solid]; +"199 features_5_13_norm1_weight" -> "2103 layer_norm_37" [label="(384,)", style=solid]; +"200 features_5_13_norm1_bias" -> "2103 layer_norm_37" [label="(384,)", style=solid]; +"201 features_5_13_mlp_0_bias" -> "2112 linear_108" [label="(1536,)", style=solid]; +"202 features_5_13_mlp_3_bias" -> "2122 linear_109" [label="(384,)", style=solid]; +"203 features_5_13_norm2_weight" -> "2124 layer_norm_38" [label="(384,)", style=solid]; +"204 features_5_13_norm2_bias" -> "2124 layer_norm_38" [label="(384,)", style=solid]; +"205 features_5_14_attn_cpb_mlp_0_bias" -> "2132 linear_110" [label="(512,)", style=solid]; +"206 features_5_14_norm1_weight" -> "2194 layer_norm_39" [label="(384,)", style=solid]; +"207 features_5_14_norm1_bias" -> "2194 layer_norm_39" [label="(384,)", style=solid]; +"208 features_5_14_mlp_0_bias" -> "2203 linear_114" [label="(1536,)", style=solid]; +"209 features_5_14_mlp_3_bias" -> "2213 linear_115" [label="(384,)", style=solid]; +"210 features_5_14_norm2_weight" -> "2215 layer_norm_40" [label="(384,)", style=solid]; +"211 features_5_14_norm2_bias" -> "2215 layer_norm_40" [label="(384,)", style=solid]; +"212 features_5_15_attn_cpb_mlp_0_bias" -> "2223 linear_116" [label="(512,)", style=solid]; +"213 features_5_15_norm1_weight" -> "2303 layer_norm_41" [label="(384,)", style=solid]; +"214 features_5_15_norm1_bias" -> "2303 layer_norm_41" [label="(384,)", style=solid]; +"215 features_5_15_mlp_0_bias" -> "2312 linear_120" [label="(1536,)", style=solid]; +"216 features_5_15_mlp_3_bias" -> "2322 linear_121" [label="(384,)", style=solid]; +"217 features_5_15_norm2_weight" -> "2324 layer_norm_42" [label="(384,)", style=solid]; +"218 features_5_15_norm2_bias" -> "2324 layer_norm_42" [label="(384,)", style=solid]; +"219 features_5_16_attn_cpb_mlp_0_bias" -> "2332 linear_122" [label="(512,)", style=solid]; +"220 features_5_16_norm1_weight" -> "2394 layer_norm_43" [label="(384,)", style=solid]; +"221 features_5_16_norm1_bias" -> "2394 layer_norm_43" [label="(384,)", style=solid]; +"222 features_5_16_mlp_0_bias" -> "2403 linear_126" [label="(1536,)", style=solid]; +"223 features_5_16_mlp_3_bias" -> "2413 linear_127" [label="(384,)", style=solid]; +"224 features_5_16_norm2_weight" -> "2415 layer_norm_44" [label="(384,)", style=solid]; +"225 features_5_16_norm2_bias" -> "2415 layer_norm_44" [label="(384,)", style=solid]; +"226 features_5_17_attn_cpb_mlp_0_bias" -> "2423 linear_128" [label="(512,)", style=solid]; +"227 features_5_17_norm1_weight" -> "2503 layer_norm_45" [label="(384,)", style=solid]; +"228 features_5_17_norm1_bias" -> "2503 layer_norm_45" [label="(384,)", style=solid]; +"229 features_5_17_mlp_0_bias" -> "2512 linear_132" [label="(1536,)", style=solid]; +"230 features_5_17_mlp_3_bias" -> "2522 linear_133" [label="(384,)", style=solid]; +"231 features_5_17_norm2_weight" -> "2524 layer_norm_46" [label="(384,)", style=solid]; +"232 features_5_17_norm2_bias" -> "2524 layer_norm_46" [label="(384,)", style=solid]; +"233 features_6_norm_weight" -> "2548 layer_norm_47" [label="(768,)", style=solid]; +"234 features_6_norm_bias" -> "2548 layer_norm_47" [label="(768,)", style=solid]; +"235 features_7_0_attn_cpb_mlp_0_bias" -> "2555 linear_135" [label="(512,)", style=solid]; +"236 features_7_0_norm1_weight" -> "2617 layer_norm_48" [label="(768,)", style=solid]; +"237 features_7_0_norm1_bias" -> "2617 layer_norm_48" [label="(768,)", style=solid]; +"238 features_7_0_mlp_0_bias" -> "2626 linear_139" [label="(3072,)", style=solid]; +"239 features_7_0_mlp_3_bias" -> "2636 linear_140" [label="(768,)", style=solid]; +"240 features_7_0_norm2_weight" -> "2638 layer_norm_49" [label="(768,)", style=solid]; +"241 features_7_0_norm2_bias" -> "2638 layer_norm_49" [label="(768,)", style=solid]; +"242 features_7_1_attn_cpb_mlp_0_bias" -> "2646 linear_141" [label="(512,)", style=solid]; +"243 features_7_1_norm1_weight" -> "2708 layer_norm_50" [label="(768,)", style=solid]; +"244 features_7_1_norm1_bias" -> "2708 layer_norm_50" [label="(768,)", style=solid]; +"245 features_7_1_mlp_0_bias" -> "2717 linear_145" [label="(3072,)", style=solid]; +"246 features_7_1_mlp_3_bias" -> "2727 linear_146" [label="(768,)", style=solid]; +"247 features_7_1_norm2_weight" -> "2729 layer_norm_51" [label="(768,)", style=solid]; +"248 features_7_1_norm2_bias" -> "2729 layer_norm_51" [label="(768,)", style=solid]; +"249 norm_weight" -> "2731 layer_norm_52" [label="(768,)", style=solid]; +"250 norm_bias" -> "2731 layer_norm_52" [label="(768,)", style=solid]; +"251 head_bias" -> "2742 linear_147" [label="(1000,)", style=solid]; +"252 features_1_0_attn_relative_position_index" -> "303 index" [label="(4096,)", style=solid]; +"253 features_1_1_attn_relative_position_index" -> "391 index_1" [label="(4096,)", style=solid]; +"254 features_3_0_attn_relative_position_index" -> "520 index_2" [label="(4096,)", style=solid]; +"255 features_3_1_attn_relative_position_index" -> "611 index_3" [label="(4096,)", style=solid]; +"256 features_5_0_attn_relative_position_index" -> "743 index_4" [label="(4096,)", style=solid]; +"257 features_5_1_attn_relative_position_index" -> "834 index_5" [label="(4096,)", style=solid]; +"258 features_5_2_attn_relative_position_index" -> "943 index_6" [label="(4096,)", style=solid]; +"259 features_5_3_attn_relative_position_index" -> "1034 index_7" [label="(4096,)", style=solid]; +"260 features_5_4_attn_relative_position_index" -> "1143 index_8" [label="(4096,)", style=solid]; +"261 features_5_5_attn_relative_position_index" -> "1234 index_9" [label="(4096,)", style=solid]; +"262 features_5_6_attn_relative_position_index" -> "1343 index_10" [label="(4096,)", style=solid]; +"263 features_5_7_attn_relative_position_index" -> "1434 index_11" [label="(4096,)", style=solid]; +"264 features_5_8_attn_relative_position_index" -> "1543 index_12" [label="(4096,)", style=solid]; +"265 features_5_9_attn_relative_position_index" -> "1634 index_13" [label="(4096,)", style=solid]; +"266 features_5_10_attn_relative_position_index" -> "1743 index_14" [label="(4096,)", style=solid]; +"267 features_5_11_attn_relative_position_index" -> "1834 index_15" [label="(4096,)", style=solid]; +"268 features_5_12_attn_relative_position_index" -> "1943 index_16" [label="(4096,)", style=solid]; +"269 features_5_13_attn_relative_position_index" -> "2034 index_17" [label="(4096,)", style=solid]; +"270 features_5_14_attn_relative_position_index" -> "2143 index_18" [label="(4096,)", style=solid]; +"271 features_5_15_attn_relative_position_index" -> "2234 index_19" [label="(4096,)", style=solid]; +"272 features_5_16_attn_relative_position_index" -> "2343 index_20" [label="(4096,)", style=solid]; +"273 features_5_17_attn_relative_position_index" -> "2434 index_21" [label="(4096,)", style=solid]; +"274 features_7_0_attn_relative_position_index" -> "2566 index_22" [label="(4096,)", style=solid]; +"275 features_7_1_attn_relative_position_index" -> "2657 index_23" [label="(4096,)", style=solid]; +"276 x" -> "277 quantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; +"277 quantize_per_tensor_default" -> "278 dequantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; +"278 dequantize_per_tensor_default" -> "283 conv2d" [label="(1, 3, 224, 224)", style=solid]; +"279 conv2d_scale_0" -> "282 dequantize_per_channel_default" [label="(96,)", style=solid]; +"280 conv2d_zero_point_0" -> "282 dequantize_per_channel_default" [label="(96,)", style=solid]; +"281 _frozen_param0" -> "282 dequantize_per_channel_default" [label="(96, 3, 4, 4)", style=solid]; +"282 dequantize_per_channel_default" -> "283 conv2d" [label="(96, 3, 4, 4)", style=solid]; +"283 conv2d" -> "284 permute" [label="(1, 96, 56, 56)", style=solid]; +"284 permute" -> "285 layer_norm" [label="(1, 56, 56, 96)", style=solid]; +"285 layer_norm" -> "310 pad" [label="(1, 56, 56, 96)", style=solid]; +"285 layer_norm" -> "352 add_1" [label="(1, 56, 56, 96)", style=solid]; +"286 _frozen_param1" -> "287 dequantize_per_tensor_default_1" [label="(1, 15, 15, 2)", style=solid]; +"287 dequantize_per_tensor_default_1" -> "292 linear" [label="(1, 15, 15, 2)", style=solid]; +"288 linear_scale_0" -> "291 dequantize_per_channel_default_1" [label="(512,)", style=solid]; +"289 linear_zero_point_0" -> "291 dequantize_per_channel_default_1" [label="(512,)", style=solid]; +"290 _frozen_param2" -> "291 dequantize_per_channel_default_1" [label="(512, 2)", style=solid]; +"291 dequantize_per_channel_default_1" -> "292 linear" [label="(512, 2)", style=solid]; +"292 linear" -> "293 relu_" [label="(1, 15, 15, 512)", style=solid]; +"293 relu_" -> "294 relu__0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"294 relu__0_0_nncf_smooth_quant_0" -> "295 quantize_per_tensor_default_2" [label="(1, 15, 15, 512)", style=solid]; +"295 quantize_per_tensor_default_2" -> "296 dequantize_per_tensor_default_2" [label="(1, 15, 15, 512)", style=solid]; +"296 dequantize_per_tensor_default_2" -> "301 linear_1" [label="(1, 15, 15, 512)", style=solid]; +"297 linear_1_scale_0" -> "300 dequantize_per_channel_default_2" [label="(3,)", style=solid]; +"298 linear_1_zero_point_0" -> "300 dequantize_per_channel_default_2" [label="(3,)", style=solid]; +"299 _frozen_param3" -> "300 dequantize_per_channel_default_2" [label="(3, 512)", style=solid]; +"300 dequantize_per_channel_default_2" -> "301 linear_1" [label="(3, 512)", style=solid]; +"301 linear_1" -> "302 view" [label="(1, 15, 15, 3)", style=solid]; +"302 view" -> "303 index" [label="(225, 3)", style=solid]; +"303 index" -> "304 view_1" [label="(4096, 3)", style=solid]; +"304 view_1" -> "305 permute_1" [label="(64, 64, 3)", style=solid]; +"305 permute_1" -> "306 contiguous" [label="(3, 64, 64)", style=solid]; +"306 contiguous" -> "307 unsqueeze" [label="(3, 64, 64)", style=solid]; +"307 unsqueeze" -> "308 sigmoid" [label="(1, 3, 64, 64)", style=solid]; +"308 sigmoid" -> "309 mul" [label="(1, 3, 64, 64)", style=solid]; +"309 mul" -> "336 add" [label="(1, 3, 64, 64)", style=solid]; +"310 pad" -> "311 view_2" [label="(1, 56, 56, 96)", style=solid]; +"311 view_2" -> "312 permute_2" [label="(1, 7, 8, 7, 8, 96)", style=solid]; +"312 permute_2" -> "313 reshape" [label="(1, 7, 7, 8, 8, 96)", style=solid]; +"313 reshape" -> "316 reshape_0_0_nncf_smooth_quant_0" [label="(49, 64, 96)", style=solid]; +"314 clone" -> "317 linear_2" [label="(288,)", style=solid]; +"315 features_1_0_attn_qkv_weight_updated_constant0" -> "317 linear_2" [label="(288, 96)", style=solid]; +"316 reshape_0_0_nncf_smooth_quant_0" -> "317 linear_2" [label="(49, 64, 96)", style=solid]; +"317 linear_2" -> "318 reshape_1" [label="(49, 64, 288)", style=solid]; +"318 reshape_1" -> "319 permute_3" [label="(49, 64, 3, 3, 32)", style=solid]; +"319 permute_3" -> "320 select" [label="(3, 49, 3, 64, 32)", style=solid]; +"319 permute_3" -> "321 select_1" [label="(3, 49, 3, 64, 32)", style=solid]; +"319 permute_3" -> "322 select_2" [label="(3, 49, 3, 64, 32)", style=solid]; +"320 select" -> "323 linalg_vector_norm" [label="(49, 3, 64, 32)", style=solid]; +"320 select" -> "325 expand_as" [label="(49, 3, 64, 32)", style=solid]; +"320 select" -> "326 div" [label="(49, 3, 64, 32)", style=solid]; +"321 select_1" -> "327 linalg_vector_norm_1" [label="(49, 3, 64, 32)", style=solid]; +"321 select_1" -> "329 expand_as_1" [label="(49, 3, 64, 32)", style=solid]; +"321 select_1" -> "330 div_1" [label="(49, 3, 64, 32)", style=solid]; +"322 select_2" -> "339 matmul_1" [label="(49, 3, 64, 32)", style=solid]; +"323 linalg_vector_norm" -> "324 clamp_min" [label="(49, 3, 64, 1)", style=solid]; +"324 clamp_min" -> "325 expand_as" [label="(49, 3, 64, 1)", style=solid]; +"325 expand_as" -> "326 div" [label="(49, 3, 64, 32)", style=solid]; +"326 div" -> "332 matmul" [label="(49, 3, 64, 32)", style=solid]; +"327 linalg_vector_norm_1" -> "328 clamp_min_1" [label="(49, 3, 64, 1)", style=solid]; +"328 clamp_min_1" -> "329 expand_as_1" [label="(49, 3, 64, 1)", style=solid]; +"329 expand_as_1" -> "330 div_1" [label="(49, 3, 64, 32)", style=solid]; +"330 div_1" -> "331 transpose" [label="(49, 3, 64, 32)", style=solid]; +"331 transpose" -> "332 matmul" [label="(49, 3, 32, 64)", style=solid]; +"332 matmul" -> "335 mul_1" [label="(49, 3, 64, 64)", style=solid]; +"333 clamp" -> "334 exp" [label="(3, 1, 1)", style=solid]; +"334 exp" -> "335 mul_1" [label="(3, 1, 1)", style=solid]; +"335 mul_1" -> "336 add" [label="(49, 3, 64, 64)", style=solid]; +"336 add" -> "337 softmax" [label="(49, 3, 64, 64)", style=solid]; +"337 softmax" -> "338 dropout" [label="(49, 3, 64, 64)", style=solid]; +"338 dropout" -> "339 matmul_1" [label="(49, 3, 64, 64)", style=solid]; +"339 matmul_1" -> "340 transpose_1" [label="(49, 3, 64, 32)", style=solid]; +"340 transpose_1" -> "341 reshape_2" [label="(49, 64, 3, 32)", style=solid]; +"341 reshape_2" -> "343 reshape_2_0_0_nncf_smooth_quant_0" [label="(49, 64, 96)", style=solid]; +"342 features_1_0_attn_proj_weight_updated_constant0" -> "344 linear_3" [label="(96, 96)", style=solid]; +"343 reshape_2_0_0_nncf_smooth_quant_0" -> "344 linear_3" [label="(49, 64, 96)", style=solid]; +"344 linear_3" -> "345 dropout_1" [label="(49, 64, 96)", style=solid]; +"345 dropout_1" -> "346 view_3" [label="(49, 64, 96)", style=solid]; +"346 view_3" -> "347 permute_4" [label="(1, 7, 7, 8, 8, 96)", style=solid]; +"347 permute_4" -> "348 reshape_3" [label="(1, 7, 8, 7, 8, 96)", style=solid]; +"348 reshape_3" -> "349 slice_2" [label="(1, 56, 56, 96)", style=solid]; +"349 slice_2" -> "350 slice_3" [label="(1, 56, 56, 96)", style=solid]; +"350 slice_3" -> "351 layer_norm_1" [label="(1, 56, 56, 96)", style=solid]; +"351 layer_norm_1" -> "352 add_1" [label="(1, 56, 56, 96)", style=solid]; +"352 add_1" -> "353 add_1_0_0_nncf_smooth_quant_0" [label="(1, 56, 56, 96)", style=solid]; +"352 add_1" -> "373 add_2" [label="(1, 56, 56, 96)", style=solid]; +"353 add_1_0_0_nncf_smooth_quant_0" -> "354 quantize_per_tensor_default_3" [label="(1, 56, 56, 96)", style=solid]; +"354 quantize_per_tensor_default_3" -> "355 dequantize_per_tensor_default_3" [label="(1, 56, 56, 96)", style=solid]; +"355 dequantize_per_tensor_default_3" -> "360 linear_4" [label="(1, 56, 56, 96)", style=solid]; +"356 linear_4_scale_0" -> "359 dequantize_per_channel_default_3" [label="(384,)", style=solid]; +"357 linear_4_zero_point_0" -> "359 dequantize_per_channel_default_3" [label="(384,)", style=solid]; +"358 _frozen_param4" -> "359 dequantize_per_channel_default_3" [label="(384, 96)", style=solid]; +"359 dequantize_per_channel_default_3" -> "360 linear_4" [label="(384, 96)", style=solid]; +"360 linear_4" -> "361 gelu" [label="(1, 56, 56, 384)", style=solid]; +"361 gelu" -> "362 dropout_2" [label="(1, 56, 56, 384)", style=solid]; +"362 dropout_2" -> "363 dropout_2_0_0_nncf_smooth_quant_0" [label="(1, 56, 56, 384)", style=solid]; +"363 dropout_2_0_0_nncf_smooth_quant_0" -> "364 quantize_per_tensor_default_4" [label="(1, 56, 56, 384)", style=solid]; +"364 quantize_per_tensor_default_4" -> "365 dequantize_per_tensor_default_4" [label="(1, 56, 56, 384)", style=solid]; +"365 dequantize_per_tensor_default_4" -> "370 linear_5" [label="(1, 56, 56, 384)", style=solid]; +"366 linear_5_scale_0" -> "369 dequantize_per_channel_default_4" [label="(96,)", style=solid]; +"367 linear_5_zero_point_0" -> "369 dequantize_per_channel_default_4" [label="(96,)", style=solid]; +"368 _frozen_param5" -> "369 dequantize_per_channel_default_4" [label="(96, 384)", style=solid]; +"369 dequantize_per_channel_default_4" -> "370 linear_5" [label="(96, 384)", style=solid]; +"370 linear_5" -> "371 dropout_3" [label="(1, 56, 56, 96)", style=solid]; +"371 dropout_3" -> "372 layer_norm_2" [label="(1, 56, 56, 96)", style=solid]; +"372 layer_norm_2" -> "373 add_2" [label="(1, 56, 56, 96)", style=solid]; +"373 add_2" -> "398 pad_1" [label="(1, 56, 56, 96)", style=solid]; +"373 add_2" -> "458 add_5" [label="(1, 56, 56, 96)", style=solid]; +"374 _frozen_param6" -> "375 dequantize_per_tensor_default_5" [label="(1, 15, 15, 2)", style=solid]; +"375 dequantize_per_tensor_default_5" -> "380 linear_6" [label="(1, 15, 15, 2)", style=solid]; +"376 linear_6_scale_0" -> "379 dequantize_per_channel_default_5" [label="(512,)", style=solid]; +"377 linear_6_zero_point_0" -> "379 dequantize_per_channel_default_5" [label="(512,)", style=solid]; +"378 _frozen_param7" -> "379 dequantize_per_channel_default_5" [label="(512, 2)", style=solid]; +"379 dequantize_per_channel_default_5" -> "380 linear_6" [label="(512, 2)", style=solid]; +"380 linear_6" -> "381 relu__1" [label="(1, 15, 15, 512)", style=solid]; +"381 relu__1" -> "382 relu__1_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"382 relu__1_0_0_nncf_smooth_quant_0" -> "383 quantize_per_tensor_default_6" [label="(1, 15, 15, 512)", style=solid]; +"383 quantize_per_tensor_default_6" -> "384 dequantize_per_tensor_default_6" [label="(1, 15, 15, 512)", style=solid]; +"384 dequantize_per_tensor_default_6" -> "389 linear_7" [label="(1, 15, 15, 512)", style=solid]; +"385 linear_7_scale_0" -> "388 dequantize_per_channel_default_6" [label="(3,)", style=solid]; +"386 linear_7_zero_point_0" -> "388 dequantize_per_channel_default_6" [label="(3,)", style=solid]; +"387 _frozen_param8" -> "388 dequantize_per_channel_default_6" [label="(3, 512)", style=solid]; +"388 dequantize_per_channel_default_6" -> "389 linear_7" [label="(3, 512)", style=solid]; +"389 linear_7" -> "390 view_4" [label="(1, 15, 15, 3)", style=solid]; +"390 view_4" -> "391 index_1" [label="(225, 3)", style=solid]; +"391 index_1" -> "392 view_5" [label="(4096, 3)", style=solid]; +"392 view_5" -> "393 permute_5" [label="(64, 64, 3)", style=solid]; +"393 permute_5" -> "394 contiguous_1" [label="(3, 64, 64)", style=solid]; +"394 contiguous_1" -> "395 unsqueeze_1" [label="(3, 64, 64)", style=solid]; +"395 unsqueeze_1" -> "396 sigmoid_1" [label="(1, 3, 64, 64)", style=solid]; +"396 sigmoid_1" -> "397 mul_2" [label="(1, 3, 64, 64)", style=solid]; +"397 mul_2" -> "425 add_3" [label="(1, 3, 64, 64)", style=solid]; +"398 pad_1" -> "399 roll" [label="(1, 56, 56, 96)", style=solid]; +"399 roll" -> "400 view_6" [label="(1, 56, 56, 96)", style=solid]; +"400 view_6" -> "401 permute_6" [label="(1, 7, 8, 7, 8, 96)", style=solid]; +"401 permute_6" -> "402 reshape_4" [label="(1, 7, 7, 8, 8, 96)", style=solid]; +"402 reshape_4" -> "405 reshape_4_0_0_nncf_smooth_quant_0" [label="(49, 64, 96)", style=solid]; +"402 reshape_4" -> "426 new_zeros" [label="(49, 64, 96)", style=solid]; +"403 clone_1" -> "406 linear_8" [label="(288,)", style=solid]; +"404 features_1_1_attn_qkv_weight_updated_constant0" -> "406 linear_8" [label="(288, 96)", style=solid]; +"405 reshape_4_0_0_nncf_smooth_quant_0" -> "406 linear_8" [label="(49, 64, 96)", style=solid]; +"406 linear_8" -> "407 reshape_5" [label="(49, 64, 288)", style=solid]; +"407 reshape_5" -> "408 permute_7" [label="(49, 64, 3, 3, 32)", style=solid]; +"408 permute_7" -> "409 select_3" [label="(3, 49, 3, 64, 32)", style=solid]; +"408 permute_7" -> "410 select_4" [label="(3, 49, 3, 64, 32)", style=solid]; +"408 permute_7" -> "411 select_5" [label="(3, 49, 3, 64, 32)", style=solid]; +"409 select_3" -> "412 linalg_vector_norm_2" [label="(49, 3, 64, 32)", style=solid]; +"409 select_3" -> "414 expand_as_2" [label="(49, 3, 64, 32)", style=solid]; +"409 select_3" -> "415 div_2" [label="(49, 3, 64, 32)", style=solid]; +"410 select_4" -> "416 linalg_vector_norm_3" [label="(49, 3, 64, 32)", style=solid]; +"410 select_4" -> "418 expand_as_3" [label="(49, 3, 64, 32)", style=solid]; +"410 select_4" -> "419 div_3" [label="(49, 3, 64, 32)", style=solid]; +"411 select_5" -> "444 matmul_3" [label="(49, 3, 64, 32)", style=solid]; +"412 linalg_vector_norm_2" -> "413 clamp_min_2" [label="(49, 3, 64, 1)", style=solid]; +"413 clamp_min_2" -> "414 expand_as_2" [label="(49, 3, 64, 1)", style=solid]; +"414 expand_as_2" -> "415 div_2" [label="(49, 3, 64, 32)", style=solid]; +"415 div_2" -> "421 matmul_2" [label="(49, 3, 64, 32)", style=solid]; +"416 linalg_vector_norm_3" -> "417 clamp_min_3" [label="(49, 3, 64, 1)", style=solid]; +"417 clamp_min_3" -> "418 expand_as_3" [label="(49, 3, 64, 1)", style=solid]; +"418 expand_as_3" -> "419 div_3" [label="(49, 3, 64, 32)", style=solid]; +"419 div_3" -> "420 transpose_2" [label="(49, 3, 64, 32)", style=solid]; +"420 transpose_2" -> "421 matmul_2" [label="(49, 3, 32, 64)", style=solid]; +"421 matmul_2" -> "424 mul_3" [label="(49, 3, 64, 64)", style=solid]; +"422 clamp_1" -> "423 exp_1" [label="(3, 1, 1)", style=solid]; +"423 exp_1" -> "424 mul_3" [label="(3, 1, 1)", style=solid]; +"424 mul_3" -> "425 add_3" [label="(49, 3, 64, 64)", style=solid]; +"425 add_3" -> "437 view_8" [label="(49, 3, 64, 64)", style=solid]; +"426 new_zeros" -> "427 view_7" [label="(56, 56)", style=solid]; +"427 view_7" -> "428 permute_8" [label="(7, 8, 7, 8)", style=solid]; +"428 permute_8" -> "429 reshape_6" [label="(7, 7, 8, 8)", style=solid]; +"429 reshape_6" -> "430 unsqueeze_2" [label="(49, 64)", style=solid]; +"429 reshape_6" -> "431 unsqueeze_3" [label="(49, 64)", style=solid]; +"430 unsqueeze_2" -> "432 sub" [label="(49, 1, 64)", style=solid]; +"431 unsqueeze_3" -> "432 sub" [label="(49, 64, 1)", style=solid]; +"432 sub" -> "433 ne" [label="(49, 64, 64)", style=solid]; +"432 sub" -> "434 masked_fill" [label="(49, 64, 64)", style=solid]; +"432 sub" -> "435 eq" [label="(49, 64, 64)", style=solid]; +"433 ne" -> "434 masked_fill" [label="(49, 64, 64)", style=solid]; +"434 masked_fill" -> "436 masked_fill_1" [label="(49, 64, 64)", style=solid]; +"435 eq" -> "436 masked_fill_1" [label="(49, 64, 64)", style=solid]; +"436 masked_fill_1" -> "438 unsqueeze_4" [label="(49, 64, 64)", style=solid]; +"437 view_8" -> "440 add_4" [label="(1, 49, 3, 64, 64)", style=solid]; +"438 unsqueeze_4" -> "439 unsqueeze_5" [label="(49, 1, 64, 64)", style=solid]; +"439 unsqueeze_5" -> "440 add_4" [label="(1, 49, 1, 64, 64)", style=solid]; +"440 add_4" -> "441 view_9" [label="(1, 49, 3, 64, 64)", style=solid]; +"441 view_9" -> "442 softmax_1" [label="(49, 3, 64, 64)", style=solid]; +"442 softmax_1" -> "443 dropout_4" [label="(49, 3, 64, 64)", style=solid]; +"443 dropout_4" -> "444 matmul_3" [label="(49, 3, 64, 64)", style=solid]; +"444 matmul_3" -> "445 transpose_3" [label="(49, 3, 64, 32)", style=solid]; +"445 transpose_3" -> "446 reshape_7" [label="(49, 64, 3, 32)", style=solid]; +"446 reshape_7" -> "448 reshape_7_0_0_nncf_smooth_quant_0" [label="(49, 64, 96)", style=solid]; +"447 features_1_1_attn_proj_weight_updated_constant0" -> "449 linear_9" [label="(96, 96)", style=solid]; +"448 reshape_7_0_0_nncf_smooth_quant_0" -> "449 linear_9" [label="(49, 64, 96)", style=solid]; +"449 linear_9" -> "450 dropout_5" [label="(49, 64, 96)", style=solid]; +"450 dropout_5" -> "451 view_10" [label="(49, 64, 96)", style=solid]; +"451 view_10" -> "452 permute_9" [label="(1, 7, 7, 8, 8, 96)", style=solid]; +"452 permute_9" -> "453 reshape_8" [label="(1, 7, 8, 7, 8, 96)", style=solid]; +"453 reshape_8" -> "454 roll_1" [label="(1, 56, 56, 96)", style=solid]; +"454 roll_1" -> "455 slice_23" [label="(1, 56, 56, 96)", style=solid]; +"455 slice_23" -> "456 slice_24" [label="(1, 56, 56, 96)", style=solid]; +"456 slice_24" -> "457 layer_norm_3" [label="(1, 56, 56, 96)", style=solid]; +"457 layer_norm_3" -> "458 add_5" [label="(1, 56, 56, 96)", style=solid]; +"458 add_5" -> "459 add_5_0_0_nncf_smooth_quant_0" [label="(1, 56, 56, 96)", style=solid]; +"458 add_5" -> "479 add_6" [label="(1, 56, 56, 96)", style=solid]; +"459 add_5_0_0_nncf_smooth_quant_0" -> "460 quantize_per_tensor_default_7" [label="(1, 56, 56, 96)", style=solid]; +"460 quantize_per_tensor_default_7" -> "461 dequantize_per_tensor_default_7" [label="(1, 56, 56, 96)", style=solid]; +"461 dequantize_per_tensor_default_7" -> "466 linear_10" [label="(1, 56, 56, 96)", style=solid]; +"462 linear_10_scale_0" -> "465 dequantize_per_channel_default_7" [label="(384,)", style=solid]; +"463 linear_10_zero_point_0" -> "465 dequantize_per_channel_default_7" [label="(384,)", style=solid]; +"464 _frozen_param9" -> "465 dequantize_per_channel_default_7" [label="(384, 96)", style=solid]; +"465 dequantize_per_channel_default_7" -> "466 linear_10" [label="(384, 96)", style=solid]; +"466 linear_10" -> "467 gelu_1" [label="(1, 56, 56, 384)", style=solid]; +"467 gelu_1" -> "468 dropout_6" [label="(1, 56, 56, 384)", style=solid]; +"468 dropout_6" -> "469 dropout_6_0_0_nncf_smooth_quant_0" [label="(1, 56, 56, 384)", style=solid]; +"469 dropout_6_0_0_nncf_smooth_quant_0" -> "470 quantize_per_tensor_default_8" [label="(1, 56, 56, 384)", style=solid]; +"470 quantize_per_tensor_default_8" -> "471 dequantize_per_tensor_default_8" [label="(1, 56, 56, 384)", style=solid]; +"471 dequantize_per_tensor_default_8" -> "476 linear_11" [label="(1, 56, 56, 384)", style=solid]; +"472 linear_11_scale_0" -> "475 dequantize_per_channel_default_8" [label="(96,)", style=solid]; +"473 linear_11_zero_point_0" -> "475 dequantize_per_channel_default_8" [label="(96,)", style=solid]; +"474 _frozen_param10" -> "475 dequantize_per_channel_default_8" [label="(96, 384)", style=solid]; +"475 dequantize_per_channel_default_8" -> "476 linear_11" [label="(96, 384)", style=solid]; +"476 linear_11" -> "477 dropout_7" [label="(1, 56, 56, 96)", style=solid]; +"477 dropout_7" -> "478 layer_norm_4" [label="(1, 56, 56, 96)", style=solid]; +"478 layer_norm_4" -> "479 add_6" [label="(1, 56, 56, 96)", style=solid]; +"479 add_6" -> "480 pad_2" [label="(1, 56, 56, 96)", style=solid]; +"480 pad_2" -> "481 slice_25" [label="(1, 56, 56, 96)", style=solid]; +"480 pad_2" -> "484 slice_28" [label="(1, 56, 56, 96)", style=solid]; +"480 pad_2" -> "487 slice_31" [label="(1, 56, 56, 96)", style=solid]; +"480 pad_2" -> "490 slice_34" [label="(1, 56, 56, 96)", style=solid]; +"481 slice_25" -> "482 slice_26" [label="(1, 28, 56, 96)", style=solid]; +"482 slice_26" -> "483 slice_27" [label="(1, 28, 28, 96)", style=solid]; +"483 slice_27" -> "493 cat" [label="(1, 28, 28, 96)", style=solid]; +"484 slice_28" -> "485 slice_29" [label="(1, 28, 56, 96)", style=solid]; +"485 slice_29" -> "486 slice_30" [label="(1, 28, 28, 96)", style=solid]; +"486 slice_30" -> "493 cat" [label="(1, 28, 28, 96)", style=solid]; +"487 slice_31" -> "488 slice_32" [label="(1, 28, 56, 96)", style=solid]; +"488 slice_32" -> "489 slice_33" [label="(1, 28, 28, 96)", style=solid]; +"489 slice_33" -> "493 cat" [label="(1, 28, 28, 96)", style=solid]; +"490 slice_34" -> "491 slice_35" [label="(1, 28, 56, 96)", style=solid]; +"491 slice_35" -> "492 slice_36" [label="(1, 28, 28, 96)", style=solid]; +"492 slice_36" -> "493 cat" [label="(1, 28, 28, 96)", style=solid]; +"493 cat" -> "494 cat_0_0_nncf_smooth_quant_0" [label="(1, 28, 28, 384)", style=solid]; +"494 cat_0_0_nncf_smooth_quant_0" -> "495 quantize_per_tensor_default_9" [label="(1, 28, 28, 384)", style=solid]; +"495 quantize_per_tensor_default_9" -> "496 dequantize_per_tensor_default_9" [label="(1, 28, 28, 384)", style=solid]; +"496 dequantize_per_tensor_default_9" -> "501 linear_12" [label="(1, 28, 28, 384)", style=solid]; +"497 linear_12_scale_0" -> "500 dequantize_per_channel_default_9" [label="(192,)", style=solid]; +"498 linear_12_zero_point_0" -> "500 dequantize_per_channel_default_9" [label="(192,)", style=solid]; +"499 _frozen_param11" -> "500 dequantize_per_channel_default_9" [label="(192, 384)", style=solid]; +"500 dequantize_per_channel_default_9" -> "501 linear_12" [label="(192, 384)", style=solid]; +"501 linear_12" -> "502 layer_norm_5" [label="(1, 28, 28, 192)", style=solid]; +"502 layer_norm_5" -> "527 pad_3" [label="(1, 28, 28, 192)", style=solid]; +"502 layer_norm_5" -> "572 add_8" [label="(1, 28, 28, 192)", style=solid]; +"503 _frozen_param12" -> "504 dequantize_per_tensor_default_10" [label="(1, 15, 15, 2)", style=solid]; +"504 dequantize_per_tensor_default_10" -> "509 linear_13" [label="(1, 15, 15, 2)", style=solid]; +"505 linear_13_scale_0" -> "508 dequantize_per_channel_default_10" [label="(512,)", style=solid]; +"506 linear_13_zero_point_0" -> "508 dequantize_per_channel_default_10" [label="(512,)", style=solid]; +"507 _frozen_param13" -> "508 dequantize_per_channel_default_10" [label="(512, 2)", style=solid]; +"508 dequantize_per_channel_default_10" -> "509 linear_13" [label="(512, 2)", style=solid]; +"509 linear_13" -> "510 relu__2" [label="(1, 15, 15, 512)", style=solid]; +"510 relu__2" -> "511 relu__2_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"511 relu__2_0_0_nncf_smooth_quant_0" -> "512 quantize_per_tensor_default_11" [label="(1, 15, 15, 512)", style=solid]; +"512 quantize_per_tensor_default_11" -> "513 dequantize_per_tensor_default_11" [label="(1, 15, 15, 512)", style=solid]; +"513 dequantize_per_tensor_default_11" -> "518 linear_14" [label="(1, 15, 15, 512)", style=solid]; +"514 linear_14_scale_0" -> "517 dequantize_per_channel_default_11" [label="(6,)", style=solid]; +"515 linear_14_zero_point_0" -> "517 dequantize_per_channel_default_11" [label="(6,)", style=solid]; +"516 _frozen_param14" -> "517 dequantize_per_channel_default_11" [label="(6, 512)", style=solid]; +"517 dequantize_per_channel_default_11" -> "518 linear_14" [label="(6, 512)", style=solid]; +"518 linear_14" -> "519 view_11" [label="(1, 15, 15, 6)", style=solid]; +"519 view_11" -> "520 index_2" [label="(225, 6)", style=solid]; +"520 index_2" -> "521 view_12" [label="(4096, 6)", style=solid]; +"521 view_12" -> "522 permute_10" [label="(64, 64, 6)", style=solid]; +"522 permute_10" -> "523 contiguous_2" [label="(6, 64, 64)", style=solid]; +"523 contiguous_2" -> "524 unsqueeze_6" [label="(6, 64, 64)", style=solid]; +"524 unsqueeze_6" -> "525 sigmoid_2" [label="(1, 6, 64, 64)", style=solid]; +"525 sigmoid_2" -> "526 mul_4" [label="(1, 6, 64, 64)", style=solid]; +"526 mul_4" -> "553 add_7" [label="(1, 6, 64, 64)", style=solid]; +"527 pad_3" -> "528 view_13" [label="(1, 32, 32, 192)", style=solid]; +"528 view_13" -> "529 permute_11" [label="(1, 4, 8, 4, 8, 192)", style=solid]; +"529 permute_11" -> "530 reshape_9" [label="(1, 4, 4, 8, 8, 192)", style=solid]; +"530 reshape_9" -> "533 reshape_9_0_0_nncf_smooth_quant_0" [label="(16, 64, 192)", style=solid]; +"531 clone_2" -> "534 linear_15" [label="(576,)", style=solid]; +"532 features_3_0_attn_qkv_weight_updated_constant0" -> "534 linear_15" [label="(576, 192)", style=solid]; +"533 reshape_9_0_0_nncf_smooth_quant_0" -> "534 linear_15" [label="(16, 64, 192)", style=solid]; +"534 linear_15" -> "535 reshape_10" [label="(16, 64, 576)", style=solid]; +"535 reshape_10" -> "536 permute_12" [label="(16, 64, 3, 6, 32)", style=solid]; +"536 permute_12" -> "537 select_6" [label="(3, 16, 6, 64, 32)", style=solid]; +"536 permute_12" -> "538 select_7" [label="(3, 16, 6, 64, 32)", style=solid]; +"536 permute_12" -> "539 select_8" [label="(3, 16, 6, 64, 32)", style=solid]; +"537 select_6" -> "540 linalg_vector_norm_4" [label="(16, 6, 64, 32)", style=solid]; +"537 select_6" -> "542 expand_as_4" [label="(16, 6, 64, 32)", style=solid]; +"537 select_6" -> "543 div_4" [label="(16, 6, 64, 32)", style=solid]; +"538 select_7" -> "544 linalg_vector_norm_5" [label="(16, 6, 64, 32)", style=solid]; +"538 select_7" -> "546 expand_as_5" [label="(16, 6, 64, 32)", style=solid]; +"538 select_7" -> "547 div_5" [label="(16, 6, 64, 32)", style=solid]; +"539 select_8" -> "556 matmul_5" [label="(16, 6, 64, 32)", style=solid]; +"540 linalg_vector_norm_4" -> "541 clamp_min_4" [label="(16, 6, 64, 1)", style=solid]; +"541 clamp_min_4" -> "542 expand_as_4" [label="(16, 6, 64, 1)", style=solid]; +"542 expand_as_4" -> "543 div_4" [label="(16, 6, 64, 32)", style=solid]; +"543 div_4" -> "549 matmul_4" [label="(16, 6, 64, 32)", style=solid]; +"544 linalg_vector_norm_5" -> "545 clamp_min_5" [label="(16, 6, 64, 1)", style=solid]; +"545 clamp_min_5" -> "546 expand_as_5" [label="(16, 6, 64, 1)", style=solid]; +"546 expand_as_5" -> "547 div_5" [label="(16, 6, 64, 32)", style=solid]; +"547 div_5" -> "548 transpose_4" [label="(16, 6, 64, 32)", style=solid]; +"548 transpose_4" -> "549 matmul_4" [label="(16, 6, 32, 64)", style=solid]; +"549 matmul_4" -> "552 mul_5" [label="(16, 6, 64, 64)", style=solid]; +"550 clamp_2" -> "551 exp_2" [label="(6, 1, 1)", style=solid]; +"551 exp_2" -> "552 mul_5" [label="(6, 1, 1)", style=solid]; +"552 mul_5" -> "553 add_7" [label="(16, 6, 64, 64)", style=solid]; +"553 add_7" -> "554 softmax_2" [label="(16, 6, 64, 64)", style=solid]; +"554 softmax_2" -> "555 dropout_8" [label="(16, 6, 64, 64)", style=solid]; +"555 dropout_8" -> "556 matmul_5" [label="(16, 6, 64, 64)", style=solid]; +"556 matmul_5" -> "557 transpose_5" [label="(16, 6, 64, 32)", style=solid]; +"557 transpose_5" -> "558 reshape_11" [label="(16, 64, 6, 32)", style=solid]; +"558 reshape_11" -> "560 reshape_11_0_0_nncf_smooth_quant_0" [label="(16, 64, 192)", style=solid]; +"559 features_3_0_attn_proj_weight_updated_constant0" -> "561 linear_16" [label="(192, 192)", style=solid]; +"560 reshape_11_0_0_nncf_smooth_quant_0" -> "561 linear_16" [label="(16, 64, 192)", style=solid]; +"561 linear_16" -> "562 dropout_9" [label="(16, 64, 192)", style=solid]; +"562 dropout_9" -> "563 view_14" [label="(16, 64, 192)", style=solid]; +"563 view_14" -> "564 permute_13" [label="(1, 4, 4, 8, 8, 192)", style=solid]; +"564 permute_13" -> "565 reshape_12" [label="(1, 4, 8, 4, 8, 192)", style=solid]; +"565 reshape_12" -> "566 slice_38" [label="(1, 32, 32, 192)", style=solid]; +"566 slice_38" -> "567 slice_39" [label="(1, 32, 32, 192)", style=solid]; +"567 slice_39" -> "568 slice_40" [label="(1, 28, 32, 192)", style=solid]; +"568 slice_40" -> "569 slice_41" [label="(1, 28, 28, 192)", style=solid]; +"569 slice_41" -> "570 contiguous_3" [label="(1, 28, 28, 192)", style=solid]; +"570 contiguous_3" -> "571 layer_norm_6" [label="(1, 28, 28, 192)", style=solid]; +"571 layer_norm_6" -> "572 add_8" [label="(1, 28, 28, 192)", style=solid]; +"572 add_8" -> "573 add_8_0_0_nncf_smooth_quant_0" [label="(1, 28, 28, 192)", style=solid]; +"572 add_8" -> "593 add_9" [label="(1, 28, 28, 192)", style=solid]; +"573 add_8_0_0_nncf_smooth_quant_0" -> "574 quantize_per_tensor_default_12" [label="(1, 28, 28, 192)", style=solid]; +"574 quantize_per_tensor_default_12" -> "575 dequantize_per_tensor_default_12" [label="(1, 28, 28, 192)", style=solid]; +"575 dequantize_per_tensor_default_12" -> "580 linear_17" [label="(1, 28, 28, 192)", style=solid]; +"576 linear_17_scale_0" -> "579 dequantize_per_channel_default_12" [label="(768,)", style=solid]; +"577 linear_17_zero_point_0" -> "579 dequantize_per_channel_default_12" [label="(768,)", style=solid]; +"578 _frozen_param15" -> "579 dequantize_per_channel_default_12" [label="(768, 192)", style=solid]; +"579 dequantize_per_channel_default_12" -> "580 linear_17" [label="(768, 192)", style=solid]; +"580 linear_17" -> "581 gelu_2" [label="(1, 28, 28, 768)", style=solid]; +"581 gelu_2" -> "582 dropout_10" [label="(1, 28, 28, 768)", style=solid]; +"582 dropout_10" -> "583 dropout_10_0_0_nncf_smooth_quant_0" [label="(1, 28, 28, 768)", style=solid]; +"583 dropout_10_0_0_nncf_smooth_quant_0" -> "584 quantize_per_tensor_default_13" [label="(1, 28, 28, 768)", style=solid]; +"584 quantize_per_tensor_default_13" -> "585 dequantize_per_tensor_default_13" [label="(1, 28, 28, 768)", style=solid]; +"585 dequantize_per_tensor_default_13" -> "590 linear_18" [label="(1, 28, 28, 768)", style=solid]; +"586 linear_18_scale_0" -> "589 dequantize_per_channel_default_13" [label="(192,)", style=solid]; +"587 linear_18_zero_point_0" -> "589 dequantize_per_channel_default_13" [label="(192,)", style=solid]; +"588 _frozen_param16" -> "589 dequantize_per_channel_default_13" [label="(192, 768)", style=solid]; +"589 dequantize_per_channel_default_13" -> "590 linear_18" [label="(192, 768)", style=solid]; +"590 linear_18" -> "591 dropout_11" [label="(1, 28, 28, 192)", style=solid]; +"591 dropout_11" -> "592 layer_norm_7" [label="(1, 28, 28, 192)", style=solid]; +"592 layer_norm_7" -> "593 add_9" [label="(1, 28, 28, 192)", style=solid]; +"593 add_9" -> "618 pad_4" [label="(1, 28, 28, 192)", style=solid]; +"593 add_9" -> "681 add_12" [label="(1, 28, 28, 192)", style=solid]; +"594 _frozen_param17" -> "595 dequantize_per_tensor_default_14" [label="(1, 15, 15, 2)", style=solid]; +"595 dequantize_per_tensor_default_14" -> "600 linear_19" [label="(1, 15, 15, 2)", style=solid]; +"596 linear_19_scale_0" -> "599 dequantize_per_channel_default_14" [label="(512,)", style=solid]; +"597 linear_19_zero_point_0" -> "599 dequantize_per_channel_default_14" [label="(512,)", style=solid]; +"598 _frozen_param18" -> "599 dequantize_per_channel_default_14" [label="(512, 2)", style=solid]; +"599 dequantize_per_channel_default_14" -> "600 linear_19" [label="(512, 2)", style=solid]; +"600 linear_19" -> "601 relu__3" [label="(1, 15, 15, 512)", style=solid]; +"601 relu__3" -> "602 relu__3_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"602 relu__3_0_0_nncf_smooth_quant_0" -> "603 quantize_per_tensor_default_15" [label="(1, 15, 15, 512)", style=solid]; +"603 quantize_per_tensor_default_15" -> "604 dequantize_per_tensor_default_15" [label="(1, 15, 15, 512)", style=solid]; +"604 dequantize_per_tensor_default_15" -> "609 linear_20" [label="(1, 15, 15, 512)", style=solid]; +"605 linear_20_scale_0" -> "608 dequantize_per_channel_default_15" [label="(6,)", style=solid]; +"606 linear_20_zero_point_0" -> "608 dequantize_per_channel_default_15" [label="(6,)", style=solid]; +"607 _frozen_param19" -> "608 dequantize_per_channel_default_15" [label="(6, 512)", style=solid]; +"608 dequantize_per_channel_default_15" -> "609 linear_20" [label="(6, 512)", style=solid]; +"609 linear_20" -> "610 view_15" [label="(1, 15, 15, 6)", style=solid]; +"610 view_15" -> "611 index_3" [label="(225, 6)", style=solid]; +"611 index_3" -> "612 view_16" [label="(4096, 6)", style=solid]; +"612 view_16" -> "613 permute_14" [label="(64, 64, 6)", style=solid]; +"613 permute_14" -> "614 contiguous_4" [label="(6, 64, 64)", style=solid]; +"614 contiguous_4" -> "615 unsqueeze_7" [label="(6, 64, 64)", style=solid]; +"615 unsqueeze_7" -> "616 sigmoid_3" [label="(1, 6, 64, 64)", style=solid]; +"616 sigmoid_3" -> "617 mul_6" [label="(1, 6, 64, 64)", style=solid]; +"617 mul_6" -> "645 add_10" [label="(1, 6, 64, 64)", style=solid]; +"618 pad_4" -> "619 roll_2" [label="(1, 32, 32, 192)", style=solid]; +"619 roll_2" -> "620 view_17" [label="(1, 32, 32, 192)", style=solid]; +"620 view_17" -> "621 permute_15" [label="(1, 4, 8, 4, 8, 192)", style=solid]; +"621 permute_15" -> "622 reshape_13" [label="(1, 4, 4, 8, 8, 192)", style=solid]; +"622 reshape_13" -> "625 reshape_13_0_0_nncf_smooth_quant_0" [label="(16, 64, 192)", style=solid]; +"622 reshape_13" -> "646 new_zeros_1" [label="(16, 64, 192)", style=solid]; +"623 clone_3" -> "626 linear_21" [label="(576,)", style=solid]; +"624 features_3_1_attn_qkv_weight_updated_constant0" -> "626 linear_21" [label="(576, 192)", style=solid]; +"625 reshape_13_0_0_nncf_smooth_quant_0" -> "626 linear_21" [label="(16, 64, 192)", style=solid]; +"626 linear_21" -> "627 reshape_14" [label="(16, 64, 576)", style=solid]; +"627 reshape_14" -> "628 permute_16" [label="(16, 64, 3, 6, 32)", style=solid]; +"628 permute_16" -> "629 select_9" [label="(3, 16, 6, 64, 32)", style=solid]; +"628 permute_16" -> "630 select_10" [label="(3, 16, 6, 64, 32)", style=solid]; +"628 permute_16" -> "631 select_11" [label="(3, 16, 6, 64, 32)", style=solid]; +"629 select_9" -> "632 linalg_vector_norm_6" [label="(16, 6, 64, 32)", style=solid]; +"629 select_9" -> "634 expand_as_6" [label="(16, 6, 64, 32)", style=solid]; +"629 select_9" -> "635 div_6" [label="(16, 6, 64, 32)", style=solid]; +"630 select_10" -> "636 linalg_vector_norm_7" [label="(16, 6, 64, 32)", style=solid]; +"630 select_10" -> "638 expand_as_7" [label="(16, 6, 64, 32)", style=solid]; +"630 select_10" -> "639 div_7" [label="(16, 6, 64, 32)", style=solid]; +"631 select_11" -> "664 matmul_7" [label="(16, 6, 64, 32)", style=solid]; +"632 linalg_vector_norm_6" -> "633 clamp_min_6" [label="(16, 6, 64, 1)", style=solid]; +"633 clamp_min_6" -> "634 expand_as_6" [label="(16, 6, 64, 1)", style=solid]; +"634 expand_as_6" -> "635 div_6" [label="(16, 6, 64, 32)", style=solid]; +"635 div_6" -> "641 matmul_6" [label="(16, 6, 64, 32)", style=solid]; +"636 linalg_vector_norm_7" -> "637 clamp_min_7" [label="(16, 6, 64, 1)", style=solid]; +"637 clamp_min_7" -> "638 expand_as_7" [label="(16, 6, 64, 1)", style=solid]; +"638 expand_as_7" -> "639 div_7" [label="(16, 6, 64, 32)", style=solid]; +"639 div_7" -> "640 transpose_6" [label="(16, 6, 64, 32)", style=solid]; +"640 transpose_6" -> "641 matmul_6" [label="(16, 6, 32, 64)", style=solid]; +"641 matmul_6" -> "644 mul_7" [label="(16, 6, 64, 64)", style=solid]; +"642 clamp_3" -> "643 exp_3" [label="(6, 1, 1)", style=solid]; +"643 exp_3" -> "644 mul_7" [label="(6, 1, 1)", style=solid]; +"644 mul_7" -> "645 add_10" [label="(16, 6, 64, 64)", style=solid]; +"645 add_10" -> "657 view_19" [label="(16, 6, 64, 64)", style=solid]; +"646 new_zeros_1" -> "647 view_18" [label="(32, 32)", style=solid]; +"647 view_18" -> "648 permute_17" [label="(4, 8, 4, 8)", style=solid]; +"648 permute_17" -> "649 reshape_15" [label="(4, 4, 8, 8)", style=solid]; +"649 reshape_15" -> "650 unsqueeze_8" [label="(16, 64)", style=solid]; +"649 reshape_15" -> "651 unsqueeze_9" [label="(16, 64)", style=solid]; +"650 unsqueeze_8" -> "652 sub_1" [label="(16, 1, 64)", style=solid]; +"651 unsqueeze_9" -> "652 sub_1" [label="(16, 64, 1)", style=solid]; +"652 sub_1" -> "653 ne_1" [label="(16, 64, 64)", style=solid]; +"652 sub_1" -> "654 masked_fill_2" [label="(16, 64, 64)", style=solid]; +"652 sub_1" -> "655 eq_1" [label="(16, 64, 64)", style=solid]; +"653 ne_1" -> "654 masked_fill_2" [label="(16, 64, 64)", style=solid]; +"654 masked_fill_2" -> "656 masked_fill_3" [label="(16, 64, 64)", style=solid]; +"655 eq_1" -> "656 masked_fill_3" [label="(16, 64, 64)", style=solid]; +"656 masked_fill_3" -> "658 unsqueeze_10" [label="(16, 64, 64)", style=solid]; +"657 view_19" -> "660 add_11" [label="(1, 16, 6, 64, 64)", style=solid]; +"658 unsqueeze_10" -> "659 unsqueeze_11" [label="(16, 1, 64, 64)", style=solid]; +"659 unsqueeze_11" -> "660 add_11" [label="(1, 16, 1, 64, 64)", style=solid]; +"660 add_11" -> "661 view_20" [label="(1, 16, 6, 64, 64)", style=solid]; +"661 view_20" -> "662 softmax_3" [label="(16, 6, 64, 64)", style=solid]; +"662 softmax_3" -> "663 dropout_12" [label="(16, 6, 64, 64)", style=solid]; +"663 dropout_12" -> "664 matmul_7" [label="(16, 6, 64, 64)", style=solid]; +"664 matmul_7" -> "665 transpose_7" [label="(16, 6, 64, 32)", style=solid]; +"665 transpose_7" -> "666 reshape_16" [label="(16, 64, 6, 32)", style=solid]; +"666 reshape_16" -> "668 reshape_16_0_0_nncf_smooth_quant_0" [label="(16, 64, 192)", style=solid]; +"667 features_3_1_attn_proj_weight_updated_constant0" -> "669 linear_22" [label="(192, 192)", style=solid]; +"668 reshape_16_0_0_nncf_smooth_quant_0" -> "669 linear_22" [label="(16, 64, 192)", style=solid]; +"669 linear_22" -> "670 dropout_13" [label="(16, 64, 192)", style=solid]; +"670 dropout_13" -> "671 view_21" [label="(16, 64, 192)", style=solid]; +"671 view_21" -> "672 permute_18" [label="(1, 4, 4, 8, 8, 192)", style=solid]; +"672 permute_18" -> "673 reshape_17" [label="(1, 4, 8, 4, 8, 192)", style=solid]; +"673 reshape_17" -> "674 roll_3" [label="(1, 32, 32, 192)", style=solid]; +"674 roll_3" -> "675 slice_61" [label="(1, 32, 32, 192)", style=solid]; +"675 slice_61" -> "676 slice_62" [label="(1, 32, 32, 192)", style=solid]; +"676 slice_62" -> "677 slice_63" [label="(1, 28, 32, 192)", style=solid]; +"677 slice_63" -> "678 slice_64" [label="(1, 28, 28, 192)", style=solid]; +"678 slice_64" -> "679 contiguous_5" [label="(1, 28, 28, 192)", style=solid]; +"679 contiguous_5" -> "680 layer_norm_8" [label="(1, 28, 28, 192)", style=solid]; +"680 layer_norm_8" -> "681 add_12" [label="(1, 28, 28, 192)", style=solid]; +"681 add_12" -> "682 add_12_0_0_nncf_smooth_quant_0" [label="(1, 28, 28, 192)", style=solid]; +"681 add_12" -> "702 add_13" [label="(1, 28, 28, 192)", style=solid]; +"682 add_12_0_0_nncf_smooth_quant_0" -> "683 quantize_per_tensor_default_16" [label="(1, 28, 28, 192)", style=solid]; +"683 quantize_per_tensor_default_16" -> "684 dequantize_per_tensor_default_16" [label="(1, 28, 28, 192)", style=solid]; +"684 dequantize_per_tensor_default_16" -> "689 linear_23" [label="(1, 28, 28, 192)", style=solid]; +"685 linear_23_scale_0" -> "688 dequantize_per_channel_default_16" [label="(768,)", style=solid]; +"686 linear_23_zero_point_0" -> "688 dequantize_per_channel_default_16" [label="(768,)", style=solid]; +"687 _frozen_param20" -> "688 dequantize_per_channel_default_16" [label="(768, 192)", style=solid]; +"688 dequantize_per_channel_default_16" -> "689 linear_23" [label="(768, 192)", style=solid]; +"689 linear_23" -> "690 gelu_3" [label="(1, 28, 28, 768)", style=solid]; +"690 gelu_3" -> "691 dropout_14" [label="(1, 28, 28, 768)", style=solid]; +"691 dropout_14" -> "692 dropout_14_0_0_nncf_smooth_quant_0" [label="(1, 28, 28, 768)", style=solid]; +"692 dropout_14_0_0_nncf_smooth_quant_0" -> "693 quantize_per_tensor_default_17" [label="(1, 28, 28, 768)", style=solid]; +"693 quantize_per_tensor_default_17" -> "694 dequantize_per_tensor_default_17" [label="(1, 28, 28, 768)", style=solid]; +"694 dequantize_per_tensor_default_17" -> "699 linear_24" [label="(1, 28, 28, 768)", style=solid]; +"695 linear_24_scale_0" -> "698 dequantize_per_channel_default_17" [label="(192,)", style=solid]; +"696 linear_24_zero_point_0" -> "698 dequantize_per_channel_default_17" [label="(192,)", style=solid]; +"697 _frozen_param21" -> "698 dequantize_per_channel_default_17" [label="(192, 768)", style=solid]; +"698 dequantize_per_channel_default_17" -> "699 linear_24" [label="(192, 768)", style=solid]; +"699 linear_24" -> "700 dropout_15" [label="(1, 28, 28, 192)", style=solid]; +"700 dropout_15" -> "701 layer_norm_9" [label="(1, 28, 28, 192)", style=solid]; +"701 layer_norm_9" -> "702 add_13" [label="(1, 28, 28, 192)", style=solid]; +"702 add_13" -> "703 pad_5" [label="(1, 28, 28, 192)", style=solid]; +"703 pad_5" -> "704 slice_65" [label="(1, 28, 28, 192)", style=solid]; +"703 pad_5" -> "707 slice_68" [label="(1, 28, 28, 192)", style=solid]; +"703 pad_5" -> "710 slice_71" [label="(1, 28, 28, 192)", style=solid]; +"703 pad_5" -> "713 slice_74" [label="(1, 28, 28, 192)", style=solid]; +"704 slice_65" -> "705 slice_66" [label="(1, 14, 28, 192)", style=solid]; +"705 slice_66" -> "706 slice_67" [label="(1, 14, 14, 192)", style=solid]; +"706 slice_67" -> "716 cat_1" [label="(1, 14, 14, 192)", style=solid]; +"707 slice_68" -> "708 slice_69" [label="(1, 14, 28, 192)", style=solid]; +"708 slice_69" -> "709 slice_70" [label="(1, 14, 14, 192)", style=solid]; +"709 slice_70" -> "716 cat_1" [label="(1, 14, 14, 192)", style=solid]; +"710 slice_71" -> "711 slice_72" [label="(1, 14, 28, 192)", style=solid]; +"711 slice_72" -> "712 slice_73" [label="(1, 14, 14, 192)", style=solid]; +"712 slice_73" -> "716 cat_1" [label="(1, 14, 14, 192)", style=solid]; +"713 slice_74" -> "714 slice_75" [label="(1, 14, 28, 192)", style=solid]; +"714 slice_75" -> "715 slice_76" [label="(1, 14, 14, 192)", style=solid]; +"715 slice_76" -> "716 cat_1" [label="(1, 14, 14, 192)", style=solid]; +"716 cat_1" -> "717 cat_1_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 768)", style=solid]; +"717 cat_1_0_0_nncf_smooth_quant_0" -> "718 quantize_per_tensor_default_18" [label="(1, 14, 14, 768)", style=solid]; +"718 quantize_per_tensor_default_18" -> "719 dequantize_per_tensor_default_18" [label="(1, 14, 14, 768)", style=solid]; +"719 dequantize_per_tensor_default_18" -> "724 linear_25" [label="(1, 14, 14, 768)", style=solid]; +"720 linear_25_scale_0" -> "723 dequantize_per_channel_default_18" [label="(384,)", style=solid]; +"721 linear_25_zero_point_0" -> "723 dequantize_per_channel_default_18" [label="(384,)", style=solid]; +"722 _frozen_param22" -> "723 dequantize_per_channel_default_18" [label="(384, 768)", style=solid]; +"723 dequantize_per_channel_default_18" -> "724 linear_25" [label="(384, 768)", style=solid]; +"724 linear_25" -> "725 layer_norm_10" [label="(1, 14, 14, 384)", style=solid]; +"725 layer_norm_10" -> "750 pad_6" [label="(1, 14, 14, 384)", style=solid]; +"725 layer_norm_10" -> "795 add_15" [label="(1, 14, 14, 384)", style=solid]; +"726 _frozen_param23" -> "727 dequantize_per_tensor_default_19" [label="(1, 15, 15, 2)", style=solid]; +"727 dequantize_per_tensor_default_19" -> "732 linear_26" [label="(1, 15, 15, 2)", style=solid]; +"728 linear_26_scale_0" -> "731 dequantize_per_channel_default_19" [label="(512,)", style=solid]; +"729 linear_26_zero_point_0" -> "731 dequantize_per_channel_default_19" [label="(512,)", style=solid]; +"730 _frozen_param24" -> "731 dequantize_per_channel_default_19" [label="(512, 2)", style=solid]; +"731 dequantize_per_channel_default_19" -> "732 linear_26" [label="(512, 2)", style=solid]; +"732 linear_26" -> "733 relu__4" [label="(1, 15, 15, 512)", style=solid]; +"733 relu__4" -> "734 relu__4_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"734 relu__4_0_0_nncf_smooth_quant_0" -> "735 quantize_per_tensor_default_20" [label="(1, 15, 15, 512)", style=solid]; +"735 quantize_per_tensor_default_20" -> "736 dequantize_per_tensor_default_20" [label="(1, 15, 15, 512)", style=solid]; +"736 dequantize_per_tensor_default_20" -> "741 linear_27" [label="(1, 15, 15, 512)", style=solid]; +"737 linear_27_scale_0" -> "740 dequantize_per_channel_default_20" [label="(12,)", style=solid]; +"738 linear_27_zero_point_0" -> "740 dequantize_per_channel_default_20" [label="(12,)", style=solid]; +"739 _frozen_param25" -> "740 dequantize_per_channel_default_20" [label="(12, 512)", style=solid]; +"740 dequantize_per_channel_default_20" -> "741 linear_27" [label="(12, 512)", style=solid]; +"741 linear_27" -> "742 view_22" [label="(1, 15, 15, 12)", style=solid]; +"742 view_22" -> "743 index_4" [label="(225, 12)", style=solid]; +"743 index_4" -> "744 view_23" [label="(4096, 12)", style=solid]; +"744 view_23" -> "745 permute_19" [label="(64, 64, 12)", style=solid]; +"745 permute_19" -> "746 contiguous_6" [label="(12, 64, 64)", style=solid]; +"746 contiguous_6" -> "747 unsqueeze_12" [label="(12, 64, 64)", style=solid]; +"747 unsqueeze_12" -> "748 sigmoid_4" [label="(1, 12, 64, 64)", style=solid]; +"748 sigmoid_4" -> "749 mul_8" [label="(1, 12, 64, 64)", style=solid]; +"749 mul_8" -> "776 add_14" [label="(1, 12, 64, 64)", style=solid]; +"750 pad_6" -> "751 view_24" [label="(1, 16, 16, 384)", style=solid]; +"751 view_24" -> "752 permute_20" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"752 permute_20" -> "753 reshape_18" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"753 reshape_18" -> "756 reshape_18_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"754 clone_4" -> "757 linear_28" [label="(1152,)", style=solid]; +"755 features_5_0_attn_qkv_weight_updated_constant0" -> "757 linear_28" [label="(1152, 384)", style=solid]; +"756 reshape_18_0_0_nncf_smooth_quant_0" -> "757 linear_28" [label="(4, 64, 384)", style=solid]; +"757 linear_28" -> "758 reshape_19" [label="(4, 64, 1152)", style=solid]; +"758 reshape_19" -> "759 permute_21" [label="(4, 64, 3, 12, 32)", style=solid]; +"759 permute_21" -> "760 select_12" [label="(3, 4, 12, 64, 32)", style=solid]; +"759 permute_21" -> "761 select_13" [label="(3, 4, 12, 64, 32)", style=solid]; +"759 permute_21" -> "762 select_14" [label="(3, 4, 12, 64, 32)", style=solid]; +"760 select_12" -> "763 linalg_vector_norm_8" [label="(4, 12, 64, 32)", style=solid]; +"760 select_12" -> "765 expand_as_8" [label="(4, 12, 64, 32)", style=solid]; +"760 select_12" -> "766 div_8" [label="(4, 12, 64, 32)", style=solid]; +"761 select_13" -> "767 linalg_vector_norm_9" [label="(4, 12, 64, 32)", style=solid]; +"761 select_13" -> "769 expand_as_9" [label="(4, 12, 64, 32)", style=solid]; +"761 select_13" -> "770 div_9" [label="(4, 12, 64, 32)", style=solid]; +"762 select_14" -> "779 matmul_9" [label="(4, 12, 64, 32)", style=solid]; +"763 linalg_vector_norm_8" -> "764 clamp_min_8" [label="(4, 12, 64, 1)", style=solid]; +"764 clamp_min_8" -> "765 expand_as_8" [label="(4, 12, 64, 1)", style=solid]; +"765 expand_as_8" -> "766 div_8" [label="(4, 12, 64, 32)", style=solid]; +"766 div_8" -> "772 matmul_8" [label="(4, 12, 64, 32)", style=solid]; +"767 linalg_vector_norm_9" -> "768 clamp_min_9" [label="(4, 12, 64, 1)", style=solid]; +"768 clamp_min_9" -> "769 expand_as_9" [label="(4, 12, 64, 1)", style=solid]; +"769 expand_as_9" -> "770 div_9" [label="(4, 12, 64, 32)", style=solid]; +"770 div_9" -> "771 transpose_8" [label="(4, 12, 64, 32)", style=solid]; +"771 transpose_8" -> "772 matmul_8" [label="(4, 12, 32, 64)", style=solid]; +"772 matmul_8" -> "775 mul_9" [label="(4, 12, 64, 64)", style=solid]; +"773 clamp_4" -> "774 exp_4" [label="(12, 1, 1)", style=solid]; +"774 exp_4" -> "775 mul_9" [label="(12, 1, 1)", style=solid]; +"775 mul_9" -> "776 add_14" [label="(4, 12, 64, 64)", style=solid]; +"776 add_14" -> "777 softmax_4" [label="(4, 12, 64, 64)", style=solid]; +"777 softmax_4" -> "778 dropout_16" [label="(4, 12, 64, 64)", style=solid]; +"778 dropout_16" -> "779 matmul_9" [label="(4, 12, 64, 64)", style=solid]; +"779 matmul_9" -> "780 transpose_9" [label="(4, 12, 64, 32)", style=solid]; +"780 transpose_9" -> "781 reshape_20" [label="(4, 64, 12, 32)", style=solid]; +"781 reshape_20" -> "783 reshape_20_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"782 features_5_0_attn_proj_weight_updated_constant0" -> "784 linear_29" [label="(384, 384)", style=solid]; +"783 reshape_20_0_0_nncf_smooth_quant_0" -> "784 linear_29" [label="(4, 64, 384)", style=solid]; +"784 linear_29" -> "785 dropout_17" [label="(4, 64, 384)", style=solid]; +"785 dropout_17" -> "786 view_25" [label="(4, 64, 384)", style=solid]; +"786 view_25" -> "787 permute_22" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"787 permute_22" -> "788 reshape_21" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"788 reshape_21" -> "789 slice_78" [label="(1, 16, 16, 384)", style=solid]; +"789 slice_78" -> "790 slice_79" [label="(1, 16, 16, 384)", style=solid]; +"790 slice_79" -> "791 slice_80" [label="(1, 14, 16, 384)", style=solid]; +"791 slice_80" -> "792 slice_81" [label="(1, 14, 14, 384)", style=solid]; +"792 slice_81" -> "793 contiguous_7" [label="(1, 14, 14, 384)", style=solid]; +"793 contiguous_7" -> "794 layer_norm_11" [label="(1, 14, 14, 384)", style=solid]; +"794 layer_norm_11" -> "795 add_15" [label="(1, 14, 14, 384)", style=solid]; +"795 add_15" -> "796 add_15_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"795 add_15" -> "816 add_16" [label="(1, 14, 14, 384)", style=solid]; +"796 add_15_0_0_nncf_smooth_quant_0" -> "797 quantize_per_tensor_default_21" [label="(1, 14, 14, 384)", style=solid]; +"797 quantize_per_tensor_default_21" -> "798 dequantize_per_tensor_default_21" [label="(1, 14, 14, 384)", style=solid]; +"798 dequantize_per_tensor_default_21" -> "803 linear_30" [label="(1, 14, 14, 384)", style=solid]; +"799 linear_30_scale_0" -> "802 dequantize_per_channel_default_21" [label="(1536,)", style=solid]; +"800 linear_30_zero_point_0" -> "802 dequantize_per_channel_default_21" [label="(1536,)", style=solid]; +"801 _frozen_param26" -> "802 dequantize_per_channel_default_21" [label="(1536, 384)", style=solid]; +"802 dequantize_per_channel_default_21" -> "803 linear_30" [label="(1536, 384)", style=solid]; +"803 linear_30" -> "804 gelu_4" [label="(1, 14, 14, 1536)", style=solid]; +"804 gelu_4" -> "805 dropout_18" [label="(1, 14, 14, 1536)", style=solid]; +"805 dropout_18" -> "806 dropout_18_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"806 dropout_18_0_0_nncf_smooth_quant_0" -> "807 quantize_per_tensor_default_22" [label="(1, 14, 14, 1536)", style=solid]; +"807 quantize_per_tensor_default_22" -> "808 dequantize_per_tensor_default_22" [label="(1, 14, 14, 1536)", style=solid]; +"808 dequantize_per_tensor_default_22" -> "813 linear_31" [label="(1, 14, 14, 1536)", style=solid]; +"809 linear_31_scale_0" -> "812 dequantize_per_channel_default_22" [label="(384,)", style=solid]; +"810 linear_31_zero_point_0" -> "812 dequantize_per_channel_default_22" [label="(384,)", style=solid]; +"811 _frozen_param27" -> "812 dequantize_per_channel_default_22" [label="(384, 1536)", style=solid]; +"812 dequantize_per_channel_default_22" -> "813 linear_31" [label="(384, 1536)", style=solid]; +"813 linear_31" -> "814 dropout_19" [label="(1, 14, 14, 384)", style=solid]; +"814 dropout_19" -> "815 layer_norm_12" [label="(1, 14, 14, 384)", style=solid]; +"815 layer_norm_12" -> "816 add_16" [label="(1, 14, 14, 384)", style=solid]; +"816 add_16" -> "841 pad_7" [label="(1, 14, 14, 384)", style=solid]; +"816 add_16" -> "904 add_19" [label="(1, 14, 14, 384)", style=solid]; +"817 _frozen_param28" -> "818 dequantize_per_tensor_default_23" [label="(1, 15, 15, 2)", style=solid]; +"818 dequantize_per_tensor_default_23" -> "823 linear_32" [label="(1, 15, 15, 2)", style=solid]; +"819 linear_32_scale_0" -> "822 dequantize_per_channel_default_23" [label="(512,)", style=solid]; +"820 linear_32_zero_point_0" -> "822 dequantize_per_channel_default_23" [label="(512,)", style=solid]; +"821 _frozen_param29" -> "822 dequantize_per_channel_default_23" [label="(512, 2)", style=solid]; +"822 dequantize_per_channel_default_23" -> "823 linear_32" [label="(512, 2)", style=solid]; +"823 linear_32" -> "824 relu__5" [label="(1, 15, 15, 512)", style=solid]; +"824 relu__5" -> "825 relu__5_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"825 relu__5_0_0_nncf_smooth_quant_0" -> "826 quantize_per_tensor_default_24" [label="(1, 15, 15, 512)", style=solid]; +"826 quantize_per_tensor_default_24" -> "827 dequantize_per_tensor_default_24" [label="(1, 15, 15, 512)", style=solid]; +"827 dequantize_per_tensor_default_24" -> "832 linear_33" [label="(1, 15, 15, 512)", style=solid]; +"828 linear_33_scale_0" -> "831 dequantize_per_channel_default_24" [label="(12,)", style=solid]; +"829 linear_33_zero_point_0" -> "831 dequantize_per_channel_default_24" [label="(12,)", style=solid]; +"830 _frozen_param30" -> "831 dequantize_per_channel_default_24" [label="(12, 512)", style=solid]; +"831 dequantize_per_channel_default_24" -> "832 linear_33" [label="(12, 512)", style=solid]; +"832 linear_33" -> "833 view_26" [label="(1, 15, 15, 12)", style=solid]; +"833 view_26" -> "834 index_5" [label="(225, 12)", style=solid]; +"834 index_5" -> "835 view_27" [label="(4096, 12)", style=solid]; +"835 view_27" -> "836 permute_23" [label="(64, 64, 12)", style=solid]; +"836 permute_23" -> "837 contiguous_8" [label="(12, 64, 64)", style=solid]; +"837 contiguous_8" -> "838 unsqueeze_13" [label="(12, 64, 64)", style=solid]; +"838 unsqueeze_13" -> "839 sigmoid_5" [label="(1, 12, 64, 64)", style=solid]; +"839 sigmoid_5" -> "840 mul_10" [label="(1, 12, 64, 64)", style=solid]; +"840 mul_10" -> "868 add_17" [label="(1, 12, 64, 64)", style=solid]; +"841 pad_7" -> "842 roll_4" [label="(1, 16, 16, 384)", style=solid]; +"842 roll_4" -> "843 view_28" [label="(1, 16, 16, 384)", style=solid]; +"843 view_28" -> "844 permute_24" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"844 permute_24" -> "845 reshape_22" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"845 reshape_22" -> "848 reshape_22_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"845 reshape_22" -> "869 new_zeros_2" [label="(4, 64, 384)", style=solid]; +"846 clone_5" -> "849 linear_34" [label="(1152,)", style=solid]; +"847 features_5_1_attn_qkv_weight_updated_constant0" -> "849 linear_34" [label="(1152, 384)", style=solid]; +"848 reshape_22_0_0_nncf_smooth_quant_0" -> "849 linear_34" [label="(4, 64, 384)", style=solid]; +"849 linear_34" -> "850 reshape_23" [label="(4, 64, 1152)", style=solid]; +"850 reshape_23" -> "851 permute_25" [label="(4, 64, 3, 12, 32)", style=solid]; +"851 permute_25" -> "852 select_15" [label="(3, 4, 12, 64, 32)", style=solid]; +"851 permute_25" -> "853 select_16" [label="(3, 4, 12, 64, 32)", style=solid]; +"851 permute_25" -> "854 select_17" [label="(3, 4, 12, 64, 32)", style=solid]; +"852 select_15" -> "855 linalg_vector_norm_10" [label="(4, 12, 64, 32)", style=solid]; +"852 select_15" -> "857 expand_as_10" [label="(4, 12, 64, 32)", style=solid]; +"852 select_15" -> "858 div_10" [label="(4, 12, 64, 32)", style=solid]; +"853 select_16" -> "859 linalg_vector_norm_11" [label="(4, 12, 64, 32)", style=solid]; +"853 select_16" -> "861 expand_as_11" [label="(4, 12, 64, 32)", style=solid]; +"853 select_16" -> "862 div_11" [label="(4, 12, 64, 32)", style=solid]; +"854 select_17" -> "887 matmul_11" [label="(4, 12, 64, 32)", style=solid]; +"855 linalg_vector_norm_10" -> "856 clamp_min_10" [label="(4, 12, 64, 1)", style=solid]; +"856 clamp_min_10" -> "857 expand_as_10" [label="(4, 12, 64, 1)", style=solid]; +"857 expand_as_10" -> "858 div_10" [label="(4, 12, 64, 32)", style=solid]; +"858 div_10" -> "864 matmul_10" [label="(4, 12, 64, 32)", style=solid]; +"859 linalg_vector_norm_11" -> "860 clamp_min_11" [label="(4, 12, 64, 1)", style=solid]; +"860 clamp_min_11" -> "861 expand_as_11" [label="(4, 12, 64, 1)", style=solid]; +"861 expand_as_11" -> "862 div_11" [label="(4, 12, 64, 32)", style=solid]; +"862 div_11" -> "863 transpose_10" [label="(4, 12, 64, 32)", style=solid]; +"863 transpose_10" -> "864 matmul_10" [label="(4, 12, 32, 64)", style=solid]; +"864 matmul_10" -> "867 mul_11" [label="(4, 12, 64, 64)", style=solid]; +"865 clamp_5" -> "866 exp_5" [label="(12, 1, 1)", style=solid]; +"866 exp_5" -> "867 mul_11" [label="(12, 1, 1)", style=solid]; +"867 mul_11" -> "868 add_17" [label="(4, 12, 64, 64)", style=solid]; +"868 add_17" -> "880 view_30" [label="(4, 12, 64, 64)", style=solid]; +"869 new_zeros_2" -> "870 view_29" [label="(16, 16)", style=solid]; +"870 view_29" -> "871 permute_26" [label="(2, 8, 2, 8)", style=solid]; +"871 permute_26" -> "872 reshape_24" [label="(2, 2, 8, 8)", style=solid]; +"872 reshape_24" -> "873 unsqueeze_14" [label="(4, 64)", style=solid]; +"872 reshape_24" -> "874 unsqueeze_15" [label="(4, 64)", style=solid]; +"873 unsqueeze_14" -> "875 sub_2" [label="(4, 1, 64)", style=solid]; +"874 unsqueeze_15" -> "875 sub_2" [label="(4, 64, 1)", style=solid]; +"875 sub_2" -> "876 ne_2" [label="(4, 64, 64)", style=solid]; +"875 sub_2" -> "877 masked_fill_4" [label="(4, 64, 64)", style=solid]; +"875 sub_2" -> "878 eq_2" [label="(4, 64, 64)", style=solid]; +"876 ne_2" -> "877 masked_fill_4" [label="(4, 64, 64)", style=solid]; +"877 masked_fill_4" -> "879 masked_fill_5" [label="(4, 64, 64)", style=solid]; +"878 eq_2" -> "879 masked_fill_5" [label="(4, 64, 64)", style=solid]; +"879 masked_fill_5" -> "881 unsqueeze_16" [label="(4, 64, 64)", style=solid]; +"880 view_30" -> "883 add_18" [label="(1, 4, 12, 64, 64)", style=solid]; +"881 unsqueeze_16" -> "882 unsqueeze_17" [label="(4, 1, 64, 64)", style=solid]; +"882 unsqueeze_17" -> "883 add_18" [label="(1, 4, 1, 64, 64)", style=solid]; +"883 add_18" -> "884 view_31" [label="(1, 4, 12, 64, 64)", style=solid]; +"884 view_31" -> "885 softmax_5" [label="(4, 12, 64, 64)", style=solid]; +"885 softmax_5" -> "886 dropout_20" [label="(4, 12, 64, 64)", style=solid]; +"886 dropout_20" -> "887 matmul_11" [label="(4, 12, 64, 64)", style=solid]; +"887 matmul_11" -> "888 transpose_11" [label="(4, 12, 64, 32)", style=solid]; +"888 transpose_11" -> "889 reshape_25" [label="(4, 64, 12, 32)", style=solid]; +"889 reshape_25" -> "891 reshape_25_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"890 features_5_1_attn_proj_weight_updated_constant0" -> "892 linear_35" [label="(384, 384)", style=solid]; +"891 reshape_25_0_0_nncf_smooth_quant_0" -> "892 linear_35" [label="(4, 64, 384)", style=solid]; +"892 linear_35" -> "893 dropout_21" [label="(4, 64, 384)", style=solid]; +"893 dropout_21" -> "894 view_32" [label="(4, 64, 384)", style=solid]; +"894 view_32" -> "895 permute_27" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"895 permute_27" -> "896 reshape_26" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"896 reshape_26" -> "897 roll_5" [label="(1, 16, 16, 384)", style=solid]; +"897 roll_5" -> "898 slice_101" [label="(1, 16, 16, 384)", style=solid]; +"898 slice_101" -> "899 slice_102" [label="(1, 16, 16, 384)", style=solid]; +"899 slice_102" -> "900 slice_103" [label="(1, 14, 16, 384)", style=solid]; +"900 slice_103" -> "901 slice_104" [label="(1, 14, 14, 384)", style=solid]; +"901 slice_104" -> "902 contiguous_9" [label="(1, 14, 14, 384)", style=solid]; +"902 contiguous_9" -> "903 layer_norm_13" [label="(1, 14, 14, 384)", style=solid]; +"903 layer_norm_13" -> "904 add_19" [label="(1, 14, 14, 384)", style=solid]; +"904 add_19" -> "905 add_19_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"904 add_19" -> "925 add_20" [label="(1, 14, 14, 384)", style=solid]; +"905 add_19_0_0_nncf_smooth_quant_0" -> "906 quantize_per_tensor_default_25" [label="(1, 14, 14, 384)", style=solid]; +"906 quantize_per_tensor_default_25" -> "907 dequantize_per_tensor_default_25" [label="(1, 14, 14, 384)", style=solid]; +"907 dequantize_per_tensor_default_25" -> "912 linear_36" [label="(1, 14, 14, 384)", style=solid]; +"908 linear_36_scale_0" -> "911 dequantize_per_channel_default_25" [label="(1536,)", style=solid]; +"909 linear_36_zero_point_0" -> "911 dequantize_per_channel_default_25" [label="(1536,)", style=solid]; +"910 _frozen_param31" -> "911 dequantize_per_channel_default_25" [label="(1536, 384)", style=solid]; +"911 dequantize_per_channel_default_25" -> "912 linear_36" [label="(1536, 384)", style=solid]; +"912 linear_36" -> "913 gelu_5" [label="(1, 14, 14, 1536)", style=solid]; +"913 gelu_5" -> "914 dropout_22" [label="(1, 14, 14, 1536)", style=solid]; +"914 dropout_22" -> "915 dropout_22_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"915 dropout_22_0_0_nncf_smooth_quant_0" -> "916 quantize_per_tensor_default_26" [label="(1, 14, 14, 1536)", style=solid]; +"916 quantize_per_tensor_default_26" -> "917 dequantize_per_tensor_default_26" [label="(1, 14, 14, 1536)", style=solid]; +"917 dequantize_per_tensor_default_26" -> "922 linear_37" [label="(1, 14, 14, 1536)", style=solid]; +"918 linear_37_scale_0" -> "921 dequantize_per_channel_default_26" [label="(384,)", style=solid]; +"919 linear_37_zero_point_0" -> "921 dequantize_per_channel_default_26" [label="(384,)", style=solid]; +"920 _frozen_param32" -> "921 dequantize_per_channel_default_26" [label="(384, 1536)", style=solid]; +"921 dequantize_per_channel_default_26" -> "922 linear_37" [label="(384, 1536)", style=solid]; +"922 linear_37" -> "923 dropout_23" [label="(1, 14, 14, 384)", style=solid]; +"923 dropout_23" -> "924 layer_norm_14" [label="(1, 14, 14, 384)", style=solid]; +"924 layer_norm_14" -> "925 add_20" [label="(1, 14, 14, 384)", style=solid]; +"925 add_20" -> "950 pad_8" [label="(1, 14, 14, 384)", style=solid]; +"925 add_20" -> "995 add_22" [label="(1, 14, 14, 384)", style=solid]; +"926 _frozen_param33" -> "927 dequantize_per_tensor_default_27" [label="(1, 15, 15, 2)", style=solid]; +"927 dequantize_per_tensor_default_27" -> "932 linear_38" [label="(1, 15, 15, 2)", style=solid]; +"928 linear_38_scale_0" -> "931 dequantize_per_channel_default_27" [label="(512,)", style=solid]; +"929 linear_38_zero_point_0" -> "931 dequantize_per_channel_default_27" [label="(512,)", style=solid]; +"930 _frozen_param34" -> "931 dequantize_per_channel_default_27" [label="(512, 2)", style=solid]; +"931 dequantize_per_channel_default_27" -> "932 linear_38" [label="(512, 2)", style=solid]; +"932 linear_38" -> "933 relu__6" [label="(1, 15, 15, 512)", style=solid]; +"933 relu__6" -> "934 relu__6_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"934 relu__6_0_0_nncf_smooth_quant_0" -> "935 quantize_per_tensor_default_28" [label="(1, 15, 15, 512)", style=solid]; +"935 quantize_per_tensor_default_28" -> "936 dequantize_per_tensor_default_28" [label="(1, 15, 15, 512)", style=solid]; +"936 dequantize_per_tensor_default_28" -> "941 linear_39" [label="(1, 15, 15, 512)", style=solid]; +"937 linear_39_scale_0" -> "940 dequantize_per_channel_default_28" [label="(12,)", style=solid]; +"938 linear_39_zero_point_0" -> "940 dequantize_per_channel_default_28" [label="(12,)", style=solid]; +"939 _frozen_param35" -> "940 dequantize_per_channel_default_28" [label="(12, 512)", style=solid]; +"940 dequantize_per_channel_default_28" -> "941 linear_39" [label="(12, 512)", style=solid]; +"941 linear_39" -> "942 view_33" [label="(1, 15, 15, 12)", style=solid]; +"942 view_33" -> "943 index_6" [label="(225, 12)", style=solid]; +"943 index_6" -> "944 view_34" [label="(4096, 12)", style=solid]; +"944 view_34" -> "945 permute_28" [label="(64, 64, 12)", style=solid]; +"945 permute_28" -> "946 contiguous_10" [label="(12, 64, 64)", style=solid]; +"946 contiguous_10" -> "947 unsqueeze_18" [label="(12, 64, 64)", style=solid]; +"947 unsqueeze_18" -> "948 sigmoid_6" [label="(1, 12, 64, 64)", style=solid]; +"948 sigmoid_6" -> "949 mul_12" [label="(1, 12, 64, 64)", style=solid]; +"949 mul_12" -> "976 add_21" [label="(1, 12, 64, 64)", style=solid]; +"950 pad_8" -> "951 view_35" [label="(1, 16, 16, 384)", style=solid]; +"951 view_35" -> "952 permute_29" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"952 permute_29" -> "953 reshape_27" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"953 reshape_27" -> "956 reshape_27_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"954 clone_6" -> "957 linear_40" [label="(1152,)", style=solid]; +"955 features_5_2_attn_qkv_weight_updated_constant0" -> "957 linear_40" [label="(1152, 384)", style=solid]; +"956 reshape_27_0_0_nncf_smooth_quant_0" -> "957 linear_40" [label="(4, 64, 384)", style=solid]; +"957 linear_40" -> "958 reshape_28" [label="(4, 64, 1152)", style=solid]; +"958 reshape_28" -> "959 permute_30" [label="(4, 64, 3, 12, 32)", style=solid]; +"959 permute_30" -> "960 select_18" [label="(3, 4, 12, 64, 32)", style=solid]; +"959 permute_30" -> "961 select_19" [label="(3, 4, 12, 64, 32)", style=solid]; +"959 permute_30" -> "962 select_20" [label="(3, 4, 12, 64, 32)", style=solid]; +"960 select_18" -> "963 linalg_vector_norm_12" [label="(4, 12, 64, 32)", style=solid]; +"960 select_18" -> "965 expand_as_12" [label="(4, 12, 64, 32)", style=solid]; +"960 select_18" -> "966 div_12" [label="(4, 12, 64, 32)", style=solid]; +"961 select_19" -> "967 linalg_vector_norm_13" [label="(4, 12, 64, 32)", style=solid]; +"961 select_19" -> "969 expand_as_13" [label="(4, 12, 64, 32)", style=solid]; +"961 select_19" -> "970 div_13" [label="(4, 12, 64, 32)", style=solid]; +"962 select_20" -> "979 matmul_13" [label="(4, 12, 64, 32)", style=solid]; +"963 linalg_vector_norm_12" -> "964 clamp_min_12" [label="(4, 12, 64, 1)", style=solid]; +"964 clamp_min_12" -> "965 expand_as_12" [label="(4, 12, 64, 1)", style=solid]; +"965 expand_as_12" -> "966 div_12" [label="(4, 12, 64, 32)", style=solid]; +"966 div_12" -> "972 matmul_12" [label="(4, 12, 64, 32)", style=solid]; +"967 linalg_vector_norm_13" -> "968 clamp_min_13" [label="(4, 12, 64, 1)", style=solid]; +"968 clamp_min_13" -> "969 expand_as_13" [label="(4, 12, 64, 1)", style=solid]; +"969 expand_as_13" -> "970 div_13" [label="(4, 12, 64, 32)", style=solid]; +"970 div_13" -> "971 transpose_12" [label="(4, 12, 64, 32)", style=solid]; +"971 transpose_12" -> "972 matmul_12" [label="(4, 12, 32, 64)", style=solid]; +"972 matmul_12" -> "975 mul_13" [label="(4, 12, 64, 64)", style=solid]; +"973 clamp_6" -> "974 exp_6" [label="(12, 1, 1)", style=solid]; +"974 exp_6" -> "975 mul_13" [label="(12, 1, 1)", style=solid]; +"975 mul_13" -> "976 add_21" [label="(4, 12, 64, 64)", style=solid]; +"976 add_21" -> "977 softmax_6" [label="(4, 12, 64, 64)", style=solid]; +"977 softmax_6" -> "978 dropout_24" [label="(4, 12, 64, 64)", style=solid]; +"978 dropout_24" -> "979 matmul_13" [label="(4, 12, 64, 64)", style=solid]; +"979 matmul_13" -> "980 transpose_13" [label="(4, 12, 64, 32)", style=solid]; +"980 transpose_13" -> "981 reshape_29" [label="(4, 64, 12, 32)", style=solid]; +"981 reshape_29" -> "983 reshape_29_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"982 features_5_2_attn_proj_weight_updated_constant0" -> "984 linear_41" [label="(384, 384)", style=solid]; +"983 reshape_29_0_0_nncf_smooth_quant_0" -> "984 linear_41" [label="(4, 64, 384)", style=solid]; +"984 linear_41" -> "985 dropout_25" [label="(4, 64, 384)", style=solid]; +"985 dropout_25" -> "986 view_36" [label="(4, 64, 384)", style=solid]; +"986 view_36" -> "987 permute_31" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"987 permute_31" -> "988 reshape_30" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"988 reshape_30" -> "989 slice_106" [label="(1, 16, 16, 384)", style=solid]; +"989 slice_106" -> "990 slice_107" [label="(1, 16, 16, 384)", style=solid]; +"990 slice_107" -> "991 slice_108" [label="(1, 14, 16, 384)", style=solid]; +"991 slice_108" -> "992 slice_109" [label="(1, 14, 14, 384)", style=solid]; +"992 slice_109" -> "993 contiguous_11" [label="(1, 14, 14, 384)", style=solid]; +"993 contiguous_11" -> "994 layer_norm_15" [label="(1, 14, 14, 384)", style=solid]; +"994 layer_norm_15" -> "995 add_22" [label="(1, 14, 14, 384)", style=solid]; +"995 add_22" -> "996 add_22_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"995 add_22" -> "1016 add_23" [label="(1, 14, 14, 384)", style=solid]; +"996 add_22_0_0_nncf_smooth_quant_0" -> "997 quantize_per_tensor_default_29" [label="(1, 14, 14, 384)", style=solid]; +"997 quantize_per_tensor_default_29" -> "998 dequantize_per_tensor_default_29" [label="(1, 14, 14, 384)", style=solid]; +"998 dequantize_per_tensor_default_29" -> "1003 linear_42" [label="(1, 14, 14, 384)", style=solid]; +"999 linear_42_scale_0" -> "1002 dequantize_per_channel_default_29" [label="(1536,)", style=solid]; +"1000 linear_42_zero_point_0" -> "1002 dequantize_per_channel_default_29" [label="(1536,)", style=solid]; +"1001 _frozen_param36" -> "1002 dequantize_per_channel_default_29" [label="(1536, 384)", style=solid]; +"1002 dequantize_per_channel_default_29" -> "1003 linear_42" [label="(1536, 384)", style=solid]; +"1003 linear_42" -> "1004 gelu_6" [label="(1, 14, 14, 1536)", style=solid]; +"1004 gelu_6" -> "1005 dropout_26" [label="(1, 14, 14, 1536)", style=solid]; +"1005 dropout_26" -> "1006 dropout_26_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"1006 dropout_26_0_0_nncf_smooth_quant_0" -> "1007 quantize_per_tensor_default_30" [label="(1, 14, 14, 1536)", style=solid]; +"1007 quantize_per_tensor_default_30" -> "1008 dequantize_per_tensor_default_30" [label="(1, 14, 14, 1536)", style=solid]; +"1008 dequantize_per_tensor_default_30" -> "1013 linear_43" [label="(1, 14, 14, 1536)", style=solid]; +"1009 linear_43_scale_0" -> "1012 dequantize_per_channel_default_30" [label="(384,)", style=solid]; +"1010 linear_43_zero_point_0" -> "1012 dequantize_per_channel_default_30" [label="(384,)", style=solid]; +"1011 _frozen_param37" -> "1012 dequantize_per_channel_default_30" [label="(384, 1536)", style=solid]; +"1012 dequantize_per_channel_default_30" -> "1013 linear_43" [label="(384, 1536)", style=solid]; +"1013 linear_43" -> "1014 dropout_27" [label="(1, 14, 14, 384)", style=solid]; +"1014 dropout_27" -> "1015 layer_norm_16" [label="(1, 14, 14, 384)", style=solid]; +"1015 layer_norm_16" -> "1016 add_23" [label="(1, 14, 14, 384)", style=solid]; +"1016 add_23" -> "1041 pad_9" [label="(1, 14, 14, 384)", style=solid]; +"1016 add_23" -> "1104 add_26" [label="(1, 14, 14, 384)", style=solid]; +"1017 _frozen_param38" -> "1018 dequantize_per_tensor_default_31" [label="(1, 15, 15, 2)", style=solid]; +"1018 dequantize_per_tensor_default_31" -> "1023 linear_44" [label="(1, 15, 15, 2)", style=solid]; +"1019 linear_44_scale_0" -> "1022 dequantize_per_channel_default_31" [label="(512,)", style=solid]; +"1020 linear_44_zero_point_0" -> "1022 dequantize_per_channel_default_31" [label="(512,)", style=solid]; +"1021 _frozen_param39" -> "1022 dequantize_per_channel_default_31" [label="(512, 2)", style=solid]; +"1022 dequantize_per_channel_default_31" -> "1023 linear_44" [label="(512, 2)", style=solid]; +"1023 linear_44" -> "1024 relu__7" [label="(1, 15, 15, 512)", style=solid]; +"1024 relu__7" -> "1025 relu__7_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"1025 relu__7_0_0_nncf_smooth_quant_0" -> "1026 quantize_per_tensor_default_32" [label="(1, 15, 15, 512)", style=solid]; +"1026 quantize_per_tensor_default_32" -> "1027 dequantize_per_tensor_default_32" [label="(1, 15, 15, 512)", style=solid]; +"1027 dequantize_per_tensor_default_32" -> "1032 linear_45" [label="(1, 15, 15, 512)", style=solid]; +"1028 linear_45_scale_0" -> "1031 dequantize_per_channel_default_32" [label="(12,)", style=solid]; +"1029 linear_45_zero_point_0" -> "1031 dequantize_per_channel_default_32" [label="(12,)", style=solid]; +"1030 _frozen_param40" -> "1031 dequantize_per_channel_default_32" [label="(12, 512)", style=solid]; +"1031 dequantize_per_channel_default_32" -> "1032 linear_45" [label="(12, 512)", style=solid]; +"1032 linear_45" -> "1033 view_37" [label="(1, 15, 15, 12)", style=solid]; +"1033 view_37" -> "1034 index_7" [label="(225, 12)", style=solid]; +"1034 index_7" -> "1035 view_38" [label="(4096, 12)", style=solid]; +"1035 view_38" -> "1036 permute_32" [label="(64, 64, 12)", style=solid]; +"1036 permute_32" -> "1037 contiguous_12" [label="(12, 64, 64)", style=solid]; +"1037 contiguous_12" -> "1038 unsqueeze_19" [label="(12, 64, 64)", style=solid]; +"1038 unsqueeze_19" -> "1039 sigmoid_7" [label="(1, 12, 64, 64)", style=solid]; +"1039 sigmoid_7" -> "1040 mul_14" [label="(1, 12, 64, 64)", style=solid]; +"1040 mul_14" -> "1068 add_24" [label="(1, 12, 64, 64)", style=solid]; +"1041 pad_9" -> "1042 roll_6" [label="(1, 16, 16, 384)", style=solid]; +"1042 roll_6" -> "1043 view_39" [label="(1, 16, 16, 384)", style=solid]; +"1043 view_39" -> "1044 permute_33" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1044 permute_33" -> "1045 reshape_31" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1045 reshape_31" -> "1048 reshape_31_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1045 reshape_31" -> "1069 new_zeros_3" [label="(4, 64, 384)", style=solid]; +"1046 clone_7" -> "1049 linear_46" [label="(1152,)", style=solid]; +"1047 features_5_3_attn_qkv_weight_updated_constant0" -> "1049 linear_46" [label="(1152, 384)", style=solid]; +"1048 reshape_31_0_0_nncf_smooth_quant_0" -> "1049 linear_46" [label="(4, 64, 384)", style=solid]; +"1049 linear_46" -> "1050 reshape_32" [label="(4, 64, 1152)", style=solid]; +"1050 reshape_32" -> "1051 permute_34" [label="(4, 64, 3, 12, 32)", style=solid]; +"1051 permute_34" -> "1052 select_21" [label="(3, 4, 12, 64, 32)", style=solid]; +"1051 permute_34" -> "1053 select_22" [label="(3, 4, 12, 64, 32)", style=solid]; +"1051 permute_34" -> "1054 select_23" [label="(3, 4, 12, 64, 32)", style=solid]; +"1052 select_21" -> "1055 linalg_vector_norm_14" [label="(4, 12, 64, 32)", style=solid]; +"1052 select_21" -> "1057 expand_as_14" [label="(4, 12, 64, 32)", style=solid]; +"1052 select_21" -> "1058 div_14" [label="(4, 12, 64, 32)", style=solid]; +"1053 select_22" -> "1059 linalg_vector_norm_15" [label="(4, 12, 64, 32)", style=solid]; +"1053 select_22" -> "1061 expand_as_15" [label="(4, 12, 64, 32)", style=solid]; +"1053 select_22" -> "1062 div_15" [label="(4, 12, 64, 32)", style=solid]; +"1054 select_23" -> "1087 matmul_15" [label="(4, 12, 64, 32)", style=solid]; +"1055 linalg_vector_norm_14" -> "1056 clamp_min_14" [label="(4, 12, 64, 1)", style=solid]; +"1056 clamp_min_14" -> "1057 expand_as_14" [label="(4, 12, 64, 1)", style=solid]; +"1057 expand_as_14" -> "1058 div_14" [label="(4, 12, 64, 32)", style=solid]; +"1058 div_14" -> "1064 matmul_14" [label="(4, 12, 64, 32)", style=solid]; +"1059 linalg_vector_norm_15" -> "1060 clamp_min_15" [label="(4, 12, 64, 1)", style=solid]; +"1060 clamp_min_15" -> "1061 expand_as_15" [label="(4, 12, 64, 1)", style=solid]; +"1061 expand_as_15" -> "1062 div_15" [label="(4, 12, 64, 32)", style=solid]; +"1062 div_15" -> "1063 transpose_14" [label="(4, 12, 64, 32)", style=solid]; +"1063 transpose_14" -> "1064 matmul_14" [label="(4, 12, 32, 64)", style=solid]; +"1064 matmul_14" -> "1067 mul_15" [label="(4, 12, 64, 64)", style=solid]; +"1065 clamp_7" -> "1066 exp_7" [label="(12, 1, 1)", style=solid]; +"1066 exp_7" -> "1067 mul_15" [label="(12, 1, 1)", style=solid]; +"1067 mul_15" -> "1068 add_24" [label="(4, 12, 64, 64)", style=solid]; +"1068 add_24" -> "1080 view_41" [label="(4, 12, 64, 64)", style=solid]; +"1069 new_zeros_3" -> "1070 view_40" [label="(16, 16)", style=solid]; +"1070 view_40" -> "1071 permute_35" [label="(2, 8, 2, 8)", style=solid]; +"1071 permute_35" -> "1072 reshape_33" [label="(2, 2, 8, 8)", style=solid]; +"1072 reshape_33" -> "1073 unsqueeze_20" [label="(4, 64)", style=solid]; +"1072 reshape_33" -> "1074 unsqueeze_21" [label="(4, 64)", style=solid]; +"1073 unsqueeze_20" -> "1075 sub_3" [label="(4, 1, 64)", style=solid]; +"1074 unsqueeze_21" -> "1075 sub_3" [label="(4, 64, 1)", style=solid]; +"1075 sub_3" -> "1076 ne_3" [label="(4, 64, 64)", style=solid]; +"1075 sub_3" -> "1077 masked_fill_6" [label="(4, 64, 64)", style=solid]; +"1075 sub_3" -> "1078 eq_3" [label="(4, 64, 64)", style=solid]; +"1076 ne_3" -> "1077 masked_fill_6" [label="(4, 64, 64)", style=solid]; +"1077 masked_fill_6" -> "1079 masked_fill_7" [label="(4, 64, 64)", style=solid]; +"1078 eq_3" -> "1079 masked_fill_7" [label="(4, 64, 64)", style=solid]; +"1079 masked_fill_7" -> "1081 unsqueeze_22" [label="(4, 64, 64)", style=solid]; +"1080 view_41" -> "1083 add_25" [label="(1, 4, 12, 64, 64)", style=solid]; +"1081 unsqueeze_22" -> "1082 unsqueeze_23" [label="(4, 1, 64, 64)", style=solid]; +"1082 unsqueeze_23" -> "1083 add_25" [label="(1, 4, 1, 64, 64)", style=solid]; +"1083 add_25" -> "1084 view_42" [label="(1, 4, 12, 64, 64)", style=solid]; +"1084 view_42" -> "1085 softmax_7" [label="(4, 12, 64, 64)", style=solid]; +"1085 softmax_7" -> "1086 dropout_28" [label="(4, 12, 64, 64)", style=solid]; +"1086 dropout_28" -> "1087 matmul_15" [label="(4, 12, 64, 64)", style=solid]; +"1087 matmul_15" -> "1088 transpose_15" [label="(4, 12, 64, 32)", style=solid]; +"1088 transpose_15" -> "1089 reshape_34" [label="(4, 64, 12, 32)", style=solid]; +"1089 reshape_34" -> "1091 reshape_34_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1090 features_5_3_attn_proj_weight_updated_constant0" -> "1092 linear_47" [label="(384, 384)", style=solid]; +"1091 reshape_34_0_0_nncf_smooth_quant_0" -> "1092 linear_47" [label="(4, 64, 384)", style=solid]; +"1092 linear_47" -> "1093 dropout_29" [label="(4, 64, 384)", style=solid]; +"1093 dropout_29" -> "1094 view_43" [label="(4, 64, 384)", style=solid]; +"1094 view_43" -> "1095 permute_36" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1095 permute_36" -> "1096 reshape_35" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1096 reshape_35" -> "1097 roll_7" [label="(1, 16, 16, 384)", style=solid]; +"1097 roll_7" -> "1098 slice_129" [label="(1, 16, 16, 384)", style=solid]; +"1098 slice_129" -> "1099 slice_130" [label="(1, 16, 16, 384)", style=solid]; +"1099 slice_130" -> "1100 slice_131" [label="(1, 14, 16, 384)", style=solid]; +"1100 slice_131" -> "1101 slice_132" [label="(1, 14, 14, 384)", style=solid]; +"1101 slice_132" -> "1102 contiguous_13" [label="(1, 14, 14, 384)", style=solid]; +"1102 contiguous_13" -> "1103 layer_norm_17" [label="(1, 14, 14, 384)", style=solid]; +"1103 layer_norm_17" -> "1104 add_26" [label="(1, 14, 14, 384)", style=solid]; +"1104 add_26" -> "1105 add_26_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"1104 add_26" -> "1125 add_27" [label="(1, 14, 14, 384)", style=solid]; +"1105 add_26_0_0_nncf_smooth_quant_0" -> "1106 quantize_per_tensor_default_33" [label="(1, 14, 14, 384)", style=solid]; +"1106 quantize_per_tensor_default_33" -> "1107 dequantize_per_tensor_default_33" [label="(1, 14, 14, 384)", style=solid]; +"1107 dequantize_per_tensor_default_33" -> "1112 linear_48" [label="(1, 14, 14, 384)", style=solid]; +"1108 linear_48_scale_0" -> "1111 dequantize_per_channel_default_33" [label="(1536,)", style=solid]; +"1109 linear_48_zero_point_0" -> "1111 dequantize_per_channel_default_33" [label="(1536,)", style=solid]; +"1110 _frozen_param41" -> "1111 dequantize_per_channel_default_33" [label="(1536, 384)", style=solid]; +"1111 dequantize_per_channel_default_33" -> "1112 linear_48" [label="(1536, 384)", style=solid]; +"1112 linear_48" -> "1113 gelu_7" [label="(1, 14, 14, 1536)", style=solid]; +"1113 gelu_7" -> "1114 dropout_30" [label="(1, 14, 14, 1536)", style=solid]; +"1114 dropout_30" -> "1115 dropout_30_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"1115 dropout_30_0_0_nncf_smooth_quant_0" -> "1116 quantize_per_tensor_default_34" [label="(1, 14, 14, 1536)", style=solid]; +"1116 quantize_per_tensor_default_34" -> "1117 dequantize_per_tensor_default_34" [label="(1, 14, 14, 1536)", style=solid]; +"1117 dequantize_per_tensor_default_34" -> "1122 linear_49" [label="(1, 14, 14, 1536)", style=solid]; +"1118 linear_49_scale_0" -> "1121 dequantize_per_channel_default_34" [label="(384,)", style=solid]; +"1119 linear_49_zero_point_0" -> "1121 dequantize_per_channel_default_34" [label="(384,)", style=solid]; +"1120 _frozen_param42" -> "1121 dequantize_per_channel_default_34" [label="(384, 1536)", style=solid]; +"1121 dequantize_per_channel_default_34" -> "1122 linear_49" [label="(384, 1536)", style=solid]; +"1122 linear_49" -> "1123 dropout_31" [label="(1, 14, 14, 384)", style=solid]; +"1123 dropout_31" -> "1124 layer_norm_18" [label="(1, 14, 14, 384)", style=solid]; +"1124 layer_norm_18" -> "1125 add_27" [label="(1, 14, 14, 384)", style=solid]; +"1125 add_27" -> "1150 pad_10" [label="(1, 14, 14, 384)", style=solid]; +"1125 add_27" -> "1195 add_29" [label="(1, 14, 14, 384)", style=solid]; +"1126 _frozen_param43" -> "1127 dequantize_per_tensor_default_35" [label="(1, 15, 15, 2)", style=solid]; +"1127 dequantize_per_tensor_default_35" -> "1132 linear_50" [label="(1, 15, 15, 2)", style=solid]; +"1128 linear_50_scale_0" -> "1131 dequantize_per_channel_default_35" [label="(512,)", style=solid]; +"1129 linear_50_zero_point_0" -> "1131 dequantize_per_channel_default_35" [label="(512,)", style=solid]; +"1130 _frozen_param44" -> "1131 dequantize_per_channel_default_35" [label="(512, 2)", style=solid]; +"1131 dequantize_per_channel_default_35" -> "1132 linear_50" [label="(512, 2)", style=solid]; +"1132 linear_50" -> "1133 relu__8" [label="(1, 15, 15, 512)", style=solid]; +"1133 relu__8" -> "1134 relu__8_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"1134 relu__8_0_0_nncf_smooth_quant_0" -> "1135 quantize_per_tensor_default_36" [label="(1, 15, 15, 512)", style=solid]; +"1135 quantize_per_tensor_default_36" -> "1136 dequantize_per_tensor_default_36" [label="(1, 15, 15, 512)", style=solid]; +"1136 dequantize_per_tensor_default_36" -> "1141 linear_51" [label="(1, 15, 15, 512)", style=solid]; +"1137 linear_51_scale_0" -> "1140 dequantize_per_channel_default_36" [label="(12,)", style=solid]; +"1138 linear_51_zero_point_0" -> "1140 dequantize_per_channel_default_36" [label="(12,)", style=solid]; +"1139 _frozen_param45" -> "1140 dequantize_per_channel_default_36" [label="(12, 512)", style=solid]; +"1140 dequantize_per_channel_default_36" -> "1141 linear_51" [label="(12, 512)", style=solid]; +"1141 linear_51" -> "1142 view_44" [label="(1, 15, 15, 12)", style=solid]; +"1142 view_44" -> "1143 index_8" [label="(225, 12)", style=solid]; +"1143 index_8" -> "1144 view_45" [label="(4096, 12)", style=solid]; +"1144 view_45" -> "1145 permute_37" [label="(64, 64, 12)", style=solid]; +"1145 permute_37" -> "1146 contiguous_14" [label="(12, 64, 64)", style=solid]; +"1146 contiguous_14" -> "1147 unsqueeze_24" [label="(12, 64, 64)", style=solid]; +"1147 unsqueeze_24" -> "1148 sigmoid_8" [label="(1, 12, 64, 64)", style=solid]; +"1148 sigmoid_8" -> "1149 mul_16" [label="(1, 12, 64, 64)", style=solid]; +"1149 mul_16" -> "1176 add_28" [label="(1, 12, 64, 64)", style=solid]; +"1150 pad_10" -> "1151 view_46" [label="(1, 16, 16, 384)", style=solid]; +"1151 view_46" -> "1152 permute_38" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1152 permute_38" -> "1153 reshape_36" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1153 reshape_36" -> "1156 reshape_36_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1154 clone_8" -> "1157 linear_52" [label="(1152,)", style=solid]; +"1155 features_5_4_attn_qkv_weight_updated_constant0" -> "1157 linear_52" [label="(1152, 384)", style=solid]; +"1156 reshape_36_0_0_nncf_smooth_quant_0" -> "1157 linear_52" [label="(4, 64, 384)", style=solid]; +"1157 linear_52" -> "1158 reshape_37" [label="(4, 64, 1152)", style=solid]; +"1158 reshape_37" -> "1159 permute_39" [label="(4, 64, 3, 12, 32)", style=solid]; +"1159 permute_39" -> "1160 select_24" [label="(3, 4, 12, 64, 32)", style=solid]; +"1159 permute_39" -> "1161 select_25" [label="(3, 4, 12, 64, 32)", style=solid]; +"1159 permute_39" -> "1162 select_26" [label="(3, 4, 12, 64, 32)", style=solid]; +"1160 select_24" -> "1163 linalg_vector_norm_16" [label="(4, 12, 64, 32)", style=solid]; +"1160 select_24" -> "1165 expand_as_16" [label="(4, 12, 64, 32)", style=solid]; +"1160 select_24" -> "1166 div_16" [label="(4, 12, 64, 32)", style=solid]; +"1161 select_25" -> "1167 linalg_vector_norm_17" [label="(4, 12, 64, 32)", style=solid]; +"1161 select_25" -> "1169 expand_as_17" [label="(4, 12, 64, 32)", style=solid]; +"1161 select_25" -> "1170 div_17" [label="(4, 12, 64, 32)", style=solid]; +"1162 select_26" -> "1179 matmul_17" [label="(4, 12, 64, 32)", style=solid]; +"1163 linalg_vector_norm_16" -> "1164 clamp_min_16" [label="(4, 12, 64, 1)", style=solid]; +"1164 clamp_min_16" -> "1165 expand_as_16" [label="(4, 12, 64, 1)", style=solid]; +"1165 expand_as_16" -> "1166 div_16" [label="(4, 12, 64, 32)", style=solid]; +"1166 div_16" -> "1172 matmul_16" [label="(4, 12, 64, 32)", style=solid]; +"1167 linalg_vector_norm_17" -> "1168 clamp_min_17" [label="(4, 12, 64, 1)", style=solid]; +"1168 clamp_min_17" -> "1169 expand_as_17" [label="(4, 12, 64, 1)", style=solid]; +"1169 expand_as_17" -> "1170 div_17" [label="(4, 12, 64, 32)", style=solid]; +"1170 div_17" -> "1171 transpose_16" [label="(4, 12, 64, 32)", style=solid]; +"1171 transpose_16" -> "1172 matmul_16" [label="(4, 12, 32, 64)", style=solid]; +"1172 matmul_16" -> "1175 mul_17" [label="(4, 12, 64, 64)", style=solid]; +"1173 clamp_8" -> "1174 exp_8" [label="(12, 1, 1)", style=solid]; +"1174 exp_8" -> "1175 mul_17" [label="(12, 1, 1)", style=solid]; +"1175 mul_17" -> "1176 add_28" [label="(4, 12, 64, 64)", style=solid]; +"1176 add_28" -> "1177 softmax_8" [label="(4, 12, 64, 64)", style=solid]; +"1177 softmax_8" -> "1178 dropout_32" [label="(4, 12, 64, 64)", style=solid]; +"1178 dropout_32" -> "1179 matmul_17" [label="(4, 12, 64, 64)", style=solid]; +"1179 matmul_17" -> "1180 transpose_17" [label="(4, 12, 64, 32)", style=solid]; +"1180 transpose_17" -> "1181 reshape_38" [label="(4, 64, 12, 32)", style=solid]; +"1181 reshape_38" -> "1183 reshape_38_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1182 features_5_4_attn_proj_weight_updated_constant0" -> "1184 linear_53" [label="(384, 384)", style=solid]; +"1183 reshape_38_0_0_nncf_smooth_quant_0" -> "1184 linear_53" [label="(4, 64, 384)", style=solid]; +"1184 linear_53" -> "1185 dropout_33" [label="(4, 64, 384)", style=solid]; +"1185 dropout_33" -> "1186 view_47" [label="(4, 64, 384)", style=solid]; +"1186 view_47" -> "1187 permute_40" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1187 permute_40" -> "1188 reshape_39" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1188 reshape_39" -> "1189 slice_134" [label="(1, 16, 16, 384)", style=solid]; +"1189 slice_134" -> "1190 slice_135" [label="(1, 16, 16, 384)", style=solid]; +"1190 slice_135" -> "1191 slice_136" [label="(1, 14, 16, 384)", style=solid]; +"1191 slice_136" -> "1192 slice_137" [label="(1, 14, 14, 384)", style=solid]; +"1192 slice_137" -> "1193 contiguous_15" [label="(1, 14, 14, 384)", style=solid]; +"1193 contiguous_15" -> "1194 layer_norm_19" [label="(1, 14, 14, 384)", style=solid]; +"1194 layer_norm_19" -> "1195 add_29" [label="(1, 14, 14, 384)", style=solid]; +"1195 add_29" -> "1196 add_29_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"1195 add_29" -> "1216 add_30" [label="(1, 14, 14, 384)", style=solid]; +"1196 add_29_0_0_nncf_smooth_quant_0" -> "1197 quantize_per_tensor_default_37" [label="(1, 14, 14, 384)", style=solid]; +"1197 quantize_per_tensor_default_37" -> "1198 dequantize_per_tensor_default_37" [label="(1, 14, 14, 384)", style=solid]; +"1198 dequantize_per_tensor_default_37" -> "1203 linear_54" [label="(1, 14, 14, 384)", style=solid]; +"1199 linear_54_scale_0" -> "1202 dequantize_per_channel_default_37" [label="(1536,)", style=solid]; +"1200 linear_54_zero_point_0" -> "1202 dequantize_per_channel_default_37" [label="(1536,)", style=solid]; +"1201 _frozen_param46" -> "1202 dequantize_per_channel_default_37" [label="(1536, 384)", style=solid]; +"1202 dequantize_per_channel_default_37" -> "1203 linear_54" [label="(1536, 384)", style=solid]; +"1203 linear_54" -> "1204 gelu_8" [label="(1, 14, 14, 1536)", style=solid]; +"1204 gelu_8" -> "1205 dropout_34" [label="(1, 14, 14, 1536)", style=solid]; +"1205 dropout_34" -> "1206 dropout_34_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"1206 dropout_34_0_0_nncf_smooth_quant_0" -> "1207 quantize_per_tensor_default_38" [label="(1, 14, 14, 1536)", style=solid]; +"1207 quantize_per_tensor_default_38" -> "1208 dequantize_per_tensor_default_38" [label="(1, 14, 14, 1536)", style=solid]; +"1208 dequantize_per_tensor_default_38" -> "1213 linear_55" [label="(1, 14, 14, 1536)", style=solid]; +"1209 linear_55_scale_0" -> "1212 dequantize_per_channel_default_38" [label="(384,)", style=solid]; +"1210 linear_55_zero_point_0" -> "1212 dequantize_per_channel_default_38" [label="(384,)", style=solid]; +"1211 _frozen_param47" -> "1212 dequantize_per_channel_default_38" [label="(384, 1536)", style=solid]; +"1212 dequantize_per_channel_default_38" -> "1213 linear_55" [label="(384, 1536)", style=solid]; +"1213 linear_55" -> "1214 dropout_35" [label="(1, 14, 14, 384)", style=solid]; +"1214 dropout_35" -> "1215 layer_norm_20" [label="(1, 14, 14, 384)", style=solid]; +"1215 layer_norm_20" -> "1216 add_30" [label="(1, 14, 14, 384)", style=solid]; +"1216 add_30" -> "1241 pad_11" [label="(1, 14, 14, 384)", style=solid]; +"1216 add_30" -> "1304 add_33" [label="(1, 14, 14, 384)", style=solid]; +"1217 _frozen_param48" -> "1218 dequantize_per_tensor_default_39" [label="(1, 15, 15, 2)", style=solid]; +"1218 dequantize_per_tensor_default_39" -> "1223 linear_56" [label="(1, 15, 15, 2)", style=solid]; +"1219 linear_56_scale_0" -> "1222 dequantize_per_channel_default_39" [label="(512,)", style=solid]; +"1220 linear_56_zero_point_0" -> "1222 dequantize_per_channel_default_39" [label="(512,)", style=solid]; +"1221 _frozen_param49" -> "1222 dequantize_per_channel_default_39" [label="(512, 2)", style=solid]; +"1222 dequantize_per_channel_default_39" -> "1223 linear_56" [label="(512, 2)", style=solid]; +"1223 linear_56" -> "1224 relu__9" [label="(1, 15, 15, 512)", style=solid]; +"1224 relu__9" -> "1225 relu__9_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"1225 relu__9_0_0_nncf_smooth_quant_0" -> "1226 quantize_per_tensor_default_40" [label="(1, 15, 15, 512)", style=solid]; +"1226 quantize_per_tensor_default_40" -> "1227 dequantize_per_tensor_default_40" [label="(1, 15, 15, 512)", style=solid]; +"1227 dequantize_per_tensor_default_40" -> "1232 linear_57" [label="(1, 15, 15, 512)", style=solid]; +"1228 linear_57_scale_0" -> "1231 dequantize_per_channel_default_40" [label="(12,)", style=solid]; +"1229 linear_57_zero_point_0" -> "1231 dequantize_per_channel_default_40" [label="(12,)", style=solid]; +"1230 _frozen_param50" -> "1231 dequantize_per_channel_default_40" [label="(12, 512)", style=solid]; +"1231 dequantize_per_channel_default_40" -> "1232 linear_57" [label="(12, 512)", style=solid]; +"1232 linear_57" -> "1233 view_48" [label="(1, 15, 15, 12)", style=solid]; +"1233 view_48" -> "1234 index_9" [label="(225, 12)", style=solid]; +"1234 index_9" -> "1235 view_49" [label="(4096, 12)", style=solid]; +"1235 view_49" -> "1236 permute_41" [label="(64, 64, 12)", style=solid]; +"1236 permute_41" -> "1237 contiguous_16" [label="(12, 64, 64)", style=solid]; +"1237 contiguous_16" -> "1238 unsqueeze_25" [label="(12, 64, 64)", style=solid]; +"1238 unsqueeze_25" -> "1239 sigmoid_9" [label="(1, 12, 64, 64)", style=solid]; +"1239 sigmoid_9" -> "1240 mul_18" [label="(1, 12, 64, 64)", style=solid]; +"1240 mul_18" -> "1268 add_31" [label="(1, 12, 64, 64)", style=solid]; +"1241 pad_11" -> "1242 roll_8" [label="(1, 16, 16, 384)", style=solid]; +"1242 roll_8" -> "1243 view_50" [label="(1, 16, 16, 384)", style=solid]; +"1243 view_50" -> "1244 permute_42" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1244 permute_42" -> "1245 reshape_40" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1245 reshape_40" -> "1248 reshape_40_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1245 reshape_40" -> "1269 new_zeros_4" [label="(4, 64, 384)", style=solid]; +"1246 clone_9" -> "1249 linear_58" [label="(1152,)", style=solid]; +"1247 features_5_5_attn_qkv_weight_updated_constant0" -> "1249 linear_58" [label="(1152, 384)", style=solid]; +"1248 reshape_40_0_0_nncf_smooth_quant_0" -> "1249 linear_58" [label="(4, 64, 384)", style=solid]; +"1249 linear_58" -> "1250 reshape_41" [label="(4, 64, 1152)", style=solid]; +"1250 reshape_41" -> "1251 permute_43" [label="(4, 64, 3, 12, 32)", style=solid]; +"1251 permute_43" -> "1252 select_27" [label="(3, 4, 12, 64, 32)", style=solid]; +"1251 permute_43" -> "1253 select_28" [label="(3, 4, 12, 64, 32)", style=solid]; +"1251 permute_43" -> "1254 select_29" [label="(3, 4, 12, 64, 32)", style=solid]; +"1252 select_27" -> "1255 linalg_vector_norm_18" [label="(4, 12, 64, 32)", style=solid]; +"1252 select_27" -> "1257 expand_as_18" [label="(4, 12, 64, 32)", style=solid]; +"1252 select_27" -> "1258 div_18" [label="(4, 12, 64, 32)", style=solid]; +"1253 select_28" -> "1259 linalg_vector_norm_19" [label="(4, 12, 64, 32)", style=solid]; +"1253 select_28" -> "1261 expand_as_19" [label="(4, 12, 64, 32)", style=solid]; +"1253 select_28" -> "1262 div_19" [label="(4, 12, 64, 32)", style=solid]; +"1254 select_29" -> "1287 matmul_19" [label="(4, 12, 64, 32)", style=solid]; +"1255 linalg_vector_norm_18" -> "1256 clamp_min_18" [label="(4, 12, 64, 1)", style=solid]; +"1256 clamp_min_18" -> "1257 expand_as_18" [label="(4, 12, 64, 1)", style=solid]; +"1257 expand_as_18" -> "1258 div_18" [label="(4, 12, 64, 32)", style=solid]; +"1258 div_18" -> "1264 matmul_18" [label="(4, 12, 64, 32)", style=solid]; +"1259 linalg_vector_norm_19" -> "1260 clamp_min_19" [label="(4, 12, 64, 1)", style=solid]; +"1260 clamp_min_19" -> "1261 expand_as_19" [label="(4, 12, 64, 1)", style=solid]; +"1261 expand_as_19" -> "1262 div_19" [label="(4, 12, 64, 32)", style=solid]; +"1262 div_19" -> "1263 transpose_18" [label="(4, 12, 64, 32)", style=solid]; +"1263 transpose_18" -> "1264 matmul_18" [label="(4, 12, 32, 64)", style=solid]; +"1264 matmul_18" -> "1267 mul_19" [label="(4, 12, 64, 64)", style=solid]; +"1265 clamp_9" -> "1266 exp_9" [label="(12, 1, 1)", style=solid]; +"1266 exp_9" -> "1267 mul_19" [label="(12, 1, 1)", style=solid]; +"1267 mul_19" -> "1268 add_31" [label="(4, 12, 64, 64)", style=solid]; +"1268 add_31" -> "1280 view_52" [label="(4, 12, 64, 64)", style=solid]; +"1269 new_zeros_4" -> "1270 view_51" [label="(16, 16)", style=solid]; +"1270 view_51" -> "1271 permute_44" [label="(2, 8, 2, 8)", style=solid]; +"1271 permute_44" -> "1272 reshape_42" [label="(2, 2, 8, 8)", style=solid]; +"1272 reshape_42" -> "1273 unsqueeze_26" [label="(4, 64)", style=solid]; +"1272 reshape_42" -> "1274 unsqueeze_27" [label="(4, 64)", style=solid]; +"1273 unsqueeze_26" -> "1275 sub_4" [label="(4, 1, 64)", style=solid]; +"1274 unsqueeze_27" -> "1275 sub_4" [label="(4, 64, 1)", style=solid]; +"1275 sub_4" -> "1276 ne_4" [label="(4, 64, 64)", style=solid]; +"1275 sub_4" -> "1277 masked_fill_8" [label="(4, 64, 64)", style=solid]; +"1275 sub_4" -> "1278 eq_4" [label="(4, 64, 64)", style=solid]; +"1276 ne_4" -> "1277 masked_fill_8" [label="(4, 64, 64)", style=solid]; +"1277 masked_fill_8" -> "1279 masked_fill_9" [label="(4, 64, 64)", style=solid]; +"1278 eq_4" -> "1279 masked_fill_9" [label="(4, 64, 64)", style=solid]; +"1279 masked_fill_9" -> "1281 unsqueeze_28" [label="(4, 64, 64)", style=solid]; +"1280 view_52" -> "1283 add_32" [label="(1, 4, 12, 64, 64)", style=solid]; +"1281 unsqueeze_28" -> "1282 unsqueeze_29" [label="(4, 1, 64, 64)", style=solid]; +"1282 unsqueeze_29" -> "1283 add_32" [label="(1, 4, 1, 64, 64)", style=solid]; +"1283 add_32" -> "1284 view_53" [label="(1, 4, 12, 64, 64)", style=solid]; +"1284 view_53" -> "1285 softmax_9" [label="(4, 12, 64, 64)", style=solid]; +"1285 softmax_9" -> "1286 dropout_36" [label="(4, 12, 64, 64)", style=solid]; +"1286 dropout_36" -> "1287 matmul_19" [label="(4, 12, 64, 64)", style=solid]; +"1287 matmul_19" -> "1288 transpose_19" [label="(4, 12, 64, 32)", style=solid]; +"1288 transpose_19" -> "1289 reshape_43" [label="(4, 64, 12, 32)", style=solid]; +"1289 reshape_43" -> "1291 reshape_43_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1290 features_5_5_attn_proj_weight_updated_constant0" -> "1292 linear_59" [label="(384, 384)", style=solid]; +"1291 reshape_43_0_0_nncf_smooth_quant_0" -> "1292 linear_59" [label="(4, 64, 384)", style=solid]; +"1292 linear_59" -> "1293 dropout_37" [label="(4, 64, 384)", style=solid]; +"1293 dropout_37" -> "1294 view_54" [label="(4, 64, 384)", style=solid]; +"1294 view_54" -> "1295 permute_45" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1295 permute_45" -> "1296 reshape_44" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1296 reshape_44" -> "1297 roll_9" [label="(1, 16, 16, 384)", style=solid]; +"1297 roll_9" -> "1298 slice_157" [label="(1, 16, 16, 384)", style=solid]; +"1298 slice_157" -> "1299 slice_158" [label="(1, 16, 16, 384)", style=solid]; +"1299 slice_158" -> "1300 slice_159" [label="(1, 14, 16, 384)", style=solid]; +"1300 slice_159" -> "1301 slice_160" [label="(1, 14, 14, 384)", style=solid]; +"1301 slice_160" -> "1302 contiguous_17" [label="(1, 14, 14, 384)", style=solid]; +"1302 contiguous_17" -> "1303 layer_norm_21" [label="(1, 14, 14, 384)", style=solid]; +"1303 layer_norm_21" -> "1304 add_33" [label="(1, 14, 14, 384)", style=solid]; +"1304 add_33" -> "1305 add_33_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"1304 add_33" -> "1325 add_34" [label="(1, 14, 14, 384)", style=solid]; +"1305 add_33_0_0_nncf_smooth_quant_0" -> "1306 quantize_per_tensor_default_41" [label="(1, 14, 14, 384)", style=solid]; +"1306 quantize_per_tensor_default_41" -> "1307 dequantize_per_tensor_default_41" [label="(1, 14, 14, 384)", style=solid]; +"1307 dequantize_per_tensor_default_41" -> "1312 linear_60" [label="(1, 14, 14, 384)", style=solid]; +"1308 linear_60_scale_0" -> "1311 dequantize_per_channel_default_41" [label="(1536,)", style=solid]; +"1309 linear_60_zero_point_0" -> "1311 dequantize_per_channel_default_41" [label="(1536,)", style=solid]; +"1310 _frozen_param51" -> "1311 dequantize_per_channel_default_41" [label="(1536, 384)", style=solid]; +"1311 dequantize_per_channel_default_41" -> "1312 linear_60" [label="(1536, 384)", style=solid]; +"1312 linear_60" -> "1313 gelu_9" [label="(1, 14, 14, 1536)", style=solid]; +"1313 gelu_9" -> "1314 dropout_38" [label="(1, 14, 14, 1536)", style=solid]; +"1314 dropout_38" -> "1315 dropout_38_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"1315 dropout_38_0_0_nncf_smooth_quant_0" -> "1316 quantize_per_tensor_default_42" [label="(1, 14, 14, 1536)", style=solid]; +"1316 quantize_per_tensor_default_42" -> "1317 dequantize_per_tensor_default_42" [label="(1, 14, 14, 1536)", style=solid]; +"1317 dequantize_per_tensor_default_42" -> "1322 linear_61" [label="(1, 14, 14, 1536)", style=solid]; +"1318 linear_61_scale_0" -> "1321 dequantize_per_channel_default_42" [label="(384,)", style=solid]; +"1319 linear_61_zero_point_0" -> "1321 dequantize_per_channel_default_42" [label="(384,)", style=solid]; +"1320 _frozen_param52" -> "1321 dequantize_per_channel_default_42" [label="(384, 1536)", style=solid]; +"1321 dequantize_per_channel_default_42" -> "1322 linear_61" [label="(384, 1536)", style=solid]; +"1322 linear_61" -> "1323 dropout_39" [label="(1, 14, 14, 384)", style=solid]; +"1323 dropout_39" -> "1324 layer_norm_22" [label="(1, 14, 14, 384)", style=solid]; +"1324 layer_norm_22" -> "1325 add_34" [label="(1, 14, 14, 384)", style=solid]; +"1325 add_34" -> "1350 pad_12" [label="(1, 14, 14, 384)", style=solid]; +"1325 add_34" -> "1395 add_36" [label="(1, 14, 14, 384)", style=solid]; +"1326 _frozen_param53" -> "1327 dequantize_per_tensor_default_43" [label="(1, 15, 15, 2)", style=solid]; +"1327 dequantize_per_tensor_default_43" -> "1332 linear_62" [label="(1, 15, 15, 2)", style=solid]; +"1328 linear_62_scale_0" -> "1331 dequantize_per_channel_default_43" [label="(512,)", style=solid]; +"1329 linear_62_zero_point_0" -> "1331 dequantize_per_channel_default_43" [label="(512,)", style=solid]; +"1330 _frozen_param54" -> "1331 dequantize_per_channel_default_43" [label="(512, 2)", style=solid]; +"1331 dequantize_per_channel_default_43" -> "1332 linear_62" [label="(512, 2)", style=solid]; +"1332 linear_62" -> "1333 relu__10" [label="(1, 15, 15, 512)", style=solid]; +"1333 relu__10" -> "1334 relu__10_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"1334 relu__10_0_0_nncf_smooth_quant_0" -> "1335 quantize_per_tensor_default_44" [label="(1, 15, 15, 512)", style=solid]; +"1335 quantize_per_tensor_default_44" -> "1336 dequantize_per_tensor_default_44" [label="(1, 15, 15, 512)", style=solid]; +"1336 dequantize_per_tensor_default_44" -> "1341 linear_63" [label="(1, 15, 15, 512)", style=solid]; +"1337 linear_63_scale_0" -> "1340 dequantize_per_channel_default_44" [label="(12,)", style=solid]; +"1338 linear_63_zero_point_0" -> "1340 dequantize_per_channel_default_44" [label="(12,)", style=solid]; +"1339 _frozen_param55" -> "1340 dequantize_per_channel_default_44" [label="(12, 512)", style=solid]; +"1340 dequantize_per_channel_default_44" -> "1341 linear_63" [label="(12, 512)", style=solid]; +"1341 linear_63" -> "1342 view_55" [label="(1, 15, 15, 12)", style=solid]; +"1342 view_55" -> "1343 index_10" [label="(225, 12)", style=solid]; +"1343 index_10" -> "1344 view_56" [label="(4096, 12)", style=solid]; +"1344 view_56" -> "1345 permute_46" [label="(64, 64, 12)", style=solid]; +"1345 permute_46" -> "1346 contiguous_18" [label="(12, 64, 64)", style=solid]; +"1346 contiguous_18" -> "1347 unsqueeze_30" [label="(12, 64, 64)", style=solid]; +"1347 unsqueeze_30" -> "1348 sigmoid_10" [label="(1, 12, 64, 64)", style=solid]; +"1348 sigmoid_10" -> "1349 mul_20" [label="(1, 12, 64, 64)", style=solid]; +"1349 mul_20" -> "1376 add_35" [label="(1, 12, 64, 64)", style=solid]; +"1350 pad_12" -> "1351 view_57" [label="(1, 16, 16, 384)", style=solid]; +"1351 view_57" -> "1352 permute_47" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1352 permute_47" -> "1353 reshape_45" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1353 reshape_45" -> "1356 reshape_45_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1354 clone_10" -> "1357 linear_64" [label="(1152,)", style=solid]; +"1355 features_5_6_attn_qkv_weight_updated_constant0" -> "1357 linear_64" [label="(1152, 384)", style=solid]; +"1356 reshape_45_0_0_nncf_smooth_quant_0" -> "1357 linear_64" [label="(4, 64, 384)", style=solid]; +"1357 linear_64" -> "1358 reshape_46" [label="(4, 64, 1152)", style=solid]; +"1358 reshape_46" -> "1359 permute_48" [label="(4, 64, 3, 12, 32)", style=solid]; +"1359 permute_48" -> "1360 select_30" [label="(3, 4, 12, 64, 32)", style=solid]; +"1359 permute_48" -> "1361 select_31" [label="(3, 4, 12, 64, 32)", style=solid]; +"1359 permute_48" -> "1362 select_32" [label="(3, 4, 12, 64, 32)", style=solid]; +"1360 select_30" -> "1363 linalg_vector_norm_20" [label="(4, 12, 64, 32)", style=solid]; +"1360 select_30" -> "1365 expand_as_20" [label="(4, 12, 64, 32)", style=solid]; +"1360 select_30" -> "1366 div_20" [label="(4, 12, 64, 32)", style=solid]; +"1361 select_31" -> "1367 linalg_vector_norm_21" [label="(4, 12, 64, 32)", style=solid]; +"1361 select_31" -> "1369 expand_as_21" [label="(4, 12, 64, 32)", style=solid]; +"1361 select_31" -> "1370 div_21" [label="(4, 12, 64, 32)", style=solid]; +"1362 select_32" -> "1379 matmul_21" [label="(4, 12, 64, 32)", style=solid]; +"1363 linalg_vector_norm_20" -> "1364 clamp_min_20" [label="(4, 12, 64, 1)", style=solid]; +"1364 clamp_min_20" -> "1365 expand_as_20" [label="(4, 12, 64, 1)", style=solid]; +"1365 expand_as_20" -> "1366 div_20" [label="(4, 12, 64, 32)", style=solid]; +"1366 div_20" -> "1372 matmul_20" [label="(4, 12, 64, 32)", style=solid]; +"1367 linalg_vector_norm_21" -> "1368 clamp_min_21" [label="(4, 12, 64, 1)", style=solid]; +"1368 clamp_min_21" -> "1369 expand_as_21" [label="(4, 12, 64, 1)", style=solid]; +"1369 expand_as_21" -> "1370 div_21" [label="(4, 12, 64, 32)", style=solid]; +"1370 div_21" -> "1371 transpose_20" [label="(4, 12, 64, 32)", style=solid]; +"1371 transpose_20" -> "1372 matmul_20" [label="(4, 12, 32, 64)", style=solid]; +"1372 matmul_20" -> "1375 mul_21" [label="(4, 12, 64, 64)", style=solid]; +"1373 clamp_10" -> "1374 exp_10" [label="(12, 1, 1)", style=solid]; +"1374 exp_10" -> "1375 mul_21" [label="(12, 1, 1)", style=solid]; +"1375 mul_21" -> "1376 add_35" [label="(4, 12, 64, 64)", style=solid]; +"1376 add_35" -> "1377 softmax_10" [label="(4, 12, 64, 64)", style=solid]; +"1377 softmax_10" -> "1378 dropout_40" [label="(4, 12, 64, 64)", style=solid]; +"1378 dropout_40" -> "1379 matmul_21" [label="(4, 12, 64, 64)", style=solid]; +"1379 matmul_21" -> "1380 transpose_21" [label="(4, 12, 64, 32)", style=solid]; +"1380 transpose_21" -> "1381 reshape_47" [label="(4, 64, 12, 32)", style=solid]; +"1381 reshape_47" -> "1383 reshape_47_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1382 features_5_6_attn_proj_weight_updated_constant0" -> "1384 linear_65" [label="(384, 384)", style=solid]; +"1383 reshape_47_0_0_nncf_smooth_quant_0" -> "1384 linear_65" [label="(4, 64, 384)", style=solid]; +"1384 linear_65" -> "1385 dropout_41" [label="(4, 64, 384)", style=solid]; +"1385 dropout_41" -> "1386 view_58" [label="(4, 64, 384)", style=solid]; +"1386 view_58" -> "1387 permute_49" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1387 permute_49" -> "1388 reshape_48" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1388 reshape_48" -> "1389 slice_162" [label="(1, 16, 16, 384)", style=solid]; +"1389 slice_162" -> "1390 slice_163" [label="(1, 16, 16, 384)", style=solid]; +"1390 slice_163" -> "1391 slice_164" [label="(1, 14, 16, 384)", style=solid]; +"1391 slice_164" -> "1392 slice_165" [label="(1, 14, 14, 384)", style=solid]; +"1392 slice_165" -> "1393 contiguous_19" [label="(1, 14, 14, 384)", style=solid]; +"1393 contiguous_19" -> "1394 layer_norm_23" [label="(1, 14, 14, 384)", style=solid]; +"1394 layer_norm_23" -> "1395 add_36" [label="(1, 14, 14, 384)", style=solid]; +"1395 add_36" -> "1396 add_36_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"1395 add_36" -> "1416 add_37" [label="(1, 14, 14, 384)", style=solid]; +"1396 add_36_0_0_nncf_smooth_quant_0" -> "1397 quantize_per_tensor_default_45" [label="(1, 14, 14, 384)", style=solid]; +"1397 quantize_per_tensor_default_45" -> "1398 dequantize_per_tensor_default_45" [label="(1, 14, 14, 384)", style=solid]; +"1398 dequantize_per_tensor_default_45" -> "1403 linear_66" [label="(1, 14, 14, 384)", style=solid]; +"1399 linear_66_scale_0" -> "1402 dequantize_per_channel_default_45" [label="(1536,)", style=solid]; +"1400 linear_66_zero_point_0" -> "1402 dequantize_per_channel_default_45" [label="(1536,)", style=solid]; +"1401 _frozen_param56" -> "1402 dequantize_per_channel_default_45" [label="(1536, 384)", style=solid]; +"1402 dequantize_per_channel_default_45" -> "1403 linear_66" [label="(1536, 384)", style=solid]; +"1403 linear_66" -> "1404 gelu_10" [label="(1, 14, 14, 1536)", style=solid]; +"1404 gelu_10" -> "1405 dropout_42" [label="(1, 14, 14, 1536)", style=solid]; +"1405 dropout_42" -> "1406 dropout_42_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"1406 dropout_42_0_0_nncf_smooth_quant_0" -> "1407 quantize_per_tensor_default_46" [label="(1, 14, 14, 1536)", style=solid]; +"1407 quantize_per_tensor_default_46" -> "1408 dequantize_per_tensor_default_46" [label="(1, 14, 14, 1536)", style=solid]; +"1408 dequantize_per_tensor_default_46" -> "1413 linear_67" [label="(1, 14, 14, 1536)", style=solid]; +"1409 linear_67_scale_0" -> "1412 dequantize_per_channel_default_46" [label="(384,)", style=solid]; +"1410 linear_67_zero_point_0" -> "1412 dequantize_per_channel_default_46" [label="(384,)", style=solid]; +"1411 _frozen_param57" -> "1412 dequantize_per_channel_default_46" [label="(384, 1536)", style=solid]; +"1412 dequantize_per_channel_default_46" -> "1413 linear_67" [label="(384, 1536)", style=solid]; +"1413 linear_67" -> "1414 dropout_43" [label="(1, 14, 14, 384)", style=solid]; +"1414 dropout_43" -> "1415 layer_norm_24" [label="(1, 14, 14, 384)", style=solid]; +"1415 layer_norm_24" -> "1416 add_37" [label="(1, 14, 14, 384)", style=solid]; +"1416 add_37" -> "1441 pad_13" [label="(1, 14, 14, 384)", style=solid]; +"1416 add_37" -> "1504 add_40" [label="(1, 14, 14, 384)", style=solid]; +"1417 _frozen_param58" -> "1418 dequantize_per_tensor_default_47" [label="(1, 15, 15, 2)", style=solid]; +"1418 dequantize_per_tensor_default_47" -> "1423 linear_68" [label="(1, 15, 15, 2)", style=solid]; +"1419 linear_68_scale_0" -> "1422 dequantize_per_channel_default_47" [label="(512,)", style=solid]; +"1420 linear_68_zero_point_0" -> "1422 dequantize_per_channel_default_47" [label="(512,)", style=solid]; +"1421 _frozen_param59" -> "1422 dequantize_per_channel_default_47" [label="(512, 2)", style=solid]; +"1422 dequantize_per_channel_default_47" -> "1423 linear_68" [label="(512, 2)", style=solid]; +"1423 linear_68" -> "1424 relu__11" [label="(1, 15, 15, 512)", style=solid]; +"1424 relu__11" -> "1425 relu__11_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"1425 relu__11_0_0_nncf_smooth_quant_0" -> "1426 quantize_per_tensor_default_48" [label="(1, 15, 15, 512)", style=solid]; +"1426 quantize_per_tensor_default_48" -> "1427 dequantize_per_tensor_default_48" [label="(1, 15, 15, 512)", style=solid]; +"1427 dequantize_per_tensor_default_48" -> "1432 linear_69" [label="(1, 15, 15, 512)", style=solid]; +"1428 linear_69_scale_0" -> "1431 dequantize_per_channel_default_48" [label="(12,)", style=solid]; +"1429 linear_69_zero_point_0" -> "1431 dequantize_per_channel_default_48" [label="(12,)", style=solid]; +"1430 _frozen_param60" -> "1431 dequantize_per_channel_default_48" [label="(12, 512)", style=solid]; +"1431 dequantize_per_channel_default_48" -> "1432 linear_69" [label="(12, 512)", style=solid]; +"1432 linear_69" -> "1433 view_59" [label="(1, 15, 15, 12)", style=solid]; +"1433 view_59" -> "1434 index_11" [label="(225, 12)", style=solid]; +"1434 index_11" -> "1435 view_60" [label="(4096, 12)", style=solid]; +"1435 view_60" -> "1436 permute_50" [label="(64, 64, 12)", style=solid]; +"1436 permute_50" -> "1437 contiguous_20" [label="(12, 64, 64)", style=solid]; +"1437 contiguous_20" -> "1438 unsqueeze_31" [label="(12, 64, 64)", style=solid]; +"1438 unsqueeze_31" -> "1439 sigmoid_11" [label="(1, 12, 64, 64)", style=solid]; +"1439 sigmoid_11" -> "1440 mul_22" [label="(1, 12, 64, 64)", style=solid]; +"1440 mul_22" -> "1468 add_38" [label="(1, 12, 64, 64)", style=solid]; +"1441 pad_13" -> "1442 roll_10" [label="(1, 16, 16, 384)", style=solid]; +"1442 roll_10" -> "1443 view_61" [label="(1, 16, 16, 384)", style=solid]; +"1443 view_61" -> "1444 permute_51" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1444 permute_51" -> "1445 reshape_49" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1445 reshape_49" -> "1448 reshape_49_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1445 reshape_49" -> "1469 new_zeros_5" [label="(4, 64, 384)", style=solid]; +"1446 clone_11" -> "1449 linear_70" [label="(1152,)", style=solid]; +"1447 features_5_7_attn_qkv_weight_updated_constant0" -> "1449 linear_70" [label="(1152, 384)", style=solid]; +"1448 reshape_49_0_0_nncf_smooth_quant_0" -> "1449 linear_70" [label="(4, 64, 384)", style=solid]; +"1449 linear_70" -> "1450 reshape_50" [label="(4, 64, 1152)", style=solid]; +"1450 reshape_50" -> "1451 permute_52" [label="(4, 64, 3, 12, 32)", style=solid]; +"1451 permute_52" -> "1452 select_33" [label="(3, 4, 12, 64, 32)", style=solid]; +"1451 permute_52" -> "1453 select_34" [label="(3, 4, 12, 64, 32)", style=solid]; +"1451 permute_52" -> "1454 select_35" [label="(3, 4, 12, 64, 32)", style=solid]; +"1452 select_33" -> "1455 linalg_vector_norm_22" [label="(4, 12, 64, 32)", style=solid]; +"1452 select_33" -> "1457 expand_as_22" [label="(4, 12, 64, 32)", style=solid]; +"1452 select_33" -> "1458 div_22" [label="(4, 12, 64, 32)", style=solid]; +"1453 select_34" -> "1459 linalg_vector_norm_23" [label="(4, 12, 64, 32)", style=solid]; +"1453 select_34" -> "1461 expand_as_23" [label="(4, 12, 64, 32)", style=solid]; +"1453 select_34" -> "1462 div_23" [label="(4, 12, 64, 32)", style=solid]; +"1454 select_35" -> "1487 matmul_23" [label="(4, 12, 64, 32)", style=solid]; +"1455 linalg_vector_norm_22" -> "1456 clamp_min_22" [label="(4, 12, 64, 1)", style=solid]; +"1456 clamp_min_22" -> "1457 expand_as_22" [label="(4, 12, 64, 1)", style=solid]; +"1457 expand_as_22" -> "1458 div_22" [label="(4, 12, 64, 32)", style=solid]; +"1458 div_22" -> "1464 matmul_22" [label="(4, 12, 64, 32)", style=solid]; +"1459 linalg_vector_norm_23" -> "1460 clamp_min_23" [label="(4, 12, 64, 1)", style=solid]; +"1460 clamp_min_23" -> "1461 expand_as_23" [label="(4, 12, 64, 1)", style=solid]; +"1461 expand_as_23" -> "1462 div_23" [label="(4, 12, 64, 32)", style=solid]; +"1462 div_23" -> "1463 transpose_22" [label="(4, 12, 64, 32)", style=solid]; +"1463 transpose_22" -> "1464 matmul_22" [label="(4, 12, 32, 64)", style=solid]; +"1464 matmul_22" -> "1467 mul_23" [label="(4, 12, 64, 64)", style=solid]; +"1465 clamp_11" -> "1466 exp_11" [label="(12, 1, 1)", style=solid]; +"1466 exp_11" -> "1467 mul_23" [label="(12, 1, 1)", style=solid]; +"1467 mul_23" -> "1468 add_38" [label="(4, 12, 64, 64)", style=solid]; +"1468 add_38" -> "1480 view_63" [label="(4, 12, 64, 64)", style=solid]; +"1469 new_zeros_5" -> "1470 view_62" [label="(16, 16)", style=solid]; +"1470 view_62" -> "1471 permute_53" [label="(2, 8, 2, 8)", style=solid]; +"1471 permute_53" -> "1472 reshape_51" [label="(2, 2, 8, 8)", style=solid]; +"1472 reshape_51" -> "1473 unsqueeze_32" [label="(4, 64)", style=solid]; +"1472 reshape_51" -> "1474 unsqueeze_33" [label="(4, 64)", style=solid]; +"1473 unsqueeze_32" -> "1475 sub_5" [label="(4, 1, 64)", style=solid]; +"1474 unsqueeze_33" -> "1475 sub_5" [label="(4, 64, 1)", style=solid]; +"1475 sub_5" -> "1476 ne_5" [label="(4, 64, 64)", style=solid]; +"1475 sub_5" -> "1477 masked_fill_10" [label="(4, 64, 64)", style=solid]; +"1475 sub_5" -> "1478 eq_5" [label="(4, 64, 64)", style=solid]; +"1476 ne_5" -> "1477 masked_fill_10" [label="(4, 64, 64)", style=solid]; +"1477 masked_fill_10" -> "1479 masked_fill_11" [label="(4, 64, 64)", style=solid]; +"1478 eq_5" -> "1479 masked_fill_11" [label="(4, 64, 64)", style=solid]; +"1479 masked_fill_11" -> "1481 unsqueeze_34" [label="(4, 64, 64)", style=solid]; +"1480 view_63" -> "1483 add_39" [label="(1, 4, 12, 64, 64)", style=solid]; +"1481 unsqueeze_34" -> "1482 unsqueeze_35" [label="(4, 1, 64, 64)", style=solid]; +"1482 unsqueeze_35" -> "1483 add_39" [label="(1, 4, 1, 64, 64)", style=solid]; +"1483 add_39" -> "1484 view_64" [label="(1, 4, 12, 64, 64)", style=solid]; +"1484 view_64" -> "1485 softmax_11" [label="(4, 12, 64, 64)", style=solid]; +"1485 softmax_11" -> "1486 dropout_44" [label="(4, 12, 64, 64)", style=solid]; +"1486 dropout_44" -> "1487 matmul_23" [label="(4, 12, 64, 64)", style=solid]; +"1487 matmul_23" -> "1488 transpose_23" [label="(4, 12, 64, 32)", style=solid]; +"1488 transpose_23" -> "1489 reshape_52" [label="(4, 64, 12, 32)", style=solid]; +"1489 reshape_52" -> "1491 reshape_52_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1490 features_5_7_attn_proj_weight_updated_constant0" -> "1492 linear_71" [label="(384, 384)", style=solid]; +"1491 reshape_52_0_0_nncf_smooth_quant_0" -> "1492 linear_71" [label="(4, 64, 384)", style=solid]; +"1492 linear_71" -> "1493 dropout_45" [label="(4, 64, 384)", style=solid]; +"1493 dropout_45" -> "1494 view_65" [label="(4, 64, 384)", style=solid]; +"1494 view_65" -> "1495 permute_54" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1495 permute_54" -> "1496 reshape_53" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1496 reshape_53" -> "1497 roll_11" [label="(1, 16, 16, 384)", style=solid]; +"1497 roll_11" -> "1498 slice_185" [label="(1, 16, 16, 384)", style=solid]; +"1498 slice_185" -> "1499 slice_186" [label="(1, 16, 16, 384)", style=solid]; +"1499 slice_186" -> "1500 slice_187" [label="(1, 14, 16, 384)", style=solid]; +"1500 slice_187" -> "1501 slice_188" [label="(1, 14, 14, 384)", style=solid]; +"1501 slice_188" -> "1502 contiguous_21" [label="(1, 14, 14, 384)", style=solid]; +"1502 contiguous_21" -> "1503 layer_norm_25" [label="(1, 14, 14, 384)", style=solid]; +"1503 layer_norm_25" -> "1504 add_40" [label="(1, 14, 14, 384)", style=solid]; +"1504 add_40" -> "1505 add_40_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"1504 add_40" -> "1525 add_41" [label="(1, 14, 14, 384)", style=solid]; +"1505 add_40_0_0_nncf_smooth_quant_0" -> "1506 quantize_per_tensor_default_49" [label="(1, 14, 14, 384)", style=solid]; +"1506 quantize_per_tensor_default_49" -> "1507 dequantize_per_tensor_default_49" [label="(1, 14, 14, 384)", style=solid]; +"1507 dequantize_per_tensor_default_49" -> "1512 linear_72" [label="(1, 14, 14, 384)", style=solid]; +"1508 linear_72_scale_0" -> "1511 dequantize_per_channel_default_49" [label="(1536,)", style=solid]; +"1509 linear_72_zero_point_0" -> "1511 dequantize_per_channel_default_49" [label="(1536,)", style=solid]; +"1510 _frozen_param61" -> "1511 dequantize_per_channel_default_49" [label="(1536, 384)", style=solid]; +"1511 dequantize_per_channel_default_49" -> "1512 linear_72" [label="(1536, 384)", style=solid]; +"1512 linear_72" -> "1513 gelu_11" [label="(1, 14, 14, 1536)", style=solid]; +"1513 gelu_11" -> "1514 dropout_46" [label="(1, 14, 14, 1536)", style=solid]; +"1514 dropout_46" -> "1515 dropout_46_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"1515 dropout_46_0_0_nncf_smooth_quant_0" -> "1516 quantize_per_tensor_default_50" [label="(1, 14, 14, 1536)", style=solid]; +"1516 quantize_per_tensor_default_50" -> "1517 dequantize_per_tensor_default_50" [label="(1, 14, 14, 1536)", style=solid]; +"1517 dequantize_per_tensor_default_50" -> "1522 linear_73" [label="(1, 14, 14, 1536)", style=solid]; +"1518 linear_73_scale_0" -> "1521 dequantize_per_channel_default_50" [label="(384,)", style=solid]; +"1519 linear_73_zero_point_0" -> "1521 dequantize_per_channel_default_50" [label="(384,)", style=solid]; +"1520 _frozen_param62" -> "1521 dequantize_per_channel_default_50" [label="(384, 1536)", style=solid]; +"1521 dequantize_per_channel_default_50" -> "1522 linear_73" [label="(384, 1536)", style=solid]; +"1522 linear_73" -> "1523 dropout_47" [label="(1, 14, 14, 384)", style=solid]; +"1523 dropout_47" -> "1524 layer_norm_26" [label="(1, 14, 14, 384)", style=solid]; +"1524 layer_norm_26" -> "1525 add_41" [label="(1, 14, 14, 384)", style=solid]; +"1525 add_41" -> "1550 pad_14" [label="(1, 14, 14, 384)", style=solid]; +"1525 add_41" -> "1595 add_43" [label="(1, 14, 14, 384)", style=solid]; +"1526 _frozen_param63" -> "1527 dequantize_per_tensor_default_51" [label="(1, 15, 15, 2)", style=solid]; +"1527 dequantize_per_tensor_default_51" -> "1532 linear_74" [label="(1, 15, 15, 2)", style=solid]; +"1528 linear_74_scale_0" -> "1531 dequantize_per_channel_default_51" [label="(512,)", style=solid]; +"1529 linear_74_zero_point_0" -> "1531 dequantize_per_channel_default_51" [label="(512,)", style=solid]; +"1530 _frozen_param64" -> "1531 dequantize_per_channel_default_51" [label="(512, 2)", style=solid]; +"1531 dequantize_per_channel_default_51" -> "1532 linear_74" [label="(512, 2)", style=solid]; +"1532 linear_74" -> "1533 relu__12" [label="(1, 15, 15, 512)", style=solid]; +"1533 relu__12" -> "1534 relu__12_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"1534 relu__12_0_0_nncf_smooth_quant_0" -> "1535 quantize_per_tensor_default_52" [label="(1, 15, 15, 512)", style=solid]; +"1535 quantize_per_tensor_default_52" -> "1536 dequantize_per_tensor_default_52" [label="(1, 15, 15, 512)", style=solid]; +"1536 dequantize_per_tensor_default_52" -> "1541 linear_75" [label="(1, 15, 15, 512)", style=solid]; +"1537 linear_75_scale_0" -> "1540 dequantize_per_channel_default_52" [label="(12,)", style=solid]; +"1538 linear_75_zero_point_0" -> "1540 dequantize_per_channel_default_52" [label="(12,)", style=solid]; +"1539 _frozen_param65" -> "1540 dequantize_per_channel_default_52" [label="(12, 512)", style=solid]; +"1540 dequantize_per_channel_default_52" -> "1541 linear_75" [label="(12, 512)", style=solid]; +"1541 linear_75" -> "1542 view_66" [label="(1, 15, 15, 12)", style=solid]; +"1542 view_66" -> "1543 index_12" [label="(225, 12)", style=solid]; +"1543 index_12" -> "1544 view_67" [label="(4096, 12)", style=solid]; +"1544 view_67" -> "1545 permute_55" [label="(64, 64, 12)", style=solid]; +"1545 permute_55" -> "1546 contiguous_22" [label="(12, 64, 64)", style=solid]; +"1546 contiguous_22" -> "1547 unsqueeze_36" [label="(12, 64, 64)", style=solid]; +"1547 unsqueeze_36" -> "1548 sigmoid_12" [label="(1, 12, 64, 64)", style=solid]; +"1548 sigmoid_12" -> "1549 mul_24" [label="(1, 12, 64, 64)", style=solid]; +"1549 mul_24" -> "1576 add_42" [label="(1, 12, 64, 64)", style=solid]; +"1550 pad_14" -> "1551 view_68" [label="(1, 16, 16, 384)", style=solid]; +"1551 view_68" -> "1552 permute_56" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1552 permute_56" -> "1553 reshape_54" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1553 reshape_54" -> "1556 reshape_54_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1554 clone_12" -> "1557 linear_76" [label="(1152,)", style=solid]; +"1555 features_5_8_attn_qkv_weight_updated_constant0" -> "1557 linear_76" [label="(1152, 384)", style=solid]; +"1556 reshape_54_0_0_nncf_smooth_quant_0" -> "1557 linear_76" [label="(4, 64, 384)", style=solid]; +"1557 linear_76" -> "1558 reshape_55" [label="(4, 64, 1152)", style=solid]; +"1558 reshape_55" -> "1559 permute_57" [label="(4, 64, 3, 12, 32)", style=solid]; +"1559 permute_57" -> "1560 select_36" [label="(3, 4, 12, 64, 32)", style=solid]; +"1559 permute_57" -> "1561 select_37" [label="(3, 4, 12, 64, 32)", style=solid]; +"1559 permute_57" -> "1562 select_38" [label="(3, 4, 12, 64, 32)", style=solid]; +"1560 select_36" -> "1563 linalg_vector_norm_24" [label="(4, 12, 64, 32)", style=solid]; +"1560 select_36" -> "1565 expand_as_24" [label="(4, 12, 64, 32)", style=solid]; +"1560 select_36" -> "1566 div_24" [label="(4, 12, 64, 32)", style=solid]; +"1561 select_37" -> "1567 linalg_vector_norm_25" [label="(4, 12, 64, 32)", style=solid]; +"1561 select_37" -> "1569 expand_as_25" [label="(4, 12, 64, 32)", style=solid]; +"1561 select_37" -> "1570 div_25" [label="(4, 12, 64, 32)", style=solid]; +"1562 select_38" -> "1579 matmul_25" [label="(4, 12, 64, 32)", style=solid]; +"1563 linalg_vector_norm_24" -> "1564 clamp_min_24" [label="(4, 12, 64, 1)", style=solid]; +"1564 clamp_min_24" -> "1565 expand_as_24" [label="(4, 12, 64, 1)", style=solid]; +"1565 expand_as_24" -> "1566 div_24" [label="(4, 12, 64, 32)", style=solid]; +"1566 div_24" -> "1572 matmul_24" [label="(4, 12, 64, 32)", style=solid]; +"1567 linalg_vector_norm_25" -> "1568 clamp_min_25" [label="(4, 12, 64, 1)", style=solid]; +"1568 clamp_min_25" -> "1569 expand_as_25" [label="(4, 12, 64, 1)", style=solid]; +"1569 expand_as_25" -> "1570 div_25" [label="(4, 12, 64, 32)", style=solid]; +"1570 div_25" -> "1571 transpose_24" [label="(4, 12, 64, 32)", style=solid]; +"1571 transpose_24" -> "1572 matmul_24" [label="(4, 12, 32, 64)", style=solid]; +"1572 matmul_24" -> "1575 mul_25" [label="(4, 12, 64, 64)", style=solid]; +"1573 clamp_12" -> "1574 exp_12" [label="(12, 1, 1)", style=solid]; +"1574 exp_12" -> "1575 mul_25" [label="(12, 1, 1)", style=solid]; +"1575 mul_25" -> "1576 add_42" [label="(4, 12, 64, 64)", style=solid]; +"1576 add_42" -> "1577 softmax_12" [label="(4, 12, 64, 64)", style=solid]; +"1577 softmax_12" -> "1578 dropout_48" [label="(4, 12, 64, 64)", style=solid]; +"1578 dropout_48" -> "1579 matmul_25" [label="(4, 12, 64, 64)", style=solid]; +"1579 matmul_25" -> "1580 transpose_25" [label="(4, 12, 64, 32)", style=solid]; +"1580 transpose_25" -> "1581 reshape_56" [label="(4, 64, 12, 32)", style=solid]; +"1581 reshape_56" -> "1583 reshape_56_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1582 features_5_8_attn_proj_weight_updated_constant0" -> "1584 linear_77" [label="(384, 384)", style=solid]; +"1583 reshape_56_0_0_nncf_smooth_quant_0" -> "1584 linear_77" [label="(4, 64, 384)", style=solid]; +"1584 linear_77" -> "1585 dropout_49" [label="(4, 64, 384)", style=solid]; +"1585 dropout_49" -> "1586 view_69" [label="(4, 64, 384)", style=solid]; +"1586 view_69" -> "1587 permute_58" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1587 permute_58" -> "1588 reshape_57" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1588 reshape_57" -> "1589 slice_190" [label="(1, 16, 16, 384)", style=solid]; +"1589 slice_190" -> "1590 slice_191" [label="(1, 16, 16, 384)", style=solid]; +"1590 slice_191" -> "1591 slice_192" [label="(1, 14, 16, 384)", style=solid]; +"1591 slice_192" -> "1592 slice_193" [label="(1, 14, 14, 384)", style=solid]; +"1592 slice_193" -> "1593 contiguous_23" [label="(1, 14, 14, 384)", style=solid]; +"1593 contiguous_23" -> "1594 layer_norm_27" [label="(1, 14, 14, 384)", style=solid]; +"1594 layer_norm_27" -> "1595 add_43" [label="(1, 14, 14, 384)", style=solid]; +"1595 add_43" -> "1596 add_43_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"1595 add_43" -> "1616 add_44" [label="(1, 14, 14, 384)", style=solid]; +"1596 add_43_0_0_nncf_smooth_quant_0" -> "1597 quantize_per_tensor_default_53" [label="(1, 14, 14, 384)", style=solid]; +"1597 quantize_per_tensor_default_53" -> "1598 dequantize_per_tensor_default_53" [label="(1, 14, 14, 384)", style=solid]; +"1598 dequantize_per_tensor_default_53" -> "1603 linear_78" [label="(1, 14, 14, 384)", style=solid]; +"1599 linear_78_scale_0" -> "1602 dequantize_per_channel_default_53" [label="(1536,)", style=solid]; +"1600 linear_78_zero_point_0" -> "1602 dequantize_per_channel_default_53" [label="(1536,)", style=solid]; +"1601 _frozen_param66" -> "1602 dequantize_per_channel_default_53" [label="(1536, 384)", style=solid]; +"1602 dequantize_per_channel_default_53" -> "1603 linear_78" [label="(1536, 384)", style=solid]; +"1603 linear_78" -> "1604 gelu_12" [label="(1, 14, 14, 1536)", style=solid]; +"1604 gelu_12" -> "1605 dropout_50" [label="(1, 14, 14, 1536)", style=solid]; +"1605 dropout_50" -> "1606 dropout_50_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"1606 dropout_50_0_0_nncf_smooth_quant_0" -> "1607 quantize_per_tensor_default_54" [label="(1, 14, 14, 1536)", style=solid]; +"1607 quantize_per_tensor_default_54" -> "1608 dequantize_per_tensor_default_54" [label="(1, 14, 14, 1536)", style=solid]; +"1608 dequantize_per_tensor_default_54" -> "1613 linear_79" [label="(1, 14, 14, 1536)", style=solid]; +"1609 linear_79_scale_0" -> "1612 dequantize_per_channel_default_54" [label="(384,)", style=solid]; +"1610 linear_79_zero_point_0" -> "1612 dequantize_per_channel_default_54" [label="(384,)", style=solid]; +"1611 _frozen_param67" -> "1612 dequantize_per_channel_default_54" [label="(384, 1536)", style=solid]; +"1612 dequantize_per_channel_default_54" -> "1613 linear_79" [label="(384, 1536)", style=solid]; +"1613 linear_79" -> "1614 dropout_51" [label="(1, 14, 14, 384)", style=solid]; +"1614 dropout_51" -> "1615 layer_norm_28" [label="(1, 14, 14, 384)", style=solid]; +"1615 layer_norm_28" -> "1616 add_44" [label="(1, 14, 14, 384)", style=solid]; +"1616 add_44" -> "1641 pad_15" [label="(1, 14, 14, 384)", style=solid]; +"1616 add_44" -> "1704 add_47" [label="(1, 14, 14, 384)", style=solid]; +"1617 _frozen_param68" -> "1618 dequantize_per_tensor_default_55" [label="(1, 15, 15, 2)", style=solid]; +"1618 dequantize_per_tensor_default_55" -> "1623 linear_80" [label="(1, 15, 15, 2)", style=solid]; +"1619 linear_80_scale_0" -> "1622 dequantize_per_channel_default_55" [label="(512,)", style=solid]; +"1620 linear_80_zero_point_0" -> "1622 dequantize_per_channel_default_55" [label="(512,)", style=solid]; +"1621 _frozen_param69" -> "1622 dequantize_per_channel_default_55" [label="(512, 2)", style=solid]; +"1622 dequantize_per_channel_default_55" -> "1623 linear_80" [label="(512, 2)", style=solid]; +"1623 linear_80" -> "1624 relu__13" [label="(1, 15, 15, 512)", style=solid]; +"1624 relu__13" -> "1625 relu__13_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"1625 relu__13_0_0_nncf_smooth_quant_0" -> "1626 quantize_per_tensor_default_56" [label="(1, 15, 15, 512)", style=solid]; +"1626 quantize_per_tensor_default_56" -> "1627 dequantize_per_tensor_default_56" [label="(1, 15, 15, 512)", style=solid]; +"1627 dequantize_per_tensor_default_56" -> "1632 linear_81" [label="(1, 15, 15, 512)", style=solid]; +"1628 linear_81_scale_0" -> "1631 dequantize_per_channel_default_56" [label="(12,)", style=solid]; +"1629 linear_81_zero_point_0" -> "1631 dequantize_per_channel_default_56" [label="(12,)", style=solid]; +"1630 _frozen_param70" -> "1631 dequantize_per_channel_default_56" [label="(12, 512)", style=solid]; +"1631 dequantize_per_channel_default_56" -> "1632 linear_81" [label="(12, 512)", style=solid]; +"1632 linear_81" -> "1633 view_70" [label="(1, 15, 15, 12)", style=solid]; +"1633 view_70" -> "1634 index_13" [label="(225, 12)", style=solid]; +"1634 index_13" -> "1635 view_71" [label="(4096, 12)", style=solid]; +"1635 view_71" -> "1636 permute_59" [label="(64, 64, 12)", style=solid]; +"1636 permute_59" -> "1637 contiguous_24" [label="(12, 64, 64)", style=solid]; +"1637 contiguous_24" -> "1638 unsqueeze_37" [label="(12, 64, 64)", style=solid]; +"1638 unsqueeze_37" -> "1639 sigmoid_13" [label="(1, 12, 64, 64)", style=solid]; +"1639 sigmoid_13" -> "1640 mul_26" [label="(1, 12, 64, 64)", style=solid]; +"1640 mul_26" -> "1668 add_45" [label="(1, 12, 64, 64)", style=solid]; +"1641 pad_15" -> "1642 roll_12" [label="(1, 16, 16, 384)", style=solid]; +"1642 roll_12" -> "1643 view_72" [label="(1, 16, 16, 384)", style=solid]; +"1643 view_72" -> "1644 permute_60" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1644 permute_60" -> "1645 reshape_58" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1645 reshape_58" -> "1648 reshape_58_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1645 reshape_58" -> "1669 new_zeros_6" [label="(4, 64, 384)", style=solid]; +"1646 clone_13" -> "1649 linear_82" [label="(1152,)", style=solid]; +"1647 features_5_9_attn_qkv_weight_updated_constant0" -> "1649 linear_82" [label="(1152, 384)", style=solid]; +"1648 reshape_58_0_0_nncf_smooth_quant_0" -> "1649 linear_82" [label="(4, 64, 384)", style=solid]; +"1649 linear_82" -> "1650 reshape_59" [label="(4, 64, 1152)", style=solid]; +"1650 reshape_59" -> "1651 permute_61" [label="(4, 64, 3, 12, 32)", style=solid]; +"1651 permute_61" -> "1652 select_39" [label="(3, 4, 12, 64, 32)", style=solid]; +"1651 permute_61" -> "1653 select_40" [label="(3, 4, 12, 64, 32)", style=solid]; +"1651 permute_61" -> "1654 select_41" [label="(3, 4, 12, 64, 32)", style=solid]; +"1652 select_39" -> "1655 linalg_vector_norm_26" [label="(4, 12, 64, 32)", style=solid]; +"1652 select_39" -> "1657 expand_as_26" [label="(4, 12, 64, 32)", style=solid]; +"1652 select_39" -> "1658 div_26" [label="(4, 12, 64, 32)", style=solid]; +"1653 select_40" -> "1659 linalg_vector_norm_27" [label="(4, 12, 64, 32)", style=solid]; +"1653 select_40" -> "1661 expand_as_27" [label="(4, 12, 64, 32)", style=solid]; +"1653 select_40" -> "1662 div_27" [label="(4, 12, 64, 32)", style=solid]; +"1654 select_41" -> "1687 matmul_27" [label="(4, 12, 64, 32)", style=solid]; +"1655 linalg_vector_norm_26" -> "1656 clamp_min_26" [label="(4, 12, 64, 1)", style=solid]; +"1656 clamp_min_26" -> "1657 expand_as_26" [label="(4, 12, 64, 1)", style=solid]; +"1657 expand_as_26" -> "1658 div_26" [label="(4, 12, 64, 32)", style=solid]; +"1658 div_26" -> "1664 matmul_26" [label="(4, 12, 64, 32)", style=solid]; +"1659 linalg_vector_norm_27" -> "1660 clamp_min_27" [label="(4, 12, 64, 1)", style=solid]; +"1660 clamp_min_27" -> "1661 expand_as_27" [label="(4, 12, 64, 1)", style=solid]; +"1661 expand_as_27" -> "1662 div_27" [label="(4, 12, 64, 32)", style=solid]; +"1662 div_27" -> "1663 transpose_26" [label="(4, 12, 64, 32)", style=solid]; +"1663 transpose_26" -> "1664 matmul_26" [label="(4, 12, 32, 64)", style=solid]; +"1664 matmul_26" -> "1667 mul_27" [label="(4, 12, 64, 64)", style=solid]; +"1665 clamp_13" -> "1666 exp_13" [label="(12, 1, 1)", style=solid]; +"1666 exp_13" -> "1667 mul_27" [label="(12, 1, 1)", style=solid]; +"1667 mul_27" -> "1668 add_45" [label="(4, 12, 64, 64)", style=solid]; +"1668 add_45" -> "1680 view_74" [label="(4, 12, 64, 64)", style=solid]; +"1669 new_zeros_6" -> "1670 view_73" [label="(16, 16)", style=solid]; +"1670 view_73" -> "1671 permute_62" [label="(2, 8, 2, 8)", style=solid]; +"1671 permute_62" -> "1672 reshape_60" [label="(2, 2, 8, 8)", style=solid]; +"1672 reshape_60" -> "1673 unsqueeze_38" [label="(4, 64)", style=solid]; +"1672 reshape_60" -> "1674 unsqueeze_39" [label="(4, 64)", style=solid]; +"1673 unsqueeze_38" -> "1675 sub_6" [label="(4, 1, 64)", style=solid]; +"1674 unsqueeze_39" -> "1675 sub_6" [label="(4, 64, 1)", style=solid]; +"1675 sub_6" -> "1676 ne_6" [label="(4, 64, 64)", style=solid]; +"1675 sub_6" -> "1677 masked_fill_12" [label="(4, 64, 64)", style=solid]; +"1675 sub_6" -> "1678 eq_6" [label="(4, 64, 64)", style=solid]; +"1676 ne_6" -> "1677 masked_fill_12" [label="(4, 64, 64)", style=solid]; +"1677 masked_fill_12" -> "1679 masked_fill_13" [label="(4, 64, 64)", style=solid]; +"1678 eq_6" -> "1679 masked_fill_13" [label="(4, 64, 64)", style=solid]; +"1679 masked_fill_13" -> "1681 unsqueeze_40" [label="(4, 64, 64)", style=solid]; +"1680 view_74" -> "1683 add_46" [label="(1, 4, 12, 64, 64)", style=solid]; +"1681 unsqueeze_40" -> "1682 unsqueeze_41" [label="(4, 1, 64, 64)", style=solid]; +"1682 unsqueeze_41" -> "1683 add_46" [label="(1, 4, 1, 64, 64)", style=solid]; +"1683 add_46" -> "1684 view_75" [label="(1, 4, 12, 64, 64)", style=solid]; +"1684 view_75" -> "1685 softmax_13" [label="(4, 12, 64, 64)", style=solid]; +"1685 softmax_13" -> "1686 dropout_52" [label="(4, 12, 64, 64)", style=solid]; +"1686 dropout_52" -> "1687 matmul_27" [label="(4, 12, 64, 64)", style=solid]; +"1687 matmul_27" -> "1688 transpose_27" [label="(4, 12, 64, 32)", style=solid]; +"1688 transpose_27" -> "1689 reshape_61" [label="(4, 64, 12, 32)", style=solid]; +"1689 reshape_61" -> "1691 reshape_61_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1690 features_5_9_attn_proj_weight_updated_constant0" -> "1692 linear_83" [label="(384, 384)", style=solid]; +"1691 reshape_61_0_0_nncf_smooth_quant_0" -> "1692 linear_83" [label="(4, 64, 384)", style=solid]; +"1692 linear_83" -> "1693 dropout_53" [label="(4, 64, 384)", style=solid]; +"1693 dropout_53" -> "1694 view_76" [label="(4, 64, 384)", style=solid]; +"1694 view_76" -> "1695 permute_63" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1695 permute_63" -> "1696 reshape_62" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1696 reshape_62" -> "1697 roll_13" [label="(1, 16, 16, 384)", style=solid]; +"1697 roll_13" -> "1698 slice_213" [label="(1, 16, 16, 384)", style=solid]; +"1698 slice_213" -> "1699 slice_214" [label="(1, 16, 16, 384)", style=solid]; +"1699 slice_214" -> "1700 slice_215" [label="(1, 14, 16, 384)", style=solid]; +"1700 slice_215" -> "1701 slice_216" [label="(1, 14, 14, 384)", style=solid]; +"1701 slice_216" -> "1702 contiguous_25" [label="(1, 14, 14, 384)", style=solid]; +"1702 contiguous_25" -> "1703 layer_norm_29" [label="(1, 14, 14, 384)", style=solid]; +"1703 layer_norm_29" -> "1704 add_47" [label="(1, 14, 14, 384)", style=solid]; +"1704 add_47" -> "1705 add_47_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"1704 add_47" -> "1725 add_48" [label="(1, 14, 14, 384)", style=solid]; +"1705 add_47_0_0_nncf_smooth_quant_0" -> "1706 quantize_per_tensor_default_57" [label="(1, 14, 14, 384)", style=solid]; +"1706 quantize_per_tensor_default_57" -> "1707 dequantize_per_tensor_default_57" [label="(1, 14, 14, 384)", style=solid]; +"1707 dequantize_per_tensor_default_57" -> "1712 linear_84" [label="(1, 14, 14, 384)", style=solid]; +"1708 linear_84_scale_0" -> "1711 dequantize_per_channel_default_57" [label="(1536,)", style=solid]; +"1709 linear_84_zero_point_0" -> "1711 dequantize_per_channel_default_57" [label="(1536,)", style=solid]; +"1710 _frozen_param71" -> "1711 dequantize_per_channel_default_57" [label="(1536, 384)", style=solid]; +"1711 dequantize_per_channel_default_57" -> "1712 linear_84" [label="(1536, 384)", style=solid]; +"1712 linear_84" -> "1713 gelu_13" [label="(1, 14, 14, 1536)", style=solid]; +"1713 gelu_13" -> "1714 dropout_54" [label="(1, 14, 14, 1536)", style=solid]; +"1714 dropout_54" -> "1715 dropout_54_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"1715 dropout_54_0_0_nncf_smooth_quant_0" -> "1716 quantize_per_tensor_default_58" [label="(1, 14, 14, 1536)", style=solid]; +"1716 quantize_per_tensor_default_58" -> "1717 dequantize_per_tensor_default_58" [label="(1, 14, 14, 1536)", style=solid]; +"1717 dequantize_per_tensor_default_58" -> "1722 linear_85" [label="(1, 14, 14, 1536)", style=solid]; +"1718 linear_85_scale_0" -> "1721 dequantize_per_channel_default_58" [label="(384,)", style=solid]; +"1719 linear_85_zero_point_0" -> "1721 dequantize_per_channel_default_58" [label="(384,)", style=solid]; +"1720 _frozen_param72" -> "1721 dequantize_per_channel_default_58" [label="(384, 1536)", style=solid]; +"1721 dequantize_per_channel_default_58" -> "1722 linear_85" [label="(384, 1536)", style=solid]; +"1722 linear_85" -> "1723 dropout_55" [label="(1, 14, 14, 384)", style=solid]; +"1723 dropout_55" -> "1724 layer_norm_30" [label="(1, 14, 14, 384)", style=solid]; +"1724 layer_norm_30" -> "1725 add_48" [label="(1, 14, 14, 384)", style=solid]; +"1725 add_48" -> "1750 pad_16" [label="(1, 14, 14, 384)", style=solid]; +"1725 add_48" -> "1795 add_50" [label="(1, 14, 14, 384)", style=solid]; +"1726 _frozen_param73" -> "1727 dequantize_per_tensor_default_59" [label="(1, 15, 15, 2)", style=solid]; +"1727 dequantize_per_tensor_default_59" -> "1732 linear_86" [label="(1, 15, 15, 2)", style=solid]; +"1728 linear_86_scale_0" -> "1731 dequantize_per_channel_default_59" [label="(512,)", style=solid]; +"1729 linear_86_zero_point_0" -> "1731 dequantize_per_channel_default_59" [label="(512,)", style=solid]; +"1730 _frozen_param74" -> "1731 dequantize_per_channel_default_59" [label="(512, 2)", style=solid]; +"1731 dequantize_per_channel_default_59" -> "1732 linear_86" [label="(512, 2)", style=solid]; +"1732 linear_86" -> "1733 relu__14" [label="(1, 15, 15, 512)", style=solid]; +"1733 relu__14" -> "1734 relu__14_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"1734 relu__14_0_0_nncf_smooth_quant_0" -> "1735 quantize_per_tensor_default_60" [label="(1, 15, 15, 512)", style=solid]; +"1735 quantize_per_tensor_default_60" -> "1736 dequantize_per_tensor_default_60" [label="(1, 15, 15, 512)", style=solid]; +"1736 dequantize_per_tensor_default_60" -> "1741 linear_87" [label="(1, 15, 15, 512)", style=solid]; +"1737 linear_87_scale_0" -> "1740 dequantize_per_channel_default_60" [label="(12,)", style=solid]; +"1738 linear_87_zero_point_0" -> "1740 dequantize_per_channel_default_60" [label="(12,)", style=solid]; +"1739 _frozen_param75" -> "1740 dequantize_per_channel_default_60" [label="(12, 512)", style=solid]; +"1740 dequantize_per_channel_default_60" -> "1741 linear_87" [label="(12, 512)", style=solid]; +"1741 linear_87" -> "1742 view_77" [label="(1, 15, 15, 12)", style=solid]; +"1742 view_77" -> "1743 index_14" [label="(225, 12)", style=solid]; +"1743 index_14" -> "1744 view_78" [label="(4096, 12)", style=solid]; +"1744 view_78" -> "1745 permute_64" [label="(64, 64, 12)", style=solid]; +"1745 permute_64" -> "1746 contiguous_26" [label="(12, 64, 64)", style=solid]; +"1746 contiguous_26" -> "1747 unsqueeze_42" [label="(12, 64, 64)", style=solid]; +"1747 unsqueeze_42" -> "1748 sigmoid_14" [label="(1, 12, 64, 64)", style=solid]; +"1748 sigmoid_14" -> "1749 mul_28" [label="(1, 12, 64, 64)", style=solid]; +"1749 mul_28" -> "1776 add_49" [label="(1, 12, 64, 64)", style=solid]; +"1750 pad_16" -> "1751 view_79" [label="(1, 16, 16, 384)", style=solid]; +"1751 view_79" -> "1752 permute_65" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1752 permute_65" -> "1753 reshape_63" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1753 reshape_63" -> "1756 reshape_63_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1754 clone_14" -> "1757 linear_88" [label="(1152,)", style=solid]; +"1755 features_5_10_attn_qkv_weight_updated_constant0" -> "1757 linear_88" [label="(1152, 384)", style=solid]; +"1756 reshape_63_0_0_nncf_smooth_quant_0" -> "1757 linear_88" [label="(4, 64, 384)", style=solid]; +"1757 linear_88" -> "1758 reshape_64" [label="(4, 64, 1152)", style=solid]; +"1758 reshape_64" -> "1759 permute_66" [label="(4, 64, 3, 12, 32)", style=solid]; +"1759 permute_66" -> "1760 select_42" [label="(3, 4, 12, 64, 32)", style=solid]; +"1759 permute_66" -> "1761 select_43" [label="(3, 4, 12, 64, 32)", style=solid]; +"1759 permute_66" -> "1762 select_44" [label="(3, 4, 12, 64, 32)", style=solid]; +"1760 select_42" -> "1763 linalg_vector_norm_28" [label="(4, 12, 64, 32)", style=solid]; +"1760 select_42" -> "1765 expand_as_28" [label="(4, 12, 64, 32)", style=solid]; +"1760 select_42" -> "1766 div_28" [label="(4, 12, 64, 32)", style=solid]; +"1761 select_43" -> "1767 linalg_vector_norm_29" [label="(4, 12, 64, 32)", style=solid]; +"1761 select_43" -> "1769 expand_as_29" [label="(4, 12, 64, 32)", style=solid]; +"1761 select_43" -> "1770 div_29" [label="(4, 12, 64, 32)", style=solid]; +"1762 select_44" -> "1779 matmul_29" [label="(4, 12, 64, 32)", style=solid]; +"1763 linalg_vector_norm_28" -> "1764 clamp_min_28" [label="(4, 12, 64, 1)", style=solid]; +"1764 clamp_min_28" -> "1765 expand_as_28" [label="(4, 12, 64, 1)", style=solid]; +"1765 expand_as_28" -> "1766 div_28" [label="(4, 12, 64, 32)", style=solid]; +"1766 div_28" -> "1772 matmul_28" [label="(4, 12, 64, 32)", style=solid]; +"1767 linalg_vector_norm_29" -> "1768 clamp_min_29" [label="(4, 12, 64, 1)", style=solid]; +"1768 clamp_min_29" -> "1769 expand_as_29" [label="(4, 12, 64, 1)", style=solid]; +"1769 expand_as_29" -> "1770 div_29" [label="(4, 12, 64, 32)", style=solid]; +"1770 div_29" -> "1771 transpose_28" [label="(4, 12, 64, 32)", style=solid]; +"1771 transpose_28" -> "1772 matmul_28" [label="(4, 12, 32, 64)", style=solid]; +"1772 matmul_28" -> "1775 mul_29" [label="(4, 12, 64, 64)", style=solid]; +"1773 clamp_14" -> "1774 exp_14" [label="(12, 1, 1)", style=solid]; +"1774 exp_14" -> "1775 mul_29" [label="(12, 1, 1)", style=solid]; +"1775 mul_29" -> "1776 add_49" [label="(4, 12, 64, 64)", style=solid]; +"1776 add_49" -> "1777 softmax_14" [label="(4, 12, 64, 64)", style=solid]; +"1777 softmax_14" -> "1778 dropout_56" [label="(4, 12, 64, 64)", style=solid]; +"1778 dropout_56" -> "1779 matmul_29" [label="(4, 12, 64, 64)", style=solid]; +"1779 matmul_29" -> "1780 transpose_29" [label="(4, 12, 64, 32)", style=solid]; +"1780 transpose_29" -> "1781 reshape_65" [label="(4, 64, 12, 32)", style=solid]; +"1781 reshape_65" -> "1783 reshape_65_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1782 features_5_10_attn_proj_weight_updated_constant0" -> "1784 linear_89" [label="(384, 384)", style=solid]; +"1783 reshape_65_0_0_nncf_smooth_quant_0" -> "1784 linear_89" [label="(4, 64, 384)", style=solid]; +"1784 linear_89" -> "1785 dropout_57" [label="(4, 64, 384)", style=solid]; +"1785 dropout_57" -> "1786 view_80" [label="(4, 64, 384)", style=solid]; +"1786 view_80" -> "1787 permute_67" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1787 permute_67" -> "1788 reshape_66" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1788 reshape_66" -> "1789 slice_218" [label="(1, 16, 16, 384)", style=solid]; +"1789 slice_218" -> "1790 slice_219" [label="(1, 16, 16, 384)", style=solid]; +"1790 slice_219" -> "1791 slice_220" [label="(1, 14, 16, 384)", style=solid]; +"1791 slice_220" -> "1792 slice_221" [label="(1, 14, 14, 384)", style=solid]; +"1792 slice_221" -> "1793 contiguous_27" [label="(1, 14, 14, 384)", style=solid]; +"1793 contiguous_27" -> "1794 layer_norm_31" [label="(1, 14, 14, 384)", style=solid]; +"1794 layer_norm_31" -> "1795 add_50" [label="(1, 14, 14, 384)", style=solid]; +"1795 add_50" -> "1796 add_50_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"1795 add_50" -> "1816 add_51" [label="(1, 14, 14, 384)", style=solid]; +"1796 add_50_0_0_nncf_smooth_quant_0" -> "1797 quantize_per_tensor_default_61" [label="(1, 14, 14, 384)", style=solid]; +"1797 quantize_per_tensor_default_61" -> "1798 dequantize_per_tensor_default_61" [label="(1, 14, 14, 384)", style=solid]; +"1798 dequantize_per_tensor_default_61" -> "1803 linear_90" [label="(1, 14, 14, 384)", style=solid]; +"1799 linear_90_scale_0" -> "1802 dequantize_per_channel_default_61" [label="(1536,)", style=solid]; +"1800 linear_90_zero_point_0" -> "1802 dequantize_per_channel_default_61" [label="(1536,)", style=solid]; +"1801 _frozen_param76" -> "1802 dequantize_per_channel_default_61" [label="(1536, 384)", style=solid]; +"1802 dequantize_per_channel_default_61" -> "1803 linear_90" [label="(1536, 384)", style=solid]; +"1803 linear_90" -> "1804 gelu_14" [label="(1, 14, 14, 1536)", style=solid]; +"1804 gelu_14" -> "1805 dropout_58" [label="(1, 14, 14, 1536)", style=solid]; +"1805 dropout_58" -> "1806 dropout_58_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"1806 dropout_58_0_0_nncf_smooth_quant_0" -> "1807 quantize_per_tensor_default_62" [label="(1, 14, 14, 1536)", style=solid]; +"1807 quantize_per_tensor_default_62" -> "1808 dequantize_per_tensor_default_62" [label="(1, 14, 14, 1536)", style=solid]; +"1808 dequantize_per_tensor_default_62" -> "1813 linear_91" [label="(1, 14, 14, 1536)", style=solid]; +"1809 linear_91_scale_0" -> "1812 dequantize_per_channel_default_62" [label="(384,)", style=solid]; +"1810 linear_91_zero_point_0" -> "1812 dequantize_per_channel_default_62" [label="(384,)", style=solid]; +"1811 _frozen_param77" -> "1812 dequantize_per_channel_default_62" [label="(384, 1536)", style=solid]; +"1812 dequantize_per_channel_default_62" -> "1813 linear_91" [label="(384, 1536)", style=solid]; +"1813 linear_91" -> "1814 dropout_59" [label="(1, 14, 14, 384)", style=solid]; +"1814 dropout_59" -> "1815 layer_norm_32" [label="(1, 14, 14, 384)", style=solid]; +"1815 layer_norm_32" -> "1816 add_51" [label="(1, 14, 14, 384)", style=solid]; +"1816 add_51" -> "1841 pad_17" [label="(1, 14, 14, 384)", style=solid]; +"1816 add_51" -> "1904 add_54" [label="(1, 14, 14, 384)", style=solid]; +"1817 _frozen_param78" -> "1818 dequantize_per_tensor_default_63" [label="(1, 15, 15, 2)", style=solid]; +"1818 dequantize_per_tensor_default_63" -> "1823 linear_92" [label="(1, 15, 15, 2)", style=solid]; +"1819 linear_92_scale_0" -> "1822 dequantize_per_channel_default_63" [label="(512,)", style=solid]; +"1820 linear_92_zero_point_0" -> "1822 dequantize_per_channel_default_63" [label="(512,)", style=solid]; +"1821 _frozen_param79" -> "1822 dequantize_per_channel_default_63" [label="(512, 2)", style=solid]; +"1822 dequantize_per_channel_default_63" -> "1823 linear_92" [label="(512, 2)", style=solid]; +"1823 linear_92" -> "1824 relu__15" [label="(1, 15, 15, 512)", style=solid]; +"1824 relu__15" -> "1825 relu__15_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"1825 relu__15_0_0_nncf_smooth_quant_0" -> "1826 quantize_per_tensor_default_64" [label="(1, 15, 15, 512)", style=solid]; +"1826 quantize_per_tensor_default_64" -> "1827 dequantize_per_tensor_default_64" [label="(1, 15, 15, 512)", style=solid]; +"1827 dequantize_per_tensor_default_64" -> "1832 linear_93" [label="(1, 15, 15, 512)", style=solid]; +"1828 linear_93_scale_0" -> "1831 dequantize_per_channel_default_64" [label="(12,)", style=solid]; +"1829 linear_93_zero_point_0" -> "1831 dequantize_per_channel_default_64" [label="(12,)", style=solid]; +"1830 _frozen_param80" -> "1831 dequantize_per_channel_default_64" [label="(12, 512)", style=solid]; +"1831 dequantize_per_channel_default_64" -> "1832 linear_93" [label="(12, 512)", style=solid]; +"1832 linear_93" -> "1833 view_81" [label="(1, 15, 15, 12)", style=solid]; +"1833 view_81" -> "1834 index_15" [label="(225, 12)", style=solid]; +"1834 index_15" -> "1835 view_82" [label="(4096, 12)", style=solid]; +"1835 view_82" -> "1836 permute_68" [label="(64, 64, 12)", style=solid]; +"1836 permute_68" -> "1837 contiguous_28" [label="(12, 64, 64)", style=solid]; +"1837 contiguous_28" -> "1838 unsqueeze_43" [label="(12, 64, 64)", style=solid]; +"1838 unsqueeze_43" -> "1839 sigmoid_15" [label="(1, 12, 64, 64)", style=solid]; +"1839 sigmoid_15" -> "1840 mul_30" [label="(1, 12, 64, 64)", style=solid]; +"1840 mul_30" -> "1868 add_52" [label="(1, 12, 64, 64)", style=solid]; +"1841 pad_17" -> "1842 roll_14" [label="(1, 16, 16, 384)", style=solid]; +"1842 roll_14" -> "1843 view_83" [label="(1, 16, 16, 384)", style=solid]; +"1843 view_83" -> "1844 permute_69" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1844 permute_69" -> "1845 reshape_67" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1845 reshape_67" -> "1848 reshape_67_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1845 reshape_67" -> "1869 new_zeros_7" [label="(4, 64, 384)", style=solid]; +"1846 clone_15" -> "1849 linear_94" [label="(1152,)", style=solid]; +"1847 features_5_11_attn_qkv_weight_updated_constant0" -> "1849 linear_94" [label="(1152, 384)", style=solid]; +"1848 reshape_67_0_0_nncf_smooth_quant_0" -> "1849 linear_94" [label="(4, 64, 384)", style=solid]; +"1849 linear_94" -> "1850 reshape_68" [label="(4, 64, 1152)", style=solid]; +"1850 reshape_68" -> "1851 permute_70" [label="(4, 64, 3, 12, 32)", style=solid]; +"1851 permute_70" -> "1852 select_45" [label="(3, 4, 12, 64, 32)", style=solid]; +"1851 permute_70" -> "1853 select_46" [label="(3, 4, 12, 64, 32)", style=solid]; +"1851 permute_70" -> "1854 select_47" [label="(3, 4, 12, 64, 32)", style=solid]; +"1852 select_45" -> "1855 linalg_vector_norm_30" [label="(4, 12, 64, 32)", style=solid]; +"1852 select_45" -> "1857 expand_as_30" [label="(4, 12, 64, 32)", style=solid]; +"1852 select_45" -> "1858 div_30" [label="(4, 12, 64, 32)", style=solid]; +"1853 select_46" -> "1859 linalg_vector_norm_31" [label="(4, 12, 64, 32)", style=solid]; +"1853 select_46" -> "1861 expand_as_31" [label="(4, 12, 64, 32)", style=solid]; +"1853 select_46" -> "1862 div_31" [label="(4, 12, 64, 32)", style=solid]; +"1854 select_47" -> "1887 matmul_31" [label="(4, 12, 64, 32)", style=solid]; +"1855 linalg_vector_norm_30" -> "1856 clamp_min_30" [label="(4, 12, 64, 1)", style=solid]; +"1856 clamp_min_30" -> "1857 expand_as_30" [label="(4, 12, 64, 1)", style=solid]; +"1857 expand_as_30" -> "1858 div_30" [label="(4, 12, 64, 32)", style=solid]; +"1858 div_30" -> "1864 matmul_30" [label="(4, 12, 64, 32)", style=solid]; +"1859 linalg_vector_norm_31" -> "1860 clamp_min_31" [label="(4, 12, 64, 1)", style=solid]; +"1860 clamp_min_31" -> "1861 expand_as_31" [label="(4, 12, 64, 1)", style=solid]; +"1861 expand_as_31" -> "1862 div_31" [label="(4, 12, 64, 32)", style=solid]; +"1862 div_31" -> "1863 transpose_30" [label="(4, 12, 64, 32)", style=solid]; +"1863 transpose_30" -> "1864 matmul_30" [label="(4, 12, 32, 64)", style=solid]; +"1864 matmul_30" -> "1867 mul_31" [label="(4, 12, 64, 64)", style=solid]; +"1865 clamp_15" -> "1866 exp_15" [label="(12, 1, 1)", style=solid]; +"1866 exp_15" -> "1867 mul_31" [label="(12, 1, 1)", style=solid]; +"1867 mul_31" -> "1868 add_52" [label="(4, 12, 64, 64)", style=solid]; +"1868 add_52" -> "1880 view_85" [label="(4, 12, 64, 64)", style=solid]; +"1869 new_zeros_7" -> "1870 view_84" [label="(16, 16)", style=solid]; +"1870 view_84" -> "1871 permute_71" [label="(2, 8, 2, 8)", style=solid]; +"1871 permute_71" -> "1872 reshape_69" [label="(2, 2, 8, 8)", style=solid]; +"1872 reshape_69" -> "1873 unsqueeze_44" [label="(4, 64)", style=solid]; +"1872 reshape_69" -> "1874 unsqueeze_45" [label="(4, 64)", style=solid]; +"1873 unsqueeze_44" -> "1875 sub_7" [label="(4, 1, 64)", style=solid]; +"1874 unsqueeze_45" -> "1875 sub_7" [label="(4, 64, 1)", style=solid]; +"1875 sub_7" -> "1876 ne_7" [label="(4, 64, 64)", style=solid]; +"1875 sub_7" -> "1877 masked_fill_14" [label="(4, 64, 64)", style=solid]; +"1875 sub_7" -> "1878 eq_7" [label="(4, 64, 64)", style=solid]; +"1876 ne_7" -> "1877 masked_fill_14" [label="(4, 64, 64)", style=solid]; +"1877 masked_fill_14" -> "1879 masked_fill_15" [label="(4, 64, 64)", style=solid]; +"1878 eq_7" -> "1879 masked_fill_15" [label="(4, 64, 64)", style=solid]; +"1879 masked_fill_15" -> "1881 unsqueeze_46" [label="(4, 64, 64)", style=solid]; +"1880 view_85" -> "1883 add_53" [label="(1, 4, 12, 64, 64)", style=solid]; +"1881 unsqueeze_46" -> "1882 unsqueeze_47" [label="(4, 1, 64, 64)", style=solid]; +"1882 unsqueeze_47" -> "1883 add_53" [label="(1, 4, 1, 64, 64)", style=solid]; +"1883 add_53" -> "1884 view_86" [label="(1, 4, 12, 64, 64)", style=solid]; +"1884 view_86" -> "1885 softmax_15" [label="(4, 12, 64, 64)", style=solid]; +"1885 softmax_15" -> "1886 dropout_60" [label="(4, 12, 64, 64)", style=solid]; +"1886 dropout_60" -> "1887 matmul_31" [label="(4, 12, 64, 64)", style=solid]; +"1887 matmul_31" -> "1888 transpose_31" [label="(4, 12, 64, 32)", style=solid]; +"1888 transpose_31" -> "1889 reshape_70" [label="(4, 64, 12, 32)", style=solid]; +"1889 reshape_70" -> "1891 reshape_70_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1890 features_5_11_attn_proj_weight_updated_constant0" -> "1892 linear_95" [label="(384, 384)", style=solid]; +"1891 reshape_70_0_0_nncf_smooth_quant_0" -> "1892 linear_95" [label="(4, 64, 384)", style=solid]; +"1892 linear_95" -> "1893 dropout_61" [label="(4, 64, 384)", style=solid]; +"1893 dropout_61" -> "1894 view_87" [label="(4, 64, 384)", style=solid]; +"1894 view_87" -> "1895 permute_72" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1895 permute_72" -> "1896 reshape_71" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1896 reshape_71" -> "1897 roll_15" [label="(1, 16, 16, 384)", style=solid]; +"1897 roll_15" -> "1898 slice_241" [label="(1, 16, 16, 384)", style=solid]; +"1898 slice_241" -> "1899 slice_242" [label="(1, 16, 16, 384)", style=solid]; +"1899 slice_242" -> "1900 slice_243" [label="(1, 14, 16, 384)", style=solid]; +"1900 slice_243" -> "1901 slice_244" [label="(1, 14, 14, 384)", style=solid]; +"1901 slice_244" -> "1902 contiguous_29" [label="(1, 14, 14, 384)", style=solid]; +"1902 contiguous_29" -> "1903 layer_norm_33" [label="(1, 14, 14, 384)", style=solid]; +"1903 layer_norm_33" -> "1904 add_54" [label="(1, 14, 14, 384)", style=solid]; +"1904 add_54" -> "1905 add_54_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"1904 add_54" -> "1925 add_55" [label="(1, 14, 14, 384)", style=solid]; +"1905 add_54_0_0_nncf_smooth_quant_0" -> "1906 quantize_per_tensor_default_65" [label="(1, 14, 14, 384)", style=solid]; +"1906 quantize_per_tensor_default_65" -> "1907 dequantize_per_tensor_default_65" [label="(1, 14, 14, 384)", style=solid]; +"1907 dequantize_per_tensor_default_65" -> "1912 linear_96" [label="(1, 14, 14, 384)", style=solid]; +"1908 linear_96_scale_0" -> "1911 dequantize_per_channel_default_65" [label="(1536,)", style=solid]; +"1909 linear_96_zero_point_0" -> "1911 dequantize_per_channel_default_65" [label="(1536,)", style=solid]; +"1910 _frozen_param81" -> "1911 dequantize_per_channel_default_65" [label="(1536, 384)", style=solid]; +"1911 dequantize_per_channel_default_65" -> "1912 linear_96" [label="(1536, 384)", style=solid]; +"1912 linear_96" -> "1913 gelu_15" [label="(1, 14, 14, 1536)", style=solid]; +"1913 gelu_15" -> "1914 dropout_62" [label="(1, 14, 14, 1536)", style=solid]; +"1914 dropout_62" -> "1915 dropout_62_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"1915 dropout_62_0_0_nncf_smooth_quant_0" -> "1916 quantize_per_tensor_default_66" [label="(1, 14, 14, 1536)", style=solid]; +"1916 quantize_per_tensor_default_66" -> "1917 dequantize_per_tensor_default_66" [label="(1, 14, 14, 1536)", style=solid]; +"1917 dequantize_per_tensor_default_66" -> "1922 linear_97" [label="(1, 14, 14, 1536)", style=solid]; +"1918 linear_97_scale_0" -> "1921 dequantize_per_channel_default_66" [label="(384,)", style=solid]; +"1919 linear_97_zero_point_0" -> "1921 dequantize_per_channel_default_66" [label="(384,)", style=solid]; +"1920 _frozen_param82" -> "1921 dequantize_per_channel_default_66" [label="(384, 1536)", style=solid]; +"1921 dequantize_per_channel_default_66" -> "1922 linear_97" [label="(384, 1536)", style=solid]; +"1922 linear_97" -> "1923 dropout_63" [label="(1, 14, 14, 384)", style=solid]; +"1923 dropout_63" -> "1924 layer_norm_34" [label="(1, 14, 14, 384)", style=solid]; +"1924 layer_norm_34" -> "1925 add_55" [label="(1, 14, 14, 384)", style=solid]; +"1925 add_55" -> "1950 pad_18" [label="(1, 14, 14, 384)", style=solid]; +"1925 add_55" -> "1995 add_57" [label="(1, 14, 14, 384)", style=solid]; +"1926 _frozen_param83" -> "1927 dequantize_per_tensor_default_67" [label="(1, 15, 15, 2)", style=solid]; +"1927 dequantize_per_tensor_default_67" -> "1932 linear_98" [label="(1, 15, 15, 2)", style=solid]; +"1928 linear_98_scale_0" -> "1931 dequantize_per_channel_default_67" [label="(512,)", style=solid]; +"1929 linear_98_zero_point_0" -> "1931 dequantize_per_channel_default_67" [label="(512,)", style=solid]; +"1930 _frozen_param84" -> "1931 dequantize_per_channel_default_67" [label="(512, 2)", style=solid]; +"1931 dequantize_per_channel_default_67" -> "1932 linear_98" [label="(512, 2)", style=solid]; +"1932 linear_98" -> "1933 relu__16" [label="(1, 15, 15, 512)", style=solid]; +"1933 relu__16" -> "1934 relu__16_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"1934 relu__16_0_0_nncf_smooth_quant_0" -> "1935 quantize_per_tensor_default_68" [label="(1, 15, 15, 512)", style=solid]; +"1935 quantize_per_tensor_default_68" -> "1936 dequantize_per_tensor_default_68" [label="(1, 15, 15, 512)", style=solid]; +"1936 dequantize_per_tensor_default_68" -> "1941 linear_99" [label="(1, 15, 15, 512)", style=solid]; +"1937 linear_99_scale_0" -> "1940 dequantize_per_channel_default_68" [label="(12,)", style=solid]; +"1938 linear_99_zero_point_0" -> "1940 dequantize_per_channel_default_68" [label="(12,)", style=solid]; +"1939 _frozen_param85" -> "1940 dequantize_per_channel_default_68" [label="(12, 512)", style=solid]; +"1940 dequantize_per_channel_default_68" -> "1941 linear_99" [label="(12, 512)", style=solid]; +"1941 linear_99" -> "1942 view_88" [label="(1, 15, 15, 12)", style=solid]; +"1942 view_88" -> "1943 index_16" [label="(225, 12)", style=solid]; +"1943 index_16" -> "1944 view_89" [label="(4096, 12)", style=solid]; +"1944 view_89" -> "1945 permute_73" [label="(64, 64, 12)", style=solid]; +"1945 permute_73" -> "1946 contiguous_30" [label="(12, 64, 64)", style=solid]; +"1946 contiguous_30" -> "1947 unsqueeze_48" [label="(12, 64, 64)", style=solid]; +"1947 unsqueeze_48" -> "1948 sigmoid_16" [label="(1, 12, 64, 64)", style=solid]; +"1948 sigmoid_16" -> "1949 mul_32" [label="(1, 12, 64, 64)", style=solid]; +"1949 mul_32" -> "1976 add_56" [label="(1, 12, 64, 64)", style=solid]; +"1950 pad_18" -> "1951 view_90" [label="(1, 16, 16, 384)", style=solid]; +"1951 view_90" -> "1952 permute_74" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1952 permute_74" -> "1953 reshape_72" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1953 reshape_72" -> "1956 reshape_72_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1954 clone_16" -> "1957 linear_100" [label="(1152,)", style=solid]; +"1955 features_5_12_attn_qkv_weight_updated_constant0" -> "1957 linear_100" [label="(1152, 384)", style=solid]; +"1956 reshape_72_0_0_nncf_smooth_quant_0" -> "1957 linear_100" [label="(4, 64, 384)", style=solid]; +"1957 linear_100" -> "1958 reshape_73" [label="(4, 64, 1152)", style=solid]; +"1958 reshape_73" -> "1959 permute_75" [label="(4, 64, 3, 12, 32)", style=solid]; +"1959 permute_75" -> "1960 select_48" [label="(3, 4, 12, 64, 32)", style=solid]; +"1959 permute_75" -> "1961 select_49" [label="(3, 4, 12, 64, 32)", style=solid]; +"1959 permute_75" -> "1962 select_50" [label="(3, 4, 12, 64, 32)", style=solid]; +"1960 select_48" -> "1963 linalg_vector_norm_32" [label="(4, 12, 64, 32)", style=solid]; +"1960 select_48" -> "1965 expand_as_32" [label="(4, 12, 64, 32)", style=solid]; +"1960 select_48" -> "1966 div_32" [label="(4, 12, 64, 32)", style=solid]; +"1961 select_49" -> "1967 linalg_vector_norm_33" [label="(4, 12, 64, 32)", style=solid]; +"1961 select_49" -> "1969 expand_as_33" [label="(4, 12, 64, 32)", style=solid]; +"1961 select_49" -> "1970 div_33" [label="(4, 12, 64, 32)", style=solid]; +"1962 select_50" -> "1979 matmul_33" [label="(4, 12, 64, 32)", style=solid]; +"1963 linalg_vector_norm_32" -> "1964 clamp_min_32" [label="(4, 12, 64, 1)", style=solid]; +"1964 clamp_min_32" -> "1965 expand_as_32" [label="(4, 12, 64, 1)", style=solid]; +"1965 expand_as_32" -> "1966 div_32" [label="(4, 12, 64, 32)", style=solid]; +"1966 div_32" -> "1972 matmul_32" [label="(4, 12, 64, 32)", style=solid]; +"1967 linalg_vector_norm_33" -> "1968 clamp_min_33" [label="(4, 12, 64, 1)", style=solid]; +"1968 clamp_min_33" -> "1969 expand_as_33" [label="(4, 12, 64, 1)", style=solid]; +"1969 expand_as_33" -> "1970 div_33" [label="(4, 12, 64, 32)", style=solid]; +"1970 div_33" -> "1971 transpose_32" [label="(4, 12, 64, 32)", style=solid]; +"1971 transpose_32" -> "1972 matmul_32" [label="(4, 12, 32, 64)", style=solid]; +"1972 matmul_32" -> "1975 mul_33" [label="(4, 12, 64, 64)", style=solid]; +"1973 clamp_16" -> "1974 exp_16" [label="(12, 1, 1)", style=solid]; +"1974 exp_16" -> "1975 mul_33" [label="(12, 1, 1)", style=solid]; +"1975 mul_33" -> "1976 add_56" [label="(4, 12, 64, 64)", style=solid]; +"1976 add_56" -> "1977 softmax_16" [label="(4, 12, 64, 64)", style=solid]; +"1977 softmax_16" -> "1978 dropout_64" [label="(4, 12, 64, 64)", style=solid]; +"1978 dropout_64" -> "1979 matmul_33" [label="(4, 12, 64, 64)", style=solid]; +"1979 matmul_33" -> "1980 transpose_33" [label="(4, 12, 64, 32)", style=solid]; +"1980 transpose_33" -> "1981 reshape_74" [label="(4, 64, 12, 32)", style=solid]; +"1981 reshape_74" -> "1983 reshape_74_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1982 features_5_12_attn_proj_weight_updated_constant0" -> "1984 linear_101" [label="(384, 384)", style=solid]; +"1983 reshape_74_0_0_nncf_smooth_quant_0" -> "1984 linear_101" [label="(4, 64, 384)", style=solid]; +"1984 linear_101" -> "1985 dropout_65" [label="(4, 64, 384)", style=solid]; +"1985 dropout_65" -> "1986 view_91" [label="(4, 64, 384)", style=solid]; +"1986 view_91" -> "1987 permute_76" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1987 permute_76" -> "1988 reshape_75" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1988 reshape_75" -> "1989 slice_246" [label="(1, 16, 16, 384)", style=solid]; +"1989 slice_246" -> "1990 slice_247" [label="(1, 16, 16, 384)", style=solid]; +"1990 slice_247" -> "1991 slice_248" [label="(1, 14, 16, 384)", style=solid]; +"1991 slice_248" -> "1992 slice_249" [label="(1, 14, 14, 384)", style=solid]; +"1992 slice_249" -> "1993 contiguous_31" [label="(1, 14, 14, 384)", style=solid]; +"1993 contiguous_31" -> "1994 layer_norm_35" [label="(1, 14, 14, 384)", style=solid]; +"1994 layer_norm_35" -> "1995 add_57" [label="(1, 14, 14, 384)", style=solid]; +"1995 add_57" -> "1996 add_57_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"1995 add_57" -> "2016 add_58" [label="(1, 14, 14, 384)", style=solid]; +"1996 add_57_0_0_nncf_smooth_quant_0" -> "1997 quantize_per_tensor_default_69" [label="(1, 14, 14, 384)", style=solid]; +"1997 quantize_per_tensor_default_69" -> "1998 dequantize_per_tensor_default_69" [label="(1, 14, 14, 384)", style=solid]; +"1998 dequantize_per_tensor_default_69" -> "2003 linear_102" [label="(1, 14, 14, 384)", style=solid]; +"1999 linear_102_scale_0" -> "2002 dequantize_per_channel_default_69" [label="(1536,)", style=solid]; +"2000 linear_102_zero_point_0" -> "2002 dequantize_per_channel_default_69" [label="(1536,)", style=solid]; +"2001 _frozen_param86" -> "2002 dequantize_per_channel_default_69" [label="(1536, 384)", style=solid]; +"2002 dequantize_per_channel_default_69" -> "2003 linear_102" [label="(1536, 384)", style=solid]; +"2003 linear_102" -> "2004 gelu_16" [label="(1, 14, 14, 1536)", style=solid]; +"2004 gelu_16" -> "2005 dropout_66" [label="(1, 14, 14, 1536)", style=solid]; +"2005 dropout_66" -> "2006 dropout_66_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"2006 dropout_66_0_0_nncf_smooth_quant_0" -> "2007 quantize_per_tensor_default_70" [label="(1, 14, 14, 1536)", style=solid]; +"2007 quantize_per_tensor_default_70" -> "2008 dequantize_per_tensor_default_70" [label="(1, 14, 14, 1536)", style=solid]; +"2008 dequantize_per_tensor_default_70" -> "2013 linear_103" [label="(1, 14, 14, 1536)", style=solid]; +"2009 linear_103_scale_0" -> "2012 dequantize_per_channel_default_70" [label="(384,)", style=solid]; +"2010 linear_103_zero_point_0" -> "2012 dequantize_per_channel_default_70" [label="(384,)", style=solid]; +"2011 _frozen_param87" -> "2012 dequantize_per_channel_default_70" [label="(384, 1536)", style=solid]; +"2012 dequantize_per_channel_default_70" -> "2013 linear_103" [label="(384, 1536)", style=solid]; +"2013 linear_103" -> "2014 dropout_67" [label="(1, 14, 14, 384)", style=solid]; +"2014 dropout_67" -> "2015 layer_norm_36" [label="(1, 14, 14, 384)", style=solid]; +"2015 layer_norm_36" -> "2016 add_58" [label="(1, 14, 14, 384)", style=solid]; +"2016 add_58" -> "2041 pad_19" [label="(1, 14, 14, 384)", style=solid]; +"2016 add_58" -> "2104 add_61" [label="(1, 14, 14, 384)", style=solid]; +"2017 _frozen_param88" -> "2018 dequantize_per_tensor_default_71" [label="(1, 15, 15, 2)", style=solid]; +"2018 dequantize_per_tensor_default_71" -> "2023 linear_104" [label="(1, 15, 15, 2)", style=solid]; +"2019 linear_104_scale_0" -> "2022 dequantize_per_channel_default_71" [label="(512,)", style=solid]; +"2020 linear_104_zero_point_0" -> "2022 dequantize_per_channel_default_71" [label="(512,)", style=solid]; +"2021 _frozen_param89" -> "2022 dequantize_per_channel_default_71" [label="(512, 2)", style=solid]; +"2022 dequantize_per_channel_default_71" -> "2023 linear_104" [label="(512, 2)", style=solid]; +"2023 linear_104" -> "2024 relu__17" [label="(1, 15, 15, 512)", style=solid]; +"2024 relu__17" -> "2025 relu__17_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"2025 relu__17_0_0_nncf_smooth_quant_0" -> "2026 quantize_per_tensor_default_72" [label="(1, 15, 15, 512)", style=solid]; +"2026 quantize_per_tensor_default_72" -> "2027 dequantize_per_tensor_default_72" [label="(1, 15, 15, 512)", style=solid]; +"2027 dequantize_per_tensor_default_72" -> "2032 linear_105" [label="(1, 15, 15, 512)", style=solid]; +"2028 linear_105_scale_0" -> "2031 dequantize_per_channel_default_72" [label="(12,)", style=solid]; +"2029 linear_105_zero_point_0" -> "2031 dequantize_per_channel_default_72" [label="(12,)", style=solid]; +"2030 _frozen_param90" -> "2031 dequantize_per_channel_default_72" [label="(12, 512)", style=solid]; +"2031 dequantize_per_channel_default_72" -> "2032 linear_105" [label="(12, 512)", style=solid]; +"2032 linear_105" -> "2033 view_92" [label="(1, 15, 15, 12)", style=solid]; +"2033 view_92" -> "2034 index_17" [label="(225, 12)", style=solid]; +"2034 index_17" -> "2035 view_93" [label="(4096, 12)", style=solid]; +"2035 view_93" -> "2036 permute_77" [label="(64, 64, 12)", style=solid]; +"2036 permute_77" -> "2037 contiguous_32" [label="(12, 64, 64)", style=solid]; +"2037 contiguous_32" -> "2038 unsqueeze_49" [label="(12, 64, 64)", style=solid]; +"2038 unsqueeze_49" -> "2039 sigmoid_17" [label="(1, 12, 64, 64)", style=solid]; +"2039 sigmoid_17" -> "2040 mul_34" [label="(1, 12, 64, 64)", style=solid]; +"2040 mul_34" -> "2068 add_59" [label="(1, 12, 64, 64)", style=solid]; +"2041 pad_19" -> "2042 roll_16" [label="(1, 16, 16, 384)", style=solid]; +"2042 roll_16" -> "2043 view_94" [label="(1, 16, 16, 384)", style=solid]; +"2043 view_94" -> "2044 permute_78" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2044 permute_78" -> "2045 reshape_76" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2045 reshape_76" -> "2048 reshape_76_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"2045 reshape_76" -> "2069 new_zeros_8" [label="(4, 64, 384)", style=solid]; +"2046 clone_17" -> "2049 linear_106" [label="(1152,)", style=solid]; +"2047 features_5_13_attn_qkv_weight_updated_constant0" -> "2049 linear_106" [label="(1152, 384)", style=solid]; +"2048 reshape_76_0_0_nncf_smooth_quant_0" -> "2049 linear_106" [label="(4, 64, 384)", style=solid]; +"2049 linear_106" -> "2050 reshape_77" [label="(4, 64, 1152)", style=solid]; +"2050 reshape_77" -> "2051 permute_79" [label="(4, 64, 3, 12, 32)", style=solid]; +"2051 permute_79" -> "2052 select_51" [label="(3, 4, 12, 64, 32)", style=solid]; +"2051 permute_79" -> "2053 select_52" [label="(3, 4, 12, 64, 32)", style=solid]; +"2051 permute_79" -> "2054 select_53" [label="(3, 4, 12, 64, 32)", style=solid]; +"2052 select_51" -> "2055 linalg_vector_norm_34" [label="(4, 12, 64, 32)", style=solid]; +"2052 select_51" -> "2057 expand_as_34" [label="(4, 12, 64, 32)", style=solid]; +"2052 select_51" -> "2058 div_34" [label="(4, 12, 64, 32)", style=solid]; +"2053 select_52" -> "2059 linalg_vector_norm_35" [label="(4, 12, 64, 32)", style=solid]; +"2053 select_52" -> "2061 expand_as_35" [label="(4, 12, 64, 32)", style=solid]; +"2053 select_52" -> "2062 div_35" [label="(4, 12, 64, 32)", style=solid]; +"2054 select_53" -> "2087 matmul_35" [label="(4, 12, 64, 32)", style=solid]; +"2055 linalg_vector_norm_34" -> "2056 clamp_min_34" [label="(4, 12, 64, 1)", style=solid]; +"2056 clamp_min_34" -> "2057 expand_as_34" [label="(4, 12, 64, 1)", style=solid]; +"2057 expand_as_34" -> "2058 div_34" [label="(4, 12, 64, 32)", style=solid]; +"2058 div_34" -> "2064 matmul_34" [label="(4, 12, 64, 32)", style=solid]; +"2059 linalg_vector_norm_35" -> "2060 clamp_min_35" [label="(4, 12, 64, 1)", style=solid]; +"2060 clamp_min_35" -> "2061 expand_as_35" [label="(4, 12, 64, 1)", style=solid]; +"2061 expand_as_35" -> "2062 div_35" [label="(4, 12, 64, 32)", style=solid]; +"2062 div_35" -> "2063 transpose_34" [label="(4, 12, 64, 32)", style=solid]; +"2063 transpose_34" -> "2064 matmul_34" [label="(4, 12, 32, 64)", style=solid]; +"2064 matmul_34" -> "2067 mul_35" [label="(4, 12, 64, 64)", style=solid]; +"2065 clamp_17" -> "2066 exp_17" [label="(12, 1, 1)", style=solid]; +"2066 exp_17" -> "2067 mul_35" [label="(12, 1, 1)", style=solid]; +"2067 mul_35" -> "2068 add_59" [label="(4, 12, 64, 64)", style=solid]; +"2068 add_59" -> "2080 view_96" [label="(4, 12, 64, 64)", style=solid]; +"2069 new_zeros_8" -> "2070 view_95" [label="(16, 16)", style=solid]; +"2070 view_95" -> "2071 permute_80" [label="(2, 8, 2, 8)", style=solid]; +"2071 permute_80" -> "2072 reshape_78" [label="(2, 2, 8, 8)", style=solid]; +"2072 reshape_78" -> "2073 unsqueeze_50" [label="(4, 64)", style=solid]; +"2072 reshape_78" -> "2074 unsqueeze_51" [label="(4, 64)", style=solid]; +"2073 unsqueeze_50" -> "2075 sub_8" [label="(4, 1, 64)", style=solid]; +"2074 unsqueeze_51" -> "2075 sub_8" [label="(4, 64, 1)", style=solid]; +"2075 sub_8" -> "2076 ne_8" [label="(4, 64, 64)", style=solid]; +"2075 sub_8" -> "2077 masked_fill_16" [label="(4, 64, 64)", style=solid]; +"2075 sub_8" -> "2078 eq_8" [label="(4, 64, 64)", style=solid]; +"2076 ne_8" -> "2077 masked_fill_16" [label="(4, 64, 64)", style=solid]; +"2077 masked_fill_16" -> "2079 masked_fill_17" [label="(4, 64, 64)", style=solid]; +"2078 eq_8" -> "2079 masked_fill_17" [label="(4, 64, 64)", style=solid]; +"2079 masked_fill_17" -> "2081 unsqueeze_52" [label="(4, 64, 64)", style=solid]; +"2080 view_96" -> "2083 add_60" [label="(1, 4, 12, 64, 64)", style=solid]; +"2081 unsqueeze_52" -> "2082 unsqueeze_53" [label="(4, 1, 64, 64)", style=solid]; +"2082 unsqueeze_53" -> "2083 add_60" [label="(1, 4, 1, 64, 64)", style=solid]; +"2083 add_60" -> "2084 view_97" [label="(1, 4, 12, 64, 64)", style=solid]; +"2084 view_97" -> "2085 softmax_17" [label="(4, 12, 64, 64)", style=solid]; +"2085 softmax_17" -> "2086 dropout_68" [label="(4, 12, 64, 64)", style=solid]; +"2086 dropout_68" -> "2087 matmul_35" [label="(4, 12, 64, 64)", style=solid]; +"2087 matmul_35" -> "2088 transpose_35" [label="(4, 12, 64, 32)", style=solid]; +"2088 transpose_35" -> "2089 reshape_79" [label="(4, 64, 12, 32)", style=solid]; +"2089 reshape_79" -> "2091 reshape_79_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"2090 features_5_13_attn_proj_weight_updated_constant0" -> "2092 linear_107" [label="(384, 384)", style=solid]; +"2091 reshape_79_0_0_nncf_smooth_quant_0" -> "2092 linear_107" [label="(4, 64, 384)", style=solid]; +"2092 linear_107" -> "2093 dropout_69" [label="(4, 64, 384)", style=solid]; +"2093 dropout_69" -> "2094 view_98" [label="(4, 64, 384)", style=solid]; +"2094 view_98" -> "2095 permute_81" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2095 permute_81" -> "2096 reshape_80" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2096 reshape_80" -> "2097 roll_17" [label="(1, 16, 16, 384)", style=solid]; +"2097 roll_17" -> "2098 slice_269" [label="(1, 16, 16, 384)", style=solid]; +"2098 slice_269" -> "2099 slice_270" [label="(1, 16, 16, 384)", style=solid]; +"2099 slice_270" -> "2100 slice_271" [label="(1, 14, 16, 384)", style=solid]; +"2100 slice_271" -> "2101 slice_272" [label="(1, 14, 14, 384)", style=solid]; +"2101 slice_272" -> "2102 contiguous_33" [label="(1, 14, 14, 384)", style=solid]; +"2102 contiguous_33" -> "2103 layer_norm_37" [label="(1, 14, 14, 384)", style=solid]; +"2103 layer_norm_37" -> "2104 add_61" [label="(1, 14, 14, 384)", style=solid]; +"2104 add_61" -> "2105 add_61_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"2104 add_61" -> "2125 add_62" [label="(1, 14, 14, 384)", style=solid]; +"2105 add_61_0_0_nncf_smooth_quant_0" -> "2106 quantize_per_tensor_default_73" [label="(1, 14, 14, 384)", style=solid]; +"2106 quantize_per_tensor_default_73" -> "2107 dequantize_per_tensor_default_73" [label="(1, 14, 14, 384)", style=solid]; +"2107 dequantize_per_tensor_default_73" -> "2112 linear_108" [label="(1, 14, 14, 384)", style=solid]; +"2108 linear_108_scale_0" -> "2111 dequantize_per_channel_default_73" [label="(1536,)", style=solid]; +"2109 linear_108_zero_point_0" -> "2111 dequantize_per_channel_default_73" [label="(1536,)", style=solid]; +"2110 _frozen_param91" -> "2111 dequantize_per_channel_default_73" [label="(1536, 384)", style=solid]; +"2111 dequantize_per_channel_default_73" -> "2112 linear_108" [label="(1536, 384)", style=solid]; +"2112 linear_108" -> "2113 gelu_17" [label="(1, 14, 14, 1536)", style=solid]; +"2113 gelu_17" -> "2114 dropout_70" [label="(1, 14, 14, 1536)", style=solid]; +"2114 dropout_70" -> "2115 dropout_70_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"2115 dropout_70_0_0_nncf_smooth_quant_0" -> "2116 quantize_per_tensor_default_74" [label="(1, 14, 14, 1536)", style=solid]; +"2116 quantize_per_tensor_default_74" -> "2117 dequantize_per_tensor_default_74" [label="(1, 14, 14, 1536)", style=solid]; +"2117 dequantize_per_tensor_default_74" -> "2122 linear_109" [label="(1, 14, 14, 1536)", style=solid]; +"2118 linear_109_scale_0" -> "2121 dequantize_per_channel_default_74" [label="(384,)", style=solid]; +"2119 linear_109_zero_point_0" -> "2121 dequantize_per_channel_default_74" [label="(384,)", style=solid]; +"2120 _frozen_param92" -> "2121 dequantize_per_channel_default_74" [label="(384, 1536)", style=solid]; +"2121 dequantize_per_channel_default_74" -> "2122 linear_109" [label="(384, 1536)", style=solid]; +"2122 linear_109" -> "2123 dropout_71" [label="(1, 14, 14, 384)", style=solid]; +"2123 dropout_71" -> "2124 layer_norm_38" [label="(1, 14, 14, 384)", style=solid]; +"2124 layer_norm_38" -> "2125 add_62" [label="(1, 14, 14, 384)", style=solid]; +"2125 add_62" -> "2150 pad_20" [label="(1, 14, 14, 384)", style=solid]; +"2125 add_62" -> "2195 add_64" [label="(1, 14, 14, 384)", style=solid]; +"2126 _frozen_param93" -> "2127 dequantize_per_tensor_default_75" [label="(1, 15, 15, 2)", style=solid]; +"2127 dequantize_per_tensor_default_75" -> "2132 linear_110" [label="(1, 15, 15, 2)", style=solid]; +"2128 linear_110_scale_0" -> "2131 dequantize_per_channel_default_75" [label="(512,)", style=solid]; +"2129 linear_110_zero_point_0" -> "2131 dequantize_per_channel_default_75" [label="(512,)", style=solid]; +"2130 _frozen_param94" -> "2131 dequantize_per_channel_default_75" [label="(512, 2)", style=solid]; +"2131 dequantize_per_channel_default_75" -> "2132 linear_110" [label="(512, 2)", style=solid]; +"2132 linear_110" -> "2133 relu__18" [label="(1, 15, 15, 512)", style=solid]; +"2133 relu__18" -> "2134 relu__18_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"2134 relu__18_0_0_nncf_smooth_quant_0" -> "2135 quantize_per_tensor_default_76" [label="(1, 15, 15, 512)", style=solid]; +"2135 quantize_per_tensor_default_76" -> "2136 dequantize_per_tensor_default_76" [label="(1, 15, 15, 512)", style=solid]; +"2136 dequantize_per_tensor_default_76" -> "2141 linear_111" [label="(1, 15, 15, 512)", style=solid]; +"2137 linear_111_scale_0" -> "2140 dequantize_per_channel_default_76" [label="(12,)", style=solid]; +"2138 linear_111_zero_point_0" -> "2140 dequantize_per_channel_default_76" [label="(12,)", style=solid]; +"2139 _frozen_param95" -> "2140 dequantize_per_channel_default_76" [label="(12, 512)", style=solid]; +"2140 dequantize_per_channel_default_76" -> "2141 linear_111" [label="(12, 512)", style=solid]; +"2141 linear_111" -> "2142 view_99" [label="(1, 15, 15, 12)", style=solid]; +"2142 view_99" -> "2143 index_18" [label="(225, 12)", style=solid]; +"2143 index_18" -> "2144 view_100" [label="(4096, 12)", style=solid]; +"2144 view_100" -> "2145 permute_82" [label="(64, 64, 12)", style=solid]; +"2145 permute_82" -> "2146 contiguous_34" [label="(12, 64, 64)", style=solid]; +"2146 contiguous_34" -> "2147 unsqueeze_54" [label="(12, 64, 64)", style=solid]; +"2147 unsqueeze_54" -> "2148 sigmoid_18" [label="(1, 12, 64, 64)", style=solid]; +"2148 sigmoid_18" -> "2149 mul_36" [label="(1, 12, 64, 64)", style=solid]; +"2149 mul_36" -> "2176 add_63" [label="(1, 12, 64, 64)", style=solid]; +"2150 pad_20" -> "2151 view_101" [label="(1, 16, 16, 384)", style=solid]; +"2151 view_101" -> "2152 permute_83" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2152 permute_83" -> "2153 reshape_81" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2153 reshape_81" -> "2156 reshape_81_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"2154 clone_18" -> "2157 linear_112" [label="(1152,)", style=solid]; +"2155 features_5_14_attn_qkv_weight_updated_constant0" -> "2157 linear_112" [label="(1152, 384)", style=solid]; +"2156 reshape_81_0_0_nncf_smooth_quant_0" -> "2157 linear_112" [label="(4, 64, 384)", style=solid]; +"2157 linear_112" -> "2158 reshape_82" [label="(4, 64, 1152)", style=solid]; +"2158 reshape_82" -> "2159 permute_84" [label="(4, 64, 3, 12, 32)", style=solid]; +"2159 permute_84" -> "2160 select_54" [label="(3, 4, 12, 64, 32)", style=solid]; +"2159 permute_84" -> "2161 select_55" [label="(3, 4, 12, 64, 32)", style=solid]; +"2159 permute_84" -> "2162 select_56" [label="(3, 4, 12, 64, 32)", style=solid]; +"2160 select_54" -> "2163 linalg_vector_norm_36" [label="(4, 12, 64, 32)", style=solid]; +"2160 select_54" -> "2165 expand_as_36" [label="(4, 12, 64, 32)", style=solid]; +"2160 select_54" -> "2166 div_36" [label="(4, 12, 64, 32)", style=solid]; +"2161 select_55" -> "2167 linalg_vector_norm_37" [label="(4, 12, 64, 32)", style=solid]; +"2161 select_55" -> "2169 expand_as_37" [label="(4, 12, 64, 32)", style=solid]; +"2161 select_55" -> "2170 div_37" [label="(4, 12, 64, 32)", style=solid]; +"2162 select_56" -> "2179 matmul_37" [label="(4, 12, 64, 32)", style=solid]; +"2163 linalg_vector_norm_36" -> "2164 clamp_min_36" [label="(4, 12, 64, 1)", style=solid]; +"2164 clamp_min_36" -> "2165 expand_as_36" [label="(4, 12, 64, 1)", style=solid]; +"2165 expand_as_36" -> "2166 div_36" [label="(4, 12, 64, 32)", style=solid]; +"2166 div_36" -> "2172 matmul_36" [label="(4, 12, 64, 32)", style=solid]; +"2167 linalg_vector_norm_37" -> "2168 clamp_min_37" [label="(4, 12, 64, 1)", style=solid]; +"2168 clamp_min_37" -> "2169 expand_as_37" [label="(4, 12, 64, 1)", style=solid]; +"2169 expand_as_37" -> "2170 div_37" [label="(4, 12, 64, 32)", style=solid]; +"2170 div_37" -> "2171 transpose_36" [label="(4, 12, 64, 32)", style=solid]; +"2171 transpose_36" -> "2172 matmul_36" [label="(4, 12, 32, 64)", style=solid]; +"2172 matmul_36" -> "2175 mul_37" [label="(4, 12, 64, 64)", style=solid]; +"2173 clamp_18" -> "2174 exp_18" [label="(12, 1, 1)", style=solid]; +"2174 exp_18" -> "2175 mul_37" [label="(12, 1, 1)", style=solid]; +"2175 mul_37" -> "2176 add_63" [label="(4, 12, 64, 64)", style=solid]; +"2176 add_63" -> "2177 softmax_18" [label="(4, 12, 64, 64)", style=solid]; +"2177 softmax_18" -> "2178 dropout_72" [label="(4, 12, 64, 64)", style=solid]; +"2178 dropout_72" -> "2179 matmul_37" [label="(4, 12, 64, 64)", style=solid]; +"2179 matmul_37" -> "2180 transpose_37" [label="(4, 12, 64, 32)", style=solid]; +"2180 transpose_37" -> "2181 reshape_83" [label="(4, 64, 12, 32)", style=solid]; +"2181 reshape_83" -> "2183 reshape_83_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"2182 features_5_14_attn_proj_weight_updated_constant0" -> "2184 linear_113" [label="(384, 384)", style=solid]; +"2183 reshape_83_0_0_nncf_smooth_quant_0" -> "2184 linear_113" [label="(4, 64, 384)", style=solid]; +"2184 linear_113" -> "2185 dropout_73" [label="(4, 64, 384)", style=solid]; +"2185 dropout_73" -> "2186 view_102" [label="(4, 64, 384)", style=solid]; +"2186 view_102" -> "2187 permute_85" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2187 permute_85" -> "2188 reshape_84" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2188 reshape_84" -> "2189 slice_274" [label="(1, 16, 16, 384)", style=solid]; +"2189 slice_274" -> "2190 slice_275" [label="(1, 16, 16, 384)", style=solid]; +"2190 slice_275" -> "2191 slice_276" [label="(1, 14, 16, 384)", style=solid]; +"2191 slice_276" -> "2192 slice_277" [label="(1, 14, 14, 384)", style=solid]; +"2192 slice_277" -> "2193 contiguous_35" [label="(1, 14, 14, 384)", style=solid]; +"2193 contiguous_35" -> "2194 layer_norm_39" [label="(1, 14, 14, 384)", style=solid]; +"2194 layer_norm_39" -> "2195 add_64" [label="(1, 14, 14, 384)", style=solid]; +"2195 add_64" -> "2196 add_64_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"2195 add_64" -> "2216 add_65" [label="(1, 14, 14, 384)", style=solid]; +"2196 add_64_0_0_nncf_smooth_quant_0" -> "2197 quantize_per_tensor_default_77" [label="(1, 14, 14, 384)", style=solid]; +"2197 quantize_per_tensor_default_77" -> "2198 dequantize_per_tensor_default_77" [label="(1, 14, 14, 384)", style=solid]; +"2198 dequantize_per_tensor_default_77" -> "2203 linear_114" [label="(1, 14, 14, 384)", style=solid]; +"2199 linear_114_scale_0" -> "2202 dequantize_per_channel_default_77" [label="(1536,)", style=solid]; +"2200 linear_114_zero_point_0" -> "2202 dequantize_per_channel_default_77" [label="(1536,)", style=solid]; +"2201 _frozen_param96" -> "2202 dequantize_per_channel_default_77" [label="(1536, 384)", style=solid]; +"2202 dequantize_per_channel_default_77" -> "2203 linear_114" [label="(1536, 384)", style=solid]; +"2203 linear_114" -> "2204 gelu_18" [label="(1, 14, 14, 1536)", style=solid]; +"2204 gelu_18" -> "2205 dropout_74" [label="(1, 14, 14, 1536)", style=solid]; +"2205 dropout_74" -> "2206 dropout_74_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"2206 dropout_74_0_0_nncf_smooth_quant_0" -> "2207 quantize_per_tensor_default_78" [label="(1, 14, 14, 1536)", style=solid]; +"2207 quantize_per_tensor_default_78" -> "2208 dequantize_per_tensor_default_78" [label="(1, 14, 14, 1536)", style=solid]; +"2208 dequantize_per_tensor_default_78" -> "2213 linear_115" [label="(1, 14, 14, 1536)", style=solid]; +"2209 linear_115_scale_0" -> "2212 dequantize_per_channel_default_78" [label="(384,)", style=solid]; +"2210 linear_115_zero_point_0" -> "2212 dequantize_per_channel_default_78" [label="(384,)", style=solid]; +"2211 _frozen_param97" -> "2212 dequantize_per_channel_default_78" [label="(384, 1536)", style=solid]; +"2212 dequantize_per_channel_default_78" -> "2213 linear_115" [label="(384, 1536)", style=solid]; +"2213 linear_115" -> "2214 dropout_75" [label="(1, 14, 14, 384)", style=solid]; +"2214 dropout_75" -> "2215 layer_norm_40" [label="(1, 14, 14, 384)", style=solid]; +"2215 layer_norm_40" -> "2216 add_65" [label="(1, 14, 14, 384)", style=solid]; +"2216 add_65" -> "2241 pad_21" [label="(1, 14, 14, 384)", style=solid]; +"2216 add_65" -> "2304 add_68" [label="(1, 14, 14, 384)", style=solid]; +"2217 _frozen_param98" -> "2218 dequantize_per_tensor_default_79" [label="(1, 15, 15, 2)", style=solid]; +"2218 dequantize_per_tensor_default_79" -> "2223 linear_116" [label="(1, 15, 15, 2)", style=solid]; +"2219 linear_116_scale_0" -> "2222 dequantize_per_channel_default_79" [label="(512,)", style=solid]; +"2220 linear_116_zero_point_0" -> "2222 dequantize_per_channel_default_79" [label="(512,)", style=solid]; +"2221 _frozen_param99" -> "2222 dequantize_per_channel_default_79" [label="(512, 2)", style=solid]; +"2222 dequantize_per_channel_default_79" -> "2223 linear_116" [label="(512, 2)", style=solid]; +"2223 linear_116" -> "2224 relu__19" [label="(1, 15, 15, 512)", style=solid]; +"2224 relu__19" -> "2225 relu__19_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"2225 relu__19_0_0_nncf_smooth_quant_0" -> "2226 quantize_per_tensor_default_80" [label="(1, 15, 15, 512)", style=solid]; +"2226 quantize_per_tensor_default_80" -> "2227 dequantize_per_tensor_default_80" [label="(1, 15, 15, 512)", style=solid]; +"2227 dequantize_per_tensor_default_80" -> "2232 linear_117" [label="(1, 15, 15, 512)", style=solid]; +"2228 linear_117_scale_0" -> "2231 dequantize_per_channel_default_80" [label="(12,)", style=solid]; +"2229 linear_117_zero_point_0" -> "2231 dequantize_per_channel_default_80" [label="(12,)", style=solid]; +"2230 _frozen_param100" -> "2231 dequantize_per_channel_default_80" [label="(12, 512)", style=solid]; +"2231 dequantize_per_channel_default_80" -> "2232 linear_117" [label="(12, 512)", style=solid]; +"2232 linear_117" -> "2233 view_103" [label="(1, 15, 15, 12)", style=solid]; +"2233 view_103" -> "2234 index_19" [label="(225, 12)", style=solid]; +"2234 index_19" -> "2235 view_104" [label="(4096, 12)", style=solid]; +"2235 view_104" -> "2236 permute_86" [label="(64, 64, 12)", style=solid]; +"2236 permute_86" -> "2237 contiguous_36" [label="(12, 64, 64)", style=solid]; +"2237 contiguous_36" -> "2238 unsqueeze_55" [label="(12, 64, 64)", style=solid]; +"2238 unsqueeze_55" -> "2239 sigmoid_19" [label="(1, 12, 64, 64)", style=solid]; +"2239 sigmoid_19" -> "2240 mul_38" [label="(1, 12, 64, 64)", style=solid]; +"2240 mul_38" -> "2268 add_66" [label="(1, 12, 64, 64)", style=solid]; +"2241 pad_21" -> "2242 roll_18" [label="(1, 16, 16, 384)", style=solid]; +"2242 roll_18" -> "2243 view_105" [label="(1, 16, 16, 384)", style=solid]; +"2243 view_105" -> "2244 permute_87" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2244 permute_87" -> "2245 reshape_85" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2245 reshape_85" -> "2248 reshape_85_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"2245 reshape_85" -> "2269 new_zeros_9" [label="(4, 64, 384)", style=solid]; +"2246 clone_19" -> "2249 linear_118" [label="(1152,)", style=solid]; +"2247 features_5_15_attn_qkv_weight_updated_constant0" -> "2249 linear_118" [label="(1152, 384)", style=solid]; +"2248 reshape_85_0_0_nncf_smooth_quant_0" -> "2249 linear_118" [label="(4, 64, 384)", style=solid]; +"2249 linear_118" -> "2250 reshape_86" [label="(4, 64, 1152)", style=solid]; +"2250 reshape_86" -> "2251 permute_88" [label="(4, 64, 3, 12, 32)", style=solid]; +"2251 permute_88" -> "2252 select_57" [label="(3, 4, 12, 64, 32)", style=solid]; +"2251 permute_88" -> "2253 select_58" [label="(3, 4, 12, 64, 32)", style=solid]; +"2251 permute_88" -> "2254 select_59" [label="(3, 4, 12, 64, 32)", style=solid]; +"2252 select_57" -> "2255 linalg_vector_norm_38" [label="(4, 12, 64, 32)", style=solid]; +"2252 select_57" -> "2257 expand_as_38" [label="(4, 12, 64, 32)", style=solid]; +"2252 select_57" -> "2258 div_38" [label="(4, 12, 64, 32)", style=solid]; +"2253 select_58" -> "2259 linalg_vector_norm_39" [label="(4, 12, 64, 32)", style=solid]; +"2253 select_58" -> "2261 expand_as_39" [label="(4, 12, 64, 32)", style=solid]; +"2253 select_58" -> "2262 div_39" [label="(4, 12, 64, 32)", style=solid]; +"2254 select_59" -> "2287 matmul_39" [label="(4, 12, 64, 32)", style=solid]; +"2255 linalg_vector_norm_38" -> "2256 clamp_min_38" [label="(4, 12, 64, 1)", style=solid]; +"2256 clamp_min_38" -> "2257 expand_as_38" [label="(4, 12, 64, 1)", style=solid]; +"2257 expand_as_38" -> "2258 div_38" [label="(4, 12, 64, 32)", style=solid]; +"2258 div_38" -> "2264 matmul_38" [label="(4, 12, 64, 32)", style=solid]; +"2259 linalg_vector_norm_39" -> "2260 clamp_min_39" [label="(4, 12, 64, 1)", style=solid]; +"2260 clamp_min_39" -> "2261 expand_as_39" [label="(4, 12, 64, 1)", style=solid]; +"2261 expand_as_39" -> "2262 div_39" [label="(4, 12, 64, 32)", style=solid]; +"2262 div_39" -> "2263 transpose_38" [label="(4, 12, 64, 32)", style=solid]; +"2263 transpose_38" -> "2264 matmul_38" [label="(4, 12, 32, 64)", style=solid]; +"2264 matmul_38" -> "2267 mul_39" [label="(4, 12, 64, 64)", style=solid]; +"2265 clamp_19" -> "2266 exp_19" [label="(12, 1, 1)", style=solid]; +"2266 exp_19" -> "2267 mul_39" [label="(12, 1, 1)", style=solid]; +"2267 mul_39" -> "2268 add_66" [label="(4, 12, 64, 64)", style=solid]; +"2268 add_66" -> "2280 view_107" [label="(4, 12, 64, 64)", style=solid]; +"2269 new_zeros_9" -> "2270 view_106" [label="(16, 16)", style=solid]; +"2270 view_106" -> "2271 permute_89" [label="(2, 8, 2, 8)", style=solid]; +"2271 permute_89" -> "2272 reshape_87" [label="(2, 2, 8, 8)", style=solid]; +"2272 reshape_87" -> "2273 unsqueeze_56" [label="(4, 64)", style=solid]; +"2272 reshape_87" -> "2274 unsqueeze_57" [label="(4, 64)", style=solid]; +"2273 unsqueeze_56" -> "2275 sub_9" [label="(4, 1, 64)", style=solid]; +"2274 unsqueeze_57" -> "2275 sub_9" [label="(4, 64, 1)", style=solid]; +"2275 sub_9" -> "2276 ne_9" [label="(4, 64, 64)", style=solid]; +"2275 sub_9" -> "2277 masked_fill_18" [label="(4, 64, 64)", style=solid]; +"2275 sub_9" -> "2278 eq_9" [label="(4, 64, 64)", style=solid]; +"2276 ne_9" -> "2277 masked_fill_18" [label="(4, 64, 64)", style=solid]; +"2277 masked_fill_18" -> "2279 masked_fill_19" [label="(4, 64, 64)", style=solid]; +"2278 eq_9" -> "2279 masked_fill_19" [label="(4, 64, 64)", style=solid]; +"2279 masked_fill_19" -> "2281 unsqueeze_58" [label="(4, 64, 64)", style=solid]; +"2280 view_107" -> "2283 add_67" [label="(1, 4, 12, 64, 64)", style=solid]; +"2281 unsqueeze_58" -> "2282 unsqueeze_59" [label="(4, 1, 64, 64)", style=solid]; +"2282 unsqueeze_59" -> "2283 add_67" [label="(1, 4, 1, 64, 64)", style=solid]; +"2283 add_67" -> "2284 view_108" [label="(1, 4, 12, 64, 64)", style=solid]; +"2284 view_108" -> "2285 softmax_19" [label="(4, 12, 64, 64)", style=solid]; +"2285 softmax_19" -> "2286 dropout_76" [label="(4, 12, 64, 64)", style=solid]; +"2286 dropout_76" -> "2287 matmul_39" [label="(4, 12, 64, 64)", style=solid]; +"2287 matmul_39" -> "2288 transpose_39" [label="(4, 12, 64, 32)", style=solid]; +"2288 transpose_39" -> "2289 reshape_88" [label="(4, 64, 12, 32)", style=solid]; +"2289 reshape_88" -> "2291 reshape_88_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"2290 features_5_15_attn_proj_weight_updated_constant0" -> "2292 linear_119" [label="(384, 384)", style=solid]; +"2291 reshape_88_0_0_nncf_smooth_quant_0" -> "2292 linear_119" [label="(4, 64, 384)", style=solid]; +"2292 linear_119" -> "2293 dropout_77" [label="(4, 64, 384)", style=solid]; +"2293 dropout_77" -> "2294 view_109" [label="(4, 64, 384)", style=solid]; +"2294 view_109" -> "2295 permute_90" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2295 permute_90" -> "2296 reshape_89" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2296 reshape_89" -> "2297 roll_19" [label="(1, 16, 16, 384)", style=solid]; +"2297 roll_19" -> "2298 slice_297" [label="(1, 16, 16, 384)", style=solid]; +"2298 slice_297" -> "2299 slice_298" [label="(1, 16, 16, 384)", style=solid]; +"2299 slice_298" -> "2300 slice_299" [label="(1, 14, 16, 384)", style=solid]; +"2300 slice_299" -> "2301 slice_300" [label="(1, 14, 14, 384)", style=solid]; +"2301 slice_300" -> "2302 contiguous_37" [label="(1, 14, 14, 384)", style=solid]; +"2302 contiguous_37" -> "2303 layer_norm_41" [label="(1, 14, 14, 384)", style=solid]; +"2303 layer_norm_41" -> "2304 add_68" [label="(1, 14, 14, 384)", style=solid]; +"2304 add_68" -> "2305 add_68_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"2304 add_68" -> "2325 add_69" [label="(1, 14, 14, 384)", style=solid]; +"2305 add_68_0_0_nncf_smooth_quant_0" -> "2306 quantize_per_tensor_default_81" [label="(1, 14, 14, 384)", style=solid]; +"2306 quantize_per_tensor_default_81" -> "2307 dequantize_per_tensor_default_81" [label="(1, 14, 14, 384)", style=solid]; +"2307 dequantize_per_tensor_default_81" -> "2312 linear_120" [label="(1, 14, 14, 384)", style=solid]; +"2308 linear_120_scale_0" -> "2311 dequantize_per_channel_default_81" [label="(1536,)", style=solid]; +"2309 linear_120_zero_point_0" -> "2311 dequantize_per_channel_default_81" [label="(1536,)", style=solid]; +"2310 _frozen_param101" -> "2311 dequantize_per_channel_default_81" [label="(1536, 384)", style=solid]; +"2311 dequantize_per_channel_default_81" -> "2312 linear_120" [label="(1536, 384)", style=solid]; +"2312 linear_120" -> "2313 gelu_19" [label="(1, 14, 14, 1536)", style=solid]; +"2313 gelu_19" -> "2314 dropout_78" [label="(1, 14, 14, 1536)", style=solid]; +"2314 dropout_78" -> "2315 dropout_78_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"2315 dropout_78_0_0_nncf_smooth_quant_0" -> "2316 quantize_per_tensor_default_82" [label="(1, 14, 14, 1536)", style=solid]; +"2316 quantize_per_tensor_default_82" -> "2317 dequantize_per_tensor_default_82" [label="(1, 14, 14, 1536)", style=solid]; +"2317 dequantize_per_tensor_default_82" -> "2322 linear_121" [label="(1, 14, 14, 1536)", style=solid]; +"2318 linear_121_scale_0" -> "2321 dequantize_per_channel_default_82" [label="(384,)", style=solid]; +"2319 linear_121_zero_point_0" -> "2321 dequantize_per_channel_default_82" [label="(384,)", style=solid]; +"2320 _frozen_param102" -> "2321 dequantize_per_channel_default_82" [label="(384, 1536)", style=solid]; +"2321 dequantize_per_channel_default_82" -> "2322 linear_121" [label="(384, 1536)", style=solid]; +"2322 linear_121" -> "2323 dropout_79" [label="(1, 14, 14, 384)", style=solid]; +"2323 dropout_79" -> "2324 layer_norm_42" [label="(1, 14, 14, 384)", style=solid]; +"2324 layer_norm_42" -> "2325 add_69" [label="(1, 14, 14, 384)", style=solid]; +"2325 add_69" -> "2350 pad_22" [label="(1, 14, 14, 384)", style=solid]; +"2325 add_69" -> "2395 add_71" [label="(1, 14, 14, 384)", style=solid]; +"2326 _frozen_param103" -> "2327 dequantize_per_tensor_default_83" [label="(1, 15, 15, 2)", style=solid]; +"2327 dequantize_per_tensor_default_83" -> "2332 linear_122" [label="(1, 15, 15, 2)", style=solid]; +"2328 linear_122_scale_0" -> "2331 dequantize_per_channel_default_83" [label="(512,)", style=solid]; +"2329 linear_122_zero_point_0" -> "2331 dequantize_per_channel_default_83" [label="(512,)", style=solid]; +"2330 _frozen_param104" -> "2331 dequantize_per_channel_default_83" [label="(512, 2)", style=solid]; +"2331 dequantize_per_channel_default_83" -> "2332 linear_122" [label="(512, 2)", style=solid]; +"2332 linear_122" -> "2333 relu__20" [label="(1, 15, 15, 512)", style=solid]; +"2333 relu__20" -> "2334 relu__20_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"2334 relu__20_0_0_nncf_smooth_quant_0" -> "2335 quantize_per_tensor_default_84" [label="(1, 15, 15, 512)", style=solid]; +"2335 quantize_per_tensor_default_84" -> "2336 dequantize_per_tensor_default_84" [label="(1, 15, 15, 512)", style=solid]; +"2336 dequantize_per_tensor_default_84" -> "2341 linear_123" [label="(1, 15, 15, 512)", style=solid]; +"2337 linear_123_scale_0" -> "2340 dequantize_per_channel_default_84" [label="(12,)", style=solid]; +"2338 linear_123_zero_point_0" -> "2340 dequantize_per_channel_default_84" [label="(12,)", style=solid]; +"2339 _frozen_param105" -> "2340 dequantize_per_channel_default_84" [label="(12, 512)", style=solid]; +"2340 dequantize_per_channel_default_84" -> "2341 linear_123" [label="(12, 512)", style=solid]; +"2341 linear_123" -> "2342 view_110" [label="(1, 15, 15, 12)", style=solid]; +"2342 view_110" -> "2343 index_20" [label="(225, 12)", style=solid]; +"2343 index_20" -> "2344 view_111" [label="(4096, 12)", style=solid]; +"2344 view_111" -> "2345 permute_91" [label="(64, 64, 12)", style=solid]; +"2345 permute_91" -> "2346 contiguous_38" [label="(12, 64, 64)", style=solid]; +"2346 contiguous_38" -> "2347 unsqueeze_60" [label="(12, 64, 64)", style=solid]; +"2347 unsqueeze_60" -> "2348 sigmoid_20" [label="(1, 12, 64, 64)", style=solid]; +"2348 sigmoid_20" -> "2349 mul_40" [label="(1, 12, 64, 64)", style=solid]; +"2349 mul_40" -> "2376 add_70" [label="(1, 12, 64, 64)", style=solid]; +"2350 pad_22" -> "2351 view_112" [label="(1, 16, 16, 384)", style=solid]; +"2351 view_112" -> "2352 permute_92" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2352 permute_92" -> "2353 reshape_90" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2353 reshape_90" -> "2356 reshape_90_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"2354 clone_20" -> "2357 linear_124" [label="(1152,)", style=solid]; +"2355 features_5_16_attn_qkv_weight_updated_constant0" -> "2357 linear_124" [label="(1152, 384)", style=solid]; +"2356 reshape_90_0_0_nncf_smooth_quant_0" -> "2357 linear_124" [label="(4, 64, 384)", style=solid]; +"2357 linear_124" -> "2358 reshape_91" [label="(4, 64, 1152)", style=solid]; +"2358 reshape_91" -> "2359 permute_93" [label="(4, 64, 3, 12, 32)", style=solid]; +"2359 permute_93" -> "2360 select_60" [label="(3, 4, 12, 64, 32)", style=solid]; +"2359 permute_93" -> "2361 select_61" [label="(3, 4, 12, 64, 32)", style=solid]; +"2359 permute_93" -> "2362 select_62" [label="(3, 4, 12, 64, 32)", style=solid]; +"2360 select_60" -> "2363 linalg_vector_norm_40" [label="(4, 12, 64, 32)", style=solid]; +"2360 select_60" -> "2365 expand_as_40" [label="(4, 12, 64, 32)", style=solid]; +"2360 select_60" -> "2366 div_40" [label="(4, 12, 64, 32)", style=solid]; +"2361 select_61" -> "2367 linalg_vector_norm_41" [label="(4, 12, 64, 32)", style=solid]; +"2361 select_61" -> "2369 expand_as_41" [label="(4, 12, 64, 32)", style=solid]; +"2361 select_61" -> "2370 div_41" [label="(4, 12, 64, 32)", style=solid]; +"2362 select_62" -> "2379 matmul_41" [label="(4, 12, 64, 32)", style=solid]; +"2363 linalg_vector_norm_40" -> "2364 clamp_min_40" [label="(4, 12, 64, 1)", style=solid]; +"2364 clamp_min_40" -> "2365 expand_as_40" [label="(4, 12, 64, 1)", style=solid]; +"2365 expand_as_40" -> "2366 div_40" [label="(4, 12, 64, 32)", style=solid]; +"2366 div_40" -> "2372 matmul_40" [label="(4, 12, 64, 32)", style=solid]; +"2367 linalg_vector_norm_41" -> "2368 clamp_min_41" [label="(4, 12, 64, 1)", style=solid]; +"2368 clamp_min_41" -> "2369 expand_as_41" [label="(4, 12, 64, 1)", style=solid]; +"2369 expand_as_41" -> "2370 div_41" [label="(4, 12, 64, 32)", style=solid]; +"2370 div_41" -> "2371 transpose_40" [label="(4, 12, 64, 32)", style=solid]; +"2371 transpose_40" -> "2372 matmul_40" [label="(4, 12, 32, 64)", style=solid]; +"2372 matmul_40" -> "2375 mul_41" [label="(4, 12, 64, 64)", style=solid]; +"2373 clamp_20" -> "2374 exp_20" [label="(12, 1, 1)", style=solid]; +"2374 exp_20" -> "2375 mul_41" [label="(12, 1, 1)", style=solid]; +"2375 mul_41" -> "2376 add_70" [label="(4, 12, 64, 64)", style=solid]; +"2376 add_70" -> "2377 softmax_20" [label="(4, 12, 64, 64)", style=solid]; +"2377 softmax_20" -> "2378 dropout_80" [label="(4, 12, 64, 64)", style=solid]; +"2378 dropout_80" -> "2379 matmul_41" [label="(4, 12, 64, 64)", style=solid]; +"2379 matmul_41" -> "2380 transpose_41" [label="(4, 12, 64, 32)", style=solid]; +"2380 transpose_41" -> "2381 reshape_92" [label="(4, 64, 12, 32)", style=solid]; +"2381 reshape_92" -> "2383 reshape_92_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"2382 features_5_16_attn_proj_weight_updated_constant0" -> "2384 linear_125" [label="(384, 384)", style=solid]; +"2383 reshape_92_0_0_nncf_smooth_quant_0" -> "2384 linear_125" [label="(4, 64, 384)", style=solid]; +"2384 linear_125" -> "2385 dropout_81" [label="(4, 64, 384)", style=solid]; +"2385 dropout_81" -> "2386 view_113" [label="(4, 64, 384)", style=solid]; +"2386 view_113" -> "2387 permute_94" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2387 permute_94" -> "2388 reshape_93" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2388 reshape_93" -> "2389 slice_302" [label="(1, 16, 16, 384)", style=solid]; +"2389 slice_302" -> "2390 slice_303" [label="(1, 16, 16, 384)", style=solid]; +"2390 slice_303" -> "2391 slice_304" [label="(1, 14, 16, 384)", style=solid]; +"2391 slice_304" -> "2392 slice_305" [label="(1, 14, 14, 384)", style=solid]; +"2392 slice_305" -> "2393 contiguous_39" [label="(1, 14, 14, 384)", style=solid]; +"2393 contiguous_39" -> "2394 layer_norm_43" [label="(1, 14, 14, 384)", style=solid]; +"2394 layer_norm_43" -> "2395 add_71" [label="(1, 14, 14, 384)", style=solid]; +"2395 add_71" -> "2396 add_71_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"2395 add_71" -> "2416 add_72" [label="(1, 14, 14, 384)", style=solid]; +"2396 add_71_0_0_nncf_smooth_quant_0" -> "2397 quantize_per_tensor_default_85" [label="(1, 14, 14, 384)", style=solid]; +"2397 quantize_per_tensor_default_85" -> "2398 dequantize_per_tensor_default_85" [label="(1, 14, 14, 384)", style=solid]; +"2398 dequantize_per_tensor_default_85" -> "2403 linear_126" [label="(1, 14, 14, 384)", style=solid]; +"2399 linear_126_scale_0" -> "2402 dequantize_per_channel_default_85" [label="(1536,)", style=solid]; +"2400 linear_126_zero_point_0" -> "2402 dequantize_per_channel_default_85" [label="(1536,)", style=solid]; +"2401 _frozen_param106" -> "2402 dequantize_per_channel_default_85" [label="(1536, 384)", style=solid]; +"2402 dequantize_per_channel_default_85" -> "2403 linear_126" [label="(1536, 384)", style=solid]; +"2403 linear_126" -> "2404 gelu_20" [label="(1, 14, 14, 1536)", style=solid]; +"2404 gelu_20" -> "2405 dropout_82" [label="(1, 14, 14, 1536)", style=solid]; +"2405 dropout_82" -> "2406 dropout_82_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"2406 dropout_82_0_0_nncf_smooth_quant_0" -> "2407 quantize_per_tensor_default_86" [label="(1, 14, 14, 1536)", style=solid]; +"2407 quantize_per_tensor_default_86" -> "2408 dequantize_per_tensor_default_86" [label="(1, 14, 14, 1536)", style=solid]; +"2408 dequantize_per_tensor_default_86" -> "2413 linear_127" [label="(1, 14, 14, 1536)", style=solid]; +"2409 linear_127_scale_0" -> "2412 dequantize_per_channel_default_86" [label="(384,)", style=solid]; +"2410 linear_127_zero_point_0" -> "2412 dequantize_per_channel_default_86" [label="(384,)", style=solid]; +"2411 _frozen_param107" -> "2412 dequantize_per_channel_default_86" [label="(384, 1536)", style=solid]; +"2412 dequantize_per_channel_default_86" -> "2413 linear_127" [label="(384, 1536)", style=solid]; +"2413 linear_127" -> "2414 dropout_83" [label="(1, 14, 14, 384)", style=solid]; +"2414 dropout_83" -> "2415 layer_norm_44" [label="(1, 14, 14, 384)", style=solid]; +"2415 layer_norm_44" -> "2416 add_72" [label="(1, 14, 14, 384)", style=solid]; +"2416 add_72" -> "2441 pad_23" [label="(1, 14, 14, 384)", style=solid]; +"2416 add_72" -> "2504 add_75" [label="(1, 14, 14, 384)", style=solid]; +"2417 _frozen_param108" -> "2418 dequantize_per_tensor_default_87" [label="(1, 15, 15, 2)", style=solid]; +"2418 dequantize_per_tensor_default_87" -> "2423 linear_128" [label="(1, 15, 15, 2)", style=solid]; +"2419 linear_128_scale_0" -> "2422 dequantize_per_channel_default_87" [label="(512,)", style=solid]; +"2420 linear_128_zero_point_0" -> "2422 dequantize_per_channel_default_87" [label="(512,)", style=solid]; +"2421 _frozen_param109" -> "2422 dequantize_per_channel_default_87" [label="(512, 2)", style=solid]; +"2422 dequantize_per_channel_default_87" -> "2423 linear_128" [label="(512, 2)", style=solid]; +"2423 linear_128" -> "2424 relu__21" [label="(1, 15, 15, 512)", style=solid]; +"2424 relu__21" -> "2425 relu__21_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"2425 relu__21_0_0_nncf_smooth_quant_0" -> "2426 quantize_per_tensor_default_88" [label="(1, 15, 15, 512)", style=solid]; +"2426 quantize_per_tensor_default_88" -> "2427 dequantize_per_tensor_default_88" [label="(1, 15, 15, 512)", style=solid]; +"2427 dequantize_per_tensor_default_88" -> "2432 linear_129" [label="(1, 15, 15, 512)", style=solid]; +"2428 linear_129_scale_0" -> "2431 dequantize_per_channel_default_88" [label="(12,)", style=solid]; +"2429 linear_129_zero_point_0" -> "2431 dequantize_per_channel_default_88" [label="(12,)", style=solid]; +"2430 _frozen_param110" -> "2431 dequantize_per_channel_default_88" [label="(12, 512)", style=solid]; +"2431 dequantize_per_channel_default_88" -> "2432 linear_129" [label="(12, 512)", style=solid]; +"2432 linear_129" -> "2433 view_114" [label="(1, 15, 15, 12)", style=solid]; +"2433 view_114" -> "2434 index_21" [label="(225, 12)", style=solid]; +"2434 index_21" -> "2435 view_115" [label="(4096, 12)", style=solid]; +"2435 view_115" -> "2436 permute_95" [label="(64, 64, 12)", style=solid]; +"2436 permute_95" -> "2437 contiguous_40" [label="(12, 64, 64)", style=solid]; +"2437 contiguous_40" -> "2438 unsqueeze_61" [label="(12, 64, 64)", style=solid]; +"2438 unsqueeze_61" -> "2439 sigmoid_21" [label="(1, 12, 64, 64)", style=solid]; +"2439 sigmoid_21" -> "2440 mul_42" [label="(1, 12, 64, 64)", style=solid]; +"2440 mul_42" -> "2468 add_73" [label="(1, 12, 64, 64)", style=solid]; +"2441 pad_23" -> "2442 roll_20" [label="(1, 16, 16, 384)", style=solid]; +"2442 roll_20" -> "2443 view_116" [label="(1, 16, 16, 384)", style=solid]; +"2443 view_116" -> "2444 permute_96" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2444 permute_96" -> "2445 reshape_94" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2445 reshape_94" -> "2448 reshape_94_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"2445 reshape_94" -> "2469 new_zeros_10" [label="(4, 64, 384)", style=solid]; +"2446 clone_21" -> "2449 linear_130" [label="(1152,)", style=solid]; +"2447 features_5_17_attn_qkv_weight_updated_constant0" -> "2449 linear_130" [label="(1152, 384)", style=solid]; +"2448 reshape_94_0_0_nncf_smooth_quant_0" -> "2449 linear_130" [label="(4, 64, 384)", style=solid]; +"2449 linear_130" -> "2450 reshape_95" [label="(4, 64, 1152)", style=solid]; +"2450 reshape_95" -> "2451 permute_97" [label="(4, 64, 3, 12, 32)", style=solid]; +"2451 permute_97" -> "2452 select_63" [label="(3, 4, 12, 64, 32)", style=solid]; +"2451 permute_97" -> "2453 select_64" [label="(3, 4, 12, 64, 32)", style=solid]; +"2451 permute_97" -> "2454 select_65" [label="(3, 4, 12, 64, 32)", style=solid]; +"2452 select_63" -> "2455 linalg_vector_norm_42" [label="(4, 12, 64, 32)", style=solid]; +"2452 select_63" -> "2457 expand_as_42" [label="(4, 12, 64, 32)", style=solid]; +"2452 select_63" -> "2458 div_42" [label="(4, 12, 64, 32)", style=solid]; +"2453 select_64" -> "2459 linalg_vector_norm_43" [label="(4, 12, 64, 32)", style=solid]; +"2453 select_64" -> "2461 expand_as_43" [label="(4, 12, 64, 32)", style=solid]; +"2453 select_64" -> "2462 div_43" [label="(4, 12, 64, 32)", style=solid]; +"2454 select_65" -> "2487 matmul_43" [label="(4, 12, 64, 32)", style=solid]; +"2455 linalg_vector_norm_42" -> "2456 clamp_min_42" [label="(4, 12, 64, 1)", style=solid]; +"2456 clamp_min_42" -> "2457 expand_as_42" [label="(4, 12, 64, 1)", style=solid]; +"2457 expand_as_42" -> "2458 div_42" [label="(4, 12, 64, 32)", style=solid]; +"2458 div_42" -> "2464 matmul_42" [label="(4, 12, 64, 32)", style=solid]; +"2459 linalg_vector_norm_43" -> "2460 clamp_min_43" [label="(4, 12, 64, 1)", style=solid]; +"2460 clamp_min_43" -> "2461 expand_as_43" [label="(4, 12, 64, 1)", style=solid]; +"2461 expand_as_43" -> "2462 div_43" [label="(4, 12, 64, 32)", style=solid]; +"2462 div_43" -> "2463 transpose_42" [label="(4, 12, 64, 32)", style=solid]; +"2463 transpose_42" -> "2464 matmul_42" [label="(4, 12, 32, 64)", style=solid]; +"2464 matmul_42" -> "2467 mul_43" [label="(4, 12, 64, 64)", style=solid]; +"2465 clamp_21" -> "2466 exp_21" [label="(12, 1, 1)", style=solid]; +"2466 exp_21" -> "2467 mul_43" [label="(12, 1, 1)", style=solid]; +"2467 mul_43" -> "2468 add_73" [label="(4, 12, 64, 64)", style=solid]; +"2468 add_73" -> "2480 view_118" [label="(4, 12, 64, 64)", style=solid]; +"2469 new_zeros_10" -> "2470 view_117" [label="(16, 16)", style=solid]; +"2470 view_117" -> "2471 permute_98" [label="(2, 8, 2, 8)", style=solid]; +"2471 permute_98" -> "2472 reshape_96" [label="(2, 2, 8, 8)", style=solid]; +"2472 reshape_96" -> "2473 unsqueeze_62" [label="(4, 64)", style=solid]; +"2472 reshape_96" -> "2474 unsqueeze_63" [label="(4, 64)", style=solid]; +"2473 unsqueeze_62" -> "2475 sub_10" [label="(4, 1, 64)", style=solid]; +"2474 unsqueeze_63" -> "2475 sub_10" [label="(4, 64, 1)", style=solid]; +"2475 sub_10" -> "2476 ne_10" [label="(4, 64, 64)", style=solid]; +"2475 sub_10" -> "2477 masked_fill_20" [label="(4, 64, 64)", style=solid]; +"2475 sub_10" -> "2478 eq_10" [label="(4, 64, 64)", style=solid]; +"2476 ne_10" -> "2477 masked_fill_20" [label="(4, 64, 64)", style=solid]; +"2477 masked_fill_20" -> "2479 masked_fill_21" [label="(4, 64, 64)", style=solid]; +"2478 eq_10" -> "2479 masked_fill_21" [label="(4, 64, 64)", style=solid]; +"2479 masked_fill_21" -> "2481 unsqueeze_64" [label="(4, 64, 64)", style=solid]; +"2480 view_118" -> "2483 add_74" [label="(1, 4, 12, 64, 64)", style=solid]; +"2481 unsqueeze_64" -> "2482 unsqueeze_65" [label="(4, 1, 64, 64)", style=solid]; +"2482 unsqueeze_65" -> "2483 add_74" [label="(1, 4, 1, 64, 64)", style=solid]; +"2483 add_74" -> "2484 view_119" [label="(1, 4, 12, 64, 64)", style=solid]; +"2484 view_119" -> "2485 softmax_21" [label="(4, 12, 64, 64)", style=solid]; +"2485 softmax_21" -> "2486 dropout_84" [label="(4, 12, 64, 64)", style=solid]; +"2486 dropout_84" -> "2487 matmul_43" [label="(4, 12, 64, 64)", style=solid]; +"2487 matmul_43" -> "2488 transpose_43" [label="(4, 12, 64, 32)", style=solid]; +"2488 transpose_43" -> "2489 reshape_97" [label="(4, 64, 12, 32)", style=solid]; +"2489 reshape_97" -> "2491 reshape_97_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"2490 features_5_17_attn_proj_weight_updated_constant0" -> "2492 linear_131" [label="(384, 384)", style=solid]; +"2491 reshape_97_0_0_nncf_smooth_quant_0" -> "2492 linear_131" [label="(4, 64, 384)", style=solid]; +"2492 linear_131" -> "2493 dropout_85" [label="(4, 64, 384)", style=solid]; +"2493 dropout_85" -> "2494 view_120" [label="(4, 64, 384)", style=solid]; +"2494 view_120" -> "2495 permute_99" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2495 permute_99" -> "2496 reshape_98" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2496 reshape_98" -> "2497 roll_21" [label="(1, 16, 16, 384)", style=solid]; +"2497 roll_21" -> "2498 slice_325" [label="(1, 16, 16, 384)", style=solid]; +"2498 slice_325" -> "2499 slice_326" [label="(1, 16, 16, 384)", style=solid]; +"2499 slice_326" -> "2500 slice_327" [label="(1, 14, 16, 384)", style=solid]; +"2500 slice_327" -> "2501 slice_328" [label="(1, 14, 14, 384)", style=solid]; +"2501 slice_328" -> "2502 contiguous_41" [label="(1, 14, 14, 384)", style=solid]; +"2502 contiguous_41" -> "2503 layer_norm_45" [label="(1, 14, 14, 384)", style=solid]; +"2503 layer_norm_45" -> "2504 add_75" [label="(1, 14, 14, 384)", style=solid]; +"2504 add_75" -> "2505 add_75_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"2504 add_75" -> "2525 add_76" [label="(1, 14, 14, 384)", style=solid]; +"2505 add_75_0_0_nncf_smooth_quant_0" -> "2506 quantize_per_tensor_default_89" [label="(1, 14, 14, 384)", style=solid]; +"2506 quantize_per_tensor_default_89" -> "2507 dequantize_per_tensor_default_89" [label="(1, 14, 14, 384)", style=solid]; +"2507 dequantize_per_tensor_default_89" -> "2512 linear_132" [label="(1, 14, 14, 384)", style=solid]; +"2508 linear_132_scale_0" -> "2511 dequantize_per_channel_default_89" [label="(1536,)", style=solid]; +"2509 linear_132_zero_point_0" -> "2511 dequantize_per_channel_default_89" [label="(1536,)", style=solid]; +"2510 _frozen_param111" -> "2511 dequantize_per_channel_default_89" [label="(1536, 384)", style=solid]; +"2511 dequantize_per_channel_default_89" -> "2512 linear_132" [label="(1536, 384)", style=solid]; +"2512 linear_132" -> "2513 gelu_21" [label="(1, 14, 14, 1536)", style=solid]; +"2513 gelu_21" -> "2514 dropout_86" [label="(1, 14, 14, 1536)", style=solid]; +"2514 dropout_86" -> "2515 dropout_86_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"2515 dropout_86_0_0_nncf_smooth_quant_0" -> "2516 quantize_per_tensor_default_90" [label="(1, 14, 14, 1536)", style=solid]; +"2516 quantize_per_tensor_default_90" -> "2517 dequantize_per_tensor_default_90" [label="(1, 14, 14, 1536)", style=solid]; +"2517 dequantize_per_tensor_default_90" -> "2522 linear_133" [label="(1, 14, 14, 1536)", style=solid]; +"2518 linear_133_scale_0" -> "2521 dequantize_per_channel_default_90" [label="(384,)", style=solid]; +"2519 linear_133_zero_point_0" -> "2521 dequantize_per_channel_default_90" [label="(384,)", style=solid]; +"2520 _frozen_param112" -> "2521 dequantize_per_channel_default_90" [label="(384, 1536)", style=solid]; +"2521 dequantize_per_channel_default_90" -> "2522 linear_133" [label="(384, 1536)", style=solid]; +"2522 linear_133" -> "2523 dropout_87" [label="(1, 14, 14, 384)", style=solid]; +"2523 dropout_87" -> "2524 layer_norm_46" [label="(1, 14, 14, 384)", style=solid]; +"2524 layer_norm_46" -> "2525 add_76" [label="(1, 14, 14, 384)", style=solid]; +"2525 add_76" -> "2526 pad_24" [label="(1, 14, 14, 384)", style=solid]; +"2526 pad_24" -> "2527 slice_329" [label="(1, 14, 14, 384)", style=solid]; +"2526 pad_24" -> "2530 slice_332" [label="(1, 14, 14, 384)", style=solid]; +"2526 pad_24" -> "2533 slice_335" [label="(1, 14, 14, 384)", style=solid]; +"2526 pad_24" -> "2536 slice_338" [label="(1, 14, 14, 384)", style=solid]; +"2527 slice_329" -> "2528 slice_330" [label="(1, 7, 14, 384)", style=solid]; +"2528 slice_330" -> "2529 slice_331" [label="(1, 7, 7, 384)", style=solid]; +"2529 slice_331" -> "2539 cat_2" [label="(1, 7, 7, 384)", style=solid]; +"2530 slice_332" -> "2531 slice_333" [label="(1, 7, 14, 384)", style=solid]; +"2531 slice_333" -> "2532 slice_334" [label="(1, 7, 7, 384)", style=solid]; +"2532 slice_334" -> "2539 cat_2" [label="(1, 7, 7, 384)", style=solid]; +"2533 slice_335" -> "2534 slice_336" [label="(1, 7, 14, 384)", style=solid]; +"2534 slice_336" -> "2535 slice_337" [label="(1, 7, 7, 384)", style=solid]; +"2535 slice_337" -> "2539 cat_2" [label="(1, 7, 7, 384)", style=solid]; +"2536 slice_338" -> "2537 slice_339" [label="(1, 7, 14, 384)", style=solid]; +"2537 slice_339" -> "2538 slice_340" [label="(1, 7, 7, 384)", style=solid]; +"2538 slice_340" -> "2539 cat_2" [label="(1, 7, 7, 384)", style=solid]; +"2539 cat_2" -> "2540 cat_2_0_0_nncf_smooth_quant_0" [label="(1, 7, 7, 1536)", style=solid]; +"2540 cat_2_0_0_nncf_smooth_quant_0" -> "2541 quantize_per_tensor_default_91" [label="(1, 7, 7, 1536)", style=solid]; +"2541 quantize_per_tensor_default_91" -> "2542 dequantize_per_tensor_default_91" [label="(1, 7, 7, 1536)", style=solid]; +"2542 dequantize_per_tensor_default_91" -> "2547 linear_134" [label="(1, 7, 7, 1536)", style=solid]; +"2543 linear_134_scale_0" -> "2546 dequantize_per_channel_default_91" [label="(768,)", style=solid]; +"2544 linear_134_zero_point_0" -> "2546 dequantize_per_channel_default_91" [label="(768,)", style=solid]; +"2545 _frozen_param113" -> "2546 dequantize_per_channel_default_91" [label="(768, 1536)", style=solid]; +"2546 dequantize_per_channel_default_91" -> "2547 linear_134" [label="(768, 1536)", style=solid]; +"2547 linear_134" -> "2548 layer_norm_47" [label="(1, 7, 7, 768)", style=solid]; +"2548 layer_norm_47" -> "2573 pad_25" [label="(1, 7, 7, 768)", style=solid]; +"2548 layer_norm_47" -> "2618 add_78" [label="(1, 7, 7, 768)", style=solid]; +"2549 _frozen_param114" -> "2550 dequantize_per_tensor_default_92" [label="(1, 15, 15, 2)", style=solid]; +"2550 dequantize_per_tensor_default_92" -> "2555 linear_135" [label="(1, 15, 15, 2)", style=solid]; +"2551 linear_135_scale_0" -> "2554 dequantize_per_channel_default_92" [label="(512,)", style=solid]; +"2552 linear_135_zero_point_0" -> "2554 dequantize_per_channel_default_92" [label="(512,)", style=solid]; +"2553 _frozen_param115" -> "2554 dequantize_per_channel_default_92" [label="(512, 2)", style=solid]; +"2554 dequantize_per_channel_default_92" -> "2555 linear_135" [label="(512, 2)", style=solid]; +"2555 linear_135" -> "2556 relu__22" [label="(1, 15, 15, 512)", style=solid]; +"2556 relu__22" -> "2557 relu__22_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"2557 relu__22_0_0_nncf_smooth_quant_0" -> "2558 quantize_per_tensor_default_93" [label="(1, 15, 15, 512)", style=solid]; +"2558 quantize_per_tensor_default_93" -> "2559 dequantize_per_tensor_default_93" [label="(1, 15, 15, 512)", style=solid]; +"2559 dequantize_per_tensor_default_93" -> "2564 linear_136" [label="(1, 15, 15, 512)", style=solid]; +"2560 linear_136_scale_0" -> "2563 dequantize_per_channel_default_93" [label="(24,)", style=solid]; +"2561 linear_136_zero_point_0" -> "2563 dequantize_per_channel_default_93" [label="(24,)", style=solid]; +"2562 _frozen_param116" -> "2563 dequantize_per_channel_default_93" [label="(24, 512)", style=solid]; +"2563 dequantize_per_channel_default_93" -> "2564 linear_136" [label="(24, 512)", style=solid]; +"2564 linear_136" -> "2565 view_121" [label="(1, 15, 15, 24)", style=solid]; +"2565 view_121" -> "2566 index_22" [label="(225, 24)", style=solid]; +"2566 index_22" -> "2567 view_122" [label="(4096, 24)", style=solid]; +"2567 view_122" -> "2568 permute_100" [label="(64, 64, 24)", style=solid]; +"2568 permute_100" -> "2569 contiguous_42" [label="(24, 64, 64)", style=solid]; +"2569 contiguous_42" -> "2570 unsqueeze_66" [label="(24, 64, 64)", style=solid]; +"2570 unsqueeze_66" -> "2571 sigmoid_22" [label="(1, 24, 64, 64)", style=solid]; +"2571 sigmoid_22" -> "2572 mul_44" [label="(1, 24, 64, 64)", style=solid]; +"2572 mul_44" -> "2599 add_77" [label="(1, 24, 64, 64)", style=solid]; +"2573 pad_25" -> "2574 view_123" [label="(1, 8, 8, 768)", style=solid]; +"2574 view_123" -> "2575 permute_101" [label="(1, 1, 8, 1, 8, 768)", style=solid]; +"2575 permute_101" -> "2576 reshape_99" [label="(1, 1, 1, 8, 8, 768)", style=solid]; +"2576 reshape_99" -> "2579 reshape_99_0_0_nncf_smooth_quant_0" [label="(1, 64, 768)", style=solid]; +"2577 clone_22" -> "2580 linear_137" [label="(2304,)", style=solid]; +"2578 features_7_0_attn_qkv_weight_updated_constant0" -> "2580 linear_137" [label="(2304, 768)", style=solid]; +"2579 reshape_99_0_0_nncf_smooth_quant_0" -> "2580 linear_137" [label="(1, 64, 768)", style=solid]; +"2580 linear_137" -> "2581 reshape_100" [label="(1, 64, 2304)", style=solid]; +"2581 reshape_100" -> "2582 permute_102" [label="(1, 64, 3, 24, 32)", style=solid]; +"2582 permute_102" -> "2583 select_66" [label="(3, 1, 24, 64, 32)", style=solid]; +"2582 permute_102" -> "2584 select_67" [label="(3, 1, 24, 64, 32)", style=solid]; +"2582 permute_102" -> "2585 select_68" [label="(3, 1, 24, 64, 32)", style=solid]; +"2583 select_66" -> "2586 linalg_vector_norm_44" [label="(1, 24, 64, 32)", style=solid]; +"2583 select_66" -> "2588 expand_as_44" [label="(1, 24, 64, 32)", style=solid]; +"2583 select_66" -> "2589 div_44" [label="(1, 24, 64, 32)", style=solid]; +"2584 select_67" -> "2590 linalg_vector_norm_45" [label="(1, 24, 64, 32)", style=solid]; +"2584 select_67" -> "2592 expand_as_45" [label="(1, 24, 64, 32)", style=solid]; +"2584 select_67" -> "2593 div_45" [label="(1, 24, 64, 32)", style=solid]; +"2585 select_68" -> "2602 matmul_45" [label="(1, 24, 64, 32)", style=solid]; +"2586 linalg_vector_norm_44" -> "2587 clamp_min_44" [label="(1, 24, 64, 1)", style=solid]; +"2587 clamp_min_44" -> "2588 expand_as_44" [label="(1, 24, 64, 1)", style=solid]; +"2588 expand_as_44" -> "2589 div_44" [label="(1, 24, 64, 32)", style=solid]; +"2589 div_44" -> "2595 matmul_44" [label="(1, 24, 64, 32)", style=solid]; +"2590 linalg_vector_norm_45" -> "2591 clamp_min_45" [label="(1, 24, 64, 1)", style=solid]; +"2591 clamp_min_45" -> "2592 expand_as_45" [label="(1, 24, 64, 1)", style=solid]; +"2592 expand_as_45" -> "2593 div_45" [label="(1, 24, 64, 32)", style=solid]; +"2593 div_45" -> "2594 transpose_44" [label="(1, 24, 64, 32)", style=solid]; +"2594 transpose_44" -> "2595 matmul_44" [label="(1, 24, 32, 64)", style=solid]; +"2595 matmul_44" -> "2598 mul_45" [label="(1, 24, 64, 64)", style=solid]; +"2596 clamp_22" -> "2597 exp_22" [label="(24, 1, 1)", style=solid]; +"2597 exp_22" -> "2598 mul_45" [label="(24, 1, 1)", style=solid]; +"2598 mul_45" -> "2599 add_77" [label="(1, 24, 64, 64)", style=solid]; +"2599 add_77" -> "2600 softmax_22" [label="(1, 24, 64, 64)", style=solid]; +"2600 softmax_22" -> "2601 dropout_88" [label="(1, 24, 64, 64)", style=solid]; +"2601 dropout_88" -> "2602 matmul_45" [label="(1, 24, 64, 64)", style=solid]; +"2602 matmul_45" -> "2603 transpose_45" [label="(1, 24, 64, 32)", style=solid]; +"2603 transpose_45" -> "2604 reshape_101" [label="(1, 64, 24, 32)", style=solid]; +"2604 reshape_101" -> "2606 reshape_101_0_0_nncf_smooth_quant_0" [label="(1, 64, 768)", style=solid]; +"2605 features_7_0_attn_proj_weight_updated_constant0" -> "2607 linear_138" [label="(768, 768)", style=solid]; +"2606 reshape_101_0_0_nncf_smooth_quant_0" -> "2607 linear_138" [label="(1, 64, 768)", style=solid]; +"2607 linear_138" -> "2608 dropout_89" [label="(1, 64, 768)", style=solid]; +"2608 dropout_89" -> "2609 view_124" [label="(1, 64, 768)", style=solid]; +"2609 view_124" -> "2610 permute_103" [label="(1, 1, 1, 8, 8, 768)", style=solid]; +"2610 permute_103" -> "2611 reshape_102" [label="(1, 1, 8, 1, 8, 768)", style=solid]; +"2611 reshape_102" -> "2612 slice_342" [label="(1, 8, 8, 768)", style=solid]; +"2612 slice_342" -> "2613 slice_343" [label="(1, 8, 8, 768)", style=solid]; +"2613 slice_343" -> "2614 slice_344" [label="(1, 7, 8, 768)", style=solid]; +"2614 slice_344" -> "2615 slice_345" [label="(1, 7, 7, 768)", style=solid]; +"2615 slice_345" -> "2616 contiguous_43" [label="(1, 7, 7, 768)", style=solid]; +"2616 contiguous_43" -> "2617 layer_norm_48" [label="(1, 7, 7, 768)", style=solid]; +"2617 layer_norm_48" -> "2618 add_78" [label="(1, 7, 7, 768)", style=solid]; +"2618 add_78" -> "2619 add_78_0_0_nncf_smooth_quant_0" [label="(1, 7, 7, 768)", style=solid]; +"2618 add_78" -> "2639 add_79" [label="(1, 7, 7, 768)", style=solid]; +"2619 add_78_0_0_nncf_smooth_quant_0" -> "2620 quantize_per_tensor_default_94" [label="(1, 7, 7, 768)", style=solid]; +"2620 quantize_per_tensor_default_94" -> "2621 dequantize_per_tensor_default_94" [label="(1, 7, 7, 768)", style=solid]; +"2621 dequantize_per_tensor_default_94" -> "2626 linear_139" [label="(1, 7, 7, 768)", style=solid]; +"2622 linear_139_scale_0" -> "2625 dequantize_per_channel_default_94" [label="(3072,)", style=solid]; +"2623 linear_139_zero_point_0" -> "2625 dequantize_per_channel_default_94" [label="(3072,)", style=solid]; +"2624 _frozen_param117" -> "2625 dequantize_per_channel_default_94" [label="(3072, 768)", style=solid]; +"2625 dequantize_per_channel_default_94" -> "2626 linear_139" [label="(3072, 768)", style=solid]; +"2626 linear_139" -> "2627 gelu_22" [label="(1, 7, 7, 3072)", style=solid]; +"2627 gelu_22" -> "2628 dropout_90" [label="(1, 7, 7, 3072)", style=solid]; +"2628 dropout_90" -> "2629 dropout_90_0_0_nncf_smooth_quant_0" [label="(1, 7, 7, 3072)", style=solid]; +"2629 dropout_90_0_0_nncf_smooth_quant_0" -> "2630 quantize_per_tensor_default_95" [label="(1, 7, 7, 3072)", style=solid]; +"2630 quantize_per_tensor_default_95" -> "2631 dequantize_per_tensor_default_95" [label="(1, 7, 7, 3072)", style=solid]; +"2631 dequantize_per_tensor_default_95" -> "2636 linear_140" [label="(1, 7, 7, 3072)", style=solid]; +"2632 linear_140_scale_0" -> "2635 dequantize_per_channel_default_95" [label="(768,)", style=solid]; +"2633 linear_140_zero_point_0" -> "2635 dequantize_per_channel_default_95" [label="(768,)", style=solid]; +"2634 _frozen_param118" -> "2635 dequantize_per_channel_default_95" [label="(768, 3072)", style=solid]; +"2635 dequantize_per_channel_default_95" -> "2636 linear_140" [label="(768, 3072)", style=solid]; +"2636 linear_140" -> "2637 dropout_91" [label="(1, 7, 7, 768)", style=solid]; +"2637 dropout_91" -> "2638 layer_norm_49" [label="(1, 7, 7, 768)", style=solid]; +"2638 layer_norm_49" -> "2639 add_79" [label="(1, 7, 7, 768)", style=solid]; +"2639 add_79" -> "2664 pad_26" [label="(1, 7, 7, 768)", style=solid]; +"2639 add_79" -> "2709 add_81" [label="(1, 7, 7, 768)", style=solid]; +"2640 _frozen_param119" -> "2641 dequantize_per_tensor_default_96" [label="(1, 15, 15, 2)", style=solid]; +"2641 dequantize_per_tensor_default_96" -> "2646 linear_141" [label="(1, 15, 15, 2)", style=solid]; +"2642 linear_141_scale_0" -> "2645 dequantize_per_channel_default_96" [label="(512,)", style=solid]; +"2643 linear_141_zero_point_0" -> "2645 dequantize_per_channel_default_96" [label="(512,)", style=solid]; +"2644 _frozen_param120" -> "2645 dequantize_per_channel_default_96" [label="(512, 2)", style=solid]; +"2645 dequantize_per_channel_default_96" -> "2646 linear_141" [label="(512, 2)", style=solid]; +"2646 linear_141" -> "2647 relu__23" [label="(1, 15, 15, 512)", style=solid]; +"2647 relu__23" -> "2648 relu__23_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"2648 relu__23_0_0_nncf_smooth_quant_0" -> "2649 quantize_per_tensor_default_97" [label="(1, 15, 15, 512)", style=solid]; +"2649 quantize_per_tensor_default_97" -> "2650 dequantize_per_tensor_default_97" [label="(1, 15, 15, 512)", style=solid]; +"2650 dequantize_per_tensor_default_97" -> "2655 linear_142" [label="(1, 15, 15, 512)", style=solid]; +"2651 linear_142_scale_0" -> "2654 dequantize_per_channel_default_97" [label="(24,)", style=solid]; +"2652 linear_142_zero_point_0" -> "2654 dequantize_per_channel_default_97" [label="(24,)", style=solid]; +"2653 _frozen_param121" -> "2654 dequantize_per_channel_default_97" [label="(24, 512)", style=solid]; +"2654 dequantize_per_channel_default_97" -> "2655 linear_142" [label="(24, 512)", style=solid]; +"2655 linear_142" -> "2656 view_125" [label="(1, 15, 15, 24)", style=solid]; +"2656 view_125" -> "2657 index_23" [label="(225, 24)", style=solid]; +"2657 index_23" -> "2658 view_126" [label="(4096, 24)", style=solid]; +"2658 view_126" -> "2659 permute_104" [label="(64, 64, 24)", style=solid]; +"2659 permute_104" -> "2660 contiguous_44" [label="(24, 64, 64)", style=solid]; +"2660 contiguous_44" -> "2661 unsqueeze_67" [label="(24, 64, 64)", style=solid]; +"2661 unsqueeze_67" -> "2662 sigmoid_23" [label="(1, 24, 64, 64)", style=solid]; +"2662 sigmoid_23" -> "2663 mul_46" [label="(1, 24, 64, 64)", style=solid]; +"2663 mul_46" -> "2690 add_80" [label="(1, 24, 64, 64)", style=solid]; +"2664 pad_26" -> "2665 view_127" [label="(1, 8, 8, 768)", style=solid]; +"2665 view_127" -> "2666 permute_105" [label="(1, 1, 8, 1, 8, 768)", style=solid]; +"2666 permute_105" -> "2667 reshape_103" [label="(1, 1, 1, 8, 8, 768)", style=solid]; +"2667 reshape_103" -> "2670 reshape_103_0_0_nncf_smooth_quant_0" [label="(1, 64, 768)", style=solid]; +"2668 clone_23" -> "2671 linear_143" [label="(2304,)", style=solid]; +"2669 features_7_1_attn_qkv_weight_updated_constant0" -> "2671 linear_143" [label="(2304, 768)", style=solid]; +"2670 reshape_103_0_0_nncf_smooth_quant_0" -> "2671 linear_143" [label="(1, 64, 768)", style=solid]; +"2671 linear_143" -> "2672 reshape_104" [label="(1, 64, 2304)", style=solid]; +"2672 reshape_104" -> "2673 permute_106" [label="(1, 64, 3, 24, 32)", style=solid]; +"2673 permute_106" -> "2674 select_69" [label="(3, 1, 24, 64, 32)", style=solid]; +"2673 permute_106" -> "2675 select_70" [label="(3, 1, 24, 64, 32)", style=solid]; +"2673 permute_106" -> "2676 select_71" [label="(3, 1, 24, 64, 32)", style=solid]; +"2674 select_69" -> "2677 linalg_vector_norm_46" [label="(1, 24, 64, 32)", style=solid]; +"2674 select_69" -> "2679 expand_as_46" [label="(1, 24, 64, 32)", style=solid]; +"2674 select_69" -> "2680 div_46" [label="(1, 24, 64, 32)", style=solid]; +"2675 select_70" -> "2681 linalg_vector_norm_47" [label="(1, 24, 64, 32)", style=solid]; +"2675 select_70" -> "2683 expand_as_47" [label="(1, 24, 64, 32)", style=solid]; +"2675 select_70" -> "2684 div_47" [label="(1, 24, 64, 32)", style=solid]; +"2676 select_71" -> "2693 matmul_47" [label="(1, 24, 64, 32)", style=solid]; +"2677 linalg_vector_norm_46" -> "2678 clamp_min_46" [label="(1, 24, 64, 1)", style=solid]; +"2678 clamp_min_46" -> "2679 expand_as_46" [label="(1, 24, 64, 1)", style=solid]; +"2679 expand_as_46" -> "2680 div_46" [label="(1, 24, 64, 32)", style=solid]; +"2680 div_46" -> "2686 matmul_46" [label="(1, 24, 64, 32)", style=solid]; +"2681 linalg_vector_norm_47" -> "2682 clamp_min_47" [label="(1, 24, 64, 1)", style=solid]; +"2682 clamp_min_47" -> "2683 expand_as_47" [label="(1, 24, 64, 1)", style=solid]; +"2683 expand_as_47" -> "2684 div_47" [label="(1, 24, 64, 32)", style=solid]; +"2684 div_47" -> "2685 transpose_46" [label="(1, 24, 64, 32)", style=solid]; +"2685 transpose_46" -> "2686 matmul_46" [label="(1, 24, 32, 64)", style=solid]; +"2686 matmul_46" -> "2689 mul_47" [label="(1, 24, 64, 64)", style=solid]; +"2687 clamp_23" -> "2688 exp_23" [label="(24, 1, 1)", style=solid]; +"2688 exp_23" -> "2689 mul_47" [label="(24, 1, 1)", style=solid]; +"2689 mul_47" -> "2690 add_80" [label="(1, 24, 64, 64)", style=solid]; +"2690 add_80" -> "2691 softmax_23" [label="(1, 24, 64, 64)", style=solid]; +"2691 softmax_23" -> "2692 dropout_92" [label="(1, 24, 64, 64)", style=solid]; +"2692 dropout_92" -> "2693 matmul_47" [label="(1, 24, 64, 64)", style=solid]; +"2693 matmul_47" -> "2694 transpose_47" [label="(1, 24, 64, 32)", style=solid]; +"2694 transpose_47" -> "2695 reshape_105" [label="(1, 64, 24, 32)", style=solid]; +"2695 reshape_105" -> "2697 reshape_105_0_0_nncf_smooth_quant_0" [label="(1, 64, 768)", style=solid]; +"2696 features_7_1_attn_proj_weight_updated_constant0" -> "2698 linear_144" [label="(768, 768)", style=solid]; +"2697 reshape_105_0_0_nncf_smooth_quant_0" -> "2698 linear_144" [label="(1, 64, 768)", style=solid]; +"2698 linear_144" -> "2699 dropout_93" [label="(1, 64, 768)", style=solid]; +"2699 dropout_93" -> "2700 view_128" [label="(1, 64, 768)", style=solid]; +"2700 view_128" -> "2701 permute_107" [label="(1, 1, 1, 8, 8, 768)", style=solid]; +"2701 permute_107" -> "2702 reshape_106" [label="(1, 1, 8, 1, 8, 768)", style=solid]; +"2702 reshape_106" -> "2703 slice_347" [label="(1, 8, 8, 768)", style=solid]; +"2703 slice_347" -> "2704 slice_348" [label="(1, 8, 8, 768)", style=solid]; +"2704 slice_348" -> "2705 slice_349" [label="(1, 7, 8, 768)", style=solid]; +"2705 slice_349" -> "2706 slice_350" [label="(1, 7, 7, 768)", style=solid]; +"2706 slice_350" -> "2707 contiguous_45" [label="(1, 7, 7, 768)", style=solid]; +"2707 contiguous_45" -> "2708 layer_norm_50" [label="(1, 7, 7, 768)", style=solid]; +"2708 layer_norm_50" -> "2709 add_81" [label="(1, 7, 7, 768)", style=solid]; +"2709 add_81" -> "2710 add_81_0_0_nncf_smooth_quant_0" [label="(1, 7, 7, 768)", style=solid]; +"2709 add_81" -> "2730 add_82" [label="(1, 7, 7, 768)", style=solid]; +"2710 add_81_0_0_nncf_smooth_quant_0" -> "2711 quantize_per_tensor_default_98" [label="(1, 7, 7, 768)", style=solid]; +"2711 quantize_per_tensor_default_98" -> "2712 dequantize_per_tensor_default_98" [label="(1, 7, 7, 768)", style=solid]; +"2712 dequantize_per_tensor_default_98" -> "2717 linear_145" [label="(1, 7, 7, 768)", style=solid]; +"2713 linear_145_scale_0" -> "2716 dequantize_per_channel_default_98" [label="(3072,)", style=solid]; +"2714 linear_145_zero_point_0" -> "2716 dequantize_per_channel_default_98" [label="(3072,)", style=solid]; +"2715 _frozen_param122" -> "2716 dequantize_per_channel_default_98" [label="(3072, 768)", style=solid]; +"2716 dequantize_per_channel_default_98" -> "2717 linear_145" [label="(3072, 768)", style=solid]; +"2717 linear_145" -> "2718 gelu_23" [label="(1, 7, 7, 3072)", style=solid]; +"2718 gelu_23" -> "2719 dropout_94" [label="(1, 7, 7, 3072)", style=solid]; +"2719 dropout_94" -> "2720 dropout_94_0_0_nncf_smooth_quant_0" [label="(1, 7, 7, 3072)", style=solid]; +"2720 dropout_94_0_0_nncf_smooth_quant_0" -> "2721 quantize_per_tensor_default_99" [label="(1, 7, 7, 3072)", style=solid]; +"2721 quantize_per_tensor_default_99" -> "2722 dequantize_per_tensor_default_99" [label="(1, 7, 7, 3072)", style=solid]; +"2722 dequantize_per_tensor_default_99" -> "2727 linear_146" [label="(1, 7, 7, 3072)", style=solid]; +"2723 linear_146_scale_0" -> "2726 dequantize_per_channel_default_99" [label="(768,)", style=solid]; +"2724 linear_146_zero_point_0" -> "2726 dequantize_per_channel_default_99" [label="(768,)", style=solid]; +"2725 _frozen_param123" -> "2726 dequantize_per_channel_default_99" [label="(768, 3072)", style=solid]; +"2726 dequantize_per_channel_default_99" -> "2727 linear_146" [label="(768, 3072)", style=solid]; +"2727 linear_146" -> "2728 dropout_95" [label="(1, 7, 7, 768)", style=solid]; +"2728 dropout_95" -> "2729 layer_norm_51" [label="(1, 7, 7, 768)", style=solid]; +"2729 layer_norm_51" -> "2730 add_82" [label="(1, 7, 7, 768)", style=solid]; +"2730 add_82" -> "2731 layer_norm_52" [label="(1, 7, 7, 768)", style=solid]; +"2731 layer_norm_52" -> "2732 permute_108" [label="(1, 7, 7, 768)", style=solid]; +"2732 permute_108" -> "2733 adaptive_avg_pool2d" [label="(1, 768, 7, 7)", style=solid]; +"2733 adaptive_avg_pool2d" -> "2734 flatten" [label="(1, 768, 1, 1)", style=solid]; +"2734 flatten" -> "2735 flatten_0_0_nncf_smooth_quant_0" [label="(1, 768)", style=solid]; +"2735 flatten_0_0_nncf_smooth_quant_0" -> "2736 quantize_per_tensor_default_100" [label="(1, 768)", style=solid]; +"2736 quantize_per_tensor_default_100" -> "2737 dequantize_per_tensor_default_100" [label="(1, 768)", style=solid]; +"2737 dequantize_per_tensor_default_100" -> "2742 linear_147" [label="(1, 768)", style=solid]; +"2738 linear_147_scale_0" -> "2741 dequantize_per_channel_default_100" [label="(1000,)", style=solid]; +"2739 linear_147_zero_point_0" -> "2741 dequantize_per_channel_default_100" [label="(1000,)", style=solid]; +"2740 _frozen_param124" -> "2741 dequantize_per_channel_default_100" [label="(1000, 768)", style=solid]; +"2741 dequantize_per_channel_default_100" -> "2742 linear_147" [label="(1000, 768)", style=solid]; +"2742 linear_147" -> "2743 output" [label="(1, 1000)", style=solid]; +} diff --git a/tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/synthetic_transformer.dot b/tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/synthetic_transformer.dot new file mode 100644 index 00000000000..462a2e45e41 --- /dev/null +++ b/tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/synthetic_transformer.dot @@ -0,0 +1,45 @@ +strict digraph { +"0 wte_weight" [id=0, type=get_attr]; +"1 linear_bias" [id=1, type=get_attr]; +"2 lm_head_bias" [id=2, type=get_attr]; +"3 input_ids" [id=3, type=input]; +"4 embedding" [id=4, type=embedding]; +"5 embedding_0_0_nncf_smooth_quant_0" [id=5, type=call_module]; +"6 quantize_per_tensor_default" [id=6, type=quantize_per_tensor]; +"7 dequantize_per_tensor_default" [id=7, type=dequantize_per_tensor]; +"8 linear_scale_0" [id=8, type=get_attr]; +"9 linear_zero_point_0" [id=9, type=get_attr]; +"10 _frozen_param0" [id=10, type=get_attr]; +"11 dequantize_per_channel_default" [id=11, type=dequantize_per_channel]; +"12 linear" [id=12, type=linear]; +"13 linear_0_0_nncf_smooth_quant_0" [id=13, type=call_module]; +"14 quantize_per_tensor_default_1" [id=14, type=quantize_per_tensor]; +"15 dequantize_per_tensor_default_1" [id=15, type=dequantize_per_tensor]; +"16 linear_1_scale_0" [id=16, type=get_attr]; +"17 linear_1_zero_point_0" [id=17, type=get_attr]; +"18 _frozen_param1" [id=18, type=get_attr]; +"19 dequantize_per_channel_default_1" [id=19, type=dequantize_per_channel]; +"20 linear_1" [id=20, type=linear]; +"21 output" [id=21, type=output]; +"0 wte_weight" -> "4 embedding" [label="(10, 5)", style=solid]; +"1 linear_bias" -> "12 linear" [label="(5,)", style=solid]; +"2 lm_head_bias" -> "20 linear_1" [label="(10,)", style=solid]; +"3 input_ids" -> "4 embedding" [label="(5,)", style=solid]; +"4 embedding" -> "5 embedding_0_0_nncf_smooth_quant_0" [label="(5, 5)", style=solid]; +"5 embedding_0_0_nncf_smooth_quant_0" -> "6 quantize_per_tensor_default" [label="(5, 5)", style=solid]; +"6 quantize_per_tensor_default" -> "7 dequantize_per_tensor_default" [label="(5, 5)", style=solid]; +"7 dequantize_per_tensor_default" -> "12 linear" [label="(5, 5)", style=solid]; +"8 linear_scale_0" -> "11 dequantize_per_channel_default" [label="(5,)", style=solid]; +"9 linear_zero_point_0" -> "11 dequantize_per_channel_default" [label="(5,)", style=solid]; +"10 _frozen_param0" -> "11 dequantize_per_channel_default" [label="(5, 5)", style=solid]; +"11 dequantize_per_channel_default" -> "12 linear" [label="(5, 5)", style=solid]; +"12 linear" -> "13 linear_0_0_nncf_smooth_quant_0" [label="(5, 5)", style=solid]; +"13 linear_0_0_nncf_smooth_quant_0" -> "14 quantize_per_tensor_default_1" [label="(5, 5)", style=solid]; +"14 quantize_per_tensor_default_1" -> "15 dequantize_per_tensor_default_1" [label="(5, 5)", style=solid]; +"15 dequantize_per_tensor_default_1" -> "20 linear_1" [label="(5, 5)", style=solid]; +"16 linear_1_scale_0" -> "19 dequantize_per_channel_default_1" [label="(10,)", style=solid]; +"17 linear_1_zero_point_0" -> "19 dequantize_per_channel_default_1" [label="(10,)", style=solid]; +"18 _frozen_param1" -> "19 dequantize_per_channel_default_1" [label="(10, 5)", style=solid]; +"19 dequantize_per_channel_default_1" -> "20 linear_1" [label="(10, 5)", style=solid]; +"20 linear_1" -> "21 output" [label="(5, 10)", style=solid]; +} diff --git a/tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/unet.dot b/tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/unet.dot new file mode 100644 index 00000000000..dfe64020019 --- /dev/null +++ b/tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/unet.dot @@ -0,0 +1,437 @@ +strict digraph { +"0 down_path_0_block_0_bias" [id=0, type=get_attr]; +"1 down_path_0_block_3_bias" [id=1, type=get_attr]; +"2 down_path_1_block_0_bias" [id=2, type=get_attr]; +"3 down_path_1_block_3_bias" [id=3, type=get_attr]; +"4 down_path_2_block_0_bias" [id=4, type=get_attr]; +"5 down_path_2_block_3_bias" [id=5, type=get_attr]; +"6 down_path_3_block_0_bias" [id=6, type=get_attr]; +"7 down_path_3_block_3_bias" [id=7, type=get_attr]; +"8 down_path_4_block_0_bias" [id=8, type=get_attr]; +"9 down_path_4_block_3_bias" [id=9, type=get_attr]; +"10 up_path_0_up_weight" [id=10, type=get_attr]; +"11 up_path_0_up_bias" [id=11, type=get_attr]; +"12 up_path_0_conv_block_block_0_bias" [id=12, type=get_attr]; +"13 up_path_0_conv_block_block_3_bias" [id=13, type=get_attr]; +"14 up_path_1_up_weight" [id=14, type=get_attr]; +"15 up_path_1_up_bias" [id=15, type=get_attr]; +"16 up_path_1_conv_block_block_0_bias" [id=16, type=get_attr]; +"17 up_path_1_conv_block_block_3_bias" [id=17, type=get_attr]; +"18 up_path_2_up_weight" [id=18, type=get_attr]; +"19 up_path_2_up_bias" [id=19, type=get_attr]; +"20 up_path_2_conv_block_block_0_bias" [id=20, type=get_attr]; +"21 up_path_2_conv_block_block_3_bias" [id=21, type=get_attr]; +"22 up_path_3_up_weight" [id=22, type=get_attr]; +"23 up_path_3_up_bias" [id=23, type=get_attr]; +"24 up_path_3_conv_block_block_0_bias" [id=24, type=get_attr]; +"25 up_path_3_conv_block_block_3_bias" [id=25, type=get_attr]; +"26 last_bias" [id=26, type=get_attr]; +"27 x" [id=27, type=input]; +"28 quantize_per_tensor_default" [id=28, type=quantize_per_tensor]; +"29 dequantize_per_tensor_default" [id=29, type=dequantize_per_tensor]; +"30 conv2d_scale_0" [id=30, type=get_attr]; +"31 conv2d_zero_point_0" [id=31, type=get_attr]; +"32 _frozen_param0" [id=32, type=get_attr]; +"33 dequantize_per_channel_default" [id=33, type=dequantize_per_channel]; +"34 conv2d" [id=34, type=conv2d]; +"35 relu" [id=35, type=relu]; +"36 quantize_per_tensor_default_1" [id=36, type=quantize_per_tensor]; +"37 dequantize_per_tensor_default_1" [id=37, type=dequantize_per_tensor]; +"38 conv2d_1_scale_0" [id=38, type=get_attr]; +"39 conv2d_1_zero_point_0" [id=39, type=get_attr]; +"40 _frozen_param1" [id=40, type=get_attr]; +"41 dequantize_per_channel_default_1" [id=41, type=dequantize_per_channel]; +"42 conv2d_1" [id=42, type=conv2d]; +"43 relu_1" [id=43, type=relu]; +"44 quantize_per_tensor_default_2" [id=44, type=quantize_per_tensor]; +"45 dequantize_per_tensor_default_2" [id=45, type=dequantize_per_tensor]; +"46 max_pool2d" [id=46, type=max_pool2d]; +"47 quantize_per_tensor_default_3" [id=47, type=quantize_per_tensor]; +"48 dequantize_per_tensor_default_3" [id=48, type=dequantize_per_tensor]; +"49 conv2d_2_scale_0" [id=49, type=get_attr]; +"50 conv2d_2_zero_point_0" [id=50, type=get_attr]; +"51 _frozen_param2" [id=51, type=get_attr]; +"52 dequantize_per_channel_default_2" [id=52, type=dequantize_per_channel]; +"53 conv2d_2" [id=53, type=conv2d]; +"54 relu_2" [id=54, type=relu]; +"55 quantize_per_tensor_default_4" [id=55, type=quantize_per_tensor]; +"56 dequantize_per_tensor_default_4" [id=56, type=dequantize_per_tensor]; +"57 conv2d_3_scale_0" [id=57, type=get_attr]; +"58 conv2d_3_zero_point_0" [id=58, type=get_attr]; +"59 _frozen_param3" [id=59, type=get_attr]; +"60 dequantize_per_channel_default_3" [id=60, type=dequantize_per_channel]; +"61 conv2d_3" [id=61, type=conv2d]; +"62 relu_3" [id=62, type=relu]; +"63 quantize_per_tensor_default_5" [id=63, type=quantize_per_tensor]; +"64 dequantize_per_tensor_default_5" [id=64, type=dequantize_per_tensor]; +"65 max_pool2d_1" [id=65, type=max_pool2d]; +"66 quantize_per_tensor_default_6" [id=66, type=quantize_per_tensor]; +"67 dequantize_per_tensor_default_6" [id=67, type=dequantize_per_tensor]; +"68 conv2d_4_scale_0" [id=68, type=get_attr]; +"69 conv2d_4_zero_point_0" [id=69, type=get_attr]; +"70 _frozen_param4" [id=70, type=get_attr]; +"71 dequantize_per_channel_default_4" [id=71, type=dequantize_per_channel]; +"72 conv2d_4" [id=72, type=conv2d]; +"73 relu_4" [id=73, type=relu]; +"74 quantize_per_tensor_default_7" [id=74, type=quantize_per_tensor]; +"75 dequantize_per_tensor_default_7" [id=75, type=dequantize_per_tensor]; +"76 conv2d_5_scale_0" [id=76, type=get_attr]; +"77 conv2d_5_zero_point_0" [id=77, type=get_attr]; +"78 _frozen_param5" [id=78, type=get_attr]; +"79 dequantize_per_channel_default_5" [id=79, type=dequantize_per_channel]; +"80 conv2d_5" [id=80, type=conv2d]; +"81 relu_5" [id=81, type=relu]; +"82 quantize_per_tensor_default_8" [id=82, type=quantize_per_tensor]; +"83 dequantize_per_tensor_default_8" [id=83, type=dequantize_per_tensor]; +"84 max_pool2d_2" [id=84, type=max_pool2d]; +"85 quantize_per_tensor_default_9" [id=85, type=quantize_per_tensor]; +"86 dequantize_per_tensor_default_9" [id=86, type=dequantize_per_tensor]; +"87 conv2d_6_scale_0" [id=87, type=get_attr]; +"88 conv2d_6_zero_point_0" [id=88, type=get_attr]; +"89 _frozen_param6" [id=89, type=get_attr]; +"90 dequantize_per_channel_default_6" [id=90, type=dequantize_per_channel]; +"91 conv2d_6" [id=91, type=conv2d]; +"92 relu_6" [id=92, type=relu]; +"93 quantize_per_tensor_default_10" [id=93, type=quantize_per_tensor]; +"94 dequantize_per_tensor_default_10" [id=94, type=dequantize_per_tensor]; +"95 conv2d_7_scale_0" [id=95, type=get_attr]; +"96 conv2d_7_zero_point_0" [id=96, type=get_attr]; +"97 _frozen_param7" [id=97, type=get_attr]; +"98 dequantize_per_channel_default_7" [id=98, type=dequantize_per_channel]; +"99 conv2d_7" [id=99, type=conv2d]; +"100 relu_7" [id=100, type=relu]; +"101 quantize_per_tensor_default_11" [id=101, type=quantize_per_tensor]; +"102 dequantize_per_tensor_default_11" [id=102, type=dequantize_per_tensor]; +"103 max_pool2d_3" [id=103, type=max_pool2d]; +"104 quantize_per_tensor_default_12" [id=104, type=quantize_per_tensor]; +"105 dequantize_per_tensor_default_12" [id=105, type=dequantize_per_tensor]; +"106 conv2d_8_scale_0" [id=106, type=get_attr]; +"107 conv2d_8_zero_point_0" [id=107, type=get_attr]; +"108 _frozen_param8" [id=108, type=get_attr]; +"109 dequantize_per_channel_default_8" [id=109, type=dequantize_per_channel]; +"110 conv2d_8" [id=110, type=conv2d]; +"111 relu_8" [id=111, type=relu]; +"112 quantize_per_tensor_default_13" [id=112, type=quantize_per_tensor]; +"113 dequantize_per_tensor_default_13" [id=113, type=dequantize_per_tensor]; +"114 conv2d_9_scale_0" [id=114, type=get_attr]; +"115 conv2d_9_zero_point_0" [id=115, type=get_attr]; +"116 _frozen_param9" [id=116, type=get_attr]; +"117 dequantize_per_channel_default_9" [id=117, type=dequantize_per_channel]; +"118 conv2d_9" [id=118, type=conv2d]; +"119 relu_9" [id=119, type=relu]; +"120 conv_transpose2d" [id=120, type=conv_transpose2d]; +"121 slice_1" [id=121, type=slice]; +"122 slice_2" [id=122, type=slice]; +"123 slice_3" [id=123, type=slice]; +"124 slice_4" [id=124, type=slice]; +"125 cat" [id=125, type=cat]; +"126 quantize_per_tensor_default_14" [id=126, type=quantize_per_tensor]; +"127 dequantize_per_tensor_default_14" [id=127, type=dequantize_per_tensor]; +"128 conv2d_10_scale_0" [id=128, type=get_attr]; +"129 conv2d_10_zero_point_0" [id=129, type=get_attr]; +"130 _frozen_param10" [id=130, type=get_attr]; +"131 dequantize_per_channel_default_10" [id=131, type=dequantize_per_channel]; +"132 conv2d_10" [id=132, type=conv2d]; +"133 relu_10" [id=133, type=relu]; +"134 quantize_per_tensor_default_15" [id=134, type=quantize_per_tensor]; +"135 dequantize_per_tensor_default_15" [id=135, type=dequantize_per_tensor]; +"136 conv2d_11_scale_0" [id=136, type=get_attr]; +"137 conv2d_11_zero_point_0" [id=137, type=get_attr]; +"138 _frozen_param11" [id=138, type=get_attr]; +"139 dequantize_per_channel_default_11" [id=139, type=dequantize_per_channel]; +"140 conv2d_11" [id=140, type=conv2d]; +"141 relu_11" [id=141, type=relu]; +"142 conv_transpose2d_1" [id=142, type=conv_transpose2d]; +"143 slice_5" [id=143, type=slice]; +"144 slice_6" [id=144, type=slice]; +"145 slice_7" [id=145, type=slice]; +"146 slice_8" [id=146, type=slice]; +"147 cat_1" [id=147, type=cat]; +"148 quantize_per_tensor_default_16" [id=148, type=quantize_per_tensor]; +"149 dequantize_per_tensor_default_16" [id=149, type=dequantize_per_tensor]; +"150 conv2d_12_scale_0" [id=150, type=get_attr]; +"151 conv2d_12_zero_point_0" [id=151, type=get_attr]; +"152 _frozen_param12" [id=152, type=get_attr]; +"153 dequantize_per_channel_default_12" [id=153, type=dequantize_per_channel]; +"154 conv2d_12" [id=154, type=conv2d]; +"155 relu_12" [id=155, type=relu]; +"156 quantize_per_tensor_default_17" [id=156, type=quantize_per_tensor]; +"157 dequantize_per_tensor_default_17" [id=157, type=dequantize_per_tensor]; +"158 conv2d_13_scale_0" [id=158, type=get_attr]; +"159 conv2d_13_zero_point_0" [id=159, type=get_attr]; +"160 _frozen_param13" [id=160, type=get_attr]; +"161 dequantize_per_channel_default_13" [id=161, type=dequantize_per_channel]; +"162 conv2d_13" [id=162, type=conv2d]; +"163 relu_13" [id=163, type=relu]; +"164 conv_transpose2d_2" [id=164, type=conv_transpose2d]; +"165 slice_9" [id=165, type=slice]; +"166 slice_10" [id=166, type=slice]; +"167 slice_11" [id=167, type=slice]; +"168 slice_12" [id=168, type=slice]; +"169 cat_2" [id=169, type=cat]; +"170 quantize_per_tensor_default_18" [id=170, type=quantize_per_tensor]; +"171 dequantize_per_tensor_default_18" [id=171, type=dequantize_per_tensor]; +"172 conv2d_14_scale_0" [id=172, type=get_attr]; +"173 conv2d_14_zero_point_0" [id=173, type=get_attr]; +"174 _frozen_param14" [id=174, type=get_attr]; +"175 dequantize_per_channel_default_14" [id=175, type=dequantize_per_channel]; +"176 conv2d_14" [id=176, type=conv2d]; +"177 relu_14" [id=177, type=relu]; +"178 quantize_per_tensor_default_19" [id=178, type=quantize_per_tensor]; +"179 dequantize_per_tensor_default_19" [id=179, type=dequantize_per_tensor]; +"180 conv2d_15_scale_0" [id=180, type=get_attr]; +"181 conv2d_15_zero_point_0" [id=181, type=get_attr]; +"182 _frozen_param15" [id=182, type=get_attr]; +"183 dequantize_per_channel_default_15" [id=183, type=dequantize_per_channel]; +"184 conv2d_15" [id=184, type=conv2d]; +"185 relu_15" [id=185, type=relu]; +"186 conv_transpose2d_3" [id=186, type=conv_transpose2d]; +"187 slice_13" [id=187, type=slice]; +"188 slice_14" [id=188, type=slice]; +"189 slice_15" [id=189, type=slice]; +"190 slice_16" [id=190, type=slice]; +"191 cat_3" [id=191, type=cat]; +"192 quantize_per_tensor_default_20" [id=192, type=quantize_per_tensor]; +"193 dequantize_per_tensor_default_20" [id=193, type=dequantize_per_tensor]; +"194 conv2d_16_scale_0" [id=194, type=get_attr]; +"195 conv2d_16_zero_point_0" [id=195, type=get_attr]; +"196 _frozen_param16" [id=196, type=get_attr]; +"197 dequantize_per_channel_default_16" [id=197, type=dequantize_per_channel]; +"198 conv2d_16" [id=198, type=conv2d]; +"199 relu_16" [id=199, type=relu]; +"200 quantize_per_tensor_default_21" [id=200, type=quantize_per_tensor]; +"201 dequantize_per_tensor_default_21" [id=201, type=dequantize_per_tensor]; +"202 conv2d_17_scale_0" [id=202, type=get_attr]; +"203 conv2d_17_zero_point_0" [id=203, type=get_attr]; +"204 _frozen_param17" [id=204, type=get_attr]; +"205 dequantize_per_channel_default_17" [id=205, type=dequantize_per_channel]; +"206 conv2d_17" [id=206, type=conv2d]; +"207 relu_17" [id=207, type=relu]; +"208 quantize_per_tensor_default_22" [id=208, type=quantize_per_tensor]; +"209 dequantize_per_tensor_default_22" [id=209, type=dequantize_per_tensor]; +"210 conv2d_18_scale_0" [id=210, type=get_attr]; +"211 conv2d_18_zero_point_0" [id=211, type=get_attr]; +"212 _frozen_param18" [id=212, type=get_attr]; +"213 dequantize_per_channel_default_18" [id=213, type=dequantize_per_channel]; +"214 conv2d_18" [id=214, type=conv2d]; +"215 output" [id=215, type=output]; +"0 down_path_0_block_0_bias" -> "34 conv2d" [label="(64,)", style=solid]; +"1 down_path_0_block_3_bias" -> "42 conv2d_1" [label="(64,)", style=solid]; +"2 down_path_1_block_0_bias" -> "53 conv2d_2" [label="(128,)", style=solid]; +"3 down_path_1_block_3_bias" -> "61 conv2d_3" [label="(128,)", style=solid]; +"4 down_path_2_block_0_bias" -> "72 conv2d_4" [label="(256,)", style=solid]; +"5 down_path_2_block_3_bias" -> "80 conv2d_5" [label="(256,)", style=solid]; +"6 down_path_3_block_0_bias" -> "91 conv2d_6" [label="(512,)", style=solid]; +"7 down_path_3_block_3_bias" -> "99 conv2d_7" [label="(512,)", style=solid]; +"8 down_path_4_block_0_bias" -> "110 conv2d_8" [label="(1024,)", style=solid]; +"9 down_path_4_block_3_bias" -> "118 conv2d_9" [label="(1024,)", style=solid]; +"10 up_path_0_up_weight" -> "120 conv_transpose2d" [label="(1024, 512, 2, 2)", style=solid]; +"11 up_path_0_up_bias" -> "120 conv_transpose2d" [label="(512,)", style=solid]; +"12 up_path_0_conv_block_block_0_bias" -> "132 conv2d_10" [label="(512,)", style=solid]; +"13 up_path_0_conv_block_block_3_bias" -> "140 conv2d_11" [label="(512,)", style=solid]; +"14 up_path_1_up_weight" -> "142 conv_transpose2d_1" [label="(512, 256, 2, 2)", style=solid]; +"15 up_path_1_up_bias" -> "142 conv_transpose2d_1" [label="(256,)", style=solid]; +"16 up_path_1_conv_block_block_0_bias" -> "154 conv2d_12" [label="(256,)", style=solid]; +"17 up_path_1_conv_block_block_3_bias" -> "162 conv2d_13" [label="(256,)", style=solid]; +"18 up_path_2_up_weight" -> "164 conv_transpose2d_2" [label="(256, 128, 2, 2)", style=solid]; +"19 up_path_2_up_bias" -> "164 conv_transpose2d_2" [label="(128,)", style=solid]; +"20 up_path_2_conv_block_block_0_bias" -> "176 conv2d_14" [label="(128,)", style=solid]; +"21 up_path_2_conv_block_block_3_bias" -> "184 conv2d_15" [label="(128,)", style=solid]; +"22 up_path_3_up_weight" -> "186 conv_transpose2d_3" [label="(128, 64, 2, 2)", style=solid]; +"23 up_path_3_up_bias" -> "186 conv_transpose2d_3" [label="(64,)", style=solid]; +"24 up_path_3_conv_block_block_0_bias" -> "198 conv2d_16" [label="(64,)", style=solid]; +"25 up_path_3_conv_block_block_3_bias" -> "206 conv2d_17" [label="(64,)", style=solid]; +"26 last_bias" -> "214 conv2d_18" [label="(12,)", style=solid]; +"27 x" -> "28 quantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; +"28 quantize_per_tensor_default" -> "29 dequantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; +"29 dequantize_per_tensor_default" -> "34 conv2d" [label="(1, 3, 224, 224)", style=solid]; +"30 conv2d_scale_0" -> "33 dequantize_per_channel_default" [label="(64,)", style=solid]; +"31 conv2d_zero_point_0" -> "33 dequantize_per_channel_default" [label="(64,)", style=solid]; +"32 _frozen_param0" -> "33 dequantize_per_channel_default" [label="(64, 3, 3, 3)", style=solid]; +"33 dequantize_per_channel_default" -> "34 conv2d" [label="(64, 3, 3, 3)", style=solid]; +"34 conv2d" -> "35 relu" [label="(1, 64, 222, 222)", style=solid]; +"35 relu" -> "36 quantize_per_tensor_default_1" [label="(1, 64, 222, 222)", style=solid]; +"36 quantize_per_tensor_default_1" -> "37 dequantize_per_tensor_default_1" [label="(1, 64, 222, 222)", style=solid]; +"37 dequantize_per_tensor_default_1" -> "42 conv2d_1" [label="(1, 64, 222, 222)", style=solid]; +"38 conv2d_1_scale_0" -> "41 dequantize_per_channel_default_1" [label="(64,)", style=solid]; +"39 conv2d_1_zero_point_0" -> "41 dequantize_per_channel_default_1" [label="(64,)", style=solid]; +"40 _frozen_param1" -> "41 dequantize_per_channel_default_1" [label="(64, 64, 3, 3)", style=solid]; +"41 dequantize_per_channel_default_1" -> "42 conv2d_1" [label="(64, 64, 3, 3)", style=solid]; +"42 conv2d_1" -> "43 relu_1" [label="(1, 64, 220, 220)", style=solid]; +"43 relu_1" -> "44 quantize_per_tensor_default_2" [label="(1, 64, 220, 220)", style=solid]; +"43 relu_1" -> "187 slice_13" [label="(1, 64, 220, 220)", style=solid]; +"44 quantize_per_tensor_default_2" -> "45 dequantize_per_tensor_default_2" [label="(1, 64, 220, 220)", style=solid]; +"45 dequantize_per_tensor_default_2" -> "46 max_pool2d" [label="(1, 64, 220, 220)", style=solid]; +"46 max_pool2d" -> "47 quantize_per_tensor_default_3" [label="(1, 64, 110, 110)", style=solid]; +"47 quantize_per_tensor_default_3" -> "48 dequantize_per_tensor_default_3" [label="(1, 64, 110, 110)", style=solid]; +"48 dequantize_per_tensor_default_3" -> "53 conv2d_2" [label="(1, 64, 110, 110)", style=solid]; +"49 conv2d_2_scale_0" -> "52 dequantize_per_channel_default_2" [label="(128,)", style=solid]; +"50 conv2d_2_zero_point_0" -> "52 dequantize_per_channel_default_2" [label="(128,)", style=solid]; +"51 _frozen_param2" -> "52 dequantize_per_channel_default_2" [label="(128, 64, 3, 3)", style=solid]; +"52 dequantize_per_channel_default_2" -> "53 conv2d_2" [label="(128, 64, 3, 3)", style=solid]; +"53 conv2d_2" -> "54 relu_2" [label="(1, 128, 108, 108)", style=solid]; +"54 relu_2" -> "55 quantize_per_tensor_default_4" [label="(1, 128, 108, 108)", style=solid]; +"55 quantize_per_tensor_default_4" -> "56 dequantize_per_tensor_default_4" [label="(1, 128, 108, 108)", style=solid]; +"56 dequantize_per_tensor_default_4" -> "61 conv2d_3" [label="(1, 128, 108, 108)", style=solid]; +"57 conv2d_3_scale_0" -> "60 dequantize_per_channel_default_3" [label="(128,)", style=solid]; +"58 conv2d_3_zero_point_0" -> "60 dequantize_per_channel_default_3" [label="(128,)", style=solid]; +"59 _frozen_param3" -> "60 dequantize_per_channel_default_3" [label="(128, 128, 3, 3)", style=solid]; +"60 dequantize_per_channel_default_3" -> "61 conv2d_3" [label="(128, 128, 3, 3)", style=solid]; +"61 conv2d_3" -> "62 relu_3" [label="(1, 128, 106, 106)", style=solid]; +"62 relu_3" -> "63 quantize_per_tensor_default_5" [label="(1, 128, 106, 106)", style=solid]; +"62 relu_3" -> "165 slice_9" [label="(1, 128, 106, 106)", style=solid]; +"63 quantize_per_tensor_default_5" -> "64 dequantize_per_tensor_default_5" [label="(1, 128, 106, 106)", style=solid]; +"64 dequantize_per_tensor_default_5" -> "65 max_pool2d_1" [label="(1, 128, 106, 106)", style=solid]; +"65 max_pool2d_1" -> "66 quantize_per_tensor_default_6" [label="(1, 128, 53, 53)", style=solid]; +"66 quantize_per_tensor_default_6" -> "67 dequantize_per_tensor_default_6" [label="(1, 128, 53, 53)", style=solid]; +"67 dequantize_per_tensor_default_6" -> "72 conv2d_4" [label="(1, 128, 53, 53)", style=solid]; +"68 conv2d_4_scale_0" -> "71 dequantize_per_channel_default_4" [label="(256,)", style=solid]; +"69 conv2d_4_zero_point_0" -> "71 dequantize_per_channel_default_4" [label="(256,)", style=solid]; +"70 _frozen_param4" -> "71 dequantize_per_channel_default_4" [label="(256, 128, 3, 3)", style=solid]; +"71 dequantize_per_channel_default_4" -> "72 conv2d_4" [label="(256, 128, 3, 3)", style=solid]; +"72 conv2d_4" -> "73 relu_4" [label="(1, 256, 51, 51)", style=solid]; +"73 relu_4" -> "74 quantize_per_tensor_default_7" [label="(1, 256, 51, 51)", style=solid]; +"74 quantize_per_tensor_default_7" -> "75 dequantize_per_tensor_default_7" [label="(1, 256, 51, 51)", style=solid]; +"75 dequantize_per_tensor_default_7" -> "80 conv2d_5" [label="(1, 256, 51, 51)", style=solid]; +"76 conv2d_5_scale_0" -> "79 dequantize_per_channel_default_5" [label="(256,)", style=solid]; +"77 conv2d_5_zero_point_0" -> "79 dequantize_per_channel_default_5" [label="(256,)", style=solid]; +"78 _frozen_param5" -> "79 dequantize_per_channel_default_5" [label="(256, 256, 3, 3)", style=solid]; +"79 dequantize_per_channel_default_5" -> "80 conv2d_5" [label="(256, 256, 3, 3)", style=solid]; +"80 conv2d_5" -> "81 relu_5" [label="(1, 256, 49, 49)", style=solid]; +"81 relu_5" -> "82 quantize_per_tensor_default_8" [label="(1, 256, 49, 49)", style=solid]; +"81 relu_5" -> "143 slice_5" [label="(1, 256, 49, 49)", style=solid]; +"82 quantize_per_tensor_default_8" -> "83 dequantize_per_tensor_default_8" [label="(1, 256, 49, 49)", style=solid]; +"83 dequantize_per_tensor_default_8" -> "84 max_pool2d_2" [label="(1, 256, 49, 49)", style=solid]; +"84 max_pool2d_2" -> "85 quantize_per_tensor_default_9" [label="(1, 256, 24, 24)", style=solid]; +"85 quantize_per_tensor_default_9" -> "86 dequantize_per_tensor_default_9" [label="(1, 256, 24, 24)", style=solid]; +"86 dequantize_per_tensor_default_9" -> "91 conv2d_6" [label="(1, 256, 24, 24)", style=solid]; +"87 conv2d_6_scale_0" -> "90 dequantize_per_channel_default_6" [label="(512,)", style=solid]; +"88 conv2d_6_zero_point_0" -> "90 dequantize_per_channel_default_6" [label="(512,)", style=solid]; +"89 _frozen_param6" -> "90 dequantize_per_channel_default_6" [label="(512, 256, 3, 3)", style=solid]; +"90 dequantize_per_channel_default_6" -> "91 conv2d_6" [label="(512, 256, 3, 3)", style=solid]; +"91 conv2d_6" -> "92 relu_6" [label="(1, 512, 22, 22)", style=solid]; +"92 relu_6" -> "93 quantize_per_tensor_default_10" [label="(1, 512, 22, 22)", style=solid]; +"93 quantize_per_tensor_default_10" -> "94 dequantize_per_tensor_default_10" [label="(1, 512, 22, 22)", style=solid]; +"94 dequantize_per_tensor_default_10" -> "99 conv2d_7" [label="(1, 512, 22, 22)", style=solid]; +"95 conv2d_7_scale_0" -> "98 dequantize_per_channel_default_7" [label="(512,)", style=solid]; +"96 conv2d_7_zero_point_0" -> "98 dequantize_per_channel_default_7" [label="(512,)", style=solid]; +"97 _frozen_param7" -> "98 dequantize_per_channel_default_7" [label="(512, 512, 3, 3)", style=solid]; +"98 dequantize_per_channel_default_7" -> "99 conv2d_7" [label="(512, 512, 3, 3)", style=solid]; +"99 conv2d_7" -> "100 relu_7" [label="(1, 512, 20, 20)", style=solid]; +"100 relu_7" -> "101 quantize_per_tensor_default_11" [label="(1, 512, 20, 20)", style=solid]; +"100 relu_7" -> "121 slice_1" [label="(1, 512, 20, 20)", style=solid]; +"101 quantize_per_tensor_default_11" -> "102 dequantize_per_tensor_default_11" [label="(1, 512, 20, 20)", style=solid]; +"102 dequantize_per_tensor_default_11" -> "103 max_pool2d_3" [label="(1, 512, 20, 20)", style=solid]; +"103 max_pool2d_3" -> "104 quantize_per_tensor_default_12" [label="(1, 512, 10, 10)", style=solid]; +"104 quantize_per_tensor_default_12" -> "105 dequantize_per_tensor_default_12" [label="(1, 512, 10, 10)", style=solid]; +"105 dequantize_per_tensor_default_12" -> "110 conv2d_8" [label="(1, 512, 10, 10)", style=solid]; +"106 conv2d_8_scale_0" -> "109 dequantize_per_channel_default_8" [label="(1024,)", style=solid]; +"107 conv2d_8_zero_point_0" -> "109 dequantize_per_channel_default_8" [label="(1024,)", style=solid]; +"108 _frozen_param8" -> "109 dequantize_per_channel_default_8" [label="(1024, 512, 3, 3)", style=solid]; +"109 dequantize_per_channel_default_8" -> "110 conv2d_8" [label="(1024, 512, 3, 3)", style=solid]; +"110 conv2d_8" -> "111 relu_8" [label="(1, 1024, 8, 8)", style=solid]; +"111 relu_8" -> "112 quantize_per_tensor_default_13" [label="(1, 1024, 8, 8)", style=solid]; +"112 quantize_per_tensor_default_13" -> "113 dequantize_per_tensor_default_13" [label="(1, 1024, 8, 8)", style=solid]; +"113 dequantize_per_tensor_default_13" -> "118 conv2d_9" [label="(1, 1024, 8, 8)", style=solid]; +"114 conv2d_9_scale_0" -> "117 dequantize_per_channel_default_9" [label="(1024,)", style=solid]; +"115 conv2d_9_zero_point_0" -> "117 dequantize_per_channel_default_9" [label="(1024,)", style=solid]; +"116 _frozen_param9" -> "117 dequantize_per_channel_default_9" [label="(1024, 1024, 3, 3)", style=solid]; +"117 dequantize_per_channel_default_9" -> "118 conv2d_9" [label="(1024, 1024, 3, 3)", style=solid]; +"118 conv2d_9" -> "119 relu_9" [label="(1, 1024, 6, 6)", style=solid]; +"119 relu_9" -> "120 conv_transpose2d" [label="(1, 1024, 6, 6)", style=solid]; +"120 conv_transpose2d" -> "125 cat" [label="(1, 512, 12, 12)", style=solid]; +"121 slice_1" -> "122 slice_2" [label="(1, 512, 20, 20)", style=solid]; +"122 slice_2" -> "123 slice_3" [label="(1, 512, 20, 20)", style=solid]; +"123 slice_3" -> "124 slice_4" [label="(1, 512, 12, 20)", style=solid]; +"124 slice_4" -> "125 cat" [label="(1, 512, 12, 12)", style=solid]; +"125 cat" -> "126 quantize_per_tensor_default_14" [label="(1, 1024, 12, 12)", style=solid]; +"126 quantize_per_tensor_default_14" -> "127 dequantize_per_tensor_default_14" [label="(1, 1024, 12, 12)", style=solid]; +"127 dequantize_per_tensor_default_14" -> "132 conv2d_10" [label="(1, 1024, 12, 12)", style=solid]; +"128 conv2d_10_scale_0" -> "131 dequantize_per_channel_default_10" [label="(512,)", style=solid]; +"129 conv2d_10_zero_point_0" -> "131 dequantize_per_channel_default_10" [label="(512,)", style=solid]; +"130 _frozen_param10" -> "131 dequantize_per_channel_default_10" [label="(512, 1024, 3, 3)", style=solid]; +"131 dequantize_per_channel_default_10" -> "132 conv2d_10" [label="(512, 1024, 3, 3)", style=solid]; +"132 conv2d_10" -> "133 relu_10" [label="(1, 512, 10, 10)", style=solid]; +"133 relu_10" -> "134 quantize_per_tensor_default_15" [label="(1, 512, 10, 10)", style=solid]; +"134 quantize_per_tensor_default_15" -> "135 dequantize_per_tensor_default_15" [label="(1, 512, 10, 10)", style=solid]; +"135 dequantize_per_tensor_default_15" -> "140 conv2d_11" [label="(1, 512, 10, 10)", style=solid]; +"136 conv2d_11_scale_0" -> "139 dequantize_per_channel_default_11" [label="(512,)", style=solid]; +"137 conv2d_11_zero_point_0" -> "139 dequantize_per_channel_default_11" [label="(512,)", style=solid]; +"138 _frozen_param11" -> "139 dequantize_per_channel_default_11" [label="(512, 512, 3, 3)", style=solid]; +"139 dequantize_per_channel_default_11" -> "140 conv2d_11" [label="(512, 512, 3, 3)", style=solid]; +"140 conv2d_11" -> "141 relu_11" [label="(1, 512, 8, 8)", style=solid]; +"141 relu_11" -> "142 conv_transpose2d_1" [label="(1, 512, 8, 8)", style=solid]; +"142 conv_transpose2d_1" -> "147 cat_1" [label="(1, 256, 16, 16)", style=solid]; +"143 slice_5" -> "144 slice_6" [label="(1, 256, 49, 49)", style=solid]; +"144 slice_6" -> "145 slice_7" [label="(1, 256, 49, 49)", style=solid]; +"145 slice_7" -> "146 slice_8" [label="(1, 256, 16, 49)", style=solid]; +"146 slice_8" -> "147 cat_1" [label="(1, 256, 16, 16)", style=solid]; +"147 cat_1" -> "148 quantize_per_tensor_default_16" [label="(1, 512, 16, 16)", style=solid]; +"148 quantize_per_tensor_default_16" -> "149 dequantize_per_tensor_default_16" [label="(1, 512, 16, 16)", style=solid]; +"149 dequantize_per_tensor_default_16" -> "154 conv2d_12" [label="(1, 512, 16, 16)", style=solid]; +"150 conv2d_12_scale_0" -> "153 dequantize_per_channel_default_12" [label="(256,)", style=solid]; +"151 conv2d_12_zero_point_0" -> "153 dequantize_per_channel_default_12" [label="(256,)", style=solid]; +"152 _frozen_param12" -> "153 dequantize_per_channel_default_12" [label="(256, 512, 3, 3)", style=solid]; +"153 dequantize_per_channel_default_12" -> "154 conv2d_12" [label="(256, 512, 3, 3)", style=solid]; +"154 conv2d_12" -> "155 relu_12" [label="(1, 256, 14, 14)", style=solid]; +"155 relu_12" -> "156 quantize_per_tensor_default_17" [label="(1, 256, 14, 14)", style=solid]; +"156 quantize_per_tensor_default_17" -> "157 dequantize_per_tensor_default_17" [label="(1, 256, 14, 14)", style=solid]; +"157 dequantize_per_tensor_default_17" -> "162 conv2d_13" [label="(1, 256, 14, 14)", style=solid]; +"158 conv2d_13_scale_0" -> "161 dequantize_per_channel_default_13" [label="(256,)", style=solid]; +"159 conv2d_13_zero_point_0" -> "161 dequantize_per_channel_default_13" [label="(256,)", style=solid]; +"160 _frozen_param13" -> "161 dequantize_per_channel_default_13" [label="(256, 256, 3, 3)", style=solid]; +"161 dequantize_per_channel_default_13" -> "162 conv2d_13" [label="(256, 256, 3, 3)", style=solid]; +"162 conv2d_13" -> "163 relu_13" [label="(1, 256, 12, 12)", style=solid]; +"163 relu_13" -> "164 conv_transpose2d_2" [label="(1, 256, 12, 12)", style=solid]; +"164 conv_transpose2d_2" -> "169 cat_2" [label="(1, 128, 24, 24)", style=solid]; +"165 slice_9" -> "166 slice_10" [label="(1, 128, 106, 106)", style=solid]; +"166 slice_10" -> "167 slice_11" [label="(1, 128, 106, 106)", style=solid]; +"167 slice_11" -> "168 slice_12" [label="(1, 128, 24, 106)", style=solid]; +"168 slice_12" -> "169 cat_2" [label="(1, 128, 24, 24)", style=solid]; +"169 cat_2" -> "170 quantize_per_tensor_default_18" [label="(1, 256, 24, 24)", style=solid]; +"170 quantize_per_tensor_default_18" -> "171 dequantize_per_tensor_default_18" [label="(1, 256, 24, 24)", style=solid]; +"171 dequantize_per_tensor_default_18" -> "176 conv2d_14" [label="(1, 256, 24, 24)", style=solid]; +"172 conv2d_14_scale_0" -> "175 dequantize_per_channel_default_14" [label="(128,)", style=solid]; +"173 conv2d_14_zero_point_0" -> "175 dequantize_per_channel_default_14" [label="(128,)", style=solid]; +"174 _frozen_param14" -> "175 dequantize_per_channel_default_14" [label="(128, 256, 3, 3)", style=solid]; +"175 dequantize_per_channel_default_14" -> "176 conv2d_14" [label="(128, 256, 3, 3)", style=solid]; +"176 conv2d_14" -> "177 relu_14" [label="(1, 128, 22, 22)", style=solid]; +"177 relu_14" -> "178 quantize_per_tensor_default_19" [label="(1, 128, 22, 22)", style=solid]; +"178 quantize_per_tensor_default_19" -> "179 dequantize_per_tensor_default_19" [label="(1, 128, 22, 22)", style=solid]; +"179 dequantize_per_tensor_default_19" -> "184 conv2d_15" [label="(1, 128, 22, 22)", style=solid]; +"180 conv2d_15_scale_0" -> "183 dequantize_per_channel_default_15" [label="(128,)", style=solid]; +"181 conv2d_15_zero_point_0" -> "183 dequantize_per_channel_default_15" [label="(128,)", style=solid]; +"182 _frozen_param15" -> "183 dequantize_per_channel_default_15" [label="(128, 128, 3, 3)", style=solid]; +"183 dequantize_per_channel_default_15" -> "184 conv2d_15" [label="(128, 128, 3, 3)", style=solid]; +"184 conv2d_15" -> "185 relu_15" [label="(1, 128, 20, 20)", style=solid]; +"185 relu_15" -> "186 conv_transpose2d_3" [label="(1, 128, 20, 20)", style=solid]; +"186 conv_transpose2d_3" -> "191 cat_3" [label="(1, 64, 40, 40)", style=solid]; +"187 slice_13" -> "188 slice_14" [label="(1, 64, 220, 220)", style=solid]; +"188 slice_14" -> "189 slice_15" [label="(1, 64, 220, 220)", style=solid]; +"189 slice_15" -> "190 slice_16" [label="(1, 64, 40, 220)", style=solid]; +"190 slice_16" -> "191 cat_3" [label="(1, 64, 40, 40)", style=solid]; +"191 cat_3" -> "192 quantize_per_tensor_default_20" [label="(1, 128, 40, 40)", style=solid]; +"192 quantize_per_tensor_default_20" -> "193 dequantize_per_tensor_default_20" [label="(1, 128, 40, 40)", style=solid]; +"193 dequantize_per_tensor_default_20" -> "198 conv2d_16" [label="(1, 128, 40, 40)", style=solid]; +"194 conv2d_16_scale_0" -> "197 dequantize_per_channel_default_16" [label="(64,)", style=solid]; +"195 conv2d_16_zero_point_0" -> "197 dequantize_per_channel_default_16" [label="(64,)", style=solid]; +"196 _frozen_param16" -> "197 dequantize_per_channel_default_16" [label="(64, 128, 3, 3)", style=solid]; +"197 dequantize_per_channel_default_16" -> "198 conv2d_16" [label="(64, 128, 3, 3)", style=solid]; +"198 conv2d_16" -> "199 relu_16" [label="(1, 64, 38, 38)", style=solid]; +"199 relu_16" -> "200 quantize_per_tensor_default_21" [label="(1, 64, 38, 38)", style=solid]; +"200 quantize_per_tensor_default_21" -> "201 dequantize_per_tensor_default_21" [label="(1, 64, 38, 38)", style=solid]; +"201 dequantize_per_tensor_default_21" -> "206 conv2d_17" [label="(1, 64, 38, 38)", style=solid]; +"202 conv2d_17_scale_0" -> "205 dequantize_per_channel_default_17" [label="(64,)", style=solid]; +"203 conv2d_17_zero_point_0" -> "205 dequantize_per_channel_default_17" [label="(64,)", style=solid]; +"204 _frozen_param17" -> "205 dequantize_per_channel_default_17" [label="(64, 64, 3, 3)", style=solid]; +"205 dequantize_per_channel_default_17" -> "206 conv2d_17" [label="(64, 64, 3, 3)", style=solid]; +"206 conv2d_17" -> "207 relu_17" [label="(1, 64, 36, 36)", style=solid]; +"207 relu_17" -> "208 quantize_per_tensor_default_22" [label="(1, 64, 36, 36)", style=solid]; +"208 quantize_per_tensor_default_22" -> "209 dequantize_per_tensor_default_22" [label="(1, 64, 36, 36)", style=solid]; +"209 dequantize_per_tensor_default_22" -> "214 conv2d_18" [label="(1, 64, 36, 36)", style=solid]; +"210 conv2d_18_scale_0" -> "213 dequantize_per_channel_default_18" [label="(12,)", style=solid]; +"211 conv2d_18_zero_point_0" -> "213 dequantize_per_channel_default_18" [label="(12,)", style=solid]; +"212 _frozen_param18" -> "213 dequantize_per_channel_default_18" [label="(12, 64, 1, 1)", style=solid]; +"213 dequantize_per_channel_default_18" -> "214 conv2d_18" [label="(12, 64, 1, 1)", style=solid]; +"214 conv2d_18" -> "215 output" [label="(1, 12, 36, 36)", style=solid]; +} diff --git a/tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/vit_b_16.dot b/tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/vit_b_16.dot new file mode 100644 index 00000000000..ef109b4e373 --- /dev/null +++ b/tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/vit_b_16.dot @@ -0,0 +1,1577 @@ +strict digraph { +"0 class_token" [id=0, type=get_attr]; +"1 encoder_pos_embedding" [id=1, type=get_attr]; +"2 conv_proj_bias" [id=2, type=get_attr]; +"3 encoder_layers_encoder_layer_0_ln_1_weight" [id=3, type=get_attr]; +"4 encoder_layers_encoder_layer_0_ln_1_bias" [id=4, type=get_attr]; +"5 encoder_layers_encoder_layer_0_self_attention_in_proj_bias" [id=5, type=get_attr]; +"6 encoder_layers_encoder_layer_0_self_attention_out_proj_bias" [id=6, type=get_attr]; +"7 encoder_layers_encoder_layer_0_ln_2_weight" [id=7, type=get_attr]; +"8 encoder_layers_encoder_layer_0_ln_2_bias" [id=8, type=get_attr]; +"9 encoder_layers_encoder_layer_0_mlp_0_bias" [id=9, type=get_attr]; +"10 encoder_layers_encoder_layer_0_mlp_3_bias" [id=10, type=get_attr]; +"11 encoder_layers_encoder_layer_1_ln_1_weight" [id=11, type=get_attr]; +"12 encoder_layers_encoder_layer_1_ln_1_bias" [id=12, type=get_attr]; +"13 encoder_layers_encoder_layer_1_self_attention_in_proj_bias" [id=13, type=get_attr]; +"14 encoder_layers_encoder_layer_1_self_attention_out_proj_bias" [id=14, type=get_attr]; +"15 encoder_layers_encoder_layer_1_ln_2_weight" [id=15, type=get_attr]; +"16 encoder_layers_encoder_layer_1_ln_2_bias" [id=16, type=get_attr]; +"17 encoder_layers_encoder_layer_1_mlp_0_bias" [id=17, type=get_attr]; +"18 encoder_layers_encoder_layer_1_mlp_3_bias" [id=18, type=get_attr]; +"19 encoder_layers_encoder_layer_2_ln_1_weight" [id=19, type=get_attr]; +"20 encoder_layers_encoder_layer_2_ln_1_bias" [id=20, type=get_attr]; +"21 encoder_layers_encoder_layer_2_self_attention_in_proj_bias" [id=21, type=get_attr]; +"22 encoder_layers_encoder_layer_2_self_attention_out_proj_bias" [id=22, type=get_attr]; +"23 encoder_layers_encoder_layer_2_ln_2_weight" [id=23, type=get_attr]; +"24 encoder_layers_encoder_layer_2_ln_2_bias" [id=24, type=get_attr]; +"25 encoder_layers_encoder_layer_2_mlp_0_bias" [id=25, type=get_attr]; +"26 encoder_layers_encoder_layer_2_mlp_3_bias" [id=26, type=get_attr]; +"27 encoder_layers_encoder_layer_3_ln_1_weight" [id=27, type=get_attr]; +"28 encoder_layers_encoder_layer_3_ln_1_bias" [id=28, type=get_attr]; +"29 encoder_layers_encoder_layer_3_self_attention_in_proj_bias" [id=29, type=get_attr]; +"30 encoder_layers_encoder_layer_3_self_attention_out_proj_bias" [id=30, type=get_attr]; +"31 encoder_layers_encoder_layer_3_ln_2_weight" [id=31, type=get_attr]; +"32 encoder_layers_encoder_layer_3_ln_2_bias" [id=32, type=get_attr]; +"33 encoder_layers_encoder_layer_3_mlp_0_bias" [id=33, type=get_attr]; +"34 encoder_layers_encoder_layer_3_mlp_3_bias" [id=34, type=get_attr]; +"35 encoder_layers_encoder_layer_4_ln_1_weight" [id=35, type=get_attr]; +"36 encoder_layers_encoder_layer_4_ln_1_bias" [id=36, type=get_attr]; +"37 encoder_layers_encoder_layer_4_self_attention_in_proj_bias" [id=37, type=get_attr]; +"38 encoder_layers_encoder_layer_4_self_attention_out_proj_bias" [id=38, type=get_attr]; +"39 encoder_layers_encoder_layer_4_ln_2_weight" [id=39, type=get_attr]; +"40 encoder_layers_encoder_layer_4_ln_2_bias" [id=40, type=get_attr]; +"41 encoder_layers_encoder_layer_4_mlp_0_bias" [id=41, type=get_attr]; +"42 encoder_layers_encoder_layer_4_mlp_3_bias" [id=42, type=get_attr]; +"43 encoder_layers_encoder_layer_5_ln_1_weight" [id=43, type=get_attr]; +"44 encoder_layers_encoder_layer_5_ln_1_bias" [id=44, type=get_attr]; +"45 encoder_layers_encoder_layer_5_self_attention_in_proj_bias" [id=45, type=get_attr]; +"46 encoder_layers_encoder_layer_5_self_attention_out_proj_bias" [id=46, type=get_attr]; +"47 encoder_layers_encoder_layer_5_ln_2_weight" [id=47, type=get_attr]; +"48 encoder_layers_encoder_layer_5_ln_2_bias" [id=48, type=get_attr]; +"49 encoder_layers_encoder_layer_5_mlp_0_bias" [id=49, type=get_attr]; +"50 encoder_layers_encoder_layer_5_mlp_3_bias" [id=50, type=get_attr]; +"51 encoder_layers_encoder_layer_6_ln_1_weight" [id=51, type=get_attr]; +"52 encoder_layers_encoder_layer_6_ln_1_bias" [id=52, type=get_attr]; +"53 encoder_layers_encoder_layer_6_self_attention_in_proj_bias" [id=53, type=get_attr]; +"54 encoder_layers_encoder_layer_6_self_attention_out_proj_bias" [id=54, type=get_attr]; +"55 encoder_layers_encoder_layer_6_ln_2_weight" [id=55, type=get_attr]; +"56 encoder_layers_encoder_layer_6_ln_2_bias" [id=56, type=get_attr]; +"57 encoder_layers_encoder_layer_6_mlp_0_bias" [id=57, type=get_attr]; +"58 encoder_layers_encoder_layer_6_mlp_3_bias" [id=58, type=get_attr]; +"59 encoder_layers_encoder_layer_7_ln_1_weight" [id=59, type=get_attr]; +"60 encoder_layers_encoder_layer_7_ln_1_bias" [id=60, type=get_attr]; +"61 encoder_layers_encoder_layer_7_self_attention_in_proj_bias" [id=61, type=get_attr]; +"62 encoder_layers_encoder_layer_7_self_attention_out_proj_bias" [id=62, type=get_attr]; +"63 encoder_layers_encoder_layer_7_ln_2_weight" [id=63, type=get_attr]; +"64 encoder_layers_encoder_layer_7_ln_2_bias" [id=64, type=get_attr]; +"65 encoder_layers_encoder_layer_7_mlp_0_bias" [id=65, type=get_attr]; +"66 encoder_layers_encoder_layer_7_mlp_3_bias" [id=66, type=get_attr]; +"67 encoder_layers_encoder_layer_8_ln_1_weight" [id=67, type=get_attr]; +"68 encoder_layers_encoder_layer_8_ln_1_bias" [id=68, type=get_attr]; +"69 encoder_layers_encoder_layer_8_self_attention_in_proj_bias" [id=69, type=get_attr]; +"70 encoder_layers_encoder_layer_8_self_attention_out_proj_bias" [id=70, type=get_attr]; +"71 encoder_layers_encoder_layer_8_ln_2_weight" [id=71, type=get_attr]; +"72 encoder_layers_encoder_layer_8_ln_2_bias" [id=72, type=get_attr]; +"73 encoder_layers_encoder_layer_8_mlp_0_bias" [id=73, type=get_attr]; +"74 encoder_layers_encoder_layer_8_mlp_3_bias" [id=74, type=get_attr]; +"75 encoder_layers_encoder_layer_9_ln_1_weight" [id=75, type=get_attr]; +"76 encoder_layers_encoder_layer_9_ln_1_bias" [id=76, type=get_attr]; +"77 encoder_layers_encoder_layer_9_self_attention_in_proj_bias" [id=77, type=get_attr]; +"78 encoder_layers_encoder_layer_9_self_attention_out_proj_bias" [id=78, type=get_attr]; +"79 encoder_layers_encoder_layer_9_ln_2_weight" [id=79, type=get_attr]; +"80 encoder_layers_encoder_layer_9_ln_2_bias" [id=80, type=get_attr]; +"81 encoder_layers_encoder_layer_9_mlp_0_bias" [id=81, type=get_attr]; +"82 encoder_layers_encoder_layer_9_mlp_3_bias" [id=82, type=get_attr]; +"83 encoder_layers_encoder_layer_10_ln_1_weight" [id=83, type=get_attr]; +"84 encoder_layers_encoder_layer_10_ln_1_bias" [id=84, type=get_attr]; +"85 encoder_layers_encoder_layer_10_self_attention_in_proj_bias" [id=85, type=get_attr]; +"86 encoder_layers_encoder_layer_10_self_attention_out_proj_bias" [id=86, type=get_attr]; +"87 encoder_layers_encoder_layer_10_ln_2_weight" [id=87, type=get_attr]; +"88 encoder_layers_encoder_layer_10_ln_2_bias" [id=88, type=get_attr]; +"89 encoder_layers_encoder_layer_10_mlp_0_bias" [id=89, type=get_attr]; +"90 encoder_layers_encoder_layer_10_mlp_3_bias" [id=90, type=get_attr]; +"91 encoder_layers_encoder_layer_11_ln_1_weight" [id=91, type=get_attr]; +"92 encoder_layers_encoder_layer_11_ln_1_bias" [id=92, type=get_attr]; +"93 encoder_layers_encoder_layer_11_self_attention_in_proj_bias" [id=93, type=get_attr]; +"94 encoder_layers_encoder_layer_11_self_attention_out_proj_bias" [id=94, type=get_attr]; +"95 encoder_layers_encoder_layer_11_ln_2_weight" [id=95, type=get_attr]; +"96 encoder_layers_encoder_layer_11_ln_2_bias" [id=96, type=get_attr]; +"97 encoder_layers_encoder_layer_11_mlp_0_bias" [id=97, type=get_attr]; +"98 encoder_layers_encoder_layer_11_mlp_3_bias" [id=98, type=get_attr]; +"99 encoder_ln_weight" [id=99, type=get_attr]; +"100 encoder_ln_bias" [id=100, type=get_attr]; +"101 heads_head_bias" [id=101, type=get_attr]; +"102 x" [id=102, type=input]; +"103 quantize_per_tensor_default" [id=103, type=quantize_per_tensor]; +"104 dequantize_per_tensor_default" [id=104, type=dequantize_per_tensor]; +"105 conv2d_scale_0" [id=105, type=get_attr]; +"106 conv2d_zero_point_0" [id=106, type=get_attr]; +"107 _frozen_param0" [id=107, type=get_attr]; +"108 dequantize_per_channel_default" [id=108, type=dequantize_per_channel]; +"109 conv2d" [id=109, type=conv2d]; +"110 reshape" [id=110, type=reshape]; +"111 permute" [id=111, type=permute]; +"112 expand" [id=112, type=expand]; +"113 cat" [id=113, type=cat]; +"114 add" [id=114, type=add]; +"115 dropout" [id=115, type=dropout]; +"116 layer_norm" [id=116, type=layer_norm]; +"117 transpose" [id=117, type=transpose]; +"118 encoder_layers_encoder_layer_0_self_attention_in_proj_weight_updated_constant0" [id=118, type=get_attr]; +"119 transpose_0_0_nncf_smooth_quant_0" [id=119, type=call_module]; +"120 linear" [id=120, type=linear]; +"121 unflatten" [id=121, type=unflatten]; +"122 unsqueeze" [id=122, type=unsqueeze]; +"123 transpose_1" [id=123, type=transpose]; +"124 squeeze" [id=124, type=squeeze]; +"125 contiguous" [id=125, type=contiguous]; +"126 select" [id=126, type=select]; +"127 select_1" [id=127, type=select]; +"128 select_2" [id=128, type=select]; +"129 view" [id=129, type=view]; +"130 transpose_2" [id=130, type=transpose]; +"131 view_1" [id=131, type=view]; +"132 transpose_3" [id=132, type=transpose]; +"133 view_2" [id=133, type=view]; +"134 transpose_4" [id=134, type=transpose]; +"135 view_3" [id=135, type=view]; +"136 view_4" [id=136, type=view]; +"137 view_5" [id=137, type=view]; +"138 scaled_dot_product_attention" [id=138, type=scaled_dot_product_attention]; +"139 permute_1" [id=139, type=permute]; +"140 view_6" [id=140, type=view]; +"141 encoder_layers_encoder_layer_0_self_attention_out_proj_weight_updated_constant0" [id=141, type=get_attr]; +"142 view_6_0_0_nncf_smooth_quant_0" [id=142, type=call_module]; +"143 linear_1" [id=143, type=linear]; +"144 view_7" [id=144, type=view]; +"145 transpose_5" [id=145, type=transpose]; +"146 dropout_1" [id=146, type=dropout]; +"147 add_1" [id=147, type=add]; +"148 layer_norm_1" [id=148, type=layer_norm]; +"149 layer_norm_1_0_0_nncf_smooth_quant_0" [id=149, type=call_module]; +"150 quantize_per_tensor_default_1" [id=150, type=quantize_per_tensor]; +"151 dequantize_per_tensor_default_1" [id=151, type=dequantize_per_tensor]; +"152 linear_2_scale_0" [id=152, type=get_attr]; +"153 linear_2_zero_point_0" [id=153, type=get_attr]; +"154 _frozen_param1" [id=154, type=get_attr]; +"155 dequantize_per_channel_default_1" [id=155, type=dequantize_per_channel]; +"156 linear_2" [id=156, type=linear]; +"157 gelu" [id=157, type=gelu]; +"158 dropout_2" [id=158, type=dropout]; +"159 dropout_2_0_0_nncf_smooth_quant_0" [id=159, type=call_module]; +"160 quantize_per_tensor_default_2" [id=160, type=quantize_per_tensor]; +"161 dequantize_per_tensor_default_2" [id=161, type=dequantize_per_tensor]; +"162 linear_3_scale_0" [id=162, type=get_attr]; +"163 linear_3_zero_point_0" [id=163, type=get_attr]; +"164 _frozen_param2" [id=164, type=get_attr]; +"165 dequantize_per_channel_default_2" [id=165, type=dequantize_per_channel]; +"166 linear_3" [id=166, type=linear]; +"167 dropout_3" [id=167, type=dropout]; +"168 add_2" [id=168, type=add]; +"169 layer_norm_2" [id=169, type=layer_norm]; +"170 transpose_6" [id=170, type=transpose]; +"171 encoder_layers_encoder_layer_1_self_attention_in_proj_weight_updated_constant0" [id=171, type=get_attr]; +"172 transpose_6_0_0_nncf_smooth_quant_0" [id=172, type=call_module]; +"173 linear_4" [id=173, type=linear]; +"174 unflatten_1" [id=174, type=unflatten]; +"175 unsqueeze_1" [id=175, type=unsqueeze]; +"176 transpose_7" [id=176, type=transpose]; +"177 squeeze_1" [id=177, type=squeeze]; +"178 contiguous_1" [id=178, type=contiguous]; +"179 select_3" [id=179, type=select]; +"180 select_4" [id=180, type=select]; +"181 select_5" [id=181, type=select]; +"182 view_8" [id=182, type=view]; +"183 transpose_8" [id=183, type=transpose]; +"184 view_9" [id=184, type=view]; +"185 transpose_9" [id=185, type=transpose]; +"186 view_10" [id=186, type=view]; +"187 transpose_10" [id=187, type=transpose]; +"188 view_11" [id=188, type=view]; +"189 view_12" [id=189, type=view]; +"190 view_13" [id=190, type=view]; +"191 scaled_dot_product_attention_1" [id=191, type=scaled_dot_product_attention]; +"192 permute_2" [id=192, type=permute]; +"193 view_14" [id=193, type=view]; +"194 encoder_layers_encoder_layer_1_self_attention_out_proj_weight_updated_constant0" [id=194, type=get_attr]; +"195 view_14_0_0_nncf_smooth_quant_0" [id=195, type=call_module]; +"196 linear_5" [id=196, type=linear]; +"197 view_15" [id=197, type=view]; +"198 transpose_11" [id=198, type=transpose]; +"199 dropout_4" [id=199, type=dropout]; +"200 add_3" [id=200, type=add]; +"201 layer_norm_3" [id=201, type=layer_norm]; +"202 layer_norm_3_0_0_nncf_smooth_quant_0" [id=202, type=call_module]; +"203 quantize_per_tensor_default_3" [id=203, type=quantize_per_tensor]; +"204 dequantize_per_tensor_default_3" [id=204, type=dequantize_per_tensor]; +"205 linear_6_scale_0" [id=205, type=get_attr]; +"206 linear_6_zero_point_0" [id=206, type=get_attr]; +"207 _frozen_param3" [id=207, type=get_attr]; +"208 dequantize_per_channel_default_3" [id=208, type=dequantize_per_channel]; +"209 linear_6" [id=209, type=linear]; +"210 gelu_1" [id=210, type=gelu]; +"211 dropout_5" [id=211, type=dropout]; +"212 dropout_5_0_0_nncf_smooth_quant_0" [id=212, type=call_module]; +"213 quantize_per_tensor_default_4" [id=213, type=quantize_per_tensor]; +"214 dequantize_per_tensor_default_4" [id=214, type=dequantize_per_tensor]; +"215 linear_7_scale_0" [id=215, type=get_attr]; +"216 linear_7_zero_point_0" [id=216, type=get_attr]; +"217 _frozen_param4" [id=217, type=get_attr]; +"218 dequantize_per_channel_default_4" [id=218, type=dequantize_per_channel]; +"219 linear_7" [id=219, type=linear]; +"220 dropout_6" [id=220, type=dropout]; +"221 add_4" [id=221, type=add]; +"222 layer_norm_4" [id=222, type=layer_norm]; +"223 transpose_12" [id=223, type=transpose]; +"224 encoder_layers_encoder_layer_2_self_attention_in_proj_weight_updated_constant0" [id=224, type=get_attr]; +"225 transpose_12_0_0_nncf_smooth_quant_0" [id=225, type=call_module]; +"226 linear_8" [id=226, type=linear]; +"227 unflatten_2" [id=227, type=unflatten]; +"228 unsqueeze_2" [id=228, type=unsqueeze]; +"229 transpose_13" [id=229, type=transpose]; +"230 squeeze_2" [id=230, type=squeeze]; +"231 contiguous_2" [id=231, type=contiguous]; +"232 select_6" [id=232, type=select]; +"233 select_7" [id=233, type=select]; +"234 select_8" [id=234, type=select]; +"235 view_16" [id=235, type=view]; +"236 transpose_14" [id=236, type=transpose]; +"237 view_17" [id=237, type=view]; +"238 transpose_15" [id=238, type=transpose]; +"239 view_18" [id=239, type=view]; +"240 transpose_16" [id=240, type=transpose]; +"241 view_19" [id=241, type=view]; +"242 view_20" [id=242, type=view]; +"243 view_21" [id=243, type=view]; +"244 scaled_dot_product_attention_2" [id=244, type=scaled_dot_product_attention]; +"245 permute_3" [id=245, type=permute]; +"246 view_22" [id=246, type=view]; +"247 encoder_layers_encoder_layer_2_self_attention_out_proj_weight_updated_constant0" [id=247, type=get_attr]; +"248 view_22_0_0_nncf_smooth_quant_0" [id=248, type=call_module]; +"249 linear_9" [id=249, type=linear]; +"250 view_23" [id=250, type=view]; +"251 transpose_17" [id=251, type=transpose]; +"252 dropout_7" [id=252, type=dropout]; +"253 add_5" [id=253, type=add]; +"254 layer_norm_5" [id=254, type=layer_norm]; +"255 layer_norm_5_0_0_nncf_smooth_quant_0" [id=255, type=call_module]; +"256 quantize_per_tensor_default_5" [id=256, type=quantize_per_tensor]; +"257 dequantize_per_tensor_default_5" [id=257, type=dequantize_per_tensor]; +"258 linear_10_scale_0" [id=258, type=get_attr]; +"259 linear_10_zero_point_0" [id=259, type=get_attr]; +"260 _frozen_param5" [id=260, type=get_attr]; +"261 dequantize_per_channel_default_5" [id=261, type=dequantize_per_channel]; +"262 linear_10" [id=262, type=linear]; +"263 gelu_2" [id=263, type=gelu]; +"264 dropout_8" [id=264, type=dropout]; +"265 dropout_8_0_0_nncf_smooth_quant_0" [id=265, type=call_module]; +"266 quantize_per_tensor_default_6" [id=266, type=quantize_per_tensor]; +"267 dequantize_per_tensor_default_6" [id=267, type=dequantize_per_tensor]; +"268 linear_11_scale_0" [id=268, type=get_attr]; +"269 linear_11_zero_point_0" [id=269, type=get_attr]; +"270 _frozen_param6" [id=270, type=get_attr]; +"271 dequantize_per_channel_default_6" [id=271, type=dequantize_per_channel]; +"272 linear_11" [id=272, type=linear]; +"273 dropout_9" [id=273, type=dropout]; +"274 add_6" [id=274, type=add]; +"275 layer_norm_6" [id=275, type=layer_norm]; +"276 transpose_18" [id=276, type=transpose]; +"277 encoder_layers_encoder_layer_3_self_attention_in_proj_weight_updated_constant0" [id=277, type=get_attr]; +"278 transpose_18_0_0_nncf_smooth_quant_0" [id=278, type=call_module]; +"279 linear_12" [id=279, type=linear]; +"280 unflatten_3" [id=280, type=unflatten]; +"281 unsqueeze_3" [id=281, type=unsqueeze]; +"282 transpose_19" [id=282, type=transpose]; +"283 squeeze_3" [id=283, type=squeeze]; +"284 contiguous_3" [id=284, type=contiguous]; +"285 select_9" [id=285, type=select]; +"286 select_10" [id=286, type=select]; +"287 select_11" [id=287, type=select]; +"288 view_24" [id=288, type=view]; +"289 transpose_20" [id=289, type=transpose]; +"290 view_25" [id=290, type=view]; +"291 transpose_21" [id=291, type=transpose]; +"292 view_26" [id=292, type=view]; +"293 transpose_22" [id=293, type=transpose]; +"294 view_27" [id=294, type=view]; +"295 view_28" [id=295, type=view]; +"296 view_29" [id=296, type=view]; +"297 scaled_dot_product_attention_3" [id=297, type=scaled_dot_product_attention]; +"298 permute_4" [id=298, type=permute]; +"299 view_30" [id=299, type=view]; +"300 encoder_layers_encoder_layer_3_self_attention_out_proj_weight_updated_constant0" [id=300, type=get_attr]; +"301 view_30_0_0_nncf_smooth_quant_0" [id=301, type=call_module]; +"302 linear_13" [id=302, type=linear]; +"303 view_31" [id=303, type=view]; +"304 transpose_23" [id=304, type=transpose]; +"305 dropout_10" [id=305, type=dropout]; +"306 add_7" [id=306, type=add]; +"307 layer_norm_7" [id=307, type=layer_norm]; +"308 layer_norm_7_0_0_nncf_smooth_quant_0" [id=308, type=call_module]; +"309 quantize_per_tensor_default_7" [id=309, type=quantize_per_tensor]; +"310 dequantize_per_tensor_default_7" [id=310, type=dequantize_per_tensor]; +"311 linear_14_scale_0" [id=311, type=get_attr]; +"312 linear_14_zero_point_0" [id=312, type=get_attr]; +"313 _frozen_param7" [id=313, type=get_attr]; +"314 dequantize_per_channel_default_7" [id=314, type=dequantize_per_channel]; +"315 linear_14" [id=315, type=linear]; +"316 gelu_3" [id=316, type=gelu]; +"317 dropout_11" [id=317, type=dropout]; +"318 dropout_11_0_0_nncf_smooth_quant_0" [id=318, type=call_module]; +"319 quantize_per_tensor_default_8" [id=319, type=quantize_per_tensor]; +"320 dequantize_per_tensor_default_8" [id=320, type=dequantize_per_tensor]; +"321 linear_15_scale_0" [id=321, type=get_attr]; +"322 linear_15_zero_point_0" [id=322, type=get_attr]; +"323 _frozen_param8" [id=323, type=get_attr]; +"324 dequantize_per_channel_default_8" [id=324, type=dequantize_per_channel]; +"325 linear_15" [id=325, type=linear]; +"326 dropout_12" [id=326, type=dropout]; +"327 add_8" [id=327, type=add]; +"328 layer_norm_8" [id=328, type=layer_norm]; +"329 transpose_24" [id=329, type=transpose]; +"330 encoder_layers_encoder_layer_4_self_attention_in_proj_weight_updated_constant0" [id=330, type=get_attr]; +"331 transpose_24_0_0_nncf_smooth_quant_0" [id=331, type=call_module]; +"332 linear_16" [id=332, type=linear]; +"333 unflatten_4" [id=333, type=unflatten]; +"334 unsqueeze_4" [id=334, type=unsqueeze]; +"335 transpose_25" [id=335, type=transpose]; +"336 squeeze_4" [id=336, type=squeeze]; +"337 contiguous_4" [id=337, type=contiguous]; +"338 select_12" [id=338, type=select]; +"339 select_13" [id=339, type=select]; +"340 select_14" [id=340, type=select]; +"341 view_32" [id=341, type=view]; +"342 transpose_26" [id=342, type=transpose]; +"343 view_33" [id=343, type=view]; +"344 transpose_27" [id=344, type=transpose]; +"345 view_34" [id=345, type=view]; +"346 transpose_28" [id=346, type=transpose]; +"347 view_35" [id=347, type=view]; +"348 view_36" [id=348, type=view]; +"349 view_37" [id=349, type=view]; +"350 scaled_dot_product_attention_4" [id=350, type=scaled_dot_product_attention]; +"351 permute_5" [id=351, type=permute]; +"352 view_38" [id=352, type=view]; +"353 encoder_layers_encoder_layer_4_self_attention_out_proj_weight_updated_constant0" [id=353, type=get_attr]; +"354 view_38_0_0_nncf_smooth_quant_0" [id=354, type=call_module]; +"355 linear_17" [id=355, type=linear]; +"356 view_39" [id=356, type=view]; +"357 transpose_29" [id=357, type=transpose]; +"358 dropout_13" [id=358, type=dropout]; +"359 add_9" [id=359, type=add]; +"360 layer_norm_9" [id=360, type=layer_norm]; +"361 layer_norm_9_0_0_nncf_smooth_quant_0" [id=361, type=call_module]; +"362 quantize_per_tensor_default_9" [id=362, type=quantize_per_tensor]; +"363 dequantize_per_tensor_default_9" [id=363, type=dequantize_per_tensor]; +"364 linear_18_scale_0" [id=364, type=get_attr]; +"365 linear_18_zero_point_0" [id=365, type=get_attr]; +"366 _frozen_param9" [id=366, type=get_attr]; +"367 dequantize_per_channel_default_9" [id=367, type=dequantize_per_channel]; +"368 linear_18" [id=368, type=linear]; +"369 gelu_4" [id=369, type=gelu]; +"370 dropout_14" [id=370, type=dropout]; +"371 dropout_14_0_0_nncf_smooth_quant_0" [id=371, type=call_module]; +"372 quantize_per_tensor_default_10" [id=372, type=quantize_per_tensor]; +"373 dequantize_per_tensor_default_10" [id=373, type=dequantize_per_tensor]; +"374 linear_19_scale_0" [id=374, type=get_attr]; +"375 linear_19_zero_point_0" [id=375, type=get_attr]; +"376 _frozen_param10" [id=376, type=get_attr]; +"377 dequantize_per_channel_default_10" [id=377, type=dequantize_per_channel]; +"378 linear_19" [id=378, type=linear]; +"379 dropout_15" [id=379, type=dropout]; +"380 add_10" [id=380, type=add]; +"381 layer_norm_10" [id=381, type=layer_norm]; +"382 transpose_30" [id=382, type=transpose]; +"383 encoder_layers_encoder_layer_5_self_attention_in_proj_weight_updated_constant0" [id=383, type=get_attr]; +"384 transpose_30_0_0_nncf_smooth_quant_0" [id=384, type=call_module]; +"385 linear_20" [id=385, type=linear]; +"386 unflatten_5" [id=386, type=unflatten]; +"387 unsqueeze_5" [id=387, type=unsqueeze]; +"388 transpose_31" [id=388, type=transpose]; +"389 squeeze_5" [id=389, type=squeeze]; +"390 contiguous_5" [id=390, type=contiguous]; +"391 select_15" [id=391, type=select]; +"392 select_16" [id=392, type=select]; +"393 select_17" [id=393, type=select]; +"394 view_40" [id=394, type=view]; +"395 transpose_32" [id=395, type=transpose]; +"396 view_41" [id=396, type=view]; +"397 transpose_33" [id=397, type=transpose]; +"398 view_42" [id=398, type=view]; +"399 transpose_34" [id=399, type=transpose]; +"400 view_43" [id=400, type=view]; +"401 view_44" [id=401, type=view]; +"402 view_45" [id=402, type=view]; +"403 scaled_dot_product_attention_5" [id=403, type=scaled_dot_product_attention]; +"404 permute_6" [id=404, type=permute]; +"405 view_46" [id=405, type=view]; +"406 encoder_layers_encoder_layer_5_self_attention_out_proj_weight_updated_constant0" [id=406, type=get_attr]; +"407 view_46_0_0_nncf_smooth_quant_0" [id=407, type=call_module]; +"408 linear_21" [id=408, type=linear]; +"409 view_47" [id=409, type=view]; +"410 transpose_35" [id=410, type=transpose]; +"411 dropout_16" [id=411, type=dropout]; +"412 add_11" [id=412, type=add]; +"413 layer_norm_11" [id=413, type=layer_norm]; +"414 layer_norm_11_0_0_nncf_smooth_quant_0" [id=414, type=call_module]; +"415 quantize_per_tensor_default_11" [id=415, type=quantize_per_tensor]; +"416 dequantize_per_tensor_default_11" [id=416, type=dequantize_per_tensor]; +"417 linear_22_scale_0" [id=417, type=get_attr]; +"418 linear_22_zero_point_0" [id=418, type=get_attr]; +"419 _frozen_param11" [id=419, type=get_attr]; +"420 dequantize_per_channel_default_11" [id=420, type=dequantize_per_channel]; +"421 linear_22" [id=421, type=linear]; +"422 gelu_5" [id=422, type=gelu]; +"423 dropout_17" [id=423, type=dropout]; +"424 dropout_17_0_0_nncf_smooth_quant_0" [id=424, type=call_module]; +"425 quantize_per_tensor_default_12" [id=425, type=quantize_per_tensor]; +"426 dequantize_per_tensor_default_12" [id=426, type=dequantize_per_tensor]; +"427 linear_23_scale_0" [id=427, type=get_attr]; +"428 linear_23_zero_point_0" [id=428, type=get_attr]; +"429 _frozen_param12" [id=429, type=get_attr]; +"430 dequantize_per_channel_default_12" [id=430, type=dequantize_per_channel]; +"431 linear_23" [id=431, type=linear]; +"432 dropout_18" [id=432, type=dropout]; +"433 add_12" [id=433, type=add]; +"434 layer_norm_12" [id=434, type=layer_norm]; +"435 transpose_36" [id=435, type=transpose]; +"436 encoder_layers_encoder_layer_6_self_attention_in_proj_weight_updated_constant0" [id=436, type=get_attr]; +"437 transpose_36_0_0_nncf_smooth_quant_0" [id=437, type=call_module]; +"438 linear_24" [id=438, type=linear]; +"439 unflatten_6" [id=439, type=unflatten]; +"440 unsqueeze_6" [id=440, type=unsqueeze]; +"441 transpose_37" [id=441, type=transpose]; +"442 squeeze_6" [id=442, type=squeeze]; +"443 contiguous_6" [id=443, type=contiguous]; +"444 select_18" [id=444, type=select]; +"445 select_19" [id=445, type=select]; +"446 select_20" [id=446, type=select]; +"447 view_48" [id=447, type=view]; +"448 transpose_38" [id=448, type=transpose]; +"449 view_49" [id=449, type=view]; +"450 transpose_39" [id=450, type=transpose]; +"451 view_50" [id=451, type=view]; +"452 transpose_40" [id=452, type=transpose]; +"453 view_51" [id=453, type=view]; +"454 view_52" [id=454, type=view]; +"455 view_53" [id=455, type=view]; +"456 scaled_dot_product_attention_6" [id=456, type=scaled_dot_product_attention]; +"457 permute_7" [id=457, type=permute]; +"458 view_54" [id=458, type=view]; +"459 encoder_layers_encoder_layer_6_self_attention_out_proj_weight_updated_constant0" [id=459, type=get_attr]; +"460 view_54_0_0_nncf_smooth_quant_0" [id=460, type=call_module]; +"461 linear_25" [id=461, type=linear]; +"462 view_55" [id=462, type=view]; +"463 transpose_41" [id=463, type=transpose]; +"464 dropout_19" [id=464, type=dropout]; +"465 add_13" [id=465, type=add]; +"466 layer_norm_13" [id=466, type=layer_norm]; +"467 layer_norm_13_0_0_nncf_smooth_quant_0" [id=467, type=call_module]; +"468 quantize_per_tensor_default_13" [id=468, type=quantize_per_tensor]; +"469 dequantize_per_tensor_default_13" [id=469, type=dequantize_per_tensor]; +"470 linear_26_scale_0" [id=470, type=get_attr]; +"471 linear_26_zero_point_0" [id=471, type=get_attr]; +"472 _frozen_param13" [id=472, type=get_attr]; +"473 dequantize_per_channel_default_13" [id=473, type=dequantize_per_channel]; +"474 linear_26" [id=474, type=linear]; +"475 gelu_6" [id=475, type=gelu]; +"476 dropout_20" [id=476, type=dropout]; +"477 dropout_20_0_0_nncf_smooth_quant_0" [id=477, type=call_module]; +"478 quantize_per_tensor_default_14" [id=478, type=quantize_per_tensor]; +"479 dequantize_per_tensor_default_14" [id=479, type=dequantize_per_tensor]; +"480 linear_27_scale_0" [id=480, type=get_attr]; +"481 linear_27_zero_point_0" [id=481, type=get_attr]; +"482 _frozen_param14" [id=482, type=get_attr]; +"483 dequantize_per_channel_default_14" [id=483, type=dequantize_per_channel]; +"484 linear_27" [id=484, type=linear]; +"485 dropout_21" [id=485, type=dropout]; +"486 add_14" [id=486, type=add]; +"487 layer_norm_14" [id=487, type=layer_norm]; +"488 transpose_42" [id=488, type=transpose]; +"489 encoder_layers_encoder_layer_7_self_attention_in_proj_weight_updated_constant0" [id=489, type=get_attr]; +"490 transpose_42_0_0_nncf_smooth_quant_0" [id=490, type=call_module]; +"491 linear_28" [id=491, type=linear]; +"492 unflatten_7" [id=492, type=unflatten]; +"493 unsqueeze_7" [id=493, type=unsqueeze]; +"494 transpose_43" [id=494, type=transpose]; +"495 squeeze_7" [id=495, type=squeeze]; +"496 contiguous_7" [id=496, type=contiguous]; +"497 select_21" [id=497, type=select]; +"498 select_22" [id=498, type=select]; +"499 select_23" [id=499, type=select]; +"500 view_56" [id=500, type=view]; +"501 transpose_44" [id=501, type=transpose]; +"502 view_57" [id=502, type=view]; +"503 transpose_45" [id=503, type=transpose]; +"504 view_58" [id=504, type=view]; +"505 transpose_46" [id=505, type=transpose]; +"506 view_59" [id=506, type=view]; +"507 view_60" [id=507, type=view]; +"508 view_61" [id=508, type=view]; +"509 scaled_dot_product_attention_7" [id=509, type=scaled_dot_product_attention]; +"510 permute_8" [id=510, type=permute]; +"511 view_62" [id=511, type=view]; +"512 encoder_layers_encoder_layer_7_self_attention_out_proj_weight_updated_constant0" [id=512, type=get_attr]; +"513 view_62_0_0_nncf_smooth_quant_0" [id=513, type=call_module]; +"514 linear_29" [id=514, type=linear]; +"515 view_63" [id=515, type=view]; +"516 transpose_47" [id=516, type=transpose]; +"517 dropout_22" [id=517, type=dropout]; +"518 add_15" [id=518, type=add]; +"519 layer_norm_15" [id=519, type=layer_norm]; +"520 layer_norm_15_0_0_nncf_smooth_quant_0" [id=520, type=call_module]; +"521 quantize_per_tensor_default_15" [id=521, type=quantize_per_tensor]; +"522 dequantize_per_tensor_default_15" [id=522, type=dequantize_per_tensor]; +"523 linear_30_scale_0" [id=523, type=get_attr]; +"524 linear_30_zero_point_0" [id=524, type=get_attr]; +"525 _frozen_param15" [id=525, type=get_attr]; +"526 dequantize_per_channel_default_15" [id=526, type=dequantize_per_channel]; +"527 linear_30" [id=527, type=linear]; +"528 gelu_7" [id=528, type=gelu]; +"529 dropout_23" [id=529, type=dropout]; +"530 dropout_23_0_0_nncf_smooth_quant_0" [id=530, type=call_module]; +"531 quantize_per_tensor_default_16" [id=531, type=quantize_per_tensor]; +"532 dequantize_per_tensor_default_16" [id=532, type=dequantize_per_tensor]; +"533 linear_31_scale_0" [id=533, type=get_attr]; +"534 linear_31_zero_point_0" [id=534, type=get_attr]; +"535 _frozen_param16" [id=535, type=get_attr]; +"536 dequantize_per_channel_default_16" [id=536, type=dequantize_per_channel]; +"537 linear_31" [id=537, type=linear]; +"538 dropout_24" [id=538, type=dropout]; +"539 add_16" [id=539, type=add]; +"540 layer_norm_16" [id=540, type=layer_norm]; +"541 transpose_48" [id=541, type=transpose]; +"542 encoder_layers_encoder_layer_8_self_attention_in_proj_weight_updated_constant0" [id=542, type=get_attr]; +"543 transpose_48_0_0_nncf_smooth_quant_0" [id=543, type=call_module]; +"544 linear_32" [id=544, type=linear]; +"545 unflatten_8" [id=545, type=unflatten]; +"546 unsqueeze_8" [id=546, type=unsqueeze]; +"547 transpose_49" [id=547, type=transpose]; +"548 squeeze_8" [id=548, type=squeeze]; +"549 contiguous_8" [id=549, type=contiguous]; +"550 select_24" [id=550, type=select]; +"551 select_25" [id=551, type=select]; +"552 select_26" [id=552, type=select]; +"553 view_64" [id=553, type=view]; +"554 transpose_50" [id=554, type=transpose]; +"555 view_65" [id=555, type=view]; +"556 transpose_51" [id=556, type=transpose]; +"557 view_66" [id=557, type=view]; +"558 transpose_52" [id=558, type=transpose]; +"559 view_67" [id=559, type=view]; +"560 view_68" [id=560, type=view]; +"561 view_69" [id=561, type=view]; +"562 scaled_dot_product_attention_8" [id=562, type=scaled_dot_product_attention]; +"563 permute_9" [id=563, type=permute]; +"564 view_70" [id=564, type=view]; +"565 encoder_layers_encoder_layer_8_self_attention_out_proj_weight_updated_constant0" [id=565, type=get_attr]; +"566 view_70_0_0_nncf_smooth_quant_0" [id=566, type=call_module]; +"567 linear_33" [id=567, type=linear]; +"568 view_71" [id=568, type=view]; +"569 transpose_53" [id=569, type=transpose]; +"570 dropout_25" [id=570, type=dropout]; +"571 add_17" [id=571, type=add]; +"572 layer_norm_17" [id=572, type=layer_norm]; +"573 layer_norm_17_0_0_nncf_smooth_quant_0" [id=573, type=call_module]; +"574 quantize_per_tensor_default_17" [id=574, type=quantize_per_tensor]; +"575 dequantize_per_tensor_default_17" [id=575, type=dequantize_per_tensor]; +"576 linear_34_scale_0" [id=576, type=get_attr]; +"577 linear_34_zero_point_0" [id=577, type=get_attr]; +"578 _frozen_param17" [id=578, type=get_attr]; +"579 dequantize_per_channel_default_17" [id=579, type=dequantize_per_channel]; +"580 linear_34" [id=580, type=linear]; +"581 gelu_8" [id=581, type=gelu]; +"582 dropout_26" [id=582, type=dropout]; +"583 dropout_26_0_0_nncf_smooth_quant_0" [id=583, type=call_module]; +"584 quantize_per_tensor_default_18" [id=584, type=quantize_per_tensor]; +"585 dequantize_per_tensor_default_18" [id=585, type=dequantize_per_tensor]; +"586 linear_35_scale_0" [id=586, type=get_attr]; +"587 linear_35_zero_point_0" [id=587, type=get_attr]; +"588 _frozen_param18" [id=588, type=get_attr]; +"589 dequantize_per_channel_default_18" [id=589, type=dequantize_per_channel]; +"590 linear_35" [id=590, type=linear]; +"591 dropout_27" [id=591, type=dropout]; +"592 add_18" [id=592, type=add]; +"593 layer_norm_18" [id=593, type=layer_norm]; +"594 transpose_54" [id=594, type=transpose]; +"595 encoder_layers_encoder_layer_9_self_attention_in_proj_weight_updated_constant0" [id=595, type=get_attr]; +"596 transpose_54_0_0_nncf_smooth_quant_0" [id=596, type=call_module]; +"597 linear_36" [id=597, type=linear]; +"598 unflatten_9" [id=598, type=unflatten]; +"599 unsqueeze_9" [id=599, type=unsqueeze]; +"600 transpose_55" [id=600, type=transpose]; +"601 squeeze_9" [id=601, type=squeeze]; +"602 contiguous_9" [id=602, type=contiguous]; +"603 select_27" [id=603, type=select]; +"604 select_28" [id=604, type=select]; +"605 select_29" [id=605, type=select]; +"606 view_72" [id=606, type=view]; +"607 transpose_56" [id=607, type=transpose]; +"608 view_73" [id=608, type=view]; +"609 transpose_57" [id=609, type=transpose]; +"610 view_74" [id=610, type=view]; +"611 transpose_58" [id=611, type=transpose]; +"612 view_75" [id=612, type=view]; +"613 view_76" [id=613, type=view]; +"614 view_77" [id=614, type=view]; +"615 scaled_dot_product_attention_9" [id=615, type=scaled_dot_product_attention]; +"616 permute_10" [id=616, type=permute]; +"617 view_78" [id=617, type=view]; +"618 encoder_layers_encoder_layer_9_self_attention_out_proj_weight_updated_constant0" [id=618, type=get_attr]; +"619 view_78_0_0_nncf_smooth_quant_0" [id=619, type=call_module]; +"620 linear_37" [id=620, type=linear]; +"621 view_79" [id=621, type=view]; +"622 transpose_59" [id=622, type=transpose]; +"623 dropout_28" [id=623, type=dropout]; +"624 add_19" [id=624, type=add]; +"625 layer_norm_19" [id=625, type=layer_norm]; +"626 layer_norm_19_0_0_nncf_smooth_quant_0" [id=626, type=call_module]; +"627 quantize_per_tensor_default_19" [id=627, type=quantize_per_tensor]; +"628 dequantize_per_tensor_default_19" [id=628, type=dequantize_per_tensor]; +"629 linear_38_scale_0" [id=629, type=get_attr]; +"630 linear_38_zero_point_0" [id=630, type=get_attr]; +"631 _frozen_param19" [id=631, type=get_attr]; +"632 dequantize_per_channel_default_19" [id=632, type=dequantize_per_channel]; +"633 linear_38" [id=633, type=linear]; +"634 gelu_9" [id=634, type=gelu]; +"635 dropout_29" [id=635, type=dropout]; +"636 dropout_29_0_0_nncf_smooth_quant_0" [id=636, type=call_module]; +"637 quantize_per_tensor_default_20" [id=637, type=quantize_per_tensor]; +"638 dequantize_per_tensor_default_20" [id=638, type=dequantize_per_tensor]; +"639 linear_39_scale_0" [id=639, type=get_attr]; +"640 linear_39_zero_point_0" [id=640, type=get_attr]; +"641 _frozen_param20" [id=641, type=get_attr]; +"642 dequantize_per_channel_default_20" [id=642, type=dequantize_per_channel]; +"643 linear_39" [id=643, type=linear]; +"644 dropout_30" [id=644, type=dropout]; +"645 add_20" [id=645, type=add]; +"646 layer_norm_20" [id=646, type=layer_norm]; +"647 transpose_60" [id=647, type=transpose]; +"648 encoder_layers_encoder_layer_10_self_attention_in_proj_weight_updated_constant0" [id=648, type=get_attr]; +"649 transpose_60_0_0_nncf_smooth_quant_0" [id=649, type=call_module]; +"650 linear_40" [id=650, type=linear]; +"651 unflatten_10" [id=651, type=unflatten]; +"652 unsqueeze_10" [id=652, type=unsqueeze]; +"653 transpose_61" [id=653, type=transpose]; +"654 squeeze_10" [id=654, type=squeeze]; +"655 contiguous_10" [id=655, type=contiguous]; +"656 select_30" [id=656, type=select]; +"657 select_31" [id=657, type=select]; +"658 select_32" [id=658, type=select]; +"659 view_80" [id=659, type=view]; +"660 transpose_62" [id=660, type=transpose]; +"661 view_81" [id=661, type=view]; +"662 transpose_63" [id=662, type=transpose]; +"663 view_82" [id=663, type=view]; +"664 transpose_64" [id=664, type=transpose]; +"665 view_83" [id=665, type=view]; +"666 view_84" [id=666, type=view]; +"667 view_85" [id=667, type=view]; +"668 scaled_dot_product_attention_10" [id=668, type=scaled_dot_product_attention]; +"669 permute_11" [id=669, type=permute]; +"670 view_86" [id=670, type=view]; +"671 encoder_layers_encoder_layer_10_self_attention_out_proj_weight_updated_constant0" [id=671, type=get_attr]; +"672 view_86_0_0_nncf_smooth_quant_0" [id=672, type=call_module]; +"673 linear_41" [id=673, type=linear]; +"674 view_87" [id=674, type=view]; +"675 transpose_65" [id=675, type=transpose]; +"676 dropout_31" [id=676, type=dropout]; +"677 add_21" [id=677, type=add]; +"678 layer_norm_21" [id=678, type=layer_norm]; +"679 layer_norm_21_0_0_nncf_smooth_quant_0" [id=679, type=call_module]; +"680 quantize_per_tensor_default_21" [id=680, type=quantize_per_tensor]; +"681 dequantize_per_tensor_default_21" [id=681, type=dequantize_per_tensor]; +"682 linear_42_scale_0" [id=682, type=get_attr]; +"683 linear_42_zero_point_0" [id=683, type=get_attr]; +"684 _frozen_param21" [id=684, type=get_attr]; +"685 dequantize_per_channel_default_21" [id=685, type=dequantize_per_channel]; +"686 linear_42" [id=686, type=linear]; +"687 gelu_10" [id=687, type=gelu]; +"688 dropout_32" [id=688, type=dropout]; +"689 dropout_32_0_0_nncf_smooth_quant_0" [id=689, type=call_module]; +"690 quantize_per_tensor_default_22" [id=690, type=quantize_per_tensor]; +"691 dequantize_per_tensor_default_22" [id=691, type=dequantize_per_tensor]; +"692 linear_43_scale_0" [id=692, type=get_attr]; +"693 linear_43_zero_point_0" [id=693, type=get_attr]; +"694 _frozen_param22" [id=694, type=get_attr]; +"695 dequantize_per_channel_default_22" [id=695, type=dequantize_per_channel]; +"696 linear_43" [id=696, type=linear]; +"697 dropout_33" [id=697, type=dropout]; +"698 add_22" [id=698, type=add]; +"699 layer_norm_22" [id=699, type=layer_norm]; +"700 transpose_66" [id=700, type=transpose]; +"701 encoder_layers_encoder_layer_11_self_attention_in_proj_weight_updated_constant0" [id=701, type=get_attr]; +"702 transpose_66_0_0_nncf_smooth_quant_0" [id=702, type=call_module]; +"703 linear_44" [id=703, type=linear]; +"704 unflatten_11" [id=704, type=unflatten]; +"705 unsqueeze_11" [id=705, type=unsqueeze]; +"706 transpose_67" [id=706, type=transpose]; +"707 squeeze_11" [id=707, type=squeeze]; +"708 contiguous_11" [id=708, type=contiguous]; +"709 select_33" [id=709, type=select]; +"710 select_34" [id=710, type=select]; +"711 select_35" [id=711, type=select]; +"712 view_88" [id=712, type=view]; +"713 transpose_68" [id=713, type=transpose]; +"714 view_89" [id=714, type=view]; +"715 transpose_69" [id=715, type=transpose]; +"716 view_90" [id=716, type=view]; +"717 transpose_70" [id=717, type=transpose]; +"718 view_91" [id=718, type=view]; +"719 view_92" [id=719, type=view]; +"720 view_93" [id=720, type=view]; +"721 scaled_dot_product_attention_11" [id=721, type=scaled_dot_product_attention]; +"722 permute_12" [id=722, type=permute]; +"723 view_94" [id=723, type=view]; +"724 encoder_layers_encoder_layer_11_self_attention_out_proj_weight_updated_constant0" [id=724, type=get_attr]; +"725 view_94_0_0_nncf_smooth_quant_0" [id=725, type=call_module]; +"726 linear_45" [id=726, type=linear]; +"727 view_95" [id=727, type=view]; +"728 transpose_71" [id=728, type=transpose]; +"729 dropout_34" [id=729, type=dropout]; +"730 add_23" [id=730, type=add]; +"731 layer_norm_23" [id=731, type=layer_norm]; +"732 layer_norm_23_0_0_nncf_smooth_quant_0" [id=732, type=call_module]; +"733 quantize_per_tensor_default_23" [id=733, type=quantize_per_tensor]; +"734 dequantize_per_tensor_default_23" [id=734, type=dequantize_per_tensor]; +"735 linear_46_scale_0" [id=735, type=get_attr]; +"736 linear_46_zero_point_0" [id=736, type=get_attr]; +"737 _frozen_param23" [id=737, type=get_attr]; +"738 dequantize_per_channel_default_23" [id=738, type=dequantize_per_channel]; +"739 linear_46" [id=739, type=linear]; +"740 gelu_11" [id=740, type=gelu]; +"741 dropout_35" [id=741, type=dropout]; +"742 dropout_35_0_0_nncf_smooth_quant_0" [id=742, type=call_module]; +"743 quantize_per_tensor_default_24" [id=743, type=quantize_per_tensor]; +"744 dequantize_per_tensor_default_24" [id=744, type=dequantize_per_tensor]; +"745 linear_47_scale_0" [id=745, type=get_attr]; +"746 linear_47_zero_point_0" [id=746, type=get_attr]; +"747 _frozen_param24" [id=747, type=get_attr]; +"748 dequantize_per_channel_default_24" [id=748, type=dequantize_per_channel]; +"749 linear_47" [id=749, type=linear]; +"750 dropout_36" [id=750, type=dropout]; +"751 add_24" [id=751, type=add]; +"752 layer_norm_24" [id=752, type=layer_norm]; +"753 slice_1" [id=753, type=slice]; +"754 select_36" [id=754, type=select]; +"755 select_36_0_0_nncf_smooth_quant_0" [id=755, type=call_module]; +"756 quantize_per_tensor_default_25" [id=756, type=quantize_per_tensor]; +"757 dequantize_per_tensor_default_25" [id=757, type=dequantize_per_tensor]; +"758 linear_48_scale_0" [id=758, type=get_attr]; +"759 linear_48_zero_point_0" [id=759, type=get_attr]; +"760 _frozen_param25" [id=760, type=get_attr]; +"761 dequantize_per_channel_default_25" [id=761, type=dequantize_per_channel]; +"762 linear_48" [id=762, type=linear]; +"763 output" [id=763, type=output]; +"0 class_token" -> "112 expand" [label="(1, 1, 768)", style=solid]; +"1 encoder_pos_embedding" -> "114 add" [label="(1, 197, 768)", style=solid]; +"2 conv_proj_bias" -> "109 conv2d" [label="(768,)", style=solid]; +"3 encoder_layers_encoder_layer_0_ln_1_weight" -> "116 layer_norm" [label="(768,)", style=solid]; +"4 encoder_layers_encoder_layer_0_ln_1_bias" -> "116 layer_norm" [label="(768,)", style=solid]; +"5 encoder_layers_encoder_layer_0_self_attention_in_proj_bias" -> "120 linear" [label="(2304,)", style=solid]; +"6 encoder_layers_encoder_layer_0_self_attention_out_proj_bias" -> "143 linear_1" [label="(768,)", style=solid]; +"7 encoder_layers_encoder_layer_0_ln_2_weight" -> "148 layer_norm_1" [label="(768,)", style=solid]; +"8 encoder_layers_encoder_layer_0_ln_2_bias" -> "148 layer_norm_1" [label="(768,)", style=solid]; +"9 encoder_layers_encoder_layer_0_mlp_0_bias" -> "156 linear_2" [label="(3072,)", style=solid]; +"10 encoder_layers_encoder_layer_0_mlp_3_bias" -> "166 linear_3" [label="(768,)", style=solid]; +"11 encoder_layers_encoder_layer_1_ln_1_weight" -> "169 layer_norm_2" [label="(768,)", style=solid]; +"12 encoder_layers_encoder_layer_1_ln_1_bias" -> "169 layer_norm_2" [label="(768,)", style=solid]; +"13 encoder_layers_encoder_layer_1_self_attention_in_proj_bias" -> "173 linear_4" [label="(2304,)", style=solid]; +"14 encoder_layers_encoder_layer_1_self_attention_out_proj_bias" -> "196 linear_5" [label="(768,)", style=solid]; +"15 encoder_layers_encoder_layer_1_ln_2_weight" -> "201 layer_norm_3" [label="(768,)", style=solid]; +"16 encoder_layers_encoder_layer_1_ln_2_bias" -> "201 layer_norm_3" [label="(768,)", style=solid]; +"17 encoder_layers_encoder_layer_1_mlp_0_bias" -> "209 linear_6" [label="(3072,)", style=solid]; +"18 encoder_layers_encoder_layer_1_mlp_3_bias" -> "219 linear_7" [label="(768,)", style=solid]; +"19 encoder_layers_encoder_layer_2_ln_1_weight" -> "222 layer_norm_4" [label="(768,)", style=solid]; +"20 encoder_layers_encoder_layer_2_ln_1_bias" -> "222 layer_norm_4" [label="(768,)", style=solid]; +"21 encoder_layers_encoder_layer_2_self_attention_in_proj_bias" -> "226 linear_8" [label="(2304,)", style=solid]; +"22 encoder_layers_encoder_layer_2_self_attention_out_proj_bias" -> "249 linear_9" [label="(768,)", style=solid]; +"23 encoder_layers_encoder_layer_2_ln_2_weight" -> "254 layer_norm_5" [label="(768,)", style=solid]; +"24 encoder_layers_encoder_layer_2_ln_2_bias" -> "254 layer_norm_5" [label="(768,)", style=solid]; +"25 encoder_layers_encoder_layer_2_mlp_0_bias" -> "262 linear_10" [label="(3072,)", style=solid]; +"26 encoder_layers_encoder_layer_2_mlp_3_bias" -> "272 linear_11" [label="(768,)", style=solid]; +"27 encoder_layers_encoder_layer_3_ln_1_weight" -> "275 layer_norm_6" [label="(768,)", style=solid]; +"28 encoder_layers_encoder_layer_3_ln_1_bias" -> "275 layer_norm_6" [label="(768,)", style=solid]; +"29 encoder_layers_encoder_layer_3_self_attention_in_proj_bias" -> "279 linear_12" [label="(2304,)", style=solid]; +"30 encoder_layers_encoder_layer_3_self_attention_out_proj_bias" -> "302 linear_13" [label="(768,)", style=solid]; +"31 encoder_layers_encoder_layer_3_ln_2_weight" -> "307 layer_norm_7" [label="(768,)", style=solid]; +"32 encoder_layers_encoder_layer_3_ln_2_bias" -> "307 layer_norm_7" [label="(768,)", style=solid]; +"33 encoder_layers_encoder_layer_3_mlp_0_bias" -> "315 linear_14" [label="(3072,)", style=solid]; +"34 encoder_layers_encoder_layer_3_mlp_3_bias" -> "325 linear_15" [label="(768,)", style=solid]; +"35 encoder_layers_encoder_layer_4_ln_1_weight" -> "328 layer_norm_8" [label="(768,)", style=solid]; +"36 encoder_layers_encoder_layer_4_ln_1_bias" -> "328 layer_norm_8" [label="(768,)", style=solid]; +"37 encoder_layers_encoder_layer_4_self_attention_in_proj_bias" -> "332 linear_16" [label="(2304,)", style=solid]; +"38 encoder_layers_encoder_layer_4_self_attention_out_proj_bias" -> "355 linear_17" [label="(768,)", style=solid]; +"39 encoder_layers_encoder_layer_4_ln_2_weight" -> "360 layer_norm_9" [label="(768,)", style=solid]; +"40 encoder_layers_encoder_layer_4_ln_2_bias" -> "360 layer_norm_9" [label="(768,)", style=solid]; +"41 encoder_layers_encoder_layer_4_mlp_0_bias" -> "368 linear_18" [label="(3072,)", style=solid]; +"42 encoder_layers_encoder_layer_4_mlp_3_bias" -> "378 linear_19" [label="(768,)", style=solid]; +"43 encoder_layers_encoder_layer_5_ln_1_weight" -> "381 layer_norm_10" [label="(768,)", style=solid]; +"44 encoder_layers_encoder_layer_5_ln_1_bias" -> "381 layer_norm_10" [label="(768,)", style=solid]; +"45 encoder_layers_encoder_layer_5_self_attention_in_proj_bias" -> "385 linear_20" [label="(2304,)", style=solid]; +"46 encoder_layers_encoder_layer_5_self_attention_out_proj_bias" -> "408 linear_21" [label="(768,)", style=solid]; +"47 encoder_layers_encoder_layer_5_ln_2_weight" -> "413 layer_norm_11" [label="(768,)", style=solid]; +"48 encoder_layers_encoder_layer_5_ln_2_bias" -> "413 layer_norm_11" [label="(768,)", style=solid]; +"49 encoder_layers_encoder_layer_5_mlp_0_bias" -> "421 linear_22" [label="(3072,)", style=solid]; +"50 encoder_layers_encoder_layer_5_mlp_3_bias" -> "431 linear_23" [label="(768,)", style=solid]; +"51 encoder_layers_encoder_layer_6_ln_1_weight" -> "434 layer_norm_12" [label="(768,)", style=solid]; +"52 encoder_layers_encoder_layer_6_ln_1_bias" -> "434 layer_norm_12" [label="(768,)", style=solid]; +"53 encoder_layers_encoder_layer_6_self_attention_in_proj_bias" -> "438 linear_24" [label="(2304,)", style=solid]; +"54 encoder_layers_encoder_layer_6_self_attention_out_proj_bias" -> "461 linear_25" [label="(768,)", style=solid]; +"55 encoder_layers_encoder_layer_6_ln_2_weight" -> "466 layer_norm_13" [label="(768,)", style=solid]; +"56 encoder_layers_encoder_layer_6_ln_2_bias" -> "466 layer_norm_13" [label="(768,)", style=solid]; +"57 encoder_layers_encoder_layer_6_mlp_0_bias" -> "474 linear_26" [label="(3072,)", style=solid]; +"58 encoder_layers_encoder_layer_6_mlp_3_bias" -> "484 linear_27" [label="(768,)", style=solid]; +"59 encoder_layers_encoder_layer_7_ln_1_weight" -> "487 layer_norm_14" [label="(768,)", style=solid]; +"60 encoder_layers_encoder_layer_7_ln_1_bias" -> "487 layer_norm_14" [label="(768,)", style=solid]; +"61 encoder_layers_encoder_layer_7_self_attention_in_proj_bias" -> "491 linear_28" [label="(2304,)", style=solid]; +"62 encoder_layers_encoder_layer_7_self_attention_out_proj_bias" -> "514 linear_29" [label="(768,)", style=solid]; +"63 encoder_layers_encoder_layer_7_ln_2_weight" -> "519 layer_norm_15" [label="(768,)", style=solid]; +"64 encoder_layers_encoder_layer_7_ln_2_bias" -> "519 layer_norm_15" [label="(768,)", style=solid]; +"65 encoder_layers_encoder_layer_7_mlp_0_bias" -> "527 linear_30" [label="(3072,)", style=solid]; +"66 encoder_layers_encoder_layer_7_mlp_3_bias" -> "537 linear_31" [label="(768,)", style=solid]; +"67 encoder_layers_encoder_layer_8_ln_1_weight" -> "540 layer_norm_16" [label="(768,)", style=solid]; +"68 encoder_layers_encoder_layer_8_ln_1_bias" -> "540 layer_norm_16" [label="(768,)", style=solid]; +"69 encoder_layers_encoder_layer_8_self_attention_in_proj_bias" -> "544 linear_32" [label="(2304,)", style=solid]; +"70 encoder_layers_encoder_layer_8_self_attention_out_proj_bias" -> "567 linear_33" [label="(768,)", style=solid]; +"71 encoder_layers_encoder_layer_8_ln_2_weight" -> "572 layer_norm_17" [label="(768,)", style=solid]; +"72 encoder_layers_encoder_layer_8_ln_2_bias" -> "572 layer_norm_17" [label="(768,)", style=solid]; +"73 encoder_layers_encoder_layer_8_mlp_0_bias" -> "580 linear_34" [label="(3072,)", style=solid]; +"74 encoder_layers_encoder_layer_8_mlp_3_bias" -> "590 linear_35" [label="(768,)", style=solid]; +"75 encoder_layers_encoder_layer_9_ln_1_weight" -> "593 layer_norm_18" [label="(768,)", style=solid]; +"76 encoder_layers_encoder_layer_9_ln_1_bias" -> "593 layer_norm_18" [label="(768,)", style=solid]; +"77 encoder_layers_encoder_layer_9_self_attention_in_proj_bias" -> "597 linear_36" [label="(2304,)", style=solid]; +"78 encoder_layers_encoder_layer_9_self_attention_out_proj_bias" -> "620 linear_37" [label="(768,)", style=solid]; +"79 encoder_layers_encoder_layer_9_ln_2_weight" -> "625 layer_norm_19" [label="(768,)", style=solid]; +"80 encoder_layers_encoder_layer_9_ln_2_bias" -> "625 layer_norm_19" [label="(768,)", style=solid]; +"81 encoder_layers_encoder_layer_9_mlp_0_bias" -> "633 linear_38" [label="(3072,)", style=solid]; +"82 encoder_layers_encoder_layer_9_mlp_3_bias" -> "643 linear_39" [label="(768,)", style=solid]; +"83 encoder_layers_encoder_layer_10_ln_1_weight" -> "646 layer_norm_20" [label="(768,)", style=solid]; +"84 encoder_layers_encoder_layer_10_ln_1_bias" -> "646 layer_norm_20" [label="(768,)", style=solid]; +"85 encoder_layers_encoder_layer_10_self_attention_in_proj_bias" -> "650 linear_40" [label="(2304,)", style=solid]; +"86 encoder_layers_encoder_layer_10_self_attention_out_proj_bias" -> "673 linear_41" [label="(768,)", style=solid]; +"87 encoder_layers_encoder_layer_10_ln_2_weight" -> "678 layer_norm_21" [label="(768,)", style=solid]; +"88 encoder_layers_encoder_layer_10_ln_2_bias" -> "678 layer_norm_21" [label="(768,)", style=solid]; +"89 encoder_layers_encoder_layer_10_mlp_0_bias" -> "686 linear_42" [label="(3072,)", style=solid]; +"90 encoder_layers_encoder_layer_10_mlp_3_bias" -> "696 linear_43" [label="(768,)", style=solid]; +"91 encoder_layers_encoder_layer_11_ln_1_weight" -> "699 layer_norm_22" [label="(768,)", style=solid]; +"92 encoder_layers_encoder_layer_11_ln_1_bias" -> "699 layer_norm_22" [label="(768,)", style=solid]; +"93 encoder_layers_encoder_layer_11_self_attention_in_proj_bias" -> "703 linear_44" [label="(2304,)", style=solid]; +"94 encoder_layers_encoder_layer_11_self_attention_out_proj_bias" -> "726 linear_45" [label="(768,)", style=solid]; +"95 encoder_layers_encoder_layer_11_ln_2_weight" -> "731 layer_norm_23" [label="(768,)", style=solid]; +"96 encoder_layers_encoder_layer_11_ln_2_bias" -> "731 layer_norm_23" [label="(768,)", style=solid]; +"97 encoder_layers_encoder_layer_11_mlp_0_bias" -> "739 linear_46" [label="(3072,)", style=solid]; +"98 encoder_layers_encoder_layer_11_mlp_3_bias" -> "749 linear_47" [label="(768,)", style=solid]; +"99 encoder_ln_weight" -> "752 layer_norm_24" [label="(768,)", style=solid]; +"100 encoder_ln_bias" -> "752 layer_norm_24" [label="(768,)", style=solid]; +"101 heads_head_bias" -> "762 linear_48" [label="(1000,)", style=solid]; +"102 x" -> "103 quantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; +"103 quantize_per_tensor_default" -> "104 dequantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; +"104 dequantize_per_tensor_default" -> "109 conv2d" [label="(1, 3, 224, 224)", style=solid]; +"105 conv2d_scale_0" -> "108 dequantize_per_channel_default" [label="(768,)", style=solid]; +"106 conv2d_zero_point_0" -> "108 dequantize_per_channel_default" [label="(768,)", style=solid]; +"107 _frozen_param0" -> "108 dequantize_per_channel_default" [label="(768, 3, 16, 16)", style=solid]; +"108 dequantize_per_channel_default" -> "109 conv2d" [label="(768, 3, 16, 16)", style=solid]; +"109 conv2d" -> "110 reshape" [label="(1, 768, 14, 14)", style=solid]; +"110 reshape" -> "111 permute" [label="(1, 768, 196)", style=solid]; +"111 permute" -> "113 cat" [label="(1, 196, 768)", style=solid]; +"112 expand" -> "113 cat" [label="(1, 1, 768)", style=solid]; +"113 cat" -> "114 add" [label="(1, 197, 768)", style=solid]; +"114 add" -> "115 dropout" [label="(1, 197, 768)", style=solid]; +"115 dropout" -> "116 layer_norm" [label="(1, 197, 768)", style=solid]; +"115 dropout" -> "147 add_1" [label="(1, 197, 768)", style=solid]; +"116 layer_norm" -> "117 transpose" [label="(1, 197, 768)", style=solid]; +"117 transpose" -> "119 transpose_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"118 encoder_layers_encoder_layer_0_self_attention_in_proj_weight_updated_constant0" -> "120 linear" [label="(2304, 768)", style=solid]; +"119 transpose_0_0_nncf_smooth_quant_0" -> "120 linear" [label="(197, 1, 768)", style=solid]; +"120 linear" -> "121 unflatten" [label="(197, 1, 2304)", style=solid]; +"121 unflatten" -> "122 unsqueeze" [label="(197, 1, 3, 768)", style=solid]; +"122 unsqueeze" -> "123 transpose_1" [label="(1, 197, 1, 3, 768)", style=solid]; +"123 transpose_1" -> "124 squeeze" [label="(3, 197, 1, 1, 768)", style=solid]; +"124 squeeze" -> "125 contiguous" [label="(3, 197, 1, 768)", style=solid]; +"125 contiguous" -> "126 select" [label="(3, 197, 1, 768)", style=solid]; +"125 contiguous" -> "127 select_1" [label="(3, 197, 1, 768)", style=solid]; +"125 contiguous" -> "128 select_2" [label="(3, 197, 1, 768)", style=solid]; +"126 select" -> "129 view" [label="(197, 1, 768)", style=solid]; +"127 select_1" -> "131 view_1" [label="(197, 1, 768)", style=solid]; +"128 select_2" -> "133 view_2" [label="(197, 1, 768)", style=solid]; +"129 view" -> "130 transpose_2" [label="(197, 12, 64)", style=solid]; +"130 transpose_2" -> "135 view_3" [label="(12, 197, 64)", style=solid]; +"131 view_1" -> "132 transpose_3" [label="(197, 12, 64)", style=solid]; +"132 transpose_3" -> "136 view_4" [label="(12, 197, 64)", style=solid]; +"133 view_2" -> "134 transpose_4" [label="(197, 12, 64)", style=solid]; +"134 transpose_4" -> "137 view_5" [label="(12, 197, 64)", style=solid]; +"135 view_3" -> "138 scaled_dot_product_attention" [label="(1, 12, 197, 64)", style=solid]; +"136 view_4" -> "138 scaled_dot_product_attention" [label="(1, 12, 197, 64)", style=solid]; +"137 view_5" -> "138 scaled_dot_product_attention" [label="(1, 12, 197, 64)", style=solid]; +"138 scaled_dot_product_attention" -> "139 permute_1" [label="(1, 12, 197, 64)", style=solid]; +"139 permute_1" -> "140 view_6" [label="(197, 1, 12, 64)", style=solid]; +"140 view_6" -> "142 view_6_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"141 encoder_layers_encoder_layer_0_self_attention_out_proj_weight_updated_constant0" -> "143 linear_1" [label="(768, 768)", style=solid]; +"142 view_6_0_0_nncf_smooth_quant_0" -> "143 linear_1" [label="(197, 768)", style=solid]; +"143 linear_1" -> "144 view_7" [label="(197, 768)", style=solid]; +"144 view_7" -> "145 transpose_5" [label="(197, 1, 768)", style=solid]; +"145 transpose_5" -> "146 dropout_1" [label="(1, 197, 768)", style=solid]; +"146 dropout_1" -> "147 add_1" [label="(1, 197, 768)", style=solid]; +"147 add_1" -> "148 layer_norm_1" [label="(1, 197, 768)", style=solid]; +"147 add_1" -> "168 add_2" [label="(1, 197, 768)", style=solid]; +"148 layer_norm_1" -> "149 layer_norm_1_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"149 layer_norm_1_0_0_nncf_smooth_quant_0" -> "150 quantize_per_tensor_default_1" [label="(1, 197, 768)", style=solid]; +"150 quantize_per_tensor_default_1" -> "151 dequantize_per_tensor_default_1" [label="(1, 197, 768)", style=solid]; +"151 dequantize_per_tensor_default_1" -> "156 linear_2" [label="(1, 197, 768)", style=solid]; +"152 linear_2_scale_0" -> "155 dequantize_per_channel_default_1" [label="(3072,)", style=solid]; +"153 linear_2_zero_point_0" -> "155 dequantize_per_channel_default_1" [label="(3072,)", style=solid]; +"154 _frozen_param1" -> "155 dequantize_per_channel_default_1" [label="(3072, 768)", style=solid]; +"155 dequantize_per_channel_default_1" -> "156 linear_2" [label="(3072, 768)", style=solid]; +"156 linear_2" -> "157 gelu" [label="(1, 197, 3072)", style=solid]; +"157 gelu" -> "158 dropout_2" [label="(1, 197, 3072)", style=solid]; +"158 dropout_2" -> "159 dropout_2_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"159 dropout_2_0_0_nncf_smooth_quant_0" -> "160 quantize_per_tensor_default_2" [label="(1, 197, 3072)", style=solid]; +"160 quantize_per_tensor_default_2" -> "161 dequantize_per_tensor_default_2" [label="(1, 197, 3072)", style=solid]; +"161 dequantize_per_tensor_default_2" -> "166 linear_3" [label="(1, 197, 3072)", style=solid]; +"162 linear_3_scale_0" -> "165 dequantize_per_channel_default_2" [label="(768,)", style=solid]; +"163 linear_3_zero_point_0" -> "165 dequantize_per_channel_default_2" [label="(768,)", style=solid]; +"164 _frozen_param2" -> "165 dequantize_per_channel_default_2" [label="(768, 3072)", style=solid]; +"165 dequantize_per_channel_default_2" -> "166 linear_3" [label="(768, 3072)", style=solid]; +"166 linear_3" -> "167 dropout_3" [label="(1, 197, 768)", style=solid]; +"167 dropout_3" -> "168 add_2" [label="(1, 197, 768)", style=solid]; +"168 add_2" -> "169 layer_norm_2" [label="(1, 197, 768)", style=solid]; +"168 add_2" -> "200 add_3" [label="(1, 197, 768)", style=solid]; +"169 layer_norm_2" -> "170 transpose_6" [label="(1, 197, 768)", style=solid]; +"170 transpose_6" -> "172 transpose_6_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"171 encoder_layers_encoder_layer_1_self_attention_in_proj_weight_updated_constant0" -> "173 linear_4" [label="(2304, 768)", style=solid]; +"172 transpose_6_0_0_nncf_smooth_quant_0" -> "173 linear_4" [label="(197, 1, 768)", style=solid]; +"173 linear_4" -> "174 unflatten_1" [label="(197, 1, 2304)", style=solid]; +"174 unflatten_1" -> "175 unsqueeze_1" [label="(197, 1, 3, 768)", style=solid]; +"175 unsqueeze_1" -> "176 transpose_7" [label="(1, 197, 1, 3, 768)", style=solid]; +"176 transpose_7" -> "177 squeeze_1" [label="(3, 197, 1, 1, 768)", style=solid]; +"177 squeeze_1" -> "178 contiguous_1" [label="(3, 197, 1, 768)", style=solid]; +"178 contiguous_1" -> "179 select_3" [label="(3, 197, 1, 768)", style=solid]; +"178 contiguous_1" -> "180 select_4" [label="(3, 197, 1, 768)", style=solid]; +"178 contiguous_1" -> "181 select_5" [label="(3, 197, 1, 768)", style=solid]; +"179 select_3" -> "182 view_8" [label="(197, 1, 768)", style=solid]; +"180 select_4" -> "184 view_9" [label="(197, 1, 768)", style=solid]; +"181 select_5" -> "186 view_10" [label="(197, 1, 768)", style=solid]; +"182 view_8" -> "183 transpose_8" [label="(197, 12, 64)", style=solid]; +"183 transpose_8" -> "188 view_11" [label="(12, 197, 64)", style=solid]; +"184 view_9" -> "185 transpose_9" [label="(197, 12, 64)", style=solid]; +"185 transpose_9" -> "189 view_12" [label="(12, 197, 64)", style=solid]; +"186 view_10" -> "187 transpose_10" [label="(197, 12, 64)", style=solid]; +"187 transpose_10" -> "190 view_13" [label="(12, 197, 64)", style=solid]; +"188 view_11" -> "191 scaled_dot_product_attention_1" [label="(1, 12, 197, 64)", style=solid]; +"189 view_12" -> "191 scaled_dot_product_attention_1" [label="(1, 12, 197, 64)", style=solid]; +"190 view_13" -> "191 scaled_dot_product_attention_1" [label="(1, 12, 197, 64)", style=solid]; +"191 scaled_dot_product_attention_1" -> "192 permute_2" [label="(1, 12, 197, 64)", style=solid]; +"192 permute_2" -> "193 view_14" [label="(197, 1, 12, 64)", style=solid]; +"193 view_14" -> "195 view_14_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"194 encoder_layers_encoder_layer_1_self_attention_out_proj_weight_updated_constant0" -> "196 linear_5" [label="(768, 768)", style=solid]; +"195 view_14_0_0_nncf_smooth_quant_0" -> "196 linear_5" [label="(197, 768)", style=solid]; +"196 linear_5" -> "197 view_15" [label="(197, 768)", style=solid]; +"197 view_15" -> "198 transpose_11" [label="(197, 1, 768)", style=solid]; +"198 transpose_11" -> "199 dropout_4" [label="(1, 197, 768)", style=solid]; +"199 dropout_4" -> "200 add_3" [label="(1, 197, 768)", style=solid]; +"200 add_3" -> "201 layer_norm_3" [label="(1, 197, 768)", style=solid]; +"200 add_3" -> "221 add_4" [label="(1, 197, 768)", style=solid]; +"201 layer_norm_3" -> "202 layer_norm_3_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"202 layer_norm_3_0_0_nncf_smooth_quant_0" -> "203 quantize_per_tensor_default_3" [label="(1, 197, 768)", style=solid]; +"203 quantize_per_tensor_default_3" -> "204 dequantize_per_tensor_default_3" [label="(1, 197, 768)", style=solid]; +"204 dequantize_per_tensor_default_3" -> "209 linear_6" [label="(1, 197, 768)", style=solid]; +"205 linear_6_scale_0" -> "208 dequantize_per_channel_default_3" [label="(3072,)", style=solid]; +"206 linear_6_zero_point_0" -> "208 dequantize_per_channel_default_3" [label="(3072,)", style=solid]; +"207 _frozen_param3" -> "208 dequantize_per_channel_default_3" [label="(3072, 768)", style=solid]; +"208 dequantize_per_channel_default_3" -> "209 linear_6" [label="(3072, 768)", style=solid]; +"209 linear_6" -> "210 gelu_1" [label="(1, 197, 3072)", style=solid]; +"210 gelu_1" -> "211 dropout_5" [label="(1, 197, 3072)", style=solid]; +"211 dropout_5" -> "212 dropout_5_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"212 dropout_5_0_0_nncf_smooth_quant_0" -> "213 quantize_per_tensor_default_4" [label="(1, 197, 3072)", style=solid]; +"213 quantize_per_tensor_default_4" -> "214 dequantize_per_tensor_default_4" [label="(1, 197, 3072)", style=solid]; +"214 dequantize_per_tensor_default_4" -> "219 linear_7" [label="(1, 197, 3072)", style=solid]; +"215 linear_7_scale_0" -> "218 dequantize_per_channel_default_4" [label="(768,)", style=solid]; +"216 linear_7_zero_point_0" -> "218 dequantize_per_channel_default_4" [label="(768,)", style=solid]; +"217 _frozen_param4" -> "218 dequantize_per_channel_default_4" [label="(768, 3072)", style=solid]; +"218 dequantize_per_channel_default_4" -> "219 linear_7" [label="(768, 3072)", style=solid]; +"219 linear_7" -> "220 dropout_6" [label="(1, 197, 768)", style=solid]; +"220 dropout_6" -> "221 add_4" [label="(1, 197, 768)", style=solid]; +"221 add_4" -> "222 layer_norm_4" [label="(1, 197, 768)", style=solid]; +"221 add_4" -> "253 add_5" [label="(1, 197, 768)", style=solid]; +"222 layer_norm_4" -> "223 transpose_12" [label="(1, 197, 768)", style=solid]; +"223 transpose_12" -> "225 transpose_12_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"224 encoder_layers_encoder_layer_2_self_attention_in_proj_weight_updated_constant0" -> "226 linear_8" [label="(2304, 768)", style=solid]; +"225 transpose_12_0_0_nncf_smooth_quant_0" -> "226 linear_8" [label="(197, 1, 768)", style=solid]; +"226 linear_8" -> "227 unflatten_2" [label="(197, 1, 2304)", style=solid]; +"227 unflatten_2" -> "228 unsqueeze_2" [label="(197, 1, 3, 768)", style=solid]; +"228 unsqueeze_2" -> "229 transpose_13" [label="(1, 197, 1, 3, 768)", style=solid]; +"229 transpose_13" -> "230 squeeze_2" [label="(3, 197, 1, 1, 768)", style=solid]; +"230 squeeze_2" -> "231 contiguous_2" [label="(3, 197, 1, 768)", style=solid]; +"231 contiguous_2" -> "232 select_6" [label="(3, 197, 1, 768)", style=solid]; +"231 contiguous_2" -> "233 select_7" [label="(3, 197, 1, 768)", style=solid]; +"231 contiguous_2" -> "234 select_8" [label="(3, 197, 1, 768)", style=solid]; +"232 select_6" -> "235 view_16" [label="(197, 1, 768)", style=solid]; +"233 select_7" -> "237 view_17" [label="(197, 1, 768)", style=solid]; +"234 select_8" -> "239 view_18" [label="(197, 1, 768)", style=solid]; +"235 view_16" -> "236 transpose_14" [label="(197, 12, 64)", style=solid]; +"236 transpose_14" -> "241 view_19" [label="(12, 197, 64)", style=solid]; +"237 view_17" -> "238 transpose_15" [label="(197, 12, 64)", style=solid]; +"238 transpose_15" -> "242 view_20" [label="(12, 197, 64)", style=solid]; +"239 view_18" -> "240 transpose_16" [label="(197, 12, 64)", style=solid]; +"240 transpose_16" -> "243 view_21" [label="(12, 197, 64)", style=solid]; +"241 view_19" -> "244 scaled_dot_product_attention_2" [label="(1, 12, 197, 64)", style=solid]; +"242 view_20" -> "244 scaled_dot_product_attention_2" [label="(1, 12, 197, 64)", style=solid]; +"243 view_21" -> "244 scaled_dot_product_attention_2" [label="(1, 12, 197, 64)", style=solid]; +"244 scaled_dot_product_attention_2" -> "245 permute_3" [label="(1, 12, 197, 64)", style=solid]; +"245 permute_3" -> "246 view_22" [label="(197, 1, 12, 64)", style=solid]; +"246 view_22" -> "248 view_22_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"247 encoder_layers_encoder_layer_2_self_attention_out_proj_weight_updated_constant0" -> "249 linear_9" [label="(768, 768)", style=solid]; +"248 view_22_0_0_nncf_smooth_quant_0" -> "249 linear_9" [label="(197, 768)", style=solid]; +"249 linear_9" -> "250 view_23" [label="(197, 768)", style=solid]; +"250 view_23" -> "251 transpose_17" [label="(197, 1, 768)", style=solid]; +"251 transpose_17" -> "252 dropout_7" [label="(1, 197, 768)", style=solid]; +"252 dropout_7" -> "253 add_5" [label="(1, 197, 768)", style=solid]; +"253 add_5" -> "254 layer_norm_5" [label="(1, 197, 768)", style=solid]; +"253 add_5" -> "274 add_6" [label="(1, 197, 768)", style=solid]; +"254 layer_norm_5" -> "255 layer_norm_5_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"255 layer_norm_5_0_0_nncf_smooth_quant_0" -> "256 quantize_per_tensor_default_5" [label="(1, 197, 768)", style=solid]; +"256 quantize_per_tensor_default_5" -> "257 dequantize_per_tensor_default_5" [label="(1, 197, 768)", style=solid]; +"257 dequantize_per_tensor_default_5" -> "262 linear_10" [label="(1, 197, 768)", style=solid]; +"258 linear_10_scale_0" -> "261 dequantize_per_channel_default_5" [label="(3072,)", style=solid]; +"259 linear_10_zero_point_0" -> "261 dequantize_per_channel_default_5" [label="(3072,)", style=solid]; +"260 _frozen_param5" -> "261 dequantize_per_channel_default_5" [label="(3072, 768)", style=solid]; +"261 dequantize_per_channel_default_5" -> "262 linear_10" [label="(3072, 768)", style=solid]; +"262 linear_10" -> "263 gelu_2" [label="(1, 197, 3072)", style=solid]; +"263 gelu_2" -> "264 dropout_8" [label="(1, 197, 3072)", style=solid]; +"264 dropout_8" -> "265 dropout_8_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"265 dropout_8_0_0_nncf_smooth_quant_0" -> "266 quantize_per_tensor_default_6" [label="(1, 197, 3072)", style=solid]; +"266 quantize_per_tensor_default_6" -> "267 dequantize_per_tensor_default_6" [label="(1, 197, 3072)", style=solid]; +"267 dequantize_per_tensor_default_6" -> "272 linear_11" [label="(1, 197, 3072)", style=solid]; +"268 linear_11_scale_0" -> "271 dequantize_per_channel_default_6" [label="(768,)", style=solid]; +"269 linear_11_zero_point_0" -> "271 dequantize_per_channel_default_6" [label="(768,)", style=solid]; +"270 _frozen_param6" -> "271 dequantize_per_channel_default_6" [label="(768, 3072)", style=solid]; +"271 dequantize_per_channel_default_6" -> "272 linear_11" [label="(768, 3072)", style=solid]; +"272 linear_11" -> "273 dropout_9" [label="(1, 197, 768)", style=solid]; +"273 dropout_9" -> "274 add_6" [label="(1, 197, 768)", style=solid]; +"274 add_6" -> "275 layer_norm_6" [label="(1, 197, 768)", style=solid]; +"274 add_6" -> "306 add_7" [label="(1, 197, 768)", style=solid]; +"275 layer_norm_6" -> "276 transpose_18" [label="(1, 197, 768)", style=solid]; +"276 transpose_18" -> "278 transpose_18_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"277 encoder_layers_encoder_layer_3_self_attention_in_proj_weight_updated_constant0" -> "279 linear_12" [label="(2304, 768)", style=solid]; +"278 transpose_18_0_0_nncf_smooth_quant_0" -> "279 linear_12" [label="(197, 1, 768)", style=solid]; +"279 linear_12" -> "280 unflatten_3" [label="(197, 1, 2304)", style=solid]; +"280 unflatten_3" -> "281 unsqueeze_3" [label="(197, 1, 3, 768)", style=solid]; +"281 unsqueeze_3" -> "282 transpose_19" [label="(1, 197, 1, 3, 768)", style=solid]; +"282 transpose_19" -> "283 squeeze_3" [label="(3, 197, 1, 1, 768)", style=solid]; +"283 squeeze_3" -> "284 contiguous_3" [label="(3, 197, 1, 768)", style=solid]; +"284 contiguous_3" -> "285 select_9" [label="(3, 197, 1, 768)", style=solid]; +"284 contiguous_3" -> "286 select_10" [label="(3, 197, 1, 768)", style=solid]; +"284 contiguous_3" -> "287 select_11" [label="(3, 197, 1, 768)", style=solid]; +"285 select_9" -> "288 view_24" [label="(197, 1, 768)", style=solid]; +"286 select_10" -> "290 view_25" [label="(197, 1, 768)", style=solid]; +"287 select_11" -> "292 view_26" [label="(197, 1, 768)", style=solid]; +"288 view_24" -> "289 transpose_20" [label="(197, 12, 64)", style=solid]; +"289 transpose_20" -> "294 view_27" [label="(12, 197, 64)", style=solid]; +"290 view_25" -> "291 transpose_21" [label="(197, 12, 64)", style=solid]; +"291 transpose_21" -> "295 view_28" [label="(12, 197, 64)", style=solid]; +"292 view_26" -> "293 transpose_22" [label="(197, 12, 64)", style=solid]; +"293 transpose_22" -> "296 view_29" [label="(12, 197, 64)", style=solid]; +"294 view_27" -> "297 scaled_dot_product_attention_3" [label="(1, 12, 197, 64)", style=solid]; +"295 view_28" -> "297 scaled_dot_product_attention_3" [label="(1, 12, 197, 64)", style=solid]; +"296 view_29" -> "297 scaled_dot_product_attention_3" [label="(1, 12, 197, 64)", style=solid]; +"297 scaled_dot_product_attention_3" -> "298 permute_4" [label="(1, 12, 197, 64)", style=solid]; +"298 permute_4" -> "299 view_30" [label="(197, 1, 12, 64)", style=solid]; +"299 view_30" -> "301 view_30_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"300 encoder_layers_encoder_layer_3_self_attention_out_proj_weight_updated_constant0" -> "302 linear_13" [label="(768, 768)", style=solid]; +"301 view_30_0_0_nncf_smooth_quant_0" -> "302 linear_13" [label="(197, 768)", style=solid]; +"302 linear_13" -> "303 view_31" [label="(197, 768)", style=solid]; +"303 view_31" -> "304 transpose_23" [label="(197, 1, 768)", style=solid]; +"304 transpose_23" -> "305 dropout_10" [label="(1, 197, 768)", style=solid]; +"305 dropout_10" -> "306 add_7" [label="(1, 197, 768)", style=solid]; +"306 add_7" -> "307 layer_norm_7" [label="(1, 197, 768)", style=solid]; +"306 add_7" -> "327 add_8" [label="(1, 197, 768)", style=solid]; +"307 layer_norm_7" -> "308 layer_norm_7_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"308 layer_norm_7_0_0_nncf_smooth_quant_0" -> "309 quantize_per_tensor_default_7" [label="(1, 197, 768)", style=solid]; +"309 quantize_per_tensor_default_7" -> "310 dequantize_per_tensor_default_7" [label="(1, 197, 768)", style=solid]; +"310 dequantize_per_tensor_default_7" -> "315 linear_14" [label="(1, 197, 768)", style=solid]; +"311 linear_14_scale_0" -> "314 dequantize_per_channel_default_7" [label="(3072,)", style=solid]; +"312 linear_14_zero_point_0" -> "314 dequantize_per_channel_default_7" [label="(3072,)", style=solid]; +"313 _frozen_param7" -> "314 dequantize_per_channel_default_7" [label="(3072, 768)", style=solid]; +"314 dequantize_per_channel_default_7" -> "315 linear_14" [label="(3072, 768)", style=solid]; +"315 linear_14" -> "316 gelu_3" [label="(1, 197, 3072)", style=solid]; +"316 gelu_3" -> "317 dropout_11" [label="(1, 197, 3072)", style=solid]; +"317 dropout_11" -> "318 dropout_11_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"318 dropout_11_0_0_nncf_smooth_quant_0" -> "319 quantize_per_tensor_default_8" [label="(1, 197, 3072)", style=solid]; +"319 quantize_per_tensor_default_8" -> "320 dequantize_per_tensor_default_8" [label="(1, 197, 3072)", style=solid]; +"320 dequantize_per_tensor_default_8" -> "325 linear_15" [label="(1, 197, 3072)", style=solid]; +"321 linear_15_scale_0" -> "324 dequantize_per_channel_default_8" [label="(768,)", style=solid]; +"322 linear_15_zero_point_0" -> "324 dequantize_per_channel_default_8" [label="(768,)", style=solid]; +"323 _frozen_param8" -> "324 dequantize_per_channel_default_8" [label="(768, 3072)", style=solid]; +"324 dequantize_per_channel_default_8" -> "325 linear_15" [label="(768, 3072)", style=solid]; +"325 linear_15" -> "326 dropout_12" [label="(1, 197, 768)", style=solid]; +"326 dropout_12" -> "327 add_8" [label="(1, 197, 768)", style=solid]; +"327 add_8" -> "328 layer_norm_8" [label="(1, 197, 768)", style=solid]; +"327 add_8" -> "359 add_9" [label="(1, 197, 768)", style=solid]; +"328 layer_norm_8" -> "329 transpose_24" [label="(1, 197, 768)", style=solid]; +"329 transpose_24" -> "331 transpose_24_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"330 encoder_layers_encoder_layer_4_self_attention_in_proj_weight_updated_constant0" -> "332 linear_16" [label="(2304, 768)", style=solid]; +"331 transpose_24_0_0_nncf_smooth_quant_0" -> "332 linear_16" [label="(197, 1, 768)", style=solid]; +"332 linear_16" -> "333 unflatten_4" [label="(197, 1, 2304)", style=solid]; +"333 unflatten_4" -> "334 unsqueeze_4" [label="(197, 1, 3, 768)", style=solid]; +"334 unsqueeze_4" -> "335 transpose_25" [label="(1, 197, 1, 3, 768)", style=solid]; +"335 transpose_25" -> "336 squeeze_4" [label="(3, 197, 1, 1, 768)", style=solid]; +"336 squeeze_4" -> "337 contiguous_4" [label="(3, 197, 1, 768)", style=solid]; +"337 contiguous_4" -> "338 select_12" [label="(3, 197, 1, 768)", style=solid]; +"337 contiguous_4" -> "339 select_13" [label="(3, 197, 1, 768)", style=solid]; +"337 contiguous_4" -> "340 select_14" [label="(3, 197, 1, 768)", style=solid]; +"338 select_12" -> "341 view_32" [label="(197, 1, 768)", style=solid]; +"339 select_13" -> "343 view_33" [label="(197, 1, 768)", style=solid]; +"340 select_14" -> "345 view_34" [label="(197, 1, 768)", style=solid]; +"341 view_32" -> "342 transpose_26" [label="(197, 12, 64)", style=solid]; +"342 transpose_26" -> "347 view_35" [label="(12, 197, 64)", style=solid]; +"343 view_33" -> "344 transpose_27" [label="(197, 12, 64)", style=solid]; +"344 transpose_27" -> "348 view_36" [label="(12, 197, 64)", style=solid]; +"345 view_34" -> "346 transpose_28" [label="(197, 12, 64)", style=solid]; +"346 transpose_28" -> "349 view_37" [label="(12, 197, 64)", style=solid]; +"347 view_35" -> "350 scaled_dot_product_attention_4" [label="(1, 12, 197, 64)", style=solid]; +"348 view_36" -> "350 scaled_dot_product_attention_4" [label="(1, 12, 197, 64)", style=solid]; +"349 view_37" -> "350 scaled_dot_product_attention_4" [label="(1, 12, 197, 64)", style=solid]; +"350 scaled_dot_product_attention_4" -> "351 permute_5" [label="(1, 12, 197, 64)", style=solid]; +"351 permute_5" -> "352 view_38" [label="(197, 1, 12, 64)", style=solid]; +"352 view_38" -> "354 view_38_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"353 encoder_layers_encoder_layer_4_self_attention_out_proj_weight_updated_constant0" -> "355 linear_17" [label="(768, 768)", style=solid]; +"354 view_38_0_0_nncf_smooth_quant_0" -> "355 linear_17" [label="(197, 768)", style=solid]; +"355 linear_17" -> "356 view_39" [label="(197, 768)", style=solid]; +"356 view_39" -> "357 transpose_29" [label="(197, 1, 768)", style=solid]; +"357 transpose_29" -> "358 dropout_13" [label="(1, 197, 768)", style=solid]; +"358 dropout_13" -> "359 add_9" [label="(1, 197, 768)", style=solid]; +"359 add_9" -> "360 layer_norm_9" [label="(1, 197, 768)", style=solid]; +"359 add_9" -> "380 add_10" [label="(1, 197, 768)", style=solid]; +"360 layer_norm_9" -> "361 layer_norm_9_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"361 layer_norm_9_0_0_nncf_smooth_quant_0" -> "362 quantize_per_tensor_default_9" [label="(1, 197, 768)", style=solid]; +"362 quantize_per_tensor_default_9" -> "363 dequantize_per_tensor_default_9" [label="(1, 197, 768)", style=solid]; +"363 dequantize_per_tensor_default_9" -> "368 linear_18" [label="(1, 197, 768)", style=solid]; +"364 linear_18_scale_0" -> "367 dequantize_per_channel_default_9" [label="(3072,)", style=solid]; +"365 linear_18_zero_point_0" -> "367 dequantize_per_channel_default_9" [label="(3072,)", style=solid]; +"366 _frozen_param9" -> "367 dequantize_per_channel_default_9" [label="(3072, 768)", style=solid]; +"367 dequantize_per_channel_default_9" -> "368 linear_18" [label="(3072, 768)", style=solid]; +"368 linear_18" -> "369 gelu_4" [label="(1, 197, 3072)", style=solid]; +"369 gelu_4" -> "370 dropout_14" [label="(1, 197, 3072)", style=solid]; +"370 dropout_14" -> "371 dropout_14_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"371 dropout_14_0_0_nncf_smooth_quant_0" -> "372 quantize_per_tensor_default_10" [label="(1, 197, 3072)", style=solid]; +"372 quantize_per_tensor_default_10" -> "373 dequantize_per_tensor_default_10" [label="(1, 197, 3072)", style=solid]; +"373 dequantize_per_tensor_default_10" -> "378 linear_19" [label="(1, 197, 3072)", style=solid]; +"374 linear_19_scale_0" -> "377 dequantize_per_channel_default_10" [label="(768,)", style=solid]; +"375 linear_19_zero_point_0" -> "377 dequantize_per_channel_default_10" [label="(768,)", style=solid]; +"376 _frozen_param10" -> "377 dequantize_per_channel_default_10" [label="(768, 3072)", style=solid]; +"377 dequantize_per_channel_default_10" -> "378 linear_19" [label="(768, 3072)", style=solid]; +"378 linear_19" -> "379 dropout_15" [label="(1, 197, 768)", style=solid]; +"379 dropout_15" -> "380 add_10" [label="(1, 197, 768)", style=solid]; +"380 add_10" -> "381 layer_norm_10" [label="(1, 197, 768)", style=solid]; +"380 add_10" -> "412 add_11" [label="(1, 197, 768)", style=solid]; +"381 layer_norm_10" -> "382 transpose_30" [label="(1, 197, 768)", style=solid]; +"382 transpose_30" -> "384 transpose_30_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"383 encoder_layers_encoder_layer_5_self_attention_in_proj_weight_updated_constant0" -> "385 linear_20" [label="(2304, 768)", style=solid]; +"384 transpose_30_0_0_nncf_smooth_quant_0" -> "385 linear_20" [label="(197, 1, 768)", style=solid]; +"385 linear_20" -> "386 unflatten_5" [label="(197, 1, 2304)", style=solid]; +"386 unflatten_5" -> "387 unsqueeze_5" [label="(197, 1, 3, 768)", style=solid]; +"387 unsqueeze_5" -> "388 transpose_31" [label="(1, 197, 1, 3, 768)", style=solid]; +"388 transpose_31" -> "389 squeeze_5" [label="(3, 197, 1, 1, 768)", style=solid]; +"389 squeeze_5" -> "390 contiguous_5" [label="(3, 197, 1, 768)", style=solid]; +"390 contiguous_5" -> "391 select_15" [label="(3, 197, 1, 768)", style=solid]; +"390 contiguous_5" -> "392 select_16" [label="(3, 197, 1, 768)", style=solid]; +"390 contiguous_5" -> "393 select_17" [label="(3, 197, 1, 768)", style=solid]; +"391 select_15" -> "394 view_40" [label="(197, 1, 768)", style=solid]; +"392 select_16" -> "396 view_41" [label="(197, 1, 768)", style=solid]; +"393 select_17" -> "398 view_42" [label="(197, 1, 768)", style=solid]; +"394 view_40" -> "395 transpose_32" [label="(197, 12, 64)", style=solid]; +"395 transpose_32" -> "400 view_43" [label="(12, 197, 64)", style=solid]; +"396 view_41" -> "397 transpose_33" [label="(197, 12, 64)", style=solid]; +"397 transpose_33" -> "401 view_44" [label="(12, 197, 64)", style=solid]; +"398 view_42" -> "399 transpose_34" [label="(197, 12, 64)", style=solid]; +"399 transpose_34" -> "402 view_45" [label="(12, 197, 64)", style=solid]; +"400 view_43" -> "403 scaled_dot_product_attention_5" [label="(1, 12, 197, 64)", style=solid]; +"401 view_44" -> "403 scaled_dot_product_attention_5" [label="(1, 12, 197, 64)", style=solid]; +"402 view_45" -> "403 scaled_dot_product_attention_5" [label="(1, 12, 197, 64)", style=solid]; +"403 scaled_dot_product_attention_5" -> "404 permute_6" [label="(1, 12, 197, 64)", style=solid]; +"404 permute_6" -> "405 view_46" [label="(197, 1, 12, 64)", style=solid]; +"405 view_46" -> "407 view_46_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"406 encoder_layers_encoder_layer_5_self_attention_out_proj_weight_updated_constant0" -> "408 linear_21" [label="(768, 768)", style=solid]; +"407 view_46_0_0_nncf_smooth_quant_0" -> "408 linear_21" [label="(197, 768)", style=solid]; +"408 linear_21" -> "409 view_47" [label="(197, 768)", style=solid]; +"409 view_47" -> "410 transpose_35" [label="(197, 1, 768)", style=solid]; +"410 transpose_35" -> "411 dropout_16" [label="(1, 197, 768)", style=solid]; +"411 dropout_16" -> "412 add_11" [label="(1, 197, 768)", style=solid]; +"412 add_11" -> "413 layer_norm_11" [label="(1, 197, 768)", style=solid]; +"412 add_11" -> "433 add_12" [label="(1, 197, 768)", style=solid]; +"413 layer_norm_11" -> "414 layer_norm_11_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"414 layer_norm_11_0_0_nncf_smooth_quant_0" -> "415 quantize_per_tensor_default_11" [label="(1, 197, 768)", style=solid]; +"415 quantize_per_tensor_default_11" -> "416 dequantize_per_tensor_default_11" [label="(1, 197, 768)", style=solid]; +"416 dequantize_per_tensor_default_11" -> "421 linear_22" [label="(1, 197, 768)", style=solid]; +"417 linear_22_scale_0" -> "420 dequantize_per_channel_default_11" [label="(3072,)", style=solid]; +"418 linear_22_zero_point_0" -> "420 dequantize_per_channel_default_11" [label="(3072,)", style=solid]; +"419 _frozen_param11" -> "420 dequantize_per_channel_default_11" [label="(3072, 768)", style=solid]; +"420 dequantize_per_channel_default_11" -> "421 linear_22" [label="(3072, 768)", style=solid]; +"421 linear_22" -> "422 gelu_5" [label="(1, 197, 3072)", style=solid]; +"422 gelu_5" -> "423 dropout_17" [label="(1, 197, 3072)", style=solid]; +"423 dropout_17" -> "424 dropout_17_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"424 dropout_17_0_0_nncf_smooth_quant_0" -> "425 quantize_per_tensor_default_12" [label="(1, 197, 3072)", style=solid]; +"425 quantize_per_tensor_default_12" -> "426 dequantize_per_tensor_default_12" [label="(1, 197, 3072)", style=solid]; +"426 dequantize_per_tensor_default_12" -> "431 linear_23" [label="(1, 197, 3072)", style=solid]; +"427 linear_23_scale_0" -> "430 dequantize_per_channel_default_12" [label="(768,)", style=solid]; +"428 linear_23_zero_point_0" -> "430 dequantize_per_channel_default_12" [label="(768,)", style=solid]; +"429 _frozen_param12" -> "430 dequantize_per_channel_default_12" [label="(768, 3072)", style=solid]; +"430 dequantize_per_channel_default_12" -> "431 linear_23" [label="(768, 3072)", style=solid]; +"431 linear_23" -> "432 dropout_18" [label="(1, 197, 768)", style=solid]; +"432 dropout_18" -> "433 add_12" [label="(1, 197, 768)", style=solid]; +"433 add_12" -> "434 layer_norm_12" [label="(1, 197, 768)", style=solid]; +"433 add_12" -> "465 add_13" [label="(1, 197, 768)", style=solid]; +"434 layer_norm_12" -> "435 transpose_36" [label="(1, 197, 768)", style=solid]; +"435 transpose_36" -> "437 transpose_36_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"436 encoder_layers_encoder_layer_6_self_attention_in_proj_weight_updated_constant0" -> "438 linear_24" [label="(2304, 768)", style=solid]; +"437 transpose_36_0_0_nncf_smooth_quant_0" -> "438 linear_24" [label="(197, 1, 768)", style=solid]; +"438 linear_24" -> "439 unflatten_6" [label="(197, 1, 2304)", style=solid]; +"439 unflatten_6" -> "440 unsqueeze_6" [label="(197, 1, 3, 768)", style=solid]; +"440 unsqueeze_6" -> "441 transpose_37" [label="(1, 197, 1, 3, 768)", style=solid]; +"441 transpose_37" -> "442 squeeze_6" [label="(3, 197, 1, 1, 768)", style=solid]; +"442 squeeze_6" -> "443 contiguous_6" [label="(3, 197, 1, 768)", style=solid]; +"443 contiguous_6" -> "444 select_18" [label="(3, 197, 1, 768)", style=solid]; +"443 contiguous_6" -> "445 select_19" [label="(3, 197, 1, 768)", style=solid]; +"443 contiguous_6" -> "446 select_20" [label="(3, 197, 1, 768)", style=solid]; +"444 select_18" -> "447 view_48" [label="(197, 1, 768)", style=solid]; +"445 select_19" -> "449 view_49" [label="(197, 1, 768)", style=solid]; +"446 select_20" -> "451 view_50" [label="(197, 1, 768)", style=solid]; +"447 view_48" -> "448 transpose_38" [label="(197, 12, 64)", style=solid]; +"448 transpose_38" -> "453 view_51" [label="(12, 197, 64)", style=solid]; +"449 view_49" -> "450 transpose_39" [label="(197, 12, 64)", style=solid]; +"450 transpose_39" -> "454 view_52" [label="(12, 197, 64)", style=solid]; +"451 view_50" -> "452 transpose_40" [label="(197, 12, 64)", style=solid]; +"452 transpose_40" -> "455 view_53" [label="(12, 197, 64)", style=solid]; +"453 view_51" -> "456 scaled_dot_product_attention_6" [label="(1, 12, 197, 64)", style=solid]; +"454 view_52" -> "456 scaled_dot_product_attention_6" [label="(1, 12, 197, 64)", style=solid]; +"455 view_53" -> "456 scaled_dot_product_attention_6" [label="(1, 12, 197, 64)", style=solid]; +"456 scaled_dot_product_attention_6" -> "457 permute_7" [label="(1, 12, 197, 64)", style=solid]; +"457 permute_7" -> "458 view_54" [label="(197, 1, 12, 64)", style=solid]; +"458 view_54" -> "460 view_54_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"459 encoder_layers_encoder_layer_6_self_attention_out_proj_weight_updated_constant0" -> "461 linear_25" [label="(768, 768)", style=solid]; +"460 view_54_0_0_nncf_smooth_quant_0" -> "461 linear_25" [label="(197, 768)", style=solid]; +"461 linear_25" -> "462 view_55" [label="(197, 768)", style=solid]; +"462 view_55" -> "463 transpose_41" [label="(197, 1, 768)", style=solid]; +"463 transpose_41" -> "464 dropout_19" [label="(1, 197, 768)", style=solid]; +"464 dropout_19" -> "465 add_13" [label="(1, 197, 768)", style=solid]; +"465 add_13" -> "466 layer_norm_13" [label="(1, 197, 768)", style=solid]; +"465 add_13" -> "486 add_14" [label="(1, 197, 768)", style=solid]; +"466 layer_norm_13" -> "467 layer_norm_13_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"467 layer_norm_13_0_0_nncf_smooth_quant_0" -> "468 quantize_per_tensor_default_13" [label="(1, 197, 768)", style=solid]; +"468 quantize_per_tensor_default_13" -> "469 dequantize_per_tensor_default_13" [label="(1, 197, 768)", style=solid]; +"469 dequantize_per_tensor_default_13" -> "474 linear_26" [label="(1, 197, 768)", style=solid]; +"470 linear_26_scale_0" -> "473 dequantize_per_channel_default_13" [label="(3072,)", style=solid]; +"471 linear_26_zero_point_0" -> "473 dequantize_per_channel_default_13" [label="(3072,)", style=solid]; +"472 _frozen_param13" -> "473 dequantize_per_channel_default_13" [label="(3072, 768)", style=solid]; +"473 dequantize_per_channel_default_13" -> "474 linear_26" [label="(3072, 768)", style=solid]; +"474 linear_26" -> "475 gelu_6" [label="(1, 197, 3072)", style=solid]; +"475 gelu_6" -> "476 dropout_20" [label="(1, 197, 3072)", style=solid]; +"476 dropout_20" -> "477 dropout_20_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"477 dropout_20_0_0_nncf_smooth_quant_0" -> "478 quantize_per_tensor_default_14" [label="(1, 197, 3072)", style=solid]; +"478 quantize_per_tensor_default_14" -> "479 dequantize_per_tensor_default_14" [label="(1, 197, 3072)", style=solid]; +"479 dequantize_per_tensor_default_14" -> "484 linear_27" [label="(1, 197, 3072)", style=solid]; +"480 linear_27_scale_0" -> "483 dequantize_per_channel_default_14" [label="(768,)", style=solid]; +"481 linear_27_zero_point_0" -> "483 dequantize_per_channel_default_14" [label="(768,)", style=solid]; +"482 _frozen_param14" -> "483 dequantize_per_channel_default_14" [label="(768, 3072)", style=solid]; +"483 dequantize_per_channel_default_14" -> "484 linear_27" [label="(768, 3072)", style=solid]; +"484 linear_27" -> "485 dropout_21" [label="(1, 197, 768)", style=solid]; +"485 dropout_21" -> "486 add_14" [label="(1, 197, 768)", style=solid]; +"486 add_14" -> "487 layer_norm_14" [label="(1, 197, 768)", style=solid]; +"486 add_14" -> "518 add_15" [label="(1, 197, 768)", style=solid]; +"487 layer_norm_14" -> "488 transpose_42" [label="(1, 197, 768)", style=solid]; +"488 transpose_42" -> "490 transpose_42_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"489 encoder_layers_encoder_layer_7_self_attention_in_proj_weight_updated_constant0" -> "491 linear_28" [label="(2304, 768)", style=solid]; +"490 transpose_42_0_0_nncf_smooth_quant_0" -> "491 linear_28" [label="(197, 1, 768)", style=solid]; +"491 linear_28" -> "492 unflatten_7" [label="(197, 1, 2304)", style=solid]; +"492 unflatten_7" -> "493 unsqueeze_7" [label="(197, 1, 3, 768)", style=solid]; +"493 unsqueeze_7" -> "494 transpose_43" [label="(1, 197, 1, 3, 768)", style=solid]; +"494 transpose_43" -> "495 squeeze_7" [label="(3, 197, 1, 1, 768)", style=solid]; +"495 squeeze_7" -> "496 contiguous_7" [label="(3, 197, 1, 768)", style=solid]; +"496 contiguous_7" -> "497 select_21" [label="(3, 197, 1, 768)", style=solid]; +"496 contiguous_7" -> "498 select_22" [label="(3, 197, 1, 768)", style=solid]; +"496 contiguous_7" -> "499 select_23" [label="(3, 197, 1, 768)", style=solid]; +"497 select_21" -> "500 view_56" [label="(197, 1, 768)", style=solid]; +"498 select_22" -> "502 view_57" [label="(197, 1, 768)", style=solid]; +"499 select_23" -> "504 view_58" [label="(197, 1, 768)", style=solid]; +"500 view_56" -> "501 transpose_44" [label="(197, 12, 64)", style=solid]; +"501 transpose_44" -> "506 view_59" [label="(12, 197, 64)", style=solid]; +"502 view_57" -> "503 transpose_45" [label="(197, 12, 64)", style=solid]; +"503 transpose_45" -> "507 view_60" [label="(12, 197, 64)", style=solid]; +"504 view_58" -> "505 transpose_46" [label="(197, 12, 64)", style=solid]; +"505 transpose_46" -> "508 view_61" [label="(12, 197, 64)", style=solid]; +"506 view_59" -> "509 scaled_dot_product_attention_7" [label="(1, 12, 197, 64)", style=solid]; +"507 view_60" -> "509 scaled_dot_product_attention_7" [label="(1, 12, 197, 64)", style=solid]; +"508 view_61" -> "509 scaled_dot_product_attention_7" [label="(1, 12, 197, 64)", style=solid]; +"509 scaled_dot_product_attention_7" -> "510 permute_8" [label="(1, 12, 197, 64)", style=solid]; +"510 permute_8" -> "511 view_62" [label="(197, 1, 12, 64)", style=solid]; +"511 view_62" -> "513 view_62_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"512 encoder_layers_encoder_layer_7_self_attention_out_proj_weight_updated_constant0" -> "514 linear_29" [label="(768, 768)", style=solid]; +"513 view_62_0_0_nncf_smooth_quant_0" -> "514 linear_29" [label="(197, 768)", style=solid]; +"514 linear_29" -> "515 view_63" [label="(197, 768)", style=solid]; +"515 view_63" -> "516 transpose_47" [label="(197, 1, 768)", style=solid]; +"516 transpose_47" -> "517 dropout_22" [label="(1, 197, 768)", style=solid]; +"517 dropout_22" -> "518 add_15" [label="(1, 197, 768)", style=solid]; +"518 add_15" -> "519 layer_norm_15" [label="(1, 197, 768)", style=solid]; +"518 add_15" -> "539 add_16" [label="(1, 197, 768)", style=solid]; +"519 layer_norm_15" -> "520 layer_norm_15_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"520 layer_norm_15_0_0_nncf_smooth_quant_0" -> "521 quantize_per_tensor_default_15" [label="(1, 197, 768)", style=solid]; +"521 quantize_per_tensor_default_15" -> "522 dequantize_per_tensor_default_15" [label="(1, 197, 768)", style=solid]; +"522 dequantize_per_tensor_default_15" -> "527 linear_30" [label="(1, 197, 768)", style=solid]; +"523 linear_30_scale_0" -> "526 dequantize_per_channel_default_15" [label="(3072,)", style=solid]; +"524 linear_30_zero_point_0" -> "526 dequantize_per_channel_default_15" [label="(3072,)", style=solid]; +"525 _frozen_param15" -> "526 dequantize_per_channel_default_15" [label="(3072, 768)", style=solid]; +"526 dequantize_per_channel_default_15" -> "527 linear_30" [label="(3072, 768)", style=solid]; +"527 linear_30" -> "528 gelu_7" [label="(1, 197, 3072)", style=solid]; +"528 gelu_7" -> "529 dropout_23" [label="(1, 197, 3072)", style=solid]; +"529 dropout_23" -> "530 dropout_23_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"530 dropout_23_0_0_nncf_smooth_quant_0" -> "531 quantize_per_tensor_default_16" [label="(1, 197, 3072)", style=solid]; +"531 quantize_per_tensor_default_16" -> "532 dequantize_per_tensor_default_16" [label="(1, 197, 3072)", style=solid]; +"532 dequantize_per_tensor_default_16" -> "537 linear_31" [label="(1, 197, 3072)", style=solid]; +"533 linear_31_scale_0" -> "536 dequantize_per_channel_default_16" [label="(768,)", style=solid]; +"534 linear_31_zero_point_0" -> "536 dequantize_per_channel_default_16" [label="(768,)", style=solid]; +"535 _frozen_param16" -> "536 dequantize_per_channel_default_16" [label="(768, 3072)", style=solid]; +"536 dequantize_per_channel_default_16" -> "537 linear_31" [label="(768, 3072)", style=solid]; +"537 linear_31" -> "538 dropout_24" [label="(1, 197, 768)", style=solid]; +"538 dropout_24" -> "539 add_16" [label="(1, 197, 768)", style=solid]; +"539 add_16" -> "540 layer_norm_16" [label="(1, 197, 768)", style=solid]; +"539 add_16" -> "571 add_17" [label="(1, 197, 768)", style=solid]; +"540 layer_norm_16" -> "541 transpose_48" [label="(1, 197, 768)", style=solid]; +"541 transpose_48" -> "543 transpose_48_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"542 encoder_layers_encoder_layer_8_self_attention_in_proj_weight_updated_constant0" -> "544 linear_32" [label="(2304, 768)", style=solid]; +"543 transpose_48_0_0_nncf_smooth_quant_0" -> "544 linear_32" [label="(197, 1, 768)", style=solid]; +"544 linear_32" -> "545 unflatten_8" [label="(197, 1, 2304)", style=solid]; +"545 unflatten_8" -> "546 unsqueeze_8" [label="(197, 1, 3, 768)", style=solid]; +"546 unsqueeze_8" -> "547 transpose_49" [label="(1, 197, 1, 3, 768)", style=solid]; +"547 transpose_49" -> "548 squeeze_8" [label="(3, 197, 1, 1, 768)", style=solid]; +"548 squeeze_8" -> "549 contiguous_8" [label="(3, 197, 1, 768)", style=solid]; +"549 contiguous_8" -> "550 select_24" [label="(3, 197, 1, 768)", style=solid]; +"549 contiguous_8" -> "551 select_25" [label="(3, 197, 1, 768)", style=solid]; +"549 contiguous_8" -> "552 select_26" [label="(3, 197, 1, 768)", style=solid]; +"550 select_24" -> "553 view_64" [label="(197, 1, 768)", style=solid]; +"551 select_25" -> "555 view_65" [label="(197, 1, 768)", style=solid]; +"552 select_26" -> "557 view_66" [label="(197, 1, 768)", style=solid]; +"553 view_64" -> "554 transpose_50" [label="(197, 12, 64)", style=solid]; +"554 transpose_50" -> "559 view_67" [label="(12, 197, 64)", style=solid]; +"555 view_65" -> "556 transpose_51" [label="(197, 12, 64)", style=solid]; +"556 transpose_51" -> "560 view_68" [label="(12, 197, 64)", style=solid]; +"557 view_66" -> "558 transpose_52" [label="(197, 12, 64)", style=solid]; +"558 transpose_52" -> "561 view_69" [label="(12, 197, 64)", style=solid]; +"559 view_67" -> "562 scaled_dot_product_attention_8" [label="(1, 12, 197, 64)", style=solid]; +"560 view_68" -> "562 scaled_dot_product_attention_8" [label="(1, 12, 197, 64)", style=solid]; +"561 view_69" -> "562 scaled_dot_product_attention_8" [label="(1, 12, 197, 64)", style=solid]; +"562 scaled_dot_product_attention_8" -> "563 permute_9" [label="(1, 12, 197, 64)", style=solid]; +"563 permute_9" -> "564 view_70" [label="(197, 1, 12, 64)", style=solid]; +"564 view_70" -> "566 view_70_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"565 encoder_layers_encoder_layer_8_self_attention_out_proj_weight_updated_constant0" -> "567 linear_33" [label="(768, 768)", style=solid]; +"566 view_70_0_0_nncf_smooth_quant_0" -> "567 linear_33" [label="(197, 768)", style=solid]; +"567 linear_33" -> "568 view_71" [label="(197, 768)", style=solid]; +"568 view_71" -> "569 transpose_53" [label="(197, 1, 768)", style=solid]; +"569 transpose_53" -> "570 dropout_25" [label="(1, 197, 768)", style=solid]; +"570 dropout_25" -> "571 add_17" [label="(1, 197, 768)", style=solid]; +"571 add_17" -> "572 layer_norm_17" [label="(1, 197, 768)", style=solid]; +"571 add_17" -> "592 add_18" [label="(1, 197, 768)", style=solid]; +"572 layer_norm_17" -> "573 layer_norm_17_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"573 layer_norm_17_0_0_nncf_smooth_quant_0" -> "574 quantize_per_tensor_default_17" [label="(1, 197, 768)", style=solid]; +"574 quantize_per_tensor_default_17" -> "575 dequantize_per_tensor_default_17" [label="(1, 197, 768)", style=solid]; +"575 dequantize_per_tensor_default_17" -> "580 linear_34" [label="(1, 197, 768)", style=solid]; +"576 linear_34_scale_0" -> "579 dequantize_per_channel_default_17" [label="(3072,)", style=solid]; +"577 linear_34_zero_point_0" -> "579 dequantize_per_channel_default_17" [label="(3072,)", style=solid]; +"578 _frozen_param17" -> "579 dequantize_per_channel_default_17" [label="(3072, 768)", style=solid]; +"579 dequantize_per_channel_default_17" -> "580 linear_34" [label="(3072, 768)", style=solid]; +"580 linear_34" -> "581 gelu_8" [label="(1, 197, 3072)", style=solid]; +"581 gelu_8" -> "582 dropout_26" [label="(1, 197, 3072)", style=solid]; +"582 dropout_26" -> "583 dropout_26_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"583 dropout_26_0_0_nncf_smooth_quant_0" -> "584 quantize_per_tensor_default_18" [label="(1, 197, 3072)", style=solid]; +"584 quantize_per_tensor_default_18" -> "585 dequantize_per_tensor_default_18" [label="(1, 197, 3072)", style=solid]; +"585 dequantize_per_tensor_default_18" -> "590 linear_35" [label="(1, 197, 3072)", style=solid]; +"586 linear_35_scale_0" -> "589 dequantize_per_channel_default_18" [label="(768,)", style=solid]; +"587 linear_35_zero_point_0" -> "589 dequantize_per_channel_default_18" [label="(768,)", style=solid]; +"588 _frozen_param18" -> "589 dequantize_per_channel_default_18" [label="(768, 3072)", style=solid]; +"589 dequantize_per_channel_default_18" -> "590 linear_35" [label="(768, 3072)", style=solid]; +"590 linear_35" -> "591 dropout_27" [label="(1, 197, 768)", style=solid]; +"591 dropout_27" -> "592 add_18" [label="(1, 197, 768)", style=solid]; +"592 add_18" -> "593 layer_norm_18" [label="(1, 197, 768)", style=solid]; +"592 add_18" -> "624 add_19" [label="(1, 197, 768)", style=solid]; +"593 layer_norm_18" -> "594 transpose_54" [label="(1, 197, 768)", style=solid]; +"594 transpose_54" -> "596 transpose_54_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"595 encoder_layers_encoder_layer_9_self_attention_in_proj_weight_updated_constant0" -> "597 linear_36" [label="(2304, 768)", style=solid]; +"596 transpose_54_0_0_nncf_smooth_quant_0" -> "597 linear_36" [label="(197, 1, 768)", style=solid]; +"597 linear_36" -> "598 unflatten_9" [label="(197, 1, 2304)", style=solid]; +"598 unflatten_9" -> "599 unsqueeze_9" [label="(197, 1, 3, 768)", style=solid]; +"599 unsqueeze_9" -> "600 transpose_55" [label="(1, 197, 1, 3, 768)", style=solid]; +"600 transpose_55" -> "601 squeeze_9" [label="(3, 197, 1, 1, 768)", style=solid]; +"601 squeeze_9" -> "602 contiguous_9" [label="(3, 197, 1, 768)", style=solid]; +"602 contiguous_9" -> "603 select_27" [label="(3, 197, 1, 768)", style=solid]; +"602 contiguous_9" -> "604 select_28" [label="(3, 197, 1, 768)", style=solid]; +"602 contiguous_9" -> "605 select_29" [label="(3, 197, 1, 768)", style=solid]; +"603 select_27" -> "606 view_72" [label="(197, 1, 768)", style=solid]; +"604 select_28" -> "608 view_73" [label="(197, 1, 768)", style=solid]; +"605 select_29" -> "610 view_74" [label="(197, 1, 768)", style=solid]; +"606 view_72" -> "607 transpose_56" [label="(197, 12, 64)", style=solid]; +"607 transpose_56" -> "612 view_75" [label="(12, 197, 64)", style=solid]; +"608 view_73" -> "609 transpose_57" [label="(197, 12, 64)", style=solid]; +"609 transpose_57" -> "613 view_76" [label="(12, 197, 64)", style=solid]; +"610 view_74" -> "611 transpose_58" [label="(197, 12, 64)", style=solid]; +"611 transpose_58" -> "614 view_77" [label="(12, 197, 64)", style=solid]; +"612 view_75" -> "615 scaled_dot_product_attention_9" [label="(1, 12, 197, 64)", style=solid]; +"613 view_76" -> "615 scaled_dot_product_attention_9" [label="(1, 12, 197, 64)", style=solid]; +"614 view_77" -> "615 scaled_dot_product_attention_9" [label="(1, 12, 197, 64)", style=solid]; +"615 scaled_dot_product_attention_9" -> "616 permute_10" [label="(1, 12, 197, 64)", style=solid]; +"616 permute_10" -> "617 view_78" [label="(197, 1, 12, 64)", style=solid]; +"617 view_78" -> "619 view_78_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"618 encoder_layers_encoder_layer_9_self_attention_out_proj_weight_updated_constant0" -> "620 linear_37" [label="(768, 768)", style=solid]; +"619 view_78_0_0_nncf_smooth_quant_0" -> "620 linear_37" [label="(197, 768)", style=solid]; +"620 linear_37" -> "621 view_79" [label="(197, 768)", style=solid]; +"621 view_79" -> "622 transpose_59" [label="(197, 1, 768)", style=solid]; +"622 transpose_59" -> "623 dropout_28" [label="(1, 197, 768)", style=solid]; +"623 dropout_28" -> "624 add_19" [label="(1, 197, 768)", style=solid]; +"624 add_19" -> "625 layer_norm_19" [label="(1, 197, 768)", style=solid]; +"624 add_19" -> "645 add_20" [label="(1, 197, 768)", style=solid]; +"625 layer_norm_19" -> "626 layer_norm_19_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"626 layer_norm_19_0_0_nncf_smooth_quant_0" -> "627 quantize_per_tensor_default_19" [label="(1, 197, 768)", style=solid]; +"627 quantize_per_tensor_default_19" -> "628 dequantize_per_tensor_default_19" [label="(1, 197, 768)", style=solid]; +"628 dequantize_per_tensor_default_19" -> "633 linear_38" [label="(1, 197, 768)", style=solid]; +"629 linear_38_scale_0" -> "632 dequantize_per_channel_default_19" [label="(3072,)", style=solid]; +"630 linear_38_zero_point_0" -> "632 dequantize_per_channel_default_19" [label="(3072,)", style=solid]; +"631 _frozen_param19" -> "632 dequantize_per_channel_default_19" [label="(3072, 768)", style=solid]; +"632 dequantize_per_channel_default_19" -> "633 linear_38" [label="(3072, 768)", style=solid]; +"633 linear_38" -> "634 gelu_9" [label="(1, 197, 3072)", style=solid]; +"634 gelu_9" -> "635 dropout_29" [label="(1, 197, 3072)", style=solid]; +"635 dropout_29" -> "636 dropout_29_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"636 dropout_29_0_0_nncf_smooth_quant_0" -> "637 quantize_per_tensor_default_20" [label="(1, 197, 3072)", style=solid]; +"637 quantize_per_tensor_default_20" -> "638 dequantize_per_tensor_default_20" [label="(1, 197, 3072)", style=solid]; +"638 dequantize_per_tensor_default_20" -> "643 linear_39" [label="(1, 197, 3072)", style=solid]; +"639 linear_39_scale_0" -> "642 dequantize_per_channel_default_20" [label="(768,)", style=solid]; +"640 linear_39_zero_point_0" -> "642 dequantize_per_channel_default_20" [label="(768,)", style=solid]; +"641 _frozen_param20" -> "642 dequantize_per_channel_default_20" [label="(768, 3072)", style=solid]; +"642 dequantize_per_channel_default_20" -> "643 linear_39" [label="(768, 3072)", style=solid]; +"643 linear_39" -> "644 dropout_30" [label="(1, 197, 768)", style=solid]; +"644 dropout_30" -> "645 add_20" [label="(1, 197, 768)", style=solid]; +"645 add_20" -> "646 layer_norm_20" [label="(1, 197, 768)", style=solid]; +"645 add_20" -> "677 add_21" [label="(1, 197, 768)", style=solid]; +"646 layer_norm_20" -> "647 transpose_60" [label="(1, 197, 768)", style=solid]; +"647 transpose_60" -> "649 transpose_60_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"648 encoder_layers_encoder_layer_10_self_attention_in_proj_weight_updated_constant0" -> "650 linear_40" [label="(2304, 768)", style=solid]; +"649 transpose_60_0_0_nncf_smooth_quant_0" -> "650 linear_40" [label="(197, 1, 768)", style=solid]; +"650 linear_40" -> "651 unflatten_10" [label="(197, 1, 2304)", style=solid]; +"651 unflatten_10" -> "652 unsqueeze_10" [label="(197, 1, 3, 768)", style=solid]; +"652 unsqueeze_10" -> "653 transpose_61" [label="(1, 197, 1, 3, 768)", style=solid]; +"653 transpose_61" -> "654 squeeze_10" [label="(3, 197, 1, 1, 768)", style=solid]; +"654 squeeze_10" -> "655 contiguous_10" [label="(3, 197, 1, 768)", style=solid]; +"655 contiguous_10" -> "656 select_30" [label="(3, 197, 1, 768)", style=solid]; +"655 contiguous_10" -> "657 select_31" [label="(3, 197, 1, 768)", style=solid]; +"655 contiguous_10" -> "658 select_32" [label="(3, 197, 1, 768)", style=solid]; +"656 select_30" -> "659 view_80" [label="(197, 1, 768)", style=solid]; +"657 select_31" -> "661 view_81" [label="(197, 1, 768)", style=solid]; +"658 select_32" -> "663 view_82" [label="(197, 1, 768)", style=solid]; +"659 view_80" -> "660 transpose_62" [label="(197, 12, 64)", style=solid]; +"660 transpose_62" -> "665 view_83" [label="(12, 197, 64)", style=solid]; +"661 view_81" -> "662 transpose_63" [label="(197, 12, 64)", style=solid]; +"662 transpose_63" -> "666 view_84" [label="(12, 197, 64)", style=solid]; +"663 view_82" -> "664 transpose_64" [label="(197, 12, 64)", style=solid]; +"664 transpose_64" -> "667 view_85" [label="(12, 197, 64)", style=solid]; +"665 view_83" -> "668 scaled_dot_product_attention_10" [label="(1, 12, 197, 64)", style=solid]; +"666 view_84" -> "668 scaled_dot_product_attention_10" [label="(1, 12, 197, 64)", style=solid]; +"667 view_85" -> "668 scaled_dot_product_attention_10" [label="(1, 12, 197, 64)", style=solid]; +"668 scaled_dot_product_attention_10" -> "669 permute_11" [label="(1, 12, 197, 64)", style=solid]; +"669 permute_11" -> "670 view_86" [label="(197, 1, 12, 64)", style=solid]; +"670 view_86" -> "672 view_86_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"671 encoder_layers_encoder_layer_10_self_attention_out_proj_weight_updated_constant0" -> "673 linear_41" [label="(768, 768)", style=solid]; +"672 view_86_0_0_nncf_smooth_quant_0" -> "673 linear_41" [label="(197, 768)", style=solid]; +"673 linear_41" -> "674 view_87" [label="(197, 768)", style=solid]; +"674 view_87" -> "675 transpose_65" [label="(197, 1, 768)", style=solid]; +"675 transpose_65" -> "676 dropout_31" [label="(1, 197, 768)", style=solid]; +"676 dropout_31" -> "677 add_21" [label="(1, 197, 768)", style=solid]; +"677 add_21" -> "678 layer_norm_21" [label="(1, 197, 768)", style=solid]; +"677 add_21" -> "698 add_22" [label="(1, 197, 768)", style=solid]; +"678 layer_norm_21" -> "679 layer_norm_21_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"679 layer_norm_21_0_0_nncf_smooth_quant_0" -> "680 quantize_per_tensor_default_21" [label="(1, 197, 768)", style=solid]; +"680 quantize_per_tensor_default_21" -> "681 dequantize_per_tensor_default_21" [label="(1, 197, 768)", style=solid]; +"681 dequantize_per_tensor_default_21" -> "686 linear_42" [label="(1, 197, 768)", style=solid]; +"682 linear_42_scale_0" -> "685 dequantize_per_channel_default_21" [label="(3072,)", style=solid]; +"683 linear_42_zero_point_0" -> "685 dequantize_per_channel_default_21" [label="(3072,)", style=solid]; +"684 _frozen_param21" -> "685 dequantize_per_channel_default_21" [label="(3072, 768)", style=solid]; +"685 dequantize_per_channel_default_21" -> "686 linear_42" [label="(3072, 768)", style=solid]; +"686 linear_42" -> "687 gelu_10" [label="(1, 197, 3072)", style=solid]; +"687 gelu_10" -> "688 dropout_32" [label="(1, 197, 3072)", style=solid]; +"688 dropout_32" -> "689 dropout_32_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"689 dropout_32_0_0_nncf_smooth_quant_0" -> "690 quantize_per_tensor_default_22" [label="(1, 197, 3072)", style=solid]; +"690 quantize_per_tensor_default_22" -> "691 dequantize_per_tensor_default_22" [label="(1, 197, 3072)", style=solid]; +"691 dequantize_per_tensor_default_22" -> "696 linear_43" [label="(1, 197, 3072)", style=solid]; +"692 linear_43_scale_0" -> "695 dequantize_per_channel_default_22" [label="(768,)", style=solid]; +"693 linear_43_zero_point_0" -> "695 dequantize_per_channel_default_22" [label="(768,)", style=solid]; +"694 _frozen_param22" -> "695 dequantize_per_channel_default_22" [label="(768, 3072)", style=solid]; +"695 dequantize_per_channel_default_22" -> "696 linear_43" [label="(768, 3072)", style=solid]; +"696 linear_43" -> "697 dropout_33" [label="(1, 197, 768)", style=solid]; +"697 dropout_33" -> "698 add_22" [label="(1, 197, 768)", style=solid]; +"698 add_22" -> "699 layer_norm_22" [label="(1, 197, 768)", style=solid]; +"698 add_22" -> "730 add_23" [label="(1, 197, 768)", style=solid]; +"699 layer_norm_22" -> "700 transpose_66" [label="(1, 197, 768)", style=solid]; +"700 transpose_66" -> "702 transpose_66_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"701 encoder_layers_encoder_layer_11_self_attention_in_proj_weight_updated_constant0" -> "703 linear_44" [label="(2304, 768)", style=solid]; +"702 transpose_66_0_0_nncf_smooth_quant_0" -> "703 linear_44" [label="(197, 1, 768)", style=solid]; +"703 linear_44" -> "704 unflatten_11" [label="(197, 1, 2304)", style=solid]; +"704 unflatten_11" -> "705 unsqueeze_11" [label="(197, 1, 3, 768)", style=solid]; +"705 unsqueeze_11" -> "706 transpose_67" [label="(1, 197, 1, 3, 768)", style=solid]; +"706 transpose_67" -> "707 squeeze_11" [label="(3, 197, 1, 1, 768)", style=solid]; +"707 squeeze_11" -> "708 contiguous_11" [label="(3, 197, 1, 768)", style=solid]; +"708 contiguous_11" -> "709 select_33" [label="(3, 197, 1, 768)", style=solid]; +"708 contiguous_11" -> "710 select_34" [label="(3, 197, 1, 768)", style=solid]; +"708 contiguous_11" -> "711 select_35" [label="(3, 197, 1, 768)", style=solid]; +"709 select_33" -> "712 view_88" [label="(197, 1, 768)", style=solid]; +"710 select_34" -> "714 view_89" [label="(197, 1, 768)", style=solid]; +"711 select_35" -> "716 view_90" [label="(197, 1, 768)", style=solid]; +"712 view_88" -> "713 transpose_68" [label="(197, 12, 64)", style=solid]; +"713 transpose_68" -> "718 view_91" [label="(12, 197, 64)", style=solid]; +"714 view_89" -> "715 transpose_69" [label="(197, 12, 64)", style=solid]; +"715 transpose_69" -> "719 view_92" [label="(12, 197, 64)", style=solid]; +"716 view_90" -> "717 transpose_70" [label="(197, 12, 64)", style=solid]; +"717 transpose_70" -> "720 view_93" [label="(12, 197, 64)", style=solid]; +"718 view_91" -> "721 scaled_dot_product_attention_11" [label="(1, 12, 197, 64)", style=solid]; +"719 view_92" -> "721 scaled_dot_product_attention_11" [label="(1, 12, 197, 64)", style=solid]; +"720 view_93" -> "721 scaled_dot_product_attention_11" [label="(1, 12, 197, 64)", style=solid]; +"721 scaled_dot_product_attention_11" -> "722 permute_12" [label="(1, 12, 197, 64)", style=solid]; +"722 permute_12" -> "723 view_94" [label="(197, 1, 12, 64)", style=solid]; +"723 view_94" -> "725 view_94_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"724 encoder_layers_encoder_layer_11_self_attention_out_proj_weight_updated_constant0" -> "726 linear_45" [label="(768, 768)", style=solid]; +"725 view_94_0_0_nncf_smooth_quant_0" -> "726 linear_45" [label="(197, 768)", style=solid]; +"726 linear_45" -> "727 view_95" [label="(197, 768)", style=solid]; +"727 view_95" -> "728 transpose_71" [label="(197, 1, 768)", style=solid]; +"728 transpose_71" -> "729 dropout_34" [label="(1, 197, 768)", style=solid]; +"729 dropout_34" -> "730 add_23" [label="(1, 197, 768)", style=solid]; +"730 add_23" -> "731 layer_norm_23" [label="(1, 197, 768)", style=solid]; +"730 add_23" -> "751 add_24" [label="(1, 197, 768)", style=solid]; +"731 layer_norm_23" -> "732 layer_norm_23_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"732 layer_norm_23_0_0_nncf_smooth_quant_0" -> "733 quantize_per_tensor_default_23" [label="(1, 197, 768)", style=solid]; +"733 quantize_per_tensor_default_23" -> "734 dequantize_per_tensor_default_23" [label="(1, 197, 768)", style=solid]; +"734 dequantize_per_tensor_default_23" -> "739 linear_46" [label="(1, 197, 768)", style=solid]; +"735 linear_46_scale_0" -> "738 dequantize_per_channel_default_23" [label="(3072,)", style=solid]; +"736 linear_46_zero_point_0" -> "738 dequantize_per_channel_default_23" [label="(3072,)", style=solid]; +"737 _frozen_param23" -> "738 dequantize_per_channel_default_23" [label="(3072, 768)", style=solid]; +"738 dequantize_per_channel_default_23" -> "739 linear_46" [label="(3072, 768)", style=solid]; +"739 linear_46" -> "740 gelu_11" [label="(1, 197, 3072)", style=solid]; +"740 gelu_11" -> "741 dropout_35" [label="(1, 197, 3072)", style=solid]; +"741 dropout_35" -> "742 dropout_35_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"742 dropout_35_0_0_nncf_smooth_quant_0" -> "743 quantize_per_tensor_default_24" [label="(1, 197, 3072)", style=solid]; +"743 quantize_per_tensor_default_24" -> "744 dequantize_per_tensor_default_24" [label="(1, 197, 3072)", style=solid]; +"744 dequantize_per_tensor_default_24" -> "749 linear_47" [label="(1, 197, 3072)", style=solid]; +"745 linear_47_scale_0" -> "748 dequantize_per_channel_default_24" [label="(768,)", style=solid]; +"746 linear_47_zero_point_0" -> "748 dequantize_per_channel_default_24" [label="(768,)", style=solid]; +"747 _frozen_param24" -> "748 dequantize_per_channel_default_24" [label="(768, 3072)", style=solid]; +"748 dequantize_per_channel_default_24" -> "749 linear_47" [label="(768, 3072)", style=solid]; +"749 linear_47" -> "750 dropout_36" [label="(1, 197, 768)", style=solid]; +"750 dropout_36" -> "751 add_24" [label="(1, 197, 768)", style=solid]; +"751 add_24" -> "752 layer_norm_24" [label="(1, 197, 768)", style=solid]; +"752 layer_norm_24" -> "753 slice_1" [label="(1, 197, 768)", style=solid]; +"753 slice_1" -> "754 select_36" [label="(1, 197, 768)", style=solid]; +"754 select_36" -> "755 select_36_0_0_nncf_smooth_quant_0" [label="(1, 768)", style=solid]; +"755 select_36_0_0_nncf_smooth_quant_0" -> "756 quantize_per_tensor_default_25" [label="(1, 768)", style=solid]; +"756 quantize_per_tensor_default_25" -> "757 dequantize_per_tensor_default_25" [label="(1, 768)", style=solid]; +"757 dequantize_per_tensor_default_25" -> "762 linear_48" [label="(1, 768)", style=solid]; +"758 linear_48_scale_0" -> "761 dequantize_per_channel_default_25" [label="(1000,)", style=solid]; +"759 linear_48_zero_point_0" -> "761 dequantize_per_channel_default_25" [label="(1000,)", style=solid]; +"760 _frozen_param25" -> "761 dequantize_per_channel_default_25" [label="(1000, 768)", style=solid]; +"761 dequantize_per_channel_default_25" -> "762 linear_48" [label="(1000, 768)", style=solid]; +"762 linear_48" -> "763 output" [label="(1, 1000)", style=solid]; +} diff --git a/tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/yolo11n_sdpa_block.dot b/tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/yolo11n_sdpa_block.dot new file mode 100644 index 00000000000..17951540020 --- /dev/null +++ b/tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/yolo11n_sdpa_block.dot @@ -0,0 +1,53 @@ +strict digraph { +"0 x" [id=0, type=input]; +"1 x_0_0_nncf_smooth_quant_0" [id=1, type=call_module]; +"2 quantize_per_tensor_default" [id=2, type=quantize_per_tensor]; +"3 dequantize_per_tensor_default" [id=3, type=dequantize_per_tensor]; +"4 linear_scale_0" [id=4, type=get_attr]; +"5 linear_zero_point_0" [id=5, type=get_attr]; +"6 _frozen_param0" [id=6, type=get_attr]; +"7 dequantize_per_channel_default" [id=7, type=dequantize_per_channel]; +"8 linear" [id=8, type=linear]; +"9 slice_1" [id=9, type=slice]; +"10 slice_2" [id=10, type=slice]; +"11 slice_3" [id=11, type=slice]; +"12 slice_4" [id=12, type=slice]; +"13 slice_5" [id=13, type=slice]; +"14 slice_6" [id=14, type=slice]; +"15 slice_7" [id=15, type=slice]; +"16 slice_8" [id=16, type=slice]; +"17 slice_9" [id=17, type=slice]; +"18 transpose" [id=18, type=transpose]; +"19 matmul" [id=19, type=matmul]; +"20 div_" [id=20, type=div_]; +"21 softmax" [id=21, type=softmax]; +"22 transpose_1" [id=22, type=transpose]; +"23 matmul_1" [id=23, type=matmul]; +"24 output" [id=24, type=output]; +"0 x" -> "1 x_0_0_nncf_smooth_quant_0" [label="(1, 2, 4)", style=solid]; +"1 x_0_0_nncf_smooth_quant_0" -> "2 quantize_per_tensor_default" [label="(1, 2, 4)", style=solid]; +"2 quantize_per_tensor_default" -> "3 dequantize_per_tensor_default" [label="(1, 2, 4)", style=solid]; +"3 dequantize_per_tensor_default" -> "8 linear" [label="(1, 2, 4)", style=solid]; +"4 linear_scale_0" -> "7 dequantize_per_channel_default" [label="(12,)", style=solid]; +"5 linear_zero_point_0" -> "7 dequantize_per_channel_default" [label="(12,)", style=solid]; +"6 _frozen_param0" -> "7 dequantize_per_channel_default" [label="(12, 4)", style=solid]; +"7 dequantize_per_channel_default" -> "8 linear" [label="(12, 4)", style=solid]; +"8 linear" -> "9 slice_1" [label="(1, 2, 12)", style=solid]; +"8 linear" -> "12 slice_4" [label="(1, 2, 12)", style=solid]; +"8 linear" -> "15 slice_7" [label="(1, 2, 12)", style=solid]; +"9 slice_1" -> "10 slice_2" [label="(1, 2, 12)", style=solid]; +"10 slice_2" -> "11 slice_3" [label="(1, 2, 12)", style=solid]; +"11 slice_3" -> "19 matmul" [label="(1, 2, 4)", style=solid]; +"12 slice_4" -> "13 slice_5" [label="(1, 2, 12)", style=solid]; +"13 slice_5" -> "14 slice_6" [label="(1, 2, 12)", style=solid]; +"14 slice_6" -> "18 transpose" [label="(1, 2, 4)", style=solid]; +"15 slice_7" -> "16 slice_8" [label="(1, 2, 12)", style=solid]; +"16 slice_8" -> "17 slice_9" [label="(1, 2, 12)", style=solid]; +"17 slice_9" -> "23 matmul_1" [label="(1, 2, 4)", style=solid]; +"18 transpose" -> "19 matmul" [label="(1, 4, 2)", style=solid]; +"19 matmul" -> "20 div_" [label="(1, 2, 2)", style=solid]; +"20 div_" -> "21 softmax" [label="(1, 2, 2)", style=solid]; +"21 softmax" -> "22 transpose_1" [label="(1, 2, 2)", style=solid]; +"22 transpose_1" -> "23 matmul_1" [label="(1, 2, 2)", style=solid]; +"23 matmul_1" -> "24 output" [label="(1, 2, 4)", style=solid]; +} diff --git a/tests/torch/fx/test_quantizer.py b/tests/torch/fx/test_quantizer.py new file mode 100644 index 00000000000..f95f59a4341 --- /dev/null +++ b/tests/torch/fx/test_quantizer.py @@ -0,0 +1,162 @@ +# Copyright (c) 2024 Intel Corporation +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# http://www.apache.org/licenses/LICENSE-2.0 +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from dataclasses import dataclass +from functools import partial +from pathlib import Path +from typing import Any, Callable, Tuple + +import pytest +import torch +import torch.fx +import torch.nn.parallel +import torch.optim +import torch.utils.data +import torch.utils.data.distributed +import torchvision.models as models +from torch.ao.quantization.quantizer.quantizer import Quantizer +from torch.ao.quantization.quantizer.x86_inductor_quantizer import X86InductorQuantizer +from torch.ao.quantization.quantizer.x86_inductor_quantizer import get_default_x86_inductor_quantization_config + +import nncf +from nncf.experimental.torch.fx.nncf_graph_builder import GraphConverter +from nncf.experimental.torch.fx.quantization.quantize_pt2e import quantize_pt2e +from tests.torch import test_models +from tests.torch.fx.helpers import get_torch_fx_model +from tests.torch.test_compressed_graph import check_graph +from tests.torch.test_models.synthetic import ShortTransformer +from tests.torch.test_models.synthetic import YOLO11N_SDPABlock + +FX_QUANTIZED_DIR_NAME = Path("fx") / "experimental" + + +@dataclass +class ModelCase: + model_builder: Callable[[], torch.nn.Module] + model_id: str + input_shape: Tuple[int] + + +def torchvision_model_case(model_id: str, input_shape: Tuple[int,]): + model = getattr(models, model_id) + return ModelCase(partial(model, weights=None), model_id, input_shape) + + +TEST_MODELS = ( + torchvision_model_case("resnet18", (1, 3, 224, 224)), + torchvision_model_case("mobilenet_v3_small", (1, 3, 224, 224)), + torchvision_model_case("vit_b_16", (1, 3, 224, 224)), + torchvision_model_case("swin_v2_s", (1, 3, 224, 224)), + ModelCase(test_models.UNet, "unet", [1, 3, 224, 224]), + ModelCase(partial(ShortTransformer, 5, 10), "synthetic_transformer", [5]), + ModelCase(YOLO11N_SDPABlock, "yolo11n_sdpa_block", YOLO11N_SDPABlock.INPUT_SIZE), +) + + +def get_dot_filename(model_name): + return model_name + ".dot" + + +def get_x86_quantizer(*args, **kwarsg) -> X86InductorQuantizer: + quantizer = X86InductorQuantizer() + quantizer.set_global(get_default_x86_inductor_quantization_config()) + return quantizer + + +TEST_MODELS_QUANIZED = ( + (ModelCase(test_models.UNet, "unet", [1, 3, 224, 224]), {}, {}), + (torchvision_model_case("resnet18", (1, 3, 224, 224)), {}, {}), + (torchvision_model_case("mobilenet_v3_small", (1, 3, 224, 224)), {}, {}), + ( + torchvision_model_case("vit_b_16", (1, 3, 224, 224)), + {"model_type": nncf.ModelType.TRANSFORMER}, + {"smooth_quant": True}, + ), + ( + torchvision_model_case("swin_v2_s", (1, 3, 224, 224)), + {"model_type": nncf.ModelType.TRANSFORMER}, + {"smooth_quant": True}, + ), + ( + ModelCase(partial(ShortTransformer, 5, 10), "synthetic_transformer", [5]), + {"model_type": nncf.ModelType.TRANSFORMER}, + {"smooth_quant": True}, + ), + ( + ModelCase(YOLO11N_SDPABlock, "yolo11n_sdpa_block", YOLO11N_SDPABlock.INPUT_SIZE), + {"model_type": nncf.ModelType.TRANSFORMER}, + {"smooth_quant": True}, + ), +) + + +@pytest.mark.parametrize("quantizer_builder", [get_x86_quantizer]) +class TestModelQuantization: + @staticmethod + def _build_torch_fx_model(model_case: ModelCase) -> Tuple[torch.fx.GraphModule, torch.Tensor]: + model = model_case.model_builder() + dtype = torch.int32 if model_case.model_id == "synthetic_transformer" else torch.float32 + example_input = torch.ones(model_case.input_shape, dtype=dtype) + fx_model = get_torch_fx_model(model, example_input) + return fx_model, example_input + + @staticmethod + def _get_calibration_dataset(example_input: torch.Tensor) -> nncf.Dataset: + def transform_fn(data_item): + return data_item.to("cpu") + + return nncf.Dataset([example_input], transform_fn) + + @pytest.mark.parametrize( + ("model_case", "quantizer_params", "pt2e_params"), + TEST_MODELS_QUANIZED, + ids=[m[0].model_id for m in TEST_MODELS_QUANIZED], + ) + def test_quantized_model( + self, + quantizer_builder: Callable[[Tuple[Any, ...]], Quantizer], + model_case: ModelCase, + quantizer_params, + pt2e_params, + ): + fx_model, example_input = self._build_torch_fx_model(model_case) + calibration_dataset = self._get_calibration_dataset(example_input) + + quantizer = quantizer_builder(**quantizer_params) + quantized_model = quantize_pt2e( + fx_model, + quantizer, + calibration_dataset=calibration_dataset, + fast_bias_correction=None, # BC is disabled + fold_quantize=True, + **pt2e_params, + ) + + # Uncomment to visualize torch fx graph + # from tests.torch.fx.helpers import visualize_fx_model + # visualize_fx_model(quantized_model, f"{model_case.model_id}_int8.svg") + + nncf_graph = GraphConverter.create_nncf_graph(quantized_model) + check_graph( + nncf_graph, + get_dot_filename(model_case.model_id), + FX_QUANTIZED_DIR_NAME / quantizer.__class__.__name__, + extended=True, + ) + + # Uncomment to visualize reference graphs + # from torch.ao.quantization.quantize_pt2e import convert_pt2e + # from torch.ao.quantization.quantize_pt2e import prepare_pt2e + # prepared_model = prepare_pt2e(fx_model, quantizer) + # prepared_model(example_input) + # ao_quantized_model = convert_pt2e(prepared_model) + # ao_nncf_graph = GraphConverter.create_nncf_graph(ao_quantized_model) + # ao_nncf_graph.visualize_graph("ao_" + get_dot_filename(model_case.model_id)) diff --git a/torch_fx_experimental_q.py b/torch_fx_experimental_q.py deleted file mode 100644 index 77ca8859442..00000000000 --- a/torch_fx_experimental_q.py +++ /dev/null @@ -1,107 +0,0 @@ -# Copyright (c) 2024 Intel Corporation -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# http://www.apache.org/licenses/LICENSE-2.0 -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. - -import os - -os.environ["TORCHINDUCTOR_FREEZING"] = "1" - -from time import time - -import torch -import torch.fx -from torch._export import capture_pre_autograd_graph -from torch.ao.quantization.quantize_pt2e import convert_pt2e -from torch.ao.quantization.quantize_pt2e import prepare_pt2e -from torch.ao.quantization.quantizer.x86_inductor_quantizer import X86InductorQuantizer -from torch.ao.quantization.quantizer.x86_inductor_quantizer import get_default_x86_inductor_quantization_config -from torchvision import models - -import nncf -import nncf.torch -from nncf.experimental.torch.fx.quantization.quantize_pt2e import quantize_pt2e -from tests.torch.fx.helpers import visualize_fx_model - - -def measure_time(model, example_inputs, num_iters=3000): - with torch.no_grad(): - model(*example_inputs) - total_time = 0 - for _ in range(num_iters): - start_time = time() - model(*example_inputs) - total_time += time() - start_time - average_time = (total_time / num_iters) * 1000 - return average_time - - -def main(model_cls): - model = model_cls() - example_inputs = torch.ones((1, 3, 224, 224)) - exported_model = capture_pre_autograd_graph(model.eval(), (example_inputs,)) - - quantizer = X86InductorQuantizer() - quantizer.set_global(get_default_x86_inductor_quantization_config()) - - nncf_quantizer_model = quantize_pt2e(exported_model, quantizer, calibration_dataset=nncf.Dataset([example_inputs])) - - visualize_fx_model(nncf_quantizer_model, "nncf_quantizer_before_fold_resnet.svg") - return nncf_quantizer_model - - # exported_model = capture_pre_autograd_graph(model.eval(), (example_inputs,)) - # nncf_int8 = nncf.quantize(exported_model, nncf.Dataset([example_inputs])) - # visualize_fx_model(nncf_int8, "nncf_resnet.svg") - - -def main_native(model_cls): - model = model_cls() - example_inputs = torch.ones((1, 3, 224, 224)) - exported_model = capture_pre_autograd_graph(model.eval(), (example_inputs,)) - - quantizer = X86InductorQuantizer() - quantizer.set_global(get_default_x86_inductor_quantization_config()) - - prepared_model = prepare_pt2e(exported_model, quantizer) - prepared_model(example_inputs) - converted_model = convert_pt2e(prepared_model) - visualize_fx_model(converted_model, "x86int8_resnet.svg") - return converted_model - - -def constant_fold(m): - pass - - -if __name__ == "__main__": - with nncf.torch.disable_patching(): - for model_cls in (models.resnet18, models.mobilenet_v3_small, models.vit_b_16, models.swin_v2_s): - # for model_cls in (models.mobilenet_v3_small,): - print(f"{model_cls} check!") - nncf_q_model = main(model_cls) - - constant_fold(nncf_q_model) - visualize_fx_model(nncf_q_model, "nncf_quantizer_after_constant_fold_resnet.svg") - - pt_q_model = main_native(model_cls) - print("benchmarking...") - pt_compiled = torch.compile(model_cls()) - pt_int8_compiled = torch.compile(pt_q_model) - nncf_comipled = torch.compile(nncf_q_model) - - example_inputs = (torch.ones((1, 3, 224, 224)),) - - pt_time = measure_time(pt_compiled, example_inputs) - print(f"PT fp32 performance measured: {pt_time}") - - pt_int8_time = measure_time(pt_int8_compiled, example_inputs) - print(f"PT int8 performance measured: {pt_int8_time}") - - nncf_int8_time = measure_time(nncf_comipled, example_inputs) - print(f"NNCF int8 performance measured: {nncf_int8_time}") From 8e001e346e80b2de051bf26897a3602e73472c23 Mon Sep 17 00:00:00 2001 From: dlyakhov Date: Mon, 2 Dec 2024 14:10:00 +0100 Subject: [PATCH 03/15] Reuse MinMax algo instead of copy-paste --- .../algorithms/post_training/algorithm.py | 2 +- .../algorithms/post_training/pipeline.py | 2 +- .../{quantizer.py => base_quantizer.py} | 6 +- .../algorithms/quantizer/fx_quantizer.py | 2 +- .../algorithms/range_estimator/backend.py | 154 ------ .../range_estimator/range_estimator.py | 452 +----------------- .../range_estimator/torch_fx_backend.py | 221 --------- .../algorithms/min_max/algorithm.py | 40 +- tests/common/quantization/test_minmax.py | 21 +- 9 files changed, 77 insertions(+), 823 deletions(-) rename nncf/experimental/common/quantization/algorithms/quantizer/{quantizer.py => base_quantizer.py} (77%) delete mode 100644 nncf/experimental/common/quantization/algorithms/range_estimator/backend.py delete mode 100644 nncf/experimental/common/quantization/algorithms/range_estimator/torch_fx_backend.py diff --git a/nncf/experimental/common/quantization/algorithms/post_training/algorithm.py b/nncf/experimental/common/quantization/algorithms/post_training/algorithm.py index 17558fbb1db..efbc5bb7449 100644 --- a/nncf/experimental/common/quantization/algorithms/post_training/algorithm.py +++ b/nncf/experimental/common/quantization/algorithms/post_training/algorithm.py @@ -17,7 +17,7 @@ from nncf.common.tensor_statistics.statistic_point import StatisticPointsContainer from nncf.common.utils.backend import BackendType from nncf.experimental.common.quantization.algorithms.post_training.pipeline import experimental_create_ptq_pipeline -from nncf.experimental.common.quantization.algorithms.quantizer.quantizer import NNCFQuantizer +from nncf.experimental.common.quantization.algorithms.quantizer.base_quantizer import NNCFQuantizer from nncf.quantization.advanced_parameters import AdvancedBiasCorrectionParameters from nncf.quantization.advanced_parameters import AdvancedSmoothQuantParameters from nncf.quantization.advanced_parameters import RangeEstimatorParameters diff --git a/nncf/experimental/common/quantization/algorithms/post_training/pipeline.py b/nncf/experimental/common/quantization/algorithms/post_training/pipeline.py index de496eba2b3..dab7b2be856 100644 --- a/nncf/experimental/common/quantization/algorithms/post_training/pipeline.py +++ b/nncf/experimental/common/quantization/algorithms/post_training/pipeline.py @@ -11,7 +11,7 @@ from typing import Optional, TypeVar -from nncf.experimental.common.quantization.algorithms.quantizer.quantizer import NNCFQuantizer +from nncf.experimental.common.quantization.algorithms.quantizer.base_quantizer import NNCFQuantizer from nncf.experimental.common.quantization.algorithms.range_estimator.range_estimator import MinMaxRangeEstimator from nncf.quantization.advanced_parameters import AdvancedBiasCorrectionParameters from nncf.quantization.advanced_parameters import AdvancedSmoothQuantParameters diff --git a/nncf/experimental/common/quantization/algorithms/quantizer/quantizer.py b/nncf/experimental/common/quantization/algorithms/quantizer/base_quantizer.py similarity index 77% rename from nncf/experimental/common/quantization/algorithms/quantizer/quantizer.py rename to nncf/experimental/common/quantization/algorithms/quantizer/base_quantizer.py index b0d40234210..f7a4d6018ba 100644 --- a/nncf/experimental/common/quantization/algorithms/quantizer/quantizer.py +++ b/nncf/experimental/common/quantization/algorithms/quantizer/base_quantizer.py @@ -22,5 +22,9 @@ class NNCFQuantizer: @abstractmethod def get_quantization_setup(self, model: TModel, nncf_graph: NNCFGraph) -> SingleConfigQuantizerSetup: """ - Return quantization setup. + Builds SingleConfigQuantizerSetup for the given model. + + :param model: Backend-specific model, for which Quantization Target Points are being seek. + :param nncf_graph: NNCFGraph instance. + :return: SingleConfigQuantizerSetup for the given model. """ diff --git a/nncf/experimental/common/quantization/algorithms/quantizer/fx_quantizer.py b/nncf/experimental/common/quantization/algorithms/quantizer/fx_quantizer.py index 7842e7475f0..db0ae167132 100644 --- a/nncf/experimental/common/quantization/algorithms/quantizer/fx_quantizer.py +++ b/nncf/experimental/common/quantization/algorithms/quantizer/fx_quantizer.py @@ -29,7 +29,7 @@ from nncf.common.quantization.quantizer_setup import WeightQuantizationInsertionPoint from nncf.common.quantization.structs import QuantizationScheme as QuantizationMode from nncf.common.quantization.structs import QuantizerConfig -from nncf.experimental.common.quantization.algorithms.quantizer.quantizer import NNCFQuantizer +from nncf.experimental.common.quantization.algorithms.quantizer.base_quantizer import NNCFQuantizer EdgeOrNode = Union[Tuple[torch.fx.Node, torch.fx.Node]] diff --git a/nncf/experimental/common/quantization/algorithms/range_estimator/backend.py b/nncf/experimental/common/quantization/algorithms/range_estimator/backend.py deleted file mode 100644 index dbd11f3f6b7..00000000000 --- a/nncf/experimental/common/quantization/algorithms/range_estimator/backend.py +++ /dev/null @@ -1,154 +0,0 @@ -# Copyright (c) 2024 Intel Corporation -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# http://www.apache.org/licenses/LICENSE-2.0 -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. - -from abc import ABC -from abc import abstractmethod -from typing import List, Optional, Set, Tuple, TypeVar - -from nncf.common.graph.graph import NNCFGraph -from nncf.common.graph.graph import NNCFNode -from nncf.common.graph.transformations.commands import TargetPoint -from nncf.common.graph.transformations.commands import TargetType -from nncf.common.graph.transformations.commands import TransformationCommand -from nncf.common.quantization.structs import QuantizerConfig -from nncf.common.tensor_statistics.collectors import TensorStatisticCollectorBase -from nncf.quantization.fake_quantize import FakeQuantizeParameters -from nncf.quantization.range_estimator import RangeEstimatorParameters - -TModel = TypeVar("TModel") - - -class RangeEstimatorAlgoBackend(ABC): - @staticmethod - @abstractmethod - def target_point(target_type: TargetType, target_node_name: str, port_id: int) -> TargetPoint: - """ - Returns backend-specific target point. - - :param target_type: Type of the location that should be modified. - :param target_node_name: Name of the located node. - :param port_id: Port ID of the tensor for the statistics distribution. - :return: Backend-specific TargetPoint. - """ - - @staticmethod - @abstractmethod - def create_quantizer_insertion_command( - nncf_graph: NNCFGraph, - target_point: TargetPoint, - quantizer_config: QuantizerConfig, - parameters: FakeQuantizeParameters, - ) -> TransformationCommand: - """ - Returns backend-specific quantizer insertion command. - - :param nncf_graph: NNCFGraph to get input/output shapes for the target point. - :param target_point: Target location for the quantizer insertion. - :param quantizer_config: QuantizerConfig instance for the current layer. - :param parameters: FakeQuantizeParameters to calculate activation quantization parameters. - :return: Backend-specific TransformationCommand for the quantizer insertion operation. - """ - - @staticmethod - @abstractmethod - def create_unified_scales_quantizers_insertion_commands( - nncf_graph: NNCFGraph, - target_points: List[TargetPoint], - quantizer_config: QuantizerConfig, - parameters: FakeQuantizeParameters, - ) -> List[TransformationCommand]: - """ - Returns backend-specific unified scales quantizers insertion commands. - - :param nncf_graph: NNCFGraph to get input/output shapes for the target point. - :param target_points: List of target locations for the quantizers insertion. - :param quantizer_config: QuantizerConfig instance for the current layer. - :param parameters: FakeQuantizeParameters to calculate activation quantization parameters. - :return: List of backend-specific TransformationCommands - for the quantizers with unified scales insertion operations. - """ - - @staticmethod - @abstractmethod - def get_target_point_shape(nncf_graph: NNCFGraph, node: NNCFNode, target_point: TargetPoint) -> Tuple[int, ...]: - """ - Returns shape of a target point tensor. - - :param nncf_graph: NNCFGraph instance. - :param node: NNCFNode. - :param target_point: Target point of which tensor shape is seeked. - :return: Shape of target point tensor. - """ - - @staticmethod - @abstractmethod - def get_weight_quantization_axes(node: NNCFNode, target_point: TargetPoint, ndims: int) -> Tuple[int, ...]: - """ - Returns axes for per-channel quantization of weights of the node placed on a input port_id. - - :param node: Quantized node with the weight. - :param target_point: Corresponding target point. - :param ndims: Number of dimensions of weight. - :return: Axes for per-channel quantization of weights. - """ - - @staticmethod - @abstractmethod - def get_statistic_collector( - range_estimator_params: RangeEstimatorParameters, - use_abs_max: bool, - reduction_axes: Optional[Tuple[int, ...]], - aggregation_axes: Optional[Tuple[int, ...]], - inplace: bool, - num_samples: Optional[int] = None, - ) -> TensorStatisticCollectorBase: - """ - Returns backend-specific statistic collector. - - :param range_estimator_params: Parameters that specify estimators types. - :param use_abs_max: Wheather reduce absolute values of input tensors or not. - :param reduction_axes: Axes for reducer. - :param aggregation_axes: Axes for aggregator. - :param inplace: Whether to calculate statistic inplace or not. - :param num_samples: Maximum number of samples to collect. - :return: Backend-specific TensorStatisticCollectorBase for the statistics calculation. - """ - - @staticmethod - @abstractmethod - def get_weight_tensor_port_ids(node: NNCFNode, graph: NNCFGraph) -> List[Optional[int]]: - """ - Returns node's input port indices with weight tensors. - - :param node: NNCFNode to find its weight input port indices. - :param graph: NNCFGraph instance. - :return: Weights input port indices. - """ - - @staticmethod - def get_weight_name(nncf_graph: NNCFGraph, target_point: TargetPoint) -> str: - """ - Returns node's weight name corresponding to port ID. - - :param nncf_graph: NNCFGraph instance. - :param target_point: The TargetPoint instance that contains layer's information. - :return: Weight name. - """ - - @staticmethod - def should_quantize_weight(weight_name: str, quantized_weight_names: Set[str]) -> bool: - """ - Return True if weight should be quantized. - - :param weight_name: Weight name. - :param quantized_weight_names: Set containing already quantized weight names. - :return: A boolean value specifying whether a weight should be quantized. - """ diff --git a/nncf/experimental/common/quantization/algorithms/range_estimator/range_estimator.py b/nncf/experimental/common/quantization/algorithms/range_estimator/range_estimator.py index 5431703cdb1..1b5ad8c5692 100644 --- a/nncf/experimental/common/quantization/algorithms/range_estimator/range_estimator.py +++ b/nncf/experimental/common/quantization/algorithms/range_estimator/range_estimator.py @@ -9,38 +9,16 @@ # See the License for the specific language governing permissions and # limitations under the License. -import collections -import dataclasses -from copy import deepcopy -from typing import List, Optional, OrderedDict, Tuple, TypeVar +from typing import List, Optional, TypeVar -import nncf -import nncf.tensor.functions as fns from nncf import Dataset -from nncf.common.factory import ModelTransformerFactory from nncf.common.graph.graph import NNCFGraph -from nncf.common.graph.transformations.commands import TargetPoint -from nncf.common.graph.transformations.commands import TargetType -from nncf.common.graph.transformations.layout import TransformationLayout -from nncf.common.logging import nncf_logger -from nncf.common.quantization.initialization.range import RangeInitCollectorParams -from nncf.common.quantization.quantizer_setup import SingleConfigQuantizationPoint -from nncf.common.quantization.quantizer_setup import SingleConfigQuantizerSetup -from nncf.common.quantization.structs import QuantizerConfig -from nncf.common.quantization.structs import QuantizerGroup -from nncf.common.tensor_statistics.collectors import TensorStatisticCollectorBase -from nncf.common.tensor_statistics.statistic_point import StatisticPoint from nncf.common.tensor_statistics.statistic_point import StatisticPointsContainer from nncf.common.utils.backend import BackendType -from nncf.common.utils.backend import get_backend -from nncf.experimental.common.quantization.algorithms.quantizer.quantizer import NNCFQuantizer -from nncf.experimental.common.tensor_statistics.statistics import MinMaxTensorStatistic -from nncf.quantization.advanced_parameters import changes_asdict +from nncf.experimental.common.quantization.algorithms.quantizer.base_quantizer import NNCFQuantizer from nncf.quantization.algorithms.algorithm import Algorithm -from nncf.quantization.fake_quantize import calculate_quantizer_parameters -from nncf.quantization.fake_quantize import get_quantizer_narrow_range +from nncf.quantization.algorithms.min_max.algorithm import MinMaxQuantization from nncf.quantization.range_estimator import RangeEstimatorParameters -from nncf.quantization.range_estimator import RangeEstimatorParametersSet TModel = TypeVar("TModel") @@ -69,315 +47,18 @@ def __init__( for weights. """ self._quantizer = quantizer - self._subset_size = subset_size - self._inplace_statistics = inplace_statistics - self._batchwise_statistics = batchwise_statistics - self._activations_range_estimator_params = activations_range_estimator_params - self._weights_range_estimator_params = weights_range_estimator_params - - self._range_estimator_params = { - QuantizerGroup.WEIGHTS: self._weights_range_estimator_params, - QuantizerGroup.ACTIVATIONS: self._activations_range_estimator_params, - } - # Calculates global quantizer constraints - self._reset_cache() - self._algorithm_key = f"MMQ_{hash(self)}" - - def _reset_cache(self) -> None: - """ - Marks cache by noninitialized values. Needs to be called when the new quantizer setup is needed. - """ - self._quantization_target_points_to_qconfig: OrderedDict[TargetPoint, QuantizerConfig] = None - self._unified_scale_groups = None - - def _init_cache(self) -> None: - """ - Initializes cache. - """ - self._quantization_target_points_to_qconfig: OrderedDict[TargetPoint, QuantizerConfig] = ( - collections.OrderedDict() + self._min_max_algo = MinMaxQuantization( + subset_size=subset_size, + inplace_statistics=inplace_statistics, + batchwise_statistics=batchwise_statistics, + activations_range_estimator_params=activations_range_estimator_params, + weights_range_estimator_params=weights_range_estimator_params, ) - self._unified_scale_groups = [] @property def available_backends(self) -> List[BackendType]: return [BackendType.TORCH_FX] - def _set_backend_entity(self, model: TModel) -> None: - """ - Creates a helper class with a backed-specific logic of the algorithm - - :param model: backend-specific input model - """ - model_backend = get_backend(model) - if model_backend == BackendType.TORCH_FX: - from nncf.experimental.common.quantization.algorithms.range_estimator.torch_fx_backend import ( - FXRangeEstimatorAlgoBackend, - ) - - self._backend_entity = FXRangeEstimatorAlgoBackend() - else: - raise nncf.UnsupportedBackendError( - "Cannot return backend-specific entity because {} is not supported!".format(model_backend.value) - ) - - def _get_range_estimator_parameters( - self, target_point: TargetPoint, quantizer_config: QuantizerConfig - ) -> RangeEstimatorParameters: - """ - Returns range estimator parameters. - - :param target_point: Quantizer target point. - :param quantizer_config: Quantizer config. - :return: Range estimator parameters. - """ - quantizer_group = QuantizerGroup.ACTIVATIONS - if target_point.is_weight_target_point(): - quantizer_group = QuantizerGroup.WEIGHTS - - if quantizer_group == QuantizerGroup.WEIGHTS or ( - quantizer_group == QuantizerGroup.ACTIVATIONS and quantizer_config.per_channel - ): - params = RangeEstimatorParametersSet.MINMAX - else: - params = RangeEstimatorParametersSet.MEAN_MINMAX - - user_params = self._range_estimator_params[quantizer_group] - if user_params is None: - return deepcopy(params) - - min_changes = changes_asdict(user_params.min) - min_statistic_collector = dataclasses.replace(params.min, **min_changes) - - max_changes = changes_asdict(user_params.max) - max_statistic_collector = dataclasses.replace(params.max, **max_changes) - - return RangeEstimatorParameters(min_statistic_collector, max_statistic_collector) - - def _get_stat_collector( - self, - graph: NNCFGraph, - target_point: TargetPoint, - qconfig: QuantizerConfig, - batchwise_statistics: bool, - ) -> TensorStatisticCollectorBase: - """ - Creates and returns a statistic collector based on the quantizer's configuration. - - :param graph: NNCFGraph instance. - :param target_point: Target point indicates where statistics should be collected. - :param qconfig: Configuration of a quantizer layer, - defining the configuration of created statistic collector. - :param batchwise_statistics: Determines whether quantizer statistics should be calculated - for each item of the batch or for the entire batch. - :return: Statistic Collector. - """ - is_weight = target_point.is_weight_target_point() - node = graph.get_node_by_name(target_point.target_node_name) - shape = self._backend_entity.get_target_point_shape(graph, node, target_point) - range_estimator_params = self._get_range_estimator_parameters(target_point, qconfig) - - channel_axes = () - if qconfig.per_channel: - channel_axes = ( - self._backend_entity.get_weight_quantization_axes(node, target_point, len(shape)) if is_weight else (1,) - ) - - # Weight statistics is constant, so only one collection is enough. - num_samples = self._subset_size if not is_weight else 1 - - batchwise_statistics = batchwise_statistics and not is_weight - - collector_params = RangeInitCollectorParams( - is_weights=is_weight, scheme=qconfig.mode, per_channel=qconfig.per_channel - ) - reduction_axes, aggregation_axes = None, None - if shape is not None: - reduction_axes, aggregation_axes = collector_params.get_reduction_aggregation_axes( - shape, channel_axes, batchwise_statistics - ) - - return self._backend_entity.get_statistic_collector( - range_estimator_params, - collector_params.use_abs_max, - reduction_axes, - aggregation_axes, - self._inplace_statistics, - num_samples=num_samples, - ) - - def _add_weight_quantization_target_point( - self, quantization_point: SingleConfigQuantizationPoint, nncf_graph: NNCFGraph - ) -> None: - """ - Adds weight quantization target point to the set of existing points. - - :param quantization_point: SingleConfigQuantizationPoint for the needed layer. - :param nncf_graph: The built NNCFGraph of the model. - """ - weight_quantization_target_points = self._get_weight_quantization_target_points(quantization_point, nncf_graph) - for weight_quantization_target_point in weight_quantization_target_points: - self._quantization_target_points_to_qconfig[weight_quantization_target_point] = quantization_point.qconfig - - def _add_activation_quantization_target_point( - self, quantization_point: SingleConfigQuantizationPoint, nncf_graph: NNCFGraph - ) -> None: - """ - Adds activation quantization target point to the set of existing points. - - :param quantization_point: SingleConfigQuantizationPoint for the needed layer. - :param nncf_graph: NNCFGraph instance for working with the graph and nodes. - """ - activation_quantization_target_point = self._get_activation_quantization_target_point( - quantization_point, nncf_graph - ) - self._quantization_target_points_to_qconfig[activation_quantization_target_point] = quantization_point.qconfig - - def _get_weight_quantization_target_points( - self, quantization_point: SingleConfigQuantizationPoint, nncf_graph: NNCFGraph - ) -> List[SingleConfigQuantizationPoint]: - """ - Returns weight quantization target points to the set of existing points. - - :param quantization_point: SingleConfigQuantizationPoint for the needed layer. - :param nncf_graph: NNCFGraph instance for working with the graph and nodes. - :return: List of SingleConfigQuantizationPoints for the needed layer. - """ - weight_quantization_target_points = [] - node_name = quantization_point.insertion_point.target_node_name - node = nncf_graph.get_node_by_name(node_name) - weights_port_ids = self._backend_entity.get_weight_tensor_port_ids(node, nncf_graph) - for port_id in weights_port_ids: - weight_quantization_target_points.append( - self._backend_entity.target_point(TargetType.OPERATION_WITH_WEIGHTS, node_name, port_id) - ) - return weight_quantization_target_points - - def _get_activation_quantization_target_point( - self, quantization_point: SingleConfigQuantizationPoint, nncf_graph: NNCFGraph - ) -> SingleConfigQuantizationPoint: - """ - Returns activation quantization target point to the set of existing points. - - :param quantization_point: SingleConfigQuantizationPoint for the needed layer. - :param nncf_graph: NNCFGraph instance for working with the graph and nodes. - :return: SingleConfigQuantizationPoint for the needed layer. - """ - node_name = quantization_point.insertion_point.target_node_name - # If Quantization of node's input - if quantization_point.insertion_point.input_port_id is not None: - input_port_id = quantization_point.insertion_point.input_port_id - activation_quantization_target_point = self._backend_entity.target_point( - TargetType.PRE_LAYER_OPERATION, node_name, input_port_id - ) - # If quantization of node's output or Model Input node - else: - # NOTE: Assumes that the operation has output edges only from one output port because - # we haven't encountered a model with operations that have multiple output edges with different - # output port IDs. Currently, such models are not supported. Usually, `output_port_id = 0` is used. - # However, there are operations, such as LSTMSequence, where the `output_port_id` changes from case - # to case. Therefore, the code below is required to dynamically determine the `output_port_id` where - # the quantize operation should be inserted." - node = nncf_graph.get_node_by_name(node_name) - unique_output_port_ids = set(e.output_port_id for e in nncf_graph.get_output_edges(node)) - if len(unique_output_port_ids) > 1: - nncf_logger.warning( - f"Cannot determine the output_port_id for the operation: {node_name}, " - "output_port_id = 0 will be used." - ) - output_port_id = 0 - else: - output_port_id = next(iter(unique_output_port_ids)) - - activation_quantization_target_point = self._backend_entity.target_point( - TargetType.POST_LAYER_OPERATION, node_name, output_port_id - ) - return activation_quantization_target_point - - def _find_quantization_target_points( - self, model: TModel, nncf_graph: NNCFGraph - ) -> Tuple[OrderedDict[TargetPoint, QuantizerConfig], List[List[TargetPoint]]]: - """ - Initializes a cache, finds quantization target points and them puts in the cache. - - :param model: Backend-specific model, for which Quantization Target Points are being seek. - :param nncf_graph: NNCFGraph instance. - :return: Mapping of quantization target points with associated quantization configuration, - along with target points for scale unification. - """ - quantizer_setup = self._quantizer.get_quantization_setup(model, nncf_graph) - self._unified_scale_groups = self._collect_unified_groups(quantizer_setup, nncf_graph) - quantization_points = list(quantizer_setup.quantization_points.values()) - quantization_points = self._topological_sort_quantization_points(quantization_points, nncf_graph) - for quantization_point in quantization_points: - if quantization_point.is_weight_quantization_point(): - self._add_weight_quantization_target_point(quantization_point, nncf_graph) - elif quantization_point.is_activation_quantization_point(): - self._add_activation_quantization_target_point(quantization_point, nncf_graph) - else: - raise nncf.InternalError("Incorrect quantization point") - return self._quantization_target_points_to_qconfig, self._unified_scale_groups - - def _get_quantization_target_points( - self, model: TModel, nncf_graph: NNCFGraph - ) -> Tuple[OrderedDict[TargetPoint, QuantizerConfig], List[List[TargetPoint]]]: - """ - Returns Quantization Target Points. - Returns a cache with target points if exists. Otherwise, initiates a procedure of finding them. - - :param model: Backend-specific model, for which Quantization Target Points are being seek. - :param nncf_graph: NNCFGraph instance. - :return: Mapping of quantization target points with associated quantization configuration, - along with target points for scale unification. - """ - if self._quantization_target_points_to_qconfig is not None: - return self._quantization_target_points_to_qconfig, self._unified_scale_groups - self._init_cache() - return self._find_quantization_target_points(model, nncf_graph) - - def _collect_unified_groups( - self, quantizer_setup: SingleConfigQuantizerSetup, nncf_graph: NNCFGraph - ) -> List[List[TargetPoint]]: - """ - Collects the group of quantizers for unification. - - :param quantizer_setup: SingleConfigQuantizerSetup instance. - :param nncf_graph: NNCFGraph instance. - :return: List with the groups of the TargetPoints. - """ - unified_scale_groups = [] - for quantizer_ids in quantizer_setup.unified_scale_groups.values(): - unified_scale_group = [] - for quantizer_id in quantizer_ids: - quantization_point = quantizer_setup.quantization_points[quantizer_id] - - # Only activation quantizers can be unified - if quantization_point.is_activation_quantization_point(): - activation_target_point = self._get_activation_quantization_target_point( - quantization_point, nncf_graph - ) - unified_scale_group.append(activation_target_point) - else: - weight_target_points = self._get_weight_quantization_target_points(quantization_point, nncf_graph) - for weight_target_point in weight_target_points: - unified_scale_group.append(weight_target_point) - unified_scale_groups.append(unified_scale_group) - return unified_scale_groups - - def _topological_sort_quantization_points( - self, quantization_points: List[SingleConfigQuantizationPoint], nncf_graph: NNCFGraph - ) -> List[SingleConfigQuantizationPoint]: - """ - Sorts quantization_points based on the topological order of nodes obtained form nncf_graph. - - :param quantization_points: Quantization points. - :param nncf_graph: Instance of NNCFgraph used to get topological sort. - :return: Sorted quantization_points. - """ - node_names_to_pos = {node.node_name: i for i, node in enumerate(nncf_graph.topological_sort())} - quantization_points.sort(key=lambda point: node_names_to_pos[point.insertion_point.target_node_name]) - return quantization_points - def apply( self, model: TModel, @@ -385,111 +66,16 @@ def apply( statistic_points: Optional[StatisticPointsContainer] = None, dataset: Optional[Dataset] = None, ) -> TModel: - transformation_layout = TransformationLayout() - model_transformer = ModelTransformerFactory.create(model) - quantization_target_points, unified_scale_groups = self._get_quantization_target_points(model, graph) - weight_layer_names = set() - - def filter_func(point: StatisticPoint) -> bool: - return ( - self._algorithm_key in point.algorithm_to_tensor_collectors - and point.target_point == quantization_target_point - ) - - unified_ops_list = set() - for unified_scale_group in unified_scale_groups: - group_statistics = [] - for quantization_target_point in unified_scale_group: - target_node_name = quantization_target_point.target_node_name - for tensor_collector in statistic_points.get_algo_statistics_for_node( - target_node_name, filter_func, self._algorithm_key - ): - statistics = tensor_collector.get_statistics() - if statistics.min_values is None or statistics.max_values is None: - raise nncf.InternalError(f"Statistics were not collected for the node {target_node_name}") - group_statistics.append(statistics) - - unified_values = self._unify_statistics(group_statistics) - qconfigs = [quantization_target_points[qtp] for qtp in unified_scale_group] - if any(qconfigs[0] != qconfig for qconfig in qconfigs[1:]): - raise nncf.InternalError(f"QConfigs for unified scale group {unified_scale_group} are not equal") - qconfig = qconfigs[0] - q_group = QuantizerGroup.ACTIVATIONS - narrow_range = get_quantizer_narrow_range(qconfig, q_group) - parameters = calculate_quantizer_parameters(unified_values, qconfig, q_group, narrow_range) - commands = self._backend_entity.create_unified_scales_quantizers_insertion_commands( - graph, unified_scale_group, qconfig, parameters + if self._min_max_algo._quantization_target_points_to_qconfig is None: + raise RuntimeError( + "Static points are not available." + " Please call `get_statistic_points` before calling the `apply` method." ) - for command in commands: - transformation_layout.register(command) - unified_ops_list.update(unified_scale_group) - - for quantization_target_point, qconfig in quantization_target_points.items(): - if quantization_target_point in unified_ops_list: - continue - target_node_name = quantization_target_point.target_node_name - for tensor_collector in statistic_points.get_algo_statistics_for_node( - target_node_name, filter_func, self._algorithm_key - ): - if quantization_target_point.is_weight_target_point(): - weights_name = self._backend_entity.get_weight_name(graph, quantization_target_point) - if not self._backend_entity.should_quantize_weight(weights_name, weight_layer_names): - continue - weight_layer_names.add(weights_name) - quant_group = QuantizerGroup.WEIGHTS - else: - quant_group = QuantizerGroup.ACTIVATIONS - - half_range = False - narrow_range = get_quantizer_narrow_range(qconfig, quant_group) - statistics = tensor_collector.get_statistics() - if statistics.min_values is None or statistics.max_values is None: - raise nncf.InternalError(f"Statistics were not collected for the node {target_node_name}") - parameters = calculate_quantizer_parameters(statistics, qconfig, quant_group, narrow_range, half_range) - command = self._backend_entity.create_quantizer_insertion_command( - graph, quantization_target_point, qconfig, parameters - ) - transformation_layout.register(command) - if not transformation_layout.transformations: - nncf_logger.info("The model has no operations to apply quantization.") - quantized_model = model_transformer.transform(transformation_layout) - return quantized_model + return self._min_max_algo.apply(model=model, graph=graph, statistic_points=statistic_points) def get_statistic_points(self, model: TModel, graph: NNCFGraph) -> StatisticPointsContainer: - self._set_backend_entity(model) - self._reset_cache() - quantization_target_points, _ = self._get_quantization_target_points(model, graph) - output = StatisticPointsContainer() - for quantization_target_point, qconfig in quantization_target_points.items(): - nncf_logger.debug( - f"Adding target point {quantization_target_point.target_node_name}" - f" with type {quantization_target_point.type} for statistics collection" - ) - stat_collector = self._get_stat_collector( - graph, quantization_target_point, qconfig, self._batchwise_statistics - ) - output.add_statistic_point( - StatisticPoint( - target_point=quantization_target_point, - tensor_collector=stat_collector, - algorithm=self._algorithm_key, - ) - ) - return output - - @staticmethod - def _unify_statistics(statistics: List[MinMaxTensorStatistic]) -> MinMaxTensorStatistic: - """ - Returns backend-specific unified statistics. - - :param statistics: List of MinMaxTensorStatistic instances. - :return: Unified MinMaxTensorStatistic value. - """ - - max_values, min_values = [], [] - for statistic in statistics: - max_values.append(statistic.max_values.flatten()) - min_values.append(statistic.min_values.flatten()) - max_values = fns.max(fns.stack(max_values), axis=0) - min_values = fns.min(fns.stack(min_values), axis=0) - return MinMaxTensorStatistic(min_values=min_values, max_values=max_values) + quantizer_setup = self._quantizer.get_quantization_setup(model, graph) + self._min_max_algo._set_backend_entity(model) + self._min_max_algo._init_cache() + self._min_max_algo.fill_quantization_target_points(quantizer_setup, graph) + return self._min_max_algo.get_cached_statistic_points(model, graph) diff --git a/nncf/experimental/common/quantization/algorithms/range_estimator/torch_fx_backend.py b/nncf/experimental/common/quantization/algorithms/range_estimator/torch_fx_backend.py deleted file mode 100644 index 0e30e70ae57..00000000000 --- a/nncf/experimental/common/quantization/algorithms/range_estimator/torch_fx_backend.py +++ /dev/null @@ -1,221 +0,0 @@ -# Copyright (c) 2024 Intel Corporation -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# http://www.apache.org/licenses/LICENSE-2.0 -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. - -from typing import List, Optional, Set, Tuple - -import torch -from torch.quantization.fake_quantize import FakeQuantize - -import nncf -from nncf.common.graph.graph import NNCFGraph -from nncf.common.graph.graph import NNCFNode -from nncf.common.graph.transformations.commands import TargetType -from nncf.common.quantization.structs import QuantizationScheme as QuantizationMode -from nncf.common.quantization.structs import QuantizerConfig -from nncf.experimental.common.quantization.algorithms.range_estimator.backend import RangeEstimatorAlgoBackend -from nncf.experimental.common.tensor_statistics.collectors import AGGREGATORS_MAP -from nncf.experimental.common.tensor_statistics.collectors import TensorCollector -from nncf.experimental.common.tensor_statistics.statistics import MinMaxTensorStatistic -from nncf.experimental.torch.fx.commands import FXApplyTransformationCommand -from nncf.experimental.torch.fx.model_utils import get_target_point -from nncf.experimental.torch.fx.transformations import qdq_insertion_transformation_builder -from nncf.quantization.advanced_parameters import StatisticsType -from nncf.quantization.fake_quantize import FakeQuantizeParameters -from nncf.quantization.range_estimator import AggregatorType -from nncf.quantization.range_estimator import RangeEstimatorParameters -from nncf.torch.graph.graph import PTNNCFGraph -from nncf.torch.graph.graph import PTTargetPoint -from nncf.torch.graph.transformations.commands import PTSharedFnInsertionCommand -from nncf.torch.model_graph_manager import get_weight_tensor_port_ids -from nncf.torch.quantization.layers import QUANTIZATION_MODULES -from nncf.torch.quantization.layers import AsymmetricQuantizer -from nncf.torch.quantization.layers import BaseQuantizer -from nncf.torch.quantization.layers import PTQuantizerSpec -from nncf.torch.quantization.layers import get_scale_shape -from nncf.torch.quantization.strip import convert_to_torch_fakequantizer -from nncf.torch.tensor_statistics.collectors import PT_REDUCERS_MAP - - -class FXRangeEstimatorAlgoBackend(RangeEstimatorAlgoBackend): - @staticmethod - def target_point(target_type: TargetType, target_node_name: str, port_id: int) -> PTTargetPoint: - return get_target_point(target_type, target_node_name, port_id) - - @staticmethod - def get_target_point_shape(nncf_graph: PTNNCFGraph, node: NNCFNode, target_point: PTTargetPoint) -> Tuple[int, ...]: - return nncf_graph.get_input_shape_for_insertion_point(target_point) - - @staticmethod - def get_weight_quantization_axes(node: NNCFNode, target_point: PTTargetPoint, ndims: int) -> Tuple[int]: - # TODO(dlyakhov): support transpose conv and other cases - return (0,) - - @staticmethod - def get_statistic_collector( - range_estimator_params: RangeEstimatorParameters, - use_abs_max: bool, - reduction_axes: Optional[Tuple[int, ...]], - aggregation_axes: Optional[Tuple[int, ...]], - inplace: bool, - num_samples: Optional[int] = None, - ) -> TensorCollector: - collector = TensorCollector(MinMaxTensorStatistic) - for params, container_key in zip( - [range_estimator_params.min, range_estimator_params.max], - [MinMaxTensorStatistic.MIN_STAT, MinMaxTensorStatistic.MAX_STAT], - ): - if params.statistics_type not in PT_REDUCERS_MAP: - raise nncf.InternalError( - f"Statistic type: {params.statistics_type} is not supported for Torch PTQ backend yet." - ) - - if params.aggregator_type not in AGGREGATORS_MAP: - raise nncf.InternalError( - f"Aggregator type: {params.aggregator_type} is not supported for Torch PTQ backend yet." - ) - - statistic_type = params.statistics_type - if statistic_type in [StatisticsType.QUANTILE, StatisticsType.ABS_QUANTILE]: - # TODO(dlyakhov): merge two quantile aggregators in one - if container_key == MinMaxTensorStatistic.MIN_STAT: - quantile = params.quantile_outlier_prob - else: - quantile = 1 - params.quantile_outlier_prob - reducer = PT_REDUCERS_MAP[statistic_type](reduction_axes=reduction_axes, quantile=[quantile]) - else: - if use_abs_max and statistic_type == StatisticsType.MAX: - statistic_type = StatisticsType.ABS_MAX - reducer = PT_REDUCERS_MAP[statistic_type](reduction_axes=reduction_axes) - - kwargs = { - "num_samples": num_samples, - "aggregation_axes": aggregation_axes, - } - if params.aggregator_type in [AggregatorType.MEAN_NO_OUTLIERS, AggregatorType.MEDIAN_NO_OUTLIERS]: - kwargs.update({"quantile": params.quantile_outlier_prob}) - aggregator = AGGREGATORS_MAP[params.aggregator_type](**kwargs) - - collector.register_statistic_branch(container_key, reducer, aggregator) - return collector - - @staticmethod - def get_weight_tensor_port_ids(node: NNCFNode, graph: NNCFGraph) -> List[Optional[int]]: - return get_weight_tensor_port_ids(node, graph) - - @staticmethod - def get_weight_name(nncf_graph: NNCFGraph, target_point: PTTargetPoint) -> str: - weighted_node = nncf_graph.get_node_by_name(target_point.target_node_name) - weight_edge = nncf_graph.get_input_edge_by_port_id(weighted_node, target_point.input_port_id) - weight = weight_edge.from_node - return weight.node_name - - @staticmethod - def should_quantize_weight(weight_name: str, quantized_weight_names: Set[str]) -> bool: - # If the nodes share one weight tensor, we should have only one quantizer on that - return weight_name not in quantized_weight_names - - @staticmethod - def _get_input_scale_shape( - nncf_graph: NNCFGraph, target_point: PTTargetPoint, per_channel: bool - ) -> Tuple[Tuple[int, ...], Tuple[int, ...], int]: - is_weights = target_point.is_weight_target_point() - if is_weights: - # TODO(dlyakhov): support transpose conv/ make channel_idx common - channel_idx = 0 - else: - channel_idx = 1 # channel dim for activations - - input_shape = nncf_graph.get_input_shape_for_insertion_point(target_point) - scale_shape = tuple( - get_scale_shape(input_shape, is_weights=is_weights, per_channel=per_channel, channel_idx=channel_idx) - ) - - return input_shape, scale_shape, channel_idx - - @staticmethod - def _create_quantizer( - quantizer_config: QuantizerConfig, - scale_shape: Tuple, - parameters: FakeQuantizeParameters, - target_type: TargetType, - ) -> FakeQuantize: - mode = quantizer_config.mode - quantizer_cls = QUANTIZATION_MODULES.get(mode) - narrow_range = target_type == TargetType.OPERATION_WITH_WEIGHTS and mode == QuantizationMode.SYMMETRIC - quantizer_spec = PTQuantizerSpec.from_config( - quantizer_config, - narrow_range=narrow_range, - scale_shape=scale_shape, - half_range=False, - logarithm_scale=False, - is_quantized_on_export=False, - compression_lr_multiplier=None, - ) - quantizer = quantizer_cls(quantizer_spec) - - # Fill it with minmax - # TODO(dlyakhov) Prevent creation of intermediate objects like nncf quantizer. - FXRangeEstimatorAlgoBackend._fill_quantizer_parameters(quantizer, parameters, quantizer_spec.scale_shape) - # Convert to the torch fake quantizer - torch_fq = convert_to_torch_fakequantizer(quantizer) - return torch_fq - - @staticmethod - def _fill_quantizer_parameters(quantizer: BaseQuantizer, parameters: FakeQuantizeParameters, scale_shape) -> None: - if isinstance(quantizer, AsymmetricQuantizer): - quantizer.input_low = torch.nn.Parameter(parameters.input_low.data.reshape(scale_shape)) - input_range = parameters.input_high - parameters.input_low - # Subtract eps from the input_range to make quantizer parameters equal to - # original parameters on the forward call. - quantizer.input_range = torch.nn.Parameter((input_range.data - quantizer.eps).reshape(scale_shape)) - else: - quantizer.signed = bool(torch.any(parameters.input_low.data < 0)) - # Subtract eps from the scale to make quantizer parameters equal to - # original parameters on the forward call. - quantizer.scale = torch.nn.Parameter((parameters.input_high.data - quantizer.eps).reshape(scale_shape)) - - @staticmethod - def create_quantizer_insertion_command( - nncf_graph: NNCFGraph, - target_point: PTTargetPoint, - quantizer_config: QuantizerConfig, - parameters: FakeQuantizeParameters, - ) -> FXApplyTransformationCommand: - _, scale_shape, _ = FXRangeEstimatorAlgoBackend._get_input_scale_shape( - nncf_graph, target_point, quantizer_config.per_channel - ) - - quantizer = FXRangeEstimatorAlgoBackend._create_quantizer( - quantizer_config, scale_shape, parameters, target_point.target_type - ) - transformation = qdq_insertion_transformation_builder(quantizer, [target_point]) - return FXApplyTransformationCommand(transformation) - - @staticmethod - def create_unified_scales_quantizers_insertion_commands( - nncf_graph: NNCFGraph, - target_points: List[PTTargetPoint], - quantizer_config: QuantizerConfig, - parameters: FakeQuantizeParameters, - ) -> List[PTSharedFnInsertionCommand]: - _, scale_shape, _ = FXRangeEstimatorAlgoBackend._get_input_scale_shape( - nncf_graph, target_points[0], quantizer_config.per_channel - ) - - quantizer = FXRangeEstimatorAlgoBackend._create_quantizer( - quantizer_config, scale_shape, parameters, target_points[0].target_type - ) - - transformations = [] - for tp in target_points: - transformation = qdq_insertion_transformation_builder(quantizer, [tp]) - transformations.append(FXApplyTransformationCommand(transformation)) - return transformations diff --git a/nncf/quantization/algorithms/min_max/algorithm.py b/nncf/quantization/algorithms/min_max/algorithm.py index dea9211b734..b6728b292bf 100644 --- a/nncf/quantization/algorithms/min_max/algorithm.py +++ b/nncf/quantization/algorithms/min_max/algorithm.py @@ -723,9 +723,7 @@ def _get_activation_quantization_target_point( ) return activation_quantization_target_point - def _find_quantization_target_points( - self, model: TModel, nncf_graph: NNCFGraph - ) -> Tuple[OrderedDict[TargetPoint, QuantizerConfig], List[List[TargetPoint]]]: + def find_quantization_setup(self, model: TModel, nncf_graph: NNCFGraph) -> SingleConfigQuantizerSetup: """ Initializes a cache, finds quantization target points and them puts in the cache. @@ -753,6 +751,19 @@ def _find_quantization_target_points( quantizer_setup = self._get_quantizer_setup(nncf_graph, inference_nncf_graph, hw_patterns, ignored_patterns) self._apply_model_type_pass(self._model_type, quantizer_setup, nncf_graph) self._apply_device_pass(self._target_device, quantizer_setup, inference_nncf_graph) + return quantizer_setup + + def fill_quantization_target_points( + self, quantizer_setup: SingleConfigQuantizerSetup, nncf_graph: NNCFGraph + ) -> Tuple[OrderedDict[TargetPoint, QuantizerConfig], List[List[TargetPoint]]]: + """ + Initializes a cache, finds quantization target points and them puts in the cache. + + :param model: Backend-specific model, for which Quantization Target Points are being seek. + :param nncf_graph: NNCFGraph instance. + :return: Mapping of quantization target points with associated quantization configuration, + along with target points for scale unification. + """ self._unified_scale_groups = self._collect_unified_groups(quantizer_setup, nncf_graph) quantization_points = list(quantizer_setup.quantization_points.values()) quantization_points = self._topological_sort_quantization_points(quantization_points, nncf_graph) @@ -780,7 +791,8 @@ def _get_quantization_target_points( if self._quantization_target_points_to_qconfig is not None: return self._quantization_target_points_to_qconfig, self._unified_scale_groups self._init_cache() - return self._find_quantization_target_points(model, nncf_graph) + quantizer_setup = self.find_quantization_setup(model, nncf_graph) + return self.fill_quantization_target_points(quantizer_setup, nncf_graph) def _collect_unified_groups( self, quantizer_setup: SingleConfigQuantizerSetup, nncf_graph: NNCFGraph @@ -989,10 +1001,30 @@ def filter_func(point: StatisticPoint) -> bool: quantized_model = model_transformer.transform(transformation_layout) return quantized_model + def get_cached_statistic_points(self, model: TModel, graph: NNCFGraph) -> StatisticPointsContainer: + """ + Build statistic point container using already cached target points vs qconfigs cache. + + :param model: Model instance. + :param graph: NNCFGraph instance corespondent to the passed model. + :return: Filled statistic point container. + """ + if self._quantization_target_points_to_qconfig is None: + raise RuntimeError("get_cached_statistic_points is called before statistic caching.") + self._set_backend_entity(model) + return self._get_statistic_point_container(self._quantization_target_points_to_qconfig, graph) + def get_statistic_points(self, model: TModel, graph: NNCFGraph) -> StatisticPointsContainer: self._set_backend_entity(model) self._reset_cache() quantization_target_points, _ = self._get_quantization_target_points(model, graph) + return self._get_statistic_point_container(quantization_target_points, graph) + + def _get_statistic_point_container( + self, + quantization_target_points: Tuple[OrderedDict[TargetPoint, QuantizerConfig], List[List[TargetPoint]]], + graph: NNCFGraph, + ) -> StatisticPointsContainer: output = StatisticPointsContainer() for quantization_target_point, qconfig in quantization_target_points.items(): nncf_logger.debug( diff --git a/tests/common/quantization/test_minmax.py b/tests/common/quantization/test_minmax.py index 2a63d3c7ac8..2f780ea652f 100644 --- a/tests/common/quantization/test_minmax.py +++ b/tests/common/quantization/test_minmax.py @@ -215,14 +215,20 @@ def test_min_max_caching(): Checks that the _get_quantization_target_points(...) of MinMaxQuantization called once utilizing the cache. Checks that after _reset_cache() it called one more time. """ - called = 0 + find_called = 0 + fill_called = 0 - def foo(self, *args): + def find_qsetup_mock(self, *args): + nonlocal find_called + find_called += 1 + return None + + def fill_qsetup_mock(self, *args): """ Mocked _find_quantization_target_points. """ - nonlocal called - called += 1 + nonlocal fill_called + fill_called += 1 # Set up cache self._quantization_target_points_to_qconfig = collections.OrderedDict() self._unified_scale_groups = [] @@ -230,11 +236,12 @@ def foo(self, *args): run_nums = 2 algo = MinMaxQuantization() - algo._find_quantization_target_points = types.MethodType(foo, algo) + algo.find_quantization_setup = types.MethodType(find_qsetup_mock, algo) + algo.fill_quantization_target_points = types.MethodType(fill_qsetup_mock, algo) for _ in range(run_nums): algo._get_quantization_target_points(None, None) - assert called == 1 + assert find_called == fill_called == 1 algo._reset_cache() for _ in range(run_nums): algo._get_quantization_target_points(None, None) - assert called == 2 + assert find_called == fill_called == 2 From 52e80c8398790aa789efe5ca92b747c9101dfcd1 Mon Sep 17 00:00:00 2001 From: dlyakhov Date: Mon, 2 Dec 2024 17:24:01 +0100 Subject: [PATCH 04/15] Correct use of transform_for_annotation --- .../torch/fx/quantization/quantize_pt2e.py | 17 ++++++++++++----- 1 file changed, 12 insertions(+), 5 deletions(-) diff --git a/nncf/experimental/torch/fx/quantization/quantize_pt2e.py b/nncf/experimental/torch/fx/quantization/quantize_pt2e.py index 78ba278826e..ed41166b9ca 100644 --- a/nncf/experimental/torch/fx/quantization/quantize_pt2e.py +++ b/nncf/experimental/torch/fx/quantization/quantize_pt2e.py @@ -28,6 +28,7 @@ from nncf.experimental.common.quantization.algorithms.post_training.algorithm import ( ExperimentalPostTrainingQuantization, ) +from nncf.experimental.common.quantization.algorithms.quantizer.base_quantizer import NNCFQuantizer from nncf.experimental.common.quantization.algorithms.quantizer.fx_quantizer import NNCFFXQuantizer from nncf.experimental.torch.fx.constant_folding import constant_fold from nncf.experimental.torch.fx.transformations import QUANTIZE_NODE_TARGETS @@ -63,8 +64,18 @@ def quantize_pt2e( copied_model = deepcopy(model) + # To make it easier for bias correction algorithms, + # biases are being separated by the followng calls. + fuse_conv_bn(copied_model) + # Call ao quantizer transform_for_annotation + # before the NNCFGraph creation + quantizer.transform_for_annotation(copied_model) + + if not isinstance(quantizer, NNCFQuantizer): + quantizer = NNCFFXQuantizer(quantizer) + quantization_algorithm = ExperimentalPostTrainingQuantization( - quantizer=NNCFFXQuantizer(quantizer), + quantizer=quantizer, subset_size=subset_size, fast_bias_correction=fast_bias_correction, smooth_quant=smooth_quant, @@ -74,10 +85,6 @@ def quantize_pt2e( weights_range_estimator_params=weights_range_estimator_params, ) - # To make it easier for bias correction algorithms, - # biases are being separated by the followng calls. - fuse_conv_bn(copied_model) - nncf_graph = NNCFGraphFactory.create(copied_model) quantized_model = quantization_algorithm.apply(copied_model, nncf_graph, dataset=calibration_dataset) From 43bc2516d391758f8bb117df5e53dfc35996cab7 Mon Sep 17 00:00:00 2001 From: dlyakhov Date: Thu, 5 Dec 2024 11:48:11 +0100 Subject: [PATCH 05/15] Comments/fixes --- .../algorithms/post_training/algorithm.py | 2 +- .../algorithms/post_training/pipeline.py | 4 +- .../algorithms/quantizer/fx_quantizer.py | 3 +- .../mobilenet_v3_small.dot | 76 +++++++++---------- .../X86InductorQuantizer/resnet18.dot | 36 +++++---- 5 files changed, 58 insertions(+), 63 deletions(-) diff --git a/nncf/experimental/common/quantization/algorithms/post_training/algorithm.py b/nncf/experimental/common/quantization/algorithms/post_training/algorithm.py index efbc5bb7449..ac2d86a3056 100644 --- a/nncf/experimental/common/quantization/algorithms/post_training/algorithm.py +++ b/nncf/experimental/common/quantization/algorithms/post_training/algorithm.py @@ -47,7 +47,7 @@ def __init__( weights_range_estimator_params: Optional[RangeEstimatorParameters] = None, ): """ - :param quantizer: NNCFQuantizer to use in MiMaxRageInit algorithm. + :param quantizer: NNCFQuantizer to use in MiMaxRangeInit algorithm. :param subset_size: Size of a subset to calculate activations statistics used for quantization. :param fast_bias_correction: Setting this option to `False` enables a different diff --git a/nncf/experimental/common/quantization/algorithms/post_training/pipeline.py b/nncf/experimental/common/quantization/algorithms/post_training/pipeline.py index dab7b2be856..5c8e521f65a 100644 --- a/nncf/experimental/common/quantization/algorithms/post_training/pipeline.py +++ b/nncf/experimental/common/quantization/algorithms/post_training/pipeline.py @@ -44,7 +44,7 @@ def experimental_create_ptq_pipeline( 2) MinMaxRangeInit 3) FastBiasCorrection or BiasCorrection - :param quantizer: NNCFQuantizer to use in MiMaxRageInit algorithm. + :param quantizer: NNCFQuantizer to use in MiMaxRangeInit algorithm. :param subset_size: Size of a subset to calculate activations statistics used for quantization. :param fast_bias_correction: Setting this option to `False` enables a different @@ -66,7 +66,7 @@ def experimental_create_ptq_pipeline( if smooth_quant_params is None: smooth_quant_params = AdvancedSmoothQuantParameters() - if smooth_quant and smooth_quant_params.convolution >= 0 or smooth_quant_params.matmul >= 0: + if smooth_quant and (smooth_quant_params.convolution >= 0 or smooth_quant_params.matmul >= 0): alpha_map = {"convolution": smooth_quant_params.convolution, "matmul": smooth_quant_params.matmul} pipeline_steps.append([SmoothQuant(subset_size, False, alpha_map=alpha_map)]) diff --git a/nncf/experimental/common/quantization/algorithms/quantizer/fx_quantizer.py b/nncf/experimental/common/quantization/algorithms/quantizer/fx_quantizer.py index db0ae167132..33e0ef94a79 100644 --- a/nncf/experimental/common/quantization/algorithms/quantizer/fx_quantizer.py +++ b/nncf/experimental/common/quantization/algorithms/quantizer/fx_quantizer.py @@ -41,7 +41,8 @@ def __init__(self, quantizer: Quantizer): def get_quantization_setup(self, model: torch.fx.GraphModule, nncf_graph: NNCFGraph) -> SingleConfigQuantizerSetup: anotated_model = deepcopy(model) - self._quantizer.transform_for_annotation(anotated_model) + # self._quantizer.transform_for_annotation is called in the nncf quantize_pt2e method + # before the nncf_graph building. self._quantizer.annotate(anotated_model) self._quantizer.validate(anotated_model) return self.get_quantizer_config_from_anotated_model(anotated_model) diff --git a/tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/mobilenet_v3_small.dot b/tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/mobilenet_v3_small.dot index 6afb1789a12..fd17c484b76 100644 --- a/tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/mobilenet_v3_small.dot +++ b/tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/mobilenet_v3_small.dot @@ -513,25 +513,23 @@ strict digraph { "511 quantize_per_tensor_default_62" [id=511, type=quantize_per_tensor]; "512 dequantize_per_tensor_default_68" [id=512, type=dequantize_per_tensor]; "513 flatten" [id=513, type=flatten]; -"514 flatten_0_0_nncf_smooth_quant_0" [id=514, type=call_module]; -"515 quantize_per_tensor_default_63" [id=515, type=quantize_per_tensor]; -"516 dequantize_per_tensor_default_69" [id=516, type=dequantize_per_tensor]; -"517 linear_scale_0" [id=517, type=get_attr]; -"518 linear_zero_point_0" [id=518, type=get_attr]; -"519 _frozen_param52" [id=519, type=get_attr]; -"520 dequantize_per_channel_default_52" [id=520, type=dequantize_per_channel]; -"521 linear" [id=521, type=linear]; -"522 hardswish__18" [id=522, type=hardswish_]; -"523 dropout_" [id=523, type=dropout_]; -"524 dropout__0_0_nncf_smooth_quant_0" [id=524, type=call_module]; -"525 quantize_per_tensor_default_64" [id=525, type=quantize_per_tensor]; -"526 dequantize_per_tensor_default_70" [id=526, type=dequantize_per_tensor]; -"527 linear_1_scale_0" [id=527, type=get_attr]; -"528 linear_1_zero_point_0" [id=528, type=get_attr]; -"529 _frozen_param53" [id=529, type=get_attr]; -"530 dequantize_per_channel_default_53" [id=530, type=dequantize_per_channel]; -"531 linear_1" [id=531, type=linear]; -"532 output" [id=532, type=output]; +"514 quantize_per_tensor_default_63" [id=514, type=quantize_per_tensor]; +"515 dequantize_per_tensor_default_69" [id=515, type=dequantize_per_tensor]; +"516 linear_scale_0" [id=516, type=get_attr]; +"517 linear_zero_point_0" [id=517, type=get_attr]; +"518 _frozen_param52" [id=518, type=get_attr]; +"519 dequantize_per_channel_default_52" [id=519, type=dequantize_per_channel]; +"520 linear" [id=520, type=linear]; +"521 hardswish__18" [id=521, type=hardswish_]; +"522 dropout_" [id=522, type=dropout_]; +"523 quantize_per_tensor_default_64" [id=523, type=quantize_per_tensor]; +"524 dequantize_per_tensor_default_70" [id=524, type=dequantize_per_tensor]; +"525 linear_1_scale_0" [id=525, type=get_attr]; +"526 linear_1_zero_point_0" [id=526, type=get_attr]; +"527 _frozen_param53" [id=527, type=get_attr]; +"528 dequantize_per_channel_default_53" [id=528, type=dequantize_per_channel]; +"529 linear_1" [id=529, type=linear]; +"530 output" [id=530, type=output]; "0 features_1_block_1_fc1_bias" -> "48 conv2d_2" [label="(8,)", style=solid]; "1 features_1_block_1_fc2_bias" -> "56 conv2d_3" [label="(16,)", style=solid]; "2 features_4_block_2_fc1_bias" -> "148 conv2d_13" [label="(24,)", style=solid]; @@ -550,8 +548,8 @@ strict digraph { "15 features_10_block_2_fc2_bias" -> "439 conv2d_44" [label="(576,)", style=solid]; "16 features_11_block_2_fc1_bias" -> "479 conv2d_48" [label="(144,)", style=solid]; "17 features_11_block_2_fc2_bias" -> "487 conv2d_49" [label="(576,)", style=solid]; -"18 classifier_0_bias" -> "521 linear" [label="(1024,)", style=solid]; -"19 classifier_3_bias" -> "531 linear_1" [label="(1000,)", style=solid]; +"18 classifier_0_bias" -> "520 linear" [label="(1024,)", style=solid]; +"19 classifier_3_bias" -> "529 linear_1" [label="(1000,)", style=solid]; "20 x" -> "21 quantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; "21 quantize_per_tensor_default" -> "22 dequantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; "22 dequantize_per_tensor_default" -> "28 conv2d" [label="(1, 3, 224, 224)", style=solid]; @@ -1060,23 +1058,21 @@ strict digraph { "510 adaptive_avg_pool2d_9" -> "511 quantize_per_tensor_default_62" [label="(1, 576, 1, 1)", style=solid]; "511 quantize_per_tensor_default_62" -> "512 dequantize_per_tensor_default_68" [label="(1, 576, 1, 1)", style=solid]; "512 dequantize_per_tensor_default_68" -> "513 flatten" [label="(1, 576, 1, 1)", style=solid]; -"513 flatten" -> "514 flatten_0_0_nncf_smooth_quant_0" [label="(1, 576)", style=solid]; -"514 flatten_0_0_nncf_smooth_quant_0" -> "515 quantize_per_tensor_default_63" [label="(1, 576)", style=solid]; -"515 quantize_per_tensor_default_63" -> "516 dequantize_per_tensor_default_69" [label="(1, 576)", style=solid]; -"516 dequantize_per_tensor_default_69" -> "521 linear" [label="(1, 576)", style=solid]; -"517 linear_scale_0" -> "520 dequantize_per_channel_default_52" [label="(1024,)", style=solid]; -"518 linear_zero_point_0" -> "520 dequantize_per_channel_default_52" [label="(1024,)", style=solid]; -"519 _frozen_param52" -> "520 dequantize_per_channel_default_52" [label="(1024, 576)", style=solid]; -"520 dequantize_per_channel_default_52" -> "521 linear" [label="(1024, 576)", style=solid]; -"521 linear" -> "522 hardswish__18" [label="(1, 1024)", style=solid]; -"522 hardswish__18" -> "523 dropout_" [label="(1, 1024)", style=solid]; -"523 dropout_" -> "524 dropout__0_0_nncf_smooth_quant_0" [label="(1, 1024)", style=solid]; -"524 dropout__0_0_nncf_smooth_quant_0" -> "525 quantize_per_tensor_default_64" [label="(1, 1024)", style=solid]; -"525 quantize_per_tensor_default_64" -> "526 dequantize_per_tensor_default_70" [label="(1, 1024)", style=solid]; -"526 dequantize_per_tensor_default_70" -> "531 linear_1" [label="(1, 1024)", style=solid]; -"527 linear_1_scale_0" -> "530 dequantize_per_channel_default_53" [label="(1000,)", style=solid]; -"528 linear_1_zero_point_0" -> "530 dequantize_per_channel_default_53" [label="(1000,)", style=solid]; -"529 _frozen_param53" -> "530 dequantize_per_channel_default_53" [label="(1000, 1024)", style=solid]; -"530 dequantize_per_channel_default_53" -> "531 linear_1" [label="(1000, 1024)", style=solid]; -"531 linear_1" -> "532 output" [label="(1, 1000)", style=solid]; +"513 flatten" -> "514 quantize_per_tensor_default_63" [label="(1, 576)", style=solid]; +"514 quantize_per_tensor_default_63" -> "515 dequantize_per_tensor_default_69" [label="(1, 576)", style=solid]; +"515 dequantize_per_tensor_default_69" -> "520 linear" [label="(1, 576)", style=solid]; +"516 linear_scale_0" -> "519 dequantize_per_channel_default_52" [label="(1024,)", style=solid]; +"517 linear_zero_point_0" -> "519 dequantize_per_channel_default_52" [label="(1024,)", style=solid]; +"518 _frozen_param52" -> "519 dequantize_per_channel_default_52" [label="(1024, 576)", style=solid]; +"519 dequantize_per_channel_default_52" -> "520 linear" [label="(1024, 576)", style=solid]; +"520 linear" -> "521 hardswish__18" [label="(1, 1024)", style=solid]; +"521 hardswish__18" -> "522 dropout_" [label="(1, 1024)", style=solid]; +"522 dropout_" -> "523 quantize_per_tensor_default_64" [label="(1, 1024)", style=solid]; +"523 quantize_per_tensor_default_64" -> "524 dequantize_per_tensor_default_70" [label="(1, 1024)", style=solid]; +"524 dequantize_per_tensor_default_70" -> "529 linear_1" [label="(1, 1024)", style=solid]; +"525 linear_1_scale_0" -> "528 dequantize_per_channel_default_53" [label="(1000,)", style=solid]; +"526 linear_1_zero_point_0" -> "528 dequantize_per_channel_default_53" [label="(1000,)", style=solid]; +"527 _frozen_param53" -> "528 dequantize_per_channel_default_53" [label="(1000, 1024)", style=solid]; +"528 dequantize_per_channel_default_53" -> "529 linear_1" [label="(1000, 1024)", style=solid]; +"529 linear_1" -> "530 output" [label="(1, 1000)", style=solid]; } diff --git a/tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/resnet18.dot b/tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/resnet18.dot index e9db48ad8f3..da248f33521 100644 --- a/tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/resnet18.dot +++ b/tests/torch/data/reference_graphs/fx/experimental/X86InductorQuantizer/resnet18.dot @@ -203,16 +203,15 @@ strict digraph { "201 quantize_per_tensor_default_22" [id=201, type=quantize_per_tensor]; "202 dequantize_per_tensor_default_30" [id=202, type=dequantize_per_tensor]; "203 flatten" [id=203, type=flatten]; -"204 flatten_0_0_nncf_smooth_quant_0" [id=204, type=call_module]; -"205 quantize_per_tensor_default_23" [id=205, type=quantize_per_tensor]; -"206 dequantize_per_tensor_default_31" [id=206, type=dequantize_per_tensor]; -"207 linear_scale_0" [id=207, type=get_attr]; -"208 linear_zero_point_0" [id=208, type=get_attr]; -"209 _frozen_param20" [id=209, type=get_attr]; -"210 dequantize_per_channel_default_20" [id=210, type=dequantize_per_channel]; -"211 linear" [id=211, type=linear]; -"212 output" [id=212, type=output]; -"0 fc_bias" -> "211 linear" [label="(1000,)", style=solid]; +"204 quantize_per_tensor_default_23" [id=204, type=quantize_per_tensor]; +"205 dequantize_per_tensor_default_31" [id=205, type=dequantize_per_tensor]; +"206 linear_scale_0" [id=206, type=get_attr]; +"207 linear_zero_point_0" [id=207, type=get_attr]; +"208 _frozen_param20" [id=208, type=get_attr]; +"209 dequantize_per_channel_default_20" [id=209, type=dequantize_per_channel]; +"210 linear" [id=210, type=linear]; +"211 output" [id=211, type=output]; +"0 fc_bias" -> "210 linear" [label="(1000,)", style=solid]; "1 x" -> "2 quantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; "2 quantize_per_tensor_default" -> "3 dequantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; "3 dequantize_per_tensor_default" -> "9 conv2d" [label="(1, 3, 224, 224)", style=solid]; @@ -423,13 +422,12 @@ strict digraph { "200 adaptive_avg_pool2d" -> "201 quantize_per_tensor_default_22" [label="(1, 512, 1, 1)", style=solid]; "201 quantize_per_tensor_default_22" -> "202 dequantize_per_tensor_default_30" [label="(1, 512, 1, 1)", style=solid]; "202 dequantize_per_tensor_default_30" -> "203 flatten" [label="(1, 512, 1, 1)", style=solid]; -"203 flatten" -> "204 flatten_0_0_nncf_smooth_quant_0" [label="(1, 512)", style=solid]; -"204 flatten_0_0_nncf_smooth_quant_0" -> "205 quantize_per_tensor_default_23" [label="(1, 512)", style=solid]; -"205 quantize_per_tensor_default_23" -> "206 dequantize_per_tensor_default_31" [label="(1, 512)", style=solid]; -"206 dequantize_per_tensor_default_31" -> "211 linear" [label="(1, 512)", style=solid]; -"207 linear_scale_0" -> "210 dequantize_per_channel_default_20" [label="(1000,)", style=solid]; -"208 linear_zero_point_0" -> "210 dequantize_per_channel_default_20" [label="(1000,)", style=solid]; -"209 _frozen_param20" -> "210 dequantize_per_channel_default_20" [label="(1000, 512)", style=solid]; -"210 dequantize_per_channel_default_20" -> "211 linear" [label="(1000, 512)", style=solid]; -"211 linear" -> "212 output" [label="(1, 1000)", style=solid]; +"203 flatten" -> "204 quantize_per_tensor_default_23" [label="(1, 512)", style=solid]; +"204 quantize_per_tensor_default_23" -> "205 dequantize_per_tensor_default_31" [label="(1, 512)", style=solid]; +"205 dequantize_per_tensor_default_31" -> "210 linear" [label="(1, 512)", style=solid]; +"206 linear_scale_0" -> "209 dequantize_per_channel_default_20" [label="(1000,)", style=solid]; +"207 linear_zero_point_0" -> "209 dequantize_per_channel_default_20" [label="(1000,)", style=solid]; +"208 _frozen_param20" -> "209 dequantize_per_channel_default_20" [label="(1000, 512)", style=solid]; +"209 dequantize_per_channel_default_20" -> "210 linear" [label="(1000, 512)", style=solid]; +"210 linear" -> "211 output" [label="(1, 1000)", style=solid]; } From e285dc64a2daec90d93cfb9c9ecdacb1693afcab Mon Sep 17 00:00:00 2001 From: dlyakhov Date: Thu, 5 Dec 2024 16:44:13 +0100 Subject: [PATCH 06/15] batchwise_statistics --- .../algorithms/post_training/algorithm.py | 4 +++ .../algorithms/post_training/pipeline.py | 4 +++ .../torch/fx/quantization/quantize_pt2e.py | 35 +++++++++++++++++-- 3 files changed, 40 insertions(+), 3 deletions(-) diff --git a/nncf/experimental/common/quantization/algorithms/post_training/algorithm.py b/nncf/experimental/common/quantization/algorithms/post_training/algorithm.py index ac2d86a3056..2b61689c638 100644 --- a/nncf/experimental/common/quantization/algorithms/post_training/algorithm.py +++ b/nncf/experimental/common/quantization/algorithms/post_training/algorithm.py @@ -45,6 +45,7 @@ def __init__( smooth_quant_params: Optional[AdvancedSmoothQuantParameters] = None, activations_range_estimator_params: Optional[RangeEstimatorParameters] = None, weights_range_estimator_params: Optional[RangeEstimatorParameters] = None, + batchwise_statistics: bool = False, ): """ :param quantizer: NNCFQuantizer to use in MiMaxRangeInit algorithm. @@ -60,6 +61,8 @@ def __init__( of activations of the model. :param weights_range_estimator_params: Contains parameters for estimating the range of weights of the model. + :param batchwise_statistics: Determines whether quantizer statistics should be calculated + for each item of the batch or for the entire batch, default is False. """ self._pipeline = experimental_create_ptq_pipeline( quantizer=quantizer, @@ -70,6 +73,7 @@ def __init__( smooth_quant_params=smooth_quant_params, activations_range_estimator_params=activations_range_estimator_params, weights_range_estimator_params=weights_range_estimator_params, + batchwise_statistics=batchwise_statistics, ) @property diff --git a/nncf/experimental/common/quantization/algorithms/post_training/pipeline.py b/nncf/experimental/common/quantization/algorithms/post_training/pipeline.py index 5c8e521f65a..c7dbaf796c3 100644 --- a/nncf/experimental/common/quantization/algorithms/post_training/pipeline.py +++ b/nncf/experimental/common/quantization/algorithms/post_training/pipeline.py @@ -35,6 +35,7 @@ def experimental_create_ptq_pipeline( smooth_quant_params: Optional[AdvancedSmoothQuantParameters] = None, activations_range_estimator_params: Optional[RangeEstimatorParameters] = None, weights_range_estimator_params: Optional[RangeEstimatorParameters] = None, + batchwise_statistics: bool = None, ) -> Pipeline: """ Creates an experimental post-training quantization pipeline. @@ -57,6 +58,8 @@ def experimental_create_ptq_pipeline( of activations of the model. :param weights_range_estimator_params: Contains parameters for estimating the range of weights of the model. + :param batchwise_statistics: Determines whether quantizer statistics should be calculated + for each item of the batch or for the entire batch, default is False. :return: An experimental post-training quantization pipeline. """ @@ -77,6 +80,7 @@ def experimental_create_ptq_pipeline( quantizer=quantizer, subset_size=subset_size, inplace_statistics=False, + batchwise_statistics=batchwise_statistics, activations_range_estimator_params=activations_range_estimator_params, weights_range_estimator_params=weights_range_estimator_params, ) diff --git a/nncf/experimental/torch/fx/quantization/quantize_pt2e.py b/nncf/experimental/torch/fx/quantization/quantize_pt2e.py index ed41166b9ca..f6267ecfb69 100644 --- a/nncf/experimental/torch/fx/quantization/quantize_pt2e.py +++ b/nncf/experimental/torch/fx/quantization/quantize_pt2e.py @@ -22,6 +22,7 @@ from torch.fx import GraphModule from torch.fx.passes.infra.pass_manager import PassManager +import nncf from nncf.common.factory import NNCFGraphFactory from nncf.common.logging import nncf_logger from nncf.data import Dataset @@ -43,16 +44,36 @@ def quantize_pt2e( quantizer: Quantizer, calibration_dataset: Dataset, subset_size: int = 300, - fast_bias_correction: bool = True, + fast_bias_correction: Optional[bool] = True, smooth_quant: bool = False, bias_correction_params: Optional[AdvancedBiasCorrectionParameters] = None, smooth_quant_params: Optional[AdvancedSmoothQuantParameters] = None, activations_range_estimator_params: Optional[RangeEstimatorParameters] = None, weights_range_estimator_params: Optional[RangeEstimatorParameters] = None, - fold_quantize: Optional[bool] = False, + batchwise_statistics: bool = False, + fold_quantize: bool = False, ) -> torch.fx.GraphModule: """ - Implementation of the `quantize()` method for the Torch FX backend. + Applies post-training quantization to the torch.fx.GraphModule provided model + using provided torch.ao quantizer. + + :param quantizer: Torch ao quantizer to annotate nodes in the graph with quantization setups + to convey the desired way of quantization. + :param subset_size: Size of a subset to calculate activations + statistics used for quantization. + :param fast_bias_correction: Setting this option to `False` enables a different + bias correction method which is more accurate, in general, and takes + more time but requires less memory. None disables the bias correction algorithm. + :param smooth_quant: Setting this option to `True` enables the SmoothQuant algorithm. + :param bias_correction_params: Contains advanced parameters for fine-tuning bias correction algorithm. + :param smooth_quant_params: Contains advanced alpha parameters for SmoothQuant algorithm. + :param activations_range_estimator_params: Contains parameters for estimating the range + of activations of the model. + :param weights_range_estimator_params: Contains parameters for estimating the range + of weights of the model. + :param batchwise_statistics: Determines whether quantizer statistics should be calculated + for each item of the batch or for the entire batch, default is False. + :param fold_quantize: Boolean flag for whether fold the quantize op or not. """ nncf_logger.warning( "Experimental Torch FX quantization backend is being used for the given torch.fx.GraphModule model." @@ -60,6 +81,13 @@ def quantize_pt2e( " in case of errors or a poor model performance." ) + if subset_size < 1: + raise nncf.ValidationError("Subset size must be positive.") + + batch_size = calibration_dataset.get_batch_size() + if batch_size is not None and batch_size > 1 and batchwise_statistics is not None: + batchwise_statistics = True + original_graph_meta = model.meta copied_model = deepcopy(model) @@ -83,6 +111,7 @@ def quantize_pt2e( smooth_quant_params=smooth_quant_params, activations_range_estimator_params=activations_range_estimator_params, weights_range_estimator_params=weights_range_estimator_params, + batchwise_statistics=batchwise_statistics, ) nncf_graph = NNCFGraphFactory.create(copied_model) From 4c5e27f2f27f96d8887dac7efea57cc0e054aa5e Mon Sep 17 00:00:00 2001 From: dlyakhov Date: Fri, 29 Nov 2024 11:30:16 +0100 Subject: [PATCH 07/15] WIP OpenVINOQuantizer --- .../quantizer/openvino_quantizer/backend.py | 323 + .../openvino_quantizer/openvino_quantizer.py | 593 ++ .../openvino_quantizer/torch_fx_backend.py | 366 + .../algorithms/min_max/quantizer/backend.py | 323 + .../min_max/quantizer/torch_fx_backend.py | 366 + .../OpenVINOQuantizer/mobilenet_v3_small.dot | 1042 +++ .../OpenVINOQuantizer/resnet18.dot | 457 ++ .../OpenVINOQuantizer/swin_v2_s.dot | 6310 +++++++++++++++++ .../synthetic_transformer.dot | 45 + .../experimental/OpenVINOQuantizer/unet.dot | 469 ++ .../OpenVINOQuantizer/vit_b_16.dot | 1913 +++++ .../OpenVINOQuantizer/yolo11n_sdpa_block.dot | 61 + 12 files changed, 12268 insertions(+) create mode 100644 nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer/backend.py create mode 100644 nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer/openvino_quantizer.py create mode 100644 nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer/torch_fx_backend.py create mode 100644 nncf/quantization/algorithms/min_max/quantizer/backend.py create mode 100644 nncf/quantization/algorithms/min_max/quantizer/torch_fx_backend.py create mode 100644 tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/mobilenet_v3_small.dot create mode 100644 tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/resnet18.dot create mode 100644 tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/swin_v2_s.dot create mode 100644 tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/synthetic_transformer.dot create mode 100644 tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/unet.dot create mode 100644 tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/vit_b_16.dot create mode 100644 tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/yolo11n_sdpa_block.dot diff --git a/nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer/backend.py b/nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer/backend.py new file mode 100644 index 00000000000..9f7ea5441e1 --- /dev/null +++ b/nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer/backend.py @@ -0,0 +1,323 @@ +# Copyright (c) 2024 Intel Corporation +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# http://www.apache.org/licenses/LICENSE-2.0 +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from abc import ABC +from abc import abstractmethod +from typing import Dict, List, Optional, Set, Tuple, TypeVar + +from nncf.common.graph.graph import NNCFGraph +from nncf.common.graph.graph import NNCFNode +from nncf.common.graph.operator_metatypes import OperatorMetatype +from nncf.common.graph.transformations.commands import TargetPoint +from nncf.common.graph.transformations.commands import TargetType +from nncf.common.graph.transformations.commands import TransformationCommand +from nncf.common.hardware.config import HWConfig +from nncf.common.quantization.structs import QuantizerConfig +from nncf.common.tensor_statistics.collectors import TensorStatisticCollectorBase +from nncf.parameters import ModelType +from nncf.parameters import TargetDevice +from nncf.quantization.fake_quantize import FakeConvertParameters +from nncf.quantization.fake_quantize import FakeQuantizeParameters +from nncf.quantization.range_estimator import RangeEstimatorParameters + +TModel = TypeVar("TModel") + + +class OpenVINOQuantizerBackend(ABC): + @property + @abstractmethod + def preserved_metatypes(self) -> List[OperatorMetatype]: + """ + Property for backend-specific metatypes that require preserving float subgraphs + when removing the ShapeOf subgraph. + """ + + @property + @abstractmethod + def mat_mul_metatypes(self) -> List[OperatorMetatype]: + """ + Property for the backend-specific MatMul metatypes. + """ + + @property + @abstractmethod + def post_processing_metatypes(self) -> List[OperatorMetatype]: + """ + Property for the backend-specific post-processing metatypes (NonMaximumSupression, TopK, etc.). + """ + + @property + @abstractmethod + def conv_metatypes(self) -> List[OperatorMetatype]: + """ + Property for the backend-specific Convolution metatypes. + """ + + @property + @abstractmethod + def shapeof_metatypes(self) -> List[OperatorMetatype]: + """ + Property for the backend-specific ShapeOf metatypes. + """ + + @property + @abstractmethod + def dropout_metatypes(self) -> List[OperatorMetatype]: + """ + Property for the backend-specific Dropout metatypes. + """ + + @property + @abstractmethod + def elementwise_metatypes(self) -> List[OperatorMetatype]: + """ + Property for the backend-specific Elementwises metatypes. + """ + + @property + @abstractmethod + def overflow_fix_metatypes(self) -> List[OperatorMetatype]: + """ + Property for the backend-specific metatypes for which overflow_fix is applicable. + """ + + @property + @abstractmethod + def add_metatypes(self) -> List[OperatorMetatype]: + """ + Property for the backend-specific metatypes that also can be interpreted as Add layer. + """ + + @property + @abstractmethod + def group_conv_metatypes(self) -> List[OperatorMetatype]: + """ + Property for the backend-specific Grouped Convolution metatypes. + """ + + @property + @abstractmethod + def scaled_dot_product_attention_metatypes(self) -> List[OperatorMetatype]: + """ + Property for the backend-specific Scaled Dot Product Attention metatypes. + """ + + @property + @abstractmethod + def scales_unification_map(self) -> Dict[OperatorMetatype, OperatorMetatype]: + """ + Property for the backend-specific metatypes that produces quantizers that might be unified. + """ + + @property + @abstractmethod + def hw_config(self) -> HWConfig: + """ + Property for the hardware backend-specific configuration. + """ + + @property + @abstractmethod + def quant_trait_op_dict(self) -> Dict[int, OperatorMetatype]: + """ + Property for the backend-specific dictionary that contains QuantizationTrait-specific metatypes. + """ + + @staticmethod + @abstractmethod + def target_point(target_type: TargetType, target_node_name: str, port_id: int) -> TargetPoint: + """ + Returns backend-specific target point. + + :param target_type: Type of the location that should be modified. + :param target_node_name: Name of the located node. + :param port_id: Port ID of the tensor for the statistics distribution. + :return: Backend-specific TargetPoint. + """ + + @staticmethod + @abstractmethod + def create_quantizer_insertion_command( + nncf_graph: NNCFGraph, + target_point: TargetPoint, + quantizer_config: QuantizerConfig, + parameters: FakeQuantizeParameters, + ) -> TransformationCommand: + """ + Returns backend-specific quantizer insertion command. + + :param nncf_graph: NNCFGraph to get input/output shapes for the target point. + :param target_point: Target location for the quantizer insertion. + :param quantizer_config: QuantizerConfig instance for the current layer. + :param parameters: FakeQuantizeParameters to calculate activation quantization parameters. + :return: Backend-specific TransformationCommand for the quantizer insertion operation. + """ + + @staticmethod + @abstractmethod + def create_unified_scales_quantizers_insertion_commands( + nncf_graph: NNCFGraph, + target_points: List[TargetPoint], + quantizer_config: QuantizerConfig, + parameters: FakeQuantizeParameters, + ) -> List[TransformationCommand]: + """ + Returns backend-specific unified scales quantizers insertion commands. + + :param nncf_graph: NNCFGraph to get input/output shapes for the target point. + :param target_points: List of target locations for the quantizers insertion. + :param quantizer_config: QuantizerConfig instance for the current layer. + :param parameters: FakeQuantizeParameters to calculate activation quantization parameters. + :return: List of backend-specific TransformationCommands + for the quantizers with unified scales insertion operations. + """ + + @staticmethod + @abstractmethod + def create_convert_insertion_command( + target_point: TargetPoint, + parameters: FakeConvertParameters, + ) -> TransformationCommand: + """ + Returns backend-specific convert insertion command. + + :param target_point: Target location for the correction. + :param parameters: FakeConvertParameters to calculate activation quantization parameters. + :return: Backend-specific TransformationCommand for the quantizer insertion operation. + """ + + @staticmethod + @abstractmethod + def get_start_nodes_for_activation_path_tracing(nncf_graph: NNCFGraph) -> List[NNCFNode]: + """ + Returns a list of NNCFNodes to use as start nodes for activation path tracing. + + :param nncf_graph: NNCFGraph to get the start nodes. + :return: List of NNCFNodes to use as start nodes for activation path tracing. + + """ + + @staticmethod + @abstractmethod + def get_target_point_shape(nncf_graph: NNCFGraph, node: NNCFNode, target_point: TargetPoint) -> Tuple[int, ...]: + """ + Returns shape of a target point tensor. + + :param nncf_graph: NNCFGraph instance. + :param node: NNCFNode. + :param target_point: Target point of which tensor shape is seeked. + :return: Shape of target point tensor. + """ + + @staticmethod + @abstractmethod + def get_weight_quantization_axes(node: NNCFNode, target_point: TargetPoint, ndims: int) -> Tuple[int, ...]: + """ + Returns axes for per-channel quantization of weights of the node placed on a input port_id. + + :param node: Quantized node with the weight. + :param target_point: Corresponding target point. + :param ndims: Number of dimensions of weight. + :return: Axes for per-channel quantization of weights. + """ + + @staticmethod + @abstractmethod + def get_statistic_collector( + range_estimator_params: RangeEstimatorParameters, + use_abs_max: bool, + reduction_axes: Optional[Tuple[int, ...]], + aggregation_axes: Optional[Tuple[int, ...]], + inplace: bool, + num_samples: Optional[int] = None, + ) -> TensorStatisticCollectorBase: + """ + Returns backend-specific statistic collector. + + :param range_estimator_params: Parameters that specify estimators types. + :param use_abs_max: Wheather reduce absolute values of input tensors or not. + :param reduction_axes: Axes for reducer. + :param aggregation_axes: Axes for aggregator. + :param inplace: Whether to calculate statistic inplace or not. + :param num_samples: Maximum number of samples to collect. + :return: Backend-specific TensorStatisticCollectorBase for the statistics calculation. + """ + + @staticmethod + @abstractmethod + def get_weight_tensor_port_ids(node: NNCFNode, graph: NNCFGraph) -> List[Optional[int]]: + """ + Returns node's input port indices with weight tensors. + + :param node: NNCFNode to find its weight input port indices. + :param graph: NNCFGraph instance. + :return: Weights input port indices. + """ + + @staticmethod + def get_weight_name(nncf_graph: NNCFGraph, target_point: TargetPoint) -> str: + """ + Returns node's weight name corresponding to port ID. + + :param nncf_graph: NNCFGraph instance. + :param target_point: The TargetPoint instance that contains layer's information. + :return: Weight name. + """ + + @staticmethod + def should_quantize_weight(weight_name: str, quantized_weight_names: Set[str]) -> bool: + """ + Return True if weight should be quantized. + + :param weight_name: Weight name. + :param quantized_weight_names: Set containing already quantized weight names. + :return: A boolean value specifying whether a weight should be quantized. + """ + + @staticmethod + @abstractmethod + def get_ignored_metatypes(model_type: ModelType, device: TargetDevice) -> List[OperatorMetatype]: + """ + Returns ignored metatypes based on a model type and device parameters. + + :param model_type: Model type parameter. + :param device: Target device. + :return: List of ignored metatypes. + """ + + @staticmethod + @abstractmethod + def get_ignored_names_by_layer_attributes(nncf_graph: NNCFGraph) -> Set[str]: + """ + Returns names of ignored nodes based on layer_attributes. + + :param nncf_graph: NNCFGraph instance. + :return: List of ignored names. + """ + + @abstractmethod + def get_weight_nodes(self, nncf_graph: NNCFGraph) -> List[NNCFNode]: + """ + Returns nodes that have weights. + + :param nncf_graph: Instance of NNCFGraph. + :return: All nodes with weights. + """ + + @abstractmethod + def is_matmul_with_constant(self, node: NNCFNode, nncf_graph: NNCFGraph) -> bool: + """ + Returns true if given nncf matmul node is a matmul with a constant, False otherwise. + + :param Node: Instance of NNCFNode. + :param nncf_graph: Instance of NNCFGraph. + :return: True if given nncf matmul node is a matmul with a constant, False otherwise. + """ diff --git a/nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer/openvino_quantizer.py b/nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer/openvino_quantizer.py new file mode 100644 index 00000000000..11ce83637c2 --- /dev/null +++ b/nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer/openvino_quantizer.py @@ -0,0 +1,593 @@ +# Copyright (c) 2024 Intel Corporation +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# http://www.apache.org/licenses/LICENSE-2.0 +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import dataclasses +from copy import deepcopy +from typing import Dict, List, Optional, Set, TypeVar, Union + +import numpy as np + +import nncf +from nncf.common.graph.graph import NNCFGraph +from nncf.common.graph.patterns import GraphPattern +from nncf.common.graph.patterns.manager import PatternsManager +from nncf.common.hardware.config import get_hw_config_type +from nncf.common.insertion_point_graph import InsertionPointGraph +from nncf.common.logging import nncf_logger +from nncf.common.quantization.config_assignment import assign_qconfig_lists_to_modules +from nncf.common.quantization.quantizer_propagation.solver import QuantizerPropagationRule +from nncf.common.quantization.quantizer_propagation.solver import QuantizerPropagationSolver +from nncf.common.quantization.quantizer_propagation.structs import IgnoreReason +from nncf.common.quantization.quantizer_setup import SingleConfigQuantizerSetup +from nncf.common.quantization.structs import QuantizableWeightedLayerNode +from nncf.common.quantization.structs import QuantizationConstraints +from nncf.common.quantization.structs import QuantizationPreset +from nncf.common.quantization.structs import QuantizationScheme +from nncf.common.quantization.structs import QuantizerConfig +from nncf.common.quantization.structs import QuantizerGroup +from nncf.common.utils.backend import BackendType +from nncf.common.utils.backend import get_backend +from nncf.parameters import ModelType +from nncf.parameters import QuantizationMode +from nncf.parameters import TargetDevice +from nncf.quantization.advanced_parameters import FP8QuantizationParameters +from nncf.quantization.advanced_parameters import FP8Type +from nncf.quantization.advanced_parameters import OverflowFix +from nncf.quantization.advanced_parameters import QuantizationParameters +from nncf.experimental.common.quantization.algorithms.quantizer.quantizer import NNCFQuantizer +from nncf.quantization.passes import transform_to_inference_graph +from nncf.scopes import IgnoredScope +from nncf.scopes import get_ignored_node_names_from_ignored_scope + +TModel = TypeVar("TModel") + +DEFAULT_QCONFIG = QuantizerConfig( + num_bits=8, mode=QuantizationScheme.SYMMETRIC, signedness_to_force=None, per_channel=False +) + + +@dataclasses.dataclass +class ModeBasedDefaults: + """ + Contains default values that should be set in case of abscense. + """ + + overflow_fix: OverflowFix = OverflowFix.FIRST_LAYER + activations_quantization_params: Union[QuantizationParameters, FP8QuantizationParameters] = dataclasses.field( + default_factory=QuantizationParameters + ) + weights_quantization_params: Union[QuantizationParameters, FP8QuantizationParameters] = dataclasses.field( + default_factory=QuantizationParameters + ) + + +MODE_BASED_DEFAULTS = { + None: ModeBasedDefaults(), + QuantizationMode.FP8_E4M3: ModeBasedDefaults( + overflow_fix=OverflowFix.DISABLE, + activations_quantization_params=FP8QuantizationParameters(FP8Type.E4M3), + weights_quantization_params=FP8QuantizationParameters(FP8Type.E4M3), + ), + QuantizationMode.FP8_E5M2: ModeBasedDefaults( + overflow_fix=OverflowFix.DISABLE, + activations_quantization_params=FP8QuantizationParameters(FP8Type.E5M2), + weights_quantization_params=FP8QuantizationParameters(FP8Type.E5M2), + ), +} + + +class NNCFOVQuantizer(NNCFQuantizer): + def __init__( + self, + mode: Optional[QuantizationMode] = None, + preset: Optional[QuantizationPreset] = None, + target_device: TargetDevice = TargetDevice.ANY, + model_type: Optional[ModelType] = None, + ignored_scope: Optional[IgnoredScope] = None, + overflow_fix: Optional[OverflowFix] = None, + quantize_outputs: bool = False, + activations_quantization_params: Union[QuantizationParameters, FP8QuantizationParameters] = None, + weights_quantization_params: Union[QuantizationParameters, FP8QuantizationParameters] = None, + quantizer_propagation_rule: Optional[QuantizerPropagationRule] = None, + ): + """ + :param mode: Defines optimization mode for the algorithm. None by default. + :param preset: A preset controls the quantization mode (symmetric and asymmetric). + It can take the following values: + - `performance`: Symmetric quantization of weights and activations. + - `mixed`: Symmetric quantization of weights and asymmetric quantization of activations. + Default value is None. In this case, `mixed` preset is used for `transformer` + model type otherwise `performance`. + :param target_device: A target device the specificity of which will be taken + into account while compressing in order to obtain the best performance + for this type of device, defaults to TargetDevice.ANY. + :param model_type: Model type is needed to specify additional patterns + in the model. Supported only `transformer` now. + :param ignored_scope: An ignored scope that defined the list of model control + flow graph nodes to be ignored during quantization. + :param overflow_fix: This option controls whether to apply the overflow issue + fix for the 8-bit quantization. + :param quantize_outputs: Whether to insert additional quantizers right before + each of the model outputs. + :param activations_quantization_params: Quantization parameters for model + activations. + :param weights_quantization_params: Quantization parameters for model weights. + :param quantizer_propagation_rule: The strategy to be used while propagating and merging quantizers. + """ + self._target_device = target_device + self._mode = mode + self._model_type = model_type + self._overflow_fix = overflow_fix + self._quantize_outputs = quantize_outputs + self._activations_quantization_params = activations_quantization_params + self._weights_quantization_params = weights_quantization_params + self._preset = preset + self._ignored_scope = IgnoredScope() if ignored_scope is None else ignored_scope + self.quantizer_propagation_rule = quantizer_propagation_rule + + # preset definition + if self._preset is None: + if model_type == ModelType.TRANSFORMER: + self._preset = QuantizationPreset.MIXED + else: + self._preset = QuantizationPreset.PERFORMANCE + + self._override_device() + self._set_mode_based_defaults() + self._review_mode_based_defaults() + + self._quantization_params = { + QuantizerGroup.WEIGHTS: self._weights_quantization_params, + QuantizerGroup.ACTIVATIONS: self._activations_quantization_params, + } + + # Calculates global quantizer constraints + self._global_quantizer_constraints = {} + for quantizer_group in QuantizerGroup: + self._global_quantizer_constraints[quantizer_group] = self._get_quantizer_constraints( + quantizer_group, self._preset, self._quantization_params[quantizer_group] + ) + + self._algorithm_key = f"MMQ_{hash(self)}" + + def _override_device(self) -> None: + """ + Overrides NPU device to use CPU quantization scheme. + """ + if self._target_device == TargetDevice.NPU: + act_bits, weight_bits = 8, 8 + if self._activations_quantization_params and self._activations_quantization_params.num_bits: + act_bits = self._activations_quantization_params.num_bits + if self._weights_quantization_params and self._weights_quantization_params.num_bits: + weight_bits = self._weights_quantization_params.num_bits + + if act_bits == 8 and weight_bits == 8: + self._target_device == TargetDevice.CPU + nncf_logger.debug("Target device NPU was changed to CPU!") + + def _set_mode_based_defaults(self) -> None: + """ + Sets defaults for the algorithms based on the provided mode. + """ + mode_based_defaults = MODE_BASED_DEFAULTS[self._mode] + for field in dataclasses.fields(mode_based_defaults): + self_name = "_" + field.name + default_value = getattr(mode_based_defaults, field.name) + if getattr(self, self_name) is None: + setattr(self, self_name, default_value) + + def _review_mode_based_defaults(self): + """ + Reviews default values because mode option doesn't support them. + """ + if self._mode in (QuantizationMode.FP8_E4M3, QuantizationMode.FP8_E5M2): + nncf_logger.warning(f"You're using experimental option mode with {self._mode} value.") + + if self._preset != QuantizationPreset.PERFORMANCE: + raise nncf.ParameterNotSupportedError( + f"preset option with {self._preset} value is not supported with the mode option!" + ) + + if self._target_device not in [TargetDevice.CPU, TargetDevice.ANY]: + raise nncf.ParameterNotSupportedError( + f"target_device option with {self._target_device} value is not supported with the mode option!" + ) + + if self._overflow_fix != OverflowFix.DISABLE: + raise nncf.ParameterNotSupportedError( + f"overflow_fix option with {self._overflow_fix} value is not supported with the mode option!" + ) + + if self._quantize_outputs: + raise nncf.ParameterNotSupportedError("quantize_outputs option is not supported with the mode option!") + + if isinstance(self._weights_quantization_params, QuantizationParameters): + raise nncf.ParameterNotSupportedError( + "quantization_params option for weights with " + f"{self._weights_quantization_params} " + "value is not supported with the mode option!" + ) + + if isinstance(self._activations_quantization_params, QuantizationParameters): + raise nncf.ParameterNotSupportedError( + "quantization_params option for activations with " + f"{self._activations_quantization_params} " + "value is not supported with the mode option!" + ) + elif self._mode is None: + if isinstance(self._weights_quantization_params, FP8QuantizationParameters): + raise nncf.ParameterNotSupportedError( + "quantization_params option for weights with " + f"{self._weights_quantization_params} " + "value is not supported with the mode: None option!" + ) + + if isinstance(self._activations_quantization_params, FP8QuantizationParameters): + raise nncf.ParameterNotSupportedError( + "quantization_params option for activations with " + f"{self._activations_quantization_params} " + "value is not supported with the mode: None option!" + ) + + @property + def available_backends(self) -> List[BackendType]: + return [BackendType.ONNX, BackendType.OPENVINO, BackendType.TORCH, BackendType.TORCH_FX] + + def _get_quantizer_constraints( + self, + group: QuantizerGroup, + preset: QuantizationPreset, + quantization_params: Union[QuantizationParameters, FP8QuantizationParameters], + ) -> QuantizationConstraints: + """ + Returns QuantizationConstraints for the provided quantizer group. + + :param group: Quantizer group. + :param preset: Quantization preset. + :param quantization_params: Quantization parameters. + :return: QuantizationConstraints. + """ + constraints = {"mode": preset.get_params_configured_by_preset(group)["mode"]} + if quantization_params is None: + return QuantizationConstraints(**constraints) + + if isinstance(quantization_params, FP8QuantizationParameters): + if self._mode is None: + raise nncf.InternalError( + f"FP8QuantizationParameters for {group.value} can not be used without QuantizationMode option!" + ) + return QuantizationConstraints(**constraints) + + if quantization_params.mode is not None: + constraints["mode"] = quantization_params.mode + if quantization_params.num_bits is not None: + constraints["num_bits"] = quantization_params.num_bits + if quantization_params.per_channel is not None: + constraints["per_channel"] = quantization_params.per_channel + if quantization_params.signedness_to_force is not None: + constraints["signedness_to_force"] = quantization_params.signedness_to_force + + return QuantizationConstraints(**constraints) + + def _set_backend_entity(self, model: TModel) -> None: + """ + Creates a helper class with a backed-specific logic of the algorithm + + :param model: backend-specific input model + """ + model_backend = get_backend(model) + if model_backend == BackendType.TORCH_FX: + from nncf.experimental.common.quantization.algorithms.quantizer.openvino_quantizer.torch_fx_backend import OpenVINOQuantizerBackend + + self._backend_entity = OpenVINOQuantizerBackend() + else: + raise nncf.UnsupportedBackendError( + "Cannot return backend-specific entity because {} is not supported!".format(model_backend.value) + ) + + def _get_default_qconfig(self, constraints: QuantizationConstraints = None) -> QuantizerConfig: + """ + Returns default quantizer configuration, based on the provided constraints. + + :param constraints: Quantization constraints. + :return: Quantizer config. + """ + qconfig = deepcopy(DEFAULT_QCONFIG) + if constraints is not None: + qconfig = constraints.apply_constraints_to(qconfig) + return qconfig + + def _get_ignored_names( + self, nncf_graph: NNCFGraph, inference_nncf_graph: NNCFGraph, ignored_patterns: GraphPattern + ) -> Dict[str, IgnoreReason]: + """ + Returns all node names that are ignored for quantization: + Firstly, the ignored names are obtained from user-defined ignored the scope. + Secondly, the ignored names are updated from model_type parameter. + Lastly, the ignored names are updated from ignored_patterns. + + :param nncf_graph: NNCFGraph instance. + :param inference_nncf_graph: Inference graph without constant flows. + :param ignored_patterns: Ignored patterns. + :return: Ignored node names and ignore reason for quantization. + """ + user_ignored_names = get_ignored_node_names_from_ignored_scope( + self._ignored_scope, nncf_graph, strict=self._ignored_scope.validate + ) + autogenerated_ignored_names = self._get_ignored_names_by_ignored_patterns( + inference_nncf_graph, ignored_patterns + ) + autogenerated_ignored_names |= self._backend_entity.get_ignored_names_by_layer_attributes(inference_nncf_graph) + autogenerated_ignored_names |= self._get_ignored_names_by_algorithm(inference_nncf_graph) + ignored_names = {name: IgnoreReason.AUTOGENERATED for name in autogenerated_ignored_names} + # User ignored scope has higher priority + ignored_names.update({name: IgnoreReason.USER_REQUESTED for name in user_ignored_names}) + return ignored_names + + def _get_ignored_names_by_ignored_patterns( + self, inference_nncf_graph: NNCFGraph, ignored_patterns: GraphPattern + ) -> Set[str]: + """ + Returns node names matched ignored_patterns. + + :param nncf_graph: Inference graph without constant flows. + :param ignored_patterns: Ignored patterns. + :return: IgnoredScope with all node names matched ignored_patterns. + """ + nncf_node_names = set() + for subgraph in inference_nncf_graph.find_matching_subgraphs(ignored_patterns, strict=False): + for nncf_node in subgraph: + nncf_node_names.add(nncf_node.node_name) + return nncf_node_names + + def _get_ignored_names_by_algorithm(self, inference_nncf_graph: NNCFGraph) -> Set[str]: + """ + Returns node names for ignored_algorithms matched `quantization`. + + :param inference_nncf_graph: Inference NNCFGraph instance. + :return: IgnoredScope with corresponded nodes. + """ + nncf_node_names = set() + for nncf_node in inference_nncf_graph.get_all_nodes(): + if "ptq_quantization" in nncf_node.ignored_algorithms: + nncf_node_names.add(nncf_node.node_name) + return nncf_node_names + + def _get_scope_overrides(self, inference_nncf_graph: NNCFGraph) -> Dict: + """ + Returns a dictionary of quantization configuration overrides for inputs to matching operation nodes. + + :param inference_nncf_graph: Inference NNCFGraph instance. + :return: A dictionary of quantization configuration overrides for inputs to matching operation nodes. + """ + scaled_dot_product_attention_node_names = [ + node.node_name + for node in inference_nncf_graph.get_nodes_by_metatypes( + self._backend_entity.scaled_dot_product_attention_metatypes + ) + ] + + scope_overrides_activations = {} + for node_name in scaled_dot_product_attention_node_names: + scope_overrides_activations[node_name] = {"mode": "symmetric"} + return {"activations": scope_overrides_activations} + + def _get_quantizer_setup( + self, + nncf_graph: NNCFGraph, + inference_nncf_graph: NNCFGraph, + hw_patterns: GraphPattern, + ignored_patterns: GraphPattern, + ) -> SingleConfigQuantizerSetup: + """ + Returns SingleConfigQuantizerSetup instance based on the input NNCFGraph. + + :param nncf_graph: NNCFGraph instance. + :param hw_patterns: Hardware patterns. + :param ignored_patterns: Ignored patterns. + :return: SingleConfigQuantizerSetup for the current NNCFGraph entity. + """ + hw_config_type = get_hw_config_type(self._target_device.value) + hw_config_path = self._backend_entity.hw_config.get_path_to_hw_config(hw_config_type) + hw_config = self._backend_entity.hw_config.from_json(hw_config_path) + + ignored_names = self._get_ignored_names(nncf_graph, inference_nncf_graph, ignored_patterns) + weight_nodes = self._backend_entity.get_weight_nodes(nncf_graph) + + default_weight_qconfig = self._get_default_qconfig(self._global_quantizer_constraints[QuantizerGroup.WEIGHTS]) + weighted_node_and_qconf_lists = assign_qconfig_lists_to_modules( + nodes_with_weights=weight_nodes, + default_weight_qconfig=default_weight_qconfig, + global_weight_constraints=self._global_quantizer_constraints[QuantizerGroup.WEIGHTS], + scope_overrides_dict=None, + hw_config=hw_config, + ) + quantizable_layer_nodes = [ + QuantizableWeightedLayerNode(node, qconf_list) for node, qconf_list in weighted_node_and_qconf_lists.items() + ] + + scope_overrides = self._get_scope_overrides(inference_nncf_graph) + + ip_graph = InsertionPointGraph(inference_nncf_graph) + ip_graph = ip_graph.get_ip_graph_with_merged_hw_optimized_operations(hw_patterns) + post_processing_types = self._backend_entity.post_processing_metatypes + metatypes_to_ignore = self._backend_entity.get_ignored_metatypes(self._model_type, self._target_device) + solver = QuantizerPropagationSolver( + activation_ignored_scopes=ignored_names, + weight_ignored_scopes=list(ignored_names.keys()), + hw_config=hw_config, + default_trait_to_metatype_map=self._backend_entity.quant_trait_op_dict, + propagation_strategy=self.quantizer_propagation_rule, + default_qconfig_list=[ + self._get_default_qconfig(self._global_quantizer_constraints[QuantizerGroup.ACTIVATIONS]) + ], + quantizable_layer_nodes=quantizable_layer_nodes, + quantize_outputs=self._quantize_outputs, + global_constraints=self._global_quantizer_constraints, + post_processing_marker_metatypes=post_processing_types, + metatypes_to_ignore=metatypes_to_ignore, + scales_unification_map=self._backend_entity.scales_unification_map, + scope_overrides=scope_overrides, + ) + + quantization_proposal = solver.run_on_ip_graph(ip_graph, self._backend_entity.elementwise_metatypes) + multi_config_setup = quantization_proposal.quantizer_setup + single_config_setup = multi_config_setup.select_first_qconfig_for_each_point() + finalized_proposal = quantization_proposal.finalize(single_config_setup) + final_setup = solver.get_final_quantizer_setup(finalized_proposal) + return final_setup + + def get_quantization_setup(self, model: TModel, nncf_graph: NNCFGraph) -> SingleConfigQuantizerSetup: + """ + Initializes a cache, finds quantization target points and them puts in the cache. + + :param model: Backend-specific model, for which Quantization Target Points are being seek. + :param nncf_graph: NNCFGraph instance. + :return: Mapping of quantization target points with associated quantization configuration, + along with target points for scale unification. + """ + self._set_backend_entity(model) + backend = get_backend(model) + device = self._target_device + model_type = self._model_type + ignored_patterns = PatternsManager.get_full_ignored_pattern_graph( + backend=backend, device=device, model_type=model_type + ) + hw_patterns = PatternsManager.get_full_hw_pattern_graph(backend=backend, device=device, model_type=model_type) + + inference_nncf_graph = transform_to_inference_graph( + deepcopy(nncf_graph), + self._backend_entity.get_start_nodes_for_activation_path_tracing(nncf_graph), + self._backend_entity.shapeof_metatypes, + self._backend_entity.dropout_metatypes, + self._backend_entity.preserved_metatypes, + ) + + quantizer_setup = self._get_quantizer_setup(nncf_graph, inference_nncf_graph, hw_patterns, ignored_patterns) + self._apply_model_type_pass(self._model_type, quantizer_setup, nncf_graph) + self._apply_device_pass(self._target_device, quantizer_setup, inference_nncf_graph) + return quantizer_setup + + def _apply_model_type_pass( + self, model_type: Optional[ModelType], quantizer_setup: SingleConfigQuantizerSetup, nncf_graph: NNCFGraph + ) -> None: + """ + Applies changes in-place into quantizer setup based on model_type and device parameters. + + :param model_type: Model type parameter. + :param quantizer_setup: Quantizer setup which considered to update. + :param nncf_graph: Instance of NNCFGraph. + :return: None + """ + if model_type == ModelType.TRANSFORMER: + for quantization_point in quantizer_setup.quantization_points.values(): + if quantization_point.is_activation_quantization_point(): + for node_name in quantization_point.directly_quantized_operator_node_names: + node = nncf_graph.get_node_by_name(node_name) + if node.metatype not in self._backend_entity.mat_mul_metatypes: + continue + if ( + quantization_point.qconfig.mode != QuantizationScheme.SYMMETRIC + and not self._backend_entity.is_matmul_with_constant(node, nncf_graph) + ): + quantization_point.qconfig.mode = QuantizationScheme.SYMMETRIC + nncf_logger.debug( + f"Update quantization mode for the node {node_name}" + f" to the symmetric due to ModelType parameter." + ) + + def _apply_device_pass( + self, target_device: TargetDevice, quantizer_setup: SingleConfigQuantizerSetup, nncf_graph: NNCFGraph + ) -> None: + """ + This method applies model post-processing device passes to SingleConfigQuantizerSetup in-place. + + :param target_device: TargetDevice instance. + :param quantizer_setup: SingleConfigQuantizerSetup instance to update. + :param nncf_graph: NNCFGraph. + :return: None. + """ + + passes_map = {TargetDevice.CPU_SPR: self._apply_spr_pass} + + if target_device not in passes_map: + return + + passes_map[target_device](quantizer_setup, nncf_graph) + + def _apply_spr_pass( + self, quantizer_setup: SingleConfigQuantizerSetup, nncf_graph: NNCFGraph + ) -> SingleConfigQuantizerSetup: + """ + Applies CPU_SPR-related pass. + The main action is to remove one of the quantizers before elementwise layer (e.g. Add). + This action allows to get performance boost on SPR devices. + + :param quantizer_setup: SingleConfigQuantizerSetup instance to update. + :param nncf_graph: NNCFGraph instance to update. + :return: Modified SingleConfigQuantizerSetup. + """ + + def _is_node_after_producers(node): + input_node = node + while True: + input_node = nncf_graph.get_previous_nodes(input_node) + if len(input_node) > 1: + return False + input_node = input_node[0] + if input_node.metatype in producer_metatypes: + return True + + producer_metatypes = ( + self._backend_entity.conv_metatypes + + self._backend_entity.mat_mul_metatypes + + self._backend_entity.group_conv_metatypes + ) + + quantizer_setup_map = { + p.insertion_point.target_node_name: q_key for q_key, p in quantizer_setup.quantization_points.items() + } + + # Walking through all Add layers. + for add_node in nncf_graph.get_nodes_by_metatypes(self._backend_entity.add_metatypes): + add_inputs = nncf_graph.get_previous_nodes(add_node) + + # Filtering Add based on it's input. + # Need to find Add layer only with two activations as input. + if len(add_inputs) == 2 and all(n.node_name in quantizer_setup_map for n in add_inputs): + # Sorting of the inputs based on length of input's consumer in descending order. + add_inputs.sort(key=lambda n: len(nncf_graph.get_next_nodes(n)), reverse=True) + fq_1_producer, fq_2_producer = add_inputs + fq_1_q_key = quantizer_setup_map[fq_1_producer.node_name] + fq_2_q_key = quantizer_setup_map[fq_2_producer.node_name] + + # In the case of the two quantizers where one of them produces data into branching, + # it needs to remove the quantizer without branching after it. + if ( + len(nncf_graph.get_next_nodes(fq_1_producer)) > 1 + and len(nncf_graph.get_next_nodes(fq_2_producer)) == 1 + ): + quantizer_setup.discard(fq_2_q_key, True) + continue + + # In the case of the two quantizers without the branching after them, + # it needs to check that all quantizers follows after producer nodes. + if _is_node_after_producers(fq_1_producer) and _is_node_after_producers(fq_2_producer): + fq_1_prod_shape = np.prod(nncf_graph.get_output_edges_by_port_id(fq_1_producer, 0)[0].tensor_shape) + fq_2_prod_shape = np.prod(nncf_graph.get_output_edges_by_port_id(fq_2_producer, 0)[0].tensor_shape) + + # Then it needs to remove quantizer with the smallest shape. + if fq_1_prod_shape >= fq_2_prod_shape: + quantizer_setup.discard(fq_1_q_key, True) + else: + quantizer_setup.discard(fq_2_q_key, True) + + return quantizer_setup diff --git a/nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer/torch_fx_backend.py b/nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer/torch_fx_backend.py new file mode 100644 index 00000000000..4115bec611e --- /dev/null +++ b/nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer/torch_fx_backend.py @@ -0,0 +1,366 @@ +# Copyright (c) 2024 Intel Corporation +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# http://www.apache.org/licenses/LICENSE-2.0 +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Dict, List, Optional, Set, Tuple + +import torch +from torch.quantization.fake_quantize import FakeQuantize + +import nncf +import nncf.torch.graph.operator_metatypes as om +from nncf.common.graph.graph import NNCFGraph +from nncf.common.graph.graph import NNCFNode +from nncf.common.graph.operator_metatypes import OperatorMetatype +from nncf.common.graph.transformations.commands import TargetType +from nncf.common.graph.transformations.commands import TransformationCommand +from nncf.common.hardware.config import HWConfig +from nncf.common.quantization.structs import QuantizationScheme as QuantizationMode +from nncf.common.quantization.structs import QuantizerConfig +from nncf.experimental.common.tensor_statistics.collectors import AGGREGATORS_MAP +from nncf.experimental.common.tensor_statistics.collectors import TensorCollector +from nncf.experimental.common.tensor_statistics.statistics import MinMaxTensorStatistic +from nncf.experimental.torch.fx.commands import FXApplyTransformationCommand +from nncf.experimental.torch.fx.model_utils import get_target_point +from nncf.experimental.torch.fx.transformations import qdq_insertion_transformation_builder +from nncf.parameters import ModelType +from nncf.parameters import TargetDevice +from nncf.quantization.advanced_parameters import StatisticsType +from nncf.experimental.common.quantization.algorithms.quantizer.openvino_quantizer.backend import OpenVINOQuantizerBackend +from nncf.quantization.fake_quantize import FakeConvertParameters +from nncf.quantization.fake_quantize import FakeQuantizeParameters +from nncf.quantization.range_estimator import AggregatorType +from nncf.quantization.range_estimator import RangeEstimatorParameters +from nncf.torch.graph.graph import PTNNCFGraph +from nncf.torch.graph.graph import PTTargetPoint +from nncf.torch.graph.operator_metatypes import ELEMENTWISE_OPERATIONS +from nncf.torch.graph.transformations.commands import PTSharedFnInsertionCommand +from nncf.torch.hardware.config import PTHWConfig +from nncf.torch.model_graph_manager import get_weight_tensor_port_ids +from nncf.torch.nncf_network import NNCFNetwork +from nncf.torch.quantization.default_quantization import DEFAULT_PT_QUANT_TRAIT_TO_OP_DICT +from nncf.torch.quantization.layers import QUANTIZATION_MODULES +from nncf.torch.quantization.layers import AsymmetricQuantizer +from nncf.torch.quantization.layers import BaseQuantizer +from nncf.torch.quantization.layers import PTQuantizerSpec +from nncf.torch.quantization.layers import get_scale_shape +from nncf.torch.quantization.strip import convert_to_torch_fakequantizer +from nncf.torch.tensor_statistics.collectors import PT_REDUCERS_MAP + +FXMinMaxAlgoBackend = None + + +class FXNNCFQuantizerBackend(OpenVINOQuantizerBackend): + + @property + def preserved_metatypes(self) -> List[OperatorMetatype]: + return [] + + @property + def mat_mul_metatypes(self) -> List[OperatorMetatype]: + return [om.PTLinearMetatype, om.PTMatMulMetatype] + + @property + def post_processing_metatypes(self) -> List[OperatorMetatype]: + return [] + + @property + def shapeof_metatypes(self) -> List[OperatorMetatype]: + return [] + + @property + def dropout_metatypes(self) -> List[OperatorMetatype]: + return [om.PTDropoutMetatype] + + @property + def read_variable_metatypes(self) -> List[OperatorMetatype]: + return [] + + @property + def conv_metatypes(self) -> List[OperatorMetatype]: + return [om.PTConv1dMetatype, om.PTConv2dMetatype, om.PTConv3dMetatype] + + @property + def elementwise_metatypes(self) -> List[OperatorMetatype]: + return ELEMENTWISE_OPERATIONS + + @property + def overflow_fix_metatypes(self) -> List[OperatorMetatype]: + return [ + om.PTConv1dMetatype, + om.PTConv2dMetatype, + om.PTConv3dMetatype, + om.PTLinearMetatype, + om.PTConvTranspose1dMetatype, + om.PTConvTranspose2dMetatype, + om.PTConvTranspose3dMetatype, + ] + + @property + def add_metatypes(self) -> List[OperatorMetatype]: + return [om.PTAddMetatype] + + @property + def group_conv_metatypes(self) -> List[OperatorMetatype]: + return self.conv_metatypes + + @property + def scaled_dot_product_attention_metatypes(self) -> List[OperatorMetatype]: + return [om.PTScaledDotProductAttentionMetatype] + + @property + def scales_unification_map(self) -> Dict[OperatorMetatype, OperatorMetatype]: + return {om.PTCatMetatype: self.overflow_fix_metatypes} + + @property + def hw_config(self) -> HWConfig: + return PTHWConfig + + @property + def quant_trait_op_dict(self) -> Dict[int, OperatorMetatype]: + return DEFAULT_PT_QUANT_TRAIT_TO_OP_DICT + + @staticmethod + def get_start_nodes_for_activation_path_tracing(nncf_graph: PTNNCFGraph) -> List[NNCFNode]: + return nncf_graph.get_input_nodes() + + @staticmethod + def target_point(target_type: TargetType, target_node_name: str, port_id: int) -> PTTargetPoint: + return get_target_point(target_type, target_node_name, port_id) + + @staticmethod + def create_convert_insertion_command( + target_point: PTTargetPoint, + parameters: FakeConvertParameters, + ) -> TransformationCommand: + raise nncf.InternalError("FakeConvert insertion not implemented in PyTorch backend!") + + @staticmethod + def get_target_point_shape(nncf_graph: PTNNCFGraph, node: NNCFNode, target_point: PTTargetPoint) -> Tuple[int, ...]: + return nncf_graph.get_input_shape_for_insertion_point(target_point) + + @staticmethod + def get_weight_quantization_axes(node: NNCFNode, target_point: PTTargetPoint, ndims: int) -> Tuple[int]: + # TODO(dlyakhov): support transpose conv and other cases + return (0,) + + @staticmethod + def get_statistic_collector( + range_estimator_params: RangeEstimatorParameters, + use_abs_max: bool, + reduction_axes: Optional[Tuple[int, ...]], + aggregation_axes: Optional[Tuple[int, ...]], + inplace: bool, + num_samples: Optional[int] = None, + ) -> TensorCollector: + collector = TensorCollector(MinMaxTensorStatistic) + for params, container_key in zip( + [range_estimator_params.min, range_estimator_params.max], + [MinMaxTensorStatistic.MIN_STAT, MinMaxTensorStatistic.MAX_STAT], + ): + if params.statistics_type not in PT_REDUCERS_MAP: + raise nncf.InternalError( + f"Statistic type: {params.statistics_type} is not supported for Torch PTQ backend yet." + ) + + if params.aggregator_type not in AGGREGATORS_MAP: + raise nncf.InternalError( + f"Aggregator type: {params.aggregator_type} is not supported for Torch PTQ backend yet." + ) + + statistic_type = params.statistics_type + if statistic_type in [StatisticsType.QUANTILE, StatisticsType.ABS_QUANTILE]: + # TODO(dlyakhov): merge two quantile aggregators in one + if container_key == MinMaxTensorStatistic.MIN_STAT: + quantile = params.quantile_outlier_prob + else: + quantile = 1 - params.quantile_outlier_prob + reducer = PT_REDUCERS_MAP[statistic_type](reduction_axes=reduction_axes, quantile=[quantile]) + else: + if use_abs_max and statistic_type == StatisticsType.MAX: + statistic_type = StatisticsType.ABS_MAX + reducer = PT_REDUCERS_MAP[statistic_type](reduction_axes=reduction_axes) + + kwargs = { + "num_samples": num_samples, + "aggregation_axes": aggregation_axes, + } + if params.aggregator_type in [AggregatorType.MEAN_NO_OUTLIERS, AggregatorType.MEDIAN_NO_OUTLIERS]: + kwargs.update({"quantile": params.quantile_outlier_prob}) + aggregator = AGGREGATORS_MAP[params.aggregator_type](**kwargs) + + collector.register_statistic_branch(container_key, reducer, aggregator) + return collector + + @staticmethod + def get_weight_tensor_port_ids(node: NNCFNode, graph: NNCFGraph) -> List[Optional[int]]: + return get_weight_tensor_port_ids(node, graph) + + @staticmethod + def get_weight_name(nncf_graph: NNCFGraph, target_point: PTTargetPoint) -> str: + weighted_node = nncf_graph.get_node_by_name(target_point.target_node_name) + weight_edge = nncf_graph.get_input_edge_by_port_id(weighted_node, target_point.input_port_id) + weight = weight_edge.from_node + return weight.node_name + + @staticmethod + def should_quantize_weight(weight_name: str, quantized_weight_names: Set[str]) -> bool: + # If the nodes share one weight tensor, we should have only one quantizer on that + return weight_name not in quantized_weight_names + + @staticmethod + def get_weight_config(config: QuantizerConfig, model: NNCFNetwork) -> QuantizerConfig: + return config + + @staticmethod + def _get_input_scale_shape( + nncf_graph: NNCFGraph, target_point: PTTargetPoint, per_channel: bool + ) -> Tuple[Tuple[int, ...], Tuple[int, ...], int]: + is_weights = target_point.is_weight_target_point() + if is_weights: + # TODO(dlyakhov): support transpose conv/ make channel_idx common + channel_idx = 0 + else: + channel_idx = 1 # channel dim for activations + + input_shape = nncf_graph.get_input_shape_for_insertion_point(target_point) + scale_shape = tuple( + get_scale_shape(input_shape, is_weights=is_weights, per_channel=per_channel, channel_idx=channel_idx) + ) + + return input_shape, scale_shape, channel_idx + + @staticmethod + def _create_quantizer( + quantizer_config: QuantizerConfig, + scale_shape: Tuple, + parameters: FakeQuantizeParameters, + target_type: TargetType, + ) -> FakeQuantize: + mode = quantizer_config.mode + quantizer_cls = QUANTIZATION_MODULES.get(mode) + narrow_range = target_type == TargetType.OPERATION_WITH_WEIGHTS and mode == QuantizationMode.SYMMETRIC + quantizer_spec = PTQuantizerSpec.from_config( + quantizer_config, + narrow_range=narrow_range, + scale_shape=scale_shape, + half_range=False, + logarithm_scale=False, + is_quantized_on_export=False, + compression_lr_multiplier=None, + ) + quantizer = quantizer_cls(quantizer_spec) + + # Fill it with minmax + # TODO(dlyakhov) Prevent creation of intermediate objects like nncf quantizer. + FXMinMaxAlgoBackend._fill_quantizer_parameters(quantizer, parameters, quantizer_spec.scale_shape) + # Convert to the torch fake quantizer + torch_fq = convert_to_torch_fakequantizer(quantizer) + return torch_fq + + @staticmethod + def _fill_quantizer_parameters(quantizer: BaseQuantizer, parameters: FakeQuantizeParameters, scale_shape) -> None: + if isinstance(quantizer, AsymmetricQuantizer): + quantizer.input_low = torch.nn.Parameter(parameters.input_low.data.reshape(scale_shape)) + input_range = parameters.input_high - parameters.input_low + # Subtract eps from the input_range to make quantizer parameters equal to + # original parameters on the forward call. + quantizer.input_range = torch.nn.Parameter((input_range.data - quantizer.eps).reshape(scale_shape)) + else: + quantizer.signed = bool(torch.any(parameters.input_low.data < 0)) + # Subtract eps from the scale to make quantizer parameters equal to + # original parameters on the forward call. + quantizer.scale = torch.nn.Parameter((parameters.input_high.data - quantizer.eps).reshape(scale_shape)) + + @staticmethod + def create_quantizer_insertion_command( + nncf_graph: NNCFGraph, + target_point: PTTargetPoint, + quantizer_config: QuantizerConfig, + parameters: FakeQuantizeParameters, + ) -> FXApplyTransformationCommand: + _, scale_shape, _ = FXMinMaxAlgoBackend._get_input_scale_shape( + nncf_graph, target_point, quantizer_config.per_channel + ) + + quantizer = FXMinMaxAlgoBackend._create_quantizer( + quantizer_config, scale_shape, parameters, target_point.target_type + ) + transformation = qdq_insertion_transformation_builder(quantizer, [target_point]) + return FXApplyTransformationCommand(transformation) + + @staticmethod + def create_unified_scales_quantizers_insertion_commands( + nncf_graph: NNCFGraph, + target_points: List[PTTargetPoint], + quantizer_config: QuantizerConfig, + parameters: FakeQuantizeParameters, + ) -> List[PTSharedFnInsertionCommand]: + _, scale_shape, _ = FXMinMaxAlgoBackend._get_input_scale_shape( + nncf_graph, target_points[0], quantizer_config.per_channel + ) + + quantizer = FXMinMaxAlgoBackend._create_quantizer( + quantizer_config, scale_shape, parameters, target_points[0].target_type + ) + + transformations = [] + for tp in target_points: + transformation = qdq_insertion_transformation_builder(quantizer, [tp]) + transformations.append(FXApplyTransformationCommand(transformation)) + return transformations + + @staticmethod + def get_ignored_metatypes(model_type: ModelType, device: TargetDevice) -> List[OperatorMetatype]: + types = [] + if model_type == ModelType.TRANSFORMER: + types = [ + om.PTAddMetatype, + om.PTPowerMetatype, + om.PTSubMetatype, + om.PTAvgPool2dMetatype, + om.PTAvgPool3dMetatype, + om.PTMeanMetatype, + om.PTSumMetatype, + om.PTReduceL2, + om.PTDivMetatype, + om.PTMaxMetatype, + om.PTSqueezeMetatype, + om.PTLayerNormMetatype, + om.PTModuleLayerNormMetatype, + om.PTGroupNormMetatype, + om.PTModuleGroupNormMetatype, + # Batchnorm + om.PTBatchNormMetatype, + om.PTModuleBatchNormMetatype, + ] + if device != TargetDevice.CPU_SPR: + types.append(om.PTMulMetatype) + return types + + @staticmethod + def get_ignored_names_by_layer_attributes(nncf_graph: NNCFGraph) -> Set[str]: + return set() + + def get_weight_nodes(self, nncf_graph: NNCFGraph) -> List[NNCFNode]: + weight_nodes_candidates = [ + node + for node in nncf_graph.get_all_nodes() + if issubclass(node.metatype, om.PTOperatorMetatype) and node.metatype.weight_port_ids + ] + weight_nodes = [] + for node in weight_nodes_candidates: + if node.metatype in self.mat_mul_metatypes and not self.is_matmul_with_constant(node, nncf_graph): + continue + weight_nodes.append(node) + return weight_nodes + + def is_matmul_with_constant(self, node: NNCFNode, nncf_graph: NNCFGraph) -> bool: + return node.metatype in self.mat_mul_metatypes and len(get_weight_tensor_port_ids(node, nncf_graph)) > 0 diff --git a/nncf/quantization/algorithms/min_max/quantizer/backend.py b/nncf/quantization/algorithms/min_max/quantizer/backend.py new file mode 100644 index 00000000000..821d57a82ea --- /dev/null +++ b/nncf/quantization/algorithms/min_max/quantizer/backend.py @@ -0,0 +1,323 @@ +# Copyright (c) 2024 Intel Corporation +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# http://www.apache.org/licenses/LICENSE-2.0 +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from abc import ABC +from abc import abstractmethod +from typing import Dict, List, Optional, Set, Tuple, TypeVar + +from nncf.common.graph.graph import NNCFGraph +from nncf.common.graph.graph import NNCFNode +from nncf.common.graph.operator_metatypes import OperatorMetatype +from nncf.common.graph.transformations.commands import TargetPoint +from nncf.common.graph.transformations.commands import TargetType +from nncf.common.graph.transformations.commands import TransformationCommand +from nncf.common.hardware.config import HWConfig +from nncf.common.quantization.structs import QuantizerConfig +from nncf.common.tensor_statistics.collectors import TensorStatisticCollectorBase +from nncf.parameters import ModelType +from nncf.parameters import TargetDevice +from nncf.quantization.fake_quantize import FakeConvertParameters +from nncf.quantization.fake_quantize import FakeQuantizeParameters +from nncf.quantization.range_estimator import RangeEstimatorParameters + +TModel = TypeVar("TModel") + + +class NNCFQuantizerBackend(ABC): + @property + @abstractmethod + def preserved_metatypes(self) -> List[OperatorMetatype]: + """ + Property for backend-specific metatypes that require preserving float subgraphs + when removing the ShapeOf subgraph. + """ + + @property + @abstractmethod + def mat_mul_metatypes(self) -> List[OperatorMetatype]: + """ + Property for the backend-specific MatMul metatypes. + """ + + @property + @abstractmethod + def post_processing_metatypes(self) -> List[OperatorMetatype]: + """ + Property for the backend-specific post-processing metatypes (NonMaximumSupression, TopK, etc.). + """ + + @property + @abstractmethod + def conv_metatypes(self) -> List[OperatorMetatype]: + """ + Property for the backend-specific Convolution metatypes. + """ + + @property + @abstractmethod + def shapeof_metatypes(self) -> List[OperatorMetatype]: + """ + Property for the backend-specific ShapeOf metatypes. + """ + + @property + @abstractmethod + def dropout_metatypes(self) -> List[OperatorMetatype]: + """ + Property for the backend-specific Dropout metatypes. + """ + + @property + @abstractmethod + def elementwise_metatypes(self) -> List[OperatorMetatype]: + """ + Property for the backend-specific Elementwises metatypes. + """ + + @property + @abstractmethod + def overflow_fix_metatypes(self) -> List[OperatorMetatype]: + """ + Property for the backend-specific metatypes for which overflow_fix is applicable. + """ + + @property + @abstractmethod + def add_metatypes(self) -> List[OperatorMetatype]: + """ + Property for the backend-specific metatypes that also can be interpreted as Add layer. + """ + + @property + @abstractmethod + def group_conv_metatypes(self) -> List[OperatorMetatype]: + """ + Property for the backend-specific Grouped Convolution metatypes. + """ + + @property + @abstractmethod + def scaled_dot_product_attention_metatypes(self) -> List[OperatorMetatype]: + """ + Property for the backend-specific Scaled Dot Product Attention metatypes. + """ + + @property + @abstractmethod + def scales_unification_map(self) -> Dict[OperatorMetatype, OperatorMetatype]: + """ + Property for the backend-specific metatypes that produces quantizers that might be unified. + """ + + @property + @abstractmethod + def hw_config(self) -> HWConfig: + """ + Property for the hardware backend-specific configuration. + """ + + @property + @abstractmethod + def quant_trait_op_dict(self) -> Dict[int, OperatorMetatype]: + """ + Property for the backend-specific dictionary that contains QuantizationTrait-specific metatypes. + """ + + @staticmethod + @abstractmethod + def target_point(target_type: TargetType, target_node_name: str, port_id: int) -> TargetPoint: + """ + Returns backend-specific target point. + + :param target_type: Type of the location that should be modified. + :param target_node_name: Name of the located node. + :param port_id: Port ID of the tensor for the statistics distribution. + :return: Backend-specific TargetPoint. + """ + + @staticmethod + @abstractmethod + def create_quantizer_insertion_command( + nncf_graph: NNCFGraph, + target_point: TargetPoint, + quantizer_config: QuantizerConfig, + parameters: FakeQuantizeParameters, + ) -> TransformationCommand: + """ + Returns backend-specific quantizer insertion command. + + :param nncf_graph: NNCFGraph to get input/output shapes for the target point. + :param target_point: Target location for the quantizer insertion. + :param quantizer_config: QuantizerConfig instance for the current layer. + :param parameters: FakeQuantizeParameters to calculate activation quantization parameters. + :return: Backend-specific TransformationCommand for the quantizer insertion operation. + """ + + @staticmethod + @abstractmethod + def create_unified_scales_quantizers_insertion_commands( + nncf_graph: NNCFGraph, + target_points: List[TargetPoint], + quantizer_config: QuantizerConfig, + parameters: FakeQuantizeParameters, + ) -> List[TransformationCommand]: + """ + Returns backend-specific unified scales quantizers insertion commands. + + :param nncf_graph: NNCFGraph to get input/output shapes for the target point. + :param target_points: List of target locations for the quantizers insertion. + :param quantizer_config: QuantizerConfig instance for the current layer. + :param parameters: FakeQuantizeParameters to calculate activation quantization parameters. + :return: List of backend-specific TransformationCommands + for the quantizers with unified scales insertion operations. + """ + + @staticmethod + @abstractmethod + def create_convert_insertion_command( + target_point: TargetPoint, + parameters: FakeConvertParameters, + ) -> TransformationCommand: + """ + Returns backend-specific convert insertion command. + + :param target_point: Target location for the correction. + :param parameters: FakeConvertParameters to calculate activation quantization parameters. + :return: Backend-specific TransformationCommand for the quantizer insertion operation. + """ + + @staticmethod + @abstractmethod + def get_start_nodes_for_activation_path_tracing(nncf_graph: NNCFGraph) -> List[NNCFNode]: + """ + Returns a list of NNCFNodes to use as start nodes for activation path tracing. + + :param nncf_graph: NNCFGraph to get the start nodes. + :return: List of NNCFNodes to use as start nodes for activation path tracing. + + """ + + @staticmethod + @abstractmethod + def get_target_point_shape(nncf_graph: NNCFGraph, node: NNCFNode, target_point: TargetPoint) -> Tuple[int, ...]: + """ + Returns shape of a target point tensor. + + :param nncf_graph: NNCFGraph instance. + :param node: NNCFNode. + :param target_point: Target point of which tensor shape is seeked. + :return: Shape of target point tensor. + """ + + @staticmethod + @abstractmethod + def get_weight_quantization_axes(node: NNCFNode, target_point: TargetPoint, ndims: int) -> Tuple[int, ...]: + """ + Returns axes for per-channel quantization of weights of the node placed on a input port_id. + + :param node: Quantized node with the weight. + :param target_point: Corresponding target point. + :param ndims: Number of dimensions of weight. + :return: Axes for per-channel quantization of weights. + """ + + @staticmethod + @abstractmethod + def get_statistic_collector( + range_estimator_params: RangeEstimatorParameters, + use_abs_max: bool, + reduction_axes: Optional[Tuple[int, ...]], + aggregation_axes: Optional[Tuple[int, ...]], + inplace: bool, + num_samples: Optional[int] = None, + ) -> TensorStatisticCollectorBase: + """ + Returns backend-specific statistic collector. + + :param range_estimator_params: Parameters that specify estimators types. + :param use_abs_max: Wheather reduce absolute values of input tensors or not. + :param reduction_axes: Axes for reducer. + :param aggregation_axes: Axes for aggregator. + :param inplace: Whether to calculate statistic inplace or not. + :param num_samples: Maximum number of samples to collect. + :return: Backend-specific TensorStatisticCollectorBase for the statistics calculation. + """ + + @staticmethod + @abstractmethod + def get_weight_tensor_port_ids(node: NNCFNode, graph: NNCFGraph) -> List[Optional[int]]: + """ + Returns node's input port indices with weight tensors. + + :param node: NNCFNode to find its weight input port indices. + :param graph: NNCFGraph instance. + :return: Weights input port indices. + """ + + @staticmethod + def get_weight_name(nncf_graph: NNCFGraph, target_point: TargetPoint) -> str: + """ + Returns node's weight name corresponding to port ID. + + :param nncf_graph: NNCFGraph instance. + :param target_point: The TargetPoint instance that contains layer's information. + :return: Weight name. + """ + + @staticmethod + def should_quantize_weight(weight_name: str, quantized_weight_names: Set[str]) -> bool: + """ + Return True if weight should be quantized. + + :param weight_name: Weight name. + :param quantized_weight_names: Set containing already quantized weight names. + :return: A boolean value specifying whether a weight should be quantized. + """ + + @staticmethod + @abstractmethod + def get_ignored_metatypes(model_type: ModelType, device: TargetDevice) -> List[OperatorMetatype]: + """ + Returns ignored metatypes based on a model type and device parameters. + + :param model_type: Model type parameter. + :param device: Target device. + :return: List of ignored metatypes. + """ + + @staticmethod + @abstractmethod + def get_ignored_names_by_layer_attributes(nncf_graph: NNCFGraph) -> Set[str]: + """ + Returns names of ignored nodes based on layer_attributes. + + :param nncf_graph: NNCFGraph instance. + :return: List of ignored names. + """ + + @abstractmethod + def get_weight_nodes(self, nncf_graph: NNCFGraph) -> List[NNCFNode]: + """ + Returns nodes that have weights. + + :param nncf_graph: Instance of NNCFGraph. + :return: All nodes with weights. + """ + + @abstractmethod + def is_matmul_with_constant(self, node: NNCFNode, nncf_graph: NNCFGraph) -> bool: + """ + Returns true if given nncf matmul node is a matmul with a constant, False otherwise. + + :param Node: Instance of NNCFNode. + :param nncf_graph: Instance of NNCFGraph. + :return: True if given nncf matmul node is a matmul with a constant, False otherwise. + """ diff --git a/nncf/quantization/algorithms/min_max/quantizer/torch_fx_backend.py b/nncf/quantization/algorithms/min_max/quantizer/torch_fx_backend.py new file mode 100644 index 00000000000..9ed869b543c --- /dev/null +++ b/nncf/quantization/algorithms/min_max/quantizer/torch_fx_backend.py @@ -0,0 +1,366 @@ +# Copyright (c) 2024 Intel Corporation +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# http://www.apache.org/licenses/LICENSE-2.0 +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Dict, List, Optional, Set, Tuple + +import torch +from torch.quantization.fake_quantize import FakeQuantize + +import nncf +import nncf.torch.graph.operator_metatypes as om +from nncf.common.graph.graph import NNCFGraph +from nncf.common.graph.graph import NNCFNode +from nncf.common.graph.operator_metatypes import OperatorMetatype +from nncf.common.graph.transformations.commands import TargetType +from nncf.common.graph.transformations.commands import TransformationCommand +from nncf.common.hardware.config import HWConfig +from nncf.common.quantization.structs import QuantizationScheme as QuantizationMode +from nncf.common.quantization.structs import QuantizerConfig +from nncf.experimental.common.tensor_statistics.collectors import AGGREGATORS_MAP +from nncf.experimental.common.tensor_statistics.collectors import TensorCollector +from nncf.experimental.common.tensor_statistics.statistics import MinMaxTensorStatistic +from nncf.experimental.torch.fx.commands import FXApplyTransformationCommand +from nncf.experimental.torch.fx.model_utils import get_target_point +from nncf.experimental.torch.fx.transformations import qdq_insertion_transformation_builder +from nncf.parameters import ModelType +from nncf.parameters import TargetDevice +from nncf.quantization.advanced_parameters import StatisticsType +from nncf.quantization.algorithms.min_max.quantizer.backend import NNCFQuantizerBackend +from nncf.quantization.fake_quantize import FakeConvertParameters +from nncf.quantization.fake_quantize import FakeQuantizeParameters +from nncf.quantization.range_estimator import AggregatorType +from nncf.quantization.range_estimator import RangeEstimatorParameters +from nncf.torch.graph.graph import PTNNCFGraph +from nncf.torch.graph.graph import PTTargetPoint +from nncf.torch.graph.operator_metatypes import ELEMENTWISE_OPERATIONS +from nncf.torch.graph.transformations.commands import PTSharedFnInsertionCommand +from nncf.torch.hardware.config import PTHWConfig +from nncf.torch.model_graph_manager import get_weight_tensor_port_ids +from nncf.torch.nncf_network import NNCFNetwork +from nncf.torch.quantization.default_quantization import DEFAULT_PT_QUANT_TRAIT_TO_OP_DICT +from nncf.torch.quantization.layers import QUANTIZATION_MODULES +from nncf.torch.quantization.layers import AsymmetricQuantizer +from nncf.torch.quantization.layers import BaseQuantizer +from nncf.torch.quantization.layers import PTQuantizerSpec +from nncf.torch.quantization.layers import get_scale_shape +from nncf.torch.quantization.strip import convert_to_torch_fakequantizer +from nncf.torch.tensor_statistics.collectors import PT_REDUCERS_MAP + +FXMinMaxAlgoBackend = None + + +class FXNNCFQuantizerBackend(NNCFQuantizerBackend): + + @property + def preserved_metatypes(self) -> List[OperatorMetatype]: + return [] + + @property + def mat_mul_metatypes(self) -> List[OperatorMetatype]: + return [om.PTLinearMetatype, om.PTMatMulMetatype] + + @property + def post_processing_metatypes(self) -> List[OperatorMetatype]: + return [] + + @property + def shapeof_metatypes(self) -> List[OperatorMetatype]: + return [] + + @property + def dropout_metatypes(self) -> List[OperatorMetatype]: + return [om.PTDropoutMetatype] + + @property + def read_variable_metatypes(self) -> List[OperatorMetatype]: + return [] + + @property + def conv_metatypes(self) -> List[OperatorMetatype]: + return [om.PTConv1dMetatype, om.PTConv2dMetatype, om.PTConv3dMetatype] + + @property + def elementwise_metatypes(self) -> List[OperatorMetatype]: + return ELEMENTWISE_OPERATIONS + + @property + def overflow_fix_metatypes(self) -> List[OperatorMetatype]: + return [ + om.PTConv1dMetatype, + om.PTConv2dMetatype, + om.PTConv3dMetatype, + om.PTLinearMetatype, + om.PTConvTranspose1dMetatype, + om.PTConvTranspose2dMetatype, + om.PTConvTranspose3dMetatype, + ] + + @property + def add_metatypes(self) -> List[OperatorMetatype]: + return [om.PTAddMetatype] + + @property + def group_conv_metatypes(self) -> List[OperatorMetatype]: + return self.conv_metatypes + + @property + def scaled_dot_product_attention_metatypes(self) -> List[OperatorMetatype]: + return [om.PTScaledDotProductAttentionMetatype] + + @property + def scales_unification_map(self) -> Dict[OperatorMetatype, OperatorMetatype]: + return {om.PTCatMetatype: self.overflow_fix_metatypes} + + @property + def hw_config(self) -> HWConfig: + return PTHWConfig + + @property + def quant_trait_op_dict(self) -> Dict[int, OperatorMetatype]: + return DEFAULT_PT_QUANT_TRAIT_TO_OP_DICT + + @staticmethod + def get_start_nodes_for_activation_path_tracing(nncf_graph: PTNNCFGraph) -> List[NNCFNode]: + return nncf_graph.get_input_nodes() + + @staticmethod + def target_point(target_type: TargetType, target_node_name: str, port_id: int) -> PTTargetPoint: + return get_target_point(target_type, target_node_name, port_id) + + @staticmethod + def create_convert_insertion_command( + target_point: PTTargetPoint, + parameters: FakeConvertParameters, + ) -> TransformationCommand: + raise nncf.InternalError("FakeConvert insertion not implemented in PyTorch backend!") + + @staticmethod + def get_target_point_shape(nncf_graph: PTNNCFGraph, node: NNCFNode, target_point: PTTargetPoint) -> Tuple[int, ...]: + return nncf_graph.get_input_shape_for_insertion_point(target_point) + + @staticmethod + def get_weight_quantization_axes(node: NNCFNode, target_point: PTTargetPoint, ndims: int) -> Tuple[int]: + # TODO(dlyakhov): support transpose conv and other cases + return (0,) + + @staticmethod + def get_statistic_collector( + range_estimator_params: RangeEstimatorParameters, + use_abs_max: bool, + reduction_axes: Optional[Tuple[int, ...]], + aggregation_axes: Optional[Tuple[int, ...]], + inplace: bool, + num_samples: Optional[int] = None, + ) -> TensorCollector: + collector = TensorCollector(MinMaxTensorStatistic) + for params, container_key in zip( + [range_estimator_params.min, range_estimator_params.max], + [MinMaxTensorStatistic.MIN_STAT, MinMaxTensorStatistic.MAX_STAT], + ): + if params.statistics_type not in PT_REDUCERS_MAP: + raise nncf.InternalError( + f"Statistic type: {params.statistics_type} is not supported for Torch PTQ backend yet." + ) + + if params.aggregator_type not in AGGREGATORS_MAP: + raise nncf.InternalError( + f"Aggregator type: {params.aggregator_type} is not supported for Torch PTQ backend yet." + ) + + statistic_type = params.statistics_type + if statistic_type in [StatisticsType.QUANTILE, StatisticsType.ABS_QUANTILE]: + # TODO(dlyakhov): merge two quantile aggregators in one + if container_key == MinMaxTensorStatistic.MIN_STAT: + quantile = params.quantile_outlier_prob + else: + quantile = 1 - params.quantile_outlier_prob + reducer = PT_REDUCERS_MAP[statistic_type](reduction_axes=reduction_axes, quantile=[quantile]) + else: + if use_abs_max and statistic_type == StatisticsType.MAX: + statistic_type = StatisticsType.ABS_MAX + reducer = PT_REDUCERS_MAP[statistic_type](reduction_axes=reduction_axes) + + kwargs = { + "num_samples": num_samples, + "aggregation_axes": aggregation_axes, + } + if params.aggregator_type in [AggregatorType.MEAN_NO_OUTLIERS, AggregatorType.MEDIAN_NO_OUTLIERS]: + kwargs.update({"quantile": params.quantile_outlier_prob}) + aggregator = AGGREGATORS_MAP[params.aggregator_type](**kwargs) + + collector.register_statistic_branch(container_key, reducer, aggregator) + return collector + + @staticmethod + def get_weight_tensor_port_ids(node: NNCFNode, graph: NNCFGraph) -> List[Optional[int]]: + return get_weight_tensor_port_ids(node, graph) + + @staticmethod + def get_weight_name(nncf_graph: NNCFGraph, target_point: PTTargetPoint) -> str: + weighted_node = nncf_graph.get_node_by_name(target_point.target_node_name) + weight_edge = nncf_graph.get_input_edge_by_port_id(weighted_node, target_point.input_port_id) + weight = weight_edge.from_node + return weight.node_name + + @staticmethod + def should_quantize_weight(weight_name: str, quantized_weight_names: Set[str]) -> bool: + # If the nodes share one weight tensor, we should have only one quantizer on that + return weight_name not in quantized_weight_names + + @staticmethod + def get_weight_config(config: QuantizerConfig, model: NNCFNetwork) -> QuantizerConfig: + return config + + @staticmethod + def _get_input_scale_shape( + nncf_graph: NNCFGraph, target_point: PTTargetPoint, per_channel: bool + ) -> Tuple[Tuple[int, ...], Tuple[int, ...], int]: + is_weights = target_point.is_weight_target_point() + if is_weights: + # TODO(dlyakhov): support transpose conv/ make channel_idx common + channel_idx = 0 + else: + channel_idx = 1 # channel dim for activations + + input_shape = nncf_graph.get_input_shape_for_insertion_point(target_point) + scale_shape = tuple( + get_scale_shape(input_shape, is_weights=is_weights, per_channel=per_channel, channel_idx=channel_idx) + ) + + return input_shape, scale_shape, channel_idx + + @staticmethod + def _create_quantizer( + quantizer_config: QuantizerConfig, + scale_shape: Tuple, + parameters: FakeQuantizeParameters, + target_type: TargetType, + ) -> FakeQuantize: + mode = quantizer_config.mode + quantizer_cls = QUANTIZATION_MODULES.get(mode) + narrow_range = target_type == TargetType.OPERATION_WITH_WEIGHTS and mode == QuantizationMode.SYMMETRIC + quantizer_spec = PTQuantizerSpec.from_config( + quantizer_config, + narrow_range=narrow_range, + scale_shape=scale_shape, + half_range=False, + logarithm_scale=False, + is_quantized_on_export=False, + compression_lr_multiplier=None, + ) + quantizer = quantizer_cls(quantizer_spec) + + # Fill it with minmax + # TODO(dlyakhov) Prevent creation of intermediate objects like nncf quantizer. + FXMinMaxAlgoBackend._fill_quantizer_parameters(quantizer, parameters, quantizer_spec.scale_shape) + # Convert to the torch fake quantizer + torch_fq = convert_to_torch_fakequantizer(quantizer) + return torch_fq + + @staticmethod + def _fill_quantizer_parameters(quantizer: BaseQuantizer, parameters: FakeQuantizeParameters, scale_shape) -> None: + if isinstance(quantizer, AsymmetricQuantizer): + quantizer.input_low = torch.nn.Parameter(parameters.input_low.data.reshape(scale_shape)) + input_range = parameters.input_high - parameters.input_low + # Subtract eps from the input_range to make quantizer parameters equal to + # original parameters on the forward call. + quantizer.input_range = torch.nn.Parameter((input_range.data - quantizer.eps).reshape(scale_shape)) + else: + quantizer.signed = bool(torch.any(parameters.input_low.data < 0)) + # Subtract eps from the scale to make quantizer parameters equal to + # original parameters on the forward call. + quantizer.scale = torch.nn.Parameter((parameters.input_high.data - quantizer.eps).reshape(scale_shape)) + + @staticmethod + def create_quantizer_insertion_command( + nncf_graph: NNCFGraph, + target_point: PTTargetPoint, + quantizer_config: QuantizerConfig, + parameters: FakeQuantizeParameters, + ) -> FXApplyTransformationCommand: + _, scale_shape, _ = FXMinMaxAlgoBackend._get_input_scale_shape( + nncf_graph, target_point, quantizer_config.per_channel + ) + + quantizer = FXMinMaxAlgoBackend._create_quantizer( + quantizer_config, scale_shape, parameters, target_point.target_type + ) + transformation = qdq_insertion_transformation_builder(quantizer, [target_point]) + return FXApplyTransformationCommand(transformation) + + @staticmethod + def create_unified_scales_quantizers_insertion_commands( + nncf_graph: NNCFGraph, + target_points: List[PTTargetPoint], + quantizer_config: QuantizerConfig, + parameters: FakeQuantizeParameters, + ) -> List[PTSharedFnInsertionCommand]: + _, scale_shape, _ = FXMinMaxAlgoBackend._get_input_scale_shape( + nncf_graph, target_points[0], quantizer_config.per_channel + ) + + quantizer = FXMinMaxAlgoBackend._create_quantizer( + quantizer_config, scale_shape, parameters, target_points[0].target_type + ) + + transformations = [] + for tp in target_points: + transformation = qdq_insertion_transformation_builder(quantizer, [tp]) + transformations.append(FXApplyTransformationCommand(transformation)) + return transformations + + @staticmethod + def get_ignored_metatypes(model_type: ModelType, device: TargetDevice) -> List[OperatorMetatype]: + types = [] + if model_type == ModelType.TRANSFORMER: + types = [ + om.PTAddMetatype, + om.PTPowerMetatype, + om.PTSubMetatype, + om.PTAvgPool2dMetatype, + om.PTAvgPool3dMetatype, + om.PTMeanMetatype, + om.PTSumMetatype, + om.PTReduceL2, + om.PTDivMetatype, + om.PTMaxMetatype, + om.PTSqueezeMetatype, + om.PTLayerNormMetatype, + om.PTModuleLayerNormMetatype, + om.PTGroupNormMetatype, + om.PTModuleGroupNormMetatype, + # Batchnorm + om.PTBatchNormMetatype, + om.PTModuleBatchNormMetatype, + ] + if device != TargetDevice.CPU_SPR: + types.append(om.PTMulMetatype) + return types + + @staticmethod + def get_ignored_names_by_layer_attributes(nncf_graph: NNCFGraph) -> Set[str]: + return set() + + def get_weight_nodes(self, nncf_graph: NNCFGraph) -> List[NNCFNode]: + weight_nodes_candidates = [ + node + for node in nncf_graph.get_all_nodes() + if issubclass(node.metatype, om.PTOperatorMetatype) and node.metatype.weight_port_ids + ] + weight_nodes = [] + for node in weight_nodes_candidates: + if node.metatype in self.mat_mul_metatypes and not self.is_matmul_with_constant(node, nncf_graph): + continue + weight_nodes.append(node) + return weight_nodes + + def is_matmul_with_constant(self, node: NNCFNode, nncf_graph: NNCFGraph) -> bool: + return node.metatype in self.mat_mul_metatypes and len(get_weight_tensor_port_ids(node, nncf_graph)) > 0 diff --git a/tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/mobilenet_v3_small.dot b/tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/mobilenet_v3_small.dot new file mode 100644 index 00000000000..ca81831bd41 --- /dev/null +++ b/tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/mobilenet_v3_small.dot @@ -0,0 +1,1042 @@ +strict digraph { +"0 features_1_block_1_fc1_weight" [id=0, type=get_attr]; +"1 features_1_block_1_fc1_bias" [id=1, type=get_attr]; +"2 features_1_block_1_fc2_weight" [id=2, type=get_attr]; +"3 features_1_block_1_fc2_bias" [id=3, type=get_attr]; +"4 features_4_block_2_fc1_weight" [id=4, type=get_attr]; +"5 features_4_block_2_fc1_bias" [id=5, type=get_attr]; +"6 features_4_block_2_fc2_weight" [id=6, type=get_attr]; +"7 features_4_block_2_fc2_bias" [id=7, type=get_attr]; +"8 features_5_block_2_fc1_weight" [id=8, type=get_attr]; +"9 features_5_block_2_fc1_bias" [id=9, type=get_attr]; +"10 features_5_block_2_fc2_weight" [id=10, type=get_attr]; +"11 features_5_block_2_fc2_bias" [id=11, type=get_attr]; +"12 features_6_block_2_fc1_weight" [id=12, type=get_attr]; +"13 features_6_block_2_fc1_bias" [id=13, type=get_attr]; +"14 features_6_block_2_fc2_weight" [id=14, type=get_attr]; +"15 features_6_block_2_fc2_bias" [id=15, type=get_attr]; +"16 features_7_block_2_fc1_weight" [id=16, type=get_attr]; +"17 features_7_block_2_fc1_bias" [id=17, type=get_attr]; +"18 features_7_block_2_fc2_weight" [id=18, type=get_attr]; +"19 features_7_block_2_fc2_bias" [id=19, type=get_attr]; +"20 features_8_block_2_fc1_weight" [id=20, type=get_attr]; +"21 features_8_block_2_fc1_bias" [id=21, type=get_attr]; +"22 features_8_block_2_fc2_weight" [id=22, type=get_attr]; +"23 features_8_block_2_fc2_bias" [id=23, type=get_attr]; +"24 features_9_block_2_fc1_weight" [id=24, type=get_attr]; +"25 features_9_block_2_fc1_bias" [id=25, type=get_attr]; +"26 features_9_block_2_fc2_weight" [id=26, type=get_attr]; +"27 features_9_block_2_fc2_bias" [id=27, type=get_attr]; +"28 features_10_block_2_fc1_weight" [id=28, type=get_attr]; +"29 features_10_block_2_fc1_bias" [id=29, type=get_attr]; +"30 features_10_block_2_fc2_weight" [id=30, type=get_attr]; +"31 features_10_block_2_fc2_bias" [id=31, type=get_attr]; +"32 features_11_block_2_fc1_weight" [id=32, type=get_attr]; +"33 features_11_block_2_fc1_bias" [id=33, type=get_attr]; +"34 features_11_block_2_fc2_weight" [id=34, type=get_attr]; +"35 features_11_block_2_fc2_bias" [id=35, type=get_attr]; +"36 classifier_0_bias" [id=36, type=get_attr]; +"37 classifier_3_bias" [id=37, type=get_attr]; +"38 x" [id=38, type=input]; +"39 quantize_per_tensor_default" [id=39, type=quantize_per_tensor]; +"40 dequantize_per_tensor_default" [id=40, type=dequantize_per_tensor]; +"41 features_0_0_weight_bias" [id=41, type=get_attr]; +"42 conv2d_scale_0" [id=42, type=get_attr]; +"43 conv2d_zero_point_0" [id=43, type=get_attr]; +"44 _frozen_param0" [id=44, type=get_attr]; +"45 dequantize_per_channel_default" [id=45, type=dequantize_per_channel]; +"46 conv2d" [id=46, type=conv2d]; +"47 hardswish__scale_0" [id=47, type=get_attr]; +"48 hardswish__zero_point_0" [id=48, type=get_attr]; +"49 hardswish_" [id=49, type=hardswish_]; +"50 quantize_per_channel_default_1" [id=50, type=quantize_per_channel]; +"51 dequantize_per_channel_default_1" [id=51, type=dequantize_per_channel]; +"52 features_1_block_0_0_weight_bias" [id=52, type=get_attr]; +"53 conv2d_1_scale_0" [id=53, type=get_attr]; +"54 conv2d_1_zero_point_0" [id=54, type=get_attr]; +"55 _frozen_param1" [id=55, type=get_attr]; +"56 dequantize_per_channel_default_2" [id=56, type=dequantize_per_channel]; +"57 conv2d_1" [id=57, type=conv2d]; +"58 relu_" [id=58, type=relu_]; +"59 quantize_per_tensor_default_1" [id=59, type=quantize_per_tensor]; +"60 dequantize_per_tensor_default_2" [id=60, type=dequantize_per_tensor]; +"61 dequantize_per_tensor_default_1" [id=61, type=dequantize_per_tensor]; +"62 adaptive_avg_pool2d" [id=62, type=adaptive_avg_pool2d]; +"63 conv2d_2" [id=63, type=conv2d]; +"64 relu" [id=64, type=relu]; +"65 conv2d_3" [id=65, type=conv2d]; +"66 hardsigmoid" [id=66, type=hardsigmoid]; +"67 quantize_per_tensor_default_2" [id=67, type=quantize_per_tensor]; +"68 dequantize_per_tensor_default_3" [id=68, type=dequantize_per_tensor]; +"69 mul" [id=69, type=mul]; +"70 quantize_per_tensor_default_3" [id=70, type=quantize_per_tensor]; +"71 dequantize_per_tensor_default_4" [id=71, type=dequantize_per_tensor]; +"72 features_1_block_2_0_weight_bias" [id=72, type=get_attr]; +"73 conv2d_4_scale_0" [id=73, type=get_attr]; +"74 conv2d_4_zero_point_0" [id=74, type=get_attr]; +"75 _frozen_param2" [id=75, type=get_attr]; +"76 dequantize_per_channel_default_3" [id=76, type=dequantize_per_channel]; +"77 conv2d_4" [id=77, type=conv2d]; +"78 quantize_per_tensor_default_4" [id=78, type=quantize_per_tensor]; +"79 dequantize_per_tensor_default_5" [id=79, type=dequantize_per_tensor]; +"80 features_2_block_0_0_weight_bias" [id=80, type=get_attr]; +"81 conv2d_5_scale_0" [id=81, type=get_attr]; +"82 conv2d_5_zero_point_0" [id=82, type=get_attr]; +"83 _frozen_param3" [id=83, type=get_attr]; +"84 dequantize_per_channel_default_4" [id=84, type=dequantize_per_channel]; +"85 conv2d_5" [id=85, type=conv2d]; +"86 relu__1_scale_0" [id=86, type=get_attr]; +"87 relu__1_zero_point_0" [id=87, type=get_attr]; +"88 relu__1" [id=88, type=relu_]; +"89 quantize_per_channel_default_5" [id=89, type=quantize_per_channel]; +"90 dequantize_per_channel_default_5" [id=90, type=dequantize_per_channel]; +"91 features_2_block_1_0_weight_bias" [id=91, type=get_attr]; +"92 conv2d_6_scale_0" [id=92, type=get_attr]; +"93 conv2d_6_zero_point_0" [id=93, type=get_attr]; +"94 _frozen_param4" [id=94, type=get_attr]; +"95 dequantize_per_channel_default_6" [id=95, type=dequantize_per_channel]; +"96 conv2d_6" [id=96, type=conv2d]; +"97 relu__2" [id=97, type=relu_]; +"98 quantize_per_tensor_default_5" [id=98, type=quantize_per_tensor]; +"99 dequantize_per_tensor_default_6" [id=99, type=dequantize_per_tensor]; +"100 features_2_block_2_0_weight_bias" [id=100, type=get_attr]; +"101 conv2d_7_scale_0" [id=101, type=get_attr]; +"102 conv2d_7_zero_point_0" [id=102, type=get_attr]; +"103 _frozen_param5" [id=103, type=get_attr]; +"104 dequantize_per_channel_default_7" [id=104, type=dequantize_per_channel]; +"105 conv2d_7" [id=105, type=conv2d]; +"106 quantize_per_tensor_default_6" [id=106, type=quantize_per_tensor]; +"107 dequantize_per_tensor_default_8" [id=107, type=dequantize_per_tensor]; +"108 dequantize_per_tensor_default_7" [id=108, type=dequantize_per_tensor]; +"109 features_3_block_0_0_weight_bias" [id=109, type=get_attr]; +"110 conv2d_8_scale_0" [id=110, type=get_attr]; +"111 conv2d_8_zero_point_0" [id=111, type=get_attr]; +"112 _frozen_param6" [id=112, type=get_attr]; +"113 dequantize_per_channel_default_8" [id=113, type=dequantize_per_channel]; +"114 conv2d_8" [id=114, type=conv2d]; +"115 relu__3_scale_0" [id=115, type=get_attr]; +"116 relu__3_zero_point_0" [id=116, type=get_attr]; +"117 relu__3" [id=117, type=relu_]; +"118 quantize_per_channel_default_9" [id=118, type=quantize_per_channel]; +"119 dequantize_per_channel_default_9" [id=119, type=dequantize_per_channel]; +"120 features_3_block_1_0_weight_bias" [id=120, type=get_attr]; +"121 conv2d_9_scale_0" [id=121, type=get_attr]; +"122 conv2d_9_zero_point_0" [id=122, type=get_attr]; +"123 _frozen_param7" [id=123, type=get_attr]; +"124 dequantize_per_channel_default_10" [id=124, type=dequantize_per_channel]; +"125 conv2d_9" [id=125, type=conv2d]; +"126 relu__4" [id=126, type=relu_]; +"127 quantize_per_tensor_default_7" [id=127, type=quantize_per_tensor]; +"128 dequantize_per_tensor_default_9" [id=128, type=dequantize_per_tensor]; +"129 features_3_block_2_0_weight_bias" [id=129, type=get_attr]; +"130 conv2d_10_scale_0" [id=130, type=get_attr]; +"131 conv2d_10_zero_point_0" [id=131, type=get_attr]; +"132 _frozen_param8" [id=132, type=get_attr]; +"133 dequantize_per_channel_default_11" [id=133, type=dequantize_per_channel]; +"134 conv2d_10" [id=134, type=conv2d]; +"135 quantize_per_tensor_default_8" [id=135, type=quantize_per_tensor]; +"136 dequantize_per_tensor_default_10" [id=136, type=dequantize_per_tensor]; +"137 add_" [id=137, type=add_]; +"138 quantize_per_tensor_default_9" [id=138, type=quantize_per_tensor]; +"139 dequantize_per_tensor_default_11" [id=139, type=dequantize_per_tensor]; +"140 features_4_block_0_0_weight_bias" [id=140, type=get_attr]; +"141 conv2d_11_scale_0" [id=141, type=get_attr]; +"142 conv2d_11_zero_point_0" [id=142, type=get_attr]; +"143 _frozen_param9" [id=143, type=get_attr]; +"144 dequantize_per_channel_default_12" [id=144, type=dequantize_per_channel]; +"145 conv2d_11" [id=145, type=conv2d]; +"146 hardswish__1_scale_0" [id=146, type=get_attr]; +"147 hardswish__1_zero_point_0" [id=147, type=get_attr]; +"148 hardswish__1" [id=148, type=hardswish_]; +"149 quantize_per_channel_default_13" [id=149, type=quantize_per_channel]; +"150 dequantize_per_channel_default_13" [id=150, type=dequantize_per_channel]; +"151 features_4_block_1_0_weight_bias" [id=151, type=get_attr]; +"152 conv2d_12_scale_0" [id=152, type=get_attr]; +"153 conv2d_12_zero_point_0" [id=153, type=get_attr]; +"154 _frozen_param10" [id=154, type=get_attr]; +"155 dequantize_per_channel_default_14" [id=155, type=dequantize_per_channel]; +"156 conv2d_12" [id=156, type=conv2d]; +"157 hardswish__2" [id=157, type=hardswish_]; +"158 quantize_per_tensor_default_10" [id=158, type=quantize_per_tensor]; +"159 dequantize_per_tensor_default_13" [id=159, type=dequantize_per_tensor]; +"160 dequantize_per_tensor_default_12" [id=160, type=dequantize_per_tensor]; +"161 adaptive_avg_pool2d_1" [id=161, type=adaptive_avg_pool2d]; +"162 conv2d_13" [id=162, type=conv2d]; +"163 relu_1" [id=163, type=relu]; +"164 conv2d_14" [id=164, type=conv2d]; +"165 hardsigmoid_1" [id=165, type=hardsigmoid]; +"166 quantize_per_tensor_default_11" [id=166, type=quantize_per_tensor]; +"167 dequantize_per_tensor_default_14" [id=167, type=dequantize_per_tensor]; +"168 mul_1" [id=168, type=mul]; +"169 quantize_per_tensor_default_12" [id=169, type=quantize_per_tensor]; +"170 dequantize_per_tensor_default_15" [id=170, type=dequantize_per_tensor]; +"171 features_4_block_3_0_weight_bias" [id=171, type=get_attr]; +"172 conv2d_15_scale_0" [id=172, type=get_attr]; +"173 conv2d_15_zero_point_0" [id=173, type=get_attr]; +"174 _frozen_param11" [id=174, type=get_attr]; +"175 dequantize_per_channel_default_15" [id=175, type=dequantize_per_channel]; +"176 conv2d_15" [id=176, type=conv2d]; +"177 quantize_per_tensor_default_13" [id=177, type=quantize_per_tensor]; +"178 dequantize_per_tensor_default_17" [id=178, type=dequantize_per_tensor]; +"179 dequantize_per_tensor_default_16" [id=179, type=dequantize_per_tensor]; +"180 features_5_block_0_0_weight_bias" [id=180, type=get_attr]; +"181 conv2d_16_scale_0" [id=181, type=get_attr]; +"182 conv2d_16_zero_point_0" [id=182, type=get_attr]; +"183 _frozen_param12" [id=183, type=get_attr]; +"184 dequantize_per_channel_default_16" [id=184, type=dequantize_per_channel]; +"185 conv2d_16" [id=185, type=conv2d]; +"186 hardswish__3_scale_0" [id=186, type=get_attr]; +"187 hardswish__3_zero_point_0" [id=187, type=get_attr]; +"188 hardswish__3" [id=188, type=hardswish_]; +"189 quantize_per_channel_default_17" [id=189, type=quantize_per_channel]; +"190 dequantize_per_channel_default_17" [id=190, type=dequantize_per_channel]; +"191 features_5_block_1_0_weight_bias" [id=191, type=get_attr]; +"192 conv2d_17_scale_0" [id=192, type=get_attr]; +"193 conv2d_17_zero_point_0" [id=193, type=get_attr]; +"194 _frozen_param13" [id=194, type=get_attr]; +"195 dequantize_per_channel_default_18" [id=195, type=dequantize_per_channel]; +"196 conv2d_17" [id=196, type=conv2d]; +"197 hardswish__4" [id=197, type=hardswish_]; +"198 quantize_per_tensor_default_14" [id=198, type=quantize_per_tensor]; +"199 dequantize_per_tensor_default_19" [id=199, type=dequantize_per_tensor]; +"200 dequantize_per_tensor_default_18" [id=200, type=dequantize_per_tensor]; +"201 adaptive_avg_pool2d_2" [id=201, type=adaptive_avg_pool2d]; +"202 conv2d_18" [id=202, type=conv2d]; +"203 relu_2" [id=203, type=relu]; +"204 conv2d_19" [id=204, type=conv2d]; +"205 hardsigmoid_2" [id=205, type=hardsigmoid]; +"206 quantize_per_tensor_default_15" [id=206, type=quantize_per_tensor]; +"207 dequantize_per_tensor_default_20" [id=207, type=dequantize_per_tensor]; +"208 mul_2" [id=208, type=mul]; +"209 quantize_per_tensor_default_16" [id=209, type=quantize_per_tensor]; +"210 dequantize_per_tensor_default_21" [id=210, type=dequantize_per_tensor]; +"211 features_5_block_3_0_weight_bias" [id=211, type=get_attr]; +"212 conv2d_20_scale_0" [id=212, type=get_attr]; +"213 conv2d_20_zero_point_0" [id=213, type=get_attr]; +"214 _frozen_param14" [id=214, type=get_attr]; +"215 dequantize_per_channel_default_19" [id=215, type=dequantize_per_channel]; +"216 conv2d_20" [id=216, type=conv2d]; +"217 quantize_per_tensor_default_17" [id=217, type=quantize_per_tensor]; +"218 dequantize_per_tensor_default_22" [id=218, type=dequantize_per_tensor]; +"219 add__1" [id=219, type=add_]; +"220 quantize_per_tensor_default_18" [id=220, type=quantize_per_tensor]; +"221 dequantize_per_tensor_default_24" [id=221, type=dequantize_per_tensor]; +"222 dequantize_per_tensor_default_23" [id=222, type=dequantize_per_tensor]; +"223 features_6_block_0_0_weight_bias" [id=223, type=get_attr]; +"224 conv2d_21_scale_0" [id=224, type=get_attr]; +"225 conv2d_21_zero_point_0" [id=225, type=get_attr]; +"226 _frozen_param15" [id=226, type=get_attr]; +"227 dequantize_per_channel_default_20" [id=227, type=dequantize_per_channel]; +"228 conv2d_21" [id=228, type=conv2d]; +"229 hardswish__5_scale_0" [id=229, type=get_attr]; +"230 hardswish__5_zero_point_0" [id=230, type=get_attr]; +"231 hardswish__5" [id=231, type=hardswish_]; +"232 quantize_per_channel_default_21" [id=232, type=quantize_per_channel]; +"233 dequantize_per_channel_default_21" [id=233, type=dequantize_per_channel]; +"234 features_6_block_1_0_weight_bias" [id=234, type=get_attr]; +"235 conv2d_22_scale_0" [id=235, type=get_attr]; +"236 conv2d_22_zero_point_0" [id=236, type=get_attr]; +"237 _frozen_param16" [id=237, type=get_attr]; +"238 dequantize_per_channel_default_22" [id=238, type=dequantize_per_channel]; +"239 conv2d_22" [id=239, type=conv2d]; +"240 hardswish__6" [id=240, type=hardswish_]; +"241 quantize_per_tensor_default_19" [id=241, type=quantize_per_tensor]; +"242 dequantize_per_tensor_default_26" [id=242, type=dequantize_per_tensor]; +"243 dequantize_per_tensor_default_25" [id=243, type=dequantize_per_tensor]; +"244 adaptive_avg_pool2d_3" [id=244, type=adaptive_avg_pool2d]; +"245 conv2d_23" [id=245, type=conv2d]; +"246 relu_3" [id=246, type=relu]; +"247 conv2d_24" [id=247, type=conv2d]; +"248 hardsigmoid_3" [id=248, type=hardsigmoid]; +"249 quantize_per_tensor_default_20" [id=249, type=quantize_per_tensor]; +"250 dequantize_per_tensor_default_27" [id=250, type=dequantize_per_tensor]; +"251 mul_3" [id=251, type=mul]; +"252 quantize_per_tensor_default_21" [id=252, type=quantize_per_tensor]; +"253 dequantize_per_tensor_default_28" [id=253, type=dequantize_per_tensor]; +"254 features_6_block_3_0_weight_bias" [id=254, type=get_attr]; +"255 conv2d_25_scale_0" [id=255, type=get_attr]; +"256 conv2d_25_zero_point_0" [id=256, type=get_attr]; +"257 _frozen_param17" [id=257, type=get_attr]; +"258 dequantize_per_channel_default_23" [id=258, type=dequantize_per_channel]; +"259 conv2d_25" [id=259, type=conv2d]; +"260 quantize_per_tensor_default_22" [id=260, type=quantize_per_tensor]; +"261 dequantize_per_tensor_default_29" [id=261, type=dequantize_per_tensor]; +"262 add__2" [id=262, type=add_]; +"263 quantize_per_tensor_default_23" [id=263, type=quantize_per_tensor]; +"264 dequantize_per_tensor_default_30" [id=264, type=dequantize_per_tensor]; +"265 features_7_block_0_0_weight_bias" [id=265, type=get_attr]; +"266 conv2d_26_scale_0" [id=266, type=get_attr]; +"267 conv2d_26_zero_point_0" [id=267, type=get_attr]; +"268 _frozen_param18" [id=268, type=get_attr]; +"269 dequantize_per_channel_default_24" [id=269, type=dequantize_per_channel]; +"270 conv2d_26" [id=270, type=conv2d]; +"271 hardswish__7_scale_0" [id=271, type=get_attr]; +"272 hardswish__7_zero_point_0" [id=272, type=get_attr]; +"273 hardswish__7" [id=273, type=hardswish_]; +"274 quantize_per_channel_default_25" [id=274, type=quantize_per_channel]; +"275 dequantize_per_channel_default_25" [id=275, type=dequantize_per_channel]; +"276 features_7_block_1_0_weight_bias" [id=276, type=get_attr]; +"277 conv2d_27_scale_0" [id=277, type=get_attr]; +"278 conv2d_27_zero_point_0" [id=278, type=get_attr]; +"279 _frozen_param19" [id=279, type=get_attr]; +"280 dequantize_per_channel_default_26" [id=280, type=dequantize_per_channel]; +"281 conv2d_27" [id=281, type=conv2d]; +"282 hardswish__8" [id=282, type=hardswish_]; +"283 quantize_per_tensor_default_24" [id=283, type=quantize_per_tensor]; +"284 dequantize_per_tensor_default_32" [id=284, type=dequantize_per_tensor]; +"285 dequantize_per_tensor_default_31" [id=285, type=dequantize_per_tensor]; +"286 adaptive_avg_pool2d_4" [id=286, type=adaptive_avg_pool2d]; +"287 conv2d_28" [id=287, type=conv2d]; +"288 relu_4" [id=288, type=relu]; +"289 conv2d_29" [id=289, type=conv2d]; +"290 hardsigmoid_4" [id=290, type=hardsigmoid]; +"291 quantize_per_tensor_default_25" [id=291, type=quantize_per_tensor]; +"292 dequantize_per_tensor_default_33" [id=292, type=dequantize_per_tensor]; +"293 mul_4" [id=293, type=mul]; +"294 quantize_per_tensor_default_26" [id=294, type=quantize_per_tensor]; +"295 dequantize_per_tensor_default_34" [id=295, type=dequantize_per_tensor]; +"296 features_7_block_3_0_weight_bias" [id=296, type=get_attr]; +"297 conv2d_30_scale_0" [id=297, type=get_attr]; +"298 conv2d_30_zero_point_0" [id=298, type=get_attr]; +"299 _frozen_param20" [id=299, type=get_attr]; +"300 dequantize_per_channel_default_27" [id=300, type=dequantize_per_channel]; +"301 conv2d_30" [id=301, type=conv2d]; +"302 quantize_per_tensor_default_27" [id=302, type=quantize_per_tensor]; +"303 dequantize_per_tensor_default_36" [id=303, type=dequantize_per_tensor]; +"304 dequantize_per_tensor_default_35" [id=304, type=dequantize_per_tensor]; +"305 features_8_block_0_0_weight_bias" [id=305, type=get_attr]; +"306 conv2d_31_scale_0" [id=306, type=get_attr]; +"307 conv2d_31_zero_point_0" [id=307, type=get_attr]; +"308 _frozen_param21" [id=308, type=get_attr]; +"309 dequantize_per_channel_default_28" [id=309, type=dequantize_per_channel]; +"310 conv2d_31" [id=310, type=conv2d]; +"311 hardswish__9_scale_0" [id=311, type=get_attr]; +"312 hardswish__9_zero_point_0" [id=312, type=get_attr]; +"313 hardswish__9" [id=313, type=hardswish_]; +"314 quantize_per_channel_default_29" [id=314, type=quantize_per_channel]; +"315 dequantize_per_channel_default_29" [id=315, type=dequantize_per_channel]; +"316 features_8_block_1_0_weight_bias" [id=316, type=get_attr]; +"317 conv2d_32_scale_0" [id=317, type=get_attr]; +"318 conv2d_32_zero_point_0" [id=318, type=get_attr]; +"319 _frozen_param22" [id=319, type=get_attr]; +"320 dequantize_per_channel_default_30" [id=320, type=dequantize_per_channel]; +"321 conv2d_32" [id=321, type=conv2d]; +"322 hardswish__10" [id=322, type=hardswish_]; +"323 quantize_per_tensor_default_28" [id=323, type=quantize_per_tensor]; +"324 dequantize_per_tensor_default_38" [id=324, type=dequantize_per_tensor]; +"325 dequantize_per_tensor_default_37" [id=325, type=dequantize_per_tensor]; +"326 adaptive_avg_pool2d_5" [id=326, type=adaptive_avg_pool2d]; +"327 conv2d_33" [id=327, type=conv2d]; +"328 relu_5" [id=328, type=relu]; +"329 conv2d_34" [id=329, type=conv2d]; +"330 hardsigmoid_5" [id=330, type=hardsigmoid]; +"331 quantize_per_tensor_default_29" [id=331, type=quantize_per_tensor]; +"332 dequantize_per_tensor_default_39" [id=332, type=dequantize_per_tensor]; +"333 mul_5" [id=333, type=mul]; +"334 quantize_per_tensor_default_30" [id=334, type=quantize_per_tensor]; +"335 dequantize_per_tensor_default_40" [id=335, type=dequantize_per_tensor]; +"336 features_8_block_3_0_weight_bias" [id=336, type=get_attr]; +"337 conv2d_35_scale_0" [id=337, type=get_attr]; +"338 conv2d_35_zero_point_0" [id=338, type=get_attr]; +"339 _frozen_param23" [id=339, type=get_attr]; +"340 dequantize_per_channel_default_31" [id=340, type=dequantize_per_channel]; +"341 conv2d_35" [id=341, type=conv2d]; +"342 quantize_per_tensor_default_31" [id=342, type=quantize_per_tensor]; +"343 dequantize_per_tensor_default_41" [id=343, type=dequantize_per_tensor]; +"344 add__3" [id=344, type=add_]; +"345 quantize_per_tensor_default_32" [id=345, type=quantize_per_tensor]; +"346 dequantize_per_tensor_default_42" [id=346, type=dequantize_per_tensor]; +"347 features_9_block_0_0_weight_bias" [id=347, type=get_attr]; +"348 conv2d_36_scale_0" [id=348, type=get_attr]; +"349 conv2d_36_zero_point_0" [id=349, type=get_attr]; +"350 _frozen_param24" [id=350, type=get_attr]; +"351 dequantize_per_channel_default_32" [id=351, type=dequantize_per_channel]; +"352 conv2d_36" [id=352, type=conv2d]; +"353 hardswish__11_scale_0" [id=353, type=get_attr]; +"354 hardswish__11_zero_point_0" [id=354, type=get_attr]; +"355 hardswish__11" [id=355, type=hardswish_]; +"356 quantize_per_channel_default_33" [id=356, type=quantize_per_channel]; +"357 dequantize_per_channel_default_33" [id=357, type=dequantize_per_channel]; +"358 features_9_block_1_0_weight_bias" [id=358, type=get_attr]; +"359 conv2d_37_scale_0" [id=359, type=get_attr]; +"360 conv2d_37_zero_point_0" [id=360, type=get_attr]; +"361 _frozen_param25" [id=361, type=get_attr]; +"362 dequantize_per_channel_default_34" [id=362, type=dequantize_per_channel]; +"363 conv2d_37" [id=363, type=conv2d]; +"364 hardswish__12" [id=364, type=hardswish_]; +"365 quantize_per_tensor_default_33" [id=365, type=quantize_per_tensor]; +"366 dequantize_per_tensor_default_44" [id=366, type=dequantize_per_tensor]; +"367 dequantize_per_tensor_default_43" [id=367, type=dequantize_per_tensor]; +"368 adaptive_avg_pool2d_6" [id=368, type=adaptive_avg_pool2d]; +"369 conv2d_38" [id=369, type=conv2d]; +"370 relu_6" [id=370, type=relu]; +"371 conv2d_39" [id=371, type=conv2d]; +"372 hardsigmoid_6" [id=372, type=hardsigmoid]; +"373 quantize_per_tensor_default_34" [id=373, type=quantize_per_tensor]; +"374 dequantize_per_tensor_default_45" [id=374, type=dequantize_per_tensor]; +"375 mul_6" [id=375, type=mul]; +"376 quantize_per_tensor_default_35" [id=376, type=quantize_per_tensor]; +"377 dequantize_per_tensor_default_46" [id=377, type=dequantize_per_tensor]; +"378 features_9_block_3_0_weight_bias" [id=378, type=get_attr]; +"379 conv2d_40_scale_0" [id=379, type=get_attr]; +"380 conv2d_40_zero_point_0" [id=380, type=get_attr]; +"381 _frozen_param26" [id=381, type=get_attr]; +"382 dequantize_per_channel_default_35" [id=382, type=dequantize_per_channel]; +"383 conv2d_40" [id=383, type=conv2d]; +"384 quantize_per_tensor_default_36" [id=384, type=quantize_per_tensor]; +"385 dequantize_per_tensor_default_48" [id=385, type=dequantize_per_tensor]; +"386 dequantize_per_tensor_default_47" [id=386, type=dequantize_per_tensor]; +"387 features_10_block_0_0_weight_bias" [id=387, type=get_attr]; +"388 conv2d_41_scale_0" [id=388, type=get_attr]; +"389 conv2d_41_zero_point_0" [id=389, type=get_attr]; +"390 _frozen_param27" [id=390, type=get_attr]; +"391 dequantize_per_channel_default_36" [id=391, type=dequantize_per_channel]; +"392 conv2d_41" [id=392, type=conv2d]; +"393 hardswish__13_scale_0" [id=393, type=get_attr]; +"394 hardswish__13_zero_point_0" [id=394, type=get_attr]; +"395 hardswish__13" [id=395, type=hardswish_]; +"396 quantize_per_channel_default_37" [id=396, type=quantize_per_channel]; +"397 dequantize_per_channel_default_37" [id=397, type=dequantize_per_channel]; +"398 features_10_block_1_0_weight_bias" [id=398, type=get_attr]; +"399 conv2d_42_scale_0" [id=399, type=get_attr]; +"400 conv2d_42_zero_point_0" [id=400, type=get_attr]; +"401 _frozen_param28" [id=401, type=get_attr]; +"402 dequantize_per_channel_default_38" [id=402, type=dequantize_per_channel]; +"403 conv2d_42" [id=403, type=conv2d]; +"404 hardswish__14" [id=404, type=hardswish_]; +"405 quantize_per_tensor_default_37" [id=405, type=quantize_per_tensor]; +"406 dequantize_per_tensor_default_50" [id=406, type=dequantize_per_tensor]; +"407 dequantize_per_tensor_default_49" [id=407, type=dequantize_per_tensor]; +"408 adaptive_avg_pool2d_7" [id=408, type=adaptive_avg_pool2d]; +"409 conv2d_43" [id=409, type=conv2d]; +"410 relu_7" [id=410, type=relu]; +"411 conv2d_44" [id=411, type=conv2d]; +"412 hardsigmoid_7" [id=412, type=hardsigmoid]; +"413 quantize_per_tensor_default_38" [id=413, type=quantize_per_tensor]; +"414 dequantize_per_tensor_default_51" [id=414, type=dequantize_per_tensor]; +"415 mul_7" [id=415, type=mul]; +"416 quantize_per_tensor_default_39" [id=416, type=quantize_per_tensor]; +"417 dequantize_per_tensor_default_52" [id=417, type=dequantize_per_tensor]; +"418 features_10_block_3_0_weight_bias" [id=418, type=get_attr]; +"419 conv2d_45_scale_0" [id=419, type=get_attr]; +"420 conv2d_45_zero_point_0" [id=420, type=get_attr]; +"421 _frozen_param29" [id=421, type=get_attr]; +"422 dequantize_per_channel_default_39" [id=422, type=dequantize_per_channel]; +"423 conv2d_45" [id=423, type=conv2d]; +"424 quantize_per_tensor_default_40" [id=424, type=quantize_per_tensor]; +"425 dequantize_per_tensor_default_53" [id=425, type=dequantize_per_tensor]; +"426 add__4" [id=426, type=add_]; +"427 quantize_per_tensor_default_41" [id=427, type=quantize_per_tensor]; +"428 dequantize_per_tensor_default_55" [id=428, type=dequantize_per_tensor]; +"429 dequantize_per_tensor_default_54" [id=429, type=dequantize_per_tensor]; +"430 features_11_block_0_0_weight_bias" [id=430, type=get_attr]; +"431 conv2d_46_scale_0" [id=431, type=get_attr]; +"432 conv2d_46_zero_point_0" [id=432, type=get_attr]; +"433 _frozen_param30" [id=433, type=get_attr]; +"434 dequantize_per_channel_default_40" [id=434, type=dequantize_per_channel]; +"435 conv2d_46" [id=435, type=conv2d]; +"436 hardswish__15_scale_0" [id=436, type=get_attr]; +"437 hardswish__15_zero_point_0" [id=437, type=get_attr]; +"438 hardswish__15" [id=438, type=hardswish_]; +"439 quantize_per_channel_default_41" [id=439, type=quantize_per_channel]; +"440 dequantize_per_channel_default_41" [id=440, type=dequantize_per_channel]; +"441 features_11_block_1_0_weight_bias" [id=441, type=get_attr]; +"442 conv2d_47_scale_0" [id=442, type=get_attr]; +"443 conv2d_47_zero_point_0" [id=443, type=get_attr]; +"444 _frozen_param31" [id=444, type=get_attr]; +"445 dequantize_per_channel_default_42" [id=445, type=dequantize_per_channel]; +"446 conv2d_47" [id=446, type=conv2d]; +"447 hardswish__16" [id=447, type=hardswish_]; +"448 quantize_per_tensor_default_42" [id=448, type=quantize_per_tensor]; +"449 dequantize_per_tensor_default_57" [id=449, type=dequantize_per_tensor]; +"450 dequantize_per_tensor_default_56" [id=450, type=dequantize_per_tensor]; +"451 adaptive_avg_pool2d_8" [id=451, type=adaptive_avg_pool2d]; +"452 conv2d_48" [id=452, type=conv2d]; +"453 relu_8" [id=453, type=relu]; +"454 conv2d_49" [id=454, type=conv2d]; +"455 hardsigmoid_8" [id=455, type=hardsigmoid]; +"456 quantize_per_tensor_default_43" [id=456, type=quantize_per_tensor]; +"457 dequantize_per_tensor_default_58" [id=457, type=dequantize_per_tensor]; +"458 mul_8" [id=458, type=mul]; +"459 quantize_per_tensor_default_44" [id=459, type=quantize_per_tensor]; +"460 dequantize_per_tensor_default_59" [id=460, type=dequantize_per_tensor]; +"461 features_11_block_3_0_weight_bias" [id=461, type=get_attr]; +"462 conv2d_50_scale_0" [id=462, type=get_attr]; +"463 conv2d_50_zero_point_0" [id=463, type=get_attr]; +"464 _frozen_param32" [id=464, type=get_attr]; +"465 dequantize_per_channel_default_43" [id=465, type=dequantize_per_channel]; +"466 conv2d_50" [id=466, type=conv2d]; +"467 quantize_per_tensor_default_45" [id=467, type=quantize_per_tensor]; +"468 dequantize_per_tensor_default_60" [id=468, type=dequantize_per_tensor]; +"469 add__5" [id=469, type=add_]; +"470 quantize_per_tensor_default_46" [id=470, type=quantize_per_tensor]; +"471 dequantize_per_tensor_default_61" [id=471, type=dequantize_per_tensor]; +"472 features_12_0_weight_bias" [id=472, type=get_attr]; +"473 conv2d_51_scale_0" [id=473, type=get_attr]; +"474 conv2d_51_zero_point_0" [id=474, type=get_attr]; +"475 _frozen_param33" [id=475, type=get_attr]; +"476 dequantize_per_channel_default_44" [id=476, type=dequantize_per_channel]; +"477 conv2d_51" [id=477, type=conv2d]; +"478 hardswish__17" [id=478, type=hardswish_]; +"479 quantize_per_tensor_default_47" [id=479, type=quantize_per_tensor]; +"480 dequantize_per_tensor_default_62" [id=480, type=dequantize_per_tensor]; +"481 adaptive_avg_pool2d_9" [id=481, type=adaptive_avg_pool2d]; +"482 flatten" [id=482, type=flatten]; +"483 flatten_0_0_nncf_smooth_quant_0" [id=483, type=call_module]; +"484 quantize_per_tensor_default_48" [id=484, type=quantize_per_tensor]; +"485 dequantize_per_tensor_default_63" [id=485, type=dequantize_per_tensor]; +"486 linear_scale_0" [id=486, type=get_attr]; +"487 linear_zero_point_0" [id=487, type=get_attr]; +"488 _frozen_param34" [id=488, type=get_attr]; +"489 dequantize_per_channel_default_45" [id=489, type=dequantize_per_channel]; +"490 linear" [id=490, type=linear]; +"491 hardswish__18" [id=491, type=hardswish_]; +"492 dropout_" [id=492, type=dropout_]; +"493 dropout__0_0_nncf_smooth_quant_0" [id=493, type=call_module]; +"494 quantize_per_tensor_default_49" [id=494, type=quantize_per_tensor]; +"495 dequantize_per_tensor_default_64" [id=495, type=dequantize_per_tensor]; +"496 linear_1_scale_0" [id=496, type=get_attr]; +"497 linear_1_zero_point_0" [id=497, type=get_attr]; +"498 _frozen_param35" [id=498, type=get_attr]; +"499 dequantize_per_channel_default_46" [id=499, type=dequantize_per_channel]; +"500 linear_1" [id=500, type=linear]; +"501 output" [id=501, type=output]; +"0 features_1_block_1_fc1_weight" -> "63 conv2d_2" [label="(8, 16, 1, 1)", style=solid]; +"1 features_1_block_1_fc1_bias" -> "63 conv2d_2" [label="(8,)", style=solid]; +"2 features_1_block_1_fc2_weight" -> "65 conv2d_3" [label="(16, 8, 1, 1)", style=solid]; +"3 features_1_block_1_fc2_bias" -> "65 conv2d_3" [label="(16,)", style=solid]; +"4 features_4_block_2_fc1_weight" -> "162 conv2d_13" [label="(24, 96, 1, 1)", style=solid]; +"5 features_4_block_2_fc1_bias" -> "162 conv2d_13" [label="(24,)", style=solid]; +"6 features_4_block_2_fc2_weight" -> "164 conv2d_14" [label="(96, 24, 1, 1)", style=solid]; +"7 features_4_block_2_fc2_bias" -> "164 conv2d_14" [label="(96,)", style=solid]; +"8 features_5_block_2_fc1_weight" -> "202 conv2d_18" [label="(64, 240, 1, 1)", style=solid]; +"9 features_5_block_2_fc1_bias" -> "202 conv2d_18" [label="(64,)", style=solid]; +"10 features_5_block_2_fc2_weight" -> "204 conv2d_19" [label="(240, 64, 1, 1)", style=solid]; +"11 features_5_block_2_fc2_bias" -> "204 conv2d_19" [label="(240,)", style=solid]; +"12 features_6_block_2_fc1_weight" -> "245 conv2d_23" [label="(64, 240, 1, 1)", style=solid]; +"13 features_6_block_2_fc1_bias" -> "245 conv2d_23" [label="(64,)", style=solid]; +"14 features_6_block_2_fc2_weight" -> "247 conv2d_24" [label="(240, 64, 1, 1)", style=solid]; +"15 features_6_block_2_fc2_bias" -> "247 conv2d_24" [label="(240,)", style=solid]; +"16 features_7_block_2_fc1_weight" -> "287 conv2d_28" [label="(32, 120, 1, 1)", style=solid]; +"17 features_7_block_2_fc1_bias" -> "287 conv2d_28" [label="(32,)", style=solid]; +"18 features_7_block_2_fc2_weight" -> "289 conv2d_29" [label="(120, 32, 1, 1)", style=solid]; +"19 features_7_block_2_fc2_bias" -> "289 conv2d_29" [label="(120,)", style=solid]; +"20 features_8_block_2_fc1_weight" -> "327 conv2d_33" [label="(40, 144, 1, 1)", style=solid]; +"21 features_8_block_2_fc1_bias" -> "327 conv2d_33" [label="(40,)", style=solid]; +"22 features_8_block_2_fc2_weight" -> "329 conv2d_34" [label="(144, 40, 1, 1)", style=solid]; +"23 features_8_block_2_fc2_bias" -> "329 conv2d_34" [label="(144,)", style=solid]; +"24 features_9_block_2_fc1_weight" -> "369 conv2d_38" [label="(72, 288, 1, 1)", style=solid]; +"25 features_9_block_2_fc1_bias" -> "369 conv2d_38" [label="(72,)", style=solid]; +"26 features_9_block_2_fc2_weight" -> "371 conv2d_39" [label="(288, 72, 1, 1)", style=solid]; +"27 features_9_block_2_fc2_bias" -> "371 conv2d_39" [label="(288,)", style=solid]; +"28 features_10_block_2_fc1_weight" -> "409 conv2d_43" [label="(144, 576, 1, 1)", style=solid]; +"29 features_10_block_2_fc1_bias" -> "409 conv2d_43" [label="(144,)", style=solid]; +"30 features_10_block_2_fc2_weight" -> "411 conv2d_44" [label="(576, 144, 1, 1)", style=solid]; +"31 features_10_block_2_fc2_bias" -> "411 conv2d_44" [label="(576,)", style=solid]; +"32 features_11_block_2_fc1_weight" -> "452 conv2d_48" [label="(144, 576, 1, 1)", style=solid]; +"33 features_11_block_2_fc1_bias" -> "452 conv2d_48" [label="(144,)", style=solid]; +"34 features_11_block_2_fc2_weight" -> "454 conv2d_49" [label="(576, 144, 1, 1)", style=solid]; +"35 features_11_block_2_fc2_bias" -> "454 conv2d_49" [label="(576,)", style=solid]; +"36 classifier_0_bias" -> "490 linear" [label="(1024,)", style=solid]; +"37 classifier_3_bias" -> "500 linear_1" [label="(1000,)", style=solid]; +"38 x" -> "39 quantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; +"39 quantize_per_tensor_default" -> "40 dequantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; +"40 dequantize_per_tensor_default" -> "46 conv2d" [label="(1, 3, 224, 224)", style=solid]; +"41 features_0_0_weight_bias" -> "46 conv2d" [label="(16,)", style=solid]; +"42 conv2d_scale_0" -> "45 dequantize_per_channel_default" [label="(16,)", style=solid]; +"43 conv2d_zero_point_0" -> "45 dequantize_per_channel_default" [label="(16,)", style=solid]; +"44 _frozen_param0" -> "45 dequantize_per_channel_default" [label="(16, 3, 3, 3)", style=solid]; +"45 dequantize_per_channel_default" -> "46 conv2d" [label="(16, 3, 3, 3)", style=solid]; +"46 conv2d" -> "49 hardswish_" [label="(1, 16, 112, 112)", style=solid]; +"47 hardswish__scale_0" -> "50 quantize_per_channel_default_1" [label="(16,)", style=solid]; +"47 hardswish__scale_0" -> "51 dequantize_per_channel_default_1" [label="(16,)", style=solid]; +"48 hardswish__zero_point_0" -> "50 quantize_per_channel_default_1" [label="(16,)", style=solid]; +"48 hardswish__zero_point_0" -> "51 dequantize_per_channel_default_1" [label="(16,)", style=solid]; +"49 hardswish_" -> "50 quantize_per_channel_default_1" [label="(1, 16, 112, 112)", style=solid]; +"50 quantize_per_channel_default_1" -> "51 dequantize_per_channel_default_1" [label="(1, 16, 112, 112)", style=solid]; +"51 dequantize_per_channel_default_1" -> "57 conv2d_1" [label="(1, 16, 112, 112)", style=solid]; +"52 features_1_block_0_0_weight_bias" -> "57 conv2d_1" [label="(16,)", style=solid]; +"53 conv2d_1_scale_0" -> "56 dequantize_per_channel_default_2" [label="(16,)", style=solid]; +"54 conv2d_1_zero_point_0" -> "56 dequantize_per_channel_default_2" [label="(16,)", style=solid]; +"55 _frozen_param1" -> "56 dequantize_per_channel_default_2" [label="(16, 1, 3, 3)", style=solid]; +"56 dequantize_per_channel_default_2" -> "57 conv2d_1" [label="(16, 1, 3, 3)", style=solid]; +"57 conv2d_1" -> "58 relu_" [label="(1, 16, 56, 56)", style=solid]; +"58 relu_" -> "59 quantize_per_tensor_default_1" [label="(1, 16, 56, 56)", style=solid]; +"59 quantize_per_tensor_default_1" -> "60 dequantize_per_tensor_default_2" [label="(1, 16, 56, 56)", style=solid]; +"59 quantize_per_tensor_default_1" -> "61 dequantize_per_tensor_default_1" [label="(1, 16, 56, 56)", style=solid]; +"60 dequantize_per_tensor_default_2" -> "69 mul" [label="(1, 16, 56, 56)", style=solid]; +"61 dequantize_per_tensor_default_1" -> "62 adaptive_avg_pool2d" [label="(1, 16, 56, 56)", style=solid]; +"62 adaptive_avg_pool2d" -> "63 conv2d_2" [label="(1, 16, 1, 1)", style=solid]; +"63 conv2d_2" -> "64 relu" [label="(1, 8, 1, 1)", style=solid]; +"64 relu" -> "65 conv2d_3" [label="(1, 8, 1, 1)", style=solid]; +"65 conv2d_3" -> "66 hardsigmoid" [label="(1, 16, 1, 1)", style=solid]; +"66 hardsigmoid" -> "67 quantize_per_tensor_default_2" [label="(1, 16, 1, 1)", style=solid]; +"67 quantize_per_tensor_default_2" -> "68 dequantize_per_tensor_default_3" [label="(1, 16, 1, 1)", style=solid]; +"68 dequantize_per_tensor_default_3" -> "69 mul" [label="(1, 16, 1, 1)", style=solid]; +"69 mul" -> "70 quantize_per_tensor_default_3" [label="(1, 16, 56, 56)", style=solid]; +"70 quantize_per_tensor_default_3" -> "71 dequantize_per_tensor_default_4" [label="(1, 16, 56, 56)", style=solid]; +"71 dequantize_per_tensor_default_4" -> "77 conv2d_4" [label="(1, 16, 56, 56)", style=solid]; +"72 features_1_block_2_0_weight_bias" -> "77 conv2d_4" [label="(16,)", style=solid]; +"73 conv2d_4_scale_0" -> "76 dequantize_per_channel_default_3" [label="(16,)", style=solid]; +"74 conv2d_4_zero_point_0" -> "76 dequantize_per_channel_default_3" [label="(16,)", style=solid]; +"75 _frozen_param2" -> "76 dequantize_per_channel_default_3" [label="(16, 16, 1, 1)", style=solid]; +"76 dequantize_per_channel_default_3" -> "77 conv2d_4" [label="(16, 16, 1, 1)", style=solid]; +"77 conv2d_4" -> "78 quantize_per_tensor_default_4" [label="(1, 16, 56, 56)", style=solid]; +"78 quantize_per_tensor_default_4" -> "79 dequantize_per_tensor_default_5" [label="(1, 16, 56, 56)", style=solid]; +"79 dequantize_per_tensor_default_5" -> "85 conv2d_5" [label="(1, 16, 56, 56)", style=solid]; +"80 features_2_block_0_0_weight_bias" -> "85 conv2d_5" [label="(72,)", style=solid]; +"81 conv2d_5_scale_0" -> "84 dequantize_per_channel_default_4" [label="(72,)", style=solid]; +"82 conv2d_5_zero_point_0" -> "84 dequantize_per_channel_default_4" [label="(72,)", style=solid]; +"83 _frozen_param3" -> "84 dequantize_per_channel_default_4" [label="(72, 16, 1, 1)", style=solid]; +"84 dequantize_per_channel_default_4" -> "85 conv2d_5" [label="(72, 16, 1, 1)", style=solid]; +"85 conv2d_5" -> "88 relu__1" [label="(1, 72, 56, 56)", style=solid]; +"86 relu__1_scale_0" -> "89 quantize_per_channel_default_5" [label="(72,)", style=solid]; +"86 relu__1_scale_0" -> "90 dequantize_per_channel_default_5" [label="(72,)", style=solid]; +"87 relu__1_zero_point_0" -> "89 quantize_per_channel_default_5" [label="(72,)", style=solid]; +"87 relu__1_zero_point_0" -> "90 dequantize_per_channel_default_5" [label="(72,)", style=solid]; +"88 relu__1" -> "89 quantize_per_channel_default_5" [label="(1, 72, 56, 56)", style=solid]; +"89 quantize_per_channel_default_5" -> "90 dequantize_per_channel_default_5" [label="(1, 72, 56, 56)", style=solid]; +"90 dequantize_per_channel_default_5" -> "96 conv2d_6" [label="(1, 72, 56, 56)", style=solid]; +"91 features_2_block_1_0_weight_bias" -> "96 conv2d_6" [label="(72,)", style=solid]; +"92 conv2d_6_scale_0" -> "95 dequantize_per_channel_default_6" [label="(72,)", style=solid]; +"93 conv2d_6_zero_point_0" -> "95 dequantize_per_channel_default_6" [label="(72,)", style=solid]; +"94 _frozen_param4" -> "95 dequantize_per_channel_default_6" [label="(72, 1, 3, 3)", style=solid]; +"95 dequantize_per_channel_default_6" -> "96 conv2d_6" [label="(72, 1, 3, 3)", style=solid]; +"96 conv2d_6" -> "97 relu__2" [label="(1, 72, 28, 28)", style=solid]; +"97 relu__2" -> "98 quantize_per_tensor_default_5" [label="(1, 72, 28, 28)", style=solid]; +"98 quantize_per_tensor_default_5" -> "99 dequantize_per_tensor_default_6" [label="(1, 72, 28, 28)", style=solid]; +"99 dequantize_per_tensor_default_6" -> "105 conv2d_7" [label="(1, 72, 28, 28)", style=solid]; +"100 features_2_block_2_0_weight_bias" -> "105 conv2d_7" [label="(24,)", style=solid]; +"101 conv2d_7_scale_0" -> "104 dequantize_per_channel_default_7" [label="(24,)", style=solid]; +"102 conv2d_7_zero_point_0" -> "104 dequantize_per_channel_default_7" [label="(24,)", style=solid]; +"103 _frozen_param5" -> "104 dequantize_per_channel_default_7" [label="(24, 72, 1, 1)", style=solid]; +"104 dequantize_per_channel_default_7" -> "105 conv2d_7" [label="(24, 72, 1, 1)", style=solid]; +"105 conv2d_7" -> "106 quantize_per_tensor_default_6" [label="(1, 24, 28, 28)", style=solid]; +"106 quantize_per_tensor_default_6" -> "107 dequantize_per_tensor_default_8" [label="(1, 24, 28, 28)", style=solid]; +"106 quantize_per_tensor_default_6" -> "108 dequantize_per_tensor_default_7" [label="(1, 24, 28, 28)", style=solid]; +"107 dequantize_per_tensor_default_8" -> "137 add_" [label="(1, 24, 28, 28)", style=solid]; +"108 dequantize_per_tensor_default_7" -> "114 conv2d_8" [label="(1, 24, 28, 28)", style=solid]; +"109 features_3_block_0_0_weight_bias" -> "114 conv2d_8" [label="(88,)", style=solid]; +"110 conv2d_8_scale_0" -> "113 dequantize_per_channel_default_8" [label="(88,)", style=solid]; +"111 conv2d_8_zero_point_0" -> "113 dequantize_per_channel_default_8" [label="(88,)", style=solid]; +"112 _frozen_param6" -> "113 dequantize_per_channel_default_8" [label="(88, 24, 1, 1)", style=solid]; +"113 dequantize_per_channel_default_8" -> "114 conv2d_8" [label="(88, 24, 1, 1)", style=solid]; +"114 conv2d_8" -> "117 relu__3" [label="(1, 88, 28, 28)", style=solid]; +"115 relu__3_scale_0" -> "118 quantize_per_channel_default_9" [label="(88,)", style=solid]; +"115 relu__3_scale_0" -> "119 dequantize_per_channel_default_9" [label="(88,)", style=solid]; +"116 relu__3_zero_point_0" -> "118 quantize_per_channel_default_9" [label="(88,)", style=solid]; +"116 relu__3_zero_point_0" -> "119 dequantize_per_channel_default_9" [label="(88,)", style=solid]; +"117 relu__3" -> "118 quantize_per_channel_default_9" [label="(1, 88, 28, 28)", style=solid]; +"118 quantize_per_channel_default_9" -> "119 dequantize_per_channel_default_9" [label="(1, 88, 28, 28)", style=solid]; +"119 dequantize_per_channel_default_9" -> "125 conv2d_9" [label="(1, 88, 28, 28)", style=solid]; +"120 features_3_block_1_0_weight_bias" -> "125 conv2d_9" [label="(88,)", style=solid]; +"121 conv2d_9_scale_0" -> "124 dequantize_per_channel_default_10" [label="(88,)", style=solid]; +"122 conv2d_9_zero_point_0" -> "124 dequantize_per_channel_default_10" [label="(88,)", style=solid]; +"123 _frozen_param7" -> "124 dequantize_per_channel_default_10" [label="(88, 1, 3, 3)", style=solid]; +"124 dequantize_per_channel_default_10" -> "125 conv2d_9" [label="(88, 1, 3, 3)", style=solid]; +"125 conv2d_9" -> "126 relu__4" [label="(1, 88, 28, 28)", style=solid]; +"126 relu__4" -> "127 quantize_per_tensor_default_7" [label="(1, 88, 28, 28)", style=solid]; +"127 quantize_per_tensor_default_7" -> "128 dequantize_per_tensor_default_9" [label="(1, 88, 28, 28)", style=solid]; +"128 dequantize_per_tensor_default_9" -> "134 conv2d_10" [label="(1, 88, 28, 28)", style=solid]; +"129 features_3_block_2_0_weight_bias" -> "134 conv2d_10" [label="(24,)", style=solid]; +"130 conv2d_10_scale_0" -> "133 dequantize_per_channel_default_11" [label="(24,)", style=solid]; +"131 conv2d_10_zero_point_0" -> "133 dequantize_per_channel_default_11" [label="(24,)", style=solid]; +"132 _frozen_param8" -> "133 dequantize_per_channel_default_11" [label="(24, 88, 1, 1)", style=solid]; +"133 dequantize_per_channel_default_11" -> "134 conv2d_10" [label="(24, 88, 1, 1)", style=solid]; +"134 conv2d_10" -> "135 quantize_per_tensor_default_8" [label="(1, 24, 28, 28)", style=solid]; +"135 quantize_per_tensor_default_8" -> "136 dequantize_per_tensor_default_10" [label="(1, 24, 28, 28)", style=solid]; +"136 dequantize_per_tensor_default_10" -> "137 add_" [label="(1, 24, 28, 28)", style=solid]; +"137 add_" -> "138 quantize_per_tensor_default_9" [label="(1, 24, 28, 28)", style=solid]; +"138 quantize_per_tensor_default_9" -> "139 dequantize_per_tensor_default_11" [label="(1, 24, 28, 28)", style=solid]; +"139 dequantize_per_tensor_default_11" -> "145 conv2d_11" [label="(1, 24, 28, 28)", style=solid]; +"140 features_4_block_0_0_weight_bias" -> "145 conv2d_11" [label="(96,)", style=solid]; +"141 conv2d_11_scale_0" -> "144 dequantize_per_channel_default_12" [label="(96,)", style=solid]; +"142 conv2d_11_zero_point_0" -> "144 dequantize_per_channel_default_12" [label="(96,)", style=solid]; +"143 _frozen_param9" -> "144 dequantize_per_channel_default_12" [label="(96, 24, 1, 1)", style=solid]; +"144 dequantize_per_channel_default_12" -> "145 conv2d_11" [label="(96, 24, 1, 1)", style=solid]; +"145 conv2d_11" -> "148 hardswish__1" [label="(1, 96, 28, 28)", style=solid]; +"146 hardswish__1_scale_0" -> "149 quantize_per_channel_default_13" [label="(96,)", style=solid]; +"146 hardswish__1_scale_0" -> "150 dequantize_per_channel_default_13" [label="(96,)", style=solid]; +"147 hardswish__1_zero_point_0" -> "149 quantize_per_channel_default_13" [label="(96,)", style=solid]; +"147 hardswish__1_zero_point_0" -> "150 dequantize_per_channel_default_13" [label="(96,)", style=solid]; +"148 hardswish__1" -> "149 quantize_per_channel_default_13" [label="(1, 96, 28, 28)", style=solid]; +"149 quantize_per_channel_default_13" -> "150 dequantize_per_channel_default_13" [label="(1, 96, 28, 28)", style=solid]; +"150 dequantize_per_channel_default_13" -> "156 conv2d_12" [label="(1, 96, 28, 28)", style=solid]; +"151 features_4_block_1_0_weight_bias" -> "156 conv2d_12" [label="(96,)", style=solid]; +"152 conv2d_12_scale_0" -> "155 dequantize_per_channel_default_14" [label="(96,)", style=solid]; +"153 conv2d_12_zero_point_0" -> "155 dequantize_per_channel_default_14" [label="(96,)", style=solid]; +"154 _frozen_param10" -> "155 dequantize_per_channel_default_14" [label="(96, 1, 5, 5)", style=solid]; +"155 dequantize_per_channel_default_14" -> "156 conv2d_12" [label="(96, 1, 5, 5)", style=solid]; +"156 conv2d_12" -> "157 hardswish__2" [label="(1, 96, 14, 14)", style=solid]; +"157 hardswish__2" -> "158 quantize_per_tensor_default_10" [label="(1, 96, 14, 14)", style=solid]; +"158 quantize_per_tensor_default_10" -> "159 dequantize_per_tensor_default_13" [label="(1, 96, 14, 14)", style=solid]; +"158 quantize_per_tensor_default_10" -> "160 dequantize_per_tensor_default_12" [label="(1, 96, 14, 14)", style=solid]; +"159 dequantize_per_tensor_default_13" -> "168 mul_1" [label="(1, 96, 14, 14)", style=solid]; +"160 dequantize_per_tensor_default_12" -> "161 adaptive_avg_pool2d_1" [label="(1, 96, 14, 14)", style=solid]; +"161 adaptive_avg_pool2d_1" -> "162 conv2d_13" [label="(1, 96, 1, 1)", style=solid]; +"162 conv2d_13" -> "163 relu_1" [label="(1, 24, 1, 1)", style=solid]; +"163 relu_1" -> "164 conv2d_14" [label="(1, 24, 1, 1)", style=solid]; +"164 conv2d_14" -> "165 hardsigmoid_1" [label="(1, 96, 1, 1)", style=solid]; +"165 hardsigmoid_1" -> "166 quantize_per_tensor_default_11" [label="(1, 96, 1, 1)", style=solid]; +"166 quantize_per_tensor_default_11" -> "167 dequantize_per_tensor_default_14" [label="(1, 96, 1, 1)", style=solid]; +"167 dequantize_per_tensor_default_14" -> "168 mul_1" [label="(1, 96, 1, 1)", style=solid]; +"168 mul_1" -> "169 quantize_per_tensor_default_12" [label="(1, 96, 14, 14)", style=solid]; +"169 quantize_per_tensor_default_12" -> "170 dequantize_per_tensor_default_15" [label="(1, 96, 14, 14)", style=solid]; +"170 dequantize_per_tensor_default_15" -> "176 conv2d_15" [label="(1, 96, 14, 14)", style=solid]; +"171 features_4_block_3_0_weight_bias" -> "176 conv2d_15" [label="(40,)", style=solid]; +"172 conv2d_15_scale_0" -> "175 dequantize_per_channel_default_15" [label="(40,)", style=solid]; +"173 conv2d_15_zero_point_0" -> "175 dequantize_per_channel_default_15" [label="(40,)", style=solid]; +"174 _frozen_param11" -> "175 dequantize_per_channel_default_15" [label="(40, 96, 1, 1)", style=solid]; +"175 dequantize_per_channel_default_15" -> "176 conv2d_15" [label="(40, 96, 1, 1)", style=solid]; +"176 conv2d_15" -> "177 quantize_per_tensor_default_13" [label="(1, 40, 14, 14)", style=solid]; +"177 quantize_per_tensor_default_13" -> "178 dequantize_per_tensor_default_17" [label="(1, 40, 14, 14)", style=solid]; +"177 quantize_per_tensor_default_13" -> "179 dequantize_per_tensor_default_16" [label="(1, 40, 14, 14)", style=solid]; +"178 dequantize_per_tensor_default_17" -> "219 add__1" [label="(1, 40, 14, 14)", style=solid]; +"179 dequantize_per_tensor_default_16" -> "185 conv2d_16" [label="(1, 40, 14, 14)", style=solid]; +"180 features_5_block_0_0_weight_bias" -> "185 conv2d_16" [label="(240,)", style=solid]; +"181 conv2d_16_scale_0" -> "184 dequantize_per_channel_default_16" [label="(240,)", style=solid]; +"182 conv2d_16_zero_point_0" -> "184 dequantize_per_channel_default_16" [label="(240,)", style=solid]; +"183 _frozen_param12" -> "184 dequantize_per_channel_default_16" [label="(240, 40, 1, 1)", style=solid]; +"184 dequantize_per_channel_default_16" -> "185 conv2d_16" [label="(240, 40, 1, 1)", style=solid]; +"185 conv2d_16" -> "188 hardswish__3" [label="(1, 240, 14, 14)", style=solid]; +"186 hardswish__3_scale_0" -> "189 quantize_per_channel_default_17" [label="(240,)", style=solid]; +"186 hardswish__3_scale_0" -> "190 dequantize_per_channel_default_17" [label="(240,)", style=solid]; +"187 hardswish__3_zero_point_0" -> "189 quantize_per_channel_default_17" [label="(240,)", style=solid]; +"187 hardswish__3_zero_point_0" -> "190 dequantize_per_channel_default_17" [label="(240,)", style=solid]; +"188 hardswish__3" -> "189 quantize_per_channel_default_17" [label="(1, 240, 14, 14)", style=solid]; +"189 quantize_per_channel_default_17" -> "190 dequantize_per_channel_default_17" [label="(1, 240, 14, 14)", style=solid]; +"190 dequantize_per_channel_default_17" -> "196 conv2d_17" [label="(1, 240, 14, 14)", style=solid]; +"191 features_5_block_1_0_weight_bias" -> "196 conv2d_17" [label="(240,)", style=solid]; +"192 conv2d_17_scale_0" -> "195 dequantize_per_channel_default_18" [label="(240,)", style=solid]; +"193 conv2d_17_zero_point_0" -> "195 dequantize_per_channel_default_18" [label="(240,)", style=solid]; +"194 _frozen_param13" -> "195 dequantize_per_channel_default_18" [label="(240, 1, 5, 5)", style=solid]; +"195 dequantize_per_channel_default_18" -> "196 conv2d_17" [label="(240, 1, 5, 5)", style=solid]; +"196 conv2d_17" -> "197 hardswish__4" [label="(1, 240, 14, 14)", style=solid]; +"197 hardswish__4" -> "198 quantize_per_tensor_default_14" [label="(1, 240, 14, 14)", style=solid]; +"198 quantize_per_tensor_default_14" -> "199 dequantize_per_tensor_default_19" [label="(1, 240, 14, 14)", style=solid]; +"198 quantize_per_tensor_default_14" -> "200 dequantize_per_tensor_default_18" [label="(1, 240, 14, 14)", style=solid]; +"199 dequantize_per_tensor_default_19" -> "208 mul_2" [label="(1, 240, 14, 14)", style=solid]; +"200 dequantize_per_tensor_default_18" -> "201 adaptive_avg_pool2d_2" [label="(1, 240, 14, 14)", style=solid]; +"201 adaptive_avg_pool2d_2" -> "202 conv2d_18" [label="(1, 240, 1, 1)", style=solid]; +"202 conv2d_18" -> "203 relu_2" [label="(1, 64, 1, 1)", style=solid]; +"203 relu_2" -> "204 conv2d_19" [label="(1, 64, 1, 1)", style=solid]; +"204 conv2d_19" -> "205 hardsigmoid_2" [label="(1, 240, 1, 1)", style=solid]; +"205 hardsigmoid_2" -> "206 quantize_per_tensor_default_15" [label="(1, 240, 1, 1)", style=solid]; +"206 quantize_per_tensor_default_15" -> "207 dequantize_per_tensor_default_20" [label="(1, 240, 1, 1)", style=solid]; +"207 dequantize_per_tensor_default_20" -> "208 mul_2" [label="(1, 240, 1, 1)", style=solid]; +"208 mul_2" -> "209 quantize_per_tensor_default_16" [label="(1, 240, 14, 14)", style=solid]; +"209 quantize_per_tensor_default_16" -> "210 dequantize_per_tensor_default_21" [label="(1, 240, 14, 14)", style=solid]; +"210 dequantize_per_tensor_default_21" -> "216 conv2d_20" [label="(1, 240, 14, 14)", style=solid]; +"211 features_5_block_3_0_weight_bias" -> "216 conv2d_20" [label="(40,)", style=solid]; +"212 conv2d_20_scale_0" -> "215 dequantize_per_channel_default_19" [label="(40,)", style=solid]; +"213 conv2d_20_zero_point_0" -> "215 dequantize_per_channel_default_19" [label="(40,)", style=solid]; +"214 _frozen_param14" -> "215 dequantize_per_channel_default_19" [label="(40, 240, 1, 1)", style=solid]; +"215 dequantize_per_channel_default_19" -> "216 conv2d_20" [label="(40, 240, 1, 1)", style=solid]; +"216 conv2d_20" -> "217 quantize_per_tensor_default_17" [label="(1, 40, 14, 14)", style=solid]; +"217 quantize_per_tensor_default_17" -> "218 dequantize_per_tensor_default_22" [label="(1, 40, 14, 14)", style=solid]; +"218 dequantize_per_tensor_default_22" -> "219 add__1" [label="(1, 40, 14, 14)", style=solid]; +"219 add__1" -> "220 quantize_per_tensor_default_18" [label="(1, 40, 14, 14)", style=solid]; +"220 quantize_per_tensor_default_18" -> "221 dequantize_per_tensor_default_24" [label="(1, 40, 14, 14)", style=solid]; +"220 quantize_per_tensor_default_18" -> "222 dequantize_per_tensor_default_23" [label="(1, 40, 14, 14)", style=solid]; +"221 dequantize_per_tensor_default_24" -> "262 add__2" [label="(1, 40, 14, 14)", style=solid]; +"222 dequantize_per_tensor_default_23" -> "228 conv2d_21" [label="(1, 40, 14, 14)", style=solid]; +"223 features_6_block_0_0_weight_bias" -> "228 conv2d_21" [label="(240,)", style=solid]; +"224 conv2d_21_scale_0" -> "227 dequantize_per_channel_default_20" [label="(240,)", style=solid]; +"225 conv2d_21_zero_point_0" -> "227 dequantize_per_channel_default_20" [label="(240,)", style=solid]; +"226 _frozen_param15" -> "227 dequantize_per_channel_default_20" [label="(240, 40, 1, 1)", style=solid]; +"227 dequantize_per_channel_default_20" -> "228 conv2d_21" [label="(240, 40, 1, 1)", style=solid]; +"228 conv2d_21" -> "231 hardswish__5" [label="(1, 240, 14, 14)", style=solid]; +"229 hardswish__5_scale_0" -> "232 quantize_per_channel_default_21" [label="(240,)", style=solid]; +"229 hardswish__5_scale_0" -> "233 dequantize_per_channel_default_21" [label="(240,)", style=solid]; +"230 hardswish__5_zero_point_0" -> "232 quantize_per_channel_default_21" [label="(240,)", style=solid]; +"230 hardswish__5_zero_point_0" -> "233 dequantize_per_channel_default_21" [label="(240,)", style=solid]; +"231 hardswish__5" -> "232 quantize_per_channel_default_21" [label="(1, 240, 14, 14)", style=solid]; +"232 quantize_per_channel_default_21" -> "233 dequantize_per_channel_default_21" [label="(1, 240, 14, 14)", style=solid]; +"233 dequantize_per_channel_default_21" -> "239 conv2d_22" [label="(1, 240, 14, 14)", style=solid]; +"234 features_6_block_1_0_weight_bias" -> "239 conv2d_22" [label="(240,)", style=solid]; +"235 conv2d_22_scale_0" -> "238 dequantize_per_channel_default_22" [label="(240,)", style=solid]; +"236 conv2d_22_zero_point_0" -> "238 dequantize_per_channel_default_22" [label="(240,)", style=solid]; +"237 _frozen_param16" -> "238 dequantize_per_channel_default_22" [label="(240, 1, 5, 5)", style=solid]; +"238 dequantize_per_channel_default_22" -> "239 conv2d_22" [label="(240, 1, 5, 5)", style=solid]; +"239 conv2d_22" -> "240 hardswish__6" [label="(1, 240, 14, 14)", style=solid]; +"240 hardswish__6" -> "241 quantize_per_tensor_default_19" [label="(1, 240, 14, 14)", style=solid]; +"241 quantize_per_tensor_default_19" -> "242 dequantize_per_tensor_default_26" [label="(1, 240, 14, 14)", style=solid]; +"241 quantize_per_tensor_default_19" -> "243 dequantize_per_tensor_default_25" [label="(1, 240, 14, 14)", style=solid]; +"242 dequantize_per_tensor_default_26" -> "251 mul_3" [label="(1, 240, 14, 14)", style=solid]; +"243 dequantize_per_tensor_default_25" -> "244 adaptive_avg_pool2d_3" [label="(1, 240, 14, 14)", style=solid]; +"244 adaptive_avg_pool2d_3" -> "245 conv2d_23" [label="(1, 240, 1, 1)", style=solid]; +"245 conv2d_23" -> "246 relu_3" [label="(1, 64, 1, 1)", style=solid]; +"246 relu_3" -> "247 conv2d_24" [label="(1, 64, 1, 1)", style=solid]; +"247 conv2d_24" -> "248 hardsigmoid_3" [label="(1, 240, 1, 1)", style=solid]; +"248 hardsigmoid_3" -> "249 quantize_per_tensor_default_20" [label="(1, 240, 1, 1)", style=solid]; +"249 quantize_per_tensor_default_20" -> "250 dequantize_per_tensor_default_27" [label="(1, 240, 1, 1)", style=solid]; +"250 dequantize_per_tensor_default_27" -> "251 mul_3" [label="(1, 240, 1, 1)", style=solid]; +"251 mul_3" -> "252 quantize_per_tensor_default_21" [label="(1, 240, 14, 14)", style=solid]; +"252 quantize_per_tensor_default_21" -> "253 dequantize_per_tensor_default_28" [label="(1, 240, 14, 14)", style=solid]; +"253 dequantize_per_tensor_default_28" -> "259 conv2d_25" [label="(1, 240, 14, 14)", style=solid]; +"254 features_6_block_3_0_weight_bias" -> "259 conv2d_25" [label="(40,)", style=solid]; +"255 conv2d_25_scale_0" -> "258 dequantize_per_channel_default_23" [label="(40,)", style=solid]; +"256 conv2d_25_zero_point_0" -> "258 dequantize_per_channel_default_23" [label="(40,)", style=solid]; +"257 _frozen_param17" -> "258 dequantize_per_channel_default_23" [label="(40, 240, 1, 1)", style=solid]; +"258 dequantize_per_channel_default_23" -> "259 conv2d_25" [label="(40, 240, 1, 1)", style=solid]; +"259 conv2d_25" -> "260 quantize_per_tensor_default_22" [label="(1, 40, 14, 14)", style=solid]; +"260 quantize_per_tensor_default_22" -> "261 dequantize_per_tensor_default_29" [label="(1, 40, 14, 14)", style=solid]; +"261 dequantize_per_tensor_default_29" -> "262 add__2" [label="(1, 40, 14, 14)", style=solid]; +"262 add__2" -> "263 quantize_per_tensor_default_23" [label="(1, 40, 14, 14)", style=solid]; +"263 quantize_per_tensor_default_23" -> "264 dequantize_per_tensor_default_30" [label="(1, 40, 14, 14)", style=solid]; +"264 dequantize_per_tensor_default_30" -> "270 conv2d_26" [label="(1, 40, 14, 14)", style=solid]; +"265 features_7_block_0_0_weight_bias" -> "270 conv2d_26" [label="(120,)", style=solid]; +"266 conv2d_26_scale_0" -> "269 dequantize_per_channel_default_24" [label="(120,)", style=solid]; +"267 conv2d_26_zero_point_0" -> "269 dequantize_per_channel_default_24" [label="(120,)", style=solid]; +"268 _frozen_param18" -> "269 dequantize_per_channel_default_24" [label="(120, 40, 1, 1)", style=solid]; +"269 dequantize_per_channel_default_24" -> "270 conv2d_26" [label="(120, 40, 1, 1)", style=solid]; +"270 conv2d_26" -> "273 hardswish__7" [label="(1, 120, 14, 14)", style=solid]; +"271 hardswish__7_scale_0" -> "274 quantize_per_channel_default_25" [label="(120,)", style=solid]; +"271 hardswish__7_scale_0" -> "275 dequantize_per_channel_default_25" [label="(120,)", style=solid]; +"272 hardswish__7_zero_point_0" -> "274 quantize_per_channel_default_25" [label="(120,)", style=solid]; +"272 hardswish__7_zero_point_0" -> "275 dequantize_per_channel_default_25" [label="(120,)", style=solid]; +"273 hardswish__7" -> "274 quantize_per_channel_default_25" [label="(1, 120, 14, 14)", style=solid]; +"274 quantize_per_channel_default_25" -> "275 dequantize_per_channel_default_25" [label="(1, 120, 14, 14)", style=solid]; +"275 dequantize_per_channel_default_25" -> "281 conv2d_27" [label="(1, 120, 14, 14)", style=solid]; +"276 features_7_block_1_0_weight_bias" -> "281 conv2d_27" [label="(120,)", style=solid]; +"277 conv2d_27_scale_0" -> "280 dequantize_per_channel_default_26" [label="(120,)", style=solid]; +"278 conv2d_27_zero_point_0" -> "280 dequantize_per_channel_default_26" [label="(120,)", style=solid]; +"279 _frozen_param19" -> "280 dequantize_per_channel_default_26" [label="(120, 1, 5, 5)", style=solid]; +"280 dequantize_per_channel_default_26" -> "281 conv2d_27" [label="(120, 1, 5, 5)", style=solid]; +"281 conv2d_27" -> "282 hardswish__8" [label="(1, 120, 14, 14)", style=solid]; +"282 hardswish__8" -> "283 quantize_per_tensor_default_24" [label="(1, 120, 14, 14)", style=solid]; +"283 quantize_per_tensor_default_24" -> "284 dequantize_per_tensor_default_32" [label="(1, 120, 14, 14)", style=solid]; +"283 quantize_per_tensor_default_24" -> "285 dequantize_per_tensor_default_31" [label="(1, 120, 14, 14)", style=solid]; +"284 dequantize_per_tensor_default_32" -> "293 mul_4" [label="(1, 120, 14, 14)", style=solid]; +"285 dequantize_per_tensor_default_31" -> "286 adaptive_avg_pool2d_4" [label="(1, 120, 14, 14)", style=solid]; +"286 adaptive_avg_pool2d_4" -> "287 conv2d_28" [label="(1, 120, 1, 1)", style=solid]; +"287 conv2d_28" -> "288 relu_4" [label="(1, 32, 1, 1)", style=solid]; +"288 relu_4" -> "289 conv2d_29" [label="(1, 32, 1, 1)", style=solid]; +"289 conv2d_29" -> "290 hardsigmoid_4" [label="(1, 120, 1, 1)", style=solid]; +"290 hardsigmoid_4" -> "291 quantize_per_tensor_default_25" [label="(1, 120, 1, 1)", style=solid]; +"291 quantize_per_tensor_default_25" -> "292 dequantize_per_tensor_default_33" [label="(1, 120, 1, 1)", style=solid]; +"292 dequantize_per_tensor_default_33" -> "293 mul_4" [label="(1, 120, 1, 1)", style=solid]; +"293 mul_4" -> "294 quantize_per_tensor_default_26" [label="(1, 120, 14, 14)", style=solid]; +"294 quantize_per_tensor_default_26" -> "295 dequantize_per_tensor_default_34" [label="(1, 120, 14, 14)", style=solid]; +"295 dequantize_per_tensor_default_34" -> "301 conv2d_30" [label="(1, 120, 14, 14)", style=solid]; +"296 features_7_block_3_0_weight_bias" -> "301 conv2d_30" [label="(48,)", style=solid]; +"297 conv2d_30_scale_0" -> "300 dequantize_per_channel_default_27" [label="(48,)", style=solid]; +"298 conv2d_30_zero_point_0" -> "300 dequantize_per_channel_default_27" [label="(48,)", style=solid]; +"299 _frozen_param20" -> "300 dequantize_per_channel_default_27" [label="(48, 120, 1, 1)", style=solid]; +"300 dequantize_per_channel_default_27" -> "301 conv2d_30" [label="(48, 120, 1, 1)", style=solid]; +"301 conv2d_30" -> "302 quantize_per_tensor_default_27" [label="(1, 48, 14, 14)", style=solid]; +"302 quantize_per_tensor_default_27" -> "303 dequantize_per_tensor_default_36" [label="(1, 48, 14, 14)", style=solid]; +"302 quantize_per_tensor_default_27" -> "304 dequantize_per_tensor_default_35" [label="(1, 48, 14, 14)", style=solid]; +"303 dequantize_per_tensor_default_36" -> "344 add__3" [label="(1, 48, 14, 14)", style=solid]; +"304 dequantize_per_tensor_default_35" -> "310 conv2d_31" [label="(1, 48, 14, 14)", style=solid]; +"305 features_8_block_0_0_weight_bias" -> "310 conv2d_31" [label="(144,)", style=solid]; +"306 conv2d_31_scale_0" -> "309 dequantize_per_channel_default_28" [label="(144,)", style=solid]; +"307 conv2d_31_zero_point_0" -> "309 dequantize_per_channel_default_28" [label="(144,)", style=solid]; +"308 _frozen_param21" -> "309 dequantize_per_channel_default_28" [label="(144, 48, 1, 1)", style=solid]; +"309 dequantize_per_channel_default_28" -> "310 conv2d_31" [label="(144, 48, 1, 1)", style=solid]; +"310 conv2d_31" -> "313 hardswish__9" [label="(1, 144, 14, 14)", style=solid]; +"311 hardswish__9_scale_0" -> "314 quantize_per_channel_default_29" [label="(144,)", style=solid]; +"311 hardswish__9_scale_0" -> "315 dequantize_per_channel_default_29" [label="(144,)", style=solid]; +"312 hardswish__9_zero_point_0" -> "314 quantize_per_channel_default_29" [label="(144,)", style=solid]; +"312 hardswish__9_zero_point_0" -> "315 dequantize_per_channel_default_29" [label="(144,)", style=solid]; +"313 hardswish__9" -> "314 quantize_per_channel_default_29" [label="(1, 144, 14, 14)", style=solid]; +"314 quantize_per_channel_default_29" -> "315 dequantize_per_channel_default_29" [label="(1, 144, 14, 14)", style=solid]; +"315 dequantize_per_channel_default_29" -> "321 conv2d_32" [label="(1, 144, 14, 14)", style=solid]; +"316 features_8_block_1_0_weight_bias" -> "321 conv2d_32" [label="(144,)", style=solid]; +"317 conv2d_32_scale_0" -> "320 dequantize_per_channel_default_30" [label="(144,)", style=solid]; +"318 conv2d_32_zero_point_0" -> "320 dequantize_per_channel_default_30" [label="(144,)", style=solid]; +"319 _frozen_param22" -> "320 dequantize_per_channel_default_30" [label="(144, 1, 5, 5)", style=solid]; +"320 dequantize_per_channel_default_30" -> "321 conv2d_32" [label="(144, 1, 5, 5)", style=solid]; +"321 conv2d_32" -> "322 hardswish__10" [label="(1, 144, 14, 14)", style=solid]; +"322 hardswish__10" -> "323 quantize_per_tensor_default_28" [label="(1, 144, 14, 14)", style=solid]; +"323 quantize_per_tensor_default_28" -> "324 dequantize_per_tensor_default_38" [label="(1, 144, 14, 14)", style=solid]; +"323 quantize_per_tensor_default_28" -> "325 dequantize_per_tensor_default_37" [label="(1, 144, 14, 14)", style=solid]; +"324 dequantize_per_tensor_default_38" -> "333 mul_5" [label="(1, 144, 14, 14)", style=solid]; +"325 dequantize_per_tensor_default_37" -> "326 adaptive_avg_pool2d_5" [label="(1, 144, 14, 14)", style=solid]; +"326 adaptive_avg_pool2d_5" -> "327 conv2d_33" [label="(1, 144, 1, 1)", style=solid]; +"327 conv2d_33" -> "328 relu_5" [label="(1, 40, 1, 1)", style=solid]; +"328 relu_5" -> "329 conv2d_34" [label="(1, 40, 1, 1)", style=solid]; +"329 conv2d_34" -> "330 hardsigmoid_5" [label="(1, 144, 1, 1)", style=solid]; +"330 hardsigmoid_5" -> "331 quantize_per_tensor_default_29" [label="(1, 144, 1, 1)", style=solid]; +"331 quantize_per_tensor_default_29" -> "332 dequantize_per_tensor_default_39" [label="(1, 144, 1, 1)", style=solid]; +"332 dequantize_per_tensor_default_39" -> "333 mul_5" [label="(1, 144, 1, 1)", style=solid]; +"333 mul_5" -> "334 quantize_per_tensor_default_30" [label="(1, 144, 14, 14)", style=solid]; +"334 quantize_per_tensor_default_30" -> "335 dequantize_per_tensor_default_40" [label="(1, 144, 14, 14)", style=solid]; +"335 dequantize_per_tensor_default_40" -> "341 conv2d_35" [label="(1, 144, 14, 14)", style=solid]; +"336 features_8_block_3_0_weight_bias" -> "341 conv2d_35" [label="(48,)", style=solid]; +"337 conv2d_35_scale_0" -> "340 dequantize_per_channel_default_31" [label="(48,)", style=solid]; +"338 conv2d_35_zero_point_0" -> "340 dequantize_per_channel_default_31" [label="(48,)", style=solid]; +"339 _frozen_param23" -> "340 dequantize_per_channel_default_31" [label="(48, 144, 1, 1)", style=solid]; +"340 dequantize_per_channel_default_31" -> "341 conv2d_35" [label="(48, 144, 1, 1)", style=solid]; +"341 conv2d_35" -> "342 quantize_per_tensor_default_31" [label="(1, 48, 14, 14)", style=solid]; +"342 quantize_per_tensor_default_31" -> "343 dequantize_per_tensor_default_41" [label="(1, 48, 14, 14)", style=solid]; +"343 dequantize_per_tensor_default_41" -> "344 add__3" [label="(1, 48, 14, 14)", style=solid]; +"344 add__3" -> "345 quantize_per_tensor_default_32" [label="(1, 48, 14, 14)", style=solid]; +"345 quantize_per_tensor_default_32" -> "346 dequantize_per_tensor_default_42" [label="(1, 48, 14, 14)", style=solid]; +"346 dequantize_per_tensor_default_42" -> "352 conv2d_36" [label="(1, 48, 14, 14)", style=solid]; +"347 features_9_block_0_0_weight_bias" -> "352 conv2d_36" [label="(288,)", style=solid]; +"348 conv2d_36_scale_0" -> "351 dequantize_per_channel_default_32" [label="(288,)", style=solid]; +"349 conv2d_36_zero_point_0" -> "351 dequantize_per_channel_default_32" [label="(288,)", style=solid]; +"350 _frozen_param24" -> "351 dequantize_per_channel_default_32" [label="(288, 48, 1, 1)", style=solid]; +"351 dequantize_per_channel_default_32" -> "352 conv2d_36" [label="(288, 48, 1, 1)", style=solid]; +"352 conv2d_36" -> "355 hardswish__11" [label="(1, 288, 14, 14)", style=solid]; +"353 hardswish__11_scale_0" -> "356 quantize_per_channel_default_33" [label="(288,)", style=solid]; +"353 hardswish__11_scale_0" -> "357 dequantize_per_channel_default_33" [label="(288,)", style=solid]; +"354 hardswish__11_zero_point_0" -> "356 quantize_per_channel_default_33" [label="(288,)", style=solid]; +"354 hardswish__11_zero_point_0" -> "357 dequantize_per_channel_default_33" [label="(288,)", style=solid]; +"355 hardswish__11" -> "356 quantize_per_channel_default_33" [label="(1, 288, 14, 14)", style=solid]; +"356 quantize_per_channel_default_33" -> "357 dequantize_per_channel_default_33" [label="(1, 288, 14, 14)", style=solid]; +"357 dequantize_per_channel_default_33" -> "363 conv2d_37" [label="(1, 288, 14, 14)", style=solid]; +"358 features_9_block_1_0_weight_bias" -> "363 conv2d_37" [label="(288,)", style=solid]; +"359 conv2d_37_scale_0" -> "362 dequantize_per_channel_default_34" [label="(288,)", style=solid]; +"360 conv2d_37_zero_point_0" -> "362 dequantize_per_channel_default_34" [label="(288,)", style=solid]; +"361 _frozen_param25" -> "362 dequantize_per_channel_default_34" [label="(288, 1, 5, 5)", style=solid]; +"362 dequantize_per_channel_default_34" -> "363 conv2d_37" [label="(288, 1, 5, 5)", style=solid]; +"363 conv2d_37" -> "364 hardswish__12" [label="(1, 288, 7, 7)", style=solid]; +"364 hardswish__12" -> "365 quantize_per_tensor_default_33" [label="(1, 288, 7, 7)", style=solid]; +"365 quantize_per_tensor_default_33" -> "366 dequantize_per_tensor_default_44" [label="(1, 288, 7, 7)", style=solid]; +"365 quantize_per_tensor_default_33" -> "367 dequantize_per_tensor_default_43" [label="(1, 288, 7, 7)", style=solid]; +"366 dequantize_per_tensor_default_44" -> "375 mul_6" [label="(1, 288, 7, 7)", style=solid]; +"367 dequantize_per_tensor_default_43" -> "368 adaptive_avg_pool2d_6" [label="(1, 288, 7, 7)", style=solid]; +"368 adaptive_avg_pool2d_6" -> "369 conv2d_38" [label="(1, 288, 1, 1)", style=solid]; +"369 conv2d_38" -> "370 relu_6" [label="(1, 72, 1, 1)", style=solid]; +"370 relu_6" -> "371 conv2d_39" [label="(1, 72, 1, 1)", style=solid]; +"371 conv2d_39" -> "372 hardsigmoid_6" [label="(1, 288, 1, 1)", style=solid]; +"372 hardsigmoid_6" -> "373 quantize_per_tensor_default_34" [label="(1, 288, 1, 1)", style=solid]; +"373 quantize_per_tensor_default_34" -> "374 dequantize_per_tensor_default_45" [label="(1, 288, 1, 1)", style=solid]; +"374 dequantize_per_tensor_default_45" -> "375 mul_6" [label="(1, 288, 1, 1)", style=solid]; +"375 mul_6" -> "376 quantize_per_tensor_default_35" [label="(1, 288, 7, 7)", style=solid]; +"376 quantize_per_tensor_default_35" -> "377 dequantize_per_tensor_default_46" [label="(1, 288, 7, 7)", style=solid]; +"377 dequantize_per_tensor_default_46" -> "383 conv2d_40" [label="(1, 288, 7, 7)", style=solid]; +"378 features_9_block_3_0_weight_bias" -> "383 conv2d_40" [label="(96,)", style=solid]; +"379 conv2d_40_scale_0" -> "382 dequantize_per_channel_default_35" [label="(96,)", style=solid]; +"380 conv2d_40_zero_point_0" -> "382 dequantize_per_channel_default_35" [label="(96,)", style=solid]; +"381 _frozen_param26" -> "382 dequantize_per_channel_default_35" [label="(96, 288, 1, 1)", style=solid]; +"382 dequantize_per_channel_default_35" -> "383 conv2d_40" [label="(96, 288, 1, 1)", style=solid]; +"383 conv2d_40" -> "384 quantize_per_tensor_default_36" [label="(1, 96, 7, 7)", style=solid]; +"384 quantize_per_tensor_default_36" -> "385 dequantize_per_tensor_default_48" [label="(1, 96, 7, 7)", style=solid]; +"384 quantize_per_tensor_default_36" -> "386 dequantize_per_tensor_default_47" [label="(1, 96, 7, 7)", style=solid]; +"385 dequantize_per_tensor_default_48" -> "426 add__4" [label="(1, 96, 7, 7)", style=solid]; +"386 dequantize_per_tensor_default_47" -> "392 conv2d_41" [label="(1, 96, 7, 7)", style=solid]; +"387 features_10_block_0_0_weight_bias" -> "392 conv2d_41" [label="(576,)", style=solid]; +"388 conv2d_41_scale_0" -> "391 dequantize_per_channel_default_36" [label="(576,)", style=solid]; +"389 conv2d_41_zero_point_0" -> "391 dequantize_per_channel_default_36" [label="(576,)", style=solid]; +"390 _frozen_param27" -> "391 dequantize_per_channel_default_36" [label="(576, 96, 1, 1)", style=solid]; +"391 dequantize_per_channel_default_36" -> "392 conv2d_41" [label="(576, 96, 1, 1)", style=solid]; +"392 conv2d_41" -> "395 hardswish__13" [label="(1, 576, 7, 7)", style=solid]; +"393 hardswish__13_scale_0" -> "396 quantize_per_channel_default_37" [label="(576,)", style=solid]; +"393 hardswish__13_scale_0" -> "397 dequantize_per_channel_default_37" [label="(576,)", style=solid]; +"394 hardswish__13_zero_point_0" -> "396 quantize_per_channel_default_37" [label="(576,)", style=solid]; +"394 hardswish__13_zero_point_0" -> "397 dequantize_per_channel_default_37" [label="(576,)", style=solid]; +"395 hardswish__13" -> "396 quantize_per_channel_default_37" [label="(1, 576, 7, 7)", style=solid]; +"396 quantize_per_channel_default_37" -> "397 dequantize_per_channel_default_37" [label="(1, 576, 7, 7)", style=solid]; +"397 dequantize_per_channel_default_37" -> "403 conv2d_42" [label="(1, 576, 7, 7)", style=solid]; +"398 features_10_block_1_0_weight_bias" -> "403 conv2d_42" [label="(576,)", style=solid]; +"399 conv2d_42_scale_0" -> "402 dequantize_per_channel_default_38" [label="(576,)", style=solid]; +"400 conv2d_42_zero_point_0" -> "402 dequantize_per_channel_default_38" [label="(576,)", style=solid]; +"401 _frozen_param28" -> "402 dequantize_per_channel_default_38" [label="(576, 1, 5, 5)", style=solid]; +"402 dequantize_per_channel_default_38" -> "403 conv2d_42" [label="(576, 1, 5, 5)", style=solid]; +"403 conv2d_42" -> "404 hardswish__14" [label="(1, 576, 7, 7)", style=solid]; +"404 hardswish__14" -> "405 quantize_per_tensor_default_37" [label="(1, 576, 7, 7)", style=solid]; +"405 quantize_per_tensor_default_37" -> "406 dequantize_per_tensor_default_50" [label="(1, 576, 7, 7)", style=solid]; +"405 quantize_per_tensor_default_37" -> "407 dequantize_per_tensor_default_49" [label="(1, 576, 7, 7)", style=solid]; +"406 dequantize_per_tensor_default_50" -> "415 mul_7" [label="(1, 576, 7, 7)", style=solid]; +"407 dequantize_per_tensor_default_49" -> "408 adaptive_avg_pool2d_7" [label="(1, 576, 7, 7)", style=solid]; +"408 adaptive_avg_pool2d_7" -> "409 conv2d_43" [label="(1, 576, 1, 1)", style=solid]; +"409 conv2d_43" -> "410 relu_7" [label="(1, 144, 1, 1)", style=solid]; +"410 relu_7" -> "411 conv2d_44" [label="(1, 144, 1, 1)", style=solid]; +"411 conv2d_44" -> "412 hardsigmoid_7" [label="(1, 576, 1, 1)", style=solid]; +"412 hardsigmoid_7" -> "413 quantize_per_tensor_default_38" [label="(1, 576, 1, 1)", style=solid]; +"413 quantize_per_tensor_default_38" -> "414 dequantize_per_tensor_default_51" [label="(1, 576, 1, 1)", style=solid]; +"414 dequantize_per_tensor_default_51" -> "415 mul_7" [label="(1, 576, 1, 1)", style=solid]; +"415 mul_7" -> "416 quantize_per_tensor_default_39" [label="(1, 576, 7, 7)", style=solid]; +"416 quantize_per_tensor_default_39" -> "417 dequantize_per_tensor_default_52" [label="(1, 576, 7, 7)", style=solid]; +"417 dequantize_per_tensor_default_52" -> "423 conv2d_45" [label="(1, 576, 7, 7)", style=solid]; +"418 features_10_block_3_0_weight_bias" -> "423 conv2d_45" [label="(96,)", style=solid]; +"419 conv2d_45_scale_0" -> "422 dequantize_per_channel_default_39" [label="(96,)", style=solid]; +"420 conv2d_45_zero_point_0" -> "422 dequantize_per_channel_default_39" [label="(96,)", style=solid]; +"421 _frozen_param29" -> "422 dequantize_per_channel_default_39" [label="(96, 576, 1, 1)", style=solid]; +"422 dequantize_per_channel_default_39" -> "423 conv2d_45" [label="(96, 576, 1, 1)", style=solid]; +"423 conv2d_45" -> "424 quantize_per_tensor_default_40" [label="(1, 96, 7, 7)", style=solid]; +"424 quantize_per_tensor_default_40" -> "425 dequantize_per_tensor_default_53" [label="(1, 96, 7, 7)", style=solid]; +"425 dequantize_per_tensor_default_53" -> "426 add__4" [label="(1, 96, 7, 7)", style=solid]; +"426 add__4" -> "427 quantize_per_tensor_default_41" [label="(1, 96, 7, 7)", style=solid]; +"427 quantize_per_tensor_default_41" -> "428 dequantize_per_tensor_default_55" [label="(1, 96, 7, 7)", style=solid]; +"427 quantize_per_tensor_default_41" -> "429 dequantize_per_tensor_default_54" [label="(1, 96, 7, 7)", style=solid]; +"428 dequantize_per_tensor_default_55" -> "469 add__5" [label="(1, 96, 7, 7)", style=solid]; +"429 dequantize_per_tensor_default_54" -> "435 conv2d_46" [label="(1, 96, 7, 7)", style=solid]; +"430 features_11_block_0_0_weight_bias" -> "435 conv2d_46" [label="(576,)", style=solid]; +"431 conv2d_46_scale_0" -> "434 dequantize_per_channel_default_40" [label="(576,)", style=solid]; +"432 conv2d_46_zero_point_0" -> "434 dequantize_per_channel_default_40" [label="(576,)", style=solid]; +"433 _frozen_param30" -> "434 dequantize_per_channel_default_40" [label="(576, 96, 1, 1)", style=solid]; +"434 dequantize_per_channel_default_40" -> "435 conv2d_46" [label="(576, 96, 1, 1)", style=solid]; +"435 conv2d_46" -> "438 hardswish__15" [label="(1, 576, 7, 7)", style=solid]; +"436 hardswish__15_scale_0" -> "439 quantize_per_channel_default_41" [label="(576,)", style=solid]; +"436 hardswish__15_scale_0" -> "440 dequantize_per_channel_default_41" [label="(576,)", style=solid]; +"437 hardswish__15_zero_point_0" -> "439 quantize_per_channel_default_41" [label="(576,)", style=solid]; +"437 hardswish__15_zero_point_0" -> "440 dequantize_per_channel_default_41" [label="(576,)", style=solid]; +"438 hardswish__15" -> "439 quantize_per_channel_default_41" [label="(1, 576, 7, 7)", style=solid]; +"439 quantize_per_channel_default_41" -> "440 dequantize_per_channel_default_41" [label="(1, 576, 7, 7)", style=solid]; +"440 dequantize_per_channel_default_41" -> "446 conv2d_47" [label="(1, 576, 7, 7)", style=solid]; +"441 features_11_block_1_0_weight_bias" -> "446 conv2d_47" [label="(576,)", style=solid]; +"442 conv2d_47_scale_0" -> "445 dequantize_per_channel_default_42" [label="(576,)", style=solid]; +"443 conv2d_47_zero_point_0" -> "445 dequantize_per_channel_default_42" [label="(576,)", style=solid]; +"444 _frozen_param31" -> "445 dequantize_per_channel_default_42" [label="(576, 1, 5, 5)", style=solid]; +"445 dequantize_per_channel_default_42" -> "446 conv2d_47" [label="(576, 1, 5, 5)", style=solid]; +"446 conv2d_47" -> "447 hardswish__16" [label="(1, 576, 7, 7)", style=solid]; +"447 hardswish__16" -> "448 quantize_per_tensor_default_42" [label="(1, 576, 7, 7)", style=solid]; +"448 quantize_per_tensor_default_42" -> "449 dequantize_per_tensor_default_57" [label="(1, 576, 7, 7)", style=solid]; +"448 quantize_per_tensor_default_42" -> "450 dequantize_per_tensor_default_56" [label="(1, 576, 7, 7)", style=solid]; +"449 dequantize_per_tensor_default_57" -> "458 mul_8" [label="(1, 576, 7, 7)", style=solid]; +"450 dequantize_per_tensor_default_56" -> "451 adaptive_avg_pool2d_8" [label="(1, 576, 7, 7)", style=solid]; +"451 adaptive_avg_pool2d_8" -> "452 conv2d_48" [label="(1, 576, 1, 1)", style=solid]; +"452 conv2d_48" -> "453 relu_8" [label="(1, 144, 1, 1)", style=solid]; +"453 relu_8" -> "454 conv2d_49" [label="(1, 144, 1, 1)", style=solid]; +"454 conv2d_49" -> "455 hardsigmoid_8" [label="(1, 576, 1, 1)", style=solid]; +"455 hardsigmoid_8" -> "456 quantize_per_tensor_default_43" [label="(1, 576, 1, 1)", style=solid]; +"456 quantize_per_tensor_default_43" -> "457 dequantize_per_tensor_default_58" [label="(1, 576, 1, 1)", style=solid]; +"457 dequantize_per_tensor_default_58" -> "458 mul_8" [label="(1, 576, 1, 1)", style=solid]; +"458 mul_8" -> "459 quantize_per_tensor_default_44" [label="(1, 576, 7, 7)", style=solid]; +"459 quantize_per_tensor_default_44" -> "460 dequantize_per_tensor_default_59" [label="(1, 576, 7, 7)", style=solid]; +"460 dequantize_per_tensor_default_59" -> "466 conv2d_50" [label="(1, 576, 7, 7)", style=solid]; +"461 features_11_block_3_0_weight_bias" -> "466 conv2d_50" [label="(96,)", style=solid]; +"462 conv2d_50_scale_0" -> "465 dequantize_per_channel_default_43" [label="(96,)", style=solid]; +"463 conv2d_50_zero_point_0" -> "465 dequantize_per_channel_default_43" [label="(96,)", style=solid]; +"464 _frozen_param32" -> "465 dequantize_per_channel_default_43" [label="(96, 576, 1, 1)", style=solid]; +"465 dequantize_per_channel_default_43" -> "466 conv2d_50" [label="(96, 576, 1, 1)", style=solid]; +"466 conv2d_50" -> "467 quantize_per_tensor_default_45" [label="(1, 96, 7, 7)", style=solid]; +"467 quantize_per_tensor_default_45" -> "468 dequantize_per_tensor_default_60" [label="(1, 96, 7, 7)", style=solid]; +"468 dequantize_per_tensor_default_60" -> "469 add__5" [label="(1, 96, 7, 7)", style=solid]; +"469 add__5" -> "470 quantize_per_tensor_default_46" [label="(1, 96, 7, 7)", style=solid]; +"470 quantize_per_tensor_default_46" -> "471 dequantize_per_tensor_default_61" [label="(1, 96, 7, 7)", style=solid]; +"471 dequantize_per_tensor_default_61" -> "477 conv2d_51" [label="(1, 96, 7, 7)", style=solid]; +"472 features_12_0_weight_bias" -> "477 conv2d_51" [label="(576,)", style=solid]; +"473 conv2d_51_scale_0" -> "476 dequantize_per_channel_default_44" [label="(576,)", style=solid]; +"474 conv2d_51_zero_point_0" -> "476 dequantize_per_channel_default_44" [label="(576,)", style=solid]; +"475 _frozen_param33" -> "476 dequantize_per_channel_default_44" [label="(576, 96, 1, 1)", style=solid]; +"476 dequantize_per_channel_default_44" -> "477 conv2d_51" [label="(576, 96, 1, 1)", style=solid]; +"477 conv2d_51" -> "478 hardswish__17" [label="(1, 576, 7, 7)", style=solid]; +"478 hardswish__17" -> "479 quantize_per_tensor_default_47" [label="(1, 576, 7, 7)", style=solid]; +"479 quantize_per_tensor_default_47" -> "480 dequantize_per_tensor_default_62" [label="(1, 576, 7, 7)", style=solid]; +"480 dequantize_per_tensor_default_62" -> "481 adaptive_avg_pool2d_9" [label="(1, 576, 7, 7)", style=solid]; +"481 adaptive_avg_pool2d_9" -> "482 flatten" [label="(1, 576, 1, 1)", style=solid]; +"482 flatten" -> "483 flatten_0_0_nncf_smooth_quant_0" [label="(1, 576)", style=solid]; +"483 flatten_0_0_nncf_smooth_quant_0" -> "484 quantize_per_tensor_default_48" [label="(1, 576)", style=solid]; +"484 quantize_per_tensor_default_48" -> "485 dequantize_per_tensor_default_63" [label="(1, 576)", style=solid]; +"485 dequantize_per_tensor_default_63" -> "490 linear" [label="(1, 576)", style=solid]; +"486 linear_scale_0" -> "489 dequantize_per_channel_default_45" [label="(1024,)", style=solid]; +"487 linear_zero_point_0" -> "489 dequantize_per_channel_default_45" [label="(1024,)", style=solid]; +"488 _frozen_param34" -> "489 dequantize_per_channel_default_45" [label="(1024, 576)", style=solid]; +"489 dequantize_per_channel_default_45" -> "490 linear" [label="(1024, 576)", style=solid]; +"490 linear" -> "491 hardswish__18" [label="(1, 1024)", style=solid]; +"491 hardswish__18" -> "492 dropout_" [label="(1, 1024)", style=solid]; +"492 dropout_" -> "493 dropout__0_0_nncf_smooth_quant_0" [label="(1, 1024)", style=solid]; +"493 dropout__0_0_nncf_smooth_quant_0" -> "494 quantize_per_tensor_default_49" [label="(1, 1024)", style=solid]; +"494 quantize_per_tensor_default_49" -> "495 dequantize_per_tensor_default_64" [label="(1, 1024)", style=solid]; +"495 dequantize_per_tensor_default_64" -> "500 linear_1" [label="(1, 1024)", style=solid]; +"496 linear_1_scale_0" -> "499 dequantize_per_channel_default_46" [label="(1000,)", style=solid]; +"497 linear_1_zero_point_0" -> "499 dequantize_per_channel_default_46" [label="(1000,)", style=solid]; +"498 _frozen_param35" -> "499 dequantize_per_channel_default_46" [label="(1000, 1024)", style=solid]; +"499 dequantize_per_channel_default_46" -> "500 linear_1" [label="(1000, 1024)", style=solid]; +"500 linear_1" -> "501 output" [label="(1, 1000)", style=solid]; +} diff --git a/tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/resnet18.dot b/tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/resnet18.dot new file mode 100644 index 00000000000..ec542cff909 --- /dev/null +++ b/tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/resnet18.dot @@ -0,0 +1,457 @@ +strict digraph { +"0 fc_bias" [id=0, type=get_attr]; +"1 x" [id=1, type=input]; +"2 quantize_per_tensor_default" [id=2, type=quantize_per_tensor]; +"3 dequantize_per_tensor_default" [id=3, type=dequantize_per_tensor]; +"4 conv1_weight_bias" [id=4, type=get_attr]; +"5 conv2d_scale_0" [id=5, type=get_attr]; +"6 conv2d_zero_point_0" [id=6, type=get_attr]; +"7 _frozen_param0" [id=7, type=get_attr]; +"8 dequantize_per_channel_default" [id=8, type=dequantize_per_channel]; +"9 conv2d" [id=9, type=conv2d]; +"10 relu_" [id=10, type=relu_]; +"11 quantize_per_tensor_default_1" [id=11, type=quantize_per_tensor]; +"12 dequantize_per_tensor_default_1" [id=12, type=dequantize_per_tensor]; +"13 max_pool2d" [id=13, type=max_pool2d]; +"14 layer1_0_conv1_weight_bias" [id=14, type=get_attr]; +"15 conv2d_1_scale_0" [id=15, type=get_attr]; +"16 conv2d_1_zero_point_0" [id=16, type=get_attr]; +"17 _frozen_param1" [id=17, type=get_attr]; +"18 dequantize_per_channel_default_1" [id=18, type=dequantize_per_channel]; +"19 conv2d_1" [id=19, type=conv2d]; +"20 relu__1" [id=20, type=relu_]; +"21 quantize_per_tensor_default_2" [id=21, type=quantize_per_tensor]; +"22 dequantize_per_tensor_default_2" [id=22, type=dequantize_per_tensor]; +"23 layer1_0_conv2_weight_bias" [id=23, type=get_attr]; +"24 conv2d_2_scale_0" [id=24, type=get_attr]; +"25 conv2d_2_zero_point_0" [id=25, type=get_attr]; +"26 _frozen_param2" [id=26, type=get_attr]; +"27 dequantize_per_channel_default_2" [id=27, type=dequantize_per_channel]; +"28 conv2d_2" [id=28, type=conv2d]; +"29 quantize_per_tensor_default_3" [id=29, type=quantize_per_tensor]; +"30 dequantize_per_tensor_default_3" [id=30, type=dequantize_per_tensor]; +"31 add_" [id=31, type=add_]; +"32 relu__2" [id=32, type=relu_]; +"33 quantize_per_tensor_default_4" [id=33, type=quantize_per_tensor]; +"34 dequantize_per_tensor_default_5" [id=34, type=dequantize_per_tensor]; +"35 dequantize_per_tensor_default_4" [id=35, type=dequantize_per_tensor]; +"36 layer1_1_conv1_weight_bias" [id=36, type=get_attr]; +"37 conv2d_3_scale_0" [id=37, type=get_attr]; +"38 conv2d_3_zero_point_0" [id=38, type=get_attr]; +"39 _frozen_param3" [id=39, type=get_attr]; +"40 dequantize_per_channel_default_3" [id=40, type=dequantize_per_channel]; +"41 conv2d_3" [id=41, type=conv2d]; +"42 relu__3" [id=42, type=relu_]; +"43 quantize_per_tensor_default_5" [id=43, type=quantize_per_tensor]; +"44 dequantize_per_tensor_default_6" [id=44, type=dequantize_per_tensor]; +"45 layer1_1_conv2_weight_bias" [id=45, type=get_attr]; +"46 conv2d_4_scale_0" [id=46, type=get_attr]; +"47 conv2d_4_zero_point_0" [id=47, type=get_attr]; +"48 _frozen_param4" [id=48, type=get_attr]; +"49 dequantize_per_channel_default_4" [id=49, type=dequantize_per_channel]; +"50 conv2d_4" [id=50, type=conv2d]; +"51 quantize_per_tensor_default_6" [id=51, type=quantize_per_tensor]; +"52 dequantize_per_tensor_default_7" [id=52, type=dequantize_per_tensor]; +"53 add__1" [id=53, type=add_]; +"54 relu__4" [id=54, type=relu_]; +"55 quantize_per_tensor_default_7" [id=55, type=quantize_per_tensor]; +"56 dequantize_per_tensor_default_9" [id=56, type=dequantize_per_tensor]; +"57 dequantize_per_tensor_default_8" [id=57, type=dequantize_per_tensor]; +"58 layer2_0_conv1_weight_bias" [id=58, type=get_attr]; +"59 conv2d_5_scale_0" [id=59, type=get_attr]; +"60 conv2d_5_zero_point_0" [id=60, type=get_attr]; +"61 _frozen_param5" [id=61, type=get_attr]; +"62 dequantize_per_channel_default_5" [id=62, type=dequantize_per_channel]; +"63 conv2d_5" [id=63, type=conv2d]; +"64 relu__5" [id=64, type=relu_]; +"65 quantize_per_tensor_default_8" [id=65, type=quantize_per_tensor]; +"66 dequantize_per_tensor_default_10" [id=66, type=dequantize_per_tensor]; +"67 layer2_0_conv2_weight_bias" [id=67, type=get_attr]; +"68 conv2d_6_scale_0" [id=68, type=get_attr]; +"69 conv2d_6_zero_point_0" [id=69, type=get_attr]; +"70 _frozen_param6" [id=70, type=get_attr]; +"71 dequantize_per_channel_default_6" [id=71, type=dequantize_per_channel]; +"72 conv2d_6" [id=72, type=conv2d]; +"73 quantize_per_tensor_default_9" [id=73, type=quantize_per_tensor]; +"74 dequantize_per_tensor_default_11" [id=74, type=dequantize_per_tensor]; +"75 layer2_0_downsample_0_weight_bias" [id=75, type=get_attr]; +"76 conv2d_7_scale_0" [id=76, type=get_attr]; +"77 conv2d_7_zero_point_0" [id=77, type=get_attr]; +"78 _frozen_param7" [id=78, type=get_attr]; +"79 dequantize_per_channel_default_7" [id=79, type=dequantize_per_channel]; +"80 conv2d_7" [id=80, type=conv2d]; +"81 quantize_per_tensor_default_10" [id=81, type=quantize_per_tensor]; +"82 dequantize_per_tensor_default_12" [id=82, type=dequantize_per_tensor]; +"83 add__2" [id=83, type=add_]; +"84 relu__6" [id=84, type=relu_]; +"85 quantize_per_tensor_default_11" [id=85, type=quantize_per_tensor]; +"86 dequantize_per_tensor_default_14" [id=86, type=dequantize_per_tensor]; +"87 dequantize_per_tensor_default_13" [id=87, type=dequantize_per_tensor]; +"88 layer2_1_conv1_weight_bias" [id=88, type=get_attr]; +"89 conv2d_8_scale_0" [id=89, type=get_attr]; +"90 conv2d_8_zero_point_0" [id=90, type=get_attr]; +"91 _frozen_param8" [id=91, type=get_attr]; +"92 dequantize_per_channel_default_8" [id=92, type=dequantize_per_channel]; +"93 conv2d_8" [id=93, type=conv2d]; +"94 relu__7" [id=94, type=relu_]; +"95 quantize_per_tensor_default_12" [id=95, type=quantize_per_tensor]; +"96 dequantize_per_tensor_default_15" [id=96, type=dequantize_per_tensor]; +"97 layer2_1_conv2_weight_bias" [id=97, type=get_attr]; +"98 conv2d_9_scale_0" [id=98, type=get_attr]; +"99 conv2d_9_zero_point_0" [id=99, type=get_attr]; +"100 _frozen_param9" [id=100, type=get_attr]; +"101 dequantize_per_channel_default_9" [id=101, type=dequantize_per_channel]; +"102 conv2d_9" [id=102, type=conv2d]; +"103 quantize_per_tensor_default_13" [id=103, type=quantize_per_tensor]; +"104 dequantize_per_tensor_default_16" [id=104, type=dequantize_per_tensor]; +"105 add__3" [id=105, type=add_]; +"106 relu__8" [id=106, type=relu_]; +"107 quantize_per_tensor_default_14" [id=107, type=quantize_per_tensor]; +"108 dequantize_per_tensor_default_18" [id=108, type=dequantize_per_tensor]; +"109 dequantize_per_tensor_default_17" [id=109, type=dequantize_per_tensor]; +"110 layer3_0_conv1_weight_bias" [id=110, type=get_attr]; +"111 conv2d_10_scale_0" [id=111, type=get_attr]; +"112 conv2d_10_zero_point_0" [id=112, type=get_attr]; +"113 _frozen_param10" [id=113, type=get_attr]; +"114 dequantize_per_channel_default_10" [id=114, type=dequantize_per_channel]; +"115 conv2d_10" [id=115, type=conv2d]; +"116 relu__9" [id=116, type=relu_]; +"117 quantize_per_tensor_default_15" [id=117, type=quantize_per_tensor]; +"118 dequantize_per_tensor_default_19" [id=118, type=dequantize_per_tensor]; +"119 layer3_0_conv2_weight_bias" [id=119, type=get_attr]; +"120 conv2d_11_scale_0" [id=120, type=get_attr]; +"121 conv2d_11_zero_point_0" [id=121, type=get_attr]; +"122 _frozen_param11" [id=122, type=get_attr]; +"123 dequantize_per_channel_default_11" [id=123, type=dequantize_per_channel]; +"124 conv2d_11" [id=124, type=conv2d]; +"125 quantize_per_tensor_default_16" [id=125, type=quantize_per_tensor]; +"126 dequantize_per_tensor_default_20" [id=126, type=dequantize_per_tensor]; +"127 layer3_0_downsample_0_weight_bias" [id=127, type=get_attr]; +"128 conv2d_12_scale_0" [id=128, type=get_attr]; +"129 conv2d_12_zero_point_0" [id=129, type=get_attr]; +"130 _frozen_param12" [id=130, type=get_attr]; +"131 dequantize_per_channel_default_12" [id=131, type=dequantize_per_channel]; +"132 conv2d_12" [id=132, type=conv2d]; +"133 quantize_per_tensor_default_17" [id=133, type=quantize_per_tensor]; +"134 dequantize_per_tensor_default_21" [id=134, type=dequantize_per_tensor]; +"135 add__4" [id=135, type=add_]; +"136 relu__10" [id=136, type=relu_]; +"137 quantize_per_tensor_default_18" [id=137, type=quantize_per_tensor]; +"138 dequantize_per_tensor_default_23" [id=138, type=dequantize_per_tensor]; +"139 dequantize_per_tensor_default_22" [id=139, type=dequantize_per_tensor]; +"140 layer3_1_conv1_weight_bias" [id=140, type=get_attr]; +"141 conv2d_13_scale_0" [id=141, type=get_attr]; +"142 conv2d_13_zero_point_0" [id=142, type=get_attr]; +"143 _frozen_param13" [id=143, type=get_attr]; +"144 dequantize_per_channel_default_13" [id=144, type=dequantize_per_channel]; +"145 conv2d_13" [id=145, type=conv2d]; +"146 relu__11" [id=146, type=relu_]; +"147 quantize_per_tensor_default_19" [id=147, type=quantize_per_tensor]; +"148 dequantize_per_tensor_default_24" [id=148, type=dequantize_per_tensor]; +"149 layer3_1_conv2_weight_bias" [id=149, type=get_attr]; +"150 conv2d_14_scale_0" [id=150, type=get_attr]; +"151 conv2d_14_zero_point_0" [id=151, type=get_attr]; +"152 _frozen_param14" [id=152, type=get_attr]; +"153 dequantize_per_channel_default_14" [id=153, type=dequantize_per_channel]; +"154 conv2d_14" [id=154, type=conv2d]; +"155 quantize_per_tensor_default_20" [id=155, type=quantize_per_tensor]; +"156 dequantize_per_tensor_default_25" [id=156, type=dequantize_per_tensor]; +"157 add__5" [id=157, type=add_]; +"158 relu__12" [id=158, type=relu_]; +"159 quantize_per_tensor_default_21" [id=159, type=quantize_per_tensor]; +"160 dequantize_per_tensor_default_27" [id=160, type=dequantize_per_tensor]; +"161 dequantize_per_tensor_default_26" [id=161, type=dequantize_per_tensor]; +"162 layer4_0_conv1_weight_bias" [id=162, type=get_attr]; +"163 conv2d_15_scale_0" [id=163, type=get_attr]; +"164 conv2d_15_zero_point_0" [id=164, type=get_attr]; +"165 _frozen_param15" [id=165, type=get_attr]; +"166 dequantize_per_channel_default_15" [id=166, type=dequantize_per_channel]; +"167 conv2d_15" [id=167, type=conv2d]; +"168 relu__13" [id=168, type=relu_]; +"169 quantize_per_tensor_default_22" [id=169, type=quantize_per_tensor]; +"170 dequantize_per_tensor_default_28" [id=170, type=dequantize_per_tensor]; +"171 layer4_0_conv2_weight_bias" [id=171, type=get_attr]; +"172 conv2d_16_scale_0" [id=172, type=get_attr]; +"173 conv2d_16_zero_point_0" [id=173, type=get_attr]; +"174 _frozen_param16" [id=174, type=get_attr]; +"175 dequantize_per_channel_default_16" [id=175, type=dequantize_per_channel]; +"176 conv2d_16" [id=176, type=conv2d]; +"177 quantize_per_tensor_default_23" [id=177, type=quantize_per_tensor]; +"178 dequantize_per_tensor_default_29" [id=178, type=dequantize_per_tensor]; +"179 layer4_0_downsample_0_weight_bias" [id=179, type=get_attr]; +"180 conv2d_17_scale_0" [id=180, type=get_attr]; +"181 conv2d_17_zero_point_0" [id=181, type=get_attr]; +"182 _frozen_param17" [id=182, type=get_attr]; +"183 dequantize_per_channel_default_17" [id=183, type=dequantize_per_channel]; +"184 conv2d_17" [id=184, type=conv2d]; +"185 quantize_per_tensor_default_24" [id=185, type=quantize_per_tensor]; +"186 dequantize_per_tensor_default_30" [id=186, type=dequantize_per_tensor]; +"187 add__6" [id=187, type=add_]; +"188 relu__14" [id=188, type=relu_]; +"189 quantize_per_tensor_default_25" [id=189, type=quantize_per_tensor]; +"190 dequantize_per_tensor_default_32" [id=190, type=dequantize_per_tensor]; +"191 dequantize_per_tensor_default_31" [id=191, type=dequantize_per_tensor]; +"192 layer4_1_conv1_weight_bias" [id=192, type=get_attr]; +"193 conv2d_18_scale_0" [id=193, type=get_attr]; +"194 conv2d_18_zero_point_0" [id=194, type=get_attr]; +"195 _frozen_param18" [id=195, type=get_attr]; +"196 dequantize_per_channel_default_18" [id=196, type=dequantize_per_channel]; +"197 conv2d_18" [id=197, type=conv2d]; +"198 relu__15" [id=198, type=relu_]; +"199 quantize_per_tensor_default_26" [id=199, type=quantize_per_tensor]; +"200 dequantize_per_tensor_default_33" [id=200, type=dequantize_per_tensor]; +"201 layer4_1_conv2_weight_bias" [id=201, type=get_attr]; +"202 conv2d_19_scale_0" [id=202, type=get_attr]; +"203 conv2d_19_zero_point_0" [id=203, type=get_attr]; +"204 _frozen_param19" [id=204, type=get_attr]; +"205 dequantize_per_channel_default_19" [id=205, type=dequantize_per_channel]; +"206 conv2d_19" [id=206, type=conv2d]; +"207 quantize_per_tensor_default_27" [id=207, type=quantize_per_tensor]; +"208 dequantize_per_tensor_default_34" [id=208, type=dequantize_per_tensor]; +"209 add__7" [id=209, type=add_]; +"210 relu__16" [id=210, type=relu_]; +"211 quantize_per_tensor_default_28" [id=211, type=quantize_per_tensor]; +"212 dequantize_per_tensor_default_35" [id=212, type=dequantize_per_tensor]; +"213 adaptive_avg_pool2d" [id=213, type=adaptive_avg_pool2d]; +"214 flatten" [id=214, type=flatten]; +"215 flatten_0_0_nncf_smooth_quant_0" [id=215, type=call_module]; +"216 quantize_per_tensor_default_29" [id=216, type=quantize_per_tensor]; +"217 dequantize_per_tensor_default_36" [id=217, type=dequantize_per_tensor]; +"218 linear_scale_0" [id=218, type=get_attr]; +"219 linear_zero_point_0" [id=219, type=get_attr]; +"220 _frozen_param20" [id=220, type=get_attr]; +"221 dequantize_per_channel_default_20" [id=221, type=dequantize_per_channel]; +"222 linear" [id=222, type=linear]; +"223 output" [id=223, type=output]; +"0 fc_bias" -> "222 linear" [label="(1000,)", style=solid]; +"1 x" -> "2 quantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; +"2 quantize_per_tensor_default" -> "3 dequantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; +"3 dequantize_per_tensor_default" -> "9 conv2d" [label="(1, 3, 224, 224)", style=solid]; +"4 conv1_weight_bias" -> "9 conv2d" [label="(64,)", style=solid]; +"5 conv2d_scale_0" -> "8 dequantize_per_channel_default" [label="(64,)", style=solid]; +"6 conv2d_zero_point_0" -> "8 dequantize_per_channel_default" [label="(64,)", style=solid]; +"7 _frozen_param0" -> "8 dequantize_per_channel_default" [label="(64, 3, 7, 7)", style=solid]; +"8 dequantize_per_channel_default" -> "9 conv2d" [label="(64, 3, 7, 7)", style=solid]; +"9 conv2d" -> "10 relu_" [label="(1, 64, 112, 112)", style=solid]; +"10 relu_" -> "11 quantize_per_tensor_default_1" [label="(1, 64, 112, 112)", style=solid]; +"11 quantize_per_tensor_default_1" -> "12 dequantize_per_tensor_default_1" [label="(1, 64, 112, 112)", style=solid]; +"12 dequantize_per_tensor_default_1" -> "13 max_pool2d" [label="(1, 64, 112, 112)", style=solid]; +"13 max_pool2d" -> "19 conv2d_1" [label="(1, 64, 56, 56)", style=solid]; +"13 max_pool2d" -> "31 add_" [label="(1, 64, 56, 56)", style=solid]; +"14 layer1_0_conv1_weight_bias" -> "19 conv2d_1" [label="(64,)", style=solid]; +"15 conv2d_1_scale_0" -> "18 dequantize_per_channel_default_1" [label="(64,)", style=solid]; +"16 conv2d_1_zero_point_0" -> "18 dequantize_per_channel_default_1" [label="(64,)", style=solid]; +"17 _frozen_param1" -> "18 dequantize_per_channel_default_1" [label="(64, 64, 3, 3)", style=solid]; +"18 dequantize_per_channel_default_1" -> "19 conv2d_1" [label="(64, 64, 3, 3)", style=solid]; +"19 conv2d_1" -> "20 relu__1" [label="(1, 64, 56, 56)", style=solid]; +"20 relu__1" -> "21 quantize_per_tensor_default_2" [label="(1, 64, 56, 56)", style=solid]; +"21 quantize_per_tensor_default_2" -> "22 dequantize_per_tensor_default_2" [label="(1, 64, 56, 56)", style=solid]; +"22 dequantize_per_tensor_default_2" -> "28 conv2d_2" [label="(1, 64, 56, 56)", style=solid]; +"23 layer1_0_conv2_weight_bias" -> "28 conv2d_2" [label="(64,)", style=solid]; +"24 conv2d_2_scale_0" -> "27 dequantize_per_channel_default_2" [label="(64,)", style=solid]; +"25 conv2d_2_zero_point_0" -> "27 dequantize_per_channel_default_2" [label="(64,)", style=solid]; +"26 _frozen_param2" -> "27 dequantize_per_channel_default_2" [label="(64, 64, 3, 3)", style=solid]; +"27 dequantize_per_channel_default_2" -> "28 conv2d_2" [label="(64, 64, 3, 3)", style=solid]; +"28 conv2d_2" -> "29 quantize_per_tensor_default_3" [label="(1, 64, 56, 56)", style=solid]; +"29 quantize_per_tensor_default_3" -> "30 dequantize_per_tensor_default_3" [label="(1, 64, 56, 56)", style=solid]; +"30 dequantize_per_tensor_default_3" -> "31 add_" [label="(1, 64, 56, 56)", style=solid]; +"31 add_" -> "32 relu__2" [label="(1, 64, 56, 56)", style=solid]; +"32 relu__2" -> "33 quantize_per_tensor_default_4" [label="(1, 64, 56, 56)", style=solid]; +"33 quantize_per_tensor_default_4" -> "34 dequantize_per_tensor_default_5" [label="(1, 64, 56, 56)", style=solid]; +"33 quantize_per_tensor_default_4" -> "35 dequantize_per_tensor_default_4" [label="(1, 64, 56, 56)", style=solid]; +"34 dequantize_per_tensor_default_5" -> "53 add__1" [label="(1, 64, 56, 56)", style=solid]; +"35 dequantize_per_tensor_default_4" -> "41 conv2d_3" [label="(1, 64, 56, 56)", style=solid]; +"36 layer1_1_conv1_weight_bias" -> "41 conv2d_3" [label="(64,)", style=solid]; +"37 conv2d_3_scale_0" -> "40 dequantize_per_channel_default_3" [label="(64,)", style=solid]; +"38 conv2d_3_zero_point_0" -> "40 dequantize_per_channel_default_3" [label="(64,)", style=solid]; +"39 _frozen_param3" -> "40 dequantize_per_channel_default_3" [label="(64, 64, 3, 3)", style=solid]; +"40 dequantize_per_channel_default_3" -> "41 conv2d_3" [label="(64, 64, 3, 3)", style=solid]; +"41 conv2d_3" -> "42 relu__3" [label="(1, 64, 56, 56)", style=solid]; +"42 relu__3" -> "43 quantize_per_tensor_default_5" [label="(1, 64, 56, 56)", style=solid]; +"43 quantize_per_tensor_default_5" -> "44 dequantize_per_tensor_default_6" [label="(1, 64, 56, 56)", style=solid]; +"44 dequantize_per_tensor_default_6" -> "50 conv2d_4" [label="(1, 64, 56, 56)", style=solid]; +"45 layer1_1_conv2_weight_bias" -> "50 conv2d_4" [label="(64,)", style=solid]; +"46 conv2d_4_scale_0" -> "49 dequantize_per_channel_default_4" [label="(64,)", style=solid]; +"47 conv2d_4_zero_point_0" -> "49 dequantize_per_channel_default_4" [label="(64,)", style=solid]; +"48 _frozen_param4" -> "49 dequantize_per_channel_default_4" [label="(64, 64, 3, 3)", style=solid]; +"49 dequantize_per_channel_default_4" -> "50 conv2d_4" [label="(64, 64, 3, 3)", style=solid]; +"50 conv2d_4" -> "51 quantize_per_tensor_default_6" [label="(1, 64, 56, 56)", style=solid]; +"51 quantize_per_tensor_default_6" -> "52 dequantize_per_tensor_default_7" [label="(1, 64, 56, 56)", style=solid]; +"52 dequantize_per_tensor_default_7" -> "53 add__1" [label="(1, 64, 56, 56)", style=solid]; +"53 add__1" -> "54 relu__4" [label="(1, 64, 56, 56)", style=solid]; +"54 relu__4" -> "55 quantize_per_tensor_default_7" [label="(1, 64, 56, 56)", style=solid]; +"55 quantize_per_tensor_default_7" -> "56 dequantize_per_tensor_default_9" [label="(1, 64, 56, 56)", style=solid]; +"55 quantize_per_tensor_default_7" -> "57 dequantize_per_tensor_default_8" [label="(1, 64, 56, 56)", style=solid]; +"56 dequantize_per_tensor_default_9" -> "80 conv2d_7" [label="(1, 64, 56, 56)", style=solid]; +"57 dequantize_per_tensor_default_8" -> "63 conv2d_5" [label="(1, 64, 56, 56)", style=solid]; +"58 layer2_0_conv1_weight_bias" -> "63 conv2d_5" [label="(128,)", style=solid]; +"59 conv2d_5_scale_0" -> "62 dequantize_per_channel_default_5" [label="(128,)", style=solid]; +"60 conv2d_5_zero_point_0" -> "62 dequantize_per_channel_default_5" [label="(128,)", style=solid]; +"61 _frozen_param5" -> "62 dequantize_per_channel_default_5" [label="(128, 64, 3, 3)", style=solid]; +"62 dequantize_per_channel_default_5" -> "63 conv2d_5" [label="(128, 64, 3, 3)", style=solid]; +"63 conv2d_5" -> "64 relu__5" [label="(1, 128, 28, 28)", style=solid]; +"64 relu__5" -> "65 quantize_per_tensor_default_8" [label="(1, 128, 28, 28)", style=solid]; +"65 quantize_per_tensor_default_8" -> "66 dequantize_per_tensor_default_10" [label="(1, 128, 28, 28)", style=solid]; +"66 dequantize_per_tensor_default_10" -> "72 conv2d_6" [label="(1, 128, 28, 28)", style=solid]; +"67 layer2_0_conv2_weight_bias" -> "72 conv2d_6" [label="(128,)", style=solid]; +"68 conv2d_6_scale_0" -> "71 dequantize_per_channel_default_6" [label="(128,)", style=solid]; +"69 conv2d_6_zero_point_0" -> "71 dequantize_per_channel_default_6" [label="(128,)", style=solid]; +"70 _frozen_param6" -> "71 dequantize_per_channel_default_6" [label="(128, 128, 3, 3)", style=solid]; +"71 dequantize_per_channel_default_6" -> "72 conv2d_6" [label="(128, 128, 3, 3)", style=solid]; +"72 conv2d_6" -> "73 quantize_per_tensor_default_9" [label="(1, 128, 28, 28)", style=solid]; +"73 quantize_per_tensor_default_9" -> "74 dequantize_per_tensor_default_11" [label="(1, 128, 28, 28)", style=solid]; +"74 dequantize_per_tensor_default_11" -> "83 add__2" [label="(1, 128, 28, 28)", style=solid]; +"75 layer2_0_downsample_0_weight_bias" -> "80 conv2d_7" [label="(128,)", style=solid]; +"76 conv2d_7_scale_0" -> "79 dequantize_per_channel_default_7" [label="(128,)", style=solid]; +"77 conv2d_7_zero_point_0" -> "79 dequantize_per_channel_default_7" [label="(128,)", style=solid]; +"78 _frozen_param7" -> "79 dequantize_per_channel_default_7" [label="(128, 64, 1, 1)", style=solid]; +"79 dequantize_per_channel_default_7" -> "80 conv2d_7" [label="(128, 64, 1, 1)", style=solid]; +"80 conv2d_7" -> "81 quantize_per_tensor_default_10" [label="(1, 128, 28, 28)", style=solid]; +"81 quantize_per_tensor_default_10" -> "82 dequantize_per_tensor_default_12" [label="(1, 128, 28, 28)", style=solid]; +"82 dequantize_per_tensor_default_12" -> "83 add__2" [label="(1, 128, 28, 28)", style=solid]; +"83 add__2" -> "84 relu__6" [label="(1, 128, 28, 28)", style=solid]; +"84 relu__6" -> "85 quantize_per_tensor_default_11" [label="(1, 128, 28, 28)", style=solid]; +"85 quantize_per_tensor_default_11" -> "86 dequantize_per_tensor_default_14" [label="(1, 128, 28, 28)", style=solid]; +"85 quantize_per_tensor_default_11" -> "87 dequantize_per_tensor_default_13" [label="(1, 128, 28, 28)", style=solid]; +"86 dequantize_per_tensor_default_14" -> "105 add__3" [label="(1, 128, 28, 28)", style=solid]; +"87 dequantize_per_tensor_default_13" -> "93 conv2d_8" [label="(1, 128, 28, 28)", style=solid]; +"88 layer2_1_conv1_weight_bias" -> "93 conv2d_8" [label="(128,)", style=solid]; +"89 conv2d_8_scale_0" -> "92 dequantize_per_channel_default_8" [label="(128,)", style=solid]; +"90 conv2d_8_zero_point_0" -> "92 dequantize_per_channel_default_8" [label="(128,)", style=solid]; +"91 _frozen_param8" -> "92 dequantize_per_channel_default_8" [label="(128, 128, 3, 3)", style=solid]; +"92 dequantize_per_channel_default_8" -> "93 conv2d_8" [label="(128, 128, 3, 3)", style=solid]; +"93 conv2d_8" -> "94 relu__7" [label="(1, 128, 28, 28)", style=solid]; +"94 relu__7" -> "95 quantize_per_tensor_default_12" [label="(1, 128, 28, 28)", style=solid]; +"95 quantize_per_tensor_default_12" -> "96 dequantize_per_tensor_default_15" [label="(1, 128, 28, 28)", style=solid]; +"96 dequantize_per_tensor_default_15" -> "102 conv2d_9" [label="(1, 128, 28, 28)", style=solid]; +"97 layer2_1_conv2_weight_bias" -> "102 conv2d_9" [label="(128,)", style=solid]; +"98 conv2d_9_scale_0" -> "101 dequantize_per_channel_default_9" [label="(128,)", style=solid]; +"99 conv2d_9_zero_point_0" -> "101 dequantize_per_channel_default_9" [label="(128,)", style=solid]; +"100 _frozen_param9" -> "101 dequantize_per_channel_default_9" [label="(128, 128, 3, 3)", style=solid]; +"101 dequantize_per_channel_default_9" -> "102 conv2d_9" [label="(128, 128, 3, 3)", style=solid]; +"102 conv2d_9" -> "103 quantize_per_tensor_default_13" [label="(1, 128, 28, 28)", style=solid]; +"103 quantize_per_tensor_default_13" -> "104 dequantize_per_tensor_default_16" [label="(1, 128, 28, 28)", style=solid]; +"104 dequantize_per_tensor_default_16" -> "105 add__3" [label="(1, 128, 28, 28)", style=solid]; +"105 add__3" -> "106 relu__8" [label="(1, 128, 28, 28)", style=solid]; +"106 relu__8" -> "107 quantize_per_tensor_default_14" [label="(1, 128, 28, 28)", style=solid]; +"107 quantize_per_tensor_default_14" -> "108 dequantize_per_tensor_default_18" [label="(1, 128, 28, 28)", style=solid]; +"107 quantize_per_tensor_default_14" -> "109 dequantize_per_tensor_default_17" [label="(1, 128, 28, 28)", style=solid]; +"108 dequantize_per_tensor_default_18" -> "132 conv2d_12" [label="(1, 128, 28, 28)", style=solid]; +"109 dequantize_per_tensor_default_17" -> "115 conv2d_10" [label="(1, 128, 28, 28)", style=solid]; +"110 layer3_0_conv1_weight_bias" -> "115 conv2d_10" [label="(256,)", style=solid]; +"111 conv2d_10_scale_0" -> "114 dequantize_per_channel_default_10" [label="(256,)", style=solid]; +"112 conv2d_10_zero_point_0" -> "114 dequantize_per_channel_default_10" [label="(256,)", style=solid]; +"113 _frozen_param10" -> "114 dequantize_per_channel_default_10" [label="(256, 128, 3, 3)", style=solid]; +"114 dequantize_per_channel_default_10" -> "115 conv2d_10" [label="(256, 128, 3, 3)", style=solid]; +"115 conv2d_10" -> "116 relu__9" [label="(1, 256, 14, 14)", style=solid]; +"116 relu__9" -> "117 quantize_per_tensor_default_15" [label="(1, 256, 14, 14)", style=solid]; +"117 quantize_per_tensor_default_15" -> "118 dequantize_per_tensor_default_19" [label="(1, 256, 14, 14)", style=solid]; +"118 dequantize_per_tensor_default_19" -> "124 conv2d_11" [label="(1, 256, 14, 14)", style=solid]; +"119 layer3_0_conv2_weight_bias" -> "124 conv2d_11" [label="(256,)", style=solid]; +"120 conv2d_11_scale_0" -> "123 dequantize_per_channel_default_11" [label="(256,)", style=solid]; +"121 conv2d_11_zero_point_0" -> "123 dequantize_per_channel_default_11" [label="(256,)", style=solid]; +"122 _frozen_param11" -> "123 dequantize_per_channel_default_11" [label="(256, 256, 3, 3)", style=solid]; +"123 dequantize_per_channel_default_11" -> "124 conv2d_11" [label="(256, 256, 3, 3)", style=solid]; +"124 conv2d_11" -> "125 quantize_per_tensor_default_16" [label="(1, 256, 14, 14)", style=solid]; +"125 quantize_per_tensor_default_16" -> "126 dequantize_per_tensor_default_20" [label="(1, 256, 14, 14)", style=solid]; +"126 dequantize_per_tensor_default_20" -> "135 add__4" [label="(1, 256, 14, 14)", style=solid]; +"127 layer3_0_downsample_0_weight_bias" -> "132 conv2d_12" [label="(256,)", style=solid]; +"128 conv2d_12_scale_0" -> "131 dequantize_per_channel_default_12" [label="(256,)", style=solid]; +"129 conv2d_12_zero_point_0" -> "131 dequantize_per_channel_default_12" [label="(256,)", style=solid]; +"130 _frozen_param12" -> "131 dequantize_per_channel_default_12" [label="(256, 128, 1, 1)", style=solid]; +"131 dequantize_per_channel_default_12" -> "132 conv2d_12" [label="(256, 128, 1, 1)", style=solid]; +"132 conv2d_12" -> "133 quantize_per_tensor_default_17" [label="(1, 256, 14, 14)", style=solid]; +"133 quantize_per_tensor_default_17" -> "134 dequantize_per_tensor_default_21" [label="(1, 256, 14, 14)", style=solid]; +"134 dequantize_per_tensor_default_21" -> "135 add__4" [label="(1, 256, 14, 14)", style=solid]; +"135 add__4" -> "136 relu__10" [label="(1, 256, 14, 14)", style=solid]; +"136 relu__10" -> "137 quantize_per_tensor_default_18" [label="(1, 256, 14, 14)", style=solid]; +"137 quantize_per_tensor_default_18" -> "138 dequantize_per_tensor_default_23" [label="(1, 256, 14, 14)", style=solid]; +"137 quantize_per_tensor_default_18" -> "139 dequantize_per_tensor_default_22" [label="(1, 256, 14, 14)", style=solid]; +"138 dequantize_per_tensor_default_23" -> "157 add__5" [label="(1, 256, 14, 14)", style=solid]; +"139 dequantize_per_tensor_default_22" -> "145 conv2d_13" [label="(1, 256, 14, 14)", style=solid]; +"140 layer3_1_conv1_weight_bias" -> "145 conv2d_13" [label="(256,)", style=solid]; +"141 conv2d_13_scale_0" -> "144 dequantize_per_channel_default_13" [label="(256,)", style=solid]; +"142 conv2d_13_zero_point_0" -> "144 dequantize_per_channel_default_13" [label="(256,)", style=solid]; +"143 _frozen_param13" -> "144 dequantize_per_channel_default_13" [label="(256, 256, 3, 3)", style=solid]; +"144 dequantize_per_channel_default_13" -> "145 conv2d_13" [label="(256, 256, 3, 3)", style=solid]; +"145 conv2d_13" -> "146 relu__11" [label="(1, 256, 14, 14)", style=solid]; +"146 relu__11" -> "147 quantize_per_tensor_default_19" [label="(1, 256, 14, 14)", style=solid]; +"147 quantize_per_tensor_default_19" -> "148 dequantize_per_tensor_default_24" [label="(1, 256, 14, 14)", style=solid]; +"148 dequantize_per_tensor_default_24" -> "154 conv2d_14" [label="(1, 256, 14, 14)", style=solid]; +"149 layer3_1_conv2_weight_bias" -> "154 conv2d_14" [label="(256,)", style=solid]; +"150 conv2d_14_scale_0" -> "153 dequantize_per_channel_default_14" [label="(256,)", style=solid]; +"151 conv2d_14_zero_point_0" -> "153 dequantize_per_channel_default_14" [label="(256,)", style=solid]; +"152 _frozen_param14" -> "153 dequantize_per_channel_default_14" [label="(256, 256, 3, 3)", style=solid]; +"153 dequantize_per_channel_default_14" -> "154 conv2d_14" [label="(256, 256, 3, 3)", style=solid]; +"154 conv2d_14" -> "155 quantize_per_tensor_default_20" [label="(1, 256, 14, 14)", style=solid]; +"155 quantize_per_tensor_default_20" -> "156 dequantize_per_tensor_default_25" [label="(1, 256, 14, 14)", style=solid]; +"156 dequantize_per_tensor_default_25" -> "157 add__5" [label="(1, 256, 14, 14)", style=solid]; +"157 add__5" -> "158 relu__12" [label="(1, 256, 14, 14)", style=solid]; +"158 relu__12" -> "159 quantize_per_tensor_default_21" [label="(1, 256, 14, 14)", style=solid]; +"159 quantize_per_tensor_default_21" -> "160 dequantize_per_tensor_default_27" [label="(1, 256, 14, 14)", style=solid]; +"159 quantize_per_tensor_default_21" -> "161 dequantize_per_tensor_default_26" [label="(1, 256, 14, 14)", style=solid]; +"160 dequantize_per_tensor_default_27" -> "184 conv2d_17" [label="(1, 256, 14, 14)", style=solid]; +"161 dequantize_per_tensor_default_26" -> "167 conv2d_15" [label="(1, 256, 14, 14)", style=solid]; +"162 layer4_0_conv1_weight_bias" -> "167 conv2d_15" [label="(512,)", style=solid]; +"163 conv2d_15_scale_0" -> "166 dequantize_per_channel_default_15" [label="(512,)", style=solid]; +"164 conv2d_15_zero_point_0" -> "166 dequantize_per_channel_default_15" [label="(512,)", style=solid]; +"165 _frozen_param15" -> "166 dequantize_per_channel_default_15" [label="(512, 256, 3, 3)", style=solid]; +"166 dequantize_per_channel_default_15" -> "167 conv2d_15" [label="(512, 256, 3, 3)", style=solid]; +"167 conv2d_15" -> "168 relu__13" [label="(1, 512, 7, 7)", style=solid]; +"168 relu__13" -> "169 quantize_per_tensor_default_22" [label="(1, 512, 7, 7)", style=solid]; +"169 quantize_per_tensor_default_22" -> "170 dequantize_per_tensor_default_28" [label="(1, 512, 7, 7)", style=solid]; +"170 dequantize_per_tensor_default_28" -> "176 conv2d_16" [label="(1, 512, 7, 7)", style=solid]; +"171 layer4_0_conv2_weight_bias" -> "176 conv2d_16" [label="(512,)", style=solid]; +"172 conv2d_16_scale_0" -> "175 dequantize_per_channel_default_16" [label="(512,)", style=solid]; +"173 conv2d_16_zero_point_0" -> "175 dequantize_per_channel_default_16" [label="(512,)", style=solid]; +"174 _frozen_param16" -> "175 dequantize_per_channel_default_16" [label="(512, 512, 3, 3)", style=solid]; +"175 dequantize_per_channel_default_16" -> "176 conv2d_16" [label="(512, 512, 3, 3)", style=solid]; +"176 conv2d_16" -> "177 quantize_per_tensor_default_23" [label="(1, 512, 7, 7)", style=solid]; +"177 quantize_per_tensor_default_23" -> "178 dequantize_per_tensor_default_29" [label="(1, 512, 7, 7)", style=solid]; +"178 dequantize_per_tensor_default_29" -> "187 add__6" [label="(1, 512, 7, 7)", style=solid]; +"179 layer4_0_downsample_0_weight_bias" -> "184 conv2d_17" [label="(512,)", style=solid]; +"180 conv2d_17_scale_0" -> "183 dequantize_per_channel_default_17" [label="(512,)", style=solid]; +"181 conv2d_17_zero_point_0" -> "183 dequantize_per_channel_default_17" [label="(512,)", style=solid]; +"182 _frozen_param17" -> "183 dequantize_per_channel_default_17" [label="(512, 256, 1, 1)", style=solid]; +"183 dequantize_per_channel_default_17" -> "184 conv2d_17" [label="(512, 256, 1, 1)", style=solid]; +"184 conv2d_17" -> "185 quantize_per_tensor_default_24" [label="(1, 512, 7, 7)", style=solid]; +"185 quantize_per_tensor_default_24" -> "186 dequantize_per_tensor_default_30" [label="(1, 512, 7, 7)", style=solid]; +"186 dequantize_per_tensor_default_30" -> "187 add__6" [label="(1, 512, 7, 7)", style=solid]; +"187 add__6" -> "188 relu__14" [label="(1, 512, 7, 7)", style=solid]; +"188 relu__14" -> "189 quantize_per_tensor_default_25" [label="(1, 512, 7, 7)", style=solid]; +"189 quantize_per_tensor_default_25" -> "190 dequantize_per_tensor_default_32" [label="(1, 512, 7, 7)", style=solid]; +"189 quantize_per_tensor_default_25" -> "191 dequantize_per_tensor_default_31" [label="(1, 512, 7, 7)", style=solid]; +"190 dequantize_per_tensor_default_32" -> "209 add__7" [label="(1, 512, 7, 7)", style=solid]; +"191 dequantize_per_tensor_default_31" -> "197 conv2d_18" [label="(1, 512, 7, 7)", style=solid]; +"192 layer4_1_conv1_weight_bias" -> "197 conv2d_18" [label="(512,)", style=solid]; +"193 conv2d_18_scale_0" -> "196 dequantize_per_channel_default_18" [label="(512,)", style=solid]; +"194 conv2d_18_zero_point_0" -> "196 dequantize_per_channel_default_18" [label="(512,)", style=solid]; +"195 _frozen_param18" -> "196 dequantize_per_channel_default_18" [label="(512, 512, 3, 3)", style=solid]; +"196 dequantize_per_channel_default_18" -> "197 conv2d_18" [label="(512, 512, 3, 3)", style=solid]; +"197 conv2d_18" -> "198 relu__15" [label="(1, 512, 7, 7)", style=solid]; +"198 relu__15" -> "199 quantize_per_tensor_default_26" [label="(1, 512, 7, 7)", style=solid]; +"199 quantize_per_tensor_default_26" -> "200 dequantize_per_tensor_default_33" [label="(1, 512, 7, 7)", style=solid]; +"200 dequantize_per_tensor_default_33" -> "206 conv2d_19" [label="(1, 512, 7, 7)", style=solid]; +"201 layer4_1_conv2_weight_bias" -> "206 conv2d_19" [label="(512,)", style=solid]; +"202 conv2d_19_scale_0" -> "205 dequantize_per_channel_default_19" [label="(512,)", style=solid]; +"203 conv2d_19_zero_point_0" -> "205 dequantize_per_channel_default_19" [label="(512,)", style=solid]; +"204 _frozen_param19" -> "205 dequantize_per_channel_default_19" [label="(512, 512, 3, 3)", style=solid]; +"205 dequantize_per_channel_default_19" -> "206 conv2d_19" [label="(512, 512, 3, 3)", style=solid]; +"206 conv2d_19" -> "207 quantize_per_tensor_default_27" [label="(1, 512, 7, 7)", style=solid]; +"207 quantize_per_tensor_default_27" -> "208 dequantize_per_tensor_default_34" [label="(1, 512, 7, 7)", style=solid]; +"208 dequantize_per_tensor_default_34" -> "209 add__7" [label="(1, 512, 7, 7)", style=solid]; +"209 add__7" -> "210 relu__16" [label="(1, 512, 7, 7)", style=solid]; +"210 relu__16" -> "211 quantize_per_tensor_default_28" [label="(1, 512, 7, 7)", style=solid]; +"211 quantize_per_tensor_default_28" -> "212 dequantize_per_tensor_default_35" [label="(1, 512, 7, 7)", style=solid]; +"212 dequantize_per_tensor_default_35" -> "213 adaptive_avg_pool2d" [label="(1, 512, 7, 7)", style=solid]; +"213 adaptive_avg_pool2d" -> "214 flatten" [label="(1, 512, 1, 1)", style=solid]; +"214 flatten" -> "215 flatten_0_0_nncf_smooth_quant_0" [label="(1, 512)", style=solid]; +"215 flatten_0_0_nncf_smooth_quant_0" -> "216 quantize_per_tensor_default_29" [label="(1, 512)", style=solid]; +"216 quantize_per_tensor_default_29" -> "217 dequantize_per_tensor_default_36" [label="(1, 512)", style=solid]; +"217 dequantize_per_tensor_default_36" -> "222 linear" [label="(1, 512)", style=solid]; +"218 linear_scale_0" -> "221 dequantize_per_channel_default_20" [label="(1000,)", style=solid]; +"219 linear_zero_point_0" -> "221 dequantize_per_channel_default_20" [label="(1000,)", style=solid]; +"220 _frozen_param20" -> "221 dequantize_per_channel_default_20" [label="(1000, 512)", style=solid]; +"221 dequantize_per_channel_default_20" -> "222 linear" [label="(1000, 512)", style=solid]; +"222 linear" -> "223 output" [label="(1, 1000)", style=solid]; +} diff --git a/tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/swin_v2_s.dot b/tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/swin_v2_s.dot new file mode 100644 index 00000000000..0e655b8fce3 --- /dev/null +++ b/tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/swin_v2_s.dot @@ -0,0 +1,6310 @@ +strict digraph { +"0 features_1_0_attn_qkv_bias" [id=0, type=get_attr]; +"1 features_1_0_attn_proj_bias" [id=1, type=get_attr]; +"2 features_1_0_attn_logit_scale" [id=2, type=get_attr]; +"3 features_1_1_attn_qkv_bias" [id=3, type=get_attr]; +"4 features_1_1_attn_proj_bias" [id=4, type=get_attr]; +"5 features_1_1_attn_logit_scale" [id=5, type=get_attr]; +"6 features_3_0_attn_qkv_bias" [id=6, type=get_attr]; +"7 features_3_0_attn_proj_bias" [id=7, type=get_attr]; +"8 features_3_0_attn_logit_scale" [id=8, type=get_attr]; +"9 features_3_1_attn_qkv_bias" [id=9, type=get_attr]; +"10 features_3_1_attn_proj_bias" [id=10, type=get_attr]; +"11 features_3_1_attn_logit_scale" [id=11, type=get_attr]; +"12 features_5_0_attn_qkv_bias" [id=12, type=get_attr]; +"13 features_5_0_attn_proj_bias" [id=13, type=get_attr]; +"14 features_5_0_attn_logit_scale" [id=14, type=get_attr]; +"15 features_5_1_attn_qkv_bias" [id=15, type=get_attr]; +"16 features_5_1_attn_proj_bias" [id=16, type=get_attr]; +"17 features_5_1_attn_logit_scale" [id=17, type=get_attr]; +"18 features_5_2_attn_qkv_bias" [id=18, type=get_attr]; +"19 features_5_2_attn_proj_bias" [id=19, type=get_attr]; +"20 features_5_2_attn_logit_scale" [id=20, type=get_attr]; +"21 features_5_3_attn_qkv_bias" [id=21, type=get_attr]; +"22 features_5_3_attn_proj_bias" [id=22, type=get_attr]; +"23 features_5_3_attn_logit_scale" [id=23, type=get_attr]; +"24 features_5_4_attn_qkv_bias" [id=24, type=get_attr]; +"25 features_5_4_attn_proj_bias" [id=25, type=get_attr]; +"26 features_5_4_attn_logit_scale" [id=26, type=get_attr]; +"27 features_5_5_attn_qkv_bias" [id=27, type=get_attr]; +"28 features_5_5_attn_proj_bias" [id=28, type=get_attr]; +"29 features_5_5_attn_logit_scale" [id=29, type=get_attr]; +"30 features_5_6_attn_qkv_bias" [id=30, type=get_attr]; +"31 features_5_6_attn_proj_bias" [id=31, type=get_attr]; +"32 features_5_6_attn_logit_scale" [id=32, type=get_attr]; +"33 features_5_7_attn_qkv_bias" [id=33, type=get_attr]; +"34 features_5_7_attn_proj_bias" [id=34, type=get_attr]; +"35 features_5_7_attn_logit_scale" [id=35, type=get_attr]; +"36 features_5_8_attn_qkv_bias" [id=36, type=get_attr]; +"37 features_5_8_attn_proj_bias" [id=37, type=get_attr]; +"38 features_5_8_attn_logit_scale" [id=38, type=get_attr]; +"39 features_5_9_attn_qkv_bias" [id=39, type=get_attr]; +"40 features_5_9_attn_proj_bias" [id=40, type=get_attr]; +"41 features_5_9_attn_logit_scale" [id=41, type=get_attr]; +"42 features_5_10_attn_qkv_bias" [id=42, type=get_attr]; +"43 features_5_10_attn_proj_bias" [id=43, type=get_attr]; +"44 features_5_10_attn_logit_scale" [id=44, type=get_attr]; +"45 features_5_11_attn_qkv_bias" [id=45, type=get_attr]; +"46 features_5_11_attn_proj_bias" [id=46, type=get_attr]; +"47 features_5_11_attn_logit_scale" [id=47, type=get_attr]; +"48 features_5_12_attn_qkv_bias" [id=48, type=get_attr]; +"49 features_5_12_attn_proj_bias" [id=49, type=get_attr]; +"50 features_5_12_attn_logit_scale" [id=50, type=get_attr]; +"51 features_5_13_attn_qkv_bias" [id=51, type=get_attr]; +"52 features_5_13_attn_proj_bias" [id=52, type=get_attr]; +"53 features_5_13_attn_logit_scale" [id=53, type=get_attr]; +"54 features_5_14_attn_qkv_bias" [id=54, type=get_attr]; +"55 features_5_14_attn_proj_bias" [id=55, type=get_attr]; +"56 features_5_14_attn_logit_scale" [id=56, type=get_attr]; +"57 features_5_15_attn_qkv_bias" [id=57, type=get_attr]; +"58 features_5_15_attn_proj_bias" [id=58, type=get_attr]; +"59 features_5_15_attn_logit_scale" [id=59, type=get_attr]; +"60 features_5_16_attn_qkv_bias" [id=60, type=get_attr]; +"61 features_5_16_attn_proj_bias" [id=61, type=get_attr]; +"62 features_5_16_attn_logit_scale" [id=62, type=get_attr]; +"63 features_5_17_attn_qkv_bias" [id=63, type=get_attr]; +"64 features_5_17_attn_proj_bias" [id=64, type=get_attr]; +"65 features_5_17_attn_logit_scale" [id=65, type=get_attr]; +"66 features_7_0_attn_qkv_bias" [id=66, type=get_attr]; +"67 features_7_0_attn_proj_bias" [id=67, type=get_attr]; +"68 features_7_0_attn_logit_scale" [id=68, type=get_attr]; +"69 features_7_1_attn_qkv_bias" [id=69, type=get_attr]; +"70 features_7_1_attn_proj_bias" [id=70, type=get_attr]; +"71 features_7_1_attn_logit_scale" [id=71, type=get_attr]; +"72 features_0_0_bias" [id=72, type=get_attr]; +"73 features_0_2_weight" [id=73, type=get_attr]; +"74 features_0_2_bias" [id=74, type=get_attr]; +"75 features_1_0_attn_cpb_mlp_0_bias" [id=75, type=get_attr]; +"76 features_1_0_norm1_weight" [id=76, type=get_attr]; +"77 features_1_0_norm1_bias" [id=77, type=get_attr]; +"78 features_1_0_mlp_0_bias" [id=78, type=get_attr]; +"79 features_1_0_mlp_3_bias" [id=79, type=get_attr]; +"80 features_1_0_norm2_weight" [id=80, type=get_attr]; +"81 features_1_0_norm2_bias" [id=81, type=get_attr]; +"82 features_1_1_attn_cpb_mlp_0_bias" [id=82, type=get_attr]; +"83 features_1_1_norm1_weight" [id=83, type=get_attr]; +"84 features_1_1_norm1_bias" [id=84, type=get_attr]; +"85 features_1_1_mlp_0_bias" [id=85, type=get_attr]; +"86 features_1_1_mlp_3_bias" [id=86, type=get_attr]; +"87 features_1_1_norm2_weight" [id=87, type=get_attr]; +"88 features_1_1_norm2_bias" [id=88, type=get_attr]; +"89 features_2_norm_weight" [id=89, type=get_attr]; +"90 features_2_norm_bias" [id=90, type=get_attr]; +"91 features_3_0_attn_cpb_mlp_0_bias" [id=91, type=get_attr]; +"92 features_3_0_norm1_weight" [id=92, type=get_attr]; +"93 features_3_0_norm1_bias" [id=93, type=get_attr]; +"94 features_3_0_mlp_0_bias" [id=94, type=get_attr]; +"95 features_3_0_mlp_3_bias" [id=95, type=get_attr]; +"96 features_3_0_norm2_weight" [id=96, type=get_attr]; +"97 features_3_0_norm2_bias" [id=97, type=get_attr]; +"98 features_3_1_attn_cpb_mlp_0_bias" [id=98, type=get_attr]; +"99 features_3_1_norm1_weight" [id=99, type=get_attr]; +"100 features_3_1_norm1_bias" [id=100, type=get_attr]; +"101 features_3_1_mlp_0_bias" [id=101, type=get_attr]; +"102 features_3_1_mlp_3_bias" [id=102, type=get_attr]; +"103 features_3_1_norm2_weight" [id=103, type=get_attr]; +"104 features_3_1_norm2_bias" [id=104, type=get_attr]; +"105 features_4_norm_weight" [id=105, type=get_attr]; +"106 features_4_norm_bias" [id=106, type=get_attr]; +"107 features_5_0_attn_cpb_mlp_0_bias" [id=107, type=get_attr]; +"108 features_5_0_norm1_weight" [id=108, type=get_attr]; +"109 features_5_0_norm1_bias" [id=109, type=get_attr]; +"110 features_5_0_mlp_0_bias" [id=110, type=get_attr]; +"111 features_5_0_mlp_3_bias" [id=111, type=get_attr]; +"112 features_5_0_norm2_weight" [id=112, type=get_attr]; +"113 features_5_0_norm2_bias" [id=113, type=get_attr]; +"114 features_5_1_attn_cpb_mlp_0_bias" [id=114, type=get_attr]; +"115 features_5_1_norm1_weight" [id=115, type=get_attr]; +"116 features_5_1_norm1_bias" [id=116, type=get_attr]; +"117 features_5_1_mlp_0_bias" [id=117, type=get_attr]; +"118 features_5_1_mlp_3_bias" [id=118, type=get_attr]; +"119 features_5_1_norm2_weight" [id=119, type=get_attr]; +"120 features_5_1_norm2_bias" [id=120, type=get_attr]; +"121 features_5_2_attn_cpb_mlp_0_bias" [id=121, type=get_attr]; +"122 features_5_2_norm1_weight" [id=122, type=get_attr]; +"123 features_5_2_norm1_bias" [id=123, type=get_attr]; +"124 features_5_2_mlp_0_bias" [id=124, type=get_attr]; +"125 features_5_2_mlp_3_bias" [id=125, type=get_attr]; +"126 features_5_2_norm2_weight" [id=126, type=get_attr]; +"127 features_5_2_norm2_bias" [id=127, type=get_attr]; +"128 features_5_3_attn_cpb_mlp_0_bias" [id=128, type=get_attr]; +"129 features_5_3_norm1_weight" [id=129, type=get_attr]; +"130 features_5_3_norm1_bias" [id=130, type=get_attr]; +"131 features_5_3_mlp_0_bias" [id=131, type=get_attr]; +"132 features_5_3_mlp_3_bias" [id=132, type=get_attr]; +"133 features_5_3_norm2_weight" [id=133, type=get_attr]; +"134 features_5_3_norm2_bias" [id=134, type=get_attr]; +"135 features_5_4_attn_cpb_mlp_0_bias" [id=135, type=get_attr]; +"136 features_5_4_norm1_weight" [id=136, type=get_attr]; +"137 features_5_4_norm1_bias" [id=137, type=get_attr]; +"138 features_5_4_mlp_0_bias" [id=138, type=get_attr]; +"139 features_5_4_mlp_3_bias" [id=139, type=get_attr]; +"140 features_5_4_norm2_weight" [id=140, type=get_attr]; +"141 features_5_4_norm2_bias" [id=141, type=get_attr]; +"142 features_5_5_attn_cpb_mlp_0_bias" [id=142, type=get_attr]; +"143 features_5_5_norm1_weight" [id=143, type=get_attr]; +"144 features_5_5_norm1_bias" [id=144, type=get_attr]; +"145 features_5_5_mlp_0_bias" [id=145, type=get_attr]; +"146 features_5_5_mlp_3_bias" [id=146, type=get_attr]; +"147 features_5_5_norm2_weight" [id=147, type=get_attr]; +"148 features_5_5_norm2_bias" [id=148, type=get_attr]; +"149 features_5_6_attn_cpb_mlp_0_bias" [id=149, type=get_attr]; +"150 features_5_6_norm1_weight" [id=150, type=get_attr]; +"151 features_5_6_norm1_bias" [id=151, type=get_attr]; +"152 features_5_6_mlp_0_bias" [id=152, type=get_attr]; +"153 features_5_6_mlp_3_bias" [id=153, type=get_attr]; +"154 features_5_6_norm2_weight" [id=154, type=get_attr]; +"155 features_5_6_norm2_bias" [id=155, type=get_attr]; +"156 features_5_7_attn_cpb_mlp_0_bias" [id=156, type=get_attr]; +"157 features_5_7_norm1_weight" [id=157, type=get_attr]; +"158 features_5_7_norm1_bias" [id=158, type=get_attr]; +"159 features_5_7_mlp_0_bias" [id=159, type=get_attr]; +"160 features_5_7_mlp_3_bias" [id=160, type=get_attr]; +"161 features_5_7_norm2_weight" [id=161, type=get_attr]; +"162 features_5_7_norm2_bias" [id=162, type=get_attr]; +"163 features_5_8_attn_cpb_mlp_0_bias" [id=163, type=get_attr]; +"164 features_5_8_norm1_weight" [id=164, type=get_attr]; +"165 features_5_8_norm1_bias" [id=165, type=get_attr]; +"166 features_5_8_mlp_0_bias" [id=166, type=get_attr]; +"167 features_5_8_mlp_3_bias" [id=167, type=get_attr]; +"168 features_5_8_norm2_weight" [id=168, type=get_attr]; +"169 features_5_8_norm2_bias" [id=169, type=get_attr]; +"170 features_5_9_attn_cpb_mlp_0_bias" [id=170, type=get_attr]; +"171 features_5_9_norm1_weight" [id=171, type=get_attr]; +"172 features_5_9_norm1_bias" [id=172, type=get_attr]; +"173 features_5_9_mlp_0_bias" [id=173, type=get_attr]; +"174 features_5_9_mlp_3_bias" [id=174, type=get_attr]; +"175 features_5_9_norm2_weight" [id=175, type=get_attr]; +"176 features_5_9_norm2_bias" [id=176, type=get_attr]; +"177 features_5_10_attn_cpb_mlp_0_bias" [id=177, type=get_attr]; +"178 features_5_10_norm1_weight" [id=178, type=get_attr]; +"179 features_5_10_norm1_bias" [id=179, type=get_attr]; +"180 features_5_10_mlp_0_bias" [id=180, type=get_attr]; +"181 features_5_10_mlp_3_bias" [id=181, type=get_attr]; +"182 features_5_10_norm2_weight" [id=182, type=get_attr]; +"183 features_5_10_norm2_bias" [id=183, type=get_attr]; +"184 features_5_11_attn_cpb_mlp_0_bias" [id=184, type=get_attr]; +"185 features_5_11_norm1_weight" [id=185, type=get_attr]; +"186 features_5_11_norm1_bias" [id=186, type=get_attr]; +"187 features_5_11_mlp_0_bias" [id=187, type=get_attr]; +"188 features_5_11_mlp_3_bias" [id=188, type=get_attr]; +"189 features_5_11_norm2_weight" [id=189, type=get_attr]; +"190 features_5_11_norm2_bias" [id=190, type=get_attr]; +"191 features_5_12_attn_cpb_mlp_0_bias" [id=191, type=get_attr]; +"192 features_5_12_norm1_weight" [id=192, type=get_attr]; +"193 features_5_12_norm1_bias" [id=193, type=get_attr]; +"194 features_5_12_mlp_0_bias" [id=194, type=get_attr]; +"195 features_5_12_mlp_3_bias" [id=195, type=get_attr]; +"196 features_5_12_norm2_weight" [id=196, type=get_attr]; +"197 features_5_12_norm2_bias" [id=197, type=get_attr]; +"198 features_5_13_attn_cpb_mlp_0_bias" [id=198, type=get_attr]; +"199 features_5_13_norm1_weight" [id=199, type=get_attr]; +"200 features_5_13_norm1_bias" [id=200, type=get_attr]; +"201 features_5_13_mlp_0_bias" [id=201, type=get_attr]; +"202 features_5_13_mlp_3_bias" [id=202, type=get_attr]; +"203 features_5_13_norm2_weight" [id=203, type=get_attr]; +"204 features_5_13_norm2_bias" [id=204, type=get_attr]; +"205 features_5_14_attn_cpb_mlp_0_bias" [id=205, type=get_attr]; +"206 features_5_14_norm1_weight" [id=206, type=get_attr]; +"207 features_5_14_norm1_bias" [id=207, type=get_attr]; +"208 features_5_14_mlp_0_bias" [id=208, type=get_attr]; +"209 features_5_14_mlp_3_bias" [id=209, type=get_attr]; +"210 features_5_14_norm2_weight" [id=210, type=get_attr]; +"211 features_5_14_norm2_bias" [id=211, type=get_attr]; +"212 features_5_15_attn_cpb_mlp_0_bias" [id=212, type=get_attr]; +"213 features_5_15_norm1_weight" [id=213, type=get_attr]; +"214 features_5_15_norm1_bias" [id=214, type=get_attr]; +"215 features_5_15_mlp_0_bias" [id=215, type=get_attr]; +"216 features_5_15_mlp_3_bias" [id=216, type=get_attr]; +"217 features_5_15_norm2_weight" [id=217, type=get_attr]; +"218 features_5_15_norm2_bias" [id=218, type=get_attr]; +"219 features_5_16_attn_cpb_mlp_0_bias" [id=219, type=get_attr]; +"220 features_5_16_norm1_weight" [id=220, type=get_attr]; +"221 features_5_16_norm1_bias" [id=221, type=get_attr]; +"222 features_5_16_mlp_0_bias" [id=222, type=get_attr]; +"223 features_5_16_mlp_3_bias" [id=223, type=get_attr]; +"224 features_5_16_norm2_weight" [id=224, type=get_attr]; +"225 features_5_16_norm2_bias" [id=225, type=get_attr]; +"226 features_5_17_attn_cpb_mlp_0_bias" [id=226, type=get_attr]; +"227 features_5_17_norm1_weight" [id=227, type=get_attr]; +"228 features_5_17_norm1_bias" [id=228, type=get_attr]; +"229 features_5_17_mlp_0_bias" [id=229, type=get_attr]; +"230 features_5_17_mlp_3_bias" [id=230, type=get_attr]; +"231 features_5_17_norm2_weight" [id=231, type=get_attr]; +"232 features_5_17_norm2_bias" [id=232, type=get_attr]; +"233 features_6_norm_weight" [id=233, type=get_attr]; +"234 features_6_norm_bias" [id=234, type=get_attr]; +"235 features_7_0_attn_cpb_mlp_0_bias" [id=235, type=get_attr]; +"236 features_7_0_norm1_weight" [id=236, type=get_attr]; +"237 features_7_0_norm1_bias" [id=237, type=get_attr]; +"238 features_7_0_mlp_0_bias" [id=238, type=get_attr]; +"239 features_7_0_mlp_3_bias" [id=239, type=get_attr]; +"240 features_7_0_norm2_weight" [id=240, type=get_attr]; +"241 features_7_0_norm2_bias" [id=241, type=get_attr]; +"242 features_7_1_attn_cpb_mlp_0_bias" [id=242, type=get_attr]; +"243 features_7_1_norm1_weight" [id=243, type=get_attr]; +"244 features_7_1_norm1_bias" [id=244, type=get_attr]; +"245 features_7_1_mlp_0_bias" [id=245, type=get_attr]; +"246 features_7_1_mlp_3_bias" [id=246, type=get_attr]; +"247 features_7_1_norm2_weight" [id=247, type=get_attr]; +"248 features_7_1_norm2_bias" [id=248, type=get_attr]; +"249 norm_weight" [id=249, type=get_attr]; +"250 norm_bias" [id=250, type=get_attr]; +"251 head_bias" [id=251, type=get_attr]; +"252 features_1_0_attn_relative_coords_table" [id=252, type=get_attr]; +"253 features_1_0_attn_relative_position_index" [id=253, type=get_attr]; +"254 features_1_1_attn_relative_coords_table" [id=254, type=get_attr]; +"255 features_1_1_attn_relative_position_index" [id=255, type=get_attr]; +"256 features_3_0_attn_relative_coords_table" [id=256, type=get_attr]; +"257 features_3_0_attn_relative_position_index" [id=257, type=get_attr]; +"258 features_3_1_attn_relative_coords_table" [id=258, type=get_attr]; +"259 features_3_1_attn_relative_position_index" [id=259, type=get_attr]; +"260 features_5_0_attn_relative_coords_table" [id=260, type=get_attr]; +"261 features_5_0_attn_relative_position_index" [id=261, type=get_attr]; +"262 features_5_1_attn_relative_coords_table" [id=262, type=get_attr]; +"263 features_5_1_attn_relative_position_index" [id=263, type=get_attr]; +"264 features_5_2_attn_relative_coords_table" [id=264, type=get_attr]; +"265 features_5_2_attn_relative_position_index" [id=265, type=get_attr]; +"266 features_5_3_attn_relative_coords_table" [id=266, type=get_attr]; +"267 features_5_3_attn_relative_position_index" [id=267, type=get_attr]; +"268 features_5_4_attn_relative_coords_table" [id=268, type=get_attr]; +"269 features_5_4_attn_relative_position_index" [id=269, type=get_attr]; +"270 features_5_5_attn_relative_coords_table" [id=270, type=get_attr]; +"271 features_5_5_attn_relative_position_index" [id=271, type=get_attr]; +"272 features_5_6_attn_relative_coords_table" [id=272, type=get_attr]; +"273 features_5_6_attn_relative_position_index" [id=273, type=get_attr]; +"274 features_5_7_attn_relative_coords_table" [id=274, type=get_attr]; +"275 features_5_7_attn_relative_position_index" [id=275, type=get_attr]; +"276 features_5_8_attn_relative_coords_table" [id=276, type=get_attr]; +"277 features_5_8_attn_relative_position_index" [id=277, type=get_attr]; +"278 features_5_9_attn_relative_coords_table" [id=278, type=get_attr]; +"279 features_5_9_attn_relative_position_index" [id=279, type=get_attr]; +"280 features_5_10_attn_relative_coords_table" [id=280, type=get_attr]; +"281 features_5_10_attn_relative_position_index" [id=281, type=get_attr]; +"282 features_5_11_attn_relative_coords_table" [id=282, type=get_attr]; +"283 features_5_11_attn_relative_position_index" [id=283, type=get_attr]; +"284 features_5_12_attn_relative_coords_table" [id=284, type=get_attr]; +"285 features_5_12_attn_relative_position_index" [id=285, type=get_attr]; +"286 features_5_13_attn_relative_coords_table" [id=286, type=get_attr]; +"287 features_5_13_attn_relative_position_index" [id=287, type=get_attr]; +"288 features_5_14_attn_relative_coords_table" [id=288, type=get_attr]; +"289 features_5_14_attn_relative_position_index" [id=289, type=get_attr]; +"290 features_5_15_attn_relative_coords_table" [id=290, type=get_attr]; +"291 features_5_15_attn_relative_position_index" [id=291, type=get_attr]; +"292 features_5_16_attn_relative_coords_table" [id=292, type=get_attr]; +"293 features_5_16_attn_relative_position_index" [id=293, type=get_attr]; +"294 features_5_17_attn_relative_coords_table" [id=294, type=get_attr]; +"295 features_5_17_attn_relative_position_index" [id=295, type=get_attr]; +"296 features_7_0_attn_relative_coords_table" [id=296, type=get_attr]; +"297 features_7_0_attn_relative_position_index" [id=297, type=get_attr]; +"298 features_7_1_attn_relative_coords_table" [id=298, type=get_attr]; +"299 features_7_1_attn_relative_position_index" [id=299, type=get_attr]; +"300 x" [id=300, type=input]; +"301 quantize_per_tensor_default" [id=301, type=quantize_per_tensor]; +"302 dequantize_per_tensor_default" [id=302, type=dequantize_per_tensor]; +"303 conv2d_scale_0" [id=303, type=get_attr]; +"304 conv2d_zero_point_0" [id=304, type=get_attr]; +"305 _frozen_param0" [id=305, type=get_attr]; +"306 dequantize_per_channel_default" [id=306, type=dequantize_per_channel]; +"307 conv2d" [id=307, type=conv2d]; +"308 permute" [id=308, type=permute]; +"309 layer_norm" [id=309, type=layer_norm]; +"310 features_1_0_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=310, type=call_module]; +"311 linear_scale_0" [id=311, type=get_attr]; +"312 linear_zero_point_0" [id=312, type=get_attr]; +"313 _frozen_param1" [id=313, type=get_attr]; +"314 dequantize_per_channel_default_1" [id=314, type=dequantize_per_channel]; +"315 linear" [id=315, type=linear]; +"316 relu_" [id=316, type=relu_]; +"317 relu__0_0_nncf_smooth_quant_0" [id=317, type=call_module]; +"318 linear_1_scale_0" [id=318, type=get_attr]; +"319 linear_1_zero_point_0" [id=319, type=get_attr]; +"320 _frozen_param2" [id=320, type=get_attr]; +"321 dequantize_per_channel_default_2" [id=321, type=dequantize_per_channel]; +"322 linear_1" [id=322, type=linear]; +"323 view" [id=323, type=view]; +"324 index" [id=324, type=index]; +"325 view_1" [id=325, type=view]; +"326 permute_1" [id=326, type=permute]; +"327 contiguous" [id=327, type=contiguous]; +"328 unsqueeze" [id=328, type=unsqueeze]; +"329 sigmoid" [id=329, type=sigmoid]; +"330 mul" [id=330, type=mul]; +"331 pad" [id=331, type=pad]; +"332 view_2" [id=332, type=view]; +"333 permute_2" [id=333, type=permute]; +"334 reshape" [id=334, type=reshape]; +"335 clone" [id=335, type=clone]; +"336 reshape_0_0_nncf_smooth_quant_0" [id=336, type=call_module]; +"337 quantize_per_tensor_default_1" [id=337, type=quantize_per_tensor]; +"338 dequantize_per_tensor_default_1" [id=338, type=dequantize_per_tensor]; +"339 linear_2_scale_0" [id=339, type=get_attr]; +"340 linear_2_zero_point_0" [id=340, type=get_attr]; +"341 _frozen_param3" [id=341, type=get_attr]; +"342 dequantize_per_channel_default_3" [id=342, type=dequantize_per_channel]; +"343 linear_2" [id=343, type=linear]; +"344 reshape_1" [id=344, type=reshape]; +"345 permute_3" [id=345, type=permute]; +"346 select" [id=346, type=select]; +"347 select_1" [id=347, type=select]; +"348 select_2" [id=348, type=select]; +"349 linalg_vector_norm" [id=349, type=linalg_vector_norm]; +"350 clamp_min" [id=350, type=clamp_min]; +"351 expand_as" [id=351, type=expand_as]; +"352 div" [id=352, type=div]; +"353 quantize_per_tensor_default_2" [id=353, type=quantize_per_tensor]; +"354 dequantize_per_tensor_default_2" [id=354, type=dequantize_per_tensor]; +"355 linalg_vector_norm_1" [id=355, type=linalg_vector_norm]; +"356 clamp_min_1" [id=356, type=clamp_min]; +"357 expand_as_1" [id=357, type=expand_as]; +"358 div_1" [id=358, type=div]; +"359 quantize_per_tensor_default_3" [id=359, type=quantize_per_tensor]; +"360 dequantize_per_tensor_default_3" [id=360, type=dequantize_per_tensor]; +"361 transpose" [id=361, type=transpose]; +"362 matmul" [id=362, type=matmul]; +"363 clamp" [id=363, type=clamp]; +"364 exp" [id=364, type=exp]; +"365 mul_1" [id=365, type=mul]; +"366 add" [id=366, type=add]; +"367 softmax" [id=367, type=softmax]; +"368 dropout" [id=368, type=dropout]; +"369 matmul_1" [id=369, type=matmul]; +"370 transpose_1" [id=370, type=transpose]; +"371 reshape_2" [id=371, type=reshape]; +"372 reshape_2_0_0_nncf_smooth_quant_0" [id=372, type=call_module]; +"373 quantize_per_tensor_default_4" [id=373, type=quantize_per_tensor]; +"374 dequantize_per_tensor_default_4" [id=374, type=dequantize_per_tensor]; +"375 linear_3_scale_0" [id=375, type=get_attr]; +"376 linear_3_zero_point_0" [id=376, type=get_attr]; +"377 _frozen_param4" [id=377, type=get_attr]; +"378 dequantize_per_channel_default_4" [id=378, type=dequantize_per_channel]; +"379 linear_3" [id=379, type=linear]; +"380 dropout_1" [id=380, type=dropout]; +"381 view_3" [id=381, type=view]; +"382 permute_4" [id=382, type=permute]; +"383 reshape_3" [id=383, type=reshape]; +"384 slice_2" [id=384, type=slice]; +"385 slice_3" [id=385, type=slice]; +"386 layer_norm_1" [id=386, type=layer_norm]; +"387 add_1" [id=387, type=add]; +"388 add_1_0_0_nncf_smooth_quant_0" [id=388, type=call_module]; +"389 quantize_per_tensor_default_5" [id=389, type=quantize_per_tensor]; +"390 dequantize_per_tensor_default_5" [id=390, type=dequantize_per_tensor]; +"391 linear_4_scale_0" [id=391, type=get_attr]; +"392 linear_4_zero_point_0" [id=392, type=get_attr]; +"393 _frozen_param5" [id=393, type=get_attr]; +"394 dequantize_per_channel_default_5" [id=394, type=dequantize_per_channel]; +"395 linear_4" [id=395, type=linear]; +"396 gelu" [id=396, type=gelu]; +"397 dropout_2" [id=397, type=dropout]; +"398 dropout_2_0_0_nncf_smooth_quant_0" [id=398, type=call_module]; +"399 quantize_per_tensor_default_6" [id=399, type=quantize_per_tensor]; +"400 dequantize_per_tensor_default_6" [id=400, type=dequantize_per_tensor]; +"401 linear_5_scale_0" [id=401, type=get_attr]; +"402 linear_5_zero_point_0" [id=402, type=get_attr]; +"403 _frozen_param6" [id=403, type=get_attr]; +"404 dequantize_per_channel_default_6" [id=404, type=dequantize_per_channel]; +"405 linear_5" [id=405, type=linear]; +"406 dropout_3" [id=406, type=dropout]; +"407 layer_norm_2" [id=407, type=layer_norm]; +"408 add_2" [id=408, type=add]; +"409 features_1_1_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=409, type=call_module]; +"410 linear_6_scale_0" [id=410, type=get_attr]; +"411 linear_6_zero_point_0" [id=411, type=get_attr]; +"412 _frozen_param7" [id=412, type=get_attr]; +"413 dequantize_per_channel_default_7" [id=413, type=dequantize_per_channel]; +"414 linear_6" [id=414, type=linear]; +"415 relu__1" [id=415, type=relu_]; +"416 relu__1_0_0_nncf_smooth_quant_0" [id=416, type=call_module]; +"417 linear_7_scale_0" [id=417, type=get_attr]; +"418 linear_7_zero_point_0" [id=418, type=get_attr]; +"419 _frozen_param8" [id=419, type=get_attr]; +"420 dequantize_per_channel_default_8" [id=420, type=dequantize_per_channel]; +"421 linear_7" [id=421, type=linear]; +"422 view_4" [id=422, type=view]; +"423 index_1" [id=423, type=index]; +"424 view_5" [id=424, type=view]; +"425 permute_5" [id=425, type=permute]; +"426 contiguous_1" [id=426, type=contiguous]; +"427 unsqueeze_1" [id=427, type=unsqueeze]; +"428 sigmoid_1" [id=428, type=sigmoid]; +"429 mul_2" [id=429, type=mul]; +"430 pad_1" [id=430, type=pad]; +"431 roll" [id=431, type=roll]; +"432 view_6" [id=432, type=view]; +"433 permute_6" [id=433, type=permute]; +"434 reshape_4" [id=434, type=reshape]; +"435 clone_1" [id=435, type=clone]; +"436 reshape_4_0_0_nncf_smooth_quant_0" [id=436, type=call_module]; +"437 quantize_per_tensor_default_7" [id=437, type=quantize_per_tensor]; +"438 dequantize_per_tensor_default_7" [id=438, type=dequantize_per_tensor]; +"439 linear_8_scale_0" [id=439, type=get_attr]; +"440 linear_8_zero_point_0" [id=440, type=get_attr]; +"441 _frozen_param9" [id=441, type=get_attr]; +"442 dequantize_per_channel_default_9" [id=442, type=dequantize_per_channel]; +"443 linear_8" [id=443, type=linear]; +"444 reshape_5" [id=444, type=reshape]; +"445 permute_7" [id=445, type=permute]; +"446 select_3" [id=446, type=select]; +"447 select_4" [id=447, type=select]; +"448 select_5" [id=448, type=select]; +"449 linalg_vector_norm_2" [id=449, type=linalg_vector_norm]; +"450 clamp_min_2" [id=450, type=clamp_min]; +"451 expand_as_2" [id=451, type=expand_as]; +"452 div_2" [id=452, type=div]; +"453 quantize_per_tensor_default_8" [id=453, type=quantize_per_tensor]; +"454 dequantize_per_tensor_default_8" [id=454, type=dequantize_per_tensor]; +"455 linalg_vector_norm_3" [id=455, type=linalg_vector_norm]; +"456 clamp_min_3" [id=456, type=clamp_min]; +"457 expand_as_3" [id=457, type=expand_as]; +"458 div_3" [id=458, type=div]; +"459 quantize_per_tensor_default_9" [id=459, type=quantize_per_tensor]; +"460 dequantize_per_tensor_default_9" [id=460, type=dequantize_per_tensor]; +"461 transpose_2" [id=461, type=transpose]; +"462 matmul_2" [id=462, type=matmul]; +"463 clamp_1" [id=463, type=clamp]; +"464 exp_1" [id=464, type=exp]; +"465 mul_3" [id=465, type=mul]; +"466 add_3" [id=466, type=add]; +"467 new_zeros" [id=467, type=new_zeros]; +"468 view_7" [id=468, type=view]; +"469 permute_8" [id=469, type=permute]; +"470 reshape_6" [id=470, type=reshape]; +"471 unsqueeze_2" [id=471, type=unsqueeze]; +"472 unsqueeze_3" [id=472, type=unsqueeze]; +"473 sub" [id=473, type=sub]; +"474 ne" [id=474, type=ne]; +"475 masked_fill" [id=475, type=masked_fill]; +"476 eq" [id=476, type=eq]; +"477 masked_fill_1" [id=477, type=masked_fill]; +"478 view_8" [id=478, type=view]; +"479 unsqueeze_4" [id=479, type=unsqueeze]; +"480 unsqueeze_5" [id=480, type=unsqueeze]; +"481 add_4" [id=481, type=add]; +"482 view_9" [id=482, type=view]; +"483 softmax_1" [id=483, type=softmax]; +"484 dropout_4" [id=484, type=dropout]; +"485 matmul_3" [id=485, type=matmul]; +"486 transpose_3" [id=486, type=transpose]; +"487 reshape_7" [id=487, type=reshape]; +"488 reshape_7_0_0_nncf_smooth_quant_0" [id=488, type=call_module]; +"489 quantize_per_tensor_default_10" [id=489, type=quantize_per_tensor]; +"490 dequantize_per_tensor_default_10" [id=490, type=dequantize_per_tensor]; +"491 linear_9_scale_0" [id=491, type=get_attr]; +"492 linear_9_zero_point_0" [id=492, type=get_attr]; +"493 _frozen_param10" [id=493, type=get_attr]; +"494 dequantize_per_channel_default_10" [id=494, type=dequantize_per_channel]; +"495 linear_9" [id=495, type=linear]; +"496 dropout_5" [id=496, type=dropout]; +"497 view_10" [id=497, type=view]; +"498 permute_9" [id=498, type=permute]; +"499 reshape_8" [id=499, type=reshape]; +"500 roll_1" [id=500, type=roll]; +"501 slice_23" [id=501, type=slice]; +"502 slice_24" [id=502, type=slice]; +"503 layer_norm_3" [id=503, type=layer_norm]; +"504 add_5" [id=504, type=add]; +"505 add_5_0_0_nncf_smooth_quant_0" [id=505, type=call_module]; +"506 quantize_per_tensor_default_11" [id=506, type=quantize_per_tensor]; +"507 dequantize_per_tensor_default_11" [id=507, type=dequantize_per_tensor]; +"508 linear_10_scale_0" [id=508, type=get_attr]; +"509 linear_10_zero_point_0" [id=509, type=get_attr]; +"510 _frozen_param11" [id=510, type=get_attr]; +"511 dequantize_per_channel_default_11" [id=511, type=dequantize_per_channel]; +"512 linear_10" [id=512, type=linear]; +"513 gelu_1" [id=513, type=gelu]; +"514 dropout_6" [id=514, type=dropout]; +"515 dropout_6_0_0_nncf_smooth_quant_0" [id=515, type=call_module]; +"516 quantize_per_tensor_default_12" [id=516, type=quantize_per_tensor]; +"517 dequantize_per_tensor_default_12" [id=517, type=dequantize_per_tensor]; +"518 linear_11_scale_0" [id=518, type=get_attr]; +"519 linear_11_zero_point_0" [id=519, type=get_attr]; +"520 _frozen_param12" [id=520, type=get_attr]; +"521 dequantize_per_channel_default_12" [id=521, type=dequantize_per_channel]; +"522 linear_11" [id=522, type=linear]; +"523 dropout_7" [id=523, type=dropout]; +"524 layer_norm_4" [id=524, type=layer_norm]; +"525 add_6" [id=525, type=add]; +"526 pad_2" [id=526, type=pad]; +"527 slice_25" [id=527, type=slice]; +"528 slice_26" [id=528, type=slice]; +"529 slice_27" [id=529, type=slice]; +"530 slice_28" [id=530, type=slice]; +"531 slice_29" [id=531, type=slice]; +"532 slice_30" [id=532, type=slice]; +"533 slice_31" [id=533, type=slice]; +"534 slice_32" [id=534, type=slice]; +"535 slice_33" [id=535, type=slice]; +"536 slice_34" [id=536, type=slice]; +"537 slice_35" [id=537, type=slice]; +"538 slice_36" [id=538, type=slice]; +"539 cat" [id=539, type=cat]; +"540 cat_0_0_nncf_smooth_quant_0" [id=540, type=call_module]; +"541 quantize_per_tensor_default_13" [id=541, type=quantize_per_tensor]; +"542 dequantize_per_tensor_default_13" [id=542, type=dequantize_per_tensor]; +"543 linear_12_scale_0" [id=543, type=get_attr]; +"544 linear_12_zero_point_0" [id=544, type=get_attr]; +"545 _frozen_param13" [id=545, type=get_attr]; +"546 dequantize_per_channel_default_13" [id=546, type=dequantize_per_channel]; +"547 linear_12" [id=547, type=linear]; +"548 layer_norm_5" [id=548, type=layer_norm]; +"549 features_3_0_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=549, type=call_module]; +"550 linear_13_scale_0" [id=550, type=get_attr]; +"551 linear_13_zero_point_0" [id=551, type=get_attr]; +"552 _frozen_param14" [id=552, type=get_attr]; +"553 dequantize_per_channel_default_14" [id=553, type=dequantize_per_channel]; +"554 linear_13" [id=554, type=linear]; +"555 relu__2" [id=555, type=relu_]; +"556 relu__2_0_0_nncf_smooth_quant_0" [id=556, type=call_module]; +"557 linear_14_scale_0" [id=557, type=get_attr]; +"558 linear_14_zero_point_0" [id=558, type=get_attr]; +"559 _frozen_param15" [id=559, type=get_attr]; +"560 dequantize_per_channel_default_15" [id=560, type=dequantize_per_channel]; +"561 linear_14" [id=561, type=linear]; +"562 view_11" [id=562, type=view]; +"563 index_2" [id=563, type=index]; +"564 view_12" [id=564, type=view]; +"565 permute_10" [id=565, type=permute]; +"566 contiguous_2" [id=566, type=contiguous]; +"567 unsqueeze_6" [id=567, type=unsqueeze]; +"568 sigmoid_2" [id=568, type=sigmoid]; +"569 mul_4" [id=569, type=mul]; +"570 pad_3" [id=570, type=pad]; +"571 view_13" [id=571, type=view]; +"572 permute_11" [id=572, type=permute]; +"573 reshape_9" [id=573, type=reshape]; +"574 clone_2" [id=574, type=clone]; +"575 reshape_9_0_0_nncf_smooth_quant_0" [id=575, type=call_module]; +"576 quantize_per_tensor_default_14" [id=576, type=quantize_per_tensor]; +"577 dequantize_per_tensor_default_14" [id=577, type=dequantize_per_tensor]; +"578 linear_15_scale_0" [id=578, type=get_attr]; +"579 linear_15_zero_point_0" [id=579, type=get_attr]; +"580 _frozen_param16" [id=580, type=get_attr]; +"581 dequantize_per_channel_default_16" [id=581, type=dequantize_per_channel]; +"582 linear_15" [id=582, type=linear]; +"583 reshape_10" [id=583, type=reshape]; +"584 permute_12" [id=584, type=permute]; +"585 select_6" [id=585, type=select]; +"586 select_7" [id=586, type=select]; +"587 select_8" [id=587, type=select]; +"588 linalg_vector_norm_4" [id=588, type=linalg_vector_norm]; +"589 clamp_min_4" [id=589, type=clamp_min]; +"590 expand_as_4" [id=590, type=expand_as]; +"591 div_4" [id=591, type=div]; +"592 quantize_per_tensor_default_15" [id=592, type=quantize_per_tensor]; +"593 dequantize_per_tensor_default_15" [id=593, type=dequantize_per_tensor]; +"594 linalg_vector_norm_5" [id=594, type=linalg_vector_norm]; +"595 clamp_min_5" [id=595, type=clamp_min]; +"596 expand_as_5" [id=596, type=expand_as]; +"597 div_5" [id=597, type=div]; +"598 quantize_per_tensor_default_16" [id=598, type=quantize_per_tensor]; +"599 dequantize_per_tensor_default_16" [id=599, type=dequantize_per_tensor]; +"600 transpose_4" [id=600, type=transpose]; +"601 matmul_4" [id=601, type=matmul]; +"602 clamp_2" [id=602, type=clamp]; +"603 exp_2" [id=603, type=exp]; +"604 mul_5" [id=604, type=mul]; +"605 add_7" [id=605, type=add]; +"606 softmax_2" [id=606, type=softmax]; +"607 dropout_8" [id=607, type=dropout]; +"608 matmul_5" [id=608, type=matmul]; +"609 transpose_5" [id=609, type=transpose]; +"610 reshape_11" [id=610, type=reshape]; +"611 reshape_11_0_0_nncf_smooth_quant_0" [id=611, type=call_module]; +"612 quantize_per_tensor_default_17" [id=612, type=quantize_per_tensor]; +"613 dequantize_per_tensor_default_17" [id=613, type=dequantize_per_tensor]; +"614 linear_16_scale_0" [id=614, type=get_attr]; +"615 linear_16_zero_point_0" [id=615, type=get_attr]; +"616 _frozen_param17" [id=616, type=get_attr]; +"617 dequantize_per_channel_default_17" [id=617, type=dequantize_per_channel]; +"618 linear_16" [id=618, type=linear]; +"619 dropout_9" [id=619, type=dropout]; +"620 view_14" [id=620, type=view]; +"621 permute_13" [id=621, type=permute]; +"622 reshape_12" [id=622, type=reshape]; +"623 slice_38" [id=623, type=slice]; +"624 slice_39" [id=624, type=slice]; +"625 slice_40" [id=625, type=slice]; +"626 slice_41" [id=626, type=slice]; +"627 contiguous_3" [id=627, type=contiguous]; +"628 layer_norm_6" [id=628, type=layer_norm]; +"629 add_8" [id=629, type=add]; +"630 add_8_0_0_nncf_smooth_quant_0" [id=630, type=call_module]; +"631 quantize_per_tensor_default_18" [id=631, type=quantize_per_tensor]; +"632 dequantize_per_tensor_default_18" [id=632, type=dequantize_per_tensor]; +"633 linear_17_scale_0" [id=633, type=get_attr]; +"634 linear_17_zero_point_0" [id=634, type=get_attr]; +"635 _frozen_param18" [id=635, type=get_attr]; +"636 dequantize_per_channel_default_18" [id=636, type=dequantize_per_channel]; +"637 linear_17" [id=637, type=linear]; +"638 gelu_2" [id=638, type=gelu]; +"639 dropout_10" [id=639, type=dropout]; +"640 dropout_10_0_0_nncf_smooth_quant_0" [id=640, type=call_module]; +"641 quantize_per_tensor_default_19" [id=641, type=quantize_per_tensor]; +"642 dequantize_per_tensor_default_19" [id=642, type=dequantize_per_tensor]; +"643 linear_18_scale_0" [id=643, type=get_attr]; +"644 linear_18_zero_point_0" [id=644, type=get_attr]; +"645 _frozen_param19" [id=645, type=get_attr]; +"646 dequantize_per_channel_default_19" [id=646, type=dequantize_per_channel]; +"647 linear_18" [id=647, type=linear]; +"648 dropout_11" [id=648, type=dropout]; +"649 layer_norm_7" [id=649, type=layer_norm]; +"650 add_9" [id=650, type=add]; +"651 features_3_1_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=651, type=call_module]; +"652 linear_19_scale_0" [id=652, type=get_attr]; +"653 linear_19_zero_point_0" [id=653, type=get_attr]; +"654 _frozen_param20" [id=654, type=get_attr]; +"655 dequantize_per_channel_default_20" [id=655, type=dequantize_per_channel]; +"656 linear_19" [id=656, type=linear]; +"657 relu__3" [id=657, type=relu_]; +"658 relu__3_0_0_nncf_smooth_quant_0" [id=658, type=call_module]; +"659 linear_20_scale_0" [id=659, type=get_attr]; +"660 linear_20_zero_point_0" [id=660, type=get_attr]; +"661 _frozen_param21" [id=661, type=get_attr]; +"662 dequantize_per_channel_default_21" [id=662, type=dequantize_per_channel]; +"663 linear_20" [id=663, type=linear]; +"664 view_15" [id=664, type=view]; +"665 index_3" [id=665, type=index]; +"666 view_16" [id=666, type=view]; +"667 permute_14" [id=667, type=permute]; +"668 contiguous_4" [id=668, type=contiguous]; +"669 unsqueeze_7" [id=669, type=unsqueeze]; +"670 sigmoid_3" [id=670, type=sigmoid]; +"671 mul_6" [id=671, type=mul]; +"672 pad_4" [id=672, type=pad]; +"673 roll_2" [id=673, type=roll]; +"674 view_17" [id=674, type=view]; +"675 permute_15" [id=675, type=permute]; +"676 reshape_13" [id=676, type=reshape]; +"677 clone_3" [id=677, type=clone]; +"678 reshape_13_0_0_nncf_smooth_quant_0" [id=678, type=call_module]; +"679 quantize_per_tensor_default_20" [id=679, type=quantize_per_tensor]; +"680 dequantize_per_tensor_default_20" [id=680, type=dequantize_per_tensor]; +"681 linear_21_scale_0" [id=681, type=get_attr]; +"682 linear_21_zero_point_0" [id=682, type=get_attr]; +"683 _frozen_param22" [id=683, type=get_attr]; +"684 dequantize_per_channel_default_22" [id=684, type=dequantize_per_channel]; +"685 linear_21" [id=685, type=linear]; +"686 reshape_14" [id=686, type=reshape]; +"687 permute_16" [id=687, type=permute]; +"688 select_9" [id=688, type=select]; +"689 select_10" [id=689, type=select]; +"690 select_11" [id=690, type=select]; +"691 linalg_vector_norm_6" [id=691, type=linalg_vector_norm]; +"692 clamp_min_6" [id=692, type=clamp_min]; +"693 expand_as_6" [id=693, type=expand_as]; +"694 div_6" [id=694, type=div]; +"695 quantize_per_tensor_default_21" [id=695, type=quantize_per_tensor]; +"696 dequantize_per_tensor_default_21" [id=696, type=dequantize_per_tensor]; +"697 linalg_vector_norm_7" [id=697, type=linalg_vector_norm]; +"698 clamp_min_7" [id=698, type=clamp_min]; +"699 expand_as_7" [id=699, type=expand_as]; +"700 div_7" [id=700, type=div]; +"701 quantize_per_tensor_default_22" [id=701, type=quantize_per_tensor]; +"702 dequantize_per_tensor_default_22" [id=702, type=dequantize_per_tensor]; +"703 transpose_6" [id=703, type=transpose]; +"704 matmul_6" [id=704, type=matmul]; +"705 clamp_3" [id=705, type=clamp]; +"706 exp_3" [id=706, type=exp]; +"707 mul_7" [id=707, type=mul]; +"708 add_10" [id=708, type=add]; +"709 new_zeros_1" [id=709, type=new_zeros]; +"710 view_18" [id=710, type=view]; +"711 permute_17" [id=711, type=permute]; +"712 reshape_15" [id=712, type=reshape]; +"713 unsqueeze_8" [id=713, type=unsqueeze]; +"714 unsqueeze_9" [id=714, type=unsqueeze]; +"715 sub_1" [id=715, type=sub]; +"716 ne_1" [id=716, type=ne]; +"717 masked_fill_2" [id=717, type=masked_fill]; +"718 eq_1" [id=718, type=eq]; +"719 masked_fill_3" [id=719, type=masked_fill]; +"720 view_19" [id=720, type=view]; +"721 unsqueeze_10" [id=721, type=unsqueeze]; +"722 unsqueeze_11" [id=722, type=unsqueeze]; +"723 add_11" [id=723, type=add]; +"724 view_20" [id=724, type=view]; +"725 softmax_3" [id=725, type=softmax]; +"726 dropout_12" [id=726, type=dropout]; +"727 matmul_7" [id=727, type=matmul]; +"728 transpose_7" [id=728, type=transpose]; +"729 reshape_16" [id=729, type=reshape]; +"730 reshape_16_0_0_nncf_smooth_quant_0" [id=730, type=call_module]; +"731 quantize_per_tensor_default_23" [id=731, type=quantize_per_tensor]; +"732 dequantize_per_tensor_default_23" [id=732, type=dequantize_per_tensor]; +"733 linear_22_scale_0" [id=733, type=get_attr]; +"734 linear_22_zero_point_0" [id=734, type=get_attr]; +"735 _frozen_param23" [id=735, type=get_attr]; +"736 dequantize_per_channel_default_23" [id=736, type=dequantize_per_channel]; +"737 linear_22" [id=737, type=linear]; +"738 dropout_13" [id=738, type=dropout]; +"739 view_21" [id=739, type=view]; +"740 permute_18" [id=740, type=permute]; +"741 reshape_17" [id=741, type=reshape]; +"742 roll_3" [id=742, type=roll]; +"743 slice_61" [id=743, type=slice]; +"744 slice_62" [id=744, type=slice]; +"745 slice_63" [id=745, type=slice]; +"746 slice_64" [id=746, type=slice]; +"747 contiguous_5" [id=747, type=contiguous]; +"748 layer_norm_8" [id=748, type=layer_norm]; +"749 add_12" [id=749, type=add]; +"750 add_12_0_0_nncf_smooth_quant_0" [id=750, type=call_module]; +"751 quantize_per_tensor_default_24" [id=751, type=quantize_per_tensor]; +"752 dequantize_per_tensor_default_24" [id=752, type=dequantize_per_tensor]; +"753 linear_23_scale_0" [id=753, type=get_attr]; +"754 linear_23_zero_point_0" [id=754, type=get_attr]; +"755 _frozen_param24" [id=755, type=get_attr]; +"756 dequantize_per_channel_default_24" [id=756, type=dequantize_per_channel]; +"757 linear_23" [id=757, type=linear]; +"758 gelu_3" [id=758, type=gelu]; +"759 dropout_14" [id=759, type=dropout]; +"760 dropout_14_0_0_nncf_smooth_quant_0" [id=760, type=call_module]; +"761 quantize_per_tensor_default_25" [id=761, type=quantize_per_tensor]; +"762 dequantize_per_tensor_default_25" [id=762, type=dequantize_per_tensor]; +"763 linear_24_scale_0" [id=763, type=get_attr]; +"764 linear_24_zero_point_0" [id=764, type=get_attr]; +"765 _frozen_param25" [id=765, type=get_attr]; +"766 dequantize_per_channel_default_25" [id=766, type=dequantize_per_channel]; +"767 linear_24" [id=767, type=linear]; +"768 dropout_15" [id=768, type=dropout]; +"769 layer_norm_9" [id=769, type=layer_norm]; +"770 add_13" [id=770, type=add]; +"771 pad_5" [id=771, type=pad]; +"772 slice_65" [id=772, type=slice]; +"773 slice_66" [id=773, type=slice]; +"774 slice_67" [id=774, type=slice]; +"775 slice_68" [id=775, type=slice]; +"776 slice_69" [id=776, type=slice]; +"777 slice_70" [id=777, type=slice]; +"778 slice_71" [id=778, type=slice]; +"779 slice_72" [id=779, type=slice]; +"780 slice_73" [id=780, type=slice]; +"781 slice_74" [id=781, type=slice]; +"782 slice_75" [id=782, type=slice]; +"783 slice_76" [id=783, type=slice]; +"784 cat_1" [id=784, type=cat]; +"785 cat_1_0_0_nncf_smooth_quant_0" [id=785, type=call_module]; +"786 quantize_per_tensor_default_26" [id=786, type=quantize_per_tensor]; +"787 dequantize_per_tensor_default_26" [id=787, type=dequantize_per_tensor]; +"788 linear_25_scale_0" [id=788, type=get_attr]; +"789 linear_25_zero_point_0" [id=789, type=get_attr]; +"790 _frozen_param26" [id=790, type=get_attr]; +"791 dequantize_per_channel_default_26" [id=791, type=dequantize_per_channel]; +"792 linear_25" [id=792, type=linear]; +"793 layer_norm_10" [id=793, type=layer_norm]; +"794 features_5_0_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=794, type=call_module]; +"795 linear_26_scale_0" [id=795, type=get_attr]; +"796 linear_26_zero_point_0" [id=796, type=get_attr]; +"797 _frozen_param27" [id=797, type=get_attr]; +"798 dequantize_per_channel_default_27" [id=798, type=dequantize_per_channel]; +"799 linear_26" [id=799, type=linear]; +"800 relu__4" [id=800, type=relu_]; +"801 relu__4_0_0_nncf_smooth_quant_0" [id=801, type=call_module]; +"802 linear_27_scale_0" [id=802, type=get_attr]; +"803 linear_27_zero_point_0" [id=803, type=get_attr]; +"804 _frozen_param28" [id=804, type=get_attr]; +"805 dequantize_per_channel_default_28" [id=805, type=dequantize_per_channel]; +"806 linear_27" [id=806, type=linear]; +"807 view_22" [id=807, type=view]; +"808 index_4" [id=808, type=index]; +"809 view_23" [id=809, type=view]; +"810 permute_19" [id=810, type=permute]; +"811 contiguous_6" [id=811, type=contiguous]; +"812 unsqueeze_12" [id=812, type=unsqueeze]; +"813 sigmoid_4" [id=813, type=sigmoid]; +"814 mul_8" [id=814, type=mul]; +"815 pad_6" [id=815, type=pad]; +"816 view_24" [id=816, type=view]; +"817 permute_20" [id=817, type=permute]; +"818 reshape_18" [id=818, type=reshape]; +"819 clone_4" [id=819, type=clone]; +"820 reshape_18_0_0_nncf_smooth_quant_0" [id=820, type=call_module]; +"821 quantize_per_tensor_default_27" [id=821, type=quantize_per_tensor]; +"822 dequantize_per_tensor_default_27" [id=822, type=dequantize_per_tensor]; +"823 linear_28_scale_0" [id=823, type=get_attr]; +"824 linear_28_zero_point_0" [id=824, type=get_attr]; +"825 _frozen_param29" [id=825, type=get_attr]; +"826 dequantize_per_channel_default_29" [id=826, type=dequantize_per_channel]; +"827 linear_28" [id=827, type=linear]; +"828 reshape_19" [id=828, type=reshape]; +"829 permute_21" [id=829, type=permute]; +"830 select_12" [id=830, type=select]; +"831 select_13" [id=831, type=select]; +"832 select_14" [id=832, type=select]; +"833 linalg_vector_norm_8" [id=833, type=linalg_vector_norm]; +"834 clamp_min_8" [id=834, type=clamp_min]; +"835 expand_as_8" [id=835, type=expand_as]; +"836 div_8" [id=836, type=div]; +"837 quantize_per_tensor_default_28" [id=837, type=quantize_per_tensor]; +"838 dequantize_per_tensor_default_28" [id=838, type=dequantize_per_tensor]; +"839 linalg_vector_norm_9" [id=839, type=linalg_vector_norm]; +"840 clamp_min_9" [id=840, type=clamp_min]; +"841 expand_as_9" [id=841, type=expand_as]; +"842 div_9" [id=842, type=div]; +"843 quantize_per_tensor_default_29" [id=843, type=quantize_per_tensor]; +"844 dequantize_per_tensor_default_29" [id=844, type=dequantize_per_tensor]; +"845 transpose_8" [id=845, type=transpose]; +"846 matmul_8" [id=846, type=matmul]; +"847 clamp_4" [id=847, type=clamp]; +"848 exp_4" [id=848, type=exp]; +"849 mul_9" [id=849, type=mul]; +"850 add_14" [id=850, type=add]; +"851 softmax_4" [id=851, type=softmax]; +"852 dropout_16" [id=852, type=dropout]; +"853 matmul_9" [id=853, type=matmul]; +"854 transpose_9" [id=854, type=transpose]; +"855 reshape_20" [id=855, type=reshape]; +"856 reshape_20_0_0_nncf_smooth_quant_0" [id=856, type=call_module]; +"857 quantize_per_tensor_default_30" [id=857, type=quantize_per_tensor]; +"858 dequantize_per_tensor_default_30" [id=858, type=dequantize_per_tensor]; +"859 linear_29_scale_0" [id=859, type=get_attr]; +"860 linear_29_zero_point_0" [id=860, type=get_attr]; +"861 _frozen_param30" [id=861, type=get_attr]; +"862 dequantize_per_channel_default_30" [id=862, type=dequantize_per_channel]; +"863 linear_29" [id=863, type=linear]; +"864 dropout_17" [id=864, type=dropout]; +"865 view_25" [id=865, type=view]; +"866 permute_22" [id=866, type=permute]; +"867 reshape_21" [id=867, type=reshape]; +"868 slice_78" [id=868, type=slice]; +"869 slice_79" [id=869, type=slice]; +"870 slice_80" [id=870, type=slice]; +"871 slice_81" [id=871, type=slice]; +"872 contiguous_7" [id=872, type=contiguous]; +"873 layer_norm_11" [id=873, type=layer_norm]; +"874 add_15" [id=874, type=add]; +"875 add_15_0_0_nncf_smooth_quant_0" [id=875, type=call_module]; +"876 quantize_per_tensor_default_31" [id=876, type=quantize_per_tensor]; +"877 dequantize_per_tensor_default_31" [id=877, type=dequantize_per_tensor]; +"878 linear_30_scale_0" [id=878, type=get_attr]; +"879 linear_30_zero_point_0" [id=879, type=get_attr]; +"880 _frozen_param31" [id=880, type=get_attr]; +"881 dequantize_per_channel_default_31" [id=881, type=dequantize_per_channel]; +"882 linear_30" [id=882, type=linear]; +"883 gelu_4" [id=883, type=gelu]; +"884 dropout_18" [id=884, type=dropout]; +"885 dropout_18_0_0_nncf_smooth_quant_0" [id=885, type=call_module]; +"886 quantize_per_tensor_default_32" [id=886, type=quantize_per_tensor]; +"887 dequantize_per_tensor_default_32" [id=887, type=dequantize_per_tensor]; +"888 linear_31_scale_0" [id=888, type=get_attr]; +"889 linear_31_zero_point_0" [id=889, type=get_attr]; +"890 _frozen_param32" [id=890, type=get_attr]; +"891 dequantize_per_channel_default_32" [id=891, type=dequantize_per_channel]; +"892 linear_31" [id=892, type=linear]; +"893 dropout_19" [id=893, type=dropout]; +"894 layer_norm_12" [id=894, type=layer_norm]; +"895 add_16" [id=895, type=add]; +"896 features_5_1_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=896, type=call_module]; +"897 linear_32_scale_0" [id=897, type=get_attr]; +"898 linear_32_zero_point_0" [id=898, type=get_attr]; +"899 _frozen_param33" [id=899, type=get_attr]; +"900 dequantize_per_channel_default_33" [id=900, type=dequantize_per_channel]; +"901 linear_32" [id=901, type=linear]; +"902 relu__5" [id=902, type=relu_]; +"903 relu__5_0_0_nncf_smooth_quant_0" [id=903, type=call_module]; +"904 linear_33_scale_0" [id=904, type=get_attr]; +"905 linear_33_zero_point_0" [id=905, type=get_attr]; +"906 _frozen_param34" [id=906, type=get_attr]; +"907 dequantize_per_channel_default_34" [id=907, type=dequantize_per_channel]; +"908 linear_33" [id=908, type=linear]; +"909 view_26" [id=909, type=view]; +"910 index_5" [id=910, type=index]; +"911 view_27" [id=911, type=view]; +"912 permute_23" [id=912, type=permute]; +"913 contiguous_8" [id=913, type=contiguous]; +"914 unsqueeze_13" [id=914, type=unsqueeze]; +"915 sigmoid_5" [id=915, type=sigmoid]; +"916 mul_10" [id=916, type=mul]; +"917 pad_7" [id=917, type=pad]; +"918 roll_4" [id=918, type=roll]; +"919 view_28" [id=919, type=view]; +"920 permute_24" [id=920, type=permute]; +"921 reshape_22" [id=921, type=reshape]; +"922 clone_5" [id=922, type=clone]; +"923 reshape_22_0_0_nncf_smooth_quant_0" [id=923, type=call_module]; +"924 quantize_per_tensor_default_33" [id=924, type=quantize_per_tensor]; +"925 dequantize_per_tensor_default_33" [id=925, type=dequantize_per_tensor]; +"926 linear_34_scale_0" [id=926, type=get_attr]; +"927 linear_34_zero_point_0" [id=927, type=get_attr]; +"928 _frozen_param35" [id=928, type=get_attr]; +"929 dequantize_per_channel_default_35" [id=929, type=dequantize_per_channel]; +"930 linear_34" [id=930, type=linear]; +"931 reshape_23" [id=931, type=reshape]; +"932 permute_25" [id=932, type=permute]; +"933 select_15" [id=933, type=select]; +"934 select_16" [id=934, type=select]; +"935 select_17" [id=935, type=select]; +"936 linalg_vector_norm_10" [id=936, type=linalg_vector_norm]; +"937 clamp_min_10" [id=937, type=clamp_min]; +"938 expand_as_10" [id=938, type=expand_as]; +"939 div_10" [id=939, type=div]; +"940 quantize_per_tensor_default_34" [id=940, type=quantize_per_tensor]; +"941 dequantize_per_tensor_default_34" [id=941, type=dequantize_per_tensor]; +"942 linalg_vector_norm_11" [id=942, type=linalg_vector_norm]; +"943 clamp_min_11" [id=943, type=clamp_min]; +"944 expand_as_11" [id=944, type=expand_as]; +"945 div_11" [id=945, type=div]; +"946 quantize_per_tensor_default_35" [id=946, type=quantize_per_tensor]; +"947 dequantize_per_tensor_default_35" [id=947, type=dequantize_per_tensor]; +"948 transpose_10" [id=948, type=transpose]; +"949 matmul_10" [id=949, type=matmul]; +"950 clamp_5" [id=950, type=clamp]; +"951 exp_5" [id=951, type=exp]; +"952 mul_11" [id=952, type=mul]; +"953 add_17" [id=953, type=add]; +"954 new_zeros_2" [id=954, type=new_zeros]; +"955 view_29" [id=955, type=view]; +"956 permute_26" [id=956, type=permute]; +"957 reshape_24" [id=957, type=reshape]; +"958 unsqueeze_14" [id=958, type=unsqueeze]; +"959 unsqueeze_15" [id=959, type=unsqueeze]; +"960 sub_2" [id=960, type=sub]; +"961 ne_2" [id=961, type=ne]; +"962 masked_fill_4" [id=962, type=masked_fill]; +"963 eq_2" [id=963, type=eq]; +"964 masked_fill_5" [id=964, type=masked_fill]; +"965 view_30" [id=965, type=view]; +"966 unsqueeze_16" [id=966, type=unsqueeze]; +"967 unsqueeze_17" [id=967, type=unsqueeze]; +"968 add_18" [id=968, type=add]; +"969 view_31" [id=969, type=view]; +"970 softmax_5" [id=970, type=softmax]; +"971 dropout_20" [id=971, type=dropout]; +"972 matmul_11" [id=972, type=matmul]; +"973 transpose_11" [id=973, type=transpose]; +"974 reshape_25" [id=974, type=reshape]; +"975 reshape_25_0_0_nncf_smooth_quant_0" [id=975, type=call_module]; +"976 quantize_per_tensor_default_36" [id=976, type=quantize_per_tensor]; +"977 dequantize_per_tensor_default_36" [id=977, type=dequantize_per_tensor]; +"978 linear_35_scale_0" [id=978, type=get_attr]; +"979 linear_35_zero_point_0" [id=979, type=get_attr]; +"980 _frozen_param36" [id=980, type=get_attr]; +"981 dequantize_per_channel_default_36" [id=981, type=dequantize_per_channel]; +"982 linear_35" [id=982, type=linear]; +"983 dropout_21" [id=983, type=dropout]; +"984 view_32" [id=984, type=view]; +"985 permute_27" [id=985, type=permute]; +"986 reshape_26" [id=986, type=reshape]; +"987 roll_5" [id=987, type=roll]; +"988 slice_101" [id=988, type=slice]; +"989 slice_102" [id=989, type=slice]; +"990 slice_103" [id=990, type=slice]; +"991 slice_104" [id=991, type=slice]; +"992 contiguous_9" [id=992, type=contiguous]; +"993 layer_norm_13" [id=993, type=layer_norm]; +"994 add_19" [id=994, type=add]; +"995 add_19_0_0_nncf_smooth_quant_0" [id=995, type=call_module]; +"996 quantize_per_tensor_default_37" [id=996, type=quantize_per_tensor]; +"997 dequantize_per_tensor_default_37" [id=997, type=dequantize_per_tensor]; +"998 linear_36_scale_0" [id=998, type=get_attr]; +"999 linear_36_zero_point_0" [id=999, type=get_attr]; +"1000 _frozen_param37" [id=1000, type=get_attr]; +"1001 dequantize_per_channel_default_37" [id=1001, type=dequantize_per_channel]; +"1002 linear_36" [id=1002, type=linear]; +"1003 gelu_5" [id=1003, type=gelu]; +"1004 dropout_22" [id=1004, type=dropout]; +"1005 dropout_22_0_0_nncf_smooth_quant_0" [id=1005, type=call_module]; +"1006 quantize_per_tensor_default_38" [id=1006, type=quantize_per_tensor]; +"1007 dequantize_per_tensor_default_38" [id=1007, type=dequantize_per_tensor]; +"1008 linear_37_scale_0" [id=1008, type=get_attr]; +"1009 linear_37_zero_point_0" [id=1009, type=get_attr]; +"1010 _frozen_param38" [id=1010, type=get_attr]; +"1011 dequantize_per_channel_default_38" [id=1011, type=dequantize_per_channel]; +"1012 linear_37" [id=1012, type=linear]; +"1013 dropout_23" [id=1013, type=dropout]; +"1014 layer_norm_14" [id=1014, type=layer_norm]; +"1015 add_20" [id=1015, type=add]; +"1016 features_5_2_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=1016, type=call_module]; +"1017 linear_38_scale_0" [id=1017, type=get_attr]; +"1018 linear_38_zero_point_0" [id=1018, type=get_attr]; +"1019 _frozen_param39" [id=1019, type=get_attr]; +"1020 dequantize_per_channel_default_39" [id=1020, type=dequantize_per_channel]; +"1021 linear_38" [id=1021, type=linear]; +"1022 relu__6" [id=1022, type=relu_]; +"1023 relu__6_0_0_nncf_smooth_quant_0" [id=1023, type=call_module]; +"1024 linear_39_scale_0" [id=1024, type=get_attr]; +"1025 linear_39_zero_point_0" [id=1025, type=get_attr]; +"1026 _frozen_param40" [id=1026, type=get_attr]; +"1027 dequantize_per_channel_default_40" [id=1027, type=dequantize_per_channel]; +"1028 linear_39" [id=1028, type=linear]; +"1029 view_33" [id=1029, type=view]; +"1030 index_6" [id=1030, type=index]; +"1031 view_34" [id=1031, type=view]; +"1032 permute_28" [id=1032, type=permute]; +"1033 contiguous_10" [id=1033, type=contiguous]; +"1034 unsqueeze_18" [id=1034, type=unsqueeze]; +"1035 sigmoid_6" [id=1035, type=sigmoid]; +"1036 mul_12" [id=1036, type=mul]; +"1037 pad_8" [id=1037, type=pad]; +"1038 view_35" [id=1038, type=view]; +"1039 permute_29" [id=1039, type=permute]; +"1040 reshape_27" [id=1040, type=reshape]; +"1041 clone_6" [id=1041, type=clone]; +"1042 reshape_27_0_0_nncf_smooth_quant_0" [id=1042, type=call_module]; +"1043 quantize_per_tensor_default_39" [id=1043, type=quantize_per_tensor]; +"1044 dequantize_per_tensor_default_39" [id=1044, type=dequantize_per_tensor]; +"1045 linear_40_scale_0" [id=1045, type=get_attr]; +"1046 linear_40_zero_point_0" [id=1046, type=get_attr]; +"1047 _frozen_param41" [id=1047, type=get_attr]; +"1048 dequantize_per_channel_default_41" [id=1048, type=dequantize_per_channel]; +"1049 linear_40" [id=1049, type=linear]; +"1050 reshape_28" [id=1050, type=reshape]; +"1051 permute_30" [id=1051, type=permute]; +"1052 select_18" [id=1052, type=select]; +"1053 select_19" [id=1053, type=select]; +"1054 select_20" [id=1054, type=select]; +"1055 linalg_vector_norm_12" [id=1055, type=linalg_vector_norm]; +"1056 clamp_min_12" [id=1056, type=clamp_min]; +"1057 expand_as_12" [id=1057, type=expand_as]; +"1058 div_12" [id=1058, type=div]; +"1059 quantize_per_tensor_default_40" [id=1059, type=quantize_per_tensor]; +"1060 dequantize_per_tensor_default_40" [id=1060, type=dequantize_per_tensor]; +"1061 linalg_vector_norm_13" [id=1061, type=linalg_vector_norm]; +"1062 clamp_min_13" [id=1062, type=clamp_min]; +"1063 expand_as_13" [id=1063, type=expand_as]; +"1064 div_13" [id=1064, type=div]; +"1065 quantize_per_tensor_default_41" [id=1065, type=quantize_per_tensor]; +"1066 dequantize_per_tensor_default_41" [id=1066, type=dequantize_per_tensor]; +"1067 transpose_12" [id=1067, type=transpose]; +"1068 matmul_12" [id=1068, type=matmul]; +"1069 clamp_6" [id=1069, type=clamp]; +"1070 exp_6" [id=1070, type=exp]; +"1071 mul_13" [id=1071, type=mul]; +"1072 add_21" [id=1072, type=add]; +"1073 softmax_6" [id=1073, type=softmax]; +"1074 dropout_24" [id=1074, type=dropout]; +"1075 matmul_13" [id=1075, type=matmul]; +"1076 transpose_13" [id=1076, type=transpose]; +"1077 reshape_29" [id=1077, type=reshape]; +"1078 reshape_29_0_0_nncf_smooth_quant_0" [id=1078, type=call_module]; +"1079 quantize_per_tensor_default_42" [id=1079, type=quantize_per_tensor]; +"1080 dequantize_per_tensor_default_42" [id=1080, type=dequantize_per_tensor]; +"1081 linear_41_scale_0" [id=1081, type=get_attr]; +"1082 linear_41_zero_point_0" [id=1082, type=get_attr]; +"1083 _frozen_param42" [id=1083, type=get_attr]; +"1084 dequantize_per_channel_default_42" [id=1084, type=dequantize_per_channel]; +"1085 linear_41" [id=1085, type=linear]; +"1086 dropout_25" [id=1086, type=dropout]; +"1087 view_36" [id=1087, type=view]; +"1088 permute_31" [id=1088, type=permute]; +"1089 reshape_30" [id=1089, type=reshape]; +"1090 slice_106" [id=1090, type=slice]; +"1091 slice_107" [id=1091, type=slice]; +"1092 slice_108" [id=1092, type=slice]; +"1093 slice_109" [id=1093, type=slice]; +"1094 contiguous_11" [id=1094, type=contiguous]; +"1095 layer_norm_15" [id=1095, type=layer_norm]; +"1096 add_22" [id=1096, type=add]; +"1097 add_22_0_0_nncf_smooth_quant_0" [id=1097, type=call_module]; +"1098 quantize_per_tensor_default_43" [id=1098, type=quantize_per_tensor]; +"1099 dequantize_per_tensor_default_43" [id=1099, type=dequantize_per_tensor]; +"1100 linear_42_scale_0" [id=1100, type=get_attr]; +"1101 linear_42_zero_point_0" [id=1101, type=get_attr]; +"1102 _frozen_param43" [id=1102, type=get_attr]; +"1103 dequantize_per_channel_default_43" [id=1103, type=dequantize_per_channel]; +"1104 linear_42" [id=1104, type=linear]; +"1105 gelu_6" [id=1105, type=gelu]; +"1106 dropout_26" [id=1106, type=dropout]; +"1107 dropout_26_0_0_nncf_smooth_quant_0" [id=1107, type=call_module]; +"1108 quantize_per_tensor_default_44" [id=1108, type=quantize_per_tensor]; +"1109 dequantize_per_tensor_default_44" [id=1109, type=dequantize_per_tensor]; +"1110 linear_43_scale_0" [id=1110, type=get_attr]; +"1111 linear_43_zero_point_0" [id=1111, type=get_attr]; +"1112 _frozen_param44" [id=1112, type=get_attr]; +"1113 dequantize_per_channel_default_44" [id=1113, type=dequantize_per_channel]; +"1114 linear_43" [id=1114, type=linear]; +"1115 dropout_27" [id=1115, type=dropout]; +"1116 layer_norm_16" [id=1116, type=layer_norm]; +"1117 add_23" [id=1117, type=add]; +"1118 features_5_3_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=1118, type=call_module]; +"1119 linear_44_scale_0" [id=1119, type=get_attr]; +"1120 linear_44_zero_point_0" [id=1120, type=get_attr]; +"1121 _frozen_param45" [id=1121, type=get_attr]; +"1122 dequantize_per_channel_default_45" [id=1122, type=dequantize_per_channel]; +"1123 linear_44" [id=1123, type=linear]; +"1124 relu__7" [id=1124, type=relu_]; +"1125 relu__7_0_0_nncf_smooth_quant_0" [id=1125, type=call_module]; +"1126 linear_45_scale_0" [id=1126, type=get_attr]; +"1127 linear_45_zero_point_0" [id=1127, type=get_attr]; +"1128 _frozen_param46" [id=1128, type=get_attr]; +"1129 dequantize_per_channel_default_46" [id=1129, type=dequantize_per_channel]; +"1130 linear_45" [id=1130, type=linear]; +"1131 view_37" [id=1131, type=view]; +"1132 index_7" [id=1132, type=index]; +"1133 view_38" [id=1133, type=view]; +"1134 permute_32" [id=1134, type=permute]; +"1135 contiguous_12" [id=1135, type=contiguous]; +"1136 unsqueeze_19" [id=1136, type=unsqueeze]; +"1137 sigmoid_7" [id=1137, type=sigmoid]; +"1138 mul_14" [id=1138, type=mul]; +"1139 pad_9" [id=1139, type=pad]; +"1140 roll_6" [id=1140, type=roll]; +"1141 view_39" [id=1141, type=view]; +"1142 permute_33" [id=1142, type=permute]; +"1143 reshape_31" [id=1143, type=reshape]; +"1144 clone_7" [id=1144, type=clone]; +"1145 reshape_31_0_0_nncf_smooth_quant_0" [id=1145, type=call_module]; +"1146 quantize_per_tensor_default_45" [id=1146, type=quantize_per_tensor]; +"1147 dequantize_per_tensor_default_45" [id=1147, type=dequantize_per_tensor]; +"1148 linear_46_scale_0" [id=1148, type=get_attr]; +"1149 linear_46_zero_point_0" [id=1149, type=get_attr]; +"1150 _frozen_param47" [id=1150, type=get_attr]; +"1151 dequantize_per_channel_default_47" [id=1151, type=dequantize_per_channel]; +"1152 linear_46" [id=1152, type=linear]; +"1153 reshape_32" [id=1153, type=reshape]; +"1154 permute_34" [id=1154, type=permute]; +"1155 select_21" [id=1155, type=select]; +"1156 select_22" [id=1156, type=select]; +"1157 select_23" [id=1157, type=select]; +"1158 linalg_vector_norm_14" [id=1158, type=linalg_vector_norm]; +"1159 clamp_min_14" [id=1159, type=clamp_min]; +"1160 expand_as_14" [id=1160, type=expand_as]; +"1161 div_14" [id=1161, type=div]; +"1162 quantize_per_tensor_default_46" [id=1162, type=quantize_per_tensor]; +"1163 dequantize_per_tensor_default_46" [id=1163, type=dequantize_per_tensor]; +"1164 linalg_vector_norm_15" [id=1164, type=linalg_vector_norm]; +"1165 clamp_min_15" [id=1165, type=clamp_min]; +"1166 expand_as_15" [id=1166, type=expand_as]; +"1167 div_15" [id=1167, type=div]; +"1168 quantize_per_tensor_default_47" [id=1168, type=quantize_per_tensor]; +"1169 dequantize_per_tensor_default_47" [id=1169, type=dequantize_per_tensor]; +"1170 transpose_14" [id=1170, type=transpose]; +"1171 matmul_14" [id=1171, type=matmul]; +"1172 clamp_7" [id=1172, type=clamp]; +"1173 exp_7" [id=1173, type=exp]; +"1174 mul_15" [id=1174, type=mul]; +"1175 add_24" [id=1175, type=add]; +"1176 new_zeros_3" [id=1176, type=new_zeros]; +"1177 view_40" [id=1177, type=view]; +"1178 permute_35" [id=1178, type=permute]; +"1179 reshape_33" [id=1179, type=reshape]; +"1180 unsqueeze_20" [id=1180, type=unsqueeze]; +"1181 unsqueeze_21" [id=1181, type=unsqueeze]; +"1182 sub_3" [id=1182, type=sub]; +"1183 ne_3" [id=1183, type=ne]; +"1184 masked_fill_6" [id=1184, type=masked_fill]; +"1185 eq_3" [id=1185, type=eq]; +"1186 masked_fill_7" [id=1186, type=masked_fill]; +"1187 view_41" [id=1187, type=view]; +"1188 unsqueeze_22" [id=1188, type=unsqueeze]; +"1189 unsqueeze_23" [id=1189, type=unsqueeze]; +"1190 add_25" [id=1190, type=add]; +"1191 view_42" [id=1191, type=view]; +"1192 softmax_7" [id=1192, type=softmax]; +"1193 dropout_28" [id=1193, type=dropout]; +"1194 matmul_15" [id=1194, type=matmul]; +"1195 transpose_15" [id=1195, type=transpose]; +"1196 reshape_34" [id=1196, type=reshape]; +"1197 reshape_34_0_0_nncf_smooth_quant_0" [id=1197, type=call_module]; +"1198 quantize_per_tensor_default_48" [id=1198, type=quantize_per_tensor]; +"1199 dequantize_per_tensor_default_48" [id=1199, type=dequantize_per_tensor]; +"1200 linear_47_scale_0" [id=1200, type=get_attr]; +"1201 linear_47_zero_point_0" [id=1201, type=get_attr]; +"1202 _frozen_param48" [id=1202, type=get_attr]; +"1203 dequantize_per_channel_default_48" [id=1203, type=dequantize_per_channel]; +"1204 linear_47" [id=1204, type=linear]; +"1205 dropout_29" [id=1205, type=dropout]; +"1206 view_43" [id=1206, type=view]; +"1207 permute_36" [id=1207, type=permute]; +"1208 reshape_35" [id=1208, type=reshape]; +"1209 roll_7" [id=1209, type=roll]; +"1210 slice_129" [id=1210, type=slice]; +"1211 slice_130" [id=1211, type=slice]; +"1212 slice_131" [id=1212, type=slice]; +"1213 slice_132" [id=1213, type=slice]; +"1214 contiguous_13" [id=1214, type=contiguous]; +"1215 layer_norm_17" [id=1215, type=layer_norm]; +"1216 add_26" [id=1216, type=add]; +"1217 add_26_0_0_nncf_smooth_quant_0" [id=1217, type=call_module]; +"1218 quantize_per_tensor_default_49" [id=1218, type=quantize_per_tensor]; +"1219 dequantize_per_tensor_default_49" [id=1219, type=dequantize_per_tensor]; +"1220 linear_48_scale_0" [id=1220, type=get_attr]; +"1221 linear_48_zero_point_0" [id=1221, type=get_attr]; +"1222 _frozen_param49" [id=1222, type=get_attr]; +"1223 dequantize_per_channel_default_49" [id=1223, type=dequantize_per_channel]; +"1224 linear_48" [id=1224, type=linear]; +"1225 gelu_7" [id=1225, type=gelu]; +"1226 dropout_30" [id=1226, type=dropout]; +"1227 dropout_30_0_0_nncf_smooth_quant_0" [id=1227, type=call_module]; +"1228 quantize_per_tensor_default_50" [id=1228, type=quantize_per_tensor]; +"1229 dequantize_per_tensor_default_50" [id=1229, type=dequantize_per_tensor]; +"1230 linear_49_scale_0" [id=1230, type=get_attr]; +"1231 linear_49_zero_point_0" [id=1231, type=get_attr]; +"1232 _frozen_param50" [id=1232, type=get_attr]; +"1233 dequantize_per_channel_default_50" [id=1233, type=dequantize_per_channel]; +"1234 linear_49" [id=1234, type=linear]; +"1235 dropout_31" [id=1235, type=dropout]; +"1236 layer_norm_18" [id=1236, type=layer_norm]; +"1237 add_27" [id=1237, type=add]; +"1238 features_5_4_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=1238, type=call_module]; +"1239 linear_50_scale_0" [id=1239, type=get_attr]; +"1240 linear_50_zero_point_0" [id=1240, type=get_attr]; +"1241 _frozen_param51" [id=1241, type=get_attr]; +"1242 dequantize_per_channel_default_51" [id=1242, type=dequantize_per_channel]; +"1243 linear_50" [id=1243, type=linear]; +"1244 relu__8" [id=1244, type=relu_]; +"1245 relu__8_0_0_nncf_smooth_quant_0" [id=1245, type=call_module]; +"1246 linear_51_scale_0" [id=1246, type=get_attr]; +"1247 linear_51_zero_point_0" [id=1247, type=get_attr]; +"1248 _frozen_param52" [id=1248, type=get_attr]; +"1249 dequantize_per_channel_default_52" [id=1249, type=dequantize_per_channel]; +"1250 linear_51" [id=1250, type=linear]; +"1251 view_44" [id=1251, type=view]; +"1252 index_8" [id=1252, type=index]; +"1253 view_45" [id=1253, type=view]; +"1254 permute_37" [id=1254, type=permute]; +"1255 contiguous_14" [id=1255, type=contiguous]; +"1256 unsqueeze_24" [id=1256, type=unsqueeze]; +"1257 sigmoid_8" [id=1257, type=sigmoid]; +"1258 mul_16" [id=1258, type=mul]; +"1259 pad_10" [id=1259, type=pad]; +"1260 view_46" [id=1260, type=view]; +"1261 permute_38" [id=1261, type=permute]; +"1262 reshape_36" [id=1262, type=reshape]; +"1263 clone_8" [id=1263, type=clone]; +"1264 reshape_36_0_0_nncf_smooth_quant_0" [id=1264, type=call_module]; +"1265 quantize_per_tensor_default_51" [id=1265, type=quantize_per_tensor]; +"1266 dequantize_per_tensor_default_51" [id=1266, type=dequantize_per_tensor]; +"1267 linear_52_scale_0" [id=1267, type=get_attr]; +"1268 linear_52_zero_point_0" [id=1268, type=get_attr]; +"1269 _frozen_param53" [id=1269, type=get_attr]; +"1270 dequantize_per_channel_default_53" [id=1270, type=dequantize_per_channel]; +"1271 linear_52" [id=1271, type=linear]; +"1272 reshape_37" [id=1272, type=reshape]; +"1273 permute_39" [id=1273, type=permute]; +"1274 select_24" [id=1274, type=select]; +"1275 select_25" [id=1275, type=select]; +"1276 select_26" [id=1276, type=select]; +"1277 linalg_vector_norm_16" [id=1277, type=linalg_vector_norm]; +"1278 clamp_min_16" [id=1278, type=clamp_min]; +"1279 expand_as_16" [id=1279, type=expand_as]; +"1280 div_16" [id=1280, type=div]; +"1281 quantize_per_tensor_default_52" [id=1281, type=quantize_per_tensor]; +"1282 dequantize_per_tensor_default_52" [id=1282, type=dequantize_per_tensor]; +"1283 linalg_vector_norm_17" [id=1283, type=linalg_vector_norm]; +"1284 clamp_min_17" [id=1284, type=clamp_min]; +"1285 expand_as_17" [id=1285, type=expand_as]; +"1286 div_17" [id=1286, type=div]; +"1287 quantize_per_tensor_default_53" [id=1287, type=quantize_per_tensor]; +"1288 dequantize_per_tensor_default_53" [id=1288, type=dequantize_per_tensor]; +"1289 transpose_16" [id=1289, type=transpose]; +"1290 matmul_16" [id=1290, type=matmul]; +"1291 clamp_8" [id=1291, type=clamp]; +"1292 exp_8" [id=1292, type=exp]; +"1293 mul_17" [id=1293, type=mul]; +"1294 add_28" [id=1294, type=add]; +"1295 softmax_8" [id=1295, type=softmax]; +"1296 dropout_32" [id=1296, type=dropout]; +"1297 matmul_17" [id=1297, type=matmul]; +"1298 transpose_17" [id=1298, type=transpose]; +"1299 reshape_38" [id=1299, type=reshape]; +"1300 reshape_38_0_0_nncf_smooth_quant_0" [id=1300, type=call_module]; +"1301 quantize_per_tensor_default_54" [id=1301, type=quantize_per_tensor]; +"1302 dequantize_per_tensor_default_54" [id=1302, type=dequantize_per_tensor]; +"1303 linear_53_scale_0" [id=1303, type=get_attr]; +"1304 linear_53_zero_point_0" [id=1304, type=get_attr]; +"1305 _frozen_param54" [id=1305, type=get_attr]; +"1306 dequantize_per_channel_default_54" [id=1306, type=dequantize_per_channel]; +"1307 linear_53" [id=1307, type=linear]; +"1308 dropout_33" [id=1308, type=dropout]; +"1309 view_47" [id=1309, type=view]; +"1310 permute_40" [id=1310, type=permute]; +"1311 reshape_39" [id=1311, type=reshape]; +"1312 slice_134" [id=1312, type=slice]; +"1313 slice_135" [id=1313, type=slice]; +"1314 slice_136" [id=1314, type=slice]; +"1315 slice_137" [id=1315, type=slice]; +"1316 contiguous_15" [id=1316, type=contiguous]; +"1317 layer_norm_19" [id=1317, type=layer_norm]; +"1318 add_29" [id=1318, type=add]; +"1319 add_29_0_0_nncf_smooth_quant_0" [id=1319, type=call_module]; +"1320 quantize_per_tensor_default_55" [id=1320, type=quantize_per_tensor]; +"1321 dequantize_per_tensor_default_55" [id=1321, type=dequantize_per_tensor]; +"1322 linear_54_scale_0" [id=1322, type=get_attr]; +"1323 linear_54_zero_point_0" [id=1323, type=get_attr]; +"1324 _frozen_param55" [id=1324, type=get_attr]; +"1325 dequantize_per_channel_default_55" [id=1325, type=dequantize_per_channel]; +"1326 linear_54" [id=1326, type=linear]; +"1327 gelu_8" [id=1327, type=gelu]; +"1328 dropout_34" [id=1328, type=dropout]; +"1329 dropout_34_0_0_nncf_smooth_quant_0" [id=1329, type=call_module]; +"1330 quantize_per_tensor_default_56" [id=1330, type=quantize_per_tensor]; +"1331 dequantize_per_tensor_default_56" [id=1331, type=dequantize_per_tensor]; +"1332 linear_55_scale_0" [id=1332, type=get_attr]; +"1333 linear_55_zero_point_0" [id=1333, type=get_attr]; +"1334 _frozen_param56" [id=1334, type=get_attr]; +"1335 dequantize_per_channel_default_56" [id=1335, type=dequantize_per_channel]; +"1336 linear_55" [id=1336, type=linear]; +"1337 dropout_35" [id=1337, type=dropout]; +"1338 layer_norm_20" [id=1338, type=layer_norm]; +"1339 add_30" [id=1339, type=add]; +"1340 features_5_5_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=1340, type=call_module]; +"1341 linear_56_scale_0" [id=1341, type=get_attr]; +"1342 linear_56_zero_point_0" [id=1342, type=get_attr]; +"1343 _frozen_param57" [id=1343, type=get_attr]; +"1344 dequantize_per_channel_default_57" [id=1344, type=dequantize_per_channel]; +"1345 linear_56" [id=1345, type=linear]; +"1346 relu__9" [id=1346, type=relu_]; +"1347 relu__9_0_0_nncf_smooth_quant_0" [id=1347, type=call_module]; +"1348 linear_57_scale_0" [id=1348, type=get_attr]; +"1349 linear_57_zero_point_0" [id=1349, type=get_attr]; +"1350 _frozen_param58" [id=1350, type=get_attr]; +"1351 dequantize_per_channel_default_58" [id=1351, type=dequantize_per_channel]; +"1352 linear_57" [id=1352, type=linear]; +"1353 view_48" [id=1353, type=view]; +"1354 index_9" [id=1354, type=index]; +"1355 view_49" [id=1355, type=view]; +"1356 permute_41" [id=1356, type=permute]; +"1357 contiguous_16" [id=1357, type=contiguous]; +"1358 unsqueeze_25" [id=1358, type=unsqueeze]; +"1359 sigmoid_9" [id=1359, type=sigmoid]; +"1360 mul_18" [id=1360, type=mul]; +"1361 pad_11" [id=1361, type=pad]; +"1362 roll_8" [id=1362, type=roll]; +"1363 view_50" [id=1363, type=view]; +"1364 permute_42" [id=1364, type=permute]; +"1365 reshape_40" [id=1365, type=reshape]; +"1366 clone_9" [id=1366, type=clone]; +"1367 reshape_40_0_0_nncf_smooth_quant_0" [id=1367, type=call_module]; +"1368 quantize_per_tensor_default_57" [id=1368, type=quantize_per_tensor]; +"1369 dequantize_per_tensor_default_57" [id=1369, type=dequantize_per_tensor]; +"1370 linear_58_scale_0" [id=1370, type=get_attr]; +"1371 linear_58_zero_point_0" [id=1371, type=get_attr]; +"1372 _frozen_param59" [id=1372, type=get_attr]; +"1373 dequantize_per_channel_default_59" [id=1373, type=dequantize_per_channel]; +"1374 linear_58" [id=1374, type=linear]; +"1375 reshape_41" [id=1375, type=reshape]; +"1376 permute_43" [id=1376, type=permute]; +"1377 select_27" [id=1377, type=select]; +"1378 select_28" [id=1378, type=select]; +"1379 select_29" [id=1379, type=select]; +"1380 linalg_vector_norm_18" [id=1380, type=linalg_vector_norm]; +"1381 clamp_min_18" [id=1381, type=clamp_min]; +"1382 expand_as_18" [id=1382, type=expand_as]; +"1383 div_18" [id=1383, type=div]; +"1384 quantize_per_tensor_default_58" [id=1384, type=quantize_per_tensor]; +"1385 dequantize_per_tensor_default_58" [id=1385, type=dequantize_per_tensor]; +"1386 linalg_vector_norm_19" [id=1386, type=linalg_vector_norm]; +"1387 clamp_min_19" [id=1387, type=clamp_min]; +"1388 expand_as_19" [id=1388, type=expand_as]; +"1389 div_19" [id=1389, type=div]; +"1390 quantize_per_tensor_default_59" [id=1390, type=quantize_per_tensor]; +"1391 dequantize_per_tensor_default_59" [id=1391, type=dequantize_per_tensor]; +"1392 transpose_18" [id=1392, type=transpose]; +"1393 matmul_18" [id=1393, type=matmul]; +"1394 clamp_9" [id=1394, type=clamp]; +"1395 exp_9" [id=1395, type=exp]; +"1396 mul_19" [id=1396, type=mul]; +"1397 add_31" [id=1397, type=add]; +"1398 new_zeros_4" [id=1398, type=new_zeros]; +"1399 view_51" [id=1399, type=view]; +"1400 permute_44" [id=1400, type=permute]; +"1401 reshape_42" [id=1401, type=reshape]; +"1402 unsqueeze_26" [id=1402, type=unsqueeze]; +"1403 unsqueeze_27" [id=1403, type=unsqueeze]; +"1404 sub_4" [id=1404, type=sub]; +"1405 ne_4" [id=1405, type=ne]; +"1406 masked_fill_8" [id=1406, type=masked_fill]; +"1407 eq_4" [id=1407, type=eq]; +"1408 masked_fill_9" [id=1408, type=masked_fill]; +"1409 view_52" [id=1409, type=view]; +"1410 unsqueeze_28" [id=1410, type=unsqueeze]; +"1411 unsqueeze_29" [id=1411, type=unsqueeze]; +"1412 add_32" [id=1412, type=add]; +"1413 view_53" [id=1413, type=view]; +"1414 softmax_9" [id=1414, type=softmax]; +"1415 dropout_36" [id=1415, type=dropout]; +"1416 matmul_19" [id=1416, type=matmul]; +"1417 transpose_19" [id=1417, type=transpose]; +"1418 reshape_43" [id=1418, type=reshape]; +"1419 reshape_43_0_0_nncf_smooth_quant_0" [id=1419, type=call_module]; +"1420 quantize_per_tensor_default_60" [id=1420, type=quantize_per_tensor]; +"1421 dequantize_per_tensor_default_60" [id=1421, type=dequantize_per_tensor]; +"1422 linear_59_scale_0" [id=1422, type=get_attr]; +"1423 linear_59_zero_point_0" [id=1423, type=get_attr]; +"1424 _frozen_param60" [id=1424, type=get_attr]; +"1425 dequantize_per_channel_default_60" [id=1425, type=dequantize_per_channel]; +"1426 linear_59" [id=1426, type=linear]; +"1427 dropout_37" [id=1427, type=dropout]; +"1428 view_54" [id=1428, type=view]; +"1429 permute_45" [id=1429, type=permute]; +"1430 reshape_44" [id=1430, type=reshape]; +"1431 roll_9" [id=1431, type=roll]; +"1432 slice_157" [id=1432, type=slice]; +"1433 slice_158" [id=1433, type=slice]; +"1434 slice_159" [id=1434, type=slice]; +"1435 slice_160" [id=1435, type=slice]; +"1436 contiguous_17" [id=1436, type=contiguous]; +"1437 layer_norm_21" [id=1437, type=layer_norm]; +"1438 add_33" [id=1438, type=add]; +"1439 add_33_0_0_nncf_smooth_quant_0" [id=1439, type=call_module]; +"1440 quantize_per_tensor_default_61" [id=1440, type=quantize_per_tensor]; +"1441 dequantize_per_tensor_default_61" [id=1441, type=dequantize_per_tensor]; +"1442 linear_60_scale_0" [id=1442, type=get_attr]; +"1443 linear_60_zero_point_0" [id=1443, type=get_attr]; +"1444 _frozen_param61" [id=1444, type=get_attr]; +"1445 dequantize_per_channel_default_61" [id=1445, type=dequantize_per_channel]; +"1446 linear_60" [id=1446, type=linear]; +"1447 gelu_9" [id=1447, type=gelu]; +"1448 dropout_38" [id=1448, type=dropout]; +"1449 dropout_38_0_0_nncf_smooth_quant_0" [id=1449, type=call_module]; +"1450 quantize_per_tensor_default_62" [id=1450, type=quantize_per_tensor]; +"1451 dequantize_per_tensor_default_62" [id=1451, type=dequantize_per_tensor]; +"1452 linear_61_scale_0" [id=1452, type=get_attr]; +"1453 linear_61_zero_point_0" [id=1453, type=get_attr]; +"1454 _frozen_param62" [id=1454, type=get_attr]; +"1455 dequantize_per_channel_default_62" [id=1455, type=dequantize_per_channel]; +"1456 linear_61" [id=1456, type=linear]; +"1457 dropout_39" [id=1457, type=dropout]; +"1458 layer_norm_22" [id=1458, type=layer_norm]; +"1459 add_34" [id=1459, type=add]; +"1460 features_5_6_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=1460, type=call_module]; +"1461 linear_62_scale_0" [id=1461, type=get_attr]; +"1462 linear_62_zero_point_0" [id=1462, type=get_attr]; +"1463 _frozen_param63" [id=1463, type=get_attr]; +"1464 dequantize_per_channel_default_63" [id=1464, type=dequantize_per_channel]; +"1465 linear_62" [id=1465, type=linear]; +"1466 relu__10" [id=1466, type=relu_]; +"1467 relu__10_0_0_nncf_smooth_quant_0" [id=1467, type=call_module]; +"1468 linear_63_scale_0" [id=1468, type=get_attr]; +"1469 linear_63_zero_point_0" [id=1469, type=get_attr]; +"1470 _frozen_param64" [id=1470, type=get_attr]; +"1471 dequantize_per_channel_default_64" [id=1471, type=dequantize_per_channel]; +"1472 linear_63" [id=1472, type=linear]; +"1473 view_55" [id=1473, type=view]; +"1474 index_10" [id=1474, type=index]; +"1475 view_56" [id=1475, type=view]; +"1476 permute_46" [id=1476, type=permute]; +"1477 contiguous_18" [id=1477, type=contiguous]; +"1478 unsqueeze_30" [id=1478, type=unsqueeze]; +"1479 sigmoid_10" [id=1479, type=sigmoid]; +"1480 mul_20" [id=1480, type=mul]; +"1481 pad_12" [id=1481, type=pad]; +"1482 view_57" [id=1482, type=view]; +"1483 permute_47" [id=1483, type=permute]; +"1484 reshape_45" [id=1484, type=reshape]; +"1485 clone_10" [id=1485, type=clone]; +"1486 reshape_45_0_0_nncf_smooth_quant_0" [id=1486, type=call_module]; +"1487 quantize_per_tensor_default_63" [id=1487, type=quantize_per_tensor]; +"1488 dequantize_per_tensor_default_63" [id=1488, type=dequantize_per_tensor]; +"1489 linear_64_scale_0" [id=1489, type=get_attr]; +"1490 linear_64_zero_point_0" [id=1490, type=get_attr]; +"1491 _frozen_param65" [id=1491, type=get_attr]; +"1492 dequantize_per_channel_default_65" [id=1492, type=dequantize_per_channel]; +"1493 linear_64" [id=1493, type=linear]; +"1494 reshape_46" [id=1494, type=reshape]; +"1495 permute_48" [id=1495, type=permute]; +"1496 select_30" [id=1496, type=select]; +"1497 select_31" [id=1497, type=select]; +"1498 select_32" [id=1498, type=select]; +"1499 linalg_vector_norm_20" [id=1499, type=linalg_vector_norm]; +"1500 clamp_min_20" [id=1500, type=clamp_min]; +"1501 expand_as_20" [id=1501, type=expand_as]; +"1502 div_20" [id=1502, type=div]; +"1503 quantize_per_tensor_default_64" [id=1503, type=quantize_per_tensor]; +"1504 dequantize_per_tensor_default_64" [id=1504, type=dequantize_per_tensor]; +"1505 linalg_vector_norm_21" [id=1505, type=linalg_vector_norm]; +"1506 clamp_min_21" [id=1506, type=clamp_min]; +"1507 expand_as_21" [id=1507, type=expand_as]; +"1508 div_21" [id=1508, type=div]; +"1509 quantize_per_tensor_default_65" [id=1509, type=quantize_per_tensor]; +"1510 dequantize_per_tensor_default_65" [id=1510, type=dequantize_per_tensor]; +"1511 transpose_20" [id=1511, type=transpose]; +"1512 matmul_20" [id=1512, type=matmul]; +"1513 clamp_10" [id=1513, type=clamp]; +"1514 exp_10" [id=1514, type=exp]; +"1515 mul_21" [id=1515, type=mul]; +"1516 add_35" [id=1516, type=add]; +"1517 softmax_10" [id=1517, type=softmax]; +"1518 dropout_40" [id=1518, type=dropout]; +"1519 matmul_21" [id=1519, type=matmul]; +"1520 transpose_21" [id=1520, type=transpose]; +"1521 reshape_47" [id=1521, type=reshape]; +"1522 reshape_47_0_0_nncf_smooth_quant_0" [id=1522, type=call_module]; +"1523 quantize_per_tensor_default_66" [id=1523, type=quantize_per_tensor]; +"1524 dequantize_per_tensor_default_66" [id=1524, type=dequantize_per_tensor]; +"1525 linear_65_scale_0" [id=1525, type=get_attr]; +"1526 linear_65_zero_point_0" [id=1526, type=get_attr]; +"1527 _frozen_param66" [id=1527, type=get_attr]; +"1528 dequantize_per_channel_default_66" [id=1528, type=dequantize_per_channel]; +"1529 linear_65" [id=1529, type=linear]; +"1530 dropout_41" [id=1530, type=dropout]; +"1531 view_58" [id=1531, type=view]; +"1532 permute_49" [id=1532, type=permute]; +"1533 reshape_48" [id=1533, type=reshape]; +"1534 slice_162" [id=1534, type=slice]; +"1535 slice_163" [id=1535, type=slice]; +"1536 slice_164" [id=1536, type=slice]; +"1537 slice_165" [id=1537, type=slice]; +"1538 contiguous_19" [id=1538, type=contiguous]; +"1539 layer_norm_23" [id=1539, type=layer_norm]; +"1540 add_36" [id=1540, type=add]; +"1541 add_36_0_0_nncf_smooth_quant_0" [id=1541, type=call_module]; +"1542 quantize_per_tensor_default_67" [id=1542, type=quantize_per_tensor]; +"1543 dequantize_per_tensor_default_67" [id=1543, type=dequantize_per_tensor]; +"1544 linear_66_scale_0" [id=1544, type=get_attr]; +"1545 linear_66_zero_point_0" [id=1545, type=get_attr]; +"1546 _frozen_param67" [id=1546, type=get_attr]; +"1547 dequantize_per_channel_default_67" [id=1547, type=dequantize_per_channel]; +"1548 linear_66" [id=1548, type=linear]; +"1549 gelu_10" [id=1549, type=gelu]; +"1550 dropout_42" [id=1550, type=dropout]; +"1551 dropout_42_0_0_nncf_smooth_quant_0" [id=1551, type=call_module]; +"1552 quantize_per_tensor_default_68" [id=1552, type=quantize_per_tensor]; +"1553 dequantize_per_tensor_default_68" [id=1553, type=dequantize_per_tensor]; +"1554 linear_67_scale_0" [id=1554, type=get_attr]; +"1555 linear_67_zero_point_0" [id=1555, type=get_attr]; +"1556 _frozen_param68" [id=1556, type=get_attr]; +"1557 dequantize_per_channel_default_68" [id=1557, type=dequantize_per_channel]; +"1558 linear_67" [id=1558, type=linear]; +"1559 dropout_43" [id=1559, type=dropout]; +"1560 layer_norm_24" [id=1560, type=layer_norm]; +"1561 add_37" [id=1561, type=add]; +"1562 features_5_7_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=1562, type=call_module]; +"1563 linear_68_scale_0" [id=1563, type=get_attr]; +"1564 linear_68_zero_point_0" [id=1564, type=get_attr]; +"1565 _frozen_param69" [id=1565, type=get_attr]; +"1566 dequantize_per_channel_default_69" [id=1566, type=dequantize_per_channel]; +"1567 linear_68" [id=1567, type=linear]; +"1568 relu__11" [id=1568, type=relu_]; +"1569 relu__11_0_0_nncf_smooth_quant_0" [id=1569, type=call_module]; +"1570 linear_69_scale_0" [id=1570, type=get_attr]; +"1571 linear_69_zero_point_0" [id=1571, type=get_attr]; +"1572 _frozen_param70" [id=1572, type=get_attr]; +"1573 dequantize_per_channel_default_70" [id=1573, type=dequantize_per_channel]; +"1574 linear_69" [id=1574, type=linear]; +"1575 view_59" [id=1575, type=view]; +"1576 index_11" [id=1576, type=index]; +"1577 view_60" [id=1577, type=view]; +"1578 permute_50" [id=1578, type=permute]; +"1579 contiguous_20" [id=1579, type=contiguous]; +"1580 unsqueeze_31" [id=1580, type=unsqueeze]; +"1581 sigmoid_11" [id=1581, type=sigmoid]; +"1582 mul_22" [id=1582, type=mul]; +"1583 pad_13" [id=1583, type=pad]; +"1584 roll_10" [id=1584, type=roll]; +"1585 view_61" [id=1585, type=view]; +"1586 permute_51" [id=1586, type=permute]; +"1587 reshape_49" [id=1587, type=reshape]; +"1588 clone_11" [id=1588, type=clone]; +"1589 reshape_49_0_0_nncf_smooth_quant_0" [id=1589, type=call_module]; +"1590 quantize_per_tensor_default_69" [id=1590, type=quantize_per_tensor]; +"1591 dequantize_per_tensor_default_69" [id=1591, type=dequantize_per_tensor]; +"1592 linear_70_scale_0" [id=1592, type=get_attr]; +"1593 linear_70_zero_point_0" [id=1593, type=get_attr]; +"1594 _frozen_param71" [id=1594, type=get_attr]; +"1595 dequantize_per_channel_default_71" [id=1595, type=dequantize_per_channel]; +"1596 linear_70" [id=1596, type=linear]; +"1597 reshape_50" [id=1597, type=reshape]; +"1598 permute_52" [id=1598, type=permute]; +"1599 select_33" [id=1599, type=select]; +"1600 select_34" [id=1600, type=select]; +"1601 select_35" [id=1601, type=select]; +"1602 linalg_vector_norm_22" [id=1602, type=linalg_vector_norm]; +"1603 clamp_min_22" [id=1603, type=clamp_min]; +"1604 expand_as_22" [id=1604, type=expand_as]; +"1605 div_22" [id=1605, type=div]; +"1606 quantize_per_tensor_default_70" [id=1606, type=quantize_per_tensor]; +"1607 dequantize_per_tensor_default_70" [id=1607, type=dequantize_per_tensor]; +"1608 linalg_vector_norm_23" [id=1608, type=linalg_vector_norm]; +"1609 clamp_min_23" [id=1609, type=clamp_min]; +"1610 expand_as_23" [id=1610, type=expand_as]; +"1611 div_23" [id=1611, type=div]; +"1612 quantize_per_tensor_default_71" [id=1612, type=quantize_per_tensor]; +"1613 dequantize_per_tensor_default_71" [id=1613, type=dequantize_per_tensor]; +"1614 transpose_22" [id=1614, type=transpose]; +"1615 matmul_22" [id=1615, type=matmul]; +"1616 clamp_11" [id=1616, type=clamp]; +"1617 exp_11" [id=1617, type=exp]; +"1618 mul_23" [id=1618, type=mul]; +"1619 add_38" [id=1619, type=add]; +"1620 new_zeros_5" [id=1620, type=new_zeros]; +"1621 view_62" [id=1621, type=view]; +"1622 permute_53" [id=1622, type=permute]; +"1623 reshape_51" [id=1623, type=reshape]; +"1624 unsqueeze_32" [id=1624, type=unsqueeze]; +"1625 unsqueeze_33" [id=1625, type=unsqueeze]; +"1626 sub_5" [id=1626, type=sub]; +"1627 ne_5" [id=1627, type=ne]; +"1628 masked_fill_10" [id=1628, type=masked_fill]; +"1629 eq_5" [id=1629, type=eq]; +"1630 masked_fill_11" [id=1630, type=masked_fill]; +"1631 view_63" [id=1631, type=view]; +"1632 unsqueeze_34" [id=1632, type=unsqueeze]; +"1633 unsqueeze_35" [id=1633, type=unsqueeze]; +"1634 add_39" [id=1634, type=add]; +"1635 view_64" [id=1635, type=view]; +"1636 softmax_11" [id=1636, type=softmax]; +"1637 dropout_44" [id=1637, type=dropout]; +"1638 matmul_23" [id=1638, type=matmul]; +"1639 transpose_23" [id=1639, type=transpose]; +"1640 reshape_52" [id=1640, type=reshape]; +"1641 reshape_52_0_0_nncf_smooth_quant_0" [id=1641, type=call_module]; +"1642 quantize_per_tensor_default_72" [id=1642, type=quantize_per_tensor]; +"1643 dequantize_per_tensor_default_72" [id=1643, type=dequantize_per_tensor]; +"1644 linear_71_scale_0" [id=1644, type=get_attr]; +"1645 linear_71_zero_point_0" [id=1645, type=get_attr]; +"1646 _frozen_param72" [id=1646, type=get_attr]; +"1647 dequantize_per_channel_default_72" [id=1647, type=dequantize_per_channel]; +"1648 linear_71" [id=1648, type=linear]; +"1649 dropout_45" [id=1649, type=dropout]; +"1650 view_65" [id=1650, type=view]; +"1651 permute_54" [id=1651, type=permute]; +"1652 reshape_53" [id=1652, type=reshape]; +"1653 roll_11" [id=1653, type=roll]; +"1654 slice_185" [id=1654, type=slice]; +"1655 slice_186" [id=1655, type=slice]; +"1656 slice_187" [id=1656, type=slice]; +"1657 slice_188" [id=1657, type=slice]; +"1658 contiguous_21" [id=1658, type=contiguous]; +"1659 layer_norm_25" [id=1659, type=layer_norm]; +"1660 add_40" [id=1660, type=add]; +"1661 add_40_0_0_nncf_smooth_quant_0" [id=1661, type=call_module]; +"1662 quantize_per_tensor_default_73" [id=1662, type=quantize_per_tensor]; +"1663 dequantize_per_tensor_default_73" [id=1663, type=dequantize_per_tensor]; +"1664 linear_72_scale_0" [id=1664, type=get_attr]; +"1665 linear_72_zero_point_0" [id=1665, type=get_attr]; +"1666 _frozen_param73" [id=1666, type=get_attr]; +"1667 dequantize_per_channel_default_73" [id=1667, type=dequantize_per_channel]; +"1668 linear_72" [id=1668, type=linear]; +"1669 gelu_11" [id=1669, type=gelu]; +"1670 dropout_46" [id=1670, type=dropout]; +"1671 dropout_46_0_0_nncf_smooth_quant_0" [id=1671, type=call_module]; +"1672 quantize_per_tensor_default_74" [id=1672, type=quantize_per_tensor]; +"1673 dequantize_per_tensor_default_74" [id=1673, type=dequantize_per_tensor]; +"1674 linear_73_scale_0" [id=1674, type=get_attr]; +"1675 linear_73_zero_point_0" [id=1675, type=get_attr]; +"1676 _frozen_param74" [id=1676, type=get_attr]; +"1677 dequantize_per_channel_default_74" [id=1677, type=dequantize_per_channel]; +"1678 linear_73" [id=1678, type=linear]; +"1679 dropout_47" [id=1679, type=dropout]; +"1680 layer_norm_26" [id=1680, type=layer_norm]; +"1681 add_41" [id=1681, type=add]; +"1682 features_5_8_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=1682, type=call_module]; +"1683 linear_74_scale_0" [id=1683, type=get_attr]; +"1684 linear_74_zero_point_0" [id=1684, type=get_attr]; +"1685 _frozen_param75" [id=1685, type=get_attr]; +"1686 dequantize_per_channel_default_75" [id=1686, type=dequantize_per_channel]; +"1687 linear_74" [id=1687, type=linear]; +"1688 relu__12" [id=1688, type=relu_]; +"1689 relu__12_0_0_nncf_smooth_quant_0" [id=1689, type=call_module]; +"1690 linear_75_scale_0" [id=1690, type=get_attr]; +"1691 linear_75_zero_point_0" [id=1691, type=get_attr]; +"1692 _frozen_param76" [id=1692, type=get_attr]; +"1693 dequantize_per_channel_default_76" [id=1693, type=dequantize_per_channel]; +"1694 linear_75" [id=1694, type=linear]; +"1695 view_66" [id=1695, type=view]; +"1696 index_12" [id=1696, type=index]; +"1697 view_67" [id=1697, type=view]; +"1698 permute_55" [id=1698, type=permute]; +"1699 contiguous_22" [id=1699, type=contiguous]; +"1700 unsqueeze_36" [id=1700, type=unsqueeze]; +"1701 sigmoid_12" [id=1701, type=sigmoid]; +"1702 mul_24" [id=1702, type=mul]; +"1703 pad_14" [id=1703, type=pad]; +"1704 view_68" [id=1704, type=view]; +"1705 permute_56" [id=1705, type=permute]; +"1706 reshape_54" [id=1706, type=reshape]; +"1707 clone_12" [id=1707, type=clone]; +"1708 reshape_54_0_0_nncf_smooth_quant_0" [id=1708, type=call_module]; +"1709 quantize_per_tensor_default_75" [id=1709, type=quantize_per_tensor]; +"1710 dequantize_per_tensor_default_75" [id=1710, type=dequantize_per_tensor]; +"1711 linear_76_scale_0" [id=1711, type=get_attr]; +"1712 linear_76_zero_point_0" [id=1712, type=get_attr]; +"1713 _frozen_param77" [id=1713, type=get_attr]; +"1714 dequantize_per_channel_default_77" [id=1714, type=dequantize_per_channel]; +"1715 linear_76" [id=1715, type=linear]; +"1716 reshape_55" [id=1716, type=reshape]; +"1717 permute_57" [id=1717, type=permute]; +"1718 select_36" [id=1718, type=select]; +"1719 select_37" [id=1719, type=select]; +"1720 select_38" [id=1720, type=select]; +"1721 linalg_vector_norm_24" [id=1721, type=linalg_vector_norm]; +"1722 clamp_min_24" [id=1722, type=clamp_min]; +"1723 expand_as_24" [id=1723, type=expand_as]; +"1724 div_24" [id=1724, type=div]; +"1725 quantize_per_tensor_default_76" [id=1725, type=quantize_per_tensor]; +"1726 dequantize_per_tensor_default_76" [id=1726, type=dequantize_per_tensor]; +"1727 linalg_vector_norm_25" [id=1727, type=linalg_vector_norm]; +"1728 clamp_min_25" [id=1728, type=clamp_min]; +"1729 expand_as_25" [id=1729, type=expand_as]; +"1730 div_25" [id=1730, type=div]; +"1731 quantize_per_tensor_default_77" [id=1731, type=quantize_per_tensor]; +"1732 dequantize_per_tensor_default_77" [id=1732, type=dequantize_per_tensor]; +"1733 transpose_24" [id=1733, type=transpose]; +"1734 matmul_24" [id=1734, type=matmul]; +"1735 clamp_12" [id=1735, type=clamp]; +"1736 exp_12" [id=1736, type=exp]; +"1737 mul_25" [id=1737, type=mul]; +"1738 add_42" [id=1738, type=add]; +"1739 softmax_12" [id=1739, type=softmax]; +"1740 dropout_48" [id=1740, type=dropout]; +"1741 matmul_25" [id=1741, type=matmul]; +"1742 transpose_25" [id=1742, type=transpose]; +"1743 reshape_56" [id=1743, type=reshape]; +"1744 reshape_56_0_0_nncf_smooth_quant_0" [id=1744, type=call_module]; +"1745 quantize_per_tensor_default_78" [id=1745, type=quantize_per_tensor]; +"1746 dequantize_per_tensor_default_78" [id=1746, type=dequantize_per_tensor]; +"1747 linear_77_scale_0" [id=1747, type=get_attr]; +"1748 linear_77_zero_point_0" [id=1748, type=get_attr]; +"1749 _frozen_param78" [id=1749, type=get_attr]; +"1750 dequantize_per_channel_default_78" [id=1750, type=dequantize_per_channel]; +"1751 linear_77" [id=1751, type=linear]; +"1752 dropout_49" [id=1752, type=dropout]; +"1753 view_69" [id=1753, type=view]; +"1754 permute_58" [id=1754, type=permute]; +"1755 reshape_57" [id=1755, type=reshape]; +"1756 slice_190" [id=1756, type=slice]; +"1757 slice_191" [id=1757, type=slice]; +"1758 slice_192" [id=1758, type=slice]; +"1759 slice_193" [id=1759, type=slice]; +"1760 contiguous_23" [id=1760, type=contiguous]; +"1761 layer_norm_27" [id=1761, type=layer_norm]; +"1762 add_43" [id=1762, type=add]; +"1763 add_43_0_0_nncf_smooth_quant_0" [id=1763, type=call_module]; +"1764 quantize_per_tensor_default_79" [id=1764, type=quantize_per_tensor]; +"1765 dequantize_per_tensor_default_79" [id=1765, type=dequantize_per_tensor]; +"1766 linear_78_scale_0" [id=1766, type=get_attr]; +"1767 linear_78_zero_point_0" [id=1767, type=get_attr]; +"1768 _frozen_param79" [id=1768, type=get_attr]; +"1769 dequantize_per_channel_default_79" [id=1769, type=dequantize_per_channel]; +"1770 linear_78" [id=1770, type=linear]; +"1771 gelu_12" [id=1771, type=gelu]; +"1772 dropout_50" [id=1772, type=dropout]; +"1773 dropout_50_0_0_nncf_smooth_quant_0" [id=1773, type=call_module]; +"1774 quantize_per_tensor_default_80" [id=1774, type=quantize_per_tensor]; +"1775 dequantize_per_tensor_default_80" [id=1775, type=dequantize_per_tensor]; +"1776 linear_79_scale_0" [id=1776, type=get_attr]; +"1777 linear_79_zero_point_0" [id=1777, type=get_attr]; +"1778 _frozen_param80" [id=1778, type=get_attr]; +"1779 dequantize_per_channel_default_80" [id=1779, type=dequantize_per_channel]; +"1780 linear_79" [id=1780, type=linear]; +"1781 dropout_51" [id=1781, type=dropout]; +"1782 layer_norm_28" [id=1782, type=layer_norm]; +"1783 add_44" [id=1783, type=add]; +"1784 features_5_9_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=1784, type=call_module]; +"1785 linear_80_scale_0" [id=1785, type=get_attr]; +"1786 linear_80_zero_point_0" [id=1786, type=get_attr]; +"1787 _frozen_param81" [id=1787, type=get_attr]; +"1788 dequantize_per_channel_default_81" [id=1788, type=dequantize_per_channel]; +"1789 linear_80" [id=1789, type=linear]; +"1790 relu__13" [id=1790, type=relu_]; +"1791 relu__13_0_0_nncf_smooth_quant_0" [id=1791, type=call_module]; +"1792 linear_81_scale_0" [id=1792, type=get_attr]; +"1793 linear_81_zero_point_0" [id=1793, type=get_attr]; +"1794 _frozen_param82" [id=1794, type=get_attr]; +"1795 dequantize_per_channel_default_82" [id=1795, type=dequantize_per_channel]; +"1796 linear_81" [id=1796, type=linear]; +"1797 view_70" [id=1797, type=view]; +"1798 index_13" [id=1798, type=index]; +"1799 view_71" [id=1799, type=view]; +"1800 permute_59" [id=1800, type=permute]; +"1801 contiguous_24" [id=1801, type=contiguous]; +"1802 unsqueeze_37" [id=1802, type=unsqueeze]; +"1803 sigmoid_13" [id=1803, type=sigmoid]; +"1804 mul_26" [id=1804, type=mul]; +"1805 pad_15" [id=1805, type=pad]; +"1806 roll_12" [id=1806, type=roll]; +"1807 view_72" [id=1807, type=view]; +"1808 permute_60" [id=1808, type=permute]; +"1809 reshape_58" [id=1809, type=reshape]; +"1810 clone_13" [id=1810, type=clone]; +"1811 reshape_58_0_0_nncf_smooth_quant_0" [id=1811, type=call_module]; +"1812 quantize_per_tensor_default_81" [id=1812, type=quantize_per_tensor]; +"1813 dequantize_per_tensor_default_81" [id=1813, type=dequantize_per_tensor]; +"1814 linear_82_scale_0" [id=1814, type=get_attr]; +"1815 linear_82_zero_point_0" [id=1815, type=get_attr]; +"1816 _frozen_param83" [id=1816, type=get_attr]; +"1817 dequantize_per_channel_default_83" [id=1817, type=dequantize_per_channel]; +"1818 linear_82" [id=1818, type=linear]; +"1819 reshape_59" [id=1819, type=reshape]; +"1820 permute_61" [id=1820, type=permute]; +"1821 select_39" [id=1821, type=select]; +"1822 select_40" [id=1822, type=select]; +"1823 select_41" [id=1823, type=select]; +"1824 linalg_vector_norm_26" [id=1824, type=linalg_vector_norm]; +"1825 clamp_min_26" [id=1825, type=clamp_min]; +"1826 expand_as_26" [id=1826, type=expand_as]; +"1827 div_26" [id=1827, type=div]; +"1828 quantize_per_tensor_default_82" [id=1828, type=quantize_per_tensor]; +"1829 dequantize_per_tensor_default_82" [id=1829, type=dequantize_per_tensor]; +"1830 linalg_vector_norm_27" [id=1830, type=linalg_vector_norm]; +"1831 clamp_min_27" [id=1831, type=clamp_min]; +"1832 expand_as_27" [id=1832, type=expand_as]; +"1833 div_27" [id=1833, type=div]; +"1834 quantize_per_tensor_default_83" [id=1834, type=quantize_per_tensor]; +"1835 dequantize_per_tensor_default_83" [id=1835, type=dequantize_per_tensor]; +"1836 transpose_26" [id=1836, type=transpose]; +"1837 matmul_26" [id=1837, type=matmul]; +"1838 clamp_13" [id=1838, type=clamp]; +"1839 exp_13" [id=1839, type=exp]; +"1840 mul_27" [id=1840, type=mul]; +"1841 add_45" [id=1841, type=add]; +"1842 new_zeros_6" [id=1842, type=new_zeros]; +"1843 view_73" [id=1843, type=view]; +"1844 permute_62" [id=1844, type=permute]; +"1845 reshape_60" [id=1845, type=reshape]; +"1846 unsqueeze_38" [id=1846, type=unsqueeze]; +"1847 unsqueeze_39" [id=1847, type=unsqueeze]; +"1848 sub_6" [id=1848, type=sub]; +"1849 ne_6" [id=1849, type=ne]; +"1850 masked_fill_12" [id=1850, type=masked_fill]; +"1851 eq_6" [id=1851, type=eq]; +"1852 masked_fill_13" [id=1852, type=masked_fill]; +"1853 view_74" [id=1853, type=view]; +"1854 unsqueeze_40" [id=1854, type=unsqueeze]; +"1855 unsqueeze_41" [id=1855, type=unsqueeze]; +"1856 add_46" [id=1856, type=add]; +"1857 view_75" [id=1857, type=view]; +"1858 softmax_13" [id=1858, type=softmax]; +"1859 dropout_52" [id=1859, type=dropout]; +"1860 matmul_27" [id=1860, type=matmul]; +"1861 transpose_27" [id=1861, type=transpose]; +"1862 reshape_61" [id=1862, type=reshape]; +"1863 reshape_61_0_0_nncf_smooth_quant_0" [id=1863, type=call_module]; +"1864 quantize_per_tensor_default_84" [id=1864, type=quantize_per_tensor]; +"1865 dequantize_per_tensor_default_84" [id=1865, type=dequantize_per_tensor]; +"1866 linear_83_scale_0" [id=1866, type=get_attr]; +"1867 linear_83_zero_point_0" [id=1867, type=get_attr]; +"1868 _frozen_param84" [id=1868, type=get_attr]; +"1869 dequantize_per_channel_default_84" [id=1869, type=dequantize_per_channel]; +"1870 linear_83" [id=1870, type=linear]; +"1871 dropout_53" [id=1871, type=dropout]; +"1872 view_76" [id=1872, type=view]; +"1873 permute_63" [id=1873, type=permute]; +"1874 reshape_62" [id=1874, type=reshape]; +"1875 roll_13" [id=1875, type=roll]; +"1876 slice_213" [id=1876, type=slice]; +"1877 slice_214" [id=1877, type=slice]; +"1878 slice_215" [id=1878, type=slice]; +"1879 slice_216" [id=1879, type=slice]; +"1880 contiguous_25" [id=1880, type=contiguous]; +"1881 layer_norm_29" [id=1881, type=layer_norm]; +"1882 add_47" [id=1882, type=add]; +"1883 add_47_0_0_nncf_smooth_quant_0" [id=1883, type=call_module]; +"1884 quantize_per_tensor_default_85" [id=1884, type=quantize_per_tensor]; +"1885 dequantize_per_tensor_default_85" [id=1885, type=dequantize_per_tensor]; +"1886 linear_84_scale_0" [id=1886, type=get_attr]; +"1887 linear_84_zero_point_0" [id=1887, type=get_attr]; +"1888 _frozen_param85" [id=1888, type=get_attr]; +"1889 dequantize_per_channel_default_85" [id=1889, type=dequantize_per_channel]; +"1890 linear_84" [id=1890, type=linear]; +"1891 gelu_13" [id=1891, type=gelu]; +"1892 dropout_54" [id=1892, type=dropout]; +"1893 dropout_54_0_0_nncf_smooth_quant_0" [id=1893, type=call_module]; +"1894 quantize_per_tensor_default_86" [id=1894, type=quantize_per_tensor]; +"1895 dequantize_per_tensor_default_86" [id=1895, type=dequantize_per_tensor]; +"1896 linear_85_scale_0" [id=1896, type=get_attr]; +"1897 linear_85_zero_point_0" [id=1897, type=get_attr]; +"1898 _frozen_param86" [id=1898, type=get_attr]; +"1899 dequantize_per_channel_default_86" [id=1899, type=dequantize_per_channel]; +"1900 linear_85" [id=1900, type=linear]; +"1901 dropout_55" [id=1901, type=dropout]; +"1902 layer_norm_30" [id=1902, type=layer_norm]; +"1903 add_48" [id=1903, type=add]; +"1904 features_5_10_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=1904, type=call_module]; +"1905 linear_86_scale_0" [id=1905, type=get_attr]; +"1906 linear_86_zero_point_0" [id=1906, type=get_attr]; +"1907 _frozen_param87" [id=1907, type=get_attr]; +"1908 dequantize_per_channel_default_87" [id=1908, type=dequantize_per_channel]; +"1909 linear_86" [id=1909, type=linear]; +"1910 relu__14" [id=1910, type=relu_]; +"1911 relu__14_0_0_nncf_smooth_quant_0" [id=1911, type=call_module]; +"1912 linear_87_scale_0" [id=1912, type=get_attr]; +"1913 linear_87_zero_point_0" [id=1913, type=get_attr]; +"1914 _frozen_param88" [id=1914, type=get_attr]; +"1915 dequantize_per_channel_default_88" [id=1915, type=dequantize_per_channel]; +"1916 linear_87" [id=1916, type=linear]; +"1917 view_77" [id=1917, type=view]; +"1918 index_14" [id=1918, type=index]; +"1919 view_78" [id=1919, type=view]; +"1920 permute_64" [id=1920, type=permute]; +"1921 contiguous_26" [id=1921, type=contiguous]; +"1922 unsqueeze_42" [id=1922, type=unsqueeze]; +"1923 sigmoid_14" [id=1923, type=sigmoid]; +"1924 mul_28" [id=1924, type=mul]; +"1925 pad_16" [id=1925, type=pad]; +"1926 view_79" [id=1926, type=view]; +"1927 permute_65" [id=1927, type=permute]; +"1928 reshape_63" [id=1928, type=reshape]; +"1929 clone_14" [id=1929, type=clone]; +"1930 reshape_63_0_0_nncf_smooth_quant_0" [id=1930, type=call_module]; +"1931 quantize_per_tensor_default_87" [id=1931, type=quantize_per_tensor]; +"1932 dequantize_per_tensor_default_87" [id=1932, type=dequantize_per_tensor]; +"1933 linear_88_scale_0" [id=1933, type=get_attr]; +"1934 linear_88_zero_point_0" [id=1934, type=get_attr]; +"1935 _frozen_param89" [id=1935, type=get_attr]; +"1936 dequantize_per_channel_default_89" [id=1936, type=dequantize_per_channel]; +"1937 linear_88" [id=1937, type=linear]; +"1938 reshape_64" [id=1938, type=reshape]; +"1939 permute_66" [id=1939, type=permute]; +"1940 select_42" [id=1940, type=select]; +"1941 select_43" [id=1941, type=select]; +"1942 select_44" [id=1942, type=select]; +"1943 linalg_vector_norm_28" [id=1943, type=linalg_vector_norm]; +"1944 clamp_min_28" [id=1944, type=clamp_min]; +"1945 expand_as_28" [id=1945, type=expand_as]; +"1946 div_28" [id=1946, type=div]; +"1947 quantize_per_tensor_default_88" [id=1947, type=quantize_per_tensor]; +"1948 dequantize_per_tensor_default_88" [id=1948, type=dequantize_per_tensor]; +"1949 linalg_vector_norm_29" [id=1949, type=linalg_vector_norm]; +"1950 clamp_min_29" [id=1950, type=clamp_min]; +"1951 expand_as_29" [id=1951, type=expand_as]; +"1952 div_29" [id=1952, type=div]; +"1953 quantize_per_tensor_default_89" [id=1953, type=quantize_per_tensor]; +"1954 dequantize_per_tensor_default_89" [id=1954, type=dequantize_per_tensor]; +"1955 transpose_28" [id=1955, type=transpose]; +"1956 matmul_28" [id=1956, type=matmul]; +"1957 clamp_14" [id=1957, type=clamp]; +"1958 exp_14" [id=1958, type=exp]; +"1959 mul_29" [id=1959, type=mul]; +"1960 add_49" [id=1960, type=add]; +"1961 softmax_14" [id=1961, type=softmax]; +"1962 dropout_56" [id=1962, type=dropout]; +"1963 matmul_29" [id=1963, type=matmul]; +"1964 transpose_29" [id=1964, type=transpose]; +"1965 reshape_65" [id=1965, type=reshape]; +"1966 reshape_65_0_0_nncf_smooth_quant_0" [id=1966, type=call_module]; +"1967 quantize_per_tensor_default_90" [id=1967, type=quantize_per_tensor]; +"1968 dequantize_per_tensor_default_90" [id=1968, type=dequantize_per_tensor]; +"1969 linear_89_scale_0" [id=1969, type=get_attr]; +"1970 linear_89_zero_point_0" [id=1970, type=get_attr]; +"1971 _frozen_param90" [id=1971, type=get_attr]; +"1972 dequantize_per_channel_default_90" [id=1972, type=dequantize_per_channel]; +"1973 linear_89" [id=1973, type=linear]; +"1974 dropout_57" [id=1974, type=dropout]; +"1975 view_80" [id=1975, type=view]; +"1976 permute_67" [id=1976, type=permute]; +"1977 reshape_66" [id=1977, type=reshape]; +"1978 slice_218" [id=1978, type=slice]; +"1979 slice_219" [id=1979, type=slice]; +"1980 slice_220" [id=1980, type=slice]; +"1981 slice_221" [id=1981, type=slice]; +"1982 contiguous_27" [id=1982, type=contiguous]; +"1983 layer_norm_31" [id=1983, type=layer_norm]; +"1984 add_50" [id=1984, type=add]; +"1985 add_50_0_0_nncf_smooth_quant_0" [id=1985, type=call_module]; +"1986 quantize_per_tensor_default_91" [id=1986, type=quantize_per_tensor]; +"1987 dequantize_per_tensor_default_91" [id=1987, type=dequantize_per_tensor]; +"1988 linear_90_scale_0" [id=1988, type=get_attr]; +"1989 linear_90_zero_point_0" [id=1989, type=get_attr]; +"1990 _frozen_param91" [id=1990, type=get_attr]; +"1991 dequantize_per_channel_default_91" [id=1991, type=dequantize_per_channel]; +"1992 linear_90" [id=1992, type=linear]; +"1993 gelu_14" [id=1993, type=gelu]; +"1994 dropout_58" [id=1994, type=dropout]; +"1995 dropout_58_0_0_nncf_smooth_quant_0" [id=1995, type=call_module]; +"1996 quantize_per_tensor_default_92" [id=1996, type=quantize_per_tensor]; +"1997 dequantize_per_tensor_default_92" [id=1997, type=dequantize_per_tensor]; +"1998 linear_91_scale_0" [id=1998, type=get_attr]; +"1999 linear_91_zero_point_0" [id=1999, type=get_attr]; +"2000 _frozen_param92" [id=2000, type=get_attr]; +"2001 dequantize_per_channel_default_92" [id=2001, type=dequantize_per_channel]; +"2002 linear_91" [id=2002, type=linear]; +"2003 dropout_59" [id=2003, type=dropout]; +"2004 layer_norm_32" [id=2004, type=layer_norm]; +"2005 add_51" [id=2005, type=add]; +"2006 features_5_11_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=2006, type=call_module]; +"2007 linear_92_scale_0" [id=2007, type=get_attr]; +"2008 linear_92_zero_point_0" [id=2008, type=get_attr]; +"2009 _frozen_param93" [id=2009, type=get_attr]; +"2010 dequantize_per_channel_default_93" [id=2010, type=dequantize_per_channel]; +"2011 linear_92" [id=2011, type=linear]; +"2012 relu__15" [id=2012, type=relu_]; +"2013 relu__15_0_0_nncf_smooth_quant_0" [id=2013, type=call_module]; +"2014 linear_93_scale_0" [id=2014, type=get_attr]; +"2015 linear_93_zero_point_0" [id=2015, type=get_attr]; +"2016 _frozen_param94" [id=2016, type=get_attr]; +"2017 dequantize_per_channel_default_94" [id=2017, type=dequantize_per_channel]; +"2018 linear_93" [id=2018, type=linear]; +"2019 view_81" [id=2019, type=view]; +"2020 index_15" [id=2020, type=index]; +"2021 view_82" [id=2021, type=view]; +"2022 permute_68" [id=2022, type=permute]; +"2023 contiguous_28" [id=2023, type=contiguous]; +"2024 unsqueeze_43" [id=2024, type=unsqueeze]; +"2025 sigmoid_15" [id=2025, type=sigmoid]; +"2026 mul_30" [id=2026, type=mul]; +"2027 pad_17" [id=2027, type=pad]; +"2028 roll_14" [id=2028, type=roll]; +"2029 view_83" [id=2029, type=view]; +"2030 permute_69" [id=2030, type=permute]; +"2031 reshape_67" [id=2031, type=reshape]; +"2032 clone_15" [id=2032, type=clone]; +"2033 reshape_67_0_0_nncf_smooth_quant_0" [id=2033, type=call_module]; +"2034 quantize_per_tensor_default_93" [id=2034, type=quantize_per_tensor]; +"2035 dequantize_per_tensor_default_93" [id=2035, type=dequantize_per_tensor]; +"2036 linear_94_scale_0" [id=2036, type=get_attr]; +"2037 linear_94_zero_point_0" [id=2037, type=get_attr]; +"2038 _frozen_param95" [id=2038, type=get_attr]; +"2039 dequantize_per_channel_default_95" [id=2039, type=dequantize_per_channel]; +"2040 linear_94" [id=2040, type=linear]; +"2041 reshape_68" [id=2041, type=reshape]; +"2042 permute_70" [id=2042, type=permute]; +"2043 select_45" [id=2043, type=select]; +"2044 select_46" [id=2044, type=select]; +"2045 select_47" [id=2045, type=select]; +"2046 linalg_vector_norm_30" [id=2046, type=linalg_vector_norm]; +"2047 clamp_min_30" [id=2047, type=clamp_min]; +"2048 expand_as_30" [id=2048, type=expand_as]; +"2049 div_30" [id=2049, type=div]; +"2050 quantize_per_tensor_default_94" [id=2050, type=quantize_per_tensor]; +"2051 dequantize_per_tensor_default_94" [id=2051, type=dequantize_per_tensor]; +"2052 linalg_vector_norm_31" [id=2052, type=linalg_vector_norm]; +"2053 clamp_min_31" [id=2053, type=clamp_min]; +"2054 expand_as_31" [id=2054, type=expand_as]; +"2055 div_31" [id=2055, type=div]; +"2056 quantize_per_tensor_default_95" [id=2056, type=quantize_per_tensor]; +"2057 dequantize_per_tensor_default_95" [id=2057, type=dequantize_per_tensor]; +"2058 transpose_30" [id=2058, type=transpose]; +"2059 matmul_30" [id=2059, type=matmul]; +"2060 clamp_15" [id=2060, type=clamp]; +"2061 exp_15" [id=2061, type=exp]; +"2062 mul_31" [id=2062, type=mul]; +"2063 add_52" [id=2063, type=add]; +"2064 new_zeros_7" [id=2064, type=new_zeros]; +"2065 view_84" [id=2065, type=view]; +"2066 permute_71" [id=2066, type=permute]; +"2067 reshape_69" [id=2067, type=reshape]; +"2068 unsqueeze_44" [id=2068, type=unsqueeze]; +"2069 unsqueeze_45" [id=2069, type=unsqueeze]; +"2070 sub_7" [id=2070, type=sub]; +"2071 ne_7" [id=2071, type=ne]; +"2072 masked_fill_14" [id=2072, type=masked_fill]; +"2073 eq_7" [id=2073, type=eq]; +"2074 masked_fill_15" [id=2074, type=masked_fill]; +"2075 view_85" [id=2075, type=view]; +"2076 unsqueeze_46" [id=2076, type=unsqueeze]; +"2077 unsqueeze_47" [id=2077, type=unsqueeze]; +"2078 add_53" [id=2078, type=add]; +"2079 view_86" [id=2079, type=view]; +"2080 softmax_15" [id=2080, type=softmax]; +"2081 dropout_60" [id=2081, type=dropout]; +"2082 matmul_31" [id=2082, type=matmul]; +"2083 transpose_31" [id=2083, type=transpose]; +"2084 reshape_70" [id=2084, type=reshape]; +"2085 reshape_70_0_0_nncf_smooth_quant_0" [id=2085, type=call_module]; +"2086 quantize_per_tensor_default_96" [id=2086, type=quantize_per_tensor]; +"2087 dequantize_per_tensor_default_96" [id=2087, type=dequantize_per_tensor]; +"2088 linear_95_scale_0" [id=2088, type=get_attr]; +"2089 linear_95_zero_point_0" [id=2089, type=get_attr]; +"2090 _frozen_param96" [id=2090, type=get_attr]; +"2091 dequantize_per_channel_default_96" [id=2091, type=dequantize_per_channel]; +"2092 linear_95" [id=2092, type=linear]; +"2093 dropout_61" [id=2093, type=dropout]; +"2094 view_87" [id=2094, type=view]; +"2095 permute_72" [id=2095, type=permute]; +"2096 reshape_71" [id=2096, type=reshape]; +"2097 roll_15" [id=2097, type=roll]; +"2098 slice_241" [id=2098, type=slice]; +"2099 slice_242" [id=2099, type=slice]; +"2100 slice_243" [id=2100, type=slice]; +"2101 slice_244" [id=2101, type=slice]; +"2102 contiguous_29" [id=2102, type=contiguous]; +"2103 layer_norm_33" [id=2103, type=layer_norm]; +"2104 add_54" [id=2104, type=add]; +"2105 add_54_0_0_nncf_smooth_quant_0" [id=2105, type=call_module]; +"2106 quantize_per_tensor_default_97" [id=2106, type=quantize_per_tensor]; +"2107 dequantize_per_tensor_default_97" [id=2107, type=dequantize_per_tensor]; +"2108 linear_96_scale_0" [id=2108, type=get_attr]; +"2109 linear_96_zero_point_0" [id=2109, type=get_attr]; +"2110 _frozen_param97" [id=2110, type=get_attr]; +"2111 dequantize_per_channel_default_97" [id=2111, type=dequantize_per_channel]; +"2112 linear_96" [id=2112, type=linear]; +"2113 gelu_15" [id=2113, type=gelu]; +"2114 dropout_62" [id=2114, type=dropout]; +"2115 dropout_62_0_0_nncf_smooth_quant_0" [id=2115, type=call_module]; +"2116 quantize_per_tensor_default_98" [id=2116, type=quantize_per_tensor]; +"2117 dequantize_per_tensor_default_98" [id=2117, type=dequantize_per_tensor]; +"2118 linear_97_scale_0" [id=2118, type=get_attr]; +"2119 linear_97_zero_point_0" [id=2119, type=get_attr]; +"2120 _frozen_param98" [id=2120, type=get_attr]; +"2121 dequantize_per_channel_default_98" [id=2121, type=dequantize_per_channel]; +"2122 linear_97" [id=2122, type=linear]; +"2123 dropout_63" [id=2123, type=dropout]; +"2124 layer_norm_34" [id=2124, type=layer_norm]; +"2125 add_55" [id=2125, type=add]; +"2126 features_5_12_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=2126, type=call_module]; +"2127 linear_98_scale_0" [id=2127, type=get_attr]; +"2128 linear_98_zero_point_0" [id=2128, type=get_attr]; +"2129 _frozen_param99" [id=2129, type=get_attr]; +"2130 dequantize_per_channel_default_99" [id=2130, type=dequantize_per_channel]; +"2131 linear_98" [id=2131, type=linear]; +"2132 relu__16" [id=2132, type=relu_]; +"2133 relu__16_0_0_nncf_smooth_quant_0" [id=2133, type=call_module]; +"2134 linear_99_scale_0" [id=2134, type=get_attr]; +"2135 linear_99_zero_point_0" [id=2135, type=get_attr]; +"2136 _frozen_param100" [id=2136, type=get_attr]; +"2137 dequantize_per_channel_default_100" [id=2137, type=dequantize_per_channel]; +"2138 linear_99" [id=2138, type=linear]; +"2139 view_88" [id=2139, type=view]; +"2140 index_16" [id=2140, type=index]; +"2141 view_89" [id=2141, type=view]; +"2142 permute_73" [id=2142, type=permute]; +"2143 contiguous_30" [id=2143, type=contiguous]; +"2144 unsqueeze_48" [id=2144, type=unsqueeze]; +"2145 sigmoid_16" [id=2145, type=sigmoid]; +"2146 mul_32" [id=2146, type=mul]; +"2147 pad_18" [id=2147, type=pad]; +"2148 view_90" [id=2148, type=view]; +"2149 permute_74" [id=2149, type=permute]; +"2150 reshape_72" [id=2150, type=reshape]; +"2151 clone_16" [id=2151, type=clone]; +"2152 reshape_72_0_0_nncf_smooth_quant_0" [id=2152, type=call_module]; +"2153 quantize_per_tensor_default_99" [id=2153, type=quantize_per_tensor]; +"2154 dequantize_per_tensor_default_99" [id=2154, type=dequantize_per_tensor]; +"2155 linear_100_scale_0" [id=2155, type=get_attr]; +"2156 linear_100_zero_point_0" [id=2156, type=get_attr]; +"2157 _frozen_param101" [id=2157, type=get_attr]; +"2158 dequantize_per_channel_default_101" [id=2158, type=dequantize_per_channel]; +"2159 linear_100" [id=2159, type=linear]; +"2160 reshape_73" [id=2160, type=reshape]; +"2161 permute_75" [id=2161, type=permute]; +"2162 select_48" [id=2162, type=select]; +"2163 select_49" [id=2163, type=select]; +"2164 select_50" [id=2164, type=select]; +"2165 linalg_vector_norm_32" [id=2165, type=linalg_vector_norm]; +"2166 clamp_min_32" [id=2166, type=clamp_min]; +"2167 expand_as_32" [id=2167, type=expand_as]; +"2168 div_32" [id=2168, type=div]; +"2169 quantize_per_tensor_default_100" [id=2169, type=quantize_per_tensor]; +"2170 dequantize_per_tensor_default_100" [id=2170, type=dequantize_per_tensor]; +"2171 linalg_vector_norm_33" [id=2171, type=linalg_vector_norm]; +"2172 clamp_min_33" [id=2172, type=clamp_min]; +"2173 expand_as_33" [id=2173, type=expand_as]; +"2174 div_33" [id=2174, type=div]; +"2175 quantize_per_tensor_default_101" [id=2175, type=quantize_per_tensor]; +"2176 dequantize_per_tensor_default_101" [id=2176, type=dequantize_per_tensor]; +"2177 transpose_32" [id=2177, type=transpose]; +"2178 matmul_32" [id=2178, type=matmul]; +"2179 clamp_16" [id=2179, type=clamp]; +"2180 exp_16" [id=2180, type=exp]; +"2181 mul_33" [id=2181, type=mul]; +"2182 add_56" [id=2182, type=add]; +"2183 softmax_16" [id=2183, type=softmax]; +"2184 dropout_64" [id=2184, type=dropout]; +"2185 matmul_33" [id=2185, type=matmul]; +"2186 transpose_33" [id=2186, type=transpose]; +"2187 reshape_74" [id=2187, type=reshape]; +"2188 reshape_74_0_0_nncf_smooth_quant_0" [id=2188, type=call_module]; +"2189 quantize_per_tensor_default_102" [id=2189, type=quantize_per_tensor]; +"2190 dequantize_per_tensor_default_102" [id=2190, type=dequantize_per_tensor]; +"2191 linear_101_scale_0" [id=2191, type=get_attr]; +"2192 linear_101_zero_point_0" [id=2192, type=get_attr]; +"2193 _frozen_param102" [id=2193, type=get_attr]; +"2194 dequantize_per_channel_default_102" [id=2194, type=dequantize_per_channel]; +"2195 linear_101" [id=2195, type=linear]; +"2196 dropout_65" [id=2196, type=dropout]; +"2197 view_91" [id=2197, type=view]; +"2198 permute_76" [id=2198, type=permute]; +"2199 reshape_75" [id=2199, type=reshape]; +"2200 slice_246" [id=2200, type=slice]; +"2201 slice_247" [id=2201, type=slice]; +"2202 slice_248" [id=2202, type=slice]; +"2203 slice_249" [id=2203, type=slice]; +"2204 contiguous_31" [id=2204, type=contiguous]; +"2205 layer_norm_35" [id=2205, type=layer_norm]; +"2206 add_57" [id=2206, type=add]; +"2207 add_57_0_0_nncf_smooth_quant_0" [id=2207, type=call_module]; +"2208 quantize_per_tensor_default_103" [id=2208, type=quantize_per_tensor]; +"2209 dequantize_per_tensor_default_103" [id=2209, type=dequantize_per_tensor]; +"2210 linear_102_scale_0" [id=2210, type=get_attr]; +"2211 linear_102_zero_point_0" [id=2211, type=get_attr]; +"2212 _frozen_param103" [id=2212, type=get_attr]; +"2213 dequantize_per_channel_default_103" [id=2213, type=dequantize_per_channel]; +"2214 linear_102" [id=2214, type=linear]; +"2215 gelu_16" [id=2215, type=gelu]; +"2216 dropout_66" [id=2216, type=dropout]; +"2217 dropout_66_0_0_nncf_smooth_quant_0" [id=2217, type=call_module]; +"2218 quantize_per_tensor_default_104" [id=2218, type=quantize_per_tensor]; +"2219 dequantize_per_tensor_default_104" [id=2219, type=dequantize_per_tensor]; +"2220 linear_103_scale_0" [id=2220, type=get_attr]; +"2221 linear_103_zero_point_0" [id=2221, type=get_attr]; +"2222 _frozen_param104" [id=2222, type=get_attr]; +"2223 dequantize_per_channel_default_104" [id=2223, type=dequantize_per_channel]; +"2224 linear_103" [id=2224, type=linear]; +"2225 dropout_67" [id=2225, type=dropout]; +"2226 layer_norm_36" [id=2226, type=layer_norm]; +"2227 add_58" [id=2227, type=add]; +"2228 features_5_13_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=2228, type=call_module]; +"2229 linear_104_scale_0" [id=2229, type=get_attr]; +"2230 linear_104_zero_point_0" [id=2230, type=get_attr]; +"2231 _frozen_param105" [id=2231, type=get_attr]; +"2232 dequantize_per_channel_default_105" [id=2232, type=dequantize_per_channel]; +"2233 linear_104" [id=2233, type=linear]; +"2234 relu__17" [id=2234, type=relu_]; +"2235 relu__17_0_0_nncf_smooth_quant_0" [id=2235, type=call_module]; +"2236 linear_105_scale_0" [id=2236, type=get_attr]; +"2237 linear_105_zero_point_0" [id=2237, type=get_attr]; +"2238 _frozen_param106" [id=2238, type=get_attr]; +"2239 dequantize_per_channel_default_106" [id=2239, type=dequantize_per_channel]; +"2240 linear_105" [id=2240, type=linear]; +"2241 view_92" [id=2241, type=view]; +"2242 index_17" [id=2242, type=index]; +"2243 view_93" [id=2243, type=view]; +"2244 permute_77" [id=2244, type=permute]; +"2245 contiguous_32" [id=2245, type=contiguous]; +"2246 unsqueeze_49" [id=2246, type=unsqueeze]; +"2247 sigmoid_17" [id=2247, type=sigmoid]; +"2248 mul_34" [id=2248, type=mul]; +"2249 pad_19" [id=2249, type=pad]; +"2250 roll_16" [id=2250, type=roll]; +"2251 view_94" [id=2251, type=view]; +"2252 permute_78" [id=2252, type=permute]; +"2253 reshape_76" [id=2253, type=reshape]; +"2254 clone_17" [id=2254, type=clone]; +"2255 reshape_76_0_0_nncf_smooth_quant_0" [id=2255, type=call_module]; +"2256 quantize_per_tensor_default_105" [id=2256, type=quantize_per_tensor]; +"2257 dequantize_per_tensor_default_105" [id=2257, type=dequantize_per_tensor]; +"2258 linear_106_scale_0" [id=2258, type=get_attr]; +"2259 linear_106_zero_point_0" [id=2259, type=get_attr]; +"2260 _frozen_param107" [id=2260, type=get_attr]; +"2261 dequantize_per_channel_default_107" [id=2261, type=dequantize_per_channel]; +"2262 linear_106" [id=2262, type=linear]; +"2263 reshape_77" [id=2263, type=reshape]; +"2264 permute_79" [id=2264, type=permute]; +"2265 select_51" [id=2265, type=select]; +"2266 select_52" [id=2266, type=select]; +"2267 select_53" [id=2267, type=select]; +"2268 linalg_vector_norm_34" [id=2268, type=linalg_vector_norm]; +"2269 clamp_min_34" [id=2269, type=clamp_min]; +"2270 expand_as_34" [id=2270, type=expand_as]; +"2271 div_34" [id=2271, type=div]; +"2272 quantize_per_tensor_default_106" [id=2272, type=quantize_per_tensor]; +"2273 dequantize_per_tensor_default_106" [id=2273, type=dequantize_per_tensor]; +"2274 linalg_vector_norm_35" [id=2274, type=linalg_vector_norm]; +"2275 clamp_min_35" [id=2275, type=clamp_min]; +"2276 expand_as_35" [id=2276, type=expand_as]; +"2277 div_35" [id=2277, type=div]; +"2278 quantize_per_tensor_default_107" [id=2278, type=quantize_per_tensor]; +"2279 dequantize_per_tensor_default_107" [id=2279, type=dequantize_per_tensor]; +"2280 transpose_34" [id=2280, type=transpose]; +"2281 matmul_34" [id=2281, type=matmul]; +"2282 clamp_17" [id=2282, type=clamp]; +"2283 exp_17" [id=2283, type=exp]; +"2284 mul_35" [id=2284, type=mul]; +"2285 add_59" [id=2285, type=add]; +"2286 new_zeros_8" [id=2286, type=new_zeros]; +"2287 view_95" [id=2287, type=view]; +"2288 permute_80" [id=2288, type=permute]; +"2289 reshape_78" [id=2289, type=reshape]; +"2290 unsqueeze_50" [id=2290, type=unsqueeze]; +"2291 unsqueeze_51" [id=2291, type=unsqueeze]; +"2292 sub_8" [id=2292, type=sub]; +"2293 ne_8" [id=2293, type=ne]; +"2294 masked_fill_16" [id=2294, type=masked_fill]; +"2295 eq_8" [id=2295, type=eq]; +"2296 masked_fill_17" [id=2296, type=masked_fill]; +"2297 view_96" [id=2297, type=view]; +"2298 unsqueeze_52" [id=2298, type=unsqueeze]; +"2299 unsqueeze_53" [id=2299, type=unsqueeze]; +"2300 add_60" [id=2300, type=add]; +"2301 view_97" [id=2301, type=view]; +"2302 softmax_17" [id=2302, type=softmax]; +"2303 dropout_68" [id=2303, type=dropout]; +"2304 matmul_35" [id=2304, type=matmul]; +"2305 transpose_35" [id=2305, type=transpose]; +"2306 reshape_79" [id=2306, type=reshape]; +"2307 reshape_79_0_0_nncf_smooth_quant_0" [id=2307, type=call_module]; +"2308 quantize_per_tensor_default_108" [id=2308, type=quantize_per_tensor]; +"2309 dequantize_per_tensor_default_108" [id=2309, type=dequantize_per_tensor]; +"2310 linear_107_scale_0" [id=2310, type=get_attr]; +"2311 linear_107_zero_point_0" [id=2311, type=get_attr]; +"2312 _frozen_param108" [id=2312, type=get_attr]; +"2313 dequantize_per_channel_default_108" [id=2313, type=dequantize_per_channel]; +"2314 linear_107" [id=2314, type=linear]; +"2315 dropout_69" [id=2315, type=dropout]; +"2316 view_98" [id=2316, type=view]; +"2317 permute_81" [id=2317, type=permute]; +"2318 reshape_80" [id=2318, type=reshape]; +"2319 roll_17" [id=2319, type=roll]; +"2320 slice_269" [id=2320, type=slice]; +"2321 slice_270" [id=2321, type=slice]; +"2322 slice_271" [id=2322, type=slice]; +"2323 slice_272" [id=2323, type=slice]; +"2324 contiguous_33" [id=2324, type=contiguous]; +"2325 layer_norm_37" [id=2325, type=layer_norm]; +"2326 add_61" [id=2326, type=add]; +"2327 add_61_0_0_nncf_smooth_quant_0" [id=2327, type=call_module]; +"2328 quantize_per_tensor_default_109" [id=2328, type=quantize_per_tensor]; +"2329 dequantize_per_tensor_default_109" [id=2329, type=dequantize_per_tensor]; +"2330 linear_108_scale_0" [id=2330, type=get_attr]; +"2331 linear_108_zero_point_0" [id=2331, type=get_attr]; +"2332 _frozen_param109" [id=2332, type=get_attr]; +"2333 dequantize_per_channel_default_109" [id=2333, type=dequantize_per_channel]; +"2334 linear_108" [id=2334, type=linear]; +"2335 gelu_17" [id=2335, type=gelu]; +"2336 dropout_70" [id=2336, type=dropout]; +"2337 dropout_70_0_0_nncf_smooth_quant_0" [id=2337, type=call_module]; +"2338 quantize_per_tensor_default_110" [id=2338, type=quantize_per_tensor]; +"2339 dequantize_per_tensor_default_110" [id=2339, type=dequantize_per_tensor]; +"2340 linear_109_scale_0" [id=2340, type=get_attr]; +"2341 linear_109_zero_point_0" [id=2341, type=get_attr]; +"2342 _frozen_param110" [id=2342, type=get_attr]; +"2343 dequantize_per_channel_default_110" [id=2343, type=dequantize_per_channel]; +"2344 linear_109" [id=2344, type=linear]; +"2345 dropout_71" [id=2345, type=dropout]; +"2346 layer_norm_38" [id=2346, type=layer_norm]; +"2347 add_62" [id=2347, type=add]; +"2348 features_5_14_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=2348, type=call_module]; +"2349 linear_110_scale_0" [id=2349, type=get_attr]; +"2350 linear_110_zero_point_0" [id=2350, type=get_attr]; +"2351 _frozen_param111" [id=2351, type=get_attr]; +"2352 dequantize_per_channel_default_111" [id=2352, type=dequantize_per_channel]; +"2353 linear_110" [id=2353, type=linear]; +"2354 relu__18" [id=2354, type=relu_]; +"2355 relu__18_0_0_nncf_smooth_quant_0" [id=2355, type=call_module]; +"2356 linear_111_scale_0" [id=2356, type=get_attr]; +"2357 linear_111_zero_point_0" [id=2357, type=get_attr]; +"2358 _frozen_param112" [id=2358, type=get_attr]; +"2359 dequantize_per_channel_default_112" [id=2359, type=dequantize_per_channel]; +"2360 linear_111" [id=2360, type=linear]; +"2361 view_99" [id=2361, type=view]; +"2362 index_18" [id=2362, type=index]; +"2363 view_100" [id=2363, type=view]; +"2364 permute_82" [id=2364, type=permute]; +"2365 contiguous_34" [id=2365, type=contiguous]; +"2366 unsqueeze_54" [id=2366, type=unsqueeze]; +"2367 sigmoid_18" [id=2367, type=sigmoid]; +"2368 mul_36" [id=2368, type=mul]; +"2369 pad_20" [id=2369, type=pad]; +"2370 view_101" [id=2370, type=view]; +"2371 permute_83" [id=2371, type=permute]; +"2372 reshape_81" [id=2372, type=reshape]; +"2373 clone_18" [id=2373, type=clone]; +"2374 reshape_81_0_0_nncf_smooth_quant_0" [id=2374, type=call_module]; +"2375 quantize_per_tensor_default_111" [id=2375, type=quantize_per_tensor]; +"2376 dequantize_per_tensor_default_111" [id=2376, type=dequantize_per_tensor]; +"2377 linear_112_scale_0" [id=2377, type=get_attr]; +"2378 linear_112_zero_point_0" [id=2378, type=get_attr]; +"2379 _frozen_param113" [id=2379, type=get_attr]; +"2380 dequantize_per_channel_default_113" [id=2380, type=dequantize_per_channel]; +"2381 linear_112" [id=2381, type=linear]; +"2382 reshape_82" [id=2382, type=reshape]; +"2383 permute_84" [id=2383, type=permute]; +"2384 select_54" [id=2384, type=select]; +"2385 select_55" [id=2385, type=select]; +"2386 select_56" [id=2386, type=select]; +"2387 linalg_vector_norm_36" [id=2387, type=linalg_vector_norm]; +"2388 clamp_min_36" [id=2388, type=clamp_min]; +"2389 expand_as_36" [id=2389, type=expand_as]; +"2390 div_36" [id=2390, type=div]; +"2391 quantize_per_tensor_default_112" [id=2391, type=quantize_per_tensor]; +"2392 dequantize_per_tensor_default_112" [id=2392, type=dequantize_per_tensor]; +"2393 linalg_vector_norm_37" [id=2393, type=linalg_vector_norm]; +"2394 clamp_min_37" [id=2394, type=clamp_min]; +"2395 expand_as_37" [id=2395, type=expand_as]; +"2396 div_37" [id=2396, type=div]; +"2397 quantize_per_tensor_default_113" [id=2397, type=quantize_per_tensor]; +"2398 dequantize_per_tensor_default_113" [id=2398, type=dequantize_per_tensor]; +"2399 transpose_36" [id=2399, type=transpose]; +"2400 matmul_36" [id=2400, type=matmul]; +"2401 clamp_18" [id=2401, type=clamp]; +"2402 exp_18" [id=2402, type=exp]; +"2403 mul_37" [id=2403, type=mul]; +"2404 add_63" [id=2404, type=add]; +"2405 softmax_18" [id=2405, type=softmax]; +"2406 dropout_72" [id=2406, type=dropout]; +"2407 matmul_37" [id=2407, type=matmul]; +"2408 transpose_37" [id=2408, type=transpose]; +"2409 reshape_83" [id=2409, type=reshape]; +"2410 reshape_83_0_0_nncf_smooth_quant_0" [id=2410, type=call_module]; +"2411 quantize_per_tensor_default_114" [id=2411, type=quantize_per_tensor]; +"2412 dequantize_per_tensor_default_114" [id=2412, type=dequantize_per_tensor]; +"2413 linear_113_scale_0" [id=2413, type=get_attr]; +"2414 linear_113_zero_point_0" [id=2414, type=get_attr]; +"2415 _frozen_param114" [id=2415, type=get_attr]; +"2416 dequantize_per_channel_default_114" [id=2416, type=dequantize_per_channel]; +"2417 linear_113" [id=2417, type=linear]; +"2418 dropout_73" [id=2418, type=dropout]; +"2419 view_102" [id=2419, type=view]; +"2420 permute_85" [id=2420, type=permute]; +"2421 reshape_84" [id=2421, type=reshape]; +"2422 slice_274" [id=2422, type=slice]; +"2423 slice_275" [id=2423, type=slice]; +"2424 slice_276" [id=2424, type=slice]; +"2425 slice_277" [id=2425, type=slice]; +"2426 contiguous_35" [id=2426, type=contiguous]; +"2427 layer_norm_39" [id=2427, type=layer_norm]; +"2428 add_64" [id=2428, type=add]; +"2429 add_64_0_0_nncf_smooth_quant_0" [id=2429, type=call_module]; +"2430 quantize_per_tensor_default_115" [id=2430, type=quantize_per_tensor]; +"2431 dequantize_per_tensor_default_115" [id=2431, type=dequantize_per_tensor]; +"2432 linear_114_scale_0" [id=2432, type=get_attr]; +"2433 linear_114_zero_point_0" [id=2433, type=get_attr]; +"2434 _frozen_param115" [id=2434, type=get_attr]; +"2435 dequantize_per_channel_default_115" [id=2435, type=dequantize_per_channel]; +"2436 linear_114" [id=2436, type=linear]; +"2437 gelu_18" [id=2437, type=gelu]; +"2438 dropout_74" [id=2438, type=dropout]; +"2439 dropout_74_0_0_nncf_smooth_quant_0" [id=2439, type=call_module]; +"2440 quantize_per_tensor_default_116" [id=2440, type=quantize_per_tensor]; +"2441 dequantize_per_tensor_default_116" [id=2441, type=dequantize_per_tensor]; +"2442 linear_115_scale_0" [id=2442, type=get_attr]; +"2443 linear_115_zero_point_0" [id=2443, type=get_attr]; +"2444 _frozen_param116" [id=2444, type=get_attr]; +"2445 dequantize_per_channel_default_116" [id=2445, type=dequantize_per_channel]; +"2446 linear_115" [id=2446, type=linear]; +"2447 dropout_75" [id=2447, type=dropout]; +"2448 layer_norm_40" [id=2448, type=layer_norm]; +"2449 add_65" [id=2449, type=add]; +"2450 features_5_15_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=2450, type=call_module]; +"2451 linear_116_scale_0" [id=2451, type=get_attr]; +"2452 linear_116_zero_point_0" [id=2452, type=get_attr]; +"2453 _frozen_param117" [id=2453, type=get_attr]; +"2454 dequantize_per_channel_default_117" [id=2454, type=dequantize_per_channel]; +"2455 linear_116" [id=2455, type=linear]; +"2456 relu__19" [id=2456, type=relu_]; +"2457 relu__19_0_0_nncf_smooth_quant_0" [id=2457, type=call_module]; +"2458 linear_117_scale_0" [id=2458, type=get_attr]; +"2459 linear_117_zero_point_0" [id=2459, type=get_attr]; +"2460 _frozen_param118" [id=2460, type=get_attr]; +"2461 dequantize_per_channel_default_118" [id=2461, type=dequantize_per_channel]; +"2462 linear_117" [id=2462, type=linear]; +"2463 view_103" [id=2463, type=view]; +"2464 index_19" [id=2464, type=index]; +"2465 view_104" [id=2465, type=view]; +"2466 permute_86" [id=2466, type=permute]; +"2467 contiguous_36" [id=2467, type=contiguous]; +"2468 unsqueeze_55" [id=2468, type=unsqueeze]; +"2469 sigmoid_19" [id=2469, type=sigmoid]; +"2470 mul_38" [id=2470, type=mul]; +"2471 pad_21" [id=2471, type=pad]; +"2472 roll_18" [id=2472, type=roll]; +"2473 view_105" [id=2473, type=view]; +"2474 permute_87" [id=2474, type=permute]; +"2475 reshape_85" [id=2475, type=reshape]; +"2476 clone_19" [id=2476, type=clone]; +"2477 reshape_85_0_0_nncf_smooth_quant_0" [id=2477, type=call_module]; +"2478 quantize_per_tensor_default_117" [id=2478, type=quantize_per_tensor]; +"2479 dequantize_per_tensor_default_117" [id=2479, type=dequantize_per_tensor]; +"2480 linear_118_scale_0" [id=2480, type=get_attr]; +"2481 linear_118_zero_point_0" [id=2481, type=get_attr]; +"2482 _frozen_param119" [id=2482, type=get_attr]; +"2483 dequantize_per_channel_default_119" [id=2483, type=dequantize_per_channel]; +"2484 linear_118" [id=2484, type=linear]; +"2485 reshape_86" [id=2485, type=reshape]; +"2486 permute_88" [id=2486, type=permute]; +"2487 select_57" [id=2487, type=select]; +"2488 select_58" [id=2488, type=select]; +"2489 select_59" [id=2489, type=select]; +"2490 linalg_vector_norm_38" [id=2490, type=linalg_vector_norm]; +"2491 clamp_min_38" [id=2491, type=clamp_min]; +"2492 expand_as_38" [id=2492, type=expand_as]; +"2493 div_38" [id=2493, type=div]; +"2494 quantize_per_tensor_default_118" [id=2494, type=quantize_per_tensor]; +"2495 dequantize_per_tensor_default_118" [id=2495, type=dequantize_per_tensor]; +"2496 linalg_vector_norm_39" [id=2496, type=linalg_vector_norm]; +"2497 clamp_min_39" [id=2497, type=clamp_min]; +"2498 expand_as_39" [id=2498, type=expand_as]; +"2499 div_39" [id=2499, type=div]; +"2500 quantize_per_tensor_default_119" [id=2500, type=quantize_per_tensor]; +"2501 dequantize_per_tensor_default_119" [id=2501, type=dequantize_per_tensor]; +"2502 transpose_38" [id=2502, type=transpose]; +"2503 matmul_38" [id=2503, type=matmul]; +"2504 clamp_19" [id=2504, type=clamp]; +"2505 exp_19" [id=2505, type=exp]; +"2506 mul_39" [id=2506, type=mul]; +"2507 add_66" [id=2507, type=add]; +"2508 new_zeros_9" [id=2508, type=new_zeros]; +"2509 view_106" [id=2509, type=view]; +"2510 permute_89" [id=2510, type=permute]; +"2511 reshape_87" [id=2511, type=reshape]; +"2512 unsqueeze_56" [id=2512, type=unsqueeze]; +"2513 unsqueeze_57" [id=2513, type=unsqueeze]; +"2514 sub_9" [id=2514, type=sub]; +"2515 ne_9" [id=2515, type=ne]; +"2516 masked_fill_18" [id=2516, type=masked_fill]; +"2517 eq_9" [id=2517, type=eq]; +"2518 masked_fill_19" [id=2518, type=masked_fill]; +"2519 view_107" [id=2519, type=view]; +"2520 unsqueeze_58" [id=2520, type=unsqueeze]; +"2521 unsqueeze_59" [id=2521, type=unsqueeze]; +"2522 add_67" [id=2522, type=add]; +"2523 view_108" [id=2523, type=view]; +"2524 softmax_19" [id=2524, type=softmax]; +"2525 dropout_76" [id=2525, type=dropout]; +"2526 matmul_39" [id=2526, type=matmul]; +"2527 transpose_39" [id=2527, type=transpose]; +"2528 reshape_88" [id=2528, type=reshape]; +"2529 reshape_88_0_0_nncf_smooth_quant_0" [id=2529, type=call_module]; +"2530 quantize_per_tensor_default_120" [id=2530, type=quantize_per_tensor]; +"2531 dequantize_per_tensor_default_120" [id=2531, type=dequantize_per_tensor]; +"2532 linear_119_scale_0" [id=2532, type=get_attr]; +"2533 linear_119_zero_point_0" [id=2533, type=get_attr]; +"2534 _frozen_param120" [id=2534, type=get_attr]; +"2535 dequantize_per_channel_default_120" [id=2535, type=dequantize_per_channel]; +"2536 linear_119" [id=2536, type=linear]; +"2537 dropout_77" [id=2537, type=dropout]; +"2538 view_109" [id=2538, type=view]; +"2539 permute_90" [id=2539, type=permute]; +"2540 reshape_89" [id=2540, type=reshape]; +"2541 roll_19" [id=2541, type=roll]; +"2542 slice_297" [id=2542, type=slice]; +"2543 slice_298" [id=2543, type=slice]; +"2544 slice_299" [id=2544, type=slice]; +"2545 slice_300" [id=2545, type=slice]; +"2546 contiguous_37" [id=2546, type=contiguous]; +"2547 layer_norm_41" [id=2547, type=layer_norm]; +"2548 add_68" [id=2548, type=add]; +"2549 add_68_0_0_nncf_smooth_quant_0" [id=2549, type=call_module]; +"2550 quantize_per_tensor_default_121" [id=2550, type=quantize_per_tensor]; +"2551 dequantize_per_tensor_default_121" [id=2551, type=dequantize_per_tensor]; +"2552 linear_120_scale_0" [id=2552, type=get_attr]; +"2553 linear_120_zero_point_0" [id=2553, type=get_attr]; +"2554 _frozen_param121" [id=2554, type=get_attr]; +"2555 dequantize_per_channel_default_121" [id=2555, type=dequantize_per_channel]; +"2556 linear_120" [id=2556, type=linear]; +"2557 gelu_19" [id=2557, type=gelu]; +"2558 dropout_78" [id=2558, type=dropout]; +"2559 dropout_78_0_0_nncf_smooth_quant_0" [id=2559, type=call_module]; +"2560 quantize_per_tensor_default_122" [id=2560, type=quantize_per_tensor]; +"2561 dequantize_per_tensor_default_122" [id=2561, type=dequantize_per_tensor]; +"2562 linear_121_scale_0" [id=2562, type=get_attr]; +"2563 linear_121_zero_point_0" [id=2563, type=get_attr]; +"2564 _frozen_param122" [id=2564, type=get_attr]; +"2565 dequantize_per_channel_default_122" [id=2565, type=dequantize_per_channel]; +"2566 linear_121" [id=2566, type=linear]; +"2567 dropout_79" [id=2567, type=dropout]; +"2568 layer_norm_42" [id=2568, type=layer_norm]; +"2569 add_69" [id=2569, type=add]; +"2570 features_5_16_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=2570, type=call_module]; +"2571 linear_122_scale_0" [id=2571, type=get_attr]; +"2572 linear_122_zero_point_0" [id=2572, type=get_attr]; +"2573 _frozen_param123" [id=2573, type=get_attr]; +"2574 dequantize_per_channel_default_123" [id=2574, type=dequantize_per_channel]; +"2575 linear_122" [id=2575, type=linear]; +"2576 relu__20" [id=2576, type=relu_]; +"2577 relu__20_0_0_nncf_smooth_quant_0" [id=2577, type=call_module]; +"2578 linear_123_scale_0" [id=2578, type=get_attr]; +"2579 linear_123_zero_point_0" [id=2579, type=get_attr]; +"2580 _frozen_param124" [id=2580, type=get_attr]; +"2581 dequantize_per_channel_default_124" [id=2581, type=dequantize_per_channel]; +"2582 linear_123" [id=2582, type=linear]; +"2583 view_110" [id=2583, type=view]; +"2584 index_20" [id=2584, type=index]; +"2585 view_111" [id=2585, type=view]; +"2586 permute_91" [id=2586, type=permute]; +"2587 contiguous_38" [id=2587, type=contiguous]; +"2588 unsqueeze_60" [id=2588, type=unsqueeze]; +"2589 sigmoid_20" [id=2589, type=sigmoid]; +"2590 mul_40" [id=2590, type=mul]; +"2591 pad_22" [id=2591, type=pad]; +"2592 view_112" [id=2592, type=view]; +"2593 permute_92" [id=2593, type=permute]; +"2594 reshape_90" [id=2594, type=reshape]; +"2595 clone_20" [id=2595, type=clone]; +"2596 reshape_90_0_0_nncf_smooth_quant_0" [id=2596, type=call_module]; +"2597 quantize_per_tensor_default_123" [id=2597, type=quantize_per_tensor]; +"2598 dequantize_per_tensor_default_123" [id=2598, type=dequantize_per_tensor]; +"2599 linear_124_scale_0" [id=2599, type=get_attr]; +"2600 linear_124_zero_point_0" [id=2600, type=get_attr]; +"2601 _frozen_param125" [id=2601, type=get_attr]; +"2602 dequantize_per_channel_default_125" [id=2602, type=dequantize_per_channel]; +"2603 linear_124" [id=2603, type=linear]; +"2604 reshape_91" [id=2604, type=reshape]; +"2605 permute_93" [id=2605, type=permute]; +"2606 select_60" [id=2606, type=select]; +"2607 select_61" [id=2607, type=select]; +"2608 select_62" [id=2608, type=select]; +"2609 linalg_vector_norm_40" [id=2609, type=linalg_vector_norm]; +"2610 clamp_min_40" [id=2610, type=clamp_min]; +"2611 expand_as_40" [id=2611, type=expand_as]; +"2612 div_40" [id=2612, type=div]; +"2613 quantize_per_tensor_default_124" [id=2613, type=quantize_per_tensor]; +"2614 dequantize_per_tensor_default_124" [id=2614, type=dequantize_per_tensor]; +"2615 linalg_vector_norm_41" [id=2615, type=linalg_vector_norm]; +"2616 clamp_min_41" [id=2616, type=clamp_min]; +"2617 expand_as_41" [id=2617, type=expand_as]; +"2618 div_41" [id=2618, type=div]; +"2619 quantize_per_tensor_default_125" [id=2619, type=quantize_per_tensor]; +"2620 dequantize_per_tensor_default_125" [id=2620, type=dequantize_per_tensor]; +"2621 transpose_40" [id=2621, type=transpose]; +"2622 matmul_40" [id=2622, type=matmul]; +"2623 clamp_20" [id=2623, type=clamp]; +"2624 exp_20" [id=2624, type=exp]; +"2625 mul_41" [id=2625, type=mul]; +"2626 add_70" [id=2626, type=add]; +"2627 softmax_20" [id=2627, type=softmax]; +"2628 dropout_80" [id=2628, type=dropout]; +"2629 matmul_41" [id=2629, type=matmul]; +"2630 transpose_41" [id=2630, type=transpose]; +"2631 reshape_92" [id=2631, type=reshape]; +"2632 reshape_92_0_0_nncf_smooth_quant_0" [id=2632, type=call_module]; +"2633 quantize_per_tensor_default_126" [id=2633, type=quantize_per_tensor]; +"2634 dequantize_per_tensor_default_126" [id=2634, type=dequantize_per_tensor]; +"2635 linear_125_scale_0" [id=2635, type=get_attr]; +"2636 linear_125_zero_point_0" [id=2636, type=get_attr]; +"2637 _frozen_param126" [id=2637, type=get_attr]; +"2638 dequantize_per_channel_default_126" [id=2638, type=dequantize_per_channel]; +"2639 linear_125" [id=2639, type=linear]; +"2640 dropout_81" [id=2640, type=dropout]; +"2641 view_113" [id=2641, type=view]; +"2642 permute_94" [id=2642, type=permute]; +"2643 reshape_93" [id=2643, type=reshape]; +"2644 slice_302" [id=2644, type=slice]; +"2645 slice_303" [id=2645, type=slice]; +"2646 slice_304" [id=2646, type=slice]; +"2647 slice_305" [id=2647, type=slice]; +"2648 contiguous_39" [id=2648, type=contiguous]; +"2649 layer_norm_43" [id=2649, type=layer_norm]; +"2650 add_71" [id=2650, type=add]; +"2651 add_71_0_0_nncf_smooth_quant_0" [id=2651, type=call_module]; +"2652 quantize_per_tensor_default_127" [id=2652, type=quantize_per_tensor]; +"2653 dequantize_per_tensor_default_127" [id=2653, type=dequantize_per_tensor]; +"2654 linear_126_scale_0" [id=2654, type=get_attr]; +"2655 linear_126_zero_point_0" [id=2655, type=get_attr]; +"2656 _frozen_param127" [id=2656, type=get_attr]; +"2657 dequantize_per_channel_default_127" [id=2657, type=dequantize_per_channel]; +"2658 linear_126" [id=2658, type=linear]; +"2659 gelu_20" [id=2659, type=gelu]; +"2660 dropout_82" [id=2660, type=dropout]; +"2661 dropout_82_0_0_nncf_smooth_quant_0" [id=2661, type=call_module]; +"2662 quantize_per_tensor_default_128" [id=2662, type=quantize_per_tensor]; +"2663 dequantize_per_tensor_default_128" [id=2663, type=dequantize_per_tensor]; +"2664 linear_127_scale_0" [id=2664, type=get_attr]; +"2665 linear_127_zero_point_0" [id=2665, type=get_attr]; +"2666 _frozen_param128" [id=2666, type=get_attr]; +"2667 dequantize_per_channel_default_128" [id=2667, type=dequantize_per_channel]; +"2668 linear_127" [id=2668, type=linear]; +"2669 dropout_83" [id=2669, type=dropout]; +"2670 layer_norm_44" [id=2670, type=layer_norm]; +"2671 add_72" [id=2671, type=add]; +"2672 features_5_17_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=2672, type=call_module]; +"2673 linear_128_scale_0" [id=2673, type=get_attr]; +"2674 linear_128_zero_point_0" [id=2674, type=get_attr]; +"2675 _frozen_param129" [id=2675, type=get_attr]; +"2676 dequantize_per_channel_default_129" [id=2676, type=dequantize_per_channel]; +"2677 linear_128" [id=2677, type=linear]; +"2678 relu__21" [id=2678, type=relu_]; +"2679 relu__21_0_0_nncf_smooth_quant_0" [id=2679, type=call_module]; +"2680 linear_129_scale_0" [id=2680, type=get_attr]; +"2681 linear_129_zero_point_0" [id=2681, type=get_attr]; +"2682 _frozen_param130" [id=2682, type=get_attr]; +"2683 dequantize_per_channel_default_130" [id=2683, type=dequantize_per_channel]; +"2684 linear_129" [id=2684, type=linear]; +"2685 view_114" [id=2685, type=view]; +"2686 index_21" [id=2686, type=index]; +"2687 view_115" [id=2687, type=view]; +"2688 permute_95" [id=2688, type=permute]; +"2689 contiguous_40" [id=2689, type=contiguous]; +"2690 unsqueeze_61" [id=2690, type=unsqueeze]; +"2691 sigmoid_21" [id=2691, type=sigmoid]; +"2692 mul_42" [id=2692, type=mul]; +"2693 pad_23" [id=2693, type=pad]; +"2694 roll_20" [id=2694, type=roll]; +"2695 view_116" [id=2695, type=view]; +"2696 permute_96" [id=2696, type=permute]; +"2697 reshape_94" [id=2697, type=reshape]; +"2698 clone_21" [id=2698, type=clone]; +"2699 reshape_94_0_0_nncf_smooth_quant_0" [id=2699, type=call_module]; +"2700 quantize_per_tensor_default_129" [id=2700, type=quantize_per_tensor]; +"2701 dequantize_per_tensor_default_129" [id=2701, type=dequantize_per_tensor]; +"2702 linear_130_scale_0" [id=2702, type=get_attr]; +"2703 linear_130_zero_point_0" [id=2703, type=get_attr]; +"2704 _frozen_param131" [id=2704, type=get_attr]; +"2705 dequantize_per_channel_default_131" [id=2705, type=dequantize_per_channel]; +"2706 linear_130" [id=2706, type=linear]; +"2707 reshape_95" [id=2707, type=reshape]; +"2708 permute_97" [id=2708, type=permute]; +"2709 select_63" [id=2709, type=select]; +"2710 select_64" [id=2710, type=select]; +"2711 select_65" [id=2711, type=select]; +"2712 linalg_vector_norm_42" [id=2712, type=linalg_vector_norm]; +"2713 clamp_min_42" [id=2713, type=clamp_min]; +"2714 expand_as_42" [id=2714, type=expand_as]; +"2715 div_42" [id=2715, type=div]; +"2716 quantize_per_tensor_default_130" [id=2716, type=quantize_per_tensor]; +"2717 dequantize_per_tensor_default_130" [id=2717, type=dequantize_per_tensor]; +"2718 linalg_vector_norm_43" [id=2718, type=linalg_vector_norm]; +"2719 clamp_min_43" [id=2719, type=clamp_min]; +"2720 expand_as_43" [id=2720, type=expand_as]; +"2721 div_43" [id=2721, type=div]; +"2722 quantize_per_tensor_default_131" [id=2722, type=quantize_per_tensor]; +"2723 dequantize_per_tensor_default_131" [id=2723, type=dequantize_per_tensor]; +"2724 transpose_42" [id=2724, type=transpose]; +"2725 matmul_42" [id=2725, type=matmul]; +"2726 clamp_21" [id=2726, type=clamp]; +"2727 exp_21" [id=2727, type=exp]; +"2728 mul_43" [id=2728, type=mul]; +"2729 add_73" [id=2729, type=add]; +"2730 new_zeros_10" [id=2730, type=new_zeros]; +"2731 view_117" [id=2731, type=view]; +"2732 permute_98" [id=2732, type=permute]; +"2733 reshape_96" [id=2733, type=reshape]; +"2734 unsqueeze_62" [id=2734, type=unsqueeze]; +"2735 unsqueeze_63" [id=2735, type=unsqueeze]; +"2736 sub_10" [id=2736, type=sub]; +"2737 ne_10" [id=2737, type=ne]; +"2738 masked_fill_20" [id=2738, type=masked_fill]; +"2739 eq_10" [id=2739, type=eq]; +"2740 masked_fill_21" [id=2740, type=masked_fill]; +"2741 view_118" [id=2741, type=view]; +"2742 unsqueeze_64" [id=2742, type=unsqueeze]; +"2743 unsqueeze_65" [id=2743, type=unsqueeze]; +"2744 add_74" [id=2744, type=add]; +"2745 view_119" [id=2745, type=view]; +"2746 softmax_21" [id=2746, type=softmax]; +"2747 dropout_84" [id=2747, type=dropout]; +"2748 matmul_43" [id=2748, type=matmul]; +"2749 transpose_43" [id=2749, type=transpose]; +"2750 reshape_97" [id=2750, type=reshape]; +"2751 reshape_97_0_0_nncf_smooth_quant_0" [id=2751, type=call_module]; +"2752 quantize_per_tensor_default_132" [id=2752, type=quantize_per_tensor]; +"2753 dequantize_per_tensor_default_132" [id=2753, type=dequantize_per_tensor]; +"2754 linear_131_scale_0" [id=2754, type=get_attr]; +"2755 linear_131_zero_point_0" [id=2755, type=get_attr]; +"2756 _frozen_param132" [id=2756, type=get_attr]; +"2757 dequantize_per_channel_default_132" [id=2757, type=dequantize_per_channel]; +"2758 linear_131" [id=2758, type=linear]; +"2759 dropout_85" [id=2759, type=dropout]; +"2760 view_120" [id=2760, type=view]; +"2761 permute_99" [id=2761, type=permute]; +"2762 reshape_98" [id=2762, type=reshape]; +"2763 roll_21" [id=2763, type=roll]; +"2764 slice_325" [id=2764, type=slice]; +"2765 slice_326" [id=2765, type=slice]; +"2766 slice_327" [id=2766, type=slice]; +"2767 slice_328" [id=2767, type=slice]; +"2768 contiguous_41" [id=2768, type=contiguous]; +"2769 layer_norm_45" [id=2769, type=layer_norm]; +"2770 add_75" [id=2770, type=add]; +"2771 add_75_0_0_nncf_smooth_quant_0" [id=2771, type=call_module]; +"2772 quantize_per_tensor_default_133" [id=2772, type=quantize_per_tensor]; +"2773 dequantize_per_tensor_default_133" [id=2773, type=dequantize_per_tensor]; +"2774 linear_132_scale_0" [id=2774, type=get_attr]; +"2775 linear_132_zero_point_0" [id=2775, type=get_attr]; +"2776 _frozen_param133" [id=2776, type=get_attr]; +"2777 dequantize_per_channel_default_133" [id=2777, type=dequantize_per_channel]; +"2778 linear_132" [id=2778, type=linear]; +"2779 gelu_21" [id=2779, type=gelu]; +"2780 dropout_86" [id=2780, type=dropout]; +"2781 dropout_86_0_0_nncf_smooth_quant_0" [id=2781, type=call_module]; +"2782 quantize_per_tensor_default_134" [id=2782, type=quantize_per_tensor]; +"2783 dequantize_per_tensor_default_134" [id=2783, type=dequantize_per_tensor]; +"2784 linear_133_scale_0" [id=2784, type=get_attr]; +"2785 linear_133_zero_point_0" [id=2785, type=get_attr]; +"2786 _frozen_param134" [id=2786, type=get_attr]; +"2787 dequantize_per_channel_default_134" [id=2787, type=dequantize_per_channel]; +"2788 linear_133" [id=2788, type=linear]; +"2789 dropout_87" [id=2789, type=dropout]; +"2790 layer_norm_46" [id=2790, type=layer_norm]; +"2791 add_76" [id=2791, type=add]; +"2792 pad_24" [id=2792, type=pad]; +"2793 slice_329" [id=2793, type=slice]; +"2794 slice_330" [id=2794, type=slice]; +"2795 slice_331" [id=2795, type=slice]; +"2796 slice_332" [id=2796, type=slice]; +"2797 slice_333" [id=2797, type=slice]; +"2798 slice_334" [id=2798, type=slice]; +"2799 slice_335" [id=2799, type=slice]; +"2800 slice_336" [id=2800, type=slice]; +"2801 slice_337" [id=2801, type=slice]; +"2802 slice_338" [id=2802, type=slice]; +"2803 slice_339" [id=2803, type=slice]; +"2804 slice_340" [id=2804, type=slice]; +"2805 cat_2" [id=2805, type=cat]; +"2806 cat_2_0_0_nncf_smooth_quant_0" [id=2806, type=call_module]; +"2807 quantize_per_tensor_default_135" [id=2807, type=quantize_per_tensor]; +"2808 dequantize_per_tensor_default_135" [id=2808, type=dequantize_per_tensor]; +"2809 linear_134_scale_0" [id=2809, type=get_attr]; +"2810 linear_134_zero_point_0" [id=2810, type=get_attr]; +"2811 _frozen_param135" [id=2811, type=get_attr]; +"2812 dequantize_per_channel_default_135" [id=2812, type=dequantize_per_channel]; +"2813 linear_134" [id=2813, type=linear]; +"2814 layer_norm_47" [id=2814, type=layer_norm]; +"2815 features_7_0_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=2815, type=call_module]; +"2816 linear_135_scale_0" [id=2816, type=get_attr]; +"2817 linear_135_zero_point_0" [id=2817, type=get_attr]; +"2818 _frozen_param136" [id=2818, type=get_attr]; +"2819 dequantize_per_channel_default_136" [id=2819, type=dequantize_per_channel]; +"2820 linear_135" [id=2820, type=linear]; +"2821 relu__22" [id=2821, type=relu_]; +"2822 relu__22_0_0_nncf_smooth_quant_0" [id=2822, type=call_module]; +"2823 linear_136_scale_0" [id=2823, type=get_attr]; +"2824 linear_136_zero_point_0" [id=2824, type=get_attr]; +"2825 _frozen_param137" [id=2825, type=get_attr]; +"2826 dequantize_per_channel_default_137" [id=2826, type=dequantize_per_channel]; +"2827 linear_136" [id=2827, type=linear]; +"2828 view_121" [id=2828, type=view]; +"2829 index_22" [id=2829, type=index]; +"2830 view_122" [id=2830, type=view]; +"2831 permute_100" [id=2831, type=permute]; +"2832 contiguous_42" [id=2832, type=contiguous]; +"2833 unsqueeze_66" [id=2833, type=unsqueeze]; +"2834 sigmoid_22" [id=2834, type=sigmoid]; +"2835 mul_44" [id=2835, type=mul]; +"2836 pad_25" [id=2836, type=pad]; +"2837 view_123" [id=2837, type=view]; +"2838 permute_101" [id=2838, type=permute]; +"2839 reshape_99" [id=2839, type=reshape]; +"2840 clone_22" [id=2840, type=clone]; +"2841 reshape_99_0_0_nncf_smooth_quant_0" [id=2841, type=call_module]; +"2842 quantize_per_tensor_default_136" [id=2842, type=quantize_per_tensor]; +"2843 dequantize_per_tensor_default_136" [id=2843, type=dequantize_per_tensor]; +"2844 linear_137_scale_0" [id=2844, type=get_attr]; +"2845 linear_137_zero_point_0" [id=2845, type=get_attr]; +"2846 _frozen_param138" [id=2846, type=get_attr]; +"2847 dequantize_per_channel_default_138" [id=2847, type=dequantize_per_channel]; +"2848 linear_137" [id=2848, type=linear]; +"2849 reshape_100" [id=2849, type=reshape]; +"2850 permute_102" [id=2850, type=permute]; +"2851 select_66" [id=2851, type=select]; +"2852 select_67" [id=2852, type=select]; +"2853 select_68" [id=2853, type=select]; +"2854 linalg_vector_norm_44" [id=2854, type=linalg_vector_norm]; +"2855 clamp_min_44" [id=2855, type=clamp_min]; +"2856 expand_as_44" [id=2856, type=expand_as]; +"2857 div_44" [id=2857, type=div]; +"2858 quantize_per_tensor_default_137" [id=2858, type=quantize_per_tensor]; +"2859 dequantize_per_tensor_default_137" [id=2859, type=dequantize_per_tensor]; +"2860 linalg_vector_norm_45" [id=2860, type=linalg_vector_norm]; +"2861 clamp_min_45" [id=2861, type=clamp_min]; +"2862 expand_as_45" [id=2862, type=expand_as]; +"2863 div_45" [id=2863, type=div]; +"2864 quantize_per_tensor_default_138" [id=2864, type=quantize_per_tensor]; +"2865 dequantize_per_tensor_default_138" [id=2865, type=dequantize_per_tensor]; +"2866 transpose_44" [id=2866, type=transpose]; +"2867 matmul_44" [id=2867, type=matmul]; +"2868 clamp_22" [id=2868, type=clamp]; +"2869 exp_22" [id=2869, type=exp]; +"2870 mul_45" [id=2870, type=mul]; +"2871 add_77" [id=2871, type=add]; +"2872 softmax_22" [id=2872, type=softmax]; +"2873 dropout_88" [id=2873, type=dropout]; +"2874 matmul_45" [id=2874, type=matmul]; +"2875 transpose_45" [id=2875, type=transpose]; +"2876 reshape_101" [id=2876, type=reshape]; +"2877 reshape_101_0_0_nncf_smooth_quant_0" [id=2877, type=call_module]; +"2878 quantize_per_tensor_default_139" [id=2878, type=quantize_per_tensor]; +"2879 dequantize_per_tensor_default_139" [id=2879, type=dequantize_per_tensor]; +"2880 linear_138_scale_0" [id=2880, type=get_attr]; +"2881 linear_138_zero_point_0" [id=2881, type=get_attr]; +"2882 _frozen_param139" [id=2882, type=get_attr]; +"2883 dequantize_per_channel_default_139" [id=2883, type=dequantize_per_channel]; +"2884 linear_138" [id=2884, type=linear]; +"2885 dropout_89" [id=2885, type=dropout]; +"2886 view_124" [id=2886, type=view]; +"2887 permute_103" [id=2887, type=permute]; +"2888 reshape_102" [id=2888, type=reshape]; +"2889 slice_342" [id=2889, type=slice]; +"2890 slice_343" [id=2890, type=slice]; +"2891 slice_344" [id=2891, type=slice]; +"2892 slice_345" [id=2892, type=slice]; +"2893 contiguous_43" [id=2893, type=contiguous]; +"2894 layer_norm_48" [id=2894, type=layer_norm]; +"2895 add_78" [id=2895, type=add]; +"2896 add_78_0_0_nncf_smooth_quant_0" [id=2896, type=call_module]; +"2897 quantize_per_tensor_default_140" [id=2897, type=quantize_per_tensor]; +"2898 dequantize_per_tensor_default_140" [id=2898, type=dequantize_per_tensor]; +"2899 linear_139_scale_0" [id=2899, type=get_attr]; +"2900 linear_139_zero_point_0" [id=2900, type=get_attr]; +"2901 _frozen_param140" [id=2901, type=get_attr]; +"2902 dequantize_per_channel_default_140" [id=2902, type=dequantize_per_channel]; +"2903 linear_139" [id=2903, type=linear]; +"2904 gelu_22" [id=2904, type=gelu]; +"2905 dropout_90" [id=2905, type=dropout]; +"2906 dropout_90_0_0_nncf_smooth_quant_0" [id=2906, type=call_module]; +"2907 quantize_per_tensor_default_141" [id=2907, type=quantize_per_tensor]; +"2908 dequantize_per_tensor_default_141" [id=2908, type=dequantize_per_tensor]; +"2909 linear_140_scale_0" [id=2909, type=get_attr]; +"2910 linear_140_zero_point_0" [id=2910, type=get_attr]; +"2911 _frozen_param141" [id=2911, type=get_attr]; +"2912 dequantize_per_channel_default_141" [id=2912, type=dequantize_per_channel]; +"2913 linear_140" [id=2913, type=linear]; +"2914 dropout_91" [id=2914, type=dropout]; +"2915 layer_norm_49" [id=2915, type=layer_norm]; +"2916 add_79" [id=2916, type=add]; +"2917 features_7_1_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=2917, type=call_module]; +"2918 linear_141_scale_0" [id=2918, type=get_attr]; +"2919 linear_141_zero_point_0" [id=2919, type=get_attr]; +"2920 _frozen_param142" [id=2920, type=get_attr]; +"2921 dequantize_per_channel_default_142" [id=2921, type=dequantize_per_channel]; +"2922 linear_141" [id=2922, type=linear]; +"2923 relu__23" [id=2923, type=relu_]; +"2924 relu__23_0_0_nncf_smooth_quant_0" [id=2924, type=call_module]; +"2925 linear_142_scale_0" [id=2925, type=get_attr]; +"2926 linear_142_zero_point_0" [id=2926, type=get_attr]; +"2927 _frozen_param143" [id=2927, type=get_attr]; +"2928 dequantize_per_channel_default_143" [id=2928, type=dequantize_per_channel]; +"2929 linear_142" [id=2929, type=linear]; +"2930 view_125" [id=2930, type=view]; +"2931 index_23" [id=2931, type=index]; +"2932 view_126" [id=2932, type=view]; +"2933 permute_104" [id=2933, type=permute]; +"2934 contiguous_44" [id=2934, type=contiguous]; +"2935 unsqueeze_67" [id=2935, type=unsqueeze]; +"2936 sigmoid_23" [id=2936, type=sigmoid]; +"2937 mul_46" [id=2937, type=mul]; +"2938 pad_26" [id=2938, type=pad]; +"2939 view_127" [id=2939, type=view]; +"2940 permute_105" [id=2940, type=permute]; +"2941 reshape_103" [id=2941, type=reshape]; +"2942 clone_23" [id=2942, type=clone]; +"2943 reshape_103_0_0_nncf_smooth_quant_0" [id=2943, type=call_module]; +"2944 quantize_per_tensor_default_142" [id=2944, type=quantize_per_tensor]; +"2945 dequantize_per_tensor_default_142" [id=2945, type=dequantize_per_tensor]; +"2946 linear_143_scale_0" [id=2946, type=get_attr]; +"2947 linear_143_zero_point_0" [id=2947, type=get_attr]; +"2948 _frozen_param144" [id=2948, type=get_attr]; +"2949 dequantize_per_channel_default_144" [id=2949, type=dequantize_per_channel]; +"2950 linear_143" [id=2950, type=linear]; +"2951 reshape_104" [id=2951, type=reshape]; +"2952 permute_106" [id=2952, type=permute]; +"2953 select_69" [id=2953, type=select]; +"2954 select_70" [id=2954, type=select]; +"2955 select_71" [id=2955, type=select]; +"2956 linalg_vector_norm_46" [id=2956, type=linalg_vector_norm]; +"2957 clamp_min_46" [id=2957, type=clamp_min]; +"2958 expand_as_46" [id=2958, type=expand_as]; +"2959 div_46" [id=2959, type=div]; +"2960 quantize_per_tensor_default_143" [id=2960, type=quantize_per_tensor]; +"2961 dequantize_per_tensor_default_143" [id=2961, type=dequantize_per_tensor]; +"2962 linalg_vector_norm_47" [id=2962, type=linalg_vector_norm]; +"2963 clamp_min_47" [id=2963, type=clamp_min]; +"2964 expand_as_47" [id=2964, type=expand_as]; +"2965 div_47" [id=2965, type=div]; +"2966 quantize_per_tensor_default_144" [id=2966, type=quantize_per_tensor]; +"2967 dequantize_per_tensor_default_144" [id=2967, type=dequantize_per_tensor]; +"2968 transpose_46" [id=2968, type=transpose]; +"2969 matmul_46" [id=2969, type=matmul]; +"2970 clamp_23" [id=2970, type=clamp]; +"2971 exp_23" [id=2971, type=exp]; +"2972 mul_47" [id=2972, type=mul]; +"2973 add_80" [id=2973, type=add]; +"2974 softmax_23" [id=2974, type=softmax]; +"2975 dropout_92" [id=2975, type=dropout]; +"2976 matmul_47" [id=2976, type=matmul]; +"2977 transpose_47" [id=2977, type=transpose]; +"2978 reshape_105" [id=2978, type=reshape]; +"2979 reshape_105_0_0_nncf_smooth_quant_0" [id=2979, type=call_module]; +"2980 quantize_per_tensor_default_145" [id=2980, type=quantize_per_tensor]; +"2981 dequantize_per_tensor_default_145" [id=2981, type=dequantize_per_tensor]; +"2982 linear_144_scale_0" [id=2982, type=get_attr]; +"2983 linear_144_zero_point_0" [id=2983, type=get_attr]; +"2984 _frozen_param145" [id=2984, type=get_attr]; +"2985 dequantize_per_channel_default_145" [id=2985, type=dequantize_per_channel]; +"2986 linear_144" [id=2986, type=linear]; +"2987 dropout_93" [id=2987, type=dropout]; +"2988 view_128" [id=2988, type=view]; +"2989 permute_107" [id=2989, type=permute]; +"2990 reshape_106" [id=2990, type=reshape]; +"2991 slice_347" [id=2991, type=slice]; +"2992 slice_348" [id=2992, type=slice]; +"2993 slice_349" [id=2993, type=slice]; +"2994 slice_350" [id=2994, type=slice]; +"2995 contiguous_45" [id=2995, type=contiguous]; +"2996 layer_norm_50" [id=2996, type=layer_norm]; +"2997 add_81" [id=2997, type=add]; +"2998 add_81_0_0_nncf_smooth_quant_0" [id=2998, type=call_module]; +"2999 quantize_per_tensor_default_146" [id=2999, type=quantize_per_tensor]; +"3000 dequantize_per_tensor_default_146" [id=3000, type=dequantize_per_tensor]; +"3001 linear_145_scale_0" [id=3001, type=get_attr]; +"3002 linear_145_zero_point_0" [id=3002, type=get_attr]; +"3003 _frozen_param146" [id=3003, type=get_attr]; +"3004 dequantize_per_channel_default_146" [id=3004, type=dequantize_per_channel]; +"3005 linear_145" [id=3005, type=linear]; +"3006 gelu_23" [id=3006, type=gelu]; +"3007 dropout_94" [id=3007, type=dropout]; +"3008 dropout_94_0_0_nncf_smooth_quant_0" [id=3008, type=call_module]; +"3009 quantize_per_tensor_default_147" [id=3009, type=quantize_per_tensor]; +"3010 dequantize_per_tensor_default_147" [id=3010, type=dequantize_per_tensor]; +"3011 linear_146_scale_0" [id=3011, type=get_attr]; +"3012 linear_146_zero_point_0" [id=3012, type=get_attr]; +"3013 _frozen_param147" [id=3013, type=get_attr]; +"3014 dequantize_per_channel_default_147" [id=3014, type=dequantize_per_channel]; +"3015 linear_146" [id=3015, type=linear]; +"3016 dropout_95" [id=3016, type=dropout]; +"3017 layer_norm_51" [id=3017, type=layer_norm]; +"3018 add_82" [id=3018, type=add]; +"3019 layer_norm_52" [id=3019, type=layer_norm]; +"3020 permute_108" [id=3020, type=permute]; +"3021 adaptive_avg_pool2d" [id=3021, type=adaptive_avg_pool2d]; +"3022 flatten" [id=3022, type=flatten]; +"3023 flatten_0_0_nncf_smooth_quant_0" [id=3023, type=call_module]; +"3024 quantize_per_tensor_default_148" [id=3024, type=quantize_per_tensor]; +"3025 dequantize_per_tensor_default_148" [id=3025, type=dequantize_per_tensor]; +"3026 linear_147_scale_0" [id=3026, type=get_attr]; +"3027 linear_147_zero_point_0" [id=3027, type=get_attr]; +"3028 _frozen_param148" [id=3028, type=get_attr]; +"3029 dequantize_per_channel_default_148" [id=3029, type=dequantize_per_channel]; +"3030 linear_147" [id=3030, type=linear]; +"3031 output" [id=3031, type=output]; +"0 features_1_0_attn_qkv_bias" -> "335 clone" [label="(288,)", style=solid]; +"1 features_1_0_attn_proj_bias" -> "379 linear_3" [label="(96,)", style=solid]; +"2 features_1_0_attn_logit_scale" -> "363 clamp" [label="(3, 1, 1)", style=solid]; +"3 features_1_1_attn_qkv_bias" -> "435 clone_1" [label="(288,)", style=solid]; +"4 features_1_1_attn_proj_bias" -> "495 linear_9" [label="(96,)", style=solid]; +"5 features_1_1_attn_logit_scale" -> "463 clamp_1" [label="(3, 1, 1)", style=solid]; +"6 features_3_0_attn_qkv_bias" -> "574 clone_2" [label="(576,)", style=solid]; +"7 features_3_0_attn_proj_bias" -> "618 linear_16" [label="(192,)", style=solid]; +"8 features_3_0_attn_logit_scale" -> "602 clamp_2" [label="(6, 1, 1)", style=solid]; +"9 features_3_1_attn_qkv_bias" -> "677 clone_3" [label="(576,)", style=solid]; +"10 features_3_1_attn_proj_bias" -> "737 linear_22" [label="(192,)", style=solid]; +"11 features_3_1_attn_logit_scale" -> "705 clamp_3" [label="(6, 1, 1)", style=solid]; +"12 features_5_0_attn_qkv_bias" -> "819 clone_4" [label="(1152,)", style=solid]; +"13 features_5_0_attn_proj_bias" -> "863 linear_29" [label="(384,)", style=solid]; +"14 features_5_0_attn_logit_scale" -> "847 clamp_4" [label="(12, 1, 1)", style=solid]; +"15 features_5_1_attn_qkv_bias" -> "922 clone_5" [label="(1152,)", style=solid]; +"16 features_5_1_attn_proj_bias" -> "982 linear_35" [label="(384,)", style=solid]; +"17 features_5_1_attn_logit_scale" -> "950 clamp_5" [label="(12, 1, 1)", style=solid]; +"18 features_5_2_attn_qkv_bias" -> "1041 clone_6" [label="(1152,)", style=solid]; +"19 features_5_2_attn_proj_bias" -> "1085 linear_41" [label="(384,)", style=solid]; +"20 features_5_2_attn_logit_scale" -> "1069 clamp_6" [label="(12, 1, 1)", style=solid]; +"21 features_5_3_attn_qkv_bias" -> "1144 clone_7" [label="(1152,)", style=solid]; +"22 features_5_3_attn_proj_bias" -> "1204 linear_47" [label="(384,)", style=solid]; +"23 features_5_3_attn_logit_scale" -> "1172 clamp_7" [label="(12, 1, 1)", style=solid]; +"24 features_5_4_attn_qkv_bias" -> "1263 clone_8" [label="(1152,)", style=solid]; +"25 features_5_4_attn_proj_bias" -> "1307 linear_53" [label="(384,)", style=solid]; +"26 features_5_4_attn_logit_scale" -> "1291 clamp_8" [label="(12, 1, 1)", style=solid]; +"27 features_5_5_attn_qkv_bias" -> "1366 clone_9" [label="(1152,)", style=solid]; +"28 features_5_5_attn_proj_bias" -> "1426 linear_59" [label="(384,)", style=solid]; +"29 features_5_5_attn_logit_scale" -> "1394 clamp_9" [label="(12, 1, 1)", style=solid]; +"30 features_5_6_attn_qkv_bias" -> "1485 clone_10" [label="(1152,)", style=solid]; +"31 features_5_6_attn_proj_bias" -> "1529 linear_65" [label="(384,)", style=solid]; +"32 features_5_6_attn_logit_scale" -> "1513 clamp_10" [label="(12, 1, 1)", style=solid]; +"33 features_5_7_attn_qkv_bias" -> "1588 clone_11" [label="(1152,)", style=solid]; +"34 features_5_7_attn_proj_bias" -> "1648 linear_71" [label="(384,)", style=solid]; +"35 features_5_7_attn_logit_scale" -> "1616 clamp_11" [label="(12, 1, 1)", style=solid]; +"36 features_5_8_attn_qkv_bias" -> "1707 clone_12" [label="(1152,)", style=solid]; +"37 features_5_8_attn_proj_bias" -> "1751 linear_77" [label="(384,)", style=solid]; +"38 features_5_8_attn_logit_scale" -> "1735 clamp_12" [label="(12, 1, 1)", style=solid]; +"39 features_5_9_attn_qkv_bias" -> "1810 clone_13" [label="(1152,)", style=solid]; +"40 features_5_9_attn_proj_bias" -> "1870 linear_83" [label="(384,)", style=solid]; +"41 features_5_9_attn_logit_scale" -> "1838 clamp_13" [label="(12, 1, 1)", style=solid]; +"42 features_5_10_attn_qkv_bias" -> "1929 clone_14" [label="(1152,)", style=solid]; +"43 features_5_10_attn_proj_bias" -> "1973 linear_89" [label="(384,)", style=solid]; +"44 features_5_10_attn_logit_scale" -> "1957 clamp_14" [label="(12, 1, 1)", style=solid]; +"45 features_5_11_attn_qkv_bias" -> "2032 clone_15" [label="(1152,)", style=solid]; +"46 features_5_11_attn_proj_bias" -> "2092 linear_95" [label="(384,)", style=solid]; +"47 features_5_11_attn_logit_scale" -> "2060 clamp_15" [label="(12, 1, 1)", style=solid]; +"48 features_5_12_attn_qkv_bias" -> "2151 clone_16" [label="(1152,)", style=solid]; +"49 features_5_12_attn_proj_bias" -> "2195 linear_101" [label="(384,)", style=solid]; +"50 features_5_12_attn_logit_scale" -> "2179 clamp_16" [label="(12, 1, 1)", style=solid]; +"51 features_5_13_attn_qkv_bias" -> "2254 clone_17" [label="(1152,)", style=solid]; +"52 features_5_13_attn_proj_bias" -> "2314 linear_107" [label="(384,)", style=solid]; +"53 features_5_13_attn_logit_scale" -> "2282 clamp_17" [label="(12, 1, 1)", style=solid]; +"54 features_5_14_attn_qkv_bias" -> "2373 clone_18" [label="(1152,)", style=solid]; +"55 features_5_14_attn_proj_bias" -> "2417 linear_113" [label="(384,)", style=solid]; +"56 features_5_14_attn_logit_scale" -> "2401 clamp_18" [label="(12, 1, 1)", style=solid]; +"57 features_5_15_attn_qkv_bias" -> "2476 clone_19" [label="(1152,)", style=solid]; +"58 features_5_15_attn_proj_bias" -> "2536 linear_119" [label="(384,)", style=solid]; +"59 features_5_15_attn_logit_scale" -> "2504 clamp_19" [label="(12, 1, 1)", style=solid]; +"60 features_5_16_attn_qkv_bias" -> "2595 clone_20" [label="(1152,)", style=solid]; +"61 features_5_16_attn_proj_bias" -> "2639 linear_125" [label="(384,)", style=solid]; +"62 features_5_16_attn_logit_scale" -> "2623 clamp_20" [label="(12, 1, 1)", style=solid]; +"63 features_5_17_attn_qkv_bias" -> "2698 clone_21" [label="(1152,)", style=solid]; +"64 features_5_17_attn_proj_bias" -> "2758 linear_131" [label="(384,)", style=solid]; +"65 features_5_17_attn_logit_scale" -> "2726 clamp_21" [label="(12, 1, 1)", style=solid]; +"66 features_7_0_attn_qkv_bias" -> "2840 clone_22" [label="(2304,)", style=solid]; +"67 features_7_0_attn_proj_bias" -> "2884 linear_138" [label="(768,)", style=solid]; +"68 features_7_0_attn_logit_scale" -> "2868 clamp_22" [label="(24, 1, 1)", style=solid]; +"69 features_7_1_attn_qkv_bias" -> "2942 clone_23" [label="(2304,)", style=solid]; +"70 features_7_1_attn_proj_bias" -> "2986 linear_144" [label="(768,)", style=solid]; +"71 features_7_1_attn_logit_scale" -> "2970 clamp_23" [label="(24, 1, 1)", style=solid]; +"72 features_0_0_bias" -> "307 conv2d" [label="(96,)", style=solid]; +"73 features_0_2_weight" -> "309 layer_norm" [label="(96,)", style=solid]; +"74 features_0_2_bias" -> "309 layer_norm" [label="(96,)", style=solid]; +"75 features_1_0_attn_cpb_mlp_0_bias" -> "315 linear" [label="(512,)", style=solid]; +"76 features_1_0_norm1_weight" -> "386 layer_norm_1" [label="(96,)", style=solid]; +"77 features_1_0_norm1_bias" -> "386 layer_norm_1" [label="(96,)", style=solid]; +"78 features_1_0_mlp_0_bias" -> "395 linear_4" [label="(384,)", style=solid]; +"79 features_1_0_mlp_3_bias" -> "405 linear_5" [label="(96,)", style=solid]; +"80 features_1_0_norm2_weight" -> "407 layer_norm_2" [label="(96,)", style=solid]; +"81 features_1_0_norm2_bias" -> "407 layer_norm_2" [label="(96,)", style=solid]; +"82 features_1_1_attn_cpb_mlp_0_bias" -> "414 linear_6" [label="(512,)", style=solid]; +"83 features_1_1_norm1_weight" -> "503 layer_norm_3" [label="(96,)", style=solid]; +"84 features_1_1_norm1_bias" -> "503 layer_norm_3" [label="(96,)", style=solid]; +"85 features_1_1_mlp_0_bias" -> "512 linear_10" [label="(384,)", style=solid]; +"86 features_1_1_mlp_3_bias" -> "522 linear_11" [label="(96,)", style=solid]; +"87 features_1_1_norm2_weight" -> "524 layer_norm_4" [label="(96,)", style=solid]; +"88 features_1_1_norm2_bias" -> "524 layer_norm_4" [label="(96,)", style=solid]; +"89 features_2_norm_weight" -> "548 layer_norm_5" [label="(192,)", style=solid]; +"90 features_2_norm_bias" -> "548 layer_norm_5" [label="(192,)", style=solid]; +"91 features_3_0_attn_cpb_mlp_0_bias" -> "554 linear_13" [label="(512,)", style=solid]; +"92 features_3_0_norm1_weight" -> "628 layer_norm_6" [label="(192,)", style=solid]; +"93 features_3_0_norm1_bias" -> "628 layer_norm_6" [label="(192,)", style=solid]; +"94 features_3_0_mlp_0_bias" -> "637 linear_17" [label="(768,)", style=solid]; +"95 features_3_0_mlp_3_bias" -> "647 linear_18" [label="(192,)", style=solid]; +"96 features_3_0_norm2_weight" -> "649 layer_norm_7" [label="(192,)", style=solid]; +"97 features_3_0_norm2_bias" -> "649 layer_norm_7" [label="(192,)", style=solid]; +"98 features_3_1_attn_cpb_mlp_0_bias" -> "656 linear_19" [label="(512,)", style=solid]; +"99 features_3_1_norm1_weight" -> "748 layer_norm_8" [label="(192,)", style=solid]; +"100 features_3_1_norm1_bias" -> "748 layer_norm_8" [label="(192,)", style=solid]; +"101 features_3_1_mlp_0_bias" -> "757 linear_23" [label="(768,)", style=solid]; +"102 features_3_1_mlp_3_bias" -> "767 linear_24" [label="(192,)", style=solid]; +"103 features_3_1_norm2_weight" -> "769 layer_norm_9" [label="(192,)", style=solid]; +"104 features_3_1_norm2_bias" -> "769 layer_norm_9" [label="(192,)", style=solid]; +"105 features_4_norm_weight" -> "793 layer_norm_10" [label="(384,)", style=solid]; +"106 features_4_norm_bias" -> "793 layer_norm_10" [label="(384,)", style=solid]; +"107 features_5_0_attn_cpb_mlp_0_bias" -> "799 linear_26" [label="(512,)", style=solid]; +"108 features_5_0_norm1_weight" -> "873 layer_norm_11" [label="(384,)", style=solid]; +"109 features_5_0_norm1_bias" -> "873 layer_norm_11" [label="(384,)", style=solid]; +"110 features_5_0_mlp_0_bias" -> "882 linear_30" [label="(1536,)", style=solid]; +"111 features_5_0_mlp_3_bias" -> "892 linear_31" [label="(384,)", style=solid]; +"112 features_5_0_norm2_weight" -> "894 layer_norm_12" [label="(384,)", style=solid]; +"113 features_5_0_norm2_bias" -> "894 layer_norm_12" [label="(384,)", style=solid]; +"114 features_5_1_attn_cpb_mlp_0_bias" -> "901 linear_32" [label="(512,)", style=solid]; +"115 features_5_1_norm1_weight" -> "993 layer_norm_13" [label="(384,)", style=solid]; +"116 features_5_1_norm1_bias" -> "993 layer_norm_13" [label="(384,)", style=solid]; +"117 features_5_1_mlp_0_bias" -> "1002 linear_36" [label="(1536,)", style=solid]; +"118 features_5_1_mlp_3_bias" -> "1012 linear_37" [label="(384,)", style=solid]; +"119 features_5_1_norm2_weight" -> "1014 layer_norm_14" [label="(384,)", style=solid]; +"120 features_5_1_norm2_bias" -> "1014 layer_norm_14" [label="(384,)", style=solid]; +"121 features_5_2_attn_cpb_mlp_0_bias" -> "1021 linear_38" [label="(512,)", style=solid]; +"122 features_5_2_norm1_weight" -> "1095 layer_norm_15" [label="(384,)", style=solid]; +"123 features_5_2_norm1_bias" -> "1095 layer_norm_15" [label="(384,)", style=solid]; +"124 features_5_2_mlp_0_bias" -> "1104 linear_42" [label="(1536,)", style=solid]; +"125 features_5_2_mlp_3_bias" -> "1114 linear_43" [label="(384,)", style=solid]; +"126 features_5_2_norm2_weight" -> "1116 layer_norm_16" [label="(384,)", style=solid]; +"127 features_5_2_norm2_bias" -> "1116 layer_norm_16" [label="(384,)", style=solid]; +"128 features_5_3_attn_cpb_mlp_0_bias" -> "1123 linear_44" [label="(512,)", style=solid]; +"129 features_5_3_norm1_weight" -> "1215 layer_norm_17" [label="(384,)", style=solid]; +"130 features_5_3_norm1_bias" -> "1215 layer_norm_17" [label="(384,)", style=solid]; +"131 features_5_3_mlp_0_bias" -> "1224 linear_48" [label="(1536,)", style=solid]; +"132 features_5_3_mlp_3_bias" -> "1234 linear_49" [label="(384,)", style=solid]; +"133 features_5_3_norm2_weight" -> "1236 layer_norm_18" [label="(384,)", style=solid]; +"134 features_5_3_norm2_bias" -> "1236 layer_norm_18" [label="(384,)", style=solid]; +"135 features_5_4_attn_cpb_mlp_0_bias" -> "1243 linear_50" [label="(512,)", style=solid]; +"136 features_5_4_norm1_weight" -> "1317 layer_norm_19" [label="(384,)", style=solid]; +"137 features_5_4_norm1_bias" -> "1317 layer_norm_19" [label="(384,)", style=solid]; +"138 features_5_4_mlp_0_bias" -> "1326 linear_54" [label="(1536,)", style=solid]; +"139 features_5_4_mlp_3_bias" -> "1336 linear_55" [label="(384,)", style=solid]; +"140 features_5_4_norm2_weight" -> "1338 layer_norm_20" [label="(384,)", style=solid]; +"141 features_5_4_norm2_bias" -> "1338 layer_norm_20" [label="(384,)", style=solid]; +"142 features_5_5_attn_cpb_mlp_0_bias" -> "1345 linear_56" [label="(512,)", style=solid]; +"143 features_5_5_norm1_weight" -> "1437 layer_norm_21" [label="(384,)", style=solid]; +"144 features_5_5_norm1_bias" -> "1437 layer_norm_21" [label="(384,)", style=solid]; +"145 features_5_5_mlp_0_bias" -> "1446 linear_60" [label="(1536,)", style=solid]; +"146 features_5_5_mlp_3_bias" -> "1456 linear_61" [label="(384,)", style=solid]; +"147 features_5_5_norm2_weight" -> "1458 layer_norm_22" [label="(384,)", style=solid]; +"148 features_5_5_norm2_bias" -> "1458 layer_norm_22" [label="(384,)", style=solid]; +"149 features_5_6_attn_cpb_mlp_0_bias" -> "1465 linear_62" [label="(512,)", style=solid]; +"150 features_5_6_norm1_weight" -> "1539 layer_norm_23" [label="(384,)", style=solid]; +"151 features_5_6_norm1_bias" -> "1539 layer_norm_23" [label="(384,)", style=solid]; +"152 features_5_6_mlp_0_bias" -> "1548 linear_66" [label="(1536,)", style=solid]; +"153 features_5_6_mlp_3_bias" -> "1558 linear_67" [label="(384,)", style=solid]; +"154 features_5_6_norm2_weight" -> "1560 layer_norm_24" [label="(384,)", style=solid]; +"155 features_5_6_norm2_bias" -> "1560 layer_norm_24" [label="(384,)", style=solid]; +"156 features_5_7_attn_cpb_mlp_0_bias" -> "1567 linear_68" [label="(512,)", style=solid]; +"157 features_5_7_norm1_weight" -> "1659 layer_norm_25" [label="(384,)", style=solid]; +"158 features_5_7_norm1_bias" -> "1659 layer_norm_25" [label="(384,)", style=solid]; +"159 features_5_7_mlp_0_bias" -> "1668 linear_72" [label="(1536,)", style=solid]; +"160 features_5_7_mlp_3_bias" -> "1678 linear_73" [label="(384,)", style=solid]; +"161 features_5_7_norm2_weight" -> "1680 layer_norm_26" [label="(384,)", style=solid]; +"162 features_5_7_norm2_bias" -> "1680 layer_norm_26" [label="(384,)", style=solid]; +"163 features_5_8_attn_cpb_mlp_0_bias" -> "1687 linear_74" [label="(512,)", style=solid]; +"164 features_5_8_norm1_weight" -> "1761 layer_norm_27" [label="(384,)", style=solid]; +"165 features_5_8_norm1_bias" -> "1761 layer_norm_27" [label="(384,)", style=solid]; +"166 features_5_8_mlp_0_bias" -> "1770 linear_78" [label="(1536,)", style=solid]; +"167 features_5_8_mlp_3_bias" -> "1780 linear_79" [label="(384,)", style=solid]; +"168 features_5_8_norm2_weight" -> "1782 layer_norm_28" [label="(384,)", style=solid]; +"169 features_5_8_norm2_bias" -> "1782 layer_norm_28" [label="(384,)", style=solid]; +"170 features_5_9_attn_cpb_mlp_0_bias" -> "1789 linear_80" [label="(512,)", style=solid]; +"171 features_5_9_norm1_weight" -> "1881 layer_norm_29" [label="(384,)", style=solid]; +"172 features_5_9_norm1_bias" -> "1881 layer_norm_29" [label="(384,)", style=solid]; +"173 features_5_9_mlp_0_bias" -> "1890 linear_84" [label="(1536,)", style=solid]; +"174 features_5_9_mlp_3_bias" -> "1900 linear_85" [label="(384,)", style=solid]; +"175 features_5_9_norm2_weight" -> "1902 layer_norm_30" [label="(384,)", style=solid]; +"176 features_5_9_norm2_bias" -> "1902 layer_norm_30" [label="(384,)", style=solid]; +"177 features_5_10_attn_cpb_mlp_0_bias" -> "1909 linear_86" [label="(512,)", style=solid]; +"178 features_5_10_norm1_weight" -> "1983 layer_norm_31" [label="(384,)", style=solid]; +"179 features_5_10_norm1_bias" -> "1983 layer_norm_31" [label="(384,)", style=solid]; +"180 features_5_10_mlp_0_bias" -> "1992 linear_90" [label="(1536,)", style=solid]; +"181 features_5_10_mlp_3_bias" -> "2002 linear_91" [label="(384,)", style=solid]; +"182 features_5_10_norm2_weight" -> "2004 layer_norm_32" [label="(384,)", style=solid]; +"183 features_5_10_norm2_bias" -> "2004 layer_norm_32" [label="(384,)", style=solid]; +"184 features_5_11_attn_cpb_mlp_0_bias" -> "2011 linear_92" [label="(512,)", style=solid]; +"185 features_5_11_norm1_weight" -> "2103 layer_norm_33" [label="(384,)", style=solid]; +"186 features_5_11_norm1_bias" -> "2103 layer_norm_33" [label="(384,)", style=solid]; +"187 features_5_11_mlp_0_bias" -> "2112 linear_96" [label="(1536,)", style=solid]; +"188 features_5_11_mlp_3_bias" -> "2122 linear_97" [label="(384,)", style=solid]; +"189 features_5_11_norm2_weight" -> "2124 layer_norm_34" [label="(384,)", style=solid]; +"190 features_5_11_norm2_bias" -> "2124 layer_norm_34" [label="(384,)", style=solid]; +"191 features_5_12_attn_cpb_mlp_0_bias" -> "2131 linear_98" [label="(512,)", style=solid]; +"192 features_5_12_norm1_weight" -> "2205 layer_norm_35" [label="(384,)", style=solid]; +"193 features_5_12_norm1_bias" -> "2205 layer_norm_35" [label="(384,)", style=solid]; +"194 features_5_12_mlp_0_bias" -> "2214 linear_102" [label="(1536,)", style=solid]; +"195 features_5_12_mlp_3_bias" -> "2224 linear_103" [label="(384,)", style=solid]; +"196 features_5_12_norm2_weight" -> "2226 layer_norm_36" [label="(384,)", style=solid]; +"197 features_5_12_norm2_bias" -> "2226 layer_norm_36" [label="(384,)", style=solid]; +"198 features_5_13_attn_cpb_mlp_0_bias" -> "2233 linear_104" [label="(512,)", style=solid]; +"199 features_5_13_norm1_weight" -> "2325 layer_norm_37" [label="(384,)", style=solid]; +"200 features_5_13_norm1_bias" -> "2325 layer_norm_37" [label="(384,)", style=solid]; +"201 features_5_13_mlp_0_bias" -> "2334 linear_108" [label="(1536,)", style=solid]; +"202 features_5_13_mlp_3_bias" -> "2344 linear_109" [label="(384,)", style=solid]; +"203 features_5_13_norm2_weight" -> "2346 layer_norm_38" [label="(384,)", style=solid]; +"204 features_5_13_norm2_bias" -> "2346 layer_norm_38" [label="(384,)", style=solid]; +"205 features_5_14_attn_cpb_mlp_0_bias" -> "2353 linear_110" [label="(512,)", style=solid]; +"206 features_5_14_norm1_weight" -> "2427 layer_norm_39" [label="(384,)", style=solid]; +"207 features_5_14_norm1_bias" -> "2427 layer_norm_39" [label="(384,)", style=solid]; +"208 features_5_14_mlp_0_bias" -> "2436 linear_114" [label="(1536,)", style=solid]; +"209 features_5_14_mlp_3_bias" -> "2446 linear_115" [label="(384,)", style=solid]; +"210 features_5_14_norm2_weight" -> "2448 layer_norm_40" [label="(384,)", style=solid]; +"211 features_5_14_norm2_bias" -> "2448 layer_norm_40" [label="(384,)", style=solid]; +"212 features_5_15_attn_cpb_mlp_0_bias" -> "2455 linear_116" [label="(512,)", style=solid]; +"213 features_5_15_norm1_weight" -> "2547 layer_norm_41" [label="(384,)", style=solid]; +"214 features_5_15_norm1_bias" -> "2547 layer_norm_41" [label="(384,)", style=solid]; +"215 features_5_15_mlp_0_bias" -> "2556 linear_120" [label="(1536,)", style=solid]; +"216 features_5_15_mlp_3_bias" -> "2566 linear_121" [label="(384,)", style=solid]; +"217 features_5_15_norm2_weight" -> "2568 layer_norm_42" [label="(384,)", style=solid]; +"218 features_5_15_norm2_bias" -> "2568 layer_norm_42" [label="(384,)", style=solid]; +"219 features_5_16_attn_cpb_mlp_0_bias" -> "2575 linear_122" [label="(512,)", style=solid]; +"220 features_5_16_norm1_weight" -> "2649 layer_norm_43" [label="(384,)", style=solid]; +"221 features_5_16_norm1_bias" -> "2649 layer_norm_43" [label="(384,)", style=solid]; +"222 features_5_16_mlp_0_bias" -> "2658 linear_126" [label="(1536,)", style=solid]; +"223 features_5_16_mlp_3_bias" -> "2668 linear_127" [label="(384,)", style=solid]; +"224 features_5_16_norm2_weight" -> "2670 layer_norm_44" [label="(384,)", style=solid]; +"225 features_5_16_norm2_bias" -> "2670 layer_norm_44" [label="(384,)", style=solid]; +"226 features_5_17_attn_cpb_mlp_0_bias" -> "2677 linear_128" [label="(512,)", style=solid]; +"227 features_5_17_norm1_weight" -> "2769 layer_norm_45" [label="(384,)", style=solid]; +"228 features_5_17_norm1_bias" -> "2769 layer_norm_45" [label="(384,)", style=solid]; +"229 features_5_17_mlp_0_bias" -> "2778 linear_132" [label="(1536,)", style=solid]; +"230 features_5_17_mlp_3_bias" -> "2788 linear_133" [label="(384,)", style=solid]; +"231 features_5_17_norm2_weight" -> "2790 layer_norm_46" [label="(384,)", style=solid]; +"232 features_5_17_norm2_bias" -> "2790 layer_norm_46" [label="(384,)", style=solid]; +"233 features_6_norm_weight" -> "2814 layer_norm_47" [label="(768,)", style=solid]; +"234 features_6_norm_bias" -> "2814 layer_norm_47" [label="(768,)", style=solid]; +"235 features_7_0_attn_cpb_mlp_0_bias" -> "2820 linear_135" [label="(512,)", style=solid]; +"236 features_7_0_norm1_weight" -> "2894 layer_norm_48" [label="(768,)", style=solid]; +"237 features_7_0_norm1_bias" -> "2894 layer_norm_48" [label="(768,)", style=solid]; +"238 features_7_0_mlp_0_bias" -> "2903 linear_139" [label="(3072,)", style=solid]; +"239 features_7_0_mlp_3_bias" -> "2913 linear_140" [label="(768,)", style=solid]; +"240 features_7_0_norm2_weight" -> "2915 layer_norm_49" [label="(768,)", style=solid]; +"241 features_7_0_norm2_bias" -> "2915 layer_norm_49" [label="(768,)", style=solid]; +"242 features_7_1_attn_cpb_mlp_0_bias" -> "2922 linear_141" [label="(512,)", style=solid]; +"243 features_7_1_norm1_weight" -> "2996 layer_norm_50" [label="(768,)", style=solid]; +"244 features_7_1_norm1_bias" -> "2996 layer_norm_50" [label="(768,)", style=solid]; +"245 features_7_1_mlp_0_bias" -> "3005 linear_145" [label="(3072,)", style=solid]; +"246 features_7_1_mlp_3_bias" -> "3015 linear_146" [label="(768,)", style=solid]; +"247 features_7_1_norm2_weight" -> "3017 layer_norm_51" [label="(768,)", style=solid]; +"248 features_7_1_norm2_bias" -> "3017 layer_norm_51" [label="(768,)", style=solid]; +"249 norm_weight" -> "3019 layer_norm_52" [label="(768,)", style=solid]; +"250 norm_bias" -> "3019 layer_norm_52" [label="(768,)", style=solid]; +"251 head_bias" -> "3030 linear_147" [label="(1000,)", style=solid]; +"252 features_1_0_attn_relative_coords_table" -> "310 features_1_0_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; +"253 features_1_0_attn_relative_position_index" -> "324 index" [label="(4096,)", style=solid]; +"254 features_1_1_attn_relative_coords_table" -> "409 features_1_1_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; +"255 features_1_1_attn_relative_position_index" -> "423 index_1" [label="(4096,)", style=solid]; +"256 features_3_0_attn_relative_coords_table" -> "549 features_3_0_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; +"257 features_3_0_attn_relative_position_index" -> "563 index_2" [label="(4096,)", style=solid]; +"258 features_3_1_attn_relative_coords_table" -> "651 features_3_1_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; +"259 features_3_1_attn_relative_position_index" -> "665 index_3" [label="(4096,)", style=solid]; +"260 features_5_0_attn_relative_coords_table" -> "794 features_5_0_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; +"261 features_5_0_attn_relative_position_index" -> "808 index_4" [label="(4096,)", style=solid]; +"262 features_5_1_attn_relative_coords_table" -> "896 features_5_1_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; +"263 features_5_1_attn_relative_position_index" -> "910 index_5" [label="(4096,)", style=solid]; +"264 features_5_2_attn_relative_coords_table" -> "1016 features_5_2_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; +"265 features_5_2_attn_relative_position_index" -> "1030 index_6" [label="(4096,)", style=solid]; +"266 features_5_3_attn_relative_coords_table" -> "1118 features_5_3_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; +"267 features_5_3_attn_relative_position_index" -> "1132 index_7" [label="(4096,)", style=solid]; +"268 features_5_4_attn_relative_coords_table" -> "1238 features_5_4_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; +"269 features_5_4_attn_relative_position_index" -> "1252 index_8" [label="(4096,)", style=solid]; +"270 features_5_5_attn_relative_coords_table" -> "1340 features_5_5_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; +"271 features_5_5_attn_relative_position_index" -> "1354 index_9" [label="(4096,)", style=solid]; +"272 features_5_6_attn_relative_coords_table" -> "1460 features_5_6_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; +"273 features_5_6_attn_relative_position_index" -> "1474 index_10" [label="(4096,)", style=solid]; +"274 features_5_7_attn_relative_coords_table" -> "1562 features_5_7_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; +"275 features_5_7_attn_relative_position_index" -> "1576 index_11" [label="(4096,)", style=solid]; +"276 features_5_8_attn_relative_coords_table" -> "1682 features_5_8_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; +"277 features_5_8_attn_relative_position_index" -> "1696 index_12" [label="(4096,)", style=solid]; +"278 features_5_9_attn_relative_coords_table" -> "1784 features_5_9_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; +"279 features_5_9_attn_relative_position_index" -> "1798 index_13" [label="(4096,)", style=solid]; +"280 features_5_10_attn_relative_coords_table" -> "1904 features_5_10_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; +"281 features_5_10_attn_relative_position_index" -> "1918 index_14" [label="(4096,)", style=solid]; +"282 features_5_11_attn_relative_coords_table" -> "2006 features_5_11_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; +"283 features_5_11_attn_relative_position_index" -> "2020 index_15" [label="(4096,)", style=solid]; +"284 features_5_12_attn_relative_coords_table" -> "2126 features_5_12_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; +"285 features_5_12_attn_relative_position_index" -> "2140 index_16" [label="(4096,)", style=solid]; +"286 features_5_13_attn_relative_coords_table" -> "2228 features_5_13_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; +"287 features_5_13_attn_relative_position_index" -> "2242 index_17" [label="(4096,)", style=solid]; +"288 features_5_14_attn_relative_coords_table" -> "2348 features_5_14_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; +"289 features_5_14_attn_relative_position_index" -> "2362 index_18" [label="(4096,)", style=solid]; +"290 features_5_15_attn_relative_coords_table" -> "2450 features_5_15_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; +"291 features_5_15_attn_relative_position_index" -> "2464 index_19" [label="(4096,)", style=solid]; +"292 features_5_16_attn_relative_coords_table" -> "2570 features_5_16_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; +"293 features_5_16_attn_relative_position_index" -> "2584 index_20" [label="(4096,)", style=solid]; +"294 features_5_17_attn_relative_coords_table" -> "2672 features_5_17_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; +"295 features_5_17_attn_relative_position_index" -> "2686 index_21" [label="(4096,)", style=solid]; +"296 features_7_0_attn_relative_coords_table" -> "2815 features_7_0_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; +"297 features_7_0_attn_relative_position_index" -> "2829 index_22" [label="(4096,)", style=solid]; +"298 features_7_1_attn_relative_coords_table" -> "2917 features_7_1_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; +"299 features_7_1_attn_relative_position_index" -> "2931 index_23" [label="(4096,)", style=solid]; +"300 x" -> "301 quantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; +"301 quantize_per_tensor_default" -> "302 dequantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; +"302 dequantize_per_tensor_default" -> "307 conv2d" [label="(1, 3, 224, 224)", style=solid]; +"303 conv2d_scale_0" -> "306 dequantize_per_channel_default" [label="(96,)", style=solid]; +"304 conv2d_zero_point_0" -> "306 dequantize_per_channel_default" [label="(96,)", style=solid]; +"305 _frozen_param0" -> "306 dequantize_per_channel_default" [label="(96, 3, 4, 4)", style=solid]; +"306 dequantize_per_channel_default" -> "307 conv2d" [label="(96, 3, 4, 4)", style=solid]; +"307 conv2d" -> "308 permute" [label="(1, 96, 56, 56)", style=solid]; +"308 permute" -> "309 layer_norm" [label="(1, 56, 56, 96)", style=solid]; +"309 layer_norm" -> "331 pad" [label="(1, 56, 56, 96)", style=solid]; +"309 layer_norm" -> "387 add_1" [label="(1, 56, 56, 96)", style=solid]; +"310 features_1_0_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "315 linear" [label="(1, 15, 15, 2)", style=solid]; +"311 linear_scale_0" -> "314 dequantize_per_channel_default_1" [label="(512,)", style=solid]; +"312 linear_zero_point_0" -> "314 dequantize_per_channel_default_1" [label="(512,)", style=solid]; +"313 _frozen_param1" -> "314 dequantize_per_channel_default_1" [label="(512, 2)", style=solid]; +"314 dequantize_per_channel_default_1" -> "315 linear" [label="(512, 2)", style=solid]; +"315 linear" -> "316 relu_" [label="(1, 15, 15, 512)", style=solid]; +"316 relu_" -> "317 relu__0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"317 relu__0_0_nncf_smooth_quant_0" -> "322 linear_1" [label="(1, 15, 15, 512)", style=solid]; +"318 linear_1_scale_0" -> "321 dequantize_per_channel_default_2" [label="(3,)", style=solid]; +"319 linear_1_zero_point_0" -> "321 dequantize_per_channel_default_2" [label="(3,)", style=solid]; +"320 _frozen_param2" -> "321 dequantize_per_channel_default_2" [label="(3, 512)", style=solid]; +"321 dequantize_per_channel_default_2" -> "322 linear_1" [label="(3, 512)", style=solid]; +"322 linear_1" -> "323 view" [label="(1, 15, 15, 3)", style=solid]; +"323 view" -> "324 index" [label="(225, 3)", style=solid]; +"324 index" -> "325 view_1" [label="(4096, 3)", style=solid]; +"325 view_1" -> "326 permute_1" [label="(64, 64, 3)", style=solid]; +"326 permute_1" -> "327 contiguous" [label="(3, 64, 64)", style=solid]; +"327 contiguous" -> "328 unsqueeze" [label="(3, 64, 64)", style=solid]; +"328 unsqueeze" -> "329 sigmoid" [label="(1, 3, 64, 64)", style=solid]; +"329 sigmoid" -> "330 mul" [label="(1, 3, 64, 64)", style=solid]; +"330 mul" -> "366 add" [label="(1, 3, 64, 64)", style=solid]; +"331 pad" -> "332 view_2" [label="(1, 56, 56, 96)", style=solid]; +"332 view_2" -> "333 permute_2" [label="(1, 7, 8, 7, 8, 96)", style=solid]; +"333 permute_2" -> "334 reshape" [label="(1, 7, 7, 8, 8, 96)", style=solid]; +"334 reshape" -> "336 reshape_0_0_nncf_smooth_quant_0" [label="(49, 64, 96)", style=solid]; +"335 clone" -> "343 linear_2" [label="(288,)", style=solid]; +"336 reshape_0_0_nncf_smooth_quant_0" -> "337 quantize_per_tensor_default_1" [label="(49, 64, 96)", style=solid]; +"337 quantize_per_tensor_default_1" -> "338 dequantize_per_tensor_default_1" [label="(49, 64, 96)", style=solid]; +"338 dequantize_per_tensor_default_1" -> "343 linear_2" [label="(49, 64, 96)", style=solid]; +"339 linear_2_scale_0" -> "342 dequantize_per_channel_default_3" [label="(288,)", style=solid]; +"340 linear_2_zero_point_0" -> "342 dequantize_per_channel_default_3" [label="(288,)", style=solid]; +"341 _frozen_param3" -> "342 dequantize_per_channel_default_3" [label="(288, 96)", style=solid]; +"342 dequantize_per_channel_default_3" -> "343 linear_2" [label="(288, 96)", style=solid]; +"343 linear_2" -> "344 reshape_1" [label="(49, 64, 288)", style=solid]; +"344 reshape_1" -> "345 permute_3" [label="(49, 64, 3, 3, 32)", style=solid]; +"345 permute_3" -> "346 select" [label="(3, 49, 3, 64, 32)", style=solid]; +"345 permute_3" -> "347 select_1" [label="(3, 49, 3, 64, 32)", style=solid]; +"345 permute_3" -> "348 select_2" [label="(3, 49, 3, 64, 32)", style=solid]; +"346 select" -> "349 linalg_vector_norm" [label="(49, 3, 64, 32)", style=solid]; +"346 select" -> "351 expand_as" [label="(49, 3, 64, 32)", style=solid]; +"346 select" -> "352 div" [label="(49, 3, 64, 32)", style=solid]; +"347 select_1" -> "355 linalg_vector_norm_1" [label="(49, 3, 64, 32)", style=solid]; +"347 select_1" -> "357 expand_as_1" [label="(49, 3, 64, 32)", style=solid]; +"347 select_1" -> "358 div_1" [label="(49, 3, 64, 32)", style=solid]; +"348 select_2" -> "369 matmul_1" [label="(49, 3, 64, 32)", style=solid]; +"349 linalg_vector_norm" -> "350 clamp_min" [label="(49, 3, 64, 1)", style=solid]; +"350 clamp_min" -> "351 expand_as" [label="(49, 3, 64, 1)", style=solid]; +"351 expand_as" -> "352 div" [label="(49, 3, 64, 32)", style=solid]; +"352 div" -> "353 quantize_per_tensor_default_2" [label="(49, 3, 64, 32)", style=solid]; +"353 quantize_per_tensor_default_2" -> "354 dequantize_per_tensor_default_2" [label="(49, 3, 64, 32)", style=solid]; +"354 dequantize_per_tensor_default_2" -> "362 matmul" [label="(49, 3, 64, 32)", style=solid]; +"355 linalg_vector_norm_1" -> "356 clamp_min_1" [label="(49, 3, 64, 1)", style=solid]; +"356 clamp_min_1" -> "357 expand_as_1" [label="(49, 3, 64, 1)", style=solid]; +"357 expand_as_1" -> "358 div_1" [label="(49, 3, 64, 32)", style=solid]; +"358 div_1" -> "359 quantize_per_tensor_default_3" [label="(49, 3, 64, 32)", style=solid]; +"359 quantize_per_tensor_default_3" -> "360 dequantize_per_tensor_default_3" [label="(49, 3, 64, 32)", style=solid]; +"360 dequantize_per_tensor_default_3" -> "361 transpose" [label="(49, 3, 64, 32)", style=solid]; +"361 transpose" -> "362 matmul" [label="(49, 3, 32, 64)", style=solid]; +"362 matmul" -> "365 mul_1" [label="(49, 3, 64, 64)", style=solid]; +"363 clamp" -> "364 exp" [label="(3, 1, 1)", style=solid]; +"364 exp" -> "365 mul_1" [label="(3, 1, 1)", style=solid]; +"365 mul_1" -> "366 add" [label="(49, 3, 64, 64)", style=solid]; +"366 add" -> "367 softmax" [label="(49, 3, 64, 64)", style=solid]; +"367 softmax" -> "368 dropout" [label="(49, 3, 64, 64)", style=solid]; +"368 dropout" -> "369 matmul_1" [label="(49, 3, 64, 64)", style=solid]; +"369 matmul_1" -> "370 transpose_1" [label="(49, 3, 64, 32)", style=solid]; +"370 transpose_1" -> "371 reshape_2" [label="(49, 64, 3, 32)", style=solid]; +"371 reshape_2" -> "372 reshape_2_0_0_nncf_smooth_quant_0" [label="(49, 64, 96)", style=solid]; +"372 reshape_2_0_0_nncf_smooth_quant_0" -> "373 quantize_per_tensor_default_4" [label="(49, 64, 96)", style=solid]; +"373 quantize_per_tensor_default_4" -> "374 dequantize_per_tensor_default_4" [label="(49, 64, 96)", style=solid]; +"374 dequantize_per_tensor_default_4" -> "379 linear_3" [label="(49, 64, 96)", style=solid]; +"375 linear_3_scale_0" -> "378 dequantize_per_channel_default_4" [label="(96,)", style=solid]; +"376 linear_3_zero_point_0" -> "378 dequantize_per_channel_default_4" [label="(96,)", style=solid]; +"377 _frozen_param4" -> "378 dequantize_per_channel_default_4" [label="(96, 96)", style=solid]; +"378 dequantize_per_channel_default_4" -> "379 linear_3" [label="(96, 96)", style=solid]; +"379 linear_3" -> "380 dropout_1" [label="(49, 64, 96)", style=solid]; +"380 dropout_1" -> "381 view_3" [label="(49, 64, 96)", style=solid]; +"381 view_3" -> "382 permute_4" [label="(1, 7, 7, 8, 8, 96)", style=solid]; +"382 permute_4" -> "383 reshape_3" [label="(1, 7, 8, 7, 8, 96)", style=solid]; +"383 reshape_3" -> "384 slice_2" [label="(1, 56, 56, 96)", style=solid]; +"384 slice_2" -> "385 slice_3" [label="(1, 56, 56, 96)", style=solid]; +"385 slice_3" -> "386 layer_norm_1" [label="(1, 56, 56, 96)", style=solid]; +"386 layer_norm_1" -> "387 add_1" [label="(1, 56, 56, 96)", style=solid]; +"387 add_1" -> "388 add_1_0_0_nncf_smooth_quant_0" [label="(1, 56, 56, 96)", style=solid]; +"387 add_1" -> "408 add_2" [label="(1, 56, 56, 96)", style=solid]; +"388 add_1_0_0_nncf_smooth_quant_0" -> "389 quantize_per_tensor_default_5" [label="(1, 56, 56, 96)", style=solid]; +"389 quantize_per_tensor_default_5" -> "390 dequantize_per_tensor_default_5" [label="(1, 56, 56, 96)", style=solid]; +"390 dequantize_per_tensor_default_5" -> "395 linear_4" [label="(1, 56, 56, 96)", style=solid]; +"391 linear_4_scale_0" -> "394 dequantize_per_channel_default_5" [label="(384,)", style=solid]; +"392 linear_4_zero_point_0" -> "394 dequantize_per_channel_default_5" [label="(384,)", style=solid]; +"393 _frozen_param5" -> "394 dequantize_per_channel_default_5" [label="(384, 96)", style=solid]; +"394 dequantize_per_channel_default_5" -> "395 linear_4" [label="(384, 96)", style=solid]; +"395 linear_4" -> "396 gelu" [label="(1, 56, 56, 384)", style=solid]; +"396 gelu" -> "397 dropout_2" [label="(1, 56, 56, 384)", style=solid]; +"397 dropout_2" -> "398 dropout_2_0_0_nncf_smooth_quant_0" [label="(1, 56, 56, 384)", style=solid]; +"398 dropout_2_0_0_nncf_smooth_quant_0" -> "399 quantize_per_tensor_default_6" [label="(1, 56, 56, 384)", style=solid]; +"399 quantize_per_tensor_default_6" -> "400 dequantize_per_tensor_default_6" [label="(1, 56, 56, 384)", style=solid]; +"400 dequantize_per_tensor_default_6" -> "405 linear_5" [label="(1, 56, 56, 384)", style=solid]; +"401 linear_5_scale_0" -> "404 dequantize_per_channel_default_6" [label="(96,)", style=solid]; +"402 linear_5_zero_point_0" -> "404 dequantize_per_channel_default_6" [label="(96,)", style=solid]; +"403 _frozen_param6" -> "404 dequantize_per_channel_default_6" [label="(96, 384)", style=solid]; +"404 dequantize_per_channel_default_6" -> "405 linear_5" [label="(96, 384)", style=solid]; +"405 linear_5" -> "406 dropout_3" [label="(1, 56, 56, 96)", style=solid]; +"406 dropout_3" -> "407 layer_norm_2" [label="(1, 56, 56, 96)", style=solid]; +"407 layer_norm_2" -> "408 add_2" [label="(1, 56, 56, 96)", style=solid]; +"408 add_2" -> "430 pad_1" [label="(1, 56, 56, 96)", style=solid]; +"408 add_2" -> "504 add_5" [label="(1, 56, 56, 96)", style=solid]; +"409 features_1_1_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "414 linear_6" [label="(1, 15, 15, 2)", style=solid]; +"410 linear_6_scale_0" -> "413 dequantize_per_channel_default_7" [label="(512,)", style=solid]; +"411 linear_6_zero_point_0" -> "413 dequantize_per_channel_default_7" [label="(512,)", style=solid]; +"412 _frozen_param7" -> "413 dequantize_per_channel_default_7" [label="(512, 2)", style=solid]; +"413 dequantize_per_channel_default_7" -> "414 linear_6" [label="(512, 2)", style=solid]; +"414 linear_6" -> "415 relu__1" [label="(1, 15, 15, 512)", style=solid]; +"415 relu__1" -> "416 relu__1_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"416 relu__1_0_0_nncf_smooth_quant_0" -> "421 linear_7" [label="(1, 15, 15, 512)", style=solid]; +"417 linear_7_scale_0" -> "420 dequantize_per_channel_default_8" [label="(3,)", style=solid]; +"418 linear_7_zero_point_0" -> "420 dequantize_per_channel_default_8" [label="(3,)", style=solid]; +"419 _frozen_param8" -> "420 dequantize_per_channel_default_8" [label="(3, 512)", style=solid]; +"420 dequantize_per_channel_default_8" -> "421 linear_7" [label="(3, 512)", style=solid]; +"421 linear_7" -> "422 view_4" [label="(1, 15, 15, 3)", style=solid]; +"422 view_4" -> "423 index_1" [label="(225, 3)", style=solid]; +"423 index_1" -> "424 view_5" [label="(4096, 3)", style=solid]; +"424 view_5" -> "425 permute_5" [label="(64, 64, 3)", style=solid]; +"425 permute_5" -> "426 contiguous_1" [label="(3, 64, 64)", style=solid]; +"426 contiguous_1" -> "427 unsqueeze_1" [label="(3, 64, 64)", style=solid]; +"427 unsqueeze_1" -> "428 sigmoid_1" [label="(1, 3, 64, 64)", style=solid]; +"428 sigmoid_1" -> "429 mul_2" [label="(1, 3, 64, 64)", style=solid]; +"429 mul_2" -> "466 add_3" [label="(1, 3, 64, 64)", style=solid]; +"430 pad_1" -> "431 roll" [label="(1, 56, 56, 96)", style=solid]; +"431 roll" -> "432 view_6" [label="(1, 56, 56, 96)", style=solid]; +"432 view_6" -> "433 permute_6" [label="(1, 7, 8, 7, 8, 96)", style=solid]; +"433 permute_6" -> "434 reshape_4" [label="(1, 7, 7, 8, 8, 96)", style=solid]; +"434 reshape_4" -> "436 reshape_4_0_0_nncf_smooth_quant_0" [label="(49, 64, 96)", style=solid]; +"434 reshape_4" -> "467 new_zeros" [label="(49, 64, 96)", style=solid]; +"435 clone_1" -> "443 linear_8" [label="(288,)", style=solid]; +"436 reshape_4_0_0_nncf_smooth_quant_0" -> "437 quantize_per_tensor_default_7" [label="(49, 64, 96)", style=solid]; +"437 quantize_per_tensor_default_7" -> "438 dequantize_per_tensor_default_7" [label="(49, 64, 96)", style=solid]; +"438 dequantize_per_tensor_default_7" -> "443 linear_8" [label="(49, 64, 96)", style=solid]; +"439 linear_8_scale_0" -> "442 dequantize_per_channel_default_9" [label="(288,)", style=solid]; +"440 linear_8_zero_point_0" -> "442 dequantize_per_channel_default_9" [label="(288,)", style=solid]; +"441 _frozen_param9" -> "442 dequantize_per_channel_default_9" [label="(288, 96)", style=solid]; +"442 dequantize_per_channel_default_9" -> "443 linear_8" [label="(288, 96)", style=solid]; +"443 linear_8" -> "444 reshape_5" [label="(49, 64, 288)", style=solid]; +"444 reshape_5" -> "445 permute_7" [label="(49, 64, 3, 3, 32)", style=solid]; +"445 permute_7" -> "446 select_3" [label="(3, 49, 3, 64, 32)", style=solid]; +"445 permute_7" -> "447 select_4" [label="(3, 49, 3, 64, 32)", style=solid]; +"445 permute_7" -> "448 select_5" [label="(3, 49, 3, 64, 32)", style=solid]; +"446 select_3" -> "449 linalg_vector_norm_2" [label="(49, 3, 64, 32)", style=solid]; +"446 select_3" -> "451 expand_as_2" [label="(49, 3, 64, 32)", style=solid]; +"446 select_3" -> "452 div_2" [label="(49, 3, 64, 32)", style=solid]; +"447 select_4" -> "455 linalg_vector_norm_3" [label="(49, 3, 64, 32)", style=solid]; +"447 select_4" -> "457 expand_as_3" [label="(49, 3, 64, 32)", style=solid]; +"447 select_4" -> "458 div_3" [label="(49, 3, 64, 32)", style=solid]; +"448 select_5" -> "485 matmul_3" [label="(49, 3, 64, 32)", style=solid]; +"449 linalg_vector_norm_2" -> "450 clamp_min_2" [label="(49, 3, 64, 1)", style=solid]; +"450 clamp_min_2" -> "451 expand_as_2" [label="(49, 3, 64, 1)", style=solid]; +"451 expand_as_2" -> "452 div_2" [label="(49, 3, 64, 32)", style=solid]; +"452 div_2" -> "453 quantize_per_tensor_default_8" [label="(49, 3, 64, 32)", style=solid]; +"453 quantize_per_tensor_default_8" -> "454 dequantize_per_tensor_default_8" [label="(49, 3, 64, 32)", style=solid]; +"454 dequantize_per_tensor_default_8" -> "462 matmul_2" [label="(49, 3, 64, 32)", style=solid]; +"455 linalg_vector_norm_3" -> "456 clamp_min_3" [label="(49, 3, 64, 1)", style=solid]; +"456 clamp_min_3" -> "457 expand_as_3" [label="(49, 3, 64, 1)", style=solid]; +"457 expand_as_3" -> "458 div_3" [label="(49, 3, 64, 32)", style=solid]; +"458 div_3" -> "459 quantize_per_tensor_default_9" [label="(49, 3, 64, 32)", style=solid]; +"459 quantize_per_tensor_default_9" -> "460 dequantize_per_tensor_default_9" [label="(49, 3, 64, 32)", style=solid]; +"460 dequantize_per_tensor_default_9" -> "461 transpose_2" [label="(49, 3, 64, 32)", style=solid]; +"461 transpose_2" -> "462 matmul_2" [label="(49, 3, 32, 64)", style=solid]; +"462 matmul_2" -> "465 mul_3" [label="(49, 3, 64, 64)", style=solid]; +"463 clamp_1" -> "464 exp_1" [label="(3, 1, 1)", style=solid]; +"464 exp_1" -> "465 mul_3" [label="(3, 1, 1)", style=solid]; +"465 mul_3" -> "466 add_3" [label="(49, 3, 64, 64)", style=solid]; +"466 add_3" -> "478 view_8" [label="(49, 3, 64, 64)", style=solid]; +"467 new_zeros" -> "468 view_7" [label="(56, 56)", style=solid]; +"468 view_7" -> "469 permute_8" [label="(7, 8, 7, 8)", style=solid]; +"469 permute_8" -> "470 reshape_6" [label="(7, 7, 8, 8)", style=solid]; +"470 reshape_6" -> "471 unsqueeze_2" [label="(49, 64)", style=solid]; +"470 reshape_6" -> "472 unsqueeze_3" [label="(49, 64)", style=solid]; +"471 unsqueeze_2" -> "473 sub" [label="(49, 1, 64)", style=solid]; +"472 unsqueeze_3" -> "473 sub" [label="(49, 64, 1)", style=solid]; +"473 sub" -> "474 ne" [label="(49, 64, 64)", style=solid]; +"473 sub" -> "475 masked_fill" [label="(49, 64, 64)", style=solid]; +"473 sub" -> "476 eq" [label="(49, 64, 64)", style=solid]; +"474 ne" -> "475 masked_fill" [label="(49, 64, 64)", style=solid]; +"475 masked_fill" -> "477 masked_fill_1" [label="(49, 64, 64)", style=solid]; +"476 eq" -> "477 masked_fill_1" [label="(49, 64, 64)", style=solid]; +"477 masked_fill_1" -> "479 unsqueeze_4" [label="(49, 64, 64)", style=solid]; +"478 view_8" -> "481 add_4" [label="(1, 49, 3, 64, 64)", style=solid]; +"479 unsqueeze_4" -> "480 unsqueeze_5" [label="(49, 1, 64, 64)", style=solid]; +"480 unsqueeze_5" -> "481 add_4" [label="(1, 49, 1, 64, 64)", style=solid]; +"481 add_4" -> "482 view_9" [label="(1, 49, 3, 64, 64)", style=solid]; +"482 view_9" -> "483 softmax_1" [label="(49, 3, 64, 64)", style=solid]; +"483 softmax_1" -> "484 dropout_4" [label="(49, 3, 64, 64)", style=solid]; +"484 dropout_4" -> "485 matmul_3" [label="(49, 3, 64, 64)", style=solid]; +"485 matmul_3" -> "486 transpose_3" [label="(49, 3, 64, 32)", style=solid]; +"486 transpose_3" -> "487 reshape_7" [label="(49, 64, 3, 32)", style=solid]; +"487 reshape_7" -> "488 reshape_7_0_0_nncf_smooth_quant_0" [label="(49, 64, 96)", style=solid]; +"488 reshape_7_0_0_nncf_smooth_quant_0" -> "489 quantize_per_tensor_default_10" [label="(49, 64, 96)", style=solid]; +"489 quantize_per_tensor_default_10" -> "490 dequantize_per_tensor_default_10" [label="(49, 64, 96)", style=solid]; +"490 dequantize_per_tensor_default_10" -> "495 linear_9" [label="(49, 64, 96)", style=solid]; +"491 linear_9_scale_0" -> "494 dequantize_per_channel_default_10" [label="(96,)", style=solid]; +"492 linear_9_zero_point_0" -> "494 dequantize_per_channel_default_10" [label="(96,)", style=solid]; +"493 _frozen_param10" -> "494 dequantize_per_channel_default_10" [label="(96, 96)", style=solid]; +"494 dequantize_per_channel_default_10" -> "495 linear_9" [label="(96, 96)", style=solid]; +"495 linear_9" -> "496 dropout_5" [label="(49, 64, 96)", style=solid]; +"496 dropout_5" -> "497 view_10" [label="(49, 64, 96)", style=solid]; +"497 view_10" -> "498 permute_9" [label="(1, 7, 7, 8, 8, 96)", style=solid]; +"498 permute_9" -> "499 reshape_8" [label="(1, 7, 8, 7, 8, 96)", style=solid]; +"499 reshape_8" -> "500 roll_1" [label="(1, 56, 56, 96)", style=solid]; +"500 roll_1" -> "501 slice_23" [label="(1, 56, 56, 96)", style=solid]; +"501 slice_23" -> "502 slice_24" [label="(1, 56, 56, 96)", style=solid]; +"502 slice_24" -> "503 layer_norm_3" [label="(1, 56, 56, 96)", style=solid]; +"503 layer_norm_3" -> "504 add_5" [label="(1, 56, 56, 96)", style=solid]; +"504 add_5" -> "505 add_5_0_0_nncf_smooth_quant_0" [label="(1, 56, 56, 96)", style=solid]; +"504 add_5" -> "525 add_6" [label="(1, 56, 56, 96)", style=solid]; +"505 add_5_0_0_nncf_smooth_quant_0" -> "506 quantize_per_tensor_default_11" [label="(1, 56, 56, 96)", style=solid]; +"506 quantize_per_tensor_default_11" -> "507 dequantize_per_tensor_default_11" [label="(1, 56, 56, 96)", style=solid]; +"507 dequantize_per_tensor_default_11" -> "512 linear_10" [label="(1, 56, 56, 96)", style=solid]; +"508 linear_10_scale_0" -> "511 dequantize_per_channel_default_11" [label="(384,)", style=solid]; +"509 linear_10_zero_point_0" -> "511 dequantize_per_channel_default_11" [label="(384,)", style=solid]; +"510 _frozen_param11" -> "511 dequantize_per_channel_default_11" [label="(384, 96)", style=solid]; +"511 dequantize_per_channel_default_11" -> "512 linear_10" [label="(384, 96)", style=solid]; +"512 linear_10" -> "513 gelu_1" [label="(1, 56, 56, 384)", style=solid]; +"513 gelu_1" -> "514 dropout_6" [label="(1, 56, 56, 384)", style=solid]; +"514 dropout_6" -> "515 dropout_6_0_0_nncf_smooth_quant_0" [label="(1, 56, 56, 384)", style=solid]; +"515 dropout_6_0_0_nncf_smooth_quant_0" -> "516 quantize_per_tensor_default_12" [label="(1, 56, 56, 384)", style=solid]; +"516 quantize_per_tensor_default_12" -> "517 dequantize_per_tensor_default_12" [label="(1, 56, 56, 384)", style=solid]; +"517 dequantize_per_tensor_default_12" -> "522 linear_11" [label="(1, 56, 56, 384)", style=solid]; +"518 linear_11_scale_0" -> "521 dequantize_per_channel_default_12" [label="(96,)", style=solid]; +"519 linear_11_zero_point_0" -> "521 dequantize_per_channel_default_12" [label="(96,)", style=solid]; +"520 _frozen_param12" -> "521 dequantize_per_channel_default_12" [label="(96, 384)", style=solid]; +"521 dequantize_per_channel_default_12" -> "522 linear_11" [label="(96, 384)", style=solid]; +"522 linear_11" -> "523 dropout_7" [label="(1, 56, 56, 96)", style=solid]; +"523 dropout_7" -> "524 layer_norm_4" [label="(1, 56, 56, 96)", style=solid]; +"524 layer_norm_4" -> "525 add_6" [label="(1, 56, 56, 96)", style=solid]; +"525 add_6" -> "526 pad_2" [label="(1, 56, 56, 96)", style=solid]; +"526 pad_2" -> "527 slice_25" [label="(1, 56, 56, 96)", style=solid]; +"526 pad_2" -> "530 slice_28" [label="(1, 56, 56, 96)", style=solid]; +"526 pad_2" -> "533 slice_31" [label="(1, 56, 56, 96)", style=solid]; +"526 pad_2" -> "536 slice_34" [label="(1, 56, 56, 96)", style=solid]; +"527 slice_25" -> "528 slice_26" [label="(1, 28, 56, 96)", style=solid]; +"528 slice_26" -> "529 slice_27" [label="(1, 28, 28, 96)", style=solid]; +"529 slice_27" -> "539 cat" [label="(1, 28, 28, 96)", style=solid]; +"530 slice_28" -> "531 slice_29" [label="(1, 28, 56, 96)", style=solid]; +"531 slice_29" -> "532 slice_30" [label="(1, 28, 28, 96)", style=solid]; +"532 slice_30" -> "539 cat" [label="(1, 28, 28, 96)", style=solid]; +"533 slice_31" -> "534 slice_32" [label="(1, 28, 56, 96)", style=solid]; +"534 slice_32" -> "535 slice_33" [label="(1, 28, 28, 96)", style=solid]; +"535 slice_33" -> "539 cat" [label="(1, 28, 28, 96)", style=solid]; +"536 slice_34" -> "537 slice_35" [label="(1, 28, 56, 96)", style=solid]; +"537 slice_35" -> "538 slice_36" [label="(1, 28, 28, 96)", style=solid]; +"538 slice_36" -> "539 cat" [label="(1, 28, 28, 96)", style=solid]; +"539 cat" -> "540 cat_0_0_nncf_smooth_quant_0" [label="(1, 28, 28, 384)", style=solid]; +"540 cat_0_0_nncf_smooth_quant_0" -> "541 quantize_per_tensor_default_13" [label="(1, 28, 28, 384)", style=solid]; +"541 quantize_per_tensor_default_13" -> "542 dequantize_per_tensor_default_13" [label="(1, 28, 28, 384)", style=solid]; +"542 dequantize_per_tensor_default_13" -> "547 linear_12" [label="(1, 28, 28, 384)", style=solid]; +"543 linear_12_scale_0" -> "546 dequantize_per_channel_default_13" [label="(192,)", style=solid]; +"544 linear_12_zero_point_0" -> "546 dequantize_per_channel_default_13" [label="(192,)", style=solid]; +"545 _frozen_param13" -> "546 dequantize_per_channel_default_13" [label="(192, 384)", style=solid]; +"546 dequantize_per_channel_default_13" -> "547 linear_12" [label="(192, 384)", style=solid]; +"547 linear_12" -> "548 layer_norm_5" [label="(1, 28, 28, 192)", style=solid]; +"548 layer_norm_5" -> "570 pad_3" [label="(1, 28, 28, 192)", style=solid]; +"548 layer_norm_5" -> "629 add_8" [label="(1, 28, 28, 192)", style=solid]; +"549 features_3_0_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "554 linear_13" [label="(1, 15, 15, 2)", style=solid]; +"550 linear_13_scale_0" -> "553 dequantize_per_channel_default_14" [label="(512,)", style=solid]; +"551 linear_13_zero_point_0" -> "553 dequantize_per_channel_default_14" [label="(512,)", style=solid]; +"552 _frozen_param14" -> "553 dequantize_per_channel_default_14" [label="(512, 2)", style=solid]; +"553 dequantize_per_channel_default_14" -> "554 linear_13" [label="(512, 2)", style=solid]; +"554 linear_13" -> "555 relu__2" [label="(1, 15, 15, 512)", style=solid]; +"555 relu__2" -> "556 relu__2_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"556 relu__2_0_0_nncf_smooth_quant_0" -> "561 linear_14" [label="(1, 15, 15, 512)", style=solid]; +"557 linear_14_scale_0" -> "560 dequantize_per_channel_default_15" [label="(6,)", style=solid]; +"558 linear_14_zero_point_0" -> "560 dequantize_per_channel_default_15" [label="(6,)", style=solid]; +"559 _frozen_param15" -> "560 dequantize_per_channel_default_15" [label="(6, 512)", style=solid]; +"560 dequantize_per_channel_default_15" -> "561 linear_14" [label="(6, 512)", style=solid]; +"561 linear_14" -> "562 view_11" [label="(1, 15, 15, 6)", style=solid]; +"562 view_11" -> "563 index_2" [label="(225, 6)", style=solid]; +"563 index_2" -> "564 view_12" [label="(4096, 6)", style=solid]; +"564 view_12" -> "565 permute_10" [label="(64, 64, 6)", style=solid]; +"565 permute_10" -> "566 contiguous_2" [label="(6, 64, 64)", style=solid]; +"566 contiguous_2" -> "567 unsqueeze_6" [label="(6, 64, 64)", style=solid]; +"567 unsqueeze_6" -> "568 sigmoid_2" [label="(1, 6, 64, 64)", style=solid]; +"568 sigmoid_2" -> "569 mul_4" [label="(1, 6, 64, 64)", style=solid]; +"569 mul_4" -> "605 add_7" [label="(1, 6, 64, 64)", style=solid]; +"570 pad_3" -> "571 view_13" [label="(1, 32, 32, 192)", style=solid]; +"571 view_13" -> "572 permute_11" [label="(1, 4, 8, 4, 8, 192)", style=solid]; +"572 permute_11" -> "573 reshape_9" [label="(1, 4, 4, 8, 8, 192)", style=solid]; +"573 reshape_9" -> "575 reshape_9_0_0_nncf_smooth_quant_0" [label="(16, 64, 192)", style=solid]; +"574 clone_2" -> "582 linear_15" [label="(576,)", style=solid]; +"575 reshape_9_0_0_nncf_smooth_quant_0" -> "576 quantize_per_tensor_default_14" [label="(16, 64, 192)", style=solid]; +"576 quantize_per_tensor_default_14" -> "577 dequantize_per_tensor_default_14" [label="(16, 64, 192)", style=solid]; +"577 dequantize_per_tensor_default_14" -> "582 linear_15" [label="(16, 64, 192)", style=solid]; +"578 linear_15_scale_0" -> "581 dequantize_per_channel_default_16" [label="(576,)", style=solid]; +"579 linear_15_zero_point_0" -> "581 dequantize_per_channel_default_16" [label="(576,)", style=solid]; +"580 _frozen_param16" -> "581 dequantize_per_channel_default_16" [label="(576, 192)", style=solid]; +"581 dequantize_per_channel_default_16" -> "582 linear_15" [label="(576, 192)", style=solid]; +"582 linear_15" -> "583 reshape_10" [label="(16, 64, 576)", style=solid]; +"583 reshape_10" -> "584 permute_12" [label="(16, 64, 3, 6, 32)", style=solid]; +"584 permute_12" -> "585 select_6" [label="(3, 16, 6, 64, 32)", style=solid]; +"584 permute_12" -> "586 select_7" [label="(3, 16, 6, 64, 32)", style=solid]; +"584 permute_12" -> "587 select_8" [label="(3, 16, 6, 64, 32)", style=solid]; +"585 select_6" -> "588 linalg_vector_norm_4" [label="(16, 6, 64, 32)", style=solid]; +"585 select_6" -> "590 expand_as_4" [label="(16, 6, 64, 32)", style=solid]; +"585 select_6" -> "591 div_4" [label="(16, 6, 64, 32)", style=solid]; +"586 select_7" -> "594 linalg_vector_norm_5" [label="(16, 6, 64, 32)", style=solid]; +"586 select_7" -> "596 expand_as_5" [label="(16, 6, 64, 32)", style=solid]; +"586 select_7" -> "597 div_5" [label="(16, 6, 64, 32)", style=solid]; +"587 select_8" -> "608 matmul_5" [label="(16, 6, 64, 32)", style=solid]; +"588 linalg_vector_norm_4" -> "589 clamp_min_4" [label="(16, 6, 64, 1)", style=solid]; +"589 clamp_min_4" -> "590 expand_as_4" [label="(16, 6, 64, 1)", style=solid]; +"590 expand_as_4" -> "591 div_4" [label="(16, 6, 64, 32)", style=solid]; +"591 div_4" -> "592 quantize_per_tensor_default_15" [label="(16, 6, 64, 32)", style=solid]; +"592 quantize_per_tensor_default_15" -> "593 dequantize_per_tensor_default_15" [label="(16, 6, 64, 32)", style=solid]; +"593 dequantize_per_tensor_default_15" -> "601 matmul_4" [label="(16, 6, 64, 32)", style=solid]; +"594 linalg_vector_norm_5" -> "595 clamp_min_5" [label="(16, 6, 64, 1)", style=solid]; +"595 clamp_min_5" -> "596 expand_as_5" [label="(16, 6, 64, 1)", style=solid]; +"596 expand_as_5" -> "597 div_5" [label="(16, 6, 64, 32)", style=solid]; +"597 div_5" -> "598 quantize_per_tensor_default_16" [label="(16, 6, 64, 32)", style=solid]; +"598 quantize_per_tensor_default_16" -> "599 dequantize_per_tensor_default_16" [label="(16, 6, 64, 32)", style=solid]; +"599 dequantize_per_tensor_default_16" -> "600 transpose_4" [label="(16, 6, 64, 32)", style=solid]; +"600 transpose_4" -> "601 matmul_4" [label="(16, 6, 32, 64)", style=solid]; +"601 matmul_4" -> "604 mul_5" [label="(16, 6, 64, 64)", style=solid]; +"602 clamp_2" -> "603 exp_2" [label="(6, 1, 1)", style=solid]; +"603 exp_2" -> "604 mul_5" [label="(6, 1, 1)", style=solid]; +"604 mul_5" -> "605 add_7" [label="(16, 6, 64, 64)", style=solid]; +"605 add_7" -> "606 softmax_2" [label="(16, 6, 64, 64)", style=solid]; +"606 softmax_2" -> "607 dropout_8" [label="(16, 6, 64, 64)", style=solid]; +"607 dropout_8" -> "608 matmul_5" [label="(16, 6, 64, 64)", style=solid]; +"608 matmul_5" -> "609 transpose_5" [label="(16, 6, 64, 32)", style=solid]; +"609 transpose_5" -> "610 reshape_11" [label="(16, 64, 6, 32)", style=solid]; +"610 reshape_11" -> "611 reshape_11_0_0_nncf_smooth_quant_0" [label="(16, 64, 192)", style=solid]; +"611 reshape_11_0_0_nncf_smooth_quant_0" -> "612 quantize_per_tensor_default_17" [label="(16, 64, 192)", style=solid]; +"612 quantize_per_tensor_default_17" -> "613 dequantize_per_tensor_default_17" [label="(16, 64, 192)", style=solid]; +"613 dequantize_per_tensor_default_17" -> "618 linear_16" [label="(16, 64, 192)", style=solid]; +"614 linear_16_scale_0" -> "617 dequantize_per_channel_default_17" [label="(192,)", style=solid]; +"615 linear_16_zero_point_0" -> "617 dequantize_per_channel_default_17" [label="(192,)", style=solid]; +"616 _frozen_param17" -> "617 dequantize_per_channel_default_17" [label="(192, 192)", style=solid]; +"617 dequantize_per_channel_default_17" -> "618 linear_16" [label="(192, 192)", style=solid]; +"618 linear_16" -> "619 dropout_9" [label="(16, 64, 192)", style=solid]; +"619 dropout_9" -> "620 view_14" [label="(16, 64, 192)", style=solid]; +"620 view_14" -> "621 permute_13" [label="(1, 4, 4, 8, 8, 192)", style=solid]; +"621 permute_13" -> "622 reshape_12" [label="(1, 4, 8, 4, 8, 192)", style=solid]; +"622 reshape_12" -> "623 slice_38" [label="(1, 32, 32, 192)", style=solid]; +"623 slice_38" -> "624 slice_39" [label="(1, 32, 32, 192)", style=solid]; +"624 slice_39" -> "625 slice_40" [label="(1, 28, 32, 192)", style=solid]; +"625 slice_40" -> "626 slice_41" [label="(1, 28, 28, 192)", style=solid]; +"626 slice_41" -> "627 contiguous_3" [label="(1, 28, 28, 192)", style=solid]; +"627 contiguous_3" -> "628 layer_norm_6" [label="(1, 28, 28, 192)", style=solid]; +"628 layer_norm_6" -> "629 add_8" [label="(1, 28, 28, 192)", style=solid]; +"629 add_8" -> "630 add_8_0_0_nncf_smooth_quant_0" [label="(1, 28, 28, 192)", style=solid]; +"629 add_8" -> "650 add_9" [label="(1, 28, 28, 192)", style=solid]; +"630 add_8_0_0_nncf_smooth_quant_0" -> "631 quantize_per_tensor_default_18" [label="(1, 28, 28, 192)", style=solid]; +"631 quantize_per_tensor_default_18" -> "632 dequantize_per_tensor_default_18" [label="(1, 28, 28, 192)", style=solid]; +"632 dequantize_per_tensor_default_18" -> "637 linear_17" [label="(1, 28, 28, 192)", style=solid]; +"633 linear_17_scale_0" -> "636 dequantize_per_channel_default_18" [label="(768,)", style=solid]; +"634 linear_17_zero_point_0" -> "636 dequantize_per_channel_default_18" [label="(768,)", style=solid]; +"635 _frozen_param18" -> "636 dequantize_per_channel_default_18" [label="(768, 192)", style=solid]; +"636 dequantize_per_channel_default_18" -> "637 linear_17" [label="(768, 192)", style=solid]; +"637 linear_17" -> "638 gelu_2" [label="(1, 28, 28, 768)", style=solid]; +"638 gelu_2" -> "639 dropout_10" [label="(1, 28, 28, 768)", style=solid]; +"639 dropout_10" -> "640 dropout_10_0_0_nncf_smooth_quant_0" [label="(1, 28, 28, 768)", style=solid]; +"640 dropout_10_0_0_nncf_smooth_quant_0" -> "641 quantize_per_tensor_default_19" [label="(1, 28, 28, 768)", style=solid]; +"641 quantize_per_tensor_default_19" -> "642 dequantize_per_tensor_default_19" [label="(1, 28, 28, 768)", style=solid]; +"642 dequantize_per_tensor_default_19" -> "647 linear_18" [label="(1, 28, 28, 768)", style=solid]; +"643 linear_18_scale_0" -> "646 dequantize_per_channel_default_19" [label="(192,)", style=solid]; +"644 linear_18_zero_point_0" -> "646 dequantize_per_channel_default_19" [label="(192,)", style=solid]; +"645 _frozen_param19" -> "646 dequantize_per_channel_default_19" [label="(192, 768)", style=solid]; +"646 dequantize_per_channel_default_19" -> "647 linear_18" [label="(192, 768)", style=solid]; +"647 linear_18" -> "648 dropout_11" [label="(1, 28, 28, 192)", style=solid]; +"648 dropout_11" -> "649 layer_norm_7" [label="(1, 28, 28, 192)", style=solid]; +"649 layer_norm_7" -> "650 add_9" [label="(1, 28, 28, 192)", style=solid]; +"650 add_9" -> "672 pad_4" [label="(1, 28, 28, 192)", style=solid]; +"650 add_9" -> "749 add_12" [label="(1, 28, 28, 192)", style=solid]; +"651 features_3_1_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "656 linear_19" [label="(1, 15, 15, 2)", style=solid]; +"652 linear_19_scale_0" -> "655 dequantize_per_channel_default_20" [label="(512,)", style=solid]; +"653 linear_19_zero_point_0" -> "655 dequantize_per_channel_default_20" [label="(512,)", style=solid]; +"654 _frozen_param20" -> "655 dequantize_per_channel_default_20" [label="(512, 2)", style=solid]; +"655 dequantize_per_channel_default_20" -> "656 linear_19" [label="(512, 2)", style=solid]; +"656 linear_19" -> "657 relu__3" [label="(1, 15, 15, 512)", style=solid]; +"657 relu__3" -> "658 relu__3_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"658 relu__3_0_0_nncf_smooth_quant_0" -> "663 linear_20" [label="(1, 15, 15, 512)", style=solid]; +"659 linear_20_scale_0" -> "662 dequantize_per_channel_default_21" [label="(6,)", style=solid]; +"660 linear_20_zero_point_0" -> "662 dequantize_per_channel_default_21" [label="(6,)", style=solid]; +"661 _frozen_param21" -> "662 dequantize_per_channel_default_21" [label="(6, 512)", style=solid]; +"662 dequantize_per_channel_default_21" -> "663 linear_20" [label="(6, 512)", style=solid]; +"663 linear_20" -> "664 view_15" [label="(1, 15, 15, 6)", style=solid]; +"664 view_15" -> "665 index_3" [label="(225, 6)", style=solid]; +"665 index_3" -> "666 view_16" [label="(4096, 6)", style=solid]; +"666 view_16" -> "667 permute_14" [label="(64, 64, 6)", style=solid]; +"667 permute_14" -> "668 contiguous_4" [label="(6, 64, 64)", style=solid]; +"668 contiguous_4" -> "669 unsqueeze_7" [label="(6, 64, 64)", style=solid]; +"669 unsqueeze_7" -> "670 sigmoid_3" [label="(1, 6, 64, 64)", style=solid]; +"670 sigmoid_3" -> "671 mul_6" [label="(1, 6, 64, 64)", style=solid]; +"671 mul_6" -> "708 add_10" [label="(1, 6, 64, 64)", style=solid]; +"672 pad_4" -> "673 roll_2" [label="(1, 32, 32, 192)", style=solid]; +"673 roll_2" -> "674 view_17" [label="(1, 32, 32, 192)", style=solid]; +"674 view_17" -> "675 permute_15" [label="(1, 4, 8, 4, 8, 192)", style=solid]; +"675 permute_15" -> "676 reshape_13" [label="(1, 4, 4, 8, 8, 192)", style=solid]; +"676 reshape_13" -> "678 reshape_13_0_0_nncf_smooth_quant_0" [label="(16, 64, 192)", style=solid]; +"676 reshape_13" -> "709 new_zeros_1" [label="(16, 64, 192)", style=solid]; +"677 clone_3" -> "685 linear_21" [label="(576,)", style=solid]; +"678 reshape_13_0_0_nncf_smooth_quant_0" -> "679 quantize_per_tensor_default_20" [label="(16, 64, 192)", style=solid]; +"679 quantize_per_tensor_default_20" -> "680 dequantize_per_tensor_default_20" [label="(16, 64, 192)", style=solid]; +"680 dequantize_per_tensor_default_20" -> "685 linear_21" [label="(16, 64, 192)", style=solid]; +"681 linear_21_scale_0" -> "684 dequantize_per_channel_default_22" [label="(576,)", style=solid]; +"682 linear_21_zero_point_0" -> "684 dequantize_per_channel_default_22" [label="(576,)", style=solid]; +"683 _frozen_param22" -> "684 dequantize_per_channel_default_22" [label="(576, 192)", style=solid]; +"684 dequantize_per_channel_default_22" -> "685 linear_21" [label="(576, 192)", style=solid]; +"685 linear_21" -> "686 reshape_14" [label="(16, 64, 576)", style=solid]; +"686 reshape_14" -> "687 permute_16" [label="(16, 64, 3, 6, 32)", style=solid]; +"687 permute_16" -> "688 select_9" [label="(3, 16, 6, 64, 32)", style=solid]; +"687 permute_16" -> "689 select_10" [label="(3, 16, 6, 64, 32)", style=solid]; +"687 permute_16" -> "690 select_11" [label="(3, 16, 6, 64, 32)", style=solid]; +"688 select_9" -> "691 linalg_vector_norm_6" [label="(16, 6, 64, 32)", style=solid]; +"688 select_9" -> "693 expand_as_6" [label="(16, 6, 64, 32)", style=solid]; +"688 select_9" -> "694 div_6" [label="(16, 6, 64, 32)", style=solid]; +"689 select_10" -> "697 linalg_vector_norm_7" [label="(16, 6, 64, 32)", style=solid]; +"689 select_10" -> "699 expand_as_7" [label="(16, 6, 64, 32)", style=solid]; +"689 select_10" -> "700 div_7" [label="(16, 6, 64, 32)", style=solid]; +"690 select_11" -> "727 matmul_7" [label="(16, 6, 64, 32)", style=solid]; +"691 linalg_vector_norm_6" -> "692 clamp_min_6" [label="(16, 6, 64, 1)", style=solid]; +"692 clamp_min_6" -> "693 expand_as_6" [label="(16, 6, 64, 1)", style=solid]; +"693 expand_as_6" -> "694 div_6" [label="(16, 6, 64, 32)", style=solid]; +"694 div_6" -> "695 quantize_per_tensor_default_21" [label="(16, 6, 64, 32)", style=solid]; +"695 quantize_per_tensor_default_21" -> "696 dequantize_per_tensor_default_21" [label="(16, 6, 64, 32)", style=solid]; +"696 dequantize_per_tensor_default_21" -> "704 matmul_6" [label="(16, 6, 64, 32)", style=solid]; +"697 linalg_vector_norm_7" -> "698 clamp_min_7" [label="(16, 6, 64, 1)", style=solid]; +"698 clamp_min_7" -> "699 expand_as_7" [label="(16, 6, 64, 1)", style=solid]; +"699 expand_as_7" -> "700 div_7" [label="(16, 6, 64, 32)", style=solid]; +"700 div_7" -> "701 quantize_per_tensor_default_22" [label="(16, 6, 64, 32)", style=solid]; +"701 quantize_per_tensor_default_22" -> "702 dequantize_per_tensor_default_22" [label="(16, 6, 64, 32)", style=solid]; +"702 dequantize_per_tensor_default_22" -> "703 transpose_6" [label="(16, 6, 64, 32)", style=solid]; +"703 transpose_6" -> "704 matmul_6" [label="(16, 6, 32, 64)", style=solid]; +"704 matmul_6" -> "707 mul_7" [label="(16, 6, 64, 64)", style=solid]; +"705 clamp_3" -> "706 exp_3" [label="(6, 1, 1)", style=solid]; +"706 exp_3" -> "707 mul_7" [label="(6, 1, 1)", style=solid]; +"707 mul_7" -> "708 add_10" [label="(16, 6, 64, 64)", style=solid]; +"708 add_10" -> "720 view_19" [label="(16, 6, 64, 64)", style=solid]; +"709 new_zeros_1" -> "710 view_18" [label="(32, 32)", style=solid]; +"710 view_18" -> "711 permute_17" [label="(4, 8, 4, 8)", style=solid]; +"711 permute_17" -> "712 reshape_15" [label="(4, 4, 8, 8)", style=solid]; +"712 reshape_15" -> "713 unsqueeze_8" [label="(16, 64)", style=solid]; +"712 reshape_15" -> "714 unsqueeze_9" [label="(16, 64)", style=solid]; +"713 unsqueeze_8" -> "715 sub_1" [label="(16, 1, 64)", style=solid]; +"714 unsqueeze_9" -> "715 sub_1" [label="(16, 64, 1)", style=solid]; +"715 sub_1" -> "716 ne_1" [label="(16, 64, 64)", style=solid]; +"715 sub_1" -> "717 masked_fill_2" [label="(16, 64, 64)", style=solid]; +"715 sub_1" -> "718 eq_1" [label="(16, 64, 64)", style=solid]; +"716 ne_1" -> "717 masked_fill_2" [label="(16, 64, 64)", style=solid]; +"717 masked_fill_2" -> "719 masked_fill_3" [label="(16, 64, 64)", style=solid]; +"718 eq_1" -> "719 masked_fill_3" [label="(16, 64, 64)", style=solid]; +"719 masked_fill_3" -> "721 unsqueeze_10" [label="(16, 64, 64)", style=solid]; +"720 view_19" -> "723 add_11" [label="(1, 16, 6, 64, 64)", style=solid]; +"721 unsqueeze_10" -> "722 unsqueeze_11" [label="(16, 1, 64, 64)", style=solid]; +"722 unsqueeze_11" -> "723 add_11" [label="(1, 16, 1, 64, 64)", style=solid]; +"723 add_11" -> "724 view_20" [label="(1, 16, 6, 64, 64)", style=solid]; +"724 view_20" -> "725 softmax_3" [label="(16, 6, 64, 64)", style=solid]; +"725 softmax_3" -> "726 dropout_12" [label="(16, 6, 64, 64)", style=solid]; +"726 dropout_12" -> "727 matmul_7" [label="(16, 6, 64, 64)", style=solid]; +"727 matmul_7" -> "728 transpose_7" [label="(16, 6, 64, 32)", style=solid]; +"728 transpose_7" -> "729 reshape_16" [label="(16, 64, 6, 32)", style=solid]; +"729 reshape_16" -> "730 reshape_16_0_0_nncf_smooth_quant_0" [label="(16, 64, 192)", style=solid]; +"730 reshape_16_0_0_nncf_smooth_quant_0" -> "731 quantize_per_tensor_default_23" [label="(16, 64, 192)", style=solid]; +"731 quantize_per_tensor_default_23" -> "732 dequantize_per_tensor_default_23" [label="(16, 64, 192)", style=solid]; +"732 dequantize_per_tensor_default_23" -> "737 linear_22" [label="(16, 64, 192)", style=solid]; +"733 linear_22_scale_0" -> "736 dequantize_per_channel_default_23" [label="(192,)", style=solid]; +"734 linear_22_zero_point_0" -> "736 dequantize_per_channel_default_23" [label="(192,)", style=solid]; +"735 _frozen_param23" -> "736 dequantize_per_channel_default_23" [label="(192, 192)", style=solid]; +"736 dequantize_per_channel_default_23" -> "737 linear_22" [label="(192, 192)", style=solid]; +"737 linear_22" -> "738 dropout_13" [label="(16, 64, 192)", style=solid]; +"738 dropout_13" -> "739 view_21" [label="(16, 64, 192)", style=solid]; +"739 view_21" -> "740 permute_18" [label="(1, 4, 4, 8, 8, 192)", style=solid]; +"740 permute_18" -> "741 reshape_17" [label="(1, 4, 8, 4, 8, 192)", style=solid]; +"741 reshape_17" -> "742 roll_3" [label="(1, 32, 32, 192)", style=solid]; +"742 roll_3" -> "743 slice_61" [label="(1, 32, 32, 192)", style=solid]; +"743 slice_61" -> "744 slice_62" [label="(1, 32, 32, 192)", style=solid]; +"744 slice_62" -> "745 slice_63" [label="(1, 28, 32, 192)", style=solid]; +"745 slice_63" -> "746 slice_64" [label="(1, 28, 28, 192)", style=solid]; +"746 slice_64" -> "747 contiguous_5" [label="(1, 28, 28, 192)", style=solid]; +"747 contiguous_5" -> "748 layer_norm_8" [label="(1, 28, 28, 192)", style=solid]; +"748 layer_norm_8" -> "749 add_12" [label="(1, 28, 28, 192)", style=solid]; +"749 add_12" -> "750 add_12_0_0_nncf_smooth_quant_0" [label="(1, 28, 28, 192)", style=solid]; +"749 add_12" -> "770 add_13" [label="(1, 28, 28, 192)", style=solid]; +"750 add_12_0_0_nncf_smooth_quant_0" -> "751 quantize_per_tensor_default_24" [label="(1, 28, 28, 192)", style=solid]; +"751 quantize_per_tensor_default_24" -> "752 dequantize_per_tensor_default_24" [label="(1, 28, 28, 192)", style=solid]; +"752 dequantize_per_tensor_default_24" -> "757 linear_23" [label="(1, 28, 28, 192)", style=solid]; +"753 linear_23_scale_0" -> "756 dequantize_per_channel_default_24" [label="(768,)", style=solid]; +"754 linear_23_zero_point_0" -> "756 dequantize_per_channel_default_24" [label="(768,)", style=solid]; +"755 _frozen_param24" -> "756 dequantize_per_channel_default_24" [label="(768, 192)", style=solid]; +"756 dequantize_per_channel_default_24" -> "757 linear_23" [label="(768, 192)", style=solid]; +"757 linear_23" -> "758 gelu_3" [label="(1, 28, 28, 768)", style=solid]; +"758 gelu_3" -> "759 dropout_14" [label="(1, 28, 28, 768)", style=solid]; +"759 dropout_14" -> "760 dropout_14_0_0_nncf_smooth_quant_0" [label="(1, 28, 28, 768)", style=solid]; +"760 dropout_14_0_0_nncf_smooth_quant_0" -> "761 quantize_per_tensor_default_25" [label="(1, 28, 28, 768)", style=solid]; +"761 quantize_per_tensor_default_25" -> "762 dequantize_per_tensor_default_25" [label="(1, 28, 28, 768)", style=solid]; +"762 dequantize_per_tensor_default_25" -> "767 linear_24" [label="(1, 28, 28, 768)", style=solid]; +"763 linear_24_scale_0" -> "766 dequantize_per_channel_default_25" [label="(192,)", style=solid]; +"764 linear_24_zero_point_0" -> "766 dequantize_per_channel_default_25" [label="(192,)", style=solid]; +"765 _frozen_param25" -> "766 dequantize_per_channel_default_25" [label="(192, 768)", style=solid]; +"766 dequantize_per_channel_default_25" -> "767 linear_24" [label="(192, 768)", style=solid]; +"767 linear_24" -> "768 dropout_15" [label="(1, 28, 28, 192)", style=solid]; +"768 dropout_15" -> "769 layer_norm_9" [label="(1, 28, 28, 192)", style=solid]; +"769 layer_norm_9" -> "770 add_13" [label="(1, 28, 28, 192)", style=solid]; +"770 add_13" -> "771 pad_5" [label="(1, 28, 28, 192)", style=solid]; +"771 pad_5" -> "772 slice_65" [label="(1, 28, 28, 192)", style=solid]; +"771 pad_5" -> "775 slice_68" [label="(1, 28, 28, 192)", style=solid]; +"771 pad_5" -> "778 slice_71" [label="(1, 28, 28, 192)", style=solid]; +"771 pad_5" -> "781 slice_74" [label="(1, 28, 28, 192)", style=solid]; +"772 slice_65" -> "773 slice_66" [label="(1, 14, 28, 192)", style=solid]; +"773 slice_66" -> "774 slice_67" [label="(1, 14, 14, 192)", style=solid]; +"774 slice_67" -> "784 cat_1" [label="(1, 14, 14, 192)", style=solid]; +"775 slice_68" -> "776 slice_69" [label="(1, 14, 28, 192)", style=solid]; +"776 slice_69" -> "777 slice_70" [label="(1, 14, 14, 192)", style=solid]; +"777 slice_70" -> "784 cat_1" [label="(1, 14, 14, 192)", style=solid]; +"778 slice_71" -> "779 slice_72" [label="(1, 14, 28, 192)", style=solid]; +"779 slice_72" -> "780 slice_73" [label="(1, 14, 14, 192)", style=solid]; +"780 slice_73" -> "784 cat_1" [label="(1, 14, 14, 192)", style=solid]; +"781 slice_74" -> "782 slice_75" [label="(1, 14, 28, 192)", style=solid]; +"782 slice_75" -> "783 slice_76" [label="(1, 14, 14, 192)", style=solid]; +"783 slice_76" -> "784 cat_1" [label="(1, 14, 14, 192)", style=solid]; +"784 cat_1" -> "785 cat_1_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 768)", style=solid]; +"785 cat_1_0_0_nncf_smooth_quant_0" -> "786 quantize_per_tensor_default_26" [label="(1, 14, 14, 768)", style=solid]; +"786 quantize_per_tensor_default_26" -> "787 dequantize_per_tensor_default_26" [label="(1, 14, 14, 768)", style=solid]; +"787 dequantize_per_tensor_default_26" -> "792 linear_25" [label="(1, 14, 14, 768)", style=solid]; +"788 linear_25_scale_0" -> "791 dequantize_per_channel_default_26" [label="(384,)", style=solid]; +"789 linear_25_zero_point_0" -> "791 dequantize_per_channel_default_26" [label="(384,)", style=solid]; +"790 _frozen_param26" -> "791 dequantize_per_channel_default_26" [label="(384, 768)", style=solid]; +"791 dequantize_per_channel_default_26" -> "792 linear_25" [label="(384, 768)", style=solid]; +"792 linear_25" -> "793 layer_norm_10" [label="(1, 14, 14, 384)", style=solid]; +"793 layer_norm_10" -> "815 pad_6" [label="(1, 14, 14, 384)", style=solid]; +"793 layer_norm_10" -> "874 add_15" [label="(1, 14, 14, 384)", style=solid]; +"794 features_5_0_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "799 linear_26" [label="(1, 15, 15, 2)", style=solid]; +"795 linear_26_scale_0" -> "798 dequantize_per_channel_default_27" [label="(512,)", style=solid]; +"796 linear_26_zero_point_0" -> "798 dequantize_per_channel_default_27" [label="(512,)", style=solid]; +"797 _frozen_param27" -> "798 dequantize_per_channel_default_27" [label="(512, 2)", style=solid]; +"798 dequantize_per_channel_default_27" -> "799 linear_26" [label="(512, 2)", style=solid]; +"799 linear_26" -> "800 relu__4" [label="(1, 15, 15, 512)", style=solid]; +"800 relu__4" -> "801 relu__4_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"801 relu__4_0_0_nncf_smooth_quant_0" -> "806 linear_27" [label="(1, 15, 15, 512)", style=solid]; +"802 linear_27_scale_0" -> "805 dequantize_per_channel_default_28" [label="(12,)", style=solid]; +"803 linear_27_zero_point_0" -> "805 dequantize_per_channel_default_28" [label="(12,)", style=solid]; +"804 _frozen_param28" -> "805 dequantize_per_channel_default_28" [label="(12, 512)", style=solid]; +"805 dequantize_per_channel_default_28" -> "806 linear_27" [label="(12, 512)", style=solid]; +"806 linear_27" -> "807 view_22" [label="(1, 15, 15, 12)", style=solid]; +"807 view_22" -> "808 index_4" [label="(225, 12)", style=solid]; +"808 index_4" -> "809 view_23" [label="(4096, 12)", style=solid]; +"809 view_23" -> "810 permute_19" [label="(64, 64, 12)", style=solid]; +"810 permute_19" -> "811 contiguous_6" [label="(12, 64, 64)", style=solid]; +"811 contiguous_6" -> "812 unsqueeze_12" [label="(12, 64, 64)", style=solid]; +"812 unsqueeze_12" -> "813 sigmoid_4" [label="(1, 12, 64, 64)", style=solid]; +"813 sigmoid_4" -> "814 mul_8" [label="(1, 12, 64, 64)", style=solid]; +"814 mul_8" -> "850 add_14" [label="(1, 12, 64, 64)", style=solid]; +"815 pad_6" -> "816 view_24" [label="(1, 16, 16, 384)", style=solid]; +"816 view_24" -> "817 permute_20" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"817 permute_20" -> "818 reshape_18" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"818 reshape_18" -> "820 reshape_18_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"819 clone_4" -> "827 linear_28" [label="(1152,)", style=solid]; +"820 reshape_18_0_0_nncf_smooth_quant_0" -> "821 quantize_per_tensor_default_27" [label="(4, 64, 384)", style=solid]; +"821 quantize_per_tensor_default_27" -> "822 dequantize_per_tensor_default_27" [label="(4, 64, 384)", style=solid]; +"822 dequantize_per_tensor_default_27" -> "827 linear_28" [label="(4, 64, 384)", style=solid]; +"823 linear_28_scale_0" -> "826 dequantize_per_channel_default_29" [label="(1152,)", style=solid]; +"824 linear_28_zero_point_0" -> "826 dequantize_per_channel_default_29" [label="(1152,)", style=solid]; +"825 _frozen_param29" -> "826 dequantize_per_channel_default_29" [label="(1152, 384)", style=solid]; +"826 dequantize_per_channel_default_29" -> "827 linear_28" [label="(1152, 384)", style=solid]; +"827 linear_28" -> "828 reshape_19" [label="(4, 64, 1152)", style=solid]; +"828 reshape_19" -> "829 permute_21" [label="(4, 64, 3, 12, 32)", style=solid]; +"829 permute_21" -> "830 select_12" [label="(3, 4, 12, 64, 32)", style=solid]; +"829 permute_21" -> "831 select_13" [label="(3, 4, 12, 64, 32)", style=solid]; +"829 permute_21" -> "832 select_14" [label="(3, 4, 12, 64, 32)", style=solid]; +"830 select_12" -> "833 linalg_vector_norm_8" [label="(4, 12, 64, 32)", style=solid]; +"830 select_12" -> "835 expand_as_8" [label="(4, 12, 64, 32)", style=solid]; +"830 select_12" -> "836 div_8" [label="(4, 12, 64, 32)", style=solid]; +"831 select_13" -> "839 linalg_vector_norm_9" [label="(4, 12, 64, 32)", style=solid]; +"831 select_13" -> "841 expand_as_9" [label="(4, 12, 64, 32)", style=solid]; +"831 select_13" -> "842 div_9" [label="(4, 12, 64, 32)", style=solid]; +"832 select_14" -> "853 matmul_9" [label="(4, 12, 64, 32)", style=solid]; +"833 linalg_vector_norm_8" -> "834 clamp_min_8" [label="(4, 12, 64, 1)", style=solid]; +"834 clamp_min_8" -> "835 expand_as_8" [label="(4, 12, 64, 1)", style=solid]; +"835 expand_as_8" -> "836 div_8" [label="(4, 12, 64, 32)", style=solid]; +"836 div_8" -> "837 quantize_per_tensor_default_28" [label="(4, 12, 64, 32)", style=solid]; +"837 quantize_per_tensor_default_28" -> "838 dequantize_per_tensor_default_28" [label="(4, 12, 64, 32)", style=solid]; +"838 dequantize_per_tensor_default_28" -> "846 matmul_8" [label="(4, 12, 64, 32)", style=solid]; +"839 linalg_vector_norm_9" -> "840 clamp_min_9" [label="(4, 12, 64, 1)", style=solid]; +"840 clamp_min_9" -> "841 expand_as_9" [label="(4, 12, 64, 1)", style=solid]; +"841 expand_as_9" -> "842 div_9" [label="(4, 12, 64, 32)", style=solid]; +"842 div_9" -> "843 quantize_per_tensor_default_29" [label="(4, 12, 64, 32)", style=solid]; +"843 quantize_per_tensor_default_29" -> "844 dequantize_per_tensor_default_29" [label="(4, 12, 64, 32)", style=solid]; +"844 dequantize_per_tensor_default_29" -> "845 transpose_8" [label="(4, 12, 64, 32)", style=solid]; +"845 transpose_8" -> "846 matmul_8" [label="(4, 12, 32, 64)", style=solid]; +"846 matmul_8" -> "849 mul_9" [label="(4, 12, 64, 64)", style=solid]; +"847 clamp_4" -> "848 exp_4" [label="(12, 1, 1)", style=solid]; +"848 exp_4" -> "849 mul_9" [label="(12, 1, 1)", style=solid]; +"849 mul_9" -> "850 add_14" [label="(4, 12, 64, 64)", style=solid]; +"850 add_14" -> "851 softmax_4" [label="(4, 12, 64, 64)", style=solid]; +"851 softmax_4" -> "852 dropout_16" [label="(4, 12, 64, 64)", style=solid]; +"852 dropout_16" -> "853 matmul_9" [label="(4, 12, 64, 64)", style=solid]; +"853 matmul_9" -> "854 transpose_9" [label="(4, 12, 64, 32)", style=solid]; +"854 transpose_9" -> "855 reshape_20" [label="(4, 64, 12, 32)", style=solid]; +"855 reshape_20" -> "856 reshape_20_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"856 reshape_20_0_0_nncf_smooth_quant_0" -> "857 quantize_per_tensor_default_30" [label="(4, 64, 384)", style=solid]; +"857 quantize_per_tensor_default_30" -> "858 dequantize_per_tensor_default_30" [label="(4, 64, 384)", style=solid]; +"858 dequantize_per_tensor_default_30" -> "863 linear_29" [label="(4, 64, 384)", style=solid]; +"859 linear_29_scale_0" -> "862 dequantize_per_channel_default_30" [label="(384,)", style=solid]; +"860 linear_29_zero_point_0" -> "862 dequantize_per_channel_default_30" [label="(384,)", style=solid]; +"861 _frozen_param30" -> "862 dequantize_per_channel_default_30" [label="(384, 384)", style=solid]; +"862 dequantize_per_channel_default_30" -> "863 linear_29" [label="(384, 384)", style=solid]; +"863 linear_29" -> "864 dropout_17" [label="(4, 64, 384)", style=solid]; +"864 dropout_17" -> "865 view_25" [label="(4, 64, 384)", style=solid]; +"865 view_25" -> "866 permute_22" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"866 permute_22" -> "867 reshape_21" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"867 reshape_21" -> "868 slice_78" [label="(1, 16, 16, 384)", style=solid]; +"868 slice_78" -> "869 slice_79" [label="(1, 16, 16, 384)", style=solid]; +"869 slice_79" -> "870 slice_80" [label="(1, 14, 16, 384)", style=solid]; +"870 slice_80" -> "871 slice_81" [label="(1, 14, 14, 384)", style=solid]; +"871 slice_81" -> "872 contiguous_7" [label="(1, 14, 14, 384)", style=solid]; +"872 contiguous_7" -> "873 layer_norm_11" [label="(1, 14, 14, 384)", style=solid]; +"873 layer_norm_11" -> "874 add_15" [label="(1, 14, 14, 384)", style=solid]; +"874 add_15" -> "875 add_15_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"874 add_15" -> "895 add_16" [label="(1, 14, 14, 384)", style=solid]; +"875 add_15_0_0_nncf_smooth_quant_0" -> "876 quantize_per_tensor_default_31" [label="(1, 14, 14, 384)", style=solid]; +"876 quantize_per_tensor_default_31" -> "877 dequantize_per_tensor_default_31" [label="(1, 14, 14, 384)", style=solid]; +"877 dequantize_per_tensor_default_31" -> "882 linear_30" [label="(1, 14, 14, 384)", style=solid]; +"878 linear_30_scale_0" -> "881 dequantize_per_channel_default_31" [label="(1536,)", style=solid]; +"879 linear_30_zero_point_0" -> "881 dequantize_per_channel_default_31" [label="(1536,)", style=solid]; +"880 _frozen_param31" -> "881 dequantize_per_channel_default_31" [label="(1536, 384)", style=solid]; +"881 dequantize_per_channel_default_31" -> "882 linear_30" [label="(1536, 384)", style=solid]; +"882 linear_30" -> "883 gelu_4" [label="(1, 14, 14, 1536)", style=solid]; +"883 gelu_4" -> "884 dropout_18" [label="(1, 14, 14, 1536)", style=solid]; +"884 dropout_18" -> "885 dropout_18_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"885 dropout_18_0_0_nncf_smooth_quant_0" -> "886 quantize_per_tensor_default_32" [label="(1, 14, 14, 1536)", style=solid]; +"886 quantize_per_tensor_default_32" -> "887 dequantize_per_tensor_default_32" [label="(1, 14, 14, 1536)", style=solid]; +"887 dequantize_per_tensor_default_32" -> "892 linear_31" [label="(1, 14, 14, 1536)", style=solid]; +"888 linear_31_scale_0" -> "891 dequantize_per_channel_default_32" [label="(384,)", style=solid]; +"889 linear_31_zero_point_0" -> "891 dequantize_per_channel_default_32" [label="(384,)", style=solid]; +"890 _frozen_param32" -> "891 dequantize_per_channel_default_32" [label="(384, 1536)", style=solid]; +"891 dequantize_per_channel_default_32" -> "892 linear_31" [label="(384, 1536)", style=solid]; +"892 linear_31" -> "893 dropout_19" [label="(1, 14, 14, 384)", style=solid]; +"893 dropout_19" -> "894 layer_norm_12" [label="(1, 14, 14, 384)", style=solid]; +"894 layer_norm_12" -> "895 add_16" [label="(1, 14, 14, 384)", style=solid]; +"895 add_16" -> "917 pad_7" [label="(1, 14, 14, 384)", style=solid]; +"895 add_16" -> "994 add_19" [label="(1, 14, 14, 384)", style=solid]; +"896 features_5_1_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "901 linear_32" [label="(1, 15, 15, 2)", style=solid]; +"897 linear_32_scale_0" -> "900 dequantize_per_channel_default_33" [label="(512,)", style=solid]; +"898 linear_32_zero_point_0" -> "900 dequantize_per_channel_default_33" [label="(512,)", style=solid]; +"899 _frozen_param33" -> "900 dequantize_per_channel_default_33" [label="(512, 2)", style=solid]; +"900 dequantize_per_channel_default_33" -> "901 linear_32" [label="(512, 2)", style=solid]; +"901 linear_32" -> "902 relu__5" [label="(1, 15, 15, 512)", style=solid]; +"902 relu__5" -> "903 relu__5_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"903 relu__5_0_0_nncf_smooth_quant_0" -> "908 linear_33" [label="(1, 15, 15, 512)", style=solid]; +"904 linear_33_scale_0" -> "907 dequantize_per_channel_default_34" [label="(12,)", style=solid]; +"905 linear_33_zero_point_0" -> "907 dequantize_per_channel_default_34" [label="(12,)", style=solid]; +"906 _frozen_param34" -> "907 dequantize_per_channel_default_34" [label="(12, 512)", style=solid]; +"907 dequantize_per_channel_default_34" -> "908 linear_33" [label="(12, 512)", style=solid]; +"908 linear_33" -> "909 view_26" [label="(1, 15, 15, 12)", style=solid]; +"909 view_26" -> "910 index_5" [label="(225, 12)", style=solid]; +"910 index_5" -> "911 view_27" [label="(4096, 12)", style=solid]; +"911 view_27" -> "912 permute_23" [label="(64, 64, 12)", style=solid]; +"912 permute_23" -> "913 contiguous_8" [label="(12, 64, 64)", style=solid]; +"913 contiguous_8" -> "914 unsqueeze_13" [label="(12, 64, 64)", style=solid]; +"914 unsqueeze_13" -> "915 sigmoid_5" [label="(1, 12, 64, 64)", style=solid]; +"915 sigmoid_5" -> "916 mul_10" [label="(1, 12, 64, 64)", style=solid]; +"916 mul_10" -> "953 add_17" [label="(1, 12, 64, 64)", style=solid]; +"917 pad_7" -> "918 roll_4" [label="(1, 16, 16, 384)", style=solid]; +"918 roll_4" -> "919 view_28" [label="(1, 16, 16, 384)", style=solid]; +"919 view_28" -> "920 permute_24" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"920 permute_24" -> "921 reshape_22" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"921 reshape_22" -> "923 reshape_22_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"921 reshape_22" -> "954 new_zeros_2" [label="(4, 64, 384)", style=solid]; +"922 clone_5" -> "930 linear_34" [label="(1152,)", style=solid]; +"923 reshape_22_0_0_nncf_smooth_quant_0" -> "924 quantize_per_tensor_default_33" [label="(4, 64, 384)", style=solid]; +"924 quantize_per_tensor_default_33" -> "925 dequantize_per_tensor_default_33" [label="(4, 64, 384)", style=solid]; +"925 dequantize_per_tensor_default_33" -> "930 linear_34" [label="(4, 64, 384)", style=solid]; +"926 linear_34_scale_0" -> "929 dequantize_per_channel_default_35" [label="(1152,)", style=solid]; +"927 linear_34_zero_point_0" -> "929 dequantize_per_channel_default_35" [label="(1152,)", style=solid]; +"928 _frozen_param35" -> "929 dequantize_per_channel_default_35" [label="(1152, 384)", style=solid]; +"929 dequantize_per_channel_default_35" -> "930 linear_34" [label="(1152, 384)", style=solid]; +"930 linear_34" -> "931 reshape_23" [label="(4, 64, 1152)", style=solid]; +"931 reshape_23" -> "932 permute_25" [label="(4, 64, 3, 12, 32)", style=solid]; +"932 permute_25" -> "933 select_15" [label="(3, 4, 12, 64, 32)", style=solid]; +"932 permute_25" -> "934 select_16" [label="(3, 4, 12, 64, 32)", style=solid]; +"932 permute_25" -> "935 select_17" [label="(3, 4, 12, 64, 32)", style=solid]; +"933 select_15" -> "936 linalg_vector_norm_10" [label="(4, 12, 64, 32)", style=solid]; +"933 select_15" -> "938 expand_as_10" [label="(4, 12, 64, 32)", style=solid]; +"933 select_15" -> "939 div_10" [label="(4, 12, 64, 32)", style=solid]; +"934 select_16" -> "942 linalg_vector_norm_11" [label="(4, 12, 64, 32)", style=solid]; +"934 select_16" -> "944 expand_as_11" [label="(4, 12, 64, 32)", style=solid]; +"934 select_16" -> "945 div_11" [label="(4, 12, 64, 32)", style=solid]; +"935 select_17" -> "972 matmul_11" [label="(4, 12, 64, 32)", style=solid]; +"936 linalg_vector_norm_10" -> "937 clamp_min_10" [label="(4, 12, 64, 1)", style=solid]; +"937 clamp_min_10" -> "938 expand_as_10" [label="(4, 12, 64, 1)", style=solid]; +"938 expand_as_10" -> "939 div_10" [label="(4, 12, 64, 32)", style=solid]; +"939 div_10" -> "940 quantize_per_tensor_default_34" [label="(4, 12, 64, 32)", style=solid]; +"940 quantize_per_tensor_default_34" -> "941 dequantize_per_tensor_default_34" [label="(4, 12, 64, 32)", style=solid]; +"941 dequantize_per_tensor_default_34" -> "949 matmul_10" [label="(4, 12, 64, 32)", style=solid]; +"942 linalg_vector_norm_11" -> "943 clamp_min_11" [label="(4, 12, 64, 1)", style=solid]; +"943 clamp_min_11" -> "944 expand_as_11" [label="(4, 12, 64, 1)", style=solid]; +"944 expand_as_11" -> "945 div_11" [label="(4, 12, 64, 32)", style=solid]; +"945 div_11" -> "946 quantize_per_tensor_default_35" [label="(4, 12, 64, 32)", style=solid]; +"946 quantize_per_tensor_default_35" -> "947 dequantize_per_tensor_default_35" [label="(4, 12, 64, 32)", style=solid]; +"947 dequantize_per_tensor_default_35" -> "948 transpose_10" [label="(4, 12, 64, 32)", style=solid]; +"948 transpose_10" -> "949 matmul_10" [label="(4, 12, 32, 64)", style=solid]; +"949 matmul_10" -> "952 mul_11" [label="(4, 12, 64, 64)", style=solid]; +"950 clamp_5" -> "951 exp_5" [label="(12, 1, 1)", style=solid]; +"951 exp_5" -> "952 mul_11" [label="(12, 1, 1)", style=solid]; +"952 mul_11" -> "953 add_17" [label="(4, 12, 64, 64)", style=solid]; +"953 add_17" -> "965 view_30" [label="(4, 12, 64, 64)", style=solid]; +"954 new_zeros_2" -> "955 view_29" [label="(16, 16)", style=solid]; +"955 view_29" -> "956 permute_26" [label="(2, 8, 2, 8)", style=solid]; +"956 permute_26" -> "957 reshape_24" [label="(2, 2, 8, 8)", style=solid]; +"957 reshape_24" -> "958 unsqueeze_14" [label="(4, 64)", style=solid]; +"957 reshape_24" -> "959 unsqueeze_15" [label="(4, 64)", style=solid]; +"958 unsqueeze_14" -> "960 sub_2" [label="(4, 1, 64)", style=solid]; +"959 unsqueeze_15" -> "960 sub_2" [label="(4, 64, 1)", style=solid]; +"960 sub_2" -> "961 ne_2" [label="(4, 64, 64)", style=solid]; +"960 sub_2" -> "962 masked_fill_4" [label="(4, 64, 64)", style=solid]; +"960 sub_2" -> "963 eq_2" [label="(4, 64, 64)", style=solid]; +"961 ne_2" -> "962 masked_fill_4" [label="(4, 64, 64)", style=solid]; +"962 masked_fill_4" -> "964 masked_fill_5" [label="(4, 64, 64)", style=solid]; +"963 eq_2" -> "964 masked_fill_5" [label="(4, 64, 64)", style=solid]; +"964 masked_fill_5" -> "966 unsqueeze_16" [label="(4, 64, 64)", style=solid]; +"965 view_30" -> "968 add_18" [label="(1, 4, 12, 64, 64)", style=solid]; +"966 unsqueeze_16" -> "967 unsqueeze_17" [label="(4, 1, 64, 64)", style=solid]; +"967 unsqueeze_17" -> "968 add_18" [label="(1, 4, 1, 64, 64)", style=solid]; +"968 add_18" -> "969 view_31" [label="(1, 4, 12, 64, 64)", style=solid]; +"969 view_31" -> "970 softmax_5" [label="(4, 12, 64, 64)", style=solid]; +"970 softmax_5" -> "971 dropout_20" [label="(4, 12, 64, 64)", style=solid]; +"971 dropout_20" -> "972 matmul_11" [label="(4, 12, 64, 64)", style=solid]; +"972 matmul_11" -> "973 transpose_11" [label="(4, 12, 64, 32)", style=solid]; +"973 transpose_11" -> "974 reshape_25" [label="(4, 64, 12, 32)", style=solid]; +"974 reshape_25" -> "975 reshape_25_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"975 reshape_25_0_0_nncf_smooth_quant_0" -> "976 quantize_per_tensor_default_36" [label="(4, 64, 384)", style=solid]; +"976 quantize_per_tensor_default_36" -> "977 dequantize_per_tensor_default_36" [label="(4, 64, 384)", style=solid]; +"977 dequantize_per_tensor_default_36" -> "982 linear_35" [label="(4, 64, 384)", style=solid]; +"978 linear_35_scale_0" -> "981 dequantize_per_channel_default_36" [label="(384,)", style=solid]; +"979 linear_35_zero_point_0" -> "981 dequantize_per_channel_default_36" [label="(384,)", style=solid]; +"980 _frozen_param36" -> "981 dequantize_per_channel_default_36" [label="(384, 384)", style=solid]; +"981 dequantize_per_channel_default_36" -> "982 linear_35" [label="(384, 384)", style=solid]; +"982 linear_35" -> "983 dropout_21" [label="(4, 64, 384)", style=solid]; +"983 dropout_21" -> "984 view_32" [label="(4, 64, 384)", style=solid]; +"984 view_32" -> "985 permute_27" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"985 permute_27" -> "986 reshape_26" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"986 reshape_26" -> "987 roll_5" [label="(1, 16, 16, 384)", style=solid]; +"987 roll_5" -> "988 slice_101" [label="(1, 16, 16, 384)", style=solid]; +"988 slice_101" -> "989 slice_102" [label="(1, 16, 16, 384)", style=solid]; +"989 slice_102" -> "990 slice_103" [label="(1, 14, 16, 384)", style=solid]; +"990 slice_103" -> "991 slice_104" [label="(1, 14, 14, 384)", style=solid]; +"991 slice_104" -> "992 contiguous_9" [label="(1, 14, 14, 384)", style=solid]; +"992 contiguous_9" -> "993 layer_norm_13" [label="(1, 14, 14, 384)", style=solid]; +"993 layer_norm_13" -> "994 add_19" [label="(1, 14, 14, 384)", style=solid]; +"994 add_19" -> "995 add_19_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"994 add_19" -> "1015 add_20" [label="(1, 14, 14, 384)", style=solid]; +"995 add_19_0_0_nncf_smooth_quant_0" -> "996 quantize_per_tensor_default_37" [label="(1, 14, 14, 384)", style=solid]; +"996 quantize_per_tensor_default_37" -> "997 dequantize_per_tensor_default_37" [label="(1, 14, 14, 384)", style=solid]; +"997 dequantize_per_tensor_default_37" -> "1002 linear_36" [label="(1, 14, 14, 384)", style=solid]; +"998 linear_36_scale_0" -> "1001 dequantize_per_channel_default_37" [label="(1536,)", style=solid]; +"999 linear_36_zero_point_0" -> "1001 dequantize_per_channel_default_37" [label="(1536,)", style=solid]; +"1000 _frozen_param37" -> "1001 dequantize_per_channel_default_37" [label="(1536, 384)", style=solid]; +"1001 dequantize_per_channel_default_37" -> "1002 linear_36" [label="(1536, 384)", style=solid]; +"1002 linear_36" -> "1003 gelu_5" [label="(1, 14, 14, 1536)", style=solid]; +"1003 gelu_5" -> "1004 dropout_22" [label="(1, 14, 14, 1536)", style=solid]; +"1004 dropout_22" -> "1005 dropout_22_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"1005 dropout_22_0_0_nncf_smooth_quant_0" -> "1006 quantize_per_tensor_default_38" [label="(1, 14, 14, 1536)", style=solid]; +"1006 quantize_per_tensor_default_38" -> "1007 dequantize_per_tensor_default_38" [label="(1, 14, 14, 1536)", style=solid]; +"1007 dequantize_per_tensor_default_38" -> "1012 linear_37" [label="(1, 14, 14, 1536)", style=solid]; +"1008 linear_37_scale_0" -> "1011 dequantize_per_channel_default_38" [label="(384,)", style=solid]; +"1009 linear_37_zero_point_0" -> "1011 dequantize_per_channel_default_38" [label="(384,)", style=solid]; +"1010 _frozen_param38" -> "1011 dequantize_per_channel_default_38" [label="(384, 1536)", style=solid]; +"1011 dequantize_per_channel_default_38" -> "1012 linear_37" [label="(384, 1536)", style=solid]; +"1012 linear_37" -> "1013 dropout_23" [label="(1, 14, 14, 384)", style=solid]; +"1013 dropout_23" -> "1014 layer_norm_14" [label="(1, 14, 14, 384)", style=solid]; +"1014 layer_norm_14" -> "1015 add_20" [label="(1, 14, 14, 384)", style=solid]; +"1015 add_20" -> "1037 pad_8" [label="(1, 14, 14, 384)", style=solid]; +"1015 add_20" -> "1096 add_22" [label="(1, 14, 14, 384)", style=solid]; +"1016 features_5_2_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "1021 linear_38" [label="(1, 15, 15, 2)", style=solid]; +"1017 linear_38_scale_0" -> "1020 dequantize_per_channel_default_39" [label="(512,)", style=solid]; +"1018 linear_38_zero_point_0" -> "1020 dequantize_per_channel_default_39" [label="(512,)", style=solid]; +"1019 _frozen_param39" -> "1020 dequantize_per_channel_default_39" [label="(512, 2)", style=solid]; +"1020 dequantize_per_channel_default_39" -> "1021 linear_38" [label="(512, 2)", style=solid]; +"1021 linear_38" -> "1022 relu__6" [label="(1, 15, 15, 512)", style=solid]; +"1022 relu__6" -> "1023 relu__6_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"1023 relu__6_0_0_nncf_smooth_quant_0" -> "1028 linear_39" [label="(1, 15, 15, 512)", style=solid]; +"1024 linear_39_scale_0" -> "1027 dequantize_per_channel_default_40" [label="(12,)", style=solid]; +"1025 linear_39_zero_point_0" -> "1027 dequantize_per_channel_default_40" [label="(12,)", style=solid]; +"1026 _frozen_param40" -> "1027 dequantize_per_channel_default_40" [label="(12, 512)", style=solid]; +"1027 dequantize_per_channel_default_40" -> "1028 linear_39" [label="(12, 512)", style=solid]; +"1028 linear_39" -> "1029 view_33" [label="(1, 15, 15, 12)", style=solid]; +"1029 view_33" -> "1030 index_6" [label="(225, 12)", style=solid]; +"1030 index_6" -> "1031 view_34" [label="(4096, 12)", style=solid]; +"1031 view_34" -> "1032 permute_28" [label="(64, 64, 12)", style=solid]; +"1032 permute_28" -> "1033 contiguous_10" [label="(12, 64, 64)", style=solid]; +"1033 contiguous_10" -> "1034 unsqueeze_18" [label="(12, 64, 64)", style=solid]; +"1034 unsqueeze_18" -> "1035 sigmoid_6" [label="(1, 12, 64, 64)", style=solid]; +"1035 sigmoid_6" -> "1036 mul_12" [label="(1, 12, 64, 64)", style=solid]; +"1036 mul_12" -> "1072 add_21" [label="(1, 12, 64, 64)", style=solid]; +"1037 pad_8" -> "1038 view_35" [label="(1, 16, 16, 384)", style=solid]; +"1038 view_35" -> "1039 permute_29" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1039 permute_29" -> "1040 reshape_27" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1040 reshape_27" -> "1042 reshape_27_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1041 clone_6" -> "1049 linear_40" [label="(1152,)", style=solid]; +"1042 reshape_27_0_0_nncf_smooth_quant_0" -> "1043 quantize_per_tensor_default_39" [label="(4, 64, 384)", style=solid]; +"1043 quantize_per_tensor_default_39" -> "1044 dequantize_per_tensor_default_39" [label="(4, 64, 384)", style=solid]; +"1044 dequantize_per_tensor_default_39" -> "1049 linear_40" [label="(4, 64, 384)", style=solid]; +"1045 linear_40_scale_0" -> "1048 dequantize_per_channel_default_41" [label="(1152,)", style=solid]; +"1046 linear_40_zero_point_0" -> "1048 dequantize_per_channel_default_41" [label="(1152,)", style=solid]; +"1047 _frozen_param41" -> "1048 dequantize_per_channel_default_41" [label="(1152, 384)", style=solid]; +"1048 dequantize_per_channel_default_41" -> "1049 linear_40" [label="(1152, 384)", style=solid]; +"1049 linear_40" -> "1050 reshape_28" [label="(4, 64, 1152)", style=solid]; +"1050 reshape_28" -> "1051 permute_30" [label="(4, 64, 3, 12, 32)", style=solid]; +"1051 permute_30" -> "1052 select_18" [label="(3, 4, 12, 64, 32)", style=solid]; +"1051 permute_30" -> "1053 select_19" [label="(3, 4, 12, 64, 32)", style=solid]; +"1051 permute_30" -> "1054 select_20" [label="(3, 4, 12, 64, 32)", style=solid]; +"1052 select_18" -> "1055 linalg_vector_norm_12" [label="(4, 12, 64, 32)", style=solid]; +"1052 select_18" -> "1057 expand_as_12" [label="(4, 12, 64, 32)", style=solid]; +"1052 select_18" -> "1058 div_12" [label="(4, 12, 64, 32)", style=solid]; +"1053 select_19" -> "1061 linalg_vector_norm_13" [label="(4, 12, 64, 32)", style=solid]; +"1053 select_19" -> "1063 expand_as_13" [label="(4, 12, 64, 32)", style=solid]; +"1053 select_19" -> "1064 div_13" [label="(4, 12, 64, 32)", style=solid]; +"1054 select_20" -> "1075 matmul_13" [label="(4, 12, 64, 32)", style=solid]; +"1055 linalg_vector_norm_12" -> "1056 clamp_min_12" [label="(4, 12, 64, 1)", style=solid]; +"1056 clamp_min_12" -> "1057 expand_as_12" [label="(4, 12, 64, 1)", style=solid]; +"1057 expand_as_12" -> "1058 div_12" [label="(4, 12, 64, 32)", style=solid]; +"1058 div_12" -> "1059 quantize_per_tensor_default_40" [label="(4, 12, 64, 32)", style=solid]; +"1059 quantize_per_tensor_default_40" -> "1060 dequantize_per_tensor_default_40" [label="(4, 12, 64, 32)", style=solid]; +"1060 dequantize_per_tensor_default_40" -> "1068 matmul_12" [label="(4, 12, 64, 32)", style=solid]; +"1061 linalg_vector_norm_13" -> "1062 clamp_min_13" [label="(4, 12, 64, 1)", style=solid]; +"1062 clamp_min_13" -> "1063 expand_as_13" [label="(4, 12, 64, 1)", style=solid]; +"1063 expand_as_13" -> "1064 div_13" [label="(4, 12, 64, 32)", style=solid]; +"1064 div_13" -> "1065 quantize_per_tensor_default_41" [label="(4, 12, 64, 32)", style=solid]; +"1065 quantize_per_tensor_default_41" -> "1066 dequantize_per_tensor_default_41" [label="(4, 12, 64, 32)", style=solid]; +"1066 dequantize_per_tensor_default_41" -> "1067 transpose_12" [label="(4, 12, 64, 32)", style=solid]; +"1067 transpose_12" -> "1068 matmul_12" [label="(4, 12, 32, 64)", style=solid]; +"1068 matmul_12" -> "1071 mul_13" [label="(4, 12, 64, 64)", style=solid]; +"1069 clamp_6" -> "1070 exp_6" [label="(12, 1, 1)", style=solid]; +"1070 exp_6" -> "1071 mul_13" [label="(12, 1, 1)", style=solid]; +"1071 mul_13" -> "1072 add_21" [label="(4, 12, 64, 64)", style=solid]; +"1072 add_21" -> "1073 softmax_6" [label="(4, 12, 64, 64)", style=solid]; +"1073 softmax_6" -> "1074 dropout_24" [label="(4, 12, 64, 64)", style=solid]; +"1074 dropout_24" -> "1075 matmul_13" [label="(4, 12, 64, 64)", style=solid]; +"1075 matmul_13" -> "1076 transpose_13" [label="(4, 12, 64, 32)", style=solid]; +"1076 transpose_13" -> "1077 reshape_29" [label="(4, 64, 12, 32)", style=solid]; +"1077 reshape_29" -> "1078 reshape_29_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1078 reshape_29_0_0_nncf_smooth_quant_0" -> "1079 quantize_per_tensor_default_42" [label="(4, 64, 384)", style=solid]; +"1079 quantize_per_tensor_default_42" -> "1080 dequantize_per_tensor_default_42" [label="(4, 64, 384)", style=solid]; +"1080 dequantize_per_tensor_default_42" -> "1085 linear_41" [label="(4, 64, 384)", style=solid]; +"1081 linear_41_scale_0" -> "1084 dequantize_per_channel_default_42" [label="(384,)", style=solid]; +"1082 linear_41_zero_point_0" -> "1084 dequantize_per_channel_default_42" [label="(384,)", style=solid]; +"1083 _frozen_param42" -> "1084 dequantize_per_channel_default_42" [label="(384, 384)", style=solid]; +"1084 dequantize_per_channel_default_42" -> "1085 linear_41" [label="(384, 384)", style=solid]; +"1085 linear_41" -> "1086 dropout_25" [label="(4, 64, 384)", style=solid]; +"1086 dropout_25" -> "1087 view_36" [label="(4, 64, 384)", style=solid]; +"1087 view_36" -> "1088 permute_31" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1088 permute_31" -> "1089 reshape_30" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1089 reshape_30" -> "1090 slice_106" [label="(1, 16, 16, 384)", style=solid]; +"1090 slice_106" -> "1091 slice_107" [label="(1, 16, 16, 384)", style=solid]; +"1091 slice_107" -> "1092 slice_108" [label="(1, 14, 16, 384)", style=solid]; +"1092 slice_108" -> "1093 slice_109" [label="(1, 14, 14, 384)", style=solid]; +"1093 slice_109" -> "1094 contiguous_11" [label="(1, 14, 14, 384)", style=solid]; +"1094 contiguous_11" -> "1095 layer_norm_15" [label="(1, 14, 14, 384)", style=solid]; +"1095 layer_norm_15" -> "1096 add_22" [label="(1, 14, 14, 384)", style=solid]; +"1096 add_22" -> "1097 add_22_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"1096 add_22" -> "1117 add_23" [label="(1, 14, 14, 384)", style=solid]; +"1097 add_22_0_0_nncf_smooth_quant_0" -> "1098 quantize_per_tensor_default_43" [label="(1, 14, 14, 384)", style=solid]; +"1098 quantize_per_tensor_default_43" -> "1099 dequantize_per_tensor_default_43" [label="(1, 14, 14, 384)", style=solid]; +"1099 dequantize_per_tensor_default_43" -> "1104 linear_42" [label="(1, 14, 14, 384)", style=solid]; +"1100 linear_42_scale_0" -> "1103 dequantize_per_channel_default_43" [label="(1536,)", style=solid]; +"1101 linear_42_zero_point_0" -> "1103 dequantize_per_channel_default_43" [label="(1536,)", style=solid]; +"1102 _frozen_param43" -> "1103 dequantize_per_channel_default_43" [label="(1536, 384)", style=solid]; +"1103 dequantize_per_channel_default_43" -> "1104 linear_42" [label="(1536, 384)", style=solid]; +"1104 linear_42" -> "1105 gelu_6" [label="(1, 14, 14, 1536)", style=solid]; +"1105 gelu_6" -> "1106 dropout_26" [label="(1, 14, 14, 1536)", style=solid]; +"1106 dropout_26" -> "1107 dropout_26_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"1107 dropout_26_0_0_nncf_smooth_quant_0" -> "1108 quantize_per_tensor_default_44" [label="(1, 14, 14, 1536)", style=solid]; +"1108 quantize_per_tensor_default_44" -> "1109 dequantize_per_tensor_default_44" [label="(1, 14, 14, 1536)", style=solid]; +"1109 dequantize_per_tensor_default_44" -> "1114 linear_43" [label="(1, 14, 14, 1536)", style=solid]; +"1110 linear_43_scale_0" -> "1113 dequantize_per_channel_default_44" [label="(384,)", style=solid]; +"1111 linear_43_zero_point_0" -> "1113 dequantize_per_channel_default_44" [label="(384,)", style=solid]; +"1112 _frozen_param44" -> "1113 dequantize_per_channel_default_44" [label="(384, 1536)", style=solid]; +"1113 dequantize_per_channel_default_44" -> "1114 linear_43" [label="(384, 1536)", style=solid]; +"1114 linear_43" -> "1115 dropout_27" [label="(1, 14, 14, 384)", style=solid]; +"1115 dropout_27" -> "1116 layer_norm_16" [label="(1, 14, 14, 384)", style=solid]; +"1116 layer_norm_16" -> "1117 add_23" [label="(1, 14, 14, 384)", style=solid]; +"1117 add_23" -> "1139 pad_9" [label="(1, 14, 14, 384)", style=solid]; +"1117 add_23" -> "1216 add_26" [label="(1, 14, 14, 384)", style=solid]; +"1118 features_5_3_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "1123 linear_44" [label="(1, 15, 15, 2)", style=solid]; +"1119 linear_44_scale_0" -> "1122 dequantize_per_channel_default_45" [label="(512,)", style=solid]; +"1120 linear_44_zero_point_0" -> "1122 dequantize_per_channel_default_45" [label="(512,)", style=solid]; +"1121 _frozen_param45" -> "1122 dequantize_per_channel_default_45" [label="(512, 2)", style=solid]; +"1122 dequantize_per_channel_default_45" -> "1123 linear_44" [label="(512, 2)", style=solid]; +"1123 linear_44" -> "1124 relu__7" [label="(1, 15, 15, 512)", style=solid]; +"1124 relu__7" -> "1125 relu__7_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"1125 relu__7_0_0_nncf_smooth_quant_0" -> "1130 linear_45" [label="(1, 15, 15, 512)", style=solid]; +"1126 linear_45_scale_0" -> "1129 dequantize_per_channel_default_46" [label="(12,)", style=solid]; +"1127 linear_45_zero_point_0" -> "1129 dequantize_per_channel_default_46" [label="(12,)", style=solid]; +"1128 _frozen_param46" -> "1129 dequantize_per_channel_default_46" [label="(12, 512)", style=solid]; +"1129 dequantize_per_channel_default_46" -> "1130 linear_45" [label="(12, 512)", style=solid]; +"1130 linear_45" -> "1131 view_37" [label="(1, 15, 15, 12)", style=solid]; +"1131 view_37" -> "1132 index_7" [label="(225, 12)", style=solid]; +"1132 index_7" -> "1133 view_38" [label="(4096, 12)", style=solid]; +"1133 view_38" -> "1134 permute_32" [label="(64, 64, 12)", style=solid]; +"1134 permute_32" -> "1135 contiguous_12" [label="(12, 64, 64)", style=solid]; +"1135 contiguous_12" -> "1136 unsqueeze_19" [label="(12, 64, 64)", style=solid]; +"1136 unsqueeze_19" -> "1137 sigmoid_7" [label="(1, 12, 64, 64)", style=solid]; +"1137 sigmoid_7" -> "1138 mul_14" [label="(1, 12, 64, 64)", style=solid]; +"1138 mul_14" -> "1175 add_24" [label="(1, 12, 64, 64)", style=solid]; +"1139 pad_9" -> "1140 roll_6" [label="(1, 16, 16, 384)", style=solid]; +"1140 roll_6" -> "1141 view_39" [label="(1, 16, 16, 384)", style=solid]; +"1141 view_39" -> "1142 permute_33" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1142 permute_33" -> "1143 reshape_31" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1143 reshape_31" -> "1145 reshape_31_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1143 reshape_31" -> "1176 new_zeros_3" [label="(4, 64, 384)", style=solid]; +"1144 clone_7" -> "1152 linear_46" [label="(1152,)", style=solid]; +"1145 reshape_31_0_0_nncf_smooth_quant_0" -> "1146 quantize_per_tensor_default_45" [label="(4, 64, 384)", style=solid]; +"1146 quantize_per_tensor_default_45" -> "1147 dequantize_per_tensor_default_45" [label="(4, 64, 384)", style=solid]; +"1147 dequantize_per_tensor_default_45" -> "1152 linear_46" [label="(4, 64, 384)", style=solid]; +"1148 linear_46_scale_0" -> "1151 dequantize_per_channel_default_47" [label="(1152,)", style=solid]; +"1149 linear_46_zero_point_0" -> "1151 dequantize_per_channel_default_47" [label="(1152,)", style=solid]; +"1150 _frozen_param47" -> "1151 dequantize_per_channel_default_47" [label="(1152, 384)", style=solid]; +"1151 dequantize_per_channel_default_47" -> "1152 linear_46" [label="(1152, 384)", style=solid]; +"1152 linear_46" -> "1153 reshape_32" [label="(4, 64, 1152)", style=solid]; +"1153 reshape_32" -> "1154 permute_34" [label="(4, 64, 3, 12, 32)", style=solid]; +"1154 permute_34" -> "1155 select_21" [label="(3, 4, 12, 64, 32)", style=solid]; +"1154 permute_34" -> "1156 select_22" [label="(3, 4, 12, 64, 32)", style=solid]; +"1154 permute_34" -> "1157 select_23" [label="(3, 4, 12, 64, 32)", style=solid]; +"1155 select_21" -> "1158 linalg_vector_norm_14" [label="(4, 12, 64, 32)", style=solid]; +"1155 select_21" -> "1160 expand_as_14" [label="(4, 12, 64, 32)", style=solid]; +"1155 select_21" -> "1161 div_14" [label="(4, 12, 64, 32)", style=solid]; +"1156 select_22" -> "1164 linalg_vector_norm_15" [label="(4, 12, 64, 32)", style=solid]; +"1156 select_22" -> "1166 expand_as_15" [label="(4, 12, 64, 32)", style=solid]; +"1156 select_22" -> "1167 div_15" [label="(4, 12, 64, 32)", style=solid]; +"1157 select_23" -> "1194 matmul_15" [label="(4, 12, 64, 32)", style=solid]; +"1158 linalg_vector_norm_14" -> "1159 clamp_min_14" [label="(4, 12, 64, 1)", style=solid]; +"1159 clamp_min_14" -> "1160 expand_as_14" [label="(4, 12, 64, 1)", style=solid]; +"1160 expand_as_14" -> "1161 div_14" [label="(4, 12, 64, 32)", style=solid]; +"1161 div_14" -> "1162 quantize_per_tensor_default_46" [label="(4, 12, 64, 32)", style=solid]; +"1162 quantize_per_tensor_default_46" -> "1163 dequantize_per_tensor_default_46" [label="(4, 12, 64, 32)", style=solid]; +"1163 dequantize_per_tensor_default_46" -> "1171 matmul_14" [label="(4, 12, 64, 32)", style=solid]; +"1164 linalg_vector_norm_15" -> "1165 clamp_min_15" [label="(4, 12, 64, 1)", style=solid]; +"1165 clamp_min_15" -> "1166 expand_as_15" [label="(4, 12, 64, 1)", style=solid]; +"1166 expand_as_15" -> "1167 div_15" [label="(4, 12, 64, 32)", style=solid]; +"1167 div_15" -> "1168 quantize_per_tensor_default_47" [label="(4, 12, 64, 32)", style=solid]; +"1168 quantize_per_tensor_default_47" -> "1169 dequantize_per_tensor_default_47" [label="(4, 12, 64, 32)", style=solid]; +"1169 dequantize_per_tensor_default_47" -> "1170 transpose_14" [label="(4, 12, 64, 32)", style=solid]; +"1170 transpose_14" -> "1171 matmul_14" [label="(4, 12, 32, 64)", style=solid]; +"1171 matmul_14" -> "1174 mul_15" [label="(4, 12, 64, 64)", style=solid]; +"1172 clamp_7" -> "1173 exp_7" [label="(12, 1, 1)", style=solid]; +"1173 exp_7" -> "1174 mul_15" [label="(12, 1, 1)", style=solid]; +"1174 mul_15" -> "1175 add_24" [label="(4, 12, 64, 64)", style=solid]; +"1175 add_24" -> "1187 view_41" [label="(4, 12, 64, 64)", style=solid]; +"1176 new_zeros_3" -> "1177 view_40" [label="(16, 16)", style=solid]; +"1177 view_40" -> "1178 permute_35" [label="(2, 8, 2, 8)", style=solid]; +"1178 permute_35" -> "1179 reshape_33" [label="(2, 2, 8, 8)", style=solid]; +"1179 reshape_33" -> "1180 unsqueeze_20" [label="(4, 64)", style=solid]; +"1179 reshape_33" -> "1181 unsqueeze_21" [label="(4, 64)", style=solid]; +"1180 unsqueeze_20" -> "1182 sub_3" [label="(4, 1, 64)", style=solid]; +"1181 unsqueeze_21" -> "1182 sub_3" [label="(4, 64, 1)", style=solid]; +"1182 sub_3" -> "1183 ne_3" [label="(4, 64, 64)", style=solid]; +"1182 sub_3" -> "1184 masked_fill_6" [label="(4, 64, 64)", style=solid]; +"1182 sub_3" -> "1185 eq_3" [label="(4, 64, 64)", style=solid]; +"1183 ne_3" -> "1184 masked_fill_6" [label="(4, 64, 64)", style=solid]; +"1184 masked_fill_6" -> "1186 masked_fill_7" [label="(4, 64, 64)", style=solid]; +"1185 eq_3" -> "1186 masked_fill_7" [label="(4, 64, 64)", style=solid]; +"1186 masked_fill_7" -> "1188 unsqueeze_22" [label="(4, 64, 64)", style=solid]; +"1187 view_41" -> "1190 add_25" [label="(1, 4, 12, 64, 64)", style=solid]; +"1188 unsqueeze_22" -> "1189 unsqueeze_23" [label="(4, 1, 64, 64)", style=solid]; +"1189 unsqueeze_23" -> "1190 add_25" [label="(1, 4, 1, 64, 64)", style=solid]; +"1190 add_25" -> "1191 view_42" [label="(1, 4, 12, 64, 64)", style=solid]; +"1191 view_42" -> "1192 softmax_7" [label="(4, 12, 64, 64)", style=solid]; +"1192 softmax_7" -> "1193 dropout_28" [label="(4, 12, 64, 64)", style=solid]; +"1193 dropout_28" -> "1194 matmul_15" [label="(4, 12, 64, 64)", style=solid]; +"1194 matmul_15" -> "1195 transpose_15" [label="(4, 12, 64, 32)", style=solid]; +"1195 transpose_15" -> "1196 reshape_34" [label="(4, 64, 12, 32)", style=solid]; +"1196 reshape_34" -> "1197 reshape_34_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1197 reshape_34_0_0_nncf_smooth_quant_0" -> "1198 quantize_per_tensor_default_48" [label="(4, 64, 384)", style=solid]; +"1198 quantize_per_tensor_default_48" -> "1199 dequantize_per_tensor_default_48" [label="(4, 64, 384)", style=solid]; +"1199 dequantize_per_tensor_default_48" -> "1204 linear_47" [label="(4, 64, 384)", style=solid]; +"1200 linear_47_scale_0" -> "1203 dequantize_per_channel_default_48" [label="(384,)", style=solid]; +"1201 linear_47_zero_point_0" -> "1203 dequantize_per_channel_default_48" [label="(384,)", style=solid]; +"1202 _frozen_param48" -> "1203 dequantize_per_channel_default_48" [label="(384, 384)", style=solid]; +"1203 dequantize_per_channel_default_48" -> "1204 linear_47" [label="(384, 384)", style=solid]; +"1204 linear_47" -> "1205 dropout_29" [label="(4, 64, 384)", style=solid]; +"1205 dropout_29" -> "1206 view_43" [label="(4, 64, 384)", style=solid]; +"1206 view_43" -> "1207 permute_36" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1207 permute_36" -> "1208 reshape_35" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1208 reshape_35" -> "1209 roll_7" [label="(1, 16, 16, 384)", style=solid]; +"1209 roll_7" -> "1210 slice_129" [label="(1, 16, 16, 384)", style=solid]; +"1210 slice_129" -> "1211 slice_130" [label="(1, 16, 16, 384)", style=solid]; +"1211 slice_130" -> "1212 slice_131" [label="(1, 14, 16, 384)", style=solid]; +"1212 slice_131" -> "1213 slice_132" [label="(1, 14, 14, 384)", style=solid]; +"1213 slice_132" -> "1214 contiguous_13" [label="(1, 14, 14, 384)", style=solid]; +"1214 contiguous_13" -> "1215 layer_norm_17" [label="(1, 14, 14, 384)", style=solid]; +"1215 layer_norm_17" -> "1216 add_26" [label="(1, 14, 14, 384)", style=solid]; +"1216 add_26" -> "1217 add_26_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"1216 add_26" -> "1237 add_27" [label="(1, 14, 14, 384)", style=solid]; +"1217 add_26_0_0_nncf_smooth_quant_0" -> "1218 quantize_per_tensor_default_49" [label="(1, 14, 14, 384)", style=solid]; +"1218 quantize_per_tensor_default_49" -> "1219 dequantize_per_tensor_default_49" [label="(1, 14, 14, 384)", style=solid]; +"1219 dequantize_per_tensor_default_49" -> "1224 linear_48" [label="(1, 14, 14, 384)", style=solid]; +"1220 linear_48_scale_0" -> "1223 dequantize_per_channel_default_49" [label="(1536,)", style=solid]; +"1221 linear_48_zero_point_0" -> "1223 dequantize_per_channel_default_49" [label="(1536,)", style=solid]; +"1222 _frozen_param49" -> "1223 dequantize_per_channel_default_49" [label="(1536, 384)", style=solid]; +"1223 dequantize_per_channel_default_49" -> "1224 linear_48" [label="(1536, 384)", style=solid]; +"1224 linear_48" -> "1225 gelu_7" [label="(1, 14, 14, 1536)", style=solid]; +"1225 gelu_7" -> "1226 dropout_30" [label="(1, 14, 14, 1536)", style=solid]; +"1226 dropout_30" -> "1227 dropout_30_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"1227 dropout_30_0_0_nncf_smooth_quant_0" -> "1228 quantize_per_tensor_default_50" [label="(1, 14, 14, 1536)", style=solid]; +"1228 quantize_per_tensor_default_50" -> "1229 dequantize_per_tensor_default_50" [label="(1, 14, 14, 1536)", style=solid]; +"1229 dequantize_per_tensor_default_50" -> "1234 linear_49" [label="(1, 14, 14, 1536)", style=solid]; +"1230 linear_49_scale_0" -> "1233 dequantize_per_channel_default_50" [label="(384,)", style=solid]; +"1231 linear_49_zero_point_0" -> "1233 dequantize_per_channel_default_50" [label="(384,)", style=solid]; +"1232 _frozen_param50" -> "1233 dequantize_per_channel_default_50" [label="(384, 1536)", style=solid]; +"1233 dequantize_per_channel_default_50" -> "1234 linear_49" [label="(384, 1536)", style=solid]; +"1234 linear_49" -> "1235 dropout_31" [label="(1, 14, 14, 384)", style=solid]; +"1235 dropout_31" -> "1236 layer_norm_18" [label="(1, 14, 14, 384)", style=solid]; +"1236 layer_norm_18" -> "1237 add_27" [label="(1, 14, 14, 384)", style=solid]; +"1237 add_27" -> "1259 pad_10" [label="(1, 14, 14, 384)", style=solid]; +"1237 add_27" -> "1318 add_29" [label="(1, 14, 14, 384)", style=solid]; +"1238 features_5_4_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "1243 linear_50" [label="(1, 15, 15, 2)", style=solid]; +"1239 linear_50_scale_0" -> "1242 dequantize_per_channel_default_51" [label="(512,)", style=solid]; +"1240 linear_50_zero_point_0" -> "1242 dequantize_per_channel_default_51" [label="(512,)", style=solid]; +"1241 _frozen_param51" -> "1242 dequantize_per_channel_default_51" [label="(512, 2)", style=solid]; +"1242 dequantize_per_channel_default_51" -> "1243 linear_50" [label="(512, 2)", style=solid]; +"1243 linear_50" -> "1244 relu__8" [label="(1, 15, 15, 512)", style=solid]; +"1244 relu__8" -> "1245 relu__8_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"1245 relu__8_0_0_nncf_smooth_quant_0" -> "1250 linear_51" [label="(1, 15, 15, 512)", style=solid]; +"1246 linear_51_scale_0" -> "1249 dequantize_per_channel_default_52" [label="(12,)", style=solid]; +"1247 linear_51_zero_point_0" -> "1249 dequantize_per_channel_default_52" [label="(12,)", style=solid]; +"1248 _frozen_param52" -> "1249 dequantize_per_channel_default_52" [label="(12, 512)", style=solid]; +"1249 dequantize_per_channel_default_52" -> "1250 linear_51" [label="(12, 512)", style=solid]; +"1250 linear_51" -> "1251 view_44" [label="(1, 15, 15, 12)", style=solid]; +"1251 view_44" -> "1252 index_8" [label="(225, 12)", style=solid]; +"1252 index_8" -> "1253 view_45" [label="(4096, 12)", style=solid]; +"1253 view_45" -> "1254 permute_37" [label="(64, 64, 12)", style=solid]; +"1254 permute_37" -> "1255 contiguous_14" [label="(12, 64, 64)", style=solid]; +"1255 contiguous_14" -> "1256 unsqueeze_24" [label="(12, 64, 64)", style=solid]; +"1256 unsqueeze_24" -> "1257 sigmoid_8" [label="(1, 12, 64, 64)", style=solid]; +"1257 sigmoid_8" -> "1258 mul_16" [label="(1, 12, 64, 64)", style=solid]; +"1258 mul_16" -> "1294 add_28" [label="(1, 12, 64, 64)", style=solid]; +"1259 pad_10" -> "1260 view_46" [label="(1, 16, 16, 384)", style=solid]; +"1260 view_46" -> "1261 permute_38" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1261 permute_38" -> "1262 reshape_36" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1262 reshape_36" -> "1264 reshape_36_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1263 clone_8" -> "1271 linear_52" [label="(1152,)", style=solid]; +"1264 reshape_36_0_0_nncf_smooth_quant_0" -> "1265 quantize_per_tensor_default_51" [label="(4, 64, 384)", style=solid]; +"1265 quantize_per_tensor_default_51" -> "1266 dequantize_per_tensor_default_51" [label="(4, 64, 384)", style=solid]; +"1266 dequantize_per_tensor_default_51" -> "1271 linear_52" [label="(4, 64, 384)", style=solid]; +"1267 linear_52_scale_0" -> "1270 dequantize_per_channel_default_53" [label="(1152,)", style=solid]; +"1268 linear_52_zero_point_0" -> "1270 dequantize_per_channel_default_53" [label="(1152,)", style=solid]; +"1269 _frozen_param53" -> "1270 dequantize_per_channel_default_53" [label="(1152, 384)", style=solid]; +"1270 dequantize_per_channel_default_53" -> "1271 linear_52" [label="(1152, 384)", style=solid]; +"1271 linear_52" -> "1272 reshape_37" [label="(4, 64, 1152)", style=solid]; +"1272 reshape_37" -> "1273 permute_39" [label="(4, 64, 3, 12, 32)", style=solid]; +"1273 permute_39" -> "1274 select_24" [label="(3, 4, 12, 64, 32)", style=solid]; +"1273 permute_39" -> "1275 select_25" [label="(3, 4, 12, 64, 32)", style=solid]; +"1273 permute_39" -> "1276 select_26" [label="(3, 4, 12, 64, 32)", style=solid]; +"1274 select_24" -> "1277 linalg_vector_norm_16" [label="(4, 12, 64, 32)", style=solid]; +"1274 select_24" -> "1279 expand_as_16" [label="(4, 12, 64, 32)", style=solid]; +"1274 select_24" -> "1280 div_16" [label="(4, 12, 64, 32)", style=solid]; +"1275 select_25" -> "1283 linalg_vector_norm_17" [label="(4, 12, 64, 32)", style=solid]; +"1275 select_25" -> "1285 expand_as_17" [label="(4, 12, 64, 32)", style=solid]; +"1275 select_25" -> "1286 div_17" [label="(4, 12, 64, 32)", style=solid]; +"1276 select_26" -> "1297 matmul_17" [label="(4, 12, 64, 32)", style=solid]; +"1277 linalg_vector_norm_16" -> "1278 clamp_min_16" [label="(4, 12, 64, 1)", style=solid]; +"1278 clamp_min_16" -> "1279 expand_as_16" [label="(4, 12, 64, 1)", style=solid]; +"1279 expand_as_16" -> "1280 div_16" [label="(4, 12, 64, 32)", style=solid]; +"1280 div_16" -> "1281 quantize_per_tensor_default_52" [label="(4, 12, 64, 32)", style=solid]; +"1281 quantize_per_tensor_default_52" -> "1282 dequantize_per_tensor_default_52" [label="(4, 12, 64, 32)", style=solid]; +"1282 dequantize_per_tensor_default_52" -> "1290 matmul_16" [label="(4, 12, 64, 32)", style=solid]; +"1283 linalg_vector_norm_17" -> "1284 clamp_min_17" [label="(4, 12, 64, 1)", style=solid]; +"1284 clamp_min_17" -> "1285 expand_as_17" [label="(4, 12, 64, 1)", style=solid]; +"1285 expand_as_17" -> "1286 div_17" [label="(4, 12, 64, 32)", style=solid]; +"1286 div_17" -> "1287 quantize_per_tensor_default_53" [label="(4, 12, 64, 32)", style=solid]; +"1287 quantize_per_tensor_default_53" -> "1288 dequantize_per_tensor_default_53" [label="(4, 12, 64, 32)", style=solid]; +"1288 dequantize_per_tensor_default_53" -> "1289 transpose_16" [label="(4, 12, 64, 32)", style=solid]; +"1289 transpose_16" -> "1290 matmul_16" [label="(4, 12, 32, 64)", style=solid]; +"1290 matmul_16" -> "1293 mul_17" [label="(4, 12, 64, 64)", style=solid]; +"1291 clamp_8" -> "1292 exp_8" [label="(12, 1, 1)", style=solid]; +"1292 exp_8" -> "1293 mul_17" [label="(12, 1, 1)", style=solid]; +"1293 mul_17" -> "1294 add_28" [label="(4, 12, 64, 64)", style=solid]; +"1294 add_28" -> "1295 softmax_8" [label="(4, 12, 64, 64)", style=solid]; +"1295 softmax_8" -> "1296 dropout_32" [label="(4, 12, 64, 64)", style=solid]; +"1296 dropout_32" -> "1297 matmul_17" [label="(4, 12, 64, 64)", style=solid]; +"1297 matmul_17" -> "1298 transpose_17" [label="(4, 12, 64, 32)", style=solid]; +"1298 transpose_17" -> "1299 reshape_38" [label="(4, 64, 12, 32)", style=solid]; +"1299 reshape_38" -> "1300 reshape_38_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1300 reshape_38_0_0_nncf_smooth_quant_0" -> "1301 quantize_per_tensor_default_54" [label="(4, 64, 384)", style=solid]; +"1301 quantize_per_tensor_default_54" -> "1302 dequantize_per_tensor_default_54" [label="(4, 64, 384)", style=solid]; +"1302 dequantize_per_tensor_default_54" -> "1307 linear_53" [label="(4, 64, 384)", style=solid]; +"1303 linear_53_scale_0" -> "1306 dequantize_per_channel_default_54" [label="(384,)", style=solid]; +"1304 linear_53_zero_point_0" -> "1306 dequantize_per_channel_default_54" [label="(384,)", style=solid]; +"1305 _frozen_param54" -> "1306 dequantize_per_channel_default_54" [label="(384, 384)", style=solid]; +"1306 dequantize_per_channel_default_54" -> "1307 linear_53" [label="(384, 384)", style=solid]; +"1307 linear_53" -> "1308 dropout_33" [label="(4, 64, 384)", style=solid]; +"1308 dropout_33" -> "1309 view_47" [label="(4, 64, 384)", style=solid]; +"1309 view_47" -> "1310 permute_40" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1310 permute_40" -> "1311 reshape_39" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1311 reshape_39" -> "1312 slice_134" [label="(1, 16, 16, 384)", style=solid]; +"1312 slice_134" -> "1313 slice_135" [label="(1, 16, 16, 384)", style=solid]; +"1313 slice_135" -> "1314 slice_136" [label="(1, 14, 16, 384)", style=solid]; +"1314 slice_136" -> "1315 slice_137" [label="(1, 14, 14, 384)", style=solid]; +"1315 slice_137" -> "1316 contiguous_15" [label="(1, 14, 14, 384)", style=solid]; +"1316 contiguous_15" -> "1317 layer_norm_19" [label="(1, 14, 14, 384)", style=solid]; +"1317 layer_norm_19" -> "1318 add_29" [label="(1, 14, 14, 384)", style=solid]; +"1318 add_29" -> "1319 add_29_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"1318 add_29" -> "1339 add_30" [label="(1, 14, 14, 384)", style=solid]; +"1319 add_29_0_0_nncf_smooth_quant_0" -> "1320 quantize_per_tensor_default_55" [label="(1, 14, 14, 384)", style=solid]; +"1320 quantize_per_tensor_default_55" -> "1321 dequantize_per_tensor_default_55" [label="(1, 14, 14, 384)", style=solid]; +"1321 dequantize_per_tensor_default_55" -> "1326 linear_54" [label="(1, 14, 14, 384)", style=solid]; +"1322 linear_54_scale_0" -> "1325 dequantize_per_channel_default_55" [label="(1536,)", style=solid]; +"1323 linear_54_zero_point_0" -> "1325 dequantize_per_channel_default_55" [label="(1536,)", style=solid]; +"1324 _frozen_param55" -> "1325 dequantize_per_channel_default_55" [label="(1536, 384)", style=solid]; +"1325 dequantize_per_channel_default_55" -> "1326 linear_54" [label="(1536, 384)", style=solid]; +"1326 linear_54" -> "1327 gelu_8" [label="(1, 14, 14, 1536)", style=solid]; +"1327 gelu_8" -> "1328 dropout_34" [label="(1, 14, 14, 1536)", style=solid]; +"1328 dropout_34" -> "1329 dropout_34_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"1329 dropout_34_0_0_nncf_smooth_quant_0" -> "1330 quantize_per_tensor_default_56" [label="(1, 14, 14, 1536)", style=solid]; +"1330 quantize_per_tensor_default_56" -> "1331 dequantize_per_tensor_default_56" [label="(1, 14, 14, 1536)", style=solid]; +"1331 dequantize_per_tensor_default_56" -> "1336 linear_55" [label="(1, 14, 14, 1536)", style=solid]; +"1332 linear_55_scale_0" -> "1335 dequantize_per_channel_default_56" [label="(384,)", style=solid]; +"1333 linear_55_zero_point_0" -> "1335 dequantize_per_channel_default_56" [label="(384,)", style=solid]; +"1334 _frozen_param56" -> "1335 dequantize_per_channel_default_56" [label="(384, 1536)", style=solid]; +"1335 dequantize_per_channel_default_56" -> "1336 linear_55" [label="(384, 1536)", style=solid]; +"1336 linear_55" -> "1337 dropout_35" [label="(1, 14, 14, 384)", style=solid]; +"1337 dropout_35" -> "1338 layer_norm_20" [label="(1, 14, 14, 384)", style=solid]; +"1338 layer_norm_20" -> "1339 add_30" [label="(1, 14, 14, 384)", style=solid]; +"1339 add_30" -> "1361 pad_11" [label="(1, 14, 14, 384)", style=solid]; +"1339 add_30" -> "1438 add_33" [label="(1, 14, 14, 384)", style=solid]; +"1340 features_5_5_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "1345 linear_56" [label="(1, 15, 15, 2)", style=solid]; +"1341 linear_56_scale_0" -> "1344 dequantize_per_channel_default_57" [label="(512,)", style=solid]; +"1342 linear_56_zero_point_0" -> "1344 dequantize_per_channel_default_57" [label="(512,)", style=solid]; +"1343 _frozen_param57" -> "1344 dequantize_per_channel_default_57" [label="(512, 2)", style=solid]; +"1344 dequantize_per_channel_default_57" -> "1345 linear_56" [label="(512, 2)", style=solid]; +"1345 linear_56" -> "1346 relu__9" [label="(1, 15, 15, 512)", style=solid]; +"1346 relu__9" -> "1347 relu__9_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"1347 relu__9_0_0_nncf_smooth_quant_0" -> "1352 linear_57" [label="(1, 15, 15, 512)", style=solid]; +"1348 linear_57_scale_0" -> "1351 dequantize_per_channel_default_58" [label="(12,)", style=solid]; +"1349 linear_57_zero_point_0" -> "1351 dequantize_per_channel_default_58" [label="(12,)", style=solid]; +"1350 _frozen_param58" -> "1351 dequantize_per_channel_default_58" [label="(12, 512)", style=solid]; +"1351 dequantize_per_channel_default_58" -> "1352 linear_57" [label="(12, 512)", style=solid]; +"1352 linear_57" -> "1353 view_48" [label="(1, 15, 15, 12)", style=solid]; +"1353 view_48" -> "1354 index_9" [label="(225, 12)", style=solid]; +"1354 index_9" -> "1355 view_49" [label="(4096, 12)", style=solid]; +"1355 view_49" -> "1356 permute_41" [label="(64, 64, 12)", style=solid]; +"1356 permute_41" -> "1357 contiguous_16" [label="(12, 64, 64)", style=solid]; +"1357 contiguous_16" -> "1358 unsqueeze_25" [label="(12, 64, 64)", style=solid]; +"1358 unsqueeze_25" -> "1359 sigmoid_9" [label="(1, 12, 64, 64)", style=solid]; +"1359 sigmoid_9" -> "1360 mul_18" [label="(1, 12, 64, 64)", style=solid]; +"1360 mul_18" -> "1397 add_31" [label="(1, 12, 64, 64)", style=solid]; +"1361 pad_11" -> "1362 roll_8" [label="(1, 16, 16, 384)", style=solid]; +"1362 roll_8" -> "1363 view_50" [label="(1, 16, 16, 384)", style=solid]; +"1363 view_50" -> "1364 permute_42" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1364 permute_42" -> "1365 reshape_40" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1365 reshape_40" -> "1367 reshape_40_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1365 reshape_40" -> "1398 new_zeros_4" [label="(4, 64, 384)", style=solid]; +"1366 clone_9" -> "1374 linear_58" [label="(1152,)", style=solid]; +"1367 reshape_40_0_0_nncf_smooth_quant_0" -> "1368 quantize_per_tensor_default_57" [label="(4, 64, 384)", style=solid]; +"1368 quantize_per_tensor_default_57" -> "1369 dequantize_per_tensor_default_57" [label="(4, 64, 384)", style=solid]; +"1369 dequantize_per_tensor_default_57" -> "1374 linear_58" [label="(4, 64, 384)", style=solid]; +"1370 linear_58_scale_0" -> "1373 dequantize_per_channel_default_59" [label="(1152,)", style=solid]; +"1371 linear_58_zero_point_0" -> "1373 dequantize_per_channel_default_59" [label="(1152,)", style=solid]; +"1372 _frozen_param59" -> "1373 dequantize_per_channel_default_59" [label="(1152, 384)", style=solid]; +"1373 dequantize_per_channel_default_59" -> "1374 linear_58" [label="(1152, 384)", style=solid]; +"1374 linear_58" -> "1375 reshape_41" [label="(4, 64, 1152)", style=solid]; +"1375 reshape_41" -> "1376 permute_43" [label="(4, 64, 3, 12, 32)", style=solid]; +"1376 permute_43" -> "1377 select_27" [label="(3, 4, 12, 64, 32)", style=solid]; +"1376 permute_43" -> "1378 select_28" [label="(3, 4, 12, 64, 32)", style=solid]; +"1376 permute_43" -> "1379 select_29" [label="(3, 4, 12, 64, 32)", style=solid]; +"1377 select_27" -> "1380 linalg_vector_norm_18" [label="(4, 12, 64, 32)", style=solid]; +"1377 select_27" -> "1382 expand_as_18" [label="(4, 12, 64, 32)", style=solid]; +"1377 select_27" -> "1383 div_18" [label="(4, 12, 64, 32)", style=solid]; +"1378 select_28" -> "1386 linalg_vector_norm_19" [label="(4, 12, 64, 32)", style=solid]; +"1378 select_28" -> "1388 expand_as_19" [label="(4, 12, 64, 32)", style=solid]; +"1378 select_28" -> "1389 div_19" [label="(4, 12, 64, 32)", style=solid]; +"1379 select_29" -> "1416 matmul_19" [label="(4, 12, 64, 32)", style=solid]; +"1380 linalg_vector_norm_18" -> "1381 clamp_min_18" [label="(4, 12, 64, 1)", style=solid]; +"1381 clamp_min_18" -> "1382 expand_as_18" [label="(4, 12, 64, 1)", style=solid]; +"1382 expand_as_18" -> "1383 div_18" [label="(4, 12, 64, 32)", style=solid]; +"1383 div_18" -> "1384 quantize_per_tensor_default_58" [label="(4, 12, 64, 32)", style=solid]; +"1384 quantize_per_tensor_default_58" -> "1385 dequantize_per_tensor_default_58" [label="(4, 12, 64, 32)", style=solid]; +"1385 dequantize_per_tensor_default_58" -> "1393 matmul_18" [label="(4, 12, 64, 32)", style=solid]; +"1386 linalg_vector_norm_19" -> "1387 clamp_min_19" [label="(4, 12, 64, 1)", style=solid]; +"1387 clamp_min_19" -> "1388 expand_as_19" [label="(4, 12, 64, 1)", style=solid]; +"1388 expand_as_19" -> "1389 div_19" [label="(4, 12, 64, 32)", style=solid]; +"1389 div_19" -> "1390 quantize_per_tensor_default_59" [label="(4, 12, 64, 32)", style=solid]; +"1390 quantize_per_tensor_default_59" -> "1391 dequantize_per_tensor_default_59" [label="(4, 12, 64, 32)", style=solid]; +"1391 dequantize_per_tensor_default_59" -> "1392 transpose_18" [label="(4, 12, 64, 32)", style=solid]; +"1392 transpose_18" -> "1393 matmul_18" [label="(4, 12, 32, 64)", style=solid]; +"1393 matmul_18" -> "1396 mul_19" [label="(4, 12, 64, 64)", style=solid]; +"1394 clamp_9" -> "1395 exp_9" [label="(12, 1, 1)", style=solid]; +"1395 exp_9" -> "1396 mul_19" [label="(12, 1, 1)", style=solid]; +"1396 mul_19" -> "1397 add_31" [label="(4, 12, 64, 64)", style=solid]; +"1397 add_31" -> "1409 view_52" [label="(4, 12, 64, 64)", style=solid]; +"1398 new_zeros_4" -> "1399 view_51" [label="(16, 16)", style=solid]; +"1399 view_51" -> "1400 permute_44" [label="(2, 8, 2, 8)", style=solid]; +"1400 permute_44" -> "1401 reshape_42" [label="(2, 2, 8, 8)", style=solid]; +"1401 reshape_42" -> "1402 unsqueeze_26" [label="(4, 64)", style=solid]; +"1401 reshape_42" -> "1403 unsqueeze_27" [label="(4, 64)", style=solid]; +"1402 unsqueeze_26" -> "1404 sub_4" [label="(4, 1, 64)", style=solid]; +"1403 unsqueeze_27" -> "1404 sub_4" [label="(4, 64, 1)", style=solid]; +"1404 sub_4" -> "1405 ne_4" [label="(4, 64, 64)", style=solid]; +"1404 sub_4" -> "1406 masked_fill_8" [label="(4, 64, 64)", style=solid]; +"1404 sub_4" -> "1407 eq_4" [label="(4, 64, 64)", style=solid]; +"1405 ne_4" -> "1406 masked_fill_8" [label="(4, 64, 64)", style=solid]; +"1406 masked_fill_8" -> "1408 masked_fill_9" [label="(4, 64, 64)", style=solid]; +"1407 eq_4" -> "1408 masked_fill_9" [label="(4, 64, 64)", style=solid]; +"1408 masked_fill_9" -> "1410 unsqueeze_28" [label="(4, 64, 64)", style=solid]; +"1409 view_52" -> "1412 add_32" [label="(1, 4, 12, 64, 64)", style=solid]; +"1410 unsqueeze_28" -> "1411 unsqueeze_29" [label="(4, 1, 64, 64)", style=solid]; +"1411 unsqueeze_29" -> "1412 add_32" [label="(1, 4, 1, 64, 64)", style=solid]; +"1412 add_32" -> "1413 view_53" [label="(1, 4, 12, 64, 64)", style=solid]; +"1413 view_53" -> "1414 softmax_9" [label="(4, 12, 64, 64)", style=solid]; +"1414 softmax_9" -> "1415 dropout_36" [label="(4, 12, 64, 64)", style=solid]; +"1415 dropout_36" -> "1416 matmul_19" [label="(4, 12, 64, 64)", style=solid]; +"1416 matmul_19" -> "1417 transpose_19" [label="(4, 12, 64, 32)", style=solid]; +"1417 transpose_19" -> "1418 reshape_43" [label="(4, 64, 12, 32)", style=solid]; +"1418 reshape_43" -> "1419 reshape_43_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1419 reshape_43_0_0_nncf_smooth_quant_0" -> "1420 quantize_per_tensor_default_60" [label="(4, 64, 384)", style=solid]; +"1420 quantize_per_tensor_default_60" -> "1421 dequantize_per_tensor_default_60" [label="(4, 64, 384)", style=solid]; +"1421 dequantize_per_tensor_default_60" -> "1426 linear_59" [label="(4, 64, 384)", style=solid]; +"1422 linear_59_scale_0" -> "1425 dequantize_per_channel_default_60" [label="(384,)", style=solid]; +"1423 linear_59_zero_point_0" -> "1425 dequantize_per_channel_default_60" [label="(384,)", style=solid]; +"1424 _frozen_param60" -> "1425 dequantize_per_channel_default_60" [label="(384, 384)", style=solid]; +"1425 dequantize_per_channel_default_60" -> "1426 linear_59" [label="(384, 384)", style=solid]; +"1426 linear_59" -> "1427 dropout_37" [label="(4, 64, 384)", style=solid]; +"1427 dropout_37" -> "1428 view_54" [label="(4, 64, 384)", style=solid]; +"1428 view_54" -> "1429 permute_45" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1429 permute_45" -> "1430 reshape_44" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1430 reshape_44" -> "1431 roll_9" [label="(1, 16, 16, 384)", style=solid]; +"1431 roll_9" -> "1432 slice_157" [label="(1, 16, 16, 384)", style=solid]; +"1432 slice_157" -> "1433 slice_158" [label="(1, 16, 16, 384)", style=solid]; +"1433 slice_158" -> "1434 slice_159" [label="(1, 14, 16, 384)", style=solid]; +"1434 slice_159" -> "1435 slice_160" [label="(1, 14, 14, 384)", style=solid]; +"1435 slice_160" -> "1436 contiguous_17" [label="(1, 14, 14, 384)", style=solid]; +"1436 contiguous_17" -> "1437 layer_norm_21" [label="(1, 14, 14, 384)", style=solid]; +"1437 layer_norm_21" -> "1438 add_33" [label="(1, 14, 14, 384)", style=solid]; +"1438 add_33" -> "1439 add_33_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"1438 add_33" -> "1459 add_34" [label="(1, 14, 14, 384)", style=solid]; +"1439 add_33_0_0_nncf_smooth_quant_0" -> "1440 quantize_per_tensor_default_61" [label="(1, 14, 14, 384)", style=solid]; +"1440 quantize_per_tensor_default_61" -> "1441 dequantize_per_tensor_default_61" [label="(1, 14, 14, 384)", style=solid]; +"1441 dequantize_per_tensor_default_61" -> "1446 linear_60" [label="(1, 14, 14, 384)", style=solid]; +"1442 linear_60_scale_0" -> "1445 dequantize_per_channel_default_61" [label="(1536,)", style=solid]; +"1443 linear_60_zero_point_0" -> "1445 dequantize_per_channel_default_61" [label="(1536,)", style=solid]; +"1444 _frozen_param61" -> "1445 dequantize_per_channel_default_61" [label="(1536, 384)", style=solid]; +"1445 dequantize_per_channel_default_61" -> "1446 linear_60" [label="(1536, 384)", style=solid]; +"1446 linear_60" -> "1447 gelu_9" [label="(1, 14, 14, 1536)", style=solid]; +"1447 gelu_9" -> "1448 dropout_38" [label="(1, 14, 14, 1536)", style=solid]; +"1448 dropout_38" -> "1449 dropout_38_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"1449 dropout_38_0_0_nncf_smooth_quant_0" -> "1450 quantize_per_tensor_default_62" [label="(1, 14, 14, 1536)", style=solid]; +"1450 quantize_per_tensor_default_62" -> "1451 dequantize_per_tensor_default_62" [label="(1, 14, 14, 1536)", style=solid]; +"1451 dequantize_per_tensor_default_62" -> "1456 linear_61" [label="(1, 14, 14, 1536)", style=solid]; +"1452 linear_61_scale_0" -> "1455 dequantize_per_channel_default_62" [label="(384,)", style=solid]; +"1453 linear_61_zero_point_0" -> "1455 dequantize_per_channel_default_62" [label="(384,)", style=solid]; +"1454 _frozen_param62" -> "1455 dequantize_per_channel_default_62" [label="(384, 1536)", style=solid]; +"1455 dequantize_per_channel_default_62" -> "1456 linear_61" [label="(384, 1536)", style=solid]; +"1456 linear_61" -> "1457 dropout_39" [label="(1, 14, 14, 384)", style=solid]; +"1457 dropout_39" -> "1458 layer_norm_22" [label="(1, 14, 14, 384)", style=solid]; +"1458 layer_norm_22" -> "1459 add_34" [label="(1, 14, 14, 384)", style=solid]; +"1459 add_34" -> "1481 pad_12" [label="(1, 14, 14, 384)", style=solid]; +"1459 add_34" -> "1540 add_36" [label="(1, 14, 14, 384)", style=solid]; +"1460 features_5_6_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "1465 linear_62" [label="(1, 15, 15, 2)", style=solid]; +"1461 linear_62_scale_0" -> "1464 dequantize_per_channel_default_63" [label="(512,)", style=solid]; +"1462 linear_62_zero_point_0" -> "1464 dequantize_per_channel_default_63" [label="(512,)", style=solid]; +"1463 _frozen_param63" -> "1464 dequantize_per_channel_default_63" [label="(512, 2)", style=solid]; +"1464 dequantize_per_channel_default_63" -> "1465 linear_62" [label="(512, 2)", style=solid]; +"1465 linear_62" -> "1466 relu__10" [label="(1, 15, 15, 512)", style=solid]; +"1466 relu__10" -> "1467 relu__10_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"1467 relu__10_0_0_nncf_smooth_quant_0" -> "1472 linear_63" [label="(1, 15, 15, 512)", style=solid]; +"1468 linear_63_scale_0" -> "1471 dequantize_per_channel_default_64" [label="(12,)", style=solid]; +"1469 linear_63_zero_point_0" -> "1471 dequantize_per_channel_default_64" [label="(12,)", style=solid]; +"1470 _frozen_param64" -> "1471 dequantize_per_channel_default_64" [label="(12, 512)", style=solid]; +"1471 dequantize_per_channel_default_64" -> "1472 linear_63" [label="(12, 512)", style=solid]; +"1472 linear_63" -> "1473 view_55" [label="(1, 15, 15, 12)", style=solid]; +"1473 view_55" -> "1474 index_10" [label="(225, 12)", style=solid]; +"1474 index_10" -> "1475 view_56" [label="(4096, 12)", style=solid]; +"1475 view_56" -> "1476 permute_46" [label="(64, 64, 12)", style=solid]; +"1476 permute_46" -> "1477 contiguous_18" [label="(12, 64, 64)", style=solid]; +"1477 contiguous_18" -> "1478 unsqueeze_30" [label="(12, 64, 64)", style=solid]; +"1478 unsqueeze_30" -> "1479 sigmoid_10" [label="(1, 12, 64, 64)", style=solid]; +"1479 sigmoid_10" -> "1480 mul_20" [label="(1, 12, 64, 64)", style=solid]; +"1480 mul_20" -> "1516 add_35" [label="(1, 12, 64, 64)", style=solid]; +"1481 pad_12" -> "1482 view_57" [label="(1, 16, 16, 384)", style=solid]; +"1482 view_57" -> "1483 permute_47" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1483 permute_47" -> "1484 reshape_45" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1484 reshape_45" -> "1486 reshape_45_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1485 clone_10" -> "1493 linear_64" [label="(1152,)", style=solid]; +"1486 reshape_45_0_0_nncf_smooth_quant_0" -> "1487 quantize_per_tensor_default_63" [label="(4, 64, 384)", style=solid]; +"1487 quantize_per_tensor_default_63" -> "1488 dequantize_per_tensor_default_63" [label="(4, 64, 384)", style=solid]; +"1488 dequantize_per_tensor_default_63" -> "1493 linear_64" [label="(4, 64, 384)", style=solid]; +"1489 linear_64_scale_0" -> "1492 dequantize_per_channel_default_65" [label="(1152,)", style=solid]; +"1490 linear_64_zero_point_0" -> "1492 dequantize_per_channel_default_65" [label="(1152,)", style=solid]; +"1491 _frozen_param65" -> "1492 dequantize_per_channel_default_65" [label="(1152, 384)", style=solid]; +"1492 dequantize_per_channel_default_65" -> "1493 linear_64" [label="(1152, 384)", style=solid]; +"1493 linear_64" -> "1494 reshape_46" [label="(4, 64, 1152)", style=solid]; +"1494 reshape_46" -> "1495 permute_48" [label="(4, 64, 3, 12, 32)", style=solid]; +"1495 permute_48" -> "1496 select_30" [label="(3, 4, 12, 64, 32)", style=solid]; +"1495 permute_48" -> "1497 select_31" [label="(3, 4, 12, 64, 32)", style=solid]; +"1495 permute_48" -> "1498 select_32" [label="(3, 4, 12, 64, 32)", style=solid]; +"1496 select_30" -> "1499 linalg_vector_norm_20" [label="(4, 12, 64, 32)", style=solid]; +"1496 select_30" -> "1501 expand_as_20" [label="(4, 12, 64, 32)", style=solid]; +"1496 select_30" -> "1502 div_20" [label="(4, 12, 64, 32)", style=solid]; +"1497 select_31" -> "1505 linalg_vector_norm_21" [label="(4, 12, 64, 32)", style=solid]; +"1497 select_31" -> "1507 expand_as_21" [label="(4, 12, 64, 32)", style=solid]; +"1497 select_31" -> "1508 div_21" [label="(4, 12, 64, 32)", style=solid]; +"1498 select_32" -> "1519 matmul_21" [label="(4, 12, 64, 32)", style=solid]; +"1499 linalg_vector_norm_20" -> "1500 clamp_min_20" [label="(4, 12, 64, 1)", style=solid]; +"1500 clamp_min_20" -> "1501 expand_as_20" [label="(4, 12, 64, 1)", style=solid]; +"1501 expand_as_20" -> "1502 div_20" [label="(4, 12, 64, 32)", style=solid]; +"1502 div_20" -> "1503 quantize_per_tensor_default_64" [label="(4, 12, 64, 32)", style=solid]; +"1503 quantize_per_tensor_default_64" -> "1504 dequantize_per_tensor_default_64" [label="(4, 12, 64, 32)", style=solid]; +"1504 dequantize_per_tensor_default_64" -> "1512 matmul_20" [label="(4, 12, 64, 32)", style=solid]; +"1505 linalg_vector_norm_21" -> "1506 clamp_min_21" [label="(4, 12, 64, 1)", style=solid]; +"1506 clamp_min_21" -> "1507 expand_as_21" [label="(4, 12, 64, 1)", style=solid]; +"1507 expand_as_21" -> "1508 div_21" [label="(4, 12, 64, 32)", style=solid]; +"1508 div_21" -> "1509 quantize_per_tensor_default_65" [label="(4, 12, 64, 32)", style=solid]; +"1509 quantize_per_tensor_default_65" -> "1510 dequantize_per_tensor_default_65" [label="(4, 12, 64, 32)", style=solid]; +"1510 dequantize_per_tensor_default_65" -> "1511 transpose_20" [label="(4, 12, 64, 32)", style=solid]; +"1511 transpose_20" -> "1512 matmul_20" [label="(4, 12, 32, 64)", style=solid]; +"1512 matmul_20" -> "1515 mul_21" [label="(4, 12, 64, 64)", style=solid]; +"1513 clamp_10" -> "1514 exp_10" [label="(12, 1, 1)", style=solid]; +"1514 exp_10" -> "1515 mul_21" [label="(12, 1, 1)", style=solid]; +"1515 mul_21" -> "1516 add_35" [label="(4, 12, 64, 64)", style=solid]; +"1516 add_35" -> "1517 softmax_10" [label="(4, 12, 64, 64)", style=solid]; +"1517 softmax_10" -> "1518 dropout_40" [label="(4, 12, 64, 64)", style=solid]; +"1518 dropout_40" -> "1519 matmul_21" [label="(4, 12, 64, 64)", style=solid]; +"1519 matmul_21" -> "1520 transpose_21" [label="(4, 12, 64, 32)", style=solid]; +"1520 transpose_21" -> "1521 reshape_47" [label="(4, 64, 12, 32)", style=solid]; +"1521 reshape_47" -> "1522 reshape_47_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1522 reshape_47_0_0_nncf_smooth_quant_0" -> "1523 quantize_per_tensor_default_66" [label="(4, 64, 384)", style=solid]; +"1523 quantize_per_tensor_default_66" -> "1524 dequantize_per_tensor_default_66" [label="(4, 64, 384)", style=solid]; +"1524 dequantize_per_tensor_default_66" -> "1529 linear_65" [label="(4, 64, 384)", style=solid]; +"1525 linear_65_scale_0" -> "1528 dequantize_per_channel_default_66" [label="(384,)", style=solid]; +"1526 linear_65_zero_point_0" -> "1528 dequantize_per_channel_default_66" [label="(384,)", style=solid]; +"1527 _frozen_param66" -> "1528 dequantize_per_channel_default_66" [label="(384, 384)", style=solid]; +"1528 dequantize_per_channel_default_66" -> "1529 linear_65" [label="(384, 384)", style=solid]; +"1529 linear_65" -> "1530 dropout_41" [label="(4, 64, 384)", style=solid]; +"1530 dropout_41" -> "1531 view_58" [label="(4, 64, 384)", style=solid]; +"1531 view_58" -> "1532 permute_49" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1532 permute_49" -> "1533 reshape_48" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1533 reshape_48" -> "1534 slice_162" [label="(1, 16, 16, 384)", style=solid]; +"1534 slice_162" -> "1535 slice_163" [label="(1, 16, 16, 384)", style=solid]; +"1535 slice_163" -> "1536 slice_164" [label="(1, 14, 16, 384)", style=solid]; +"1536 slice_164" -> "1537 slice_165" [label="(1, 14, 14, 384)", style=solid]; +"1537 slice_165" -> "1538 contiguous_19" [label="(1, 14, 14, 384)", style=solid]; +"1538 contiguous_19" -> "1539 layer_norm_23" [label="(1, 14, 14, 384)", style=solid]; +"1539 layer_norm_23" -> "1540 add_36" [label="(1, 14, 14, 384)", style=solid]; +"1540 add_36" -> "1541 add_36_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"1540 add_36" -> "1561 add_37" [label="(1, 14, 14, 384)", style=solid]; +"1541 add_36_0_0_nncf_smooth_quant_0" -> "1542 quantize_per_tensor_default_67" [label="(1, 14, 14, 384)", style=solid]; +"1542 quantize_per_tensor_default_67" -> "1543 dequantize_per_tensor_default_67" [label="(1, 14, 14, 384)", style=solid]; +"1543 dequantize_per_tensor_default_67" -> "1548 linear_66" [label="(1, 14, 14, 384)", style=solid]; +"1544 linear_66_scale_0" -> "1547 dequantize_per_channel_default_67" [label="(1536,)", style=solid]; +"1545 linear_66_zero_point_0" -> "1547 dequantize_per_channel_default_67" [label="(1536,)", style=solid]; +"1546 _frozen_param67" -> "1547 dequantize_per_channel_default_67" [label="(1536, 384)", style=solid]; +"1547 dequantize_per_channel_default_67" -> "1548 linear_66" [label="(1536, 384)", style=solid]; +"1548 linear_66" -> "1549 gelu_10" [label="(1, 14, 14, 1536)", style=solid]; +"1549 gelu_10" -> "1550 dropout_42" [label="(1, 14, 14, 1536)", style=solid]; +"1550 dropout_42" -> "1551 dropout_42_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"1551 dropout_42_0_0_nncf_smooth_quant_0" -> "1552 quantize_per_tensor_default_68" [label="(1, 14, 14, 1536)", style=solid]; +"1552 quantize_per_tensor_default_68" -> "1553 dequantize_per_tensor_default_68" [label="(1, 14, 14, 1536)", style=solid]; +"1553 dequantize_per_tensor_default_68" -> "1558 linear_67" [label="(1, 14, 14, 1536)", style=solid]; +"1554 linear_67_scale_0" -> "1557 dequantize_per_channel_default_68" [label="(384,)", style=solid]; +"1555 linear_67_zero_point_0" -> "1557 dequantize_per_channel_default_68" [label="(384,)", style=solid]; +"1556 _frozen_param68" -> "1557 dequantize_per_channel_default_68" [label="(384, 1536)", style=solid]; +"1557 dequantize_per_channel_default_68" -> "1558 linear_67" [label="(384, 1536)", style=solid]; +"1558 linear_67" -> "1559 dropout_43" [label="(1, 14, 14, 384)", style=solid]; +"1559 dropout_43" -> "1560 layer_norm_24" [label="(1, 14, 14, 384)", style=solid]; +"1560 layer_norm_24" -> "1561 add_37" [label="(1, 14, 14, 384)", style=solid]; +"1561 add_37" -> "1583 pad_13" [label="(1, 14, 14, 384)", style=solid]; +"1561 add_37" -> "1660 add_40" [label="(1, 14, 14, 384)", style=solid]; +"1562 features_5_7_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "1567 linear_68" [label="(1, 15, 15, 2)", style=solid]; +"1563 linear_68_scale_0" -> "1566 dequantize_per_channel_default_69" [label="(512,)", style=solid]; +"1564 linear_68_zero_point_0" -> "1566 dequantize_per_channel_default_69" [label="(512,)", style=solid]; +"1565 _frozen_param69" -> "1566 dequantize_per_channel_default_69" [label="(512, 2)", style=solid]; +"1566 dequantize_per_channel_default_69" -> "1567 linear_68" [label="(512, 2)", style=solid]; +"1567 linear_68" -> "1568 relu__11" [label="(1, 15, 15, 512)", style=solid]; +"1568 relu__11" -> "1569 relu__11_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"1569 relu__11_0_0_nncf_smooth_quant_0" -> "1574 linear_69" [label="(1, 15, 15, 512)", style=solid]; +"1570 linear_69_scale_0" -> "1573 dequantize_per_channel_default_70" [label="(12,)", style=solid]; +"1571 linear_69_zero_point_0" -> "1573 dequantize_per_channel_default_70" [label="(12,)", style=solid]; +"1572 _frozen_param70" -> "1573 dequantize_per_channel_default_70" [label="(12, 512)", style=solid]; +"1573 dequantize_per_channel_default_70" -> "1574 linear_69" [label="(12, 512)", style=solid]; +"1574 linear_69" -> "1575 view_59" [label="(1, 15, 15, 12)", style=solid]; +"1575 view_59" -> "1576 index_11" [label="(225, 12)", style=solid]; +"1576 index_11" -> "1577 view_60" [label="(4096, 12)", style=solid]; +"1577 view_60" -> "1578 permute_50" [label="(64, 64, 12)", style=solid]; +"1578 permute_50" -> "1579 contiguous_20" [label="(12, 64, 64)", style=solid]; +"1579 contiguous_20" -> "1580 unsqueeze_31" [label="(12, 64, 64)", style=solid]; +"1580 unsqueeze_31" -> "1581 sigmoid_11" [label="(1, 12, 64, 64)", style=solid]; +"1581 sigmoid_11" -> "1582 mul_22" [label="(1, 12, 64, 64)", style=solid]; +"1582 mul_22" -> "1619 add_38" [label="(1, 12, 64, 64)", style=solid]; +"1583 pad_13" -> "1584 roll_10" [label="(1, 16, 16, 384)", style=solid]; +"1584 roll_10" -> "1585 view_61" [label="(1, 16, 16, 384)", style=solid]; +"1585 view_61" -> "1586 permute_51" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1586 permute_51" -> "1587 reshape_49" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1587 reshape_49" -> "1589 reshape_49_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1587 reshape_49" -> "1620 new_zeros_5" [label="(4, 64, 384)", style=solid]; +"1588 clone_11" -> "1596 linear_70" [label="(1152,)", style=solid]; +"1589 reshape_49_0_0_nncf_smooth_quant_0" -> "1590 quantize_per_tensor_default_69" [label="(4, 64, 384)", style=solid]; +"1590 quantize_per_tensor_default_69" -> "1591 dequantize_per_tensor_default_69" [label="(4, 64, 384)", style=solid]; +"1591 dequantize_per_tensor_default_69" -> "1596 linear_70" [label="(4, 64, 384)", style=solid]; +"1592 linear_70_scale_0" -> "1595 dequantize_per_channel_default_71" [label="(1152,)", style=solid]; +"1593 linear_70_zero_point_0" -> "1595 dequantize_per_channel_default_71" [label="(1152,)", style=solid]; +"1594 _frozen_param71" -> "1595 dequantize_per_channel_default_71" [label="(1152, 384)", style=solid]; +"1595 dequantize_per_channel_default_71" -> "1596 linear_70" [label="(1152, 384)", style=solid]; +"1596 linear_70" -> "1597 reshape_50" [label="(4, 64, 1152)", style=solid]; +"1597 reshape_50" -> "1598 permute_52" [label="(4, 64, 3, 12, 32)", style=solid]; +"1598 permute_52" -> "1599 select_33" [label="(3, 4, 12, 64, 32)", style=solid]; +"1598 permute_52" -> "1600 select_34" [label="(3, 4, 12, 64, 32)", style=solid]; +"1598 permute_52" -> "1601 select_35" [label="(3, 4, 12, 64, 32)", style=solid]; +"1599 select_33" -> "1602 linalg_vector_norm_22" [label="(4, 12, 64, 32)", style=solid]; +"1599 select_33" -> "1604 expand_as_22" [label="(4, 12, 64, 32)", style=solid]; +"1599 select_33" -> "1605 div_22" [label="(4, 12, 64, 32)", style=solid]; +"1600 select_34" -> "1608 linalg_vector_norm_23" [label="(4, 12, 64, 32)", style=solid]; +"1600 select_34" -> "1610 expand_as_23" [label="(4, 12, 64, 32)", style=solid]; +"1600 select_34" -> "1611 div_23" [label="(4, 12, 64, 32)", style=solid]; +"1601 select_35" -> "1638 matmul_23" [label="(4, 12, 64, 32)", style=solid]; +"1602 linalg_vector_norm_22" -> "1603 clamp_min_22" [label="(4, 12, 64, 1)", style=solid]; +"1603 clamp_min_22" -> "1604 expand_as_22" [label="(4, 12, 64, 1)", style=solid]; +"1604 expand_as_22" -> "1605 div_22" [label="(4, 12, 64, 32)", style=solid]; +"1605 div_22" -> "1606 quantize_per_tensor_default_70" [label="(4, 12, 64, 32)", style=solid]; +"1606 quantize_per_tensor_default_70" -> "1607 dequantize_per_tensor_default_70" [label="(4, 12, 64, 32)", style=solid]; +"1607 dequantize_per_tensor_default_70" -> "1615 matmul_22" [label="(4, 12, 64, 32)", style=solid]; +"1608 linalg_vector_norm_23" -> "1609 clamp_min_23" [label="(4, 12, 64, 1)", style=solid]; +"1609 clamp_min_23" -> "1610 expand_as_23" [label="(4, 12, 64, 1)", style=solid]; +"1610 expand_as_23" -> "1611 div_23" [label="(4, 12, 64, 32)", style=solid]; +"1611 div_23" -> "1612 quantize_per_tensor_default_71" [label="(4, 12, 64, 32)", style=solid]; +"1612 quantize_per_tensor_default_71" -> "1613 dequantize_per_tensor_default_71" [label="(4, 12, 64, 32)", style=solid]; +"1613 dequantize_per_tensor_default_71" -> "1614 transpose_22" [label="(4, 12, 64, 32)", style=solid]; +"1614 transpose_22" -> "1615 matmul_22" [label="(4, 12, 32, 64)", style=solid]; +"1615 matmul_22" -> "1618 mul_23" [label="(4, 12, 64, 64)", style=solid]; +"1616 clamp_11" -> "1617 exp_11" [label="(12, 1, 1)", style=solid]; +"1617 exp_11" -> "1618 mul_23" [label="(12, 1, 1)", style=solid]; +"1618 mul_23" -> "1619 add_38" [label="(4, 12, 64, 64)", style=solid]; +"1619 add_38" -> "1631 view_63" [label="(4, 12, 64, 64)", style=solid]; +"1620 new_zeros_5" -> "1621 view_62" [label="(16, 16)", style=solid]; +"1621 view_62" -> "1622 permute_53" [label="(2, 8, 2, 8)", style=solid]; +"1622 permute_53" -> "1623 reshape_51" [label="(2, 2, 8, 8)", style=solid]; +"1623 reshape_51" -> "1624 unsqueeze_32" [label="(4, 64)", style=solid]; +"1623 reshape_51" -> "1625 unsqueeze_33" [label="(4, 64)", style=solid]; +"1624 unsqueeze_32" -> "1626 sub_5" [label="(4, 1, 64)", style=solid]; +"1625 unsqueeze_33" -> "1626 sub_5" [label="(4, 64, 1)", style=solid]; +"1626 sub_5" -> "1627 ne_5" [label="(4, 64, 64)", style=solid]; +"1626 sub_5" -> "1628 masked_fill_10" [label="(4, 64, 64)", style=solid]; +"1626 sub_5" -> "1629 eq_5" [label="(4, 64, 64)", style=solid]; +"1627 ne_5" -> "1628 masked_fill_10" [label="(4, 64, 64)", style=solid]; +"1628 masked_fill_10" -> "1630 masked_fill_11" [label="(4, 64, 64)", style=solid]; +"1629 eq_5" -> "1630 masked_fill_11" [label="(4, 64, 64)", style=solid]; +"1630 masked_fill_11" -> "1632 unsqueeze_34" [label="(4, 64, 64)", style=solid]; +"1631 view_63" -> "1634 add_39" [label="(1, 4, 12, 64, 64)", style=solid]; +"1632 unsqueeze_34" -> "1633 unsqueeze_35" [label="(4, 1, 64, 64)", style=solid]; +"1633 unsqueeze_35" -> "1634 add_39" [label="(1, 4, 1, 64, 64)", style=solid]; +"1634 add_39" -> "1635 view_64" [label="(1, 4, 12, 64, 64)", style=solid]; +"1635 view_64" -> "1636 softmax_11" [label="(4, 12, 64, 64)", style=solid]; +"1636 softmax_11" -> "1637 dropout_44" [label="(4, 12, 64, 64)", style=solid]; +"1637 dropout_44" -> "1638 matmul_23" [label="(4, 12, 64, 64)", style=solid]; +"1638 matmul_23" -> "1639 transpose_23" [label="(4, 12, 64, 32)", style=solid]; +"1639 transpose_23" -> "1640 reshape_52" [label="(4, 64, 12, 32)", style=solid]; +"1640 reshape_52" -> "1641 reshape_52_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1641 reshape_52_0_0_nncf_smooth_quant_0" -> "1642 quantize_per_tensor_default_72" [label="(4, 64, 384)", style=solid]; +"1642 quantize_per_tensor_default_72" -> "1643 dequantize_per_tensor_default_72" [label="(4, 64, 384)", style=solid]; +"1643 dequantize_per_tensor_default_72" -> "1648 linear_71" [label="(4, 64, 384)", style=solid]; +"1644 linear_71_scale_0" -> "1647 dequantize_per_channel_default_72" [label="(384,)", style=solid]; +"1645 linear_71_zero_point_0" -> "1647 dequantize_per_channel_default_72" [label="(384,)", style=solid]; +"1646 _frozen_param72" -> "1647 dequantize_per_channel_default_72" [label="(384, 384)", style=solid]; +"1647 dequantize_per_channel_default_72" -> "1648 linear_71" [label="(384, 384)", style=solid]; +"1648 linear_71" -> "1649 dropout_45" [label="(4, 64, 384)", style=solid]; +"1649 dropout_45" -> "1650 view_65" [label="(4, 64, 384)", style=solid]; +"1650 view_65" -> "1651 permute_54" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1651 permute_54" -> "1652 reshape_53" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1652 reshape_53" -> "1653 roll_11" [label="(1, 16, 16, 384)", style=solid]; +"1653 roll_11" -> "1654 slice_185" [label="(1, 16, 16, 384)", style=solid]; +"1654 slice_185" -> "1655 slice_186" [label="(1, 16, 16, 384)", style=solid]; +"1655 slice_186" -> "1656 slice_187" [label="(1, 14, 16, 384)", style=solid]; +"1656 slice_187" -> "1657 slice_188" [label="(1, 14, 14, 384)", style=solid]; +"1657 slice_188" -> "1658 contiguous_21" [label="(1, 14, 14, 384)", style=solid]; +"1658 contiguous_21" -> "1659 layer_norm_25" [label="(1, 14, 14, 384)", style=solid]; +"1659 layer_norm_25" -> "1660 add_40" [label="(1, 14, 14, 384)", style=solid]; +"1660 add_40" -> "1661 add_40_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"1660 add_40" -> "1681 add_41" [label="(1, 14, 14, 384)", style=solid]; +"1661 add_40_0_0_nncf_smooth_quant_0" -> "1662 quantize_per_tensor_default_73" [label="(1, 14, 14, 384)", style=solid]; +"1662 quantize_per_tensor_default_73" -> "1663 dequantize_per_tensor_default_73" [label="(1, 14, 14, 384)", style=solid]; +"1663 dequantize_per_tensor_default_73" -> "1668 linear_72" [label="(1, 14, 14, 384)", style=solid]; +"1664 linear_72_scale_0" -> "1667 dequantize_per_channel_default_73" [label="(1536,)", style=solid]; +"1665 linear_72_zero_point_0" -> "1667 dequantize_per_channel_default_73" [label="(1536,)", style=solid]; +"1666 _frozen_param73" -> "1667 dequantize_per_channel_default_73" [label="(1536, 384)", style=solid]; +"1667 dequantize_per_channel_default_73" -> "1668 linear_72" [label="(1536, 384)", style=solid]; +"1668 linear_72" -> "1669 gelu_11" [label="(1, 14, 14, 1536)", style=solid]; +"1669 gelu_11" -> "1670 dropout_46" [label="(1, 14, 14, 1536)", style=solid]; +"1670 dropout_46" -> "1671 dropout_46_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"1671 dropout_46_0_0_nncf_smooth_quant_0" -> "1672 quantize_per_tensor_default_74" [label="(1, 14, 14, 1536)", style=solid]; +"1672 quantize_per_tensor_default_74" -> "1673 dequantize_per_tensor_default_74" [label="(1, 14, 14, 1536)", style=solid]; +"1673 dequantize_per_tensor_default_74" -> "1678 linear_73" [label="(1, 14, 14, 1536)", style=solid]; +"1674 linear_73_scale_0" -> "1677 dequantize_per_channel_default_74" [label="(384,)", style=solid]; +"1675 linear_73_zero_point_0" -> "1677 dequantize_per_channel_default_74" [label="(384,)", style=solid]; +"1676 _frozen_param74" -> "1677 dequantize_per_channel_default_74" [label="(384, 1536)", style=solid]; +"1677 dequantize_per_channel_default_74" -> "1678 linear_73" [label="(384, 1536)", style=solid]; +"1678 linear_73" -> "1679 dropout_47" [label="(1, 14, 14, 384)", style=solid]; +"1679 dropout_47" -> "1680 layer_norm_26" [label="(1, 14, 14, 384)", style=solid]; +"1680 layer_norm_26" -> "1681 add_41" [label="(1, 14, 14, 384)", style=solid]; +"1681 add_41" -> "1703 pad_14" [label="(1, 14, 14, 384)", style=solid]; +"1681 add_41" -> "1762 add_43" [label="(1, 14, 14, 384)", style=solid]; +"1682 features_5_8_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "1687 linear_74" [label="(1, 15, 15, 2)", style=solid]; +"1683 linear_74_scale_0" -> "1686 dequantize_per_channel_default_75" [label="(512,)", style=solid]; +"1684 linear_74_zero_point_0" -> "1686 dequantize_per_channel_default_75" [label="(512,)", style=solid]; +"1685 _frozen_param75" -> "1686 dequantize_per_channel_default_75" [label="(512, 2)", style=solid]; +"1686 dequantize_per_channel_default_75" -> "1687 linear_74" [label="(512, 2)", style=solid]; +"1687 linear_74" -> "1688 relu__12" [label="(1, 15, 15, 512)", style=solid]; +"1688 relu__12" -> "1689 relu__12_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"1689 relu__12_0_0_nncf_smooth_quant_0" -> "1694 linear_75" [label="(1, 15, 15, 512)", style=solid]; +"1690 linear_75_scale_0" -> "1693 dequantize_per_channel_default_76" [label="(12,)", style=solid]; +"1691 linear_75_zero_point_0" -> "1693 dequantize_per_channel_default_76" [label="(12,)", style=solid]; +"1692 _frozen_param76" -> "1693 dequantize_per_channel_default_76" [label="(12, 512)", style=solid]; +"1693 dequantize_per_channel_default_76" -> "1694 linear_75" [label="(12, 512)", style=solid]; +"1694 linear_75" -> "1695 view_66" [label="(1, 15, 15, 12)", style=solid]; +"1695 view_66" -> "1696 index_12" [label="(225, 12)", style=solid]; +"1696 index_12" -> "1697 view_67" [label="(4096, 12)", style=solid]; +"1697 view_67" -> "1698 permute_55" [label="(64, 64, 12)", style=solid]; +"1698 permute_55" -> "1699 contiguous_22" [label="(12, 64, 64)", style=solid]; +"1699 contiguous_22" -> "1700 unsqueeze_36" [label="(12, 64, 64)", style=solid]; +"1700 unsqueeze_36" -> "1701 sigmoid_12" [label="(1, 12, 64, 64)", style=solid]; +"1701 sigmoid_12" -> "1702 mul_24" [label="(1, 12, 64, 64)", style=solid]; +"1702 mul_24" -> "1738 add_42" [label="(1, 12, 64, 64)", style=solid]; +"1703 pad_14" -> "1704 view_68" [label="(1, 16, 16, 384)", style=solid]; +"1704 view_68" -> "1705 permute_56" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1705 permute_56" -> "1706 reshape_54" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1706 reshape_54" -> "1708 reshape_54_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1707 clone_12" -> "1715 linear_76" [label="(1152,)", style=solid]; +"1708 reshape_54_0_0_nncf_smooth_quant_0" -> "1709 quantize_per_tensor_default_75" [label="(4, 64, 384)", style=solid]; +"1709 quantize_per_tensor_default_75" -> "1710 dequantize_per_tensor_default_75" [label="(4, 64, 384)", style=solid]; +"1710 dequantize_per_tensor_default_75" -> "1715 linear_76" [label="(4, 64, 384)", style=solid]; +"1711 linear_76_scale_0" -> "1714 dequantize_per_channel_default_77" [label="(1152,)", style=solid]; +"1712 linear_76_zero_point_0" -> "1714 dequantize_per_channel_default_77" [label="(1152,)", style=solid]; +"1713 _frozen_param77" -> "1714 dequantize_per_channel_default_77" [label="(1152, 384)", style=solid]; +"1714 dequantize_per_channel_default_77" -> "1715 linear_76" [label="(1152, 384)", style=solid]; +"1715 linear_76" -> "1716 reshape_55" [label="(4, 64, 1152)", style=solid]; +"1716 reshape_55" -> "1717 permute_57" [label="(4, 64, 3, 12, 32)", style=solid]; +"1717 permute_57" -> "1718 select_36" [label="(3, 4, 12, 64, 32)", style=solid]; +"1717 permute_57" -> "1719 select_37" [label="(3, 4, 12, 64, 32)", style=solid]; +"1717 permute_57" -> "1720 select_38" [label="(3, 4, 12, 64, 32)", style=solid]; +"1718 select_36" -> "1721 linalg_vector_norm_24" [label="(4, 12, 64, 32)", style=solid]; +"1718 select_36" -> "1723 expand_as_24" [label="(4, 12, 64, 32)", style=solid]; +"1718 select_36" -> "1724 div_24" [label="(4, 12, 64, 32)", style=solid]; +"1719 select_37" -> "1727 linalg_vector_norm_25" [label="(4, 12, 64, 32)", style=solid]; +"1719 select_37" -> "1729 expand_as_25" [label="(4, 12, 64, 32)", style=solid]; +"1719 select_37" -> "1730 div_25" [label="(4, 12, 64, 32)", style=solid]; +"1720 select_38" -> "1741 matmul_25" [label="(4, 12, 64, 32)", style=solid]; +"1721 linalg_vector_norm_24" -> "1722 clamp_min_24" [label="(4, 12, 64, 1)", style=solid]; +"1722 clamp_min_24" -> "1723 expand_as_24" [label="(4, 12, 64, 1)", style=solid]; +"1723 expand_as_24" -> "1724 div_24" [label="(4, 12, 64, 32)", style=solid]; +"1724 div_24" -> "1725 quantize_per_tensor_default_76" [label="(4, 12, 64, 32)", style=solid]; +"1725 quantize_per_tensor_default_76" -> "1726 dequantize_per_tensor_default_76" [label="(4, 12, 64, 32)", style=solid]; +"1726 dequantize_per_tensor_default_76" -> "1734 matmul_24" [label="(4, 12, 64, 32)", style=solid]; +"1727 linalg_vector_norm_25" -> "1728 clamp_min_25" [label="(4, 12, 64, 1)", style=solid]; +"1728 clamp_min_25" -> "1729 expand_as_25" [label="(4, 12, 64, 1)", style=solid]; +"1729 expand_as_25" -> "1730 div_25" [label="(4, 12, 64, 32)", style=solid]; +"1730 div_25" -> "1731 quantize_per_tensor_default_77" [label="(4, 12, 64, 32)", style=solid]; +"1731 quantize_per_tensor_default_77" -> "1732 dequantize_per_tensor_default_77" [label="(4, 12, 64, 32)", style=solid]; +"1732 dequantize_per_tensor_default_77" -> "1733 transpose_24" [label="(4, 12, 64, 32)", style=solid]; +"1733 transpose_24" -> "1734 matmul_24" [label="(4, 12, 32, 64)", style=solid]; +"1734 matmul_24" -> "1737 mul_25" [label="(4, 12, 64, 64)", style=solid]; +"1735 clamp_12" -> "1736 exp_12" [label="(12, 1, 1)", style=solid]; +"1736 exp_12" -> "1737 mul_25" [label="(12, 1, 1)", style=solid]; +"1737 mul_25" -> "1738 add_42" [label="(4, 12, 64, 64)", style=solid]; +"1738 add_42" -> "1739 softmax_12" [label="(4, 12, 64, 64)", style=solid]; +"1739 softmax_12" -> "1740 dropout_48" [label="(4, 12, 64, 64)", style=solid]; +"1740 dropout_48" -> "1741 matmul_25" [label="(4, 12, 64, 64)", style=solid]; +"1741 matmul_25" -> "1742 transpose_25" [label="(4, 12, 64, 32)", style=solid]; +"1742 transpose_25" -> "1743 reshape_56" [label="(4, 64, 12, 32)", style=solid]; +"1743 reshape_56" -> "1744 reshape_56_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1744 reshape_56_0_0_nncf_smooth_quant_0" -> "1745 quantize_per_tensor_default_78" [label="(4, 64, 384)", style=solid]; +"1745 quantize_per_tensor_default_78" -> "1746 dequantize_per_tensor_default_78" [label="(4, 64, 384)", style=solid]; +"1746 dequantize_per_tensor_default_78" -> "1751 linear_77" [label="(4, 64, 384)", style=solid]; +"1747 linear_77_scale_0" -> "1750 dequantize_per_channel_default_78" [label="(384,)", style=solid]; +"1748 linear_77_zero_point_0" -> "1750 dequantize_per_channel_default_78" [label="(384,)", style=solid]; +"1749 _frozen_param78" -> "1750 dequantize_per_channel_default_78" [label="(384, 384)", style=solid]; +"1750 dequantize_per_channel_default_78" -> "1751 linear_77" [label="(384, 384)", style=solid]; +"1751 linear_77" -> "1752 dropout_49" [label="(4, 64, 384)", style=solid]; +"1752 dropout_49" -> "1753 view_69" [label="(4, 64, 384)", style=solid]; +"1753 view_69" -> "1754 permute_58" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1754 permute_58" -> "1755 reshape_57" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1755 reshape_57" -> "1756 slice_190" [label="(1, 16, 16, 384)", style=solid]; +"1756 slice_190" -> "1757 slice_191" [label="(1, 16, 16, 384)", style=solid]; +"1757 slice_191" -> "1758 slice_192" [label="(1, 14, 16, 384)", style=solid]; +"1758 slice_192" -> "1759 slice_193" [label="(1, 14, 14, 384)", style=solid]; +"1759 slice_193" -> "1760 contiguous_23" [label="(1, 14, 14, 384)", style=solid]; +"1760 contiguous_23" -> "1761 layer_norm_27" [label="(1, 14, 14, 384)", style=solid]; +"1761 layer_norm_27" -> "1762 add_43" [label="(1, 14, 14, 384)", style=solid]; +"1762 add_43" -> "1763 add_43_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"1762 add_43" -> "1783 add_44" [label="(1, 14, 14, 384)", style=solid]; +"1763 add_43_0_0_nncf_smooth_quant_0" -> "1764 quantize_per_tensor_default_79" [label="(1, 14, 14, 384)", style=solid]; +"1764 quantize_per_tensor_default_79" -> "1765 dequantize_per_tensor_default_79" [label="(1, 14, 14, 384)", style=solid]; +"1765 dequantize_per_tensor_default_79" -> "1770 linear_78" [label="(1, 14, 14, 384)", style=solid]; +"1766 linear_78_scale_0" -> "1769 dequantize_per_channel_default_79" [label="(1536,)", style=solid]; +"1767 linear_78_zero_point_0" -> "1769 dequantize_per_channel_default_79" [label="(1536,)", style=solid]; +"1768 _frozen_param79" -> "1769 dequantize_per_channel_default_79" [label="(1536, 384)", style=solid]; +"1769 dequantize_per_channel_default_79" -> "1770 linear_78" [label="(1536, 384)", style=solid]; +"1770 linear_78" -> "1771 gelu_12" [label="(1, 14, 14, 1536)", style=solid]; +"1771 gelu_12" -> "1772 dropout_50" [label="(1, 14, 14, 1536)", style=solid]; +"1772 dropout_50" -> "1773 dropout_50_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"1773 dropout_50_0_0_nncf_smooth_quant_0" -> "1774 quantize_per_tensor_default_80" [label="(1, 14, 14, 1536)", style=solid]; +"1774 quantize_per_tensor_default_80" -> "1775 dequantize_per_tensor_default_80" [label="(1, 14, 14, 1536)", style=solid]; +"1775 dequantize_per_tensor_default_80" -> "1780 linear_79" [label="(1, 14, 14, 1536)", style=solid]; +"1776 linear_79_scale_0" -> "1779 dequantize_per_channel_default_80" [label="(384,)", style=solid]; +"1777 linear_79_zero_point_0" -> "1779 dequantize_per_channel_default_80" [label="(384,)", style=solid]; +"1778 _frozen_param80" -> "1779 dequantize_per_channel_default_80" [label="(384, 1536)", style=solid]; +"1779 dequantize_per_channel_default_80" -> "1780 linear_79" [label="(384, 1536)", style=solid]; +"1780 linear_79" -> "1781 dropout_51" [label="(1, 14, 14, 384)", style=solid]; +"1781 dropout_51" -> "1782 layer_norm_28" [label="(1, 14, 14, 384)", style=solid]; +"1782 layer_norm_28" -> "1783 add_44" [label="(1, 14, 14, 384)", style=solid]; +"1783 add_44" -> "1805 pad_15" [label="(1, 14, 14, 384)", style=solid]; +"1783 add_44" -> "1882 add_47" [label="(1, 14, 14, 384)", style=solid]; +"1784 features_5_9_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "1789 linear_80" [label="(1, 15, 15, 2)", style=solid]; +"1785 linear_80_scale_0" -> "1788 dequantize_per_channel_default_81" [label="(512,)", style=solid]; +"1786 linear_80_zero_point_0" -> "1788 dequantize_per_channel_default_81" [label="(512,)", style=solid]; +"1787 _frozen_param81" -> "1788 dequantize_per_channel_default_81" [label="(512, 2)", style=solid]; +"1788 dequantize_per_channel_default_81" -> "1789 linear_80" [label="(512, 2)", style=solid]; +"1789 linear_80" -> "1790 relu__13" [label="(1, 15, 15, 512)", style=solid]; +"1790 relu__13" -> "1791 relu__13_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"1791 relu__13_0_0_nncf_smooth_quant_0" -> "1796 linear_81" [label="(1, 15, 15, 512)", style=solid]; +"1792 linear_81_scale_0" -> "1795 dequantize_per_channel_default_82" [label="(12,)", style=solid]; +"1793 linear_81_zero_point_0" -> "1795 dequantize_per_channel_default_82" [label="(12,)", style=solid]; +"1794 _frozen_param82" -> "1795 dequantize_per_channel_default_82" [label="(12, 512)", style=solid]; +"1795 dequantize_per_channel_default_82" -> "1796 linear_81" [label="(12, 512)", style=solid]; +"1796 linear_81" -> "1797 view_70" [label="(1, 15, 15, 12)", style=solid]; +"1797 view_70" -> "1798 index_13" [label="(225, 12)", style=solid]; +"1798 index_13" -> "1799 view_71" [label="(4096, 12)", style=solid]; +"1799 view_71" -> "1800 permute_59" [label="(64, 64, 12)", style=solid]; +"1800 permute_59" -> "1801 contiguous_24" [label="(12, 64, 64)", style=solid]; +"1801 contiguous_24" -> "1802 unsqueeze_37" [label="(12, 64, 64)", style=solid]; +"1802 unsqueeze_37" -> "1803 sigmoid_13" [label="(1, 12, 64, 64)", style=solid]; +"1803 sigmoid_13" -> "1804 mul_26" [label="(1, 12, 64, 64)", style=solid]; +"1804 mul_26" -> "1841 add_45" [label="(1, 12, 64, 64)", style=solid]; +"1805 pad_15" -> "1806 roll_12" [label="(1, 16, 16, 384)", style=solid]; +"1806 roll_12" -> "1807 view_72" [label="(1, 16, 16, 384)", style=solid]; +"1807 view_72" -> "1808 permute_60" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1808 permute_60" -> "1809 reshape_58" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1809 reshape_58" -> "1811 reshape_58_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1809 reshape_58" -> "1842 new_zeros_6" [label="(4, 64, 384)", style=solid]; +"1810 clone_13" -> "1818 linear_82" [label="(1152,)", style=solid]; +"1811 reshape_58_0_0_nncf_smooth_quant_0" -> "1812 quantize_per_tensor_default_81" [label="(4, 64, 384)", style=solid]; +"1812 quantize_per_tensor_default_81" -> "1813 dequantize_per_tensor_default_81" [label="(4, 64, 384)", style=solid]; +"1813 dequantize_per_tensor_default_81" -> "1818 linear_82" [label="(4, 64, 384)", style=solid]; +"1814 linear_82_scale_0" -> "1817 dequantize_per_channel_default_83" [label="(1152,)", style=solid]; +"1815 linear_82_zero_point_0" -> "1817 dequantize_per_channel_default_83" [label="(1152,)", style=solid]; +"1816 _frozen_param83" -> "1817 dequantize_per_channel_default_83" [label="(1152, 384)", style=solid]; +"1817 dequantize_per_channel_default_83" -> "1818 linear_82" [label="(1152, 384)", style=solid]; +"1818 linear_82" -> "1819 reshape_59" [label="(4, 64, 1152)", style=solid]; +"1819 reshape_59" -> "1820 permute_61" [label="(4, 64, 3, 12, 32)", style=solid]; +"1820 permute_61" -> "1821 select_39" [label="(3, 4, 12, 64, 32)", style=solid]; +"1820 permute_61" -> "1822 select_40" [label="(3, 4, 12, 64, 32)", style=solid]; +"1820 permute_61" -> "1823 select_41" [label="(3, 4, 12, 64, 32)", style=solid]; +"1821 select_39" -> "1824 linalg_vector_norm_26" [label="(4, 12, 64, 32)", style=solid]; +"1821 select_39" -> "1826 expand_as_26" [label="(4, 12, 64, 32)", style=solid]; +"1821 select_39" -> "1827 div_26" [label="(4, 12, 64, 32)", style=solid]; +"1822 select_40" -> "1830 linalg_vector_norm_27" [label="(4, 12, 64, 32)", style=solid]; +"1822 select_40" -> "1832 expand_as_27" [label="(4, 12, 64, 32)", style=solid]; +"1822 select_40" -> "1833 div_27" [label="(4, 12, 64, 32)", style=solid]; +"1823 select_41" -> "1860 matmul_27" [label="(4, 12, 64, 32)", style=solid]; +"1824 linalg_vector_norm_26" -> "1825 clamp_min_26" [label="(4, 12, 64, 1)", style=solid]; +"1825 clamp_min_26" -> "1826 expand_as_26" [label="(4, 12, 64, 1)", style=solid]; +"1826 expand_as_26" -> "1827 div_26" [label="(4, 12, 64, 32)", style=solid]; +"1827 div_26" -> "1828 quantize_per_tensor_default_82" [label="(4, 12, 64, 32)", style=solid]; +"1828 quantize_per_tensor_default_82" -> "1829 dequantize_per_tensor_default_82" [label="(4, 12, 64, 32)", style=solid]; +"1829 dequantize_per_tensor_default_82" -> "1837 matmul_26" [label="(4, 12, 64, 32)", style=solid]; +"1830 linalg_vector_norm_27" -> "1831 clamp_min_27" [label="(4, 12, 64, 1)", style=solid]; +"1831 clamp_min_27" -> "1832 expand_as_27" [label="(4, 12, 64, 1)", style=solid]; +"1832 expand_as_27" -> "1833 div_27" [label="(4, 12, 64, 32)", style=solid]; +"1833 div_27" -> "1834 quantize_per_tensor_default_83" [label="(4, 12, 64, 32)", style=solid]; +"1834 quantize_per_tensor_default_83" -> "1835 dequantize_per_tensor_default_83" [label="(4, 12, 64, 32)", style=solid]; +"1835 dequantize_per_tensor_default_83" -> "1836 transpose_26" [label="(4, 12, 64, 32)", style=solid]; +"1836 transpose_26" -> "1837 matmul_26" [label="(4, 12, 32, 64)", style=solid]; +"1837 matmul_26" -> "1840 mul_27" [label="(4, 12, 64, 64)", style=solid]; +"1838 clamp_13" -> "1839 exp_13" [label="(12, 1, 1)", style=solid]; +"1839 exp_13" -> "1840 mul_27" [label="(12, 1, 1)", style=solid]; +"1840 mul_27" -> "1841 add_45" [label="(4, 12, 64, 64)", style=solid]; +"1841 add_45" -> "1853 view_74" [label="(4, 12, 64, 64)", style=solid]; +"1842 new_zeros_6" -> "1843 view_73" [label="(16, 16)", style=solid]; +"1843 view_73" -> "1844 permute_62" [label="(2, 8, 2, 8)", style=solid]; +"1844 permute_62" -> "1845 reshape_60" [label="(2, 2, 8, 8)", style=solid]; +"1845 reshape_60" -> "1846 unsqueeze_38" [label="(4, 64)", style=solid]; +"1845 reshape_60" -> "1847 unsqueeze_39" [label="(4, 64)", style=solid]; +"1846 unsqueeze_38" -> "1848 sub_6" [label="(4, 1, 64)", style=solid]; +"1847 unsqueeze_39" -> "1848 sub_6" [label="(4, 64, 1)", style=solid]; +"1848 sub_6" -> "1849 ne_6" [label="(4, 64, 64)", style=solid]; +"1848 sub_6" -> "1850 masked_fill_12" [label="(4, 64, 64)", style=solid]; +"1848 sub_6" -> "1851 eq_6" [label="(4, 64, 64)", style=solid]; +"1849 ne_6" -> "1850 masked_fill_12" [label="(4, 64, 64)", style=solid]; +"1850 masked_fill_12" -> "1852 masked_fill_13" [label="(4, 64, 64)", style=solid]; +"1851 eq_6" -> "1852 masked_fill_13" [label="(4, 64, 64)", style=solid]; +"1852 masked_fill_13" -> "1854 unsqueeze_40" [label="(4, 64, 64)", style=solid]; +"1853 view_74" -> "1856 add_46" [label="(1, 4, 12, 64, 64)", style=solid]; +"1854 unsqueeze_40" -> "1855 unsqueeze_41" [label="(4, 1, 64, 64)", style=solid]; +"1855 unsqueeze_41" -> "1856 add_46" [label="(1, 4, 1, 64, 64)", style=solid]; +"1856 add_46" -> "1857 view_75" [label="(1, 4, 12, 64, 64)", style=solid]; +"1857 view_75" -> "1858 softmax_13" [label="(4, 12, 64, 64)", style=solid]; +"1858 softmax_13" -> "1859 dropout_52" [label="(4, 12, 64, 64)", style=solid]; +"1859 dropout_52" -> "1860 matmul_27" [label="(4, 12, 64, 64)", style=solid]; +"1860 matmul_27" -> "1861 transpose_27" [label="(4, 12, 64, 32)", style=solid]; +"1861 transpose_27" -> "1862 reshape_61" [label="(4, 64, 12, 32)", style=solid]; +"1862 reshape_61" -> "1863 reshape_61_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1863 reshape_61_0_0_nncf_smooth_quant_0" -> "1864 quantize_per_tensor_default_84" [label="(4, 64, 384)", style=solid]; +"1864 quantize_per_tensor_default_84" -> "1865 dequantize_per_tensor_default_84" [label="(4, 64, 384)", style=solid]; +"1865 dequantize_per_tensor_default_84" -> "1870 linear_83" [label="(4, 64, 384)", style=solid]; +"1866 linear_83_scale_0" -> "1869 dequantize_per_channel_default_84" [label="(384,)", style=solid]; +"1867 linear_83_zero_point_0" -> "1869 dequantize_per_channel_default_84" [label="(384,)", style=solid]; +"1868 _frozen_param84" -> "1869 dequantize_per_channel_default_84" [label="(384, 384)", style=solid]; +"1869 dequantize_per_channel_default_84" -> "1870 linear_83" [label="(384, 384)", style=solid]; +"1870 linear_83" -> "1871 dropout_53" [label="(4, 64, 384)", style=solid]; +"1871 dropout_53" -> "1872 view_76" [label="(4, 64, 384)", style=solid]; +"1872 view_76" -> "1873 permute_63" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1873 permute_63" -> "1874 reshape_62" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1874 reshape_62" -> "1875 roll_13" [label="(1, 16, 16, 384)", style=solid]; +"1875 roll_13" -> "1876 slice_213" [label="(1, 16, 16, 384)", style=solid]; +"1876 slice_213" -> "1877 slice_214" [label="(1, 16, 16, 384)", style=solid]; +"1877 slice_214" -> "1878 slice_215" [label="(1, 14, 16, 384)", style=solid]; +"1878 slice_215" -> "1879 slice_216" [label="(1, 14, 14, 384)", style=solid]; +"1879 slice_216" -> "1880 contiguous_25" [label="(1, 14, 14, 384)", style=solid]; +"1880 contiguous_25" -> "1881 layer_norm_29" [label="(1, 14, 14, 384)", style=solid]; +"1881 layer_norm_29" -> "1882 add_47" [label="(1, 14, 14, 384)", style=solid]; +"1882 add_47" -> "1883 add_47_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"1882 add_47" -> "1903 add_48" [label="(1, 14, 14, 384)", style=solid]; +"1883 add_47_0_0_nncf_smooth_quant_0" -> "1884 quantize_per_tensor_default_85" [label="(1, 14, 14, 384)", style=solid]; +"1884 quantize_per_tensor_default_85" -> "1885 dequantize_per_tensor_default_85" [label="(1, 14, 14, 384)", style=solid]; +"1885 dequantize_per_tensor_default_85" -> "1890 linear_84" [label="(1, 14, 14, 384)", style=solid]; +"1886 linear_84_scale_0" -> "1889 dequantize_per_channel_default_85" [label="(1536,)", style=solid]; +"1887 linear_84_zero_point_0" -> "1889 dequantize_per_channel_default_85" [label="(1536,)", style=solid]; +"1888 _frozen_param85" -> "1889 dequantize_per_channel_default_85" [label="(1536, 384)", style=solid]; +"1889 dequantize_per_channel_default_85" -> "1890 linear_84" [label="(1536, 384)", style=solid]; +"1890 linear_84" -> "1891 gelu_13" [label="(1, 14, 14, 1536)", style=solid]; +"1891 gelu_13" -> "1892 dropout_54" [label="(1, 14, 14, 1536)", style=solid]; +"1892 dropout_54" -> "1893 dropout_54_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"1893 dropout_54_0_0_nncf_smooth_quant_0" -> "1894 quantize_per_tensor_default_86" [label="(1, 14, 14, 1536)", style=solid]; +"1894 quantize_per_tensor_default_86" -> "1895 dequantize_per_tensor_default_86" [label="(1, 14, 14, 1536)", style=solid]; +"1895 dequantize_per_tensor_default_86" -> "1900 linear_85" [label="(1, 14, 14, 1536)", style=solid]; +"1896 linear_85_scale_0" -> "1899 dequantize_per_channel_default_86" [label="(384,)", style=solid]; +"1897 linear_85_zero_point_0" -> "1899 dequantize_per_channel_default_86" [label="(384,)", style=solid]; +"1898 _frozen_param86" -> "1899 dequantize_per_channel_default_86" [label="(384, 1536)", style=solid]; +"1899 dequantize_per_channel_default_86" -> "1900 linear_85" [label="(384, 1536)", style=solid]; +"1900 linear_85" -> "1901 dropout_55" [label="(1, 14, 14, 384)", style=solid]; +"1901 dropout_55" -> "1902 layer_norm_30" [label="(1, 14, 14, 384)", style=solid]; +"1902 layer_norm_30" -> "1903 add_48" [label="(1, 14, 14, 384)", style=solid]; +"1903 add_48" -> "1925 pad_16" [label="(1, 14, 14, 384)", style=solid]; +"1903 add_48" -> "1984 add_50" [label="(1, 14, 14, 384)", style=solid]; +"1904 features_5_10_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "1909 linear_86" [label="(1, 15, 15, 2)", style=solid]; +"1905 linear_86_scale_0" -> "1908 dequantize_per_channel_default_87" [label="(512,)", style=solid]; +"1906 linear_86_zero_point_0" -> "1908 dequantize_per_channel_default_87" [label="(512,)", style=solid]; +"1907 _frozen_param87" -> "1908 dequantize_per_channel_default_87" [label="(512, 2)", style=solid]; +"1908 dequantize_per_channel_default_87" -> "1909 linear_86" [label="(512, 2)", style=solid]; +"1909 linear_86" -> "1910 relu__14" [label="(1, 15, 15, 512)", style=solid]; +"1910 relu__14" -> "1911 relu__14_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"1911 relu__14_0_0_nncf_smooth_quant_0" -> "1916 linear_87" [label="(1, 15, 15, 512)", style=solid]; +"1912 linear_87_scale_0" -> "1915 dequantize_per_channel_default_88" [label="(12,)", style=solid]; +"1913 linear_87_zero_point_0" -> "1915 dequantize_per_channel_default_88" [label="(12,)", style=solid]; +"1914 _frozen_param88" -> "1915 dequantize_per_channel_default_88" [label="(12, 512)", style=solid]; +"1915 dequantize_per_channel_default_88" -> "1916 linear_87" [label="(12, 512)", style=solid]; +"1916 linear_87" -> "1917 view_77" [label="(1, 15, 15, 12)", style=solid]; +"1917 view_77" -> "1918 index_14" [label="(225, 12)", style=solid]; +"1918 index_14" -> "1919 view_78" [label="(4096, 12)", style=solid]; +"1919 view_78" -> "1920 permute_64" [label="(64, 64, 12)", style=solid]; +"1920 permute_64" -> "1921 contiguous_26" [label="(12, 64, 64)", style=solid]; +"1921 contiguous_26" -> "1922 unsqueeze_42" [label="(12, 64, 64)", style=solid]; +"1922 unsqueeze_42" -> "1923 sigmoid_14" [label="(1, 12, 64, 64)", style=solid]; +"1923 sigmoid_14" -> "1924 mul_28" [label="(1, 12, 64, 64)", style=solid]; +"1924 mul_28" -> "1960 add_49" [label="(1, 12, 64, 64)", style=solid]; +"1925 pad_16" -> "1926 view_79" [label="(1, 16, 16, 384)", style=solid]; +"1926 view_79" -> "1927 permute_65" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1927 permute_65" -> "1928 reshape_63" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1928 reshape_63" -> "1930 reshape_63_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1929 clone_14" -> "1937 linear_88" [label="(1152,)", style=solid]; +"1930 reshape_63_0_0_nncf_smooth_quant_0" -> "1931 quantize_per_tensor_default_87" [label="(4, 64, 384)", style=solid]; +"1931 quantize_per_tensor_default_87" -> "1932 dequantize_per_tensor_default_87" [label="(4, 64, 384)", style=solid]; +"1932 dequantize_per_tensor_default_87" -> "1937 linear_88" [label="(4, 64, 384)", style=solid]; +"1933 linear_88_scale_0" -> "1936 dequantize_per_channel_default_89" [label="(1152,)", style=solid]; +"1934 linear_88_zero_point_0" -> "1936 dequantize_per_channel_default_89" [label="(1152,)", style=solid]; +"1935 _frozen_param89" -> "1936 dequantize_per_channel_default_89" [label="(1152, 384)", style=solid]; +"1936 dequantize_per_channel_default_89" -> "1937 linear_88" [label="(1152, 384)", style=solid]; +"1937 linear_88" -> "1938 reshape_64" [label="(4, 64, 1152)", style=solid]; +"1938 reshape_64" -> "1939 permute_66" [label="(4, 64, 3, 12, 32)", style=solid]; +"1939 permute_66" -> "1940 select_42" [label="(3, 4, 12, 64, 32)", style=solid]; +"1939 permute_66" -> "1941 select_43" [label="(3, 4, 12, 64, 32)", style=solid]; +"1939 permute_66" -> "1942 select_44" [label="(3, 4, 12, 64, 32)", style=solid]; +"1940 select_42" -> "1943 linalg_vector_norm_28" [label="(4, 12, 64, 32)", style=solid]; +"1940 select_42" -> "1945 expand_as_28" [label="(4, 12, 64, 32)", style=solid]; +"1940 select_42" -> "1946 div_28" [label="(4, 12, 64, 32)", style=solid]; +"1941 select_43" -> "1949 linalg_vector_norm_29" [label="(4, 12, 64, 32)", style=solid]; +"1941 select_43" -> "1951 expand_as_29" [label="(4, 12, 64, 32)", style=solid]; +"1941 select_43" -> "1952 div_29" [label="(4, 12, 64, 32)", style=solid]; +"1942 select_44" -> "1963 matmul_29" [label="(4, 12, 64, 32)", style=solid]; +"1943 linalg_vector_norm_28" -> "1944 clamp_min_28" [label="(4, 12, 64, 1)", style=solid]; +"1944 clamp_min_28" -> "1945 expand_as_28" [label="(4, 12, 64, 1)", style=solid]; +"1945 expand_as_28" -> "1946 div_28" [label="(4, 12, 64, 32)", style=solid]; +"1946 div_28" -> "1947 quantize_per_tensor_default_88" [label="(4, 12, 64, 32)", style=solid]; +"1947 quantize_per_tensor_default_88" -> "1948 dequantize_per_tensor_default_88" [label="(4, 12, 64, 32)", style=solid]; +"1948 dequantize_per_tensor_default_88" -> "1956 matmul_28" [label="(4, 12, 64, 32)", style=solid]; +"1949 linalg_vector_norm_29" -> "1950 clamp_min_29" [label="(4, 12, 64, 1)", style=solid]; +"1950 clamp_min_29" -> "1951 expand_as_29" [label="(4, 12, 64, 1)", style=solid]; +"1951 expand_as_29" -> "1952 div_29" [label="(4, 12, 64, 32)", style=solid]; +"1952 div_29" -> "1953 quantize_per_tensor_default_89" [label="(4, 12, 64, 32)", style=solid]; +"1953 quantize_per_tensor_default_89" -> "1954 dequantize_per_tensor_default_89" [label="(4, 12, 64, 32)", style=solid]; +"1954 dequantize_per_tensor_default_89" -> "1955 transpose_28" [label="(4, 12, 64, 32)", style=solid]; +"1955 transpose_28" -> "1956 matmul_28" [label="(4, 12, 32, 64)", style=solid]; +"1956 matmul_28" -> "1959 mul_29" [label="(4, 12, 64, 64)", style=solid]; +"1957 clamp_14" -> "1958 exp_14" [label="(12, 1, 1)", style=solid]; +"1958 exp_14" -> "1959 mul_29" [label="(12, 1, 1)", style=solid]; +"1959 mul_29" -> "1960 add_49" [label="(4, 12, 64, 64)", style=solid]; +"1960 add_49" -> "1961 softmax_14" [label="(4, 12, 64, 64)", style=solid]; +"1961 softmax_14" -> "1962 dropout_56" [label="(4, 12, 64, 64)", style=solid]; +"1962 dropout_56" -> "1963 matmul_29" [label="(4, 12, 64, 64)", style=solid]; +"1963 matmul_29" -> "1964 transpose_29" [label="(4, 12, 64, 32)", style=solid]; +"1964 transpose_29" -> "1965 reshape_65" [label="(4, 64, 12, 32)", style=solid]; +"1965 reshape_65" -> "1966 reshape_65_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1966 reshape_65_0_0_nncf_smooth_quant_0" -> "1967 quantize_per_tensor_default_90" [label="(4, 64, 384)", style=solid]; +"1967 quantize_per_tensor_default_90" -> "1968 dequantize_per_tensor_default_90" [label="(4, 64, 384)", style=solid]; +"1968 dequantize_per_tensor_default_90" -> "1973 linear_89" [label="(4, 64, 384)", style=solid]; +"1969 linear_89_scale_0" -> "1972 dequantize_per_channel_default_90" [label="(384,)", style=solid]; +"1970 linear_89_zero_point_0" -> "1972 dequantize_per_channel_default_90" [label="(384,)", style=solid]; +"1971 _frozen_param90" -> "1972 dequantize_per_channel_default_90" [label="(384, 384)", style=solid]; +"1972 dequantize_per_channel_default_90" -> "1973 linear_89" [label="(384, 384)", style=solid]; +"1973 linear_89" -> "1974 dropout_57" [label="(4, 64, 384)", style=solid]; +"1974 dropout_57" -> "1975 view_80" [label="(4, 64, 384)", style=solid]; +"1975 view_80" -> "1976 permute_67" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1976 permute_67" -> "1977 reshape_66" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1977 reshape_66" -> "1978 slice_218" [label="(1, 16, 16, 384)", style=solid]; +"1978 slice_218" -> "1979 slice_219" [label="(1, 16, 16, 384)", style=solid]; +"1979 slice_219" -> "1980 slice_220" [label="(1, 14, 16, 384)", style=solid]; +"1980 slice_220" -> "1981 slice_221" [label="(1, 14, 14, 384)", style=solid]; +"1981 slice_221" -> "1982 contiguous_27" [label="(1, 14, 14, 384)", style=solid]; +"1982 contiguous_27" -> "1983 layer_norm_31" [label="(1, 14, 14, 384)", style=solid]; +"1983 layer_norm_31" -> "1984 add_50" [label="(1, 14, 14, 384)", style=solid]; +"1984 add_50" -> "1985 add_50_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"1984 add_50" -> "2005 add_51" [label="(1, 14, 14, 384)", style=solid]; +"1985 add_50_0_0_nncf_smooth_quant_0" -> "1986 quantize_per_tensor_default_91" [label="(1, 14, 14, 384)", style=solid]; +"1986 quantize_per_tensor_default_91" -> "1987 dequantize_per_tensor_default_91" [label="(1, 14, 14, 384)", style=solid]; +"1987 dequantize_per_tensor_default_91" -> "1992 linear_90" [label="(1, 14, 14, 384)", style=solid]; +"1988 linear_90_scale_0" -> "1991 dequantize_per_channel_default_91" [label="(1536,)", style=solid]; +"1989 linear_90_zero_point_0" -> "1991 dequantize_per_channel_default_91" [label="(1536,)", style=solid]; +"1990 _frozen_param91" -> "1991 dequantize_per_channel_default_91" [label="(1536, 384)", style=solid]; +"1991 dequantize_per_channel_default_91" -> "1992 linear_90" [label="(1536, 384)", style=solid]; +"1992 linear_90" -> "1993 gelu_14" [label="(1, 14, 14, 1536)", style=solid]; +"1993 gelu_14" -> "1994 dropout_58" [label="(1, 14, 14, 1536)", style=solid]; +"1994 dropout_58" -> "1995 dropout_58_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"1995 dropout_58_0_0_nncf_smooth_quant_0" -> "1996 quantize_per_tensor_default_92" [label="(1, 14, 14, 1536)", style=solid]; +"1996 quantize_per_tensor_default_92" -> "1997 dequantize_per_tensor_default_92" [label="(1, 14, 14, 1536)", style=solid]; +"1997 dequantize_per_tensor_default_92" -> "2002 linear_91" [label="(1, 14, 14, 1536)", style=solid]; +"1998 linear_91_scale_0" -> "2001 dequantize_per_channel_default_92" [label="(384,)", style=solid]; +"1999 linear_91_zero_point_0" -> "2001 dequantize_per_channel_default_92" [label="(384,)", style=solid]; +"2000 _frozen_param92" -> "2001 dequantize_per_channel_default_92" [label="(384, 1536)", style=solid]; +"2001 dequantize_per_channel_default_92" -> "2002 linear_91" [label="(384, 1536)", style=solid]; +"2002 linear_91" -> "2003 dropout_59" [label="(1, 14, 14, 384)", style=solid]; +"2003 dropout_59" -> "2004 layer_norm_32" [label="(1, 14, 14, 384)", style=solid]; +"2004 layer_norm_32" -> "2005 add_51" [label="(1, 14, 14, 384)", style=solid]; +"2005 add_51" -> "2027 pad_17" [label="(1, 14, 14, 384)", style=solid]; +"2005 add_51" -> "2104 add_54" [label="(1, 14, 14, 384)", style=solid]; +"2006 features_5_11_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "2011 linear_92" [label="(1, 15, 15, 2)", style=solid]; +"2007 linear_92_scale_0" -> "2010 dequantize_per_channel_default_93" [label="(512,)", style=solid]; +"2008 linear_92_zero_point_0" -> "2010 dequantize_per_channel_default_93" [label="(512,)", style=solid]; +"2009 _frozen_param93" -> "2010 dequantize_per_channel_default_93" [label="(512, 2)", style=solid]; +"2010 dequantize_per_channel_default_93" -> "2011 linear_92" [label="(512, 2)", style=solid]; +"2011 linear_92" -> "2012 relu__15" [label="(1, 15, 15, 512)", style=solid]; +"2012 relu__15" -> "2013 relu__15_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"2013 relu__15_0_0_nncf_smooth_quant_0" -> "2018 linear_93" [label="(1, 15, 15, 512)", style=solid]; +"2014 linear_93_scale_0" -> "2017 dequantize_per_channel_default_94" [label="(12,)", style=solid]; +"2015 linear_93_zero_point_0" -> "2017 dequantize_per_channel_default_94" [label="(12,)", style=solid]; +"2016 _frozen_param94" -> "2017 dequantize_per_channel_default_94" [label="(12, 512)", style=solid]; +"2017 dequantize_per_channel_default_94" -> "2018 linear_93" [label="(12, 512)", style=solid]; +"2018 linear_93" -> "2019 view_81" [label="(1, 15, 15, 12)", style=solid]; +"2019 view_81" -> "2020 index_15" [label="(225, 12)", style=solid]; +"2020 index_15" -> "2021 view_82" [label="(4096, 12)", style=solid]; +"2021 view_82" -> "2022 permute_68" [label="(64, 64, 12)", style=solid]; +"2022 permute_68" -> "2023 contiguous_28" [label="(12, 64, 64)", style=solid]; +"2023 contiguous_28" -> "2024 unsqueeze_43" [label="(12, 64, 64)", style=solid]; +"2024 unsqueeze_43" -> "2025 sigmoid_15" [label="(1, 12, 64, 64)", style=solid]; +"2025 sigmoid_15" -> "2026 mul_30" [label="(1, 12, 64, 64)", style=solid]; +"2026 mul_30" -> "2063 add_52" [label="(1, 12, 64, 64)", style=solid]; +"2027 pad_17" -> "2028 roll_14" [label="(1, 16, 16, 384)", style=solid]; +"2028 roll_14" -> "2029 view_83" [label="(1, 16, 16, 384)", style=solid]; +"2029 view_83" -> "2030 permute_69" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2030 permute_69" -> "2031 reshape_67" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2031 reshape_67" -> "2033 reshape_67_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"2031 reshape_67" -> "2064 new_zeros_7" [label="(4, 64, 384)", style=solid]; +"2032 clone_15" -> "2040 linear_94" [label="(1152,)", style=solid]; +"2033 reshape_67_0_0_nncf_smooth_quant_0" -> "2034 quantize_per_tensor_default_93" [label="(4, 64, 384)", style=solid]; +"2034 quantize_per_tensor_default_93" -> "2035 dequantize_per_tensor_default_93" [label="(4, 64, 384)", style=solid]; +"2035 dequantize_per_tensor_default_93" -> "2040 linear_94" [label="(4, 64, 384)", style=solid]; +"2036 linear_94_scale_0" -> "2039 dequantize_per_channel_default_95" [label="(1152,)", style=solid]; +"2037 linear_94_zero_point_0" -> "2039 dequantize_per_channel_default_95" [label="(1152,)", style=solid]; +"2038 _frozen_param95" -> "2039 dequantize_per_channel_default_95" [label="(1152, 384)", style=solid]; +"2039 dequantize_per_channel_default_95" -> "2040 linear_94" [label="(1152, 384)", style=solid]; +"2040 linear_94" -> "2041 reshape_68" [label="(4, 64, 1152)", style=solid]; +"2041 reshape_68" -> "2042 permute_70" [label="(4, 64, 3, 12, 32)", style=solid]; +"2042 permute_70" -> "2043 select_45" [label="(3, 4, 12, 64, 32)", style=solid]; +"2042 permute_70" -> "2044 select_46" [label="(3, 4, 12, 64, 32)", style=solid]; +"2042 permute_70" -> "2045 select_47" [label="(3, 4, 12, 64, 32)", style=solid]; +"2043 select_45" -> "2046 linalg_vector_norm_30" [label="(4, 12, 64, 32)", style=solid]; +"2043 select_45" -> "2048 expand_as_30" [label="(4, 12, 64, 32)", style=solid]; +"2043 select_45" -> "2049 div_30" [label="(4, 12, 64, 32)", style=solid]; +"2044 select_46" -> "2052 linalg_vector_norm_31" [label="(4, 12, 64, 32)", style=solid]; +"2044 select_46" -> "2054 expand_as_31" [label="(4, 12, 64, 32)", style=solid]; +"2044 select_46" -> "2055 div_31" [label="(4, 12, 64, 32)", style=solid]; +"2045 select_47" -> "2082 matmul_31" [label="(4, 12, 64, 32)", style=solid]; +"2046 linalg_vector_norm_30" -> "2047 clamp_min_30" [label="(4, 12, 64, 1)", style=solid]; +"2047 clamp_min_30" -> "2048 expand_as_30" [label="(4, 12, 64, 1)", style=solid]; +"2048 expand_as_30" -> "2049 div_30" [label="(4, 12, 64, 32)", style=solid]; +"2049 div_30" -> "2050 quantize_per_tensor_default_94" [label="(4, 12, 64, 32)", style=solid]; +"2050 quantize_per_tensor_default_94" -> "2051 dequantize_per_tensor_default_94" [label="(4, 12, 64, 32)", style=solid]; +"2051 dequantize_per_tensor_default_94" -> "2059 matmul_30" [label="(4, 12, 64, 32)", style=solid]; +"2052 linalg_vector_norm_31" -> "2053 clamp_min_31" [label="(4, 12, 64, 1)", style=solid]; +"2053 clamp_min_31" -> "2054 expand_as_31" [label="(4, 12, 64, 1)", style=solid]; +"2054 expand_as_31" -> "2055 div_31" [label="(4, 12, 64, 32)", style=solid]; +"2055 div_31" -> "2056 quantize_per_tensor_default_95" [label="(4, 12, 64, 32)", style=solid]; +"2056 quantize_per_tensor_default_95" -> "2057 dequantize_per_tensor_default_95" [label="(4, 12, 64, 32)", style=solid]; +"2057 dequantize_per_tensor_default_95" -> "2058 transpose_30" [label="(4, 12, 64, 32)", style=solid]; +"2058 transpose_30" -> "2059 matmul_30" [label="(4, 12, 32, 64)", style=solid]; +"2059 matmul_30" -> "2062 mul_31" [label="(4, 12, 64, 64)", style=solid]; +"2060 clamp_15" -> "2061 exp_15" [label="(12, 1, 1)", style=solid]; +"2061 exp_15" -> "2062 mul_31" [label="(12, 1, 1)", style=solid]; +"2062 mul_31" -> "2063 add_52" [label="(4, 12, 64, 64)", style=solid]; +"2063 add_52" -> "2075 view_85" [label="(4, 12, 64, 64)", style=solid]; +"2064 new_zeros_7" -> "2065 view_84" [label="(16, 16)", style=solid]; +"2065 view_84" -> "2066 permute_71" [label="(2, 8, 2, 8)", style=solid]; +"2066 permute_71" -> "2067 reshape_69" [label="(2, 2, 8, 8)", style=solid]; +"2067 reshape_69" -> "2068 unsqueeze_44" [label="(4, 64)", style=solid]; +"2067 reshape_69" -> "2069 unsqueeze_45" [label="(4, 64)", style=solid]; +"2068 unsqueeze_44" -> "2070 sub_7" [label="(4, 1, 64)", style=solid]; +"2069 unsqueeze_45" -> "2070 sub_7" [label="(4, 64, 1)", style=solid]; +"2070 sub_7" -> "2071 ne_7" [label="(4, 64, 64)", style=solid]; +"2070 sub_7" -> "2072 masked_fill_14" [label="(4, 64, 64)", style=solid]; +"2070 sub_7" -> "2073 eq_7" [label="(4, 64, 64)", style=solid]; +"2071 ne_7" -> "2072 masked_fill_14" [label="(4, 64, 64)", style=solid]; +"2072 masked_fill_14" -> "2074 masked_fill_15" [label="(4, 64, 64)", style=solid]; +"2073 eq_7" -> "2074 masked_fill_15" [label="(4, 64, 64)", style=solid]; +"2074 masked_fill_15" -> "2076 unsqueeze_46" [label="(4, 64, 64)", style=solid]; +"2075 view_85" -> "2078 add_53" [label="(1, 4, 12, 64, 64)", style=solid]; +"2076 unsqueeze_46" -> "2077 unsqueeze_47" [label="(4, 1, 64, 64)", style=solid]; +"2077 unsqueeze_47" -> "2078 add_53" [label="(1, 4, 1, 64, 64)", style=solid]; +"2078 add_53" -> "2079 view_86" [label="(1, 4, 12, 64, 64)", style=solid]; +"2079 view_86" -> "2080 softmax_15" [label="(4, 12, 64, 64)", style=solid]; +"2080 softmax_15" -> "2081 dropout_60" [label="(4, 12, 64, 64)", style=solid]; +"2081 dropout_60" -> "2082 matmul_31" [label="(4, 12, 64, 64)", style=solid]; +"2082 matmul_31" -> "2083 transpose_31" [label="(4, 12, 64, 32)", style=solid]; +"2083 transpose_31" -> "2084 reshape_70" [label="(4, 64, 12, 32)", style=solid]; +"2084 reshape_70" -> "2085 reshape_70_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"2085 reshape_70_0_0_nncf_smooth_quant_0" -> "2086 quantize_per_tensor_default_96" [label="(4, 64, 384)", style=solid]; +"2086 quantize_per_tensor_default_96" -> "2087 dequantize_per_tensor_default_96" [label="(4, 64, 384)", style=solid]; +"2087 dequantize_per_tensor_default_96" -> "2092 linear_95" [label="(4, 64, 384)", style=solid]; +"2088 linear_95_scale_0" -> "2091 dequantize_per_channel_default_96" [label="(384,)", style=solid]; +"2089 linear_95_zero_point_0" -> "2091 dequantize_per_channel_default_96" [label="(384,)", style=solid]; +"2090 _frozen_param96" -> "2091 dequantize_per_channel_default_96" [label="(384, 384)", style=solid]; +"2091 dequantize_per_channel_default_96" -> "2092 linear_95" [label="(384, 384)", style=solid]; +"2092 linear_95" -> "2093 dropout_61" [label="(4, 64, 384)", style=solid]; +"2093 dropout_61" -> "2094 view_87" [label="(4, 64, 384)", style=solid]; +"2094 view_87" -> "2095 permute_72" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2095 permute_72" -> "2096 reshape_71" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2096 reshape_71" -> "2097 roll_15" [label="(1, 16, 16, 384)", style=solid]; +"2097 roll_15" -> "2098 slice_241" [label="(1, 16, 16, 384)", style=solid]; +"2098 slice_241" -> "2099 slice_242" [label="(1, 16, 16, 384)", style=solid]; +"2099 slice_242" -> "2100 slice_243" [label="(1, 14, 16, 384)", style=solid]; +"2100 slice_243" -> "2101 slice_244" [label="(1, 14, 14, 384)", style=solid]; +"2101 slice_244" -> "2102 contiguous_29" [label="(1, 14, 14, 384)", style=solid]; +"2102 contiguous_29" -> "2103 layer_norm_33" [label="(1, 14, 14, 384)", style=solid]; +"2103 layer_norm_33" -> "2104 add_54" [label="(1, 14, 14, 384)", style=solid]; +"2104 add_54" -> "2105 add_54_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"2104 add_54" -> "2125 add_55" [label="(1, 14, 14, 384)", style=solid]; +"2105 add_54_0_0_nncf_smooth_quant_0" -> "2106 quantize_per_tensor_default_97" [label="(1, 14, 14, 384)", style=solid]; +"2106 quantize_per_tensor_default_97" -> "2107 dequantize_per_tensor_default_97" [label="(1, 14, 14, 384)", style=solid]; +"2107 dequantize_per_tensor_default_97" -> "2112 linear_96" [label="(1, 14, 14, 384)", style=solid]; +"2108 linear_96_scale_0" -> "2111 dequantize_per_channel_default_97" [label="(1536,)", style=solid]; +"2109 linear_96_zero_point_0" -> "2111 dequantize_per_channel_default_97" [label="(1536,)", style=solid]; +"2110 _frozen_param97" -> "2111 dequantize_per_channel_default_97" [label="(1536, 384)", style=solid]; +"2111 dequantize_per_channel_default_97" -> "2112 linear_96" [label="(1536, 384)", style=solid]; +"2112 linear_96" -> "2113 gelu_15" [label="(1, 14, 14, 1536)", style=solid]; +"2113 gelu_15" -> "2114 dropout_62" [label="(1, 14, 14, 1536)", style=solid]; +"2114 dropout_62" -> "2115 dropout_62_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"2115 dropout_62_0_0_nncf_smooth_quant_0" -> "2116 quantize_per_tensor_default_98" [label="(1, 14, 14, 1536)", style=solid]; +"2116 quantize_per_tensor_default_98" -> "2117 dequantize_per_tensor_default_98" [label="(1, 14, 14, 1536)", style=solid]; +"2117 dequantize_per_tensor_default_98" -> "2122 linear_97" [label="(1, 14, 14, 1536)", style=solid]; +"2118 linear_97_scale_0" -> "2121 dequantize_per_channel_default_98" [label="(384,)", style=solid]; +"2119 linear_97_zero_point_0" -> "2121 dequantize_per_channel_default_98" [label="(384,)", style=solid]; +"2120 _frozen_param98" -> "2121 dequantize_per_channel_default_98" [label="(384, 1536)", style=solid]; +"2121 dequantize_per_channel_default_98" -> "2122 linear_97" [label="(384, 1536)", style=solid]; +"2122 linear_97" -> "2123 dropout_63" [label="(1, 14, 14, 384)", style=solid]; +"2123 dropout_63" -> "2124 layer_norm_34" [label="(1, 14, 14, 384)", style=solid]; +"2124 layer_norm_34" -> "2125 add_55" [label="(1, 14, 14, 384)", style=solid]; +"2125 add_55" -> "2147 pad_18" [label="(1, 14, 14, 384)", style=solid]; +"2125 add_55" -> "2206 add_57" [label="(1, 14, 14, 384)", style=solid]; +"2126 features_5_12_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "2131 linear_98" [label="(1, 15, 15, 2)", style=solid]; +"2127 linear_98_scale_0" -> "2130 dequantize_per_channel_default_99" [label="(512,)", style=solid]; +"2128 linear_98_zero_point_0" -> "2130 dequantize_per_channel_default_99" [label="(512,)", style=solid]; +"2129 _frozen_param99" -> "2130 dequantize_per_channel_default_99" [label="(512, 2)", style=solid]; +"2130 dequantize_per_channel_default_99" -> "2131 linear_98" [label="(512, 2)", style=solid]; +"2131 linear_98" -> "2132 relu__16" [label="(1, 15, 15, 512)", style=solid]; +"2132 relu__16" -> "2133 relu__16_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"2133 relu__16_0_0_nncf_smooth_quant_0" -> "2138 linear_99" [label="(1, 15, 15, 512)", style=solid]; +"2134 linear_99_scale_0" -> "2137 dequantize_per_channel_default_100" [label="(12,)", style=solid]; +"2135 linear_99_zero_point_0" -> "2137 dequantize_per_channel_default_100" [label="(12,)", style=solid]; +"2136 _frozen_param100" -> "2137 dequantize_per_channel_default_100" [label="(12, 512)", style=solid]; +"2137 dequantize_per_channel_default_100" -> "2138 linear_99" [label="(12, 512)", style=solid]; +"2138 linear_99" -> "2139 view_88" [label="(1, 15, 15, 12)", style=solid]; +"2139 view_88" -> "2140 index_16" [label="(225, 12)", style=solid]; +"2140 index_16" -> "2141 view_89" [label="(4096, 12)", style=solid]; +"2141 view_89" -> "2142 permute_73" [label="(64, 64, 12)", style=solid]; +"2142 permute_73" -> "2143 contiguous_30" [label="(12, 64, 64)", style=solid]; +"2143 contiguous_30" -> "2144 unsqueeze_48" [label="(12, 64, 64)", style=solid]; +"2144 unsqueeze_48" -> "2145 sigmoid_16" [label="(1, 12, 64, 64)", style=solid]; +"2145 sigmoid_16" -> "2146 mul_32" [label="(1, 12, 64, 64)", style=solid]; +"2146 mul_32" -> "2182 add_56" [label="(1, 12, 64, 64)", style=solid]; +"2147 pad_18" -> "2148 view_90" [label="(1, 16, 16, 384)", style=solid]; +"2148 view_90" -> "2149 permute_74" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2149 permute_74" -> "2150 reshape_72" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2150 reshape_72" -> "2152 reshape_72_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"2151 clone_16" -> "2159 linear_100" [label="(1152,)", style=solid]; +"2152 reshape_72_0_0_nncf_smooth_quant_0" -> "2153 quantize_per_tensor_default_99" [label="(4, 64, 384)", style=solid]; +"2153 quantize_per_tensor_default_99" -> "2154 dequantize_per_tensor_default_99" [label="(4, 64, 384)", style=solid]; +"2154 dequantize_per_tensor_default_99" -> "2159 linear_100" [label="(4, 64, 384)", style=solid]; +"2155 linear_100_scale_0" -> "2158 dequantize_per_channel_default_101" [label="(1152,)", style=solid]; +"2156 linear_100_zero_point_0" -> "2158 dequantize_per_channel_default_101" [label="(1152,)", style=solid]; +"2157 _frozen_param101" -> "2158 dequantize_per_channel_default_101" [label="(1152, 384)", style=solid]; +"2158 dequantize_per_channel_default_101" -> "2159 linear_100" [label="(1152, 384)", style=solid]; +"2159 linear_100" -> "2160 reshape_73" [label="(4, 64, 1152)", style=solid]; +"2160 reshape_73" -> "2161 permute_75" [label="(4, 64, 3, 12, 32)", style=solid]; +"2161 permute_75" -> "2162 select_48" [label="(3, 4, 12, 64, 32)", style=solid]; +"2161 permute_75" -> "2163 select_49" [label="(3, 4, 12, 64, 32)", style=solid]; +"2161 permute_75" -> "2164 select_50" [label="(3, 4, 12, 64, 32)", style=solid]; +"2162 select_48" -> "2165 linalg_vector_norm_32" [label="(4, 12, 64, 32)", style=solid]; +"2162 select_48" -> "2167 expand_as_32" [label="(4, 12, 64, 32)", style=solid]; +"2162 select_48" -> "2168 div_32" [label="(4, 12, 64, 32)", style=solid]; +"2163 select_49" -> "2171 linalg_vector_norm_33" [label="(4, 12, 64, 32)", style=solid]; +"2163 select_49" -> "2173 expand_as_33" [label="(4, 12, 64, 32)", style=solid]; +"2163 select_49" -> "2174 div_33" [label="(4, 12, 64, 32)", style=solid]; +"2164 select_50" -> "2185 matmul_33" [label="(4, 12, 64, 32)", style=solid]; +"2165 linalg_vector_norm_32" -> "2166 clamp_min_32" [label="(4, 12, 64, 1)", style=solid]; +"2166 clamp_min_32" -> "2167 expand_as_32" [label="(4, 12, 64, 1)", style=solid]; +"2167 expand_as_32" -> "2168 div_32" [label="(4, 12, 64, 32)", style=solid]; +"2168 div_32" -> "2169 quantize_per_tensor_default_100" [label="(4, 12, 64, 32)", style=solid]; +"2169 quantize_per_tensor_default_100" -> "2170 dequantize_per_tensor_default_100" [label="(4, 12, 64, 32)", style=solid]; +"2170 dequantize_per_tensor_default_100" -> "2178 matmul_32" [label="(4, 12, 64, 32)", style=solid]; +"2171 linalg_vector_norm_33" -> "2172 clamp_min_33" [label="(4, 12, 64, 1)", style=solid]; +"2172 clamp_min_33" -> "2173 expand_as_33" [label="(4, 12, 64, 1)", style=solid]; +"2173 expand_as_33" -> "2174 div_33" [label="(4, 12, 64, 32)", style=solid]; +"2174 div_33" -> "2175 quantize_per_tensor_default_101" [label="(4, 12, 64, 32)", style=solid]; +"2175 quantize_per_tensor_default_101" -> "2176 dequantize_per_tensor_default_101" [label="(4, 12, 64, 32)", style=solid]; +"2176 dequantize_per_tensor_default_101" -> "2177 transpose_32" [label="(4, 12, 64, 32)", style=solid]; +"2177 transpose_32" -> "2178 matmul_32" [label="(4, 12, 32, 64)", style=solid]; +"2178 matmul_32" -> "2181 mul_33" [label="(4, 12, 64, 64)", style=solid]; +"2179 clamp_16" -> "2180 exp_16" [label="(12, 1, 1)", style=solid]; +"2180 exp_16" -> "2181 mul_33" [label="(12, 1, 1)", style=solid]; +"2181 mul_33" -> "2182 add_56" [label="(4, 12, 64, 64)", style=solid]; +"2182 add_56" -> "2183 softmax_16" [label="(4, 12, 64, 64)", style=solid]; +"2183 softmax_16" -> "2184 dropout_64" [label="(4, 12, 64, 64)", style=solid]; +"2184 dropout_64" -> "2185 matmul_33" [label="(4, 12, 64, 64)", style=solid]; +"2185 matmul_33" -> "2186 transpose_33" [label="(4, 12, 64, 32)", style=solid]; +"2186 transpose_33" -> "2187 reshape_74" [label="(4, 64, 12, 32)", style=solid]; +"2187 reshape_74" -> "2188 reshape_74_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"2188 reshape_74_0_0_nncf_smooth_quant_0" -> "2189 quantize_per_tensor_default_102" [label="(4, 64, 384)", style=solid]; +"2189 quantize_per_tensor_default_102" -> "2190 dequantize_per_tensor_default_102" [label="(4, 64, 384)", style=solid]; +"2190 dequantize_per_tensor_default_102" -> "2195 linear_101" [label="(4, 64, 384)", style=solid]; +"2191 linear_101_scale_0" -> "2194 dequantize_per_channel_default_102" [label="(384,)", style=solid]; +"2192 linear_101_zero_point_0" -> "2194 dequantize_per_channel_default_102" [label="(384,)", style=solid]; +"2193 _frozen_param102" -> "2194 dequantize_per_channel_default_102" [label="(384, 384)", style=solid]; +"2194 dequantize_per_channel_default_102" -> "2195 linear_101" [label="(384, 384)", style=solid]; +"2195 linear_101" -> "2196 dropout_65" [label="(4, 64, 384)", style=solid]; +"2196 dropout_65" -> "2197 view_91" [label="(4, 64, 384)", style=solid]; +"2197 view_91" -> "2198 permute_76" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2198 permute_76" -> "2199 reshape_75" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2199 reshape_75" -> "2200 slice_246" [label="(1, 16, 16, 384)", style=solid]; +"2200 slice_246" -> "2201 slice_247" [label="(1, 16, 16, 384)", style=solid]; +"2201 slice_247" -> "2202 slice_248" [label="(1, 14, 16, 384)", style=solid]; +"2202 slice_248" -> "2203 slice_249" [label="(1, 14, 14, 384)", style=solid]; +"2203 slice_249" -> "2204 contiguous_31" [label="(1, 14, 14, 384)", style=solid]; +"2204 contiguous_31" -> "2205 layer_norm_35" [label="(1, 14, 14, 384)", style=solid]; +"2205 layer_norm_35" -> "2206 add_57" [label="(1, 14, 14, 384)", style=solid]; +"2206 add_57" -> "2207 add_57_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"2206 add_57" -> "2227 add_58" [label="(1, 14, 14, 384)", style=solid]; +"2207 add_57_0_0_nncf_smooth_quant_0" -> "2208 quantize_per_tensor_default_103" [label="(1, 14, 14, 384)", style=solid]; +"2208 quantize_per_tensor_default_103" -> "2209 dequantize_per_tensor_default_103" [label="(1, 14, 14, 384)", style=solid]; +"2209 dequantize_per_tensor_default_103" -> "2214 linear_102" [label="(1, 14, 14, 384)", style=solid]; +"2210 linear_102_scale_0" -> "2213 dequantize_per_channel_default_103" [label="(1536,)", style=solid]; +"2211 linear_102_zero_point_0" -> "2213 dequantize_per_channel_default_103" [label="(1536,)", style=solid]; +"2212 _frozen_param103" -> "2213 dequantize_per_channel_default_103" [label="(1536, 384)", style=solid]; +"2213 dequantize_per_channel_default_103" -> "2214 linear_102" [label="(1536, 384)", style=solid]; +"2214 linear_102" -> "2215 gelu_16" [label="(1, 14, 14, 1536)", style=solid]; +"2215 gelu_16" -> "2216 dropout_66" [label="(1, 14, 14, 1536)", style=solid]; +"2216 dropout_66" -> "2217 dropout_66_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"2217 dropout_66_0_0_nncf_smooth_quant_0" -> "2218 quantize_per_tensor_default_104" [label="(1, 14, 14, 1536)", style=solid]; +"2218 quantize_per_tensor_default_104" -> "2219 dequantize_per_tensor_default_104" [label="(1, 14, 14, 1536)", style=solid]; +"2219 dequantize_per_tensor_default_104" -> "2224 linear_103" [label="(1, 14, 14, 1536)", style=solid]; +"2220 linear_103_scale_0" -> "2223 dequantize_per_channel_default_104" [label="(384,)", style=solid]; +"2221 linear_103_zero_point_0" -> "2223 dequantize_per_channel_default_104" [label="(384,)", style=solid]; +"2222 _frozen_param104" -> "2223 dequantize_per_channel_default_104" [label="(384, 1536)", style=solid]; +"2223 dequantize_per_channel_default_104" -> "2224 linear_103" [label="(384, 1536)", style=solid]; +"2224 linear_103" -> "2225 dropout_67" [label="(1, 14, 14, 384)", style=solid]; +"2225 dropout_67" -> "2226 layer_norm_36" [label="(1, 14, 14, 384)", style=solid]; +"2226 layer_norm_36" -> "2227 add_58" [label="(1, 14, 14, 384)", style=solid]; +"2227 add_58" -> "2249 pad_19" [label="(1, 14, 14, 384)", style=solid]; +"2227 add_58" -> "2326 add_61" [label="(1, 14, 14, 384)", style=solid]; +"2228 features_5_13_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "2233 linear_104" [label="(1, 15, 15, 2)", style=solid]; +"2229 linear_104_scale_0" -> "2232 dequantize_per_channel_default_105" [label="(512,)", style=solid]; +"2230 linear_104_zero_point_0" -> "2232 dequantize_per_channel_default_105" [label="(512,)", style=solid]; +"2231 _frozen_param105" -> "2232 dequantize_per_channel_default_105" [label="(512, 2)", style=solid]; +"2232 dequantize_per_channel_default_105" -> "2233 linear_104" [label="(512, 2)", style=solid]; +"2233 linear_104" -> "2234 relu__17" [label="(1, 15, 15, 512)", style=solid]; +"2234 relu__17" -> "2235 relu__17_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"2235 relu__17_0_0_nncf_smooth_quant_0" -> "2240 linear_105" [label="(1, 15, 15, 512)", style=solid]; +"2236 linear_105_scale_0" -> "2239 dequantize_per_channel_default_106" [label="(12,)", style=solid]; +"2237 linear_105_zero_point_0" -> "2239 dequantize_per_channel_default_106" [label="(12,)", style=solid]; +"2238 _frozen_param106" -> "2239 dequantize_per_channel_default_106" [label="(12, 512)", style=solid]; +"2239 dequantize_per_channel_default_106" -> "2240 linear_105" [label="(12, 512)", style=solid]; +"2240 linear_105" -> "2241 view_92" [label="(1, 15, 15, 12)", style=solid]; +"2241 view_92" -> "2242 index_17" [label="(225, 12)", style=solid]; +"2242 index_17" -> "2243 view_93" [label="(4096, 12)", style=solid]; +"2243 view_93" -> "2244 permute_77" [label="(64, 64, 12)", style=solid]; +"2244 permute_77" -> "2245 contiguous_32" [label="(12, 64, 64)", style=solid]; +"2245 contiguous_32" -> "2246 unsqueeze_49" [label="(12, 64, 64)", style=solid]; +"2246 unsqueeze_49" -> "2247 sigmoid_17" [label="(1, 12, 64, 64)", style=solid]; +"2247 sigmoid_17" -> "2248 mul_34" [label="(1, 12, 64, 64)", style=solid]; +"2248 mul_34" -> "2285 add_59" [label="(1, 12, 64, 64)", style=solid]; +"2249 pad_19" -> "2250 roll_16" [label="(1, 16, 16, 384)", style=solid]; +"2250 roll_16" -> "2251 view_94" [label="(1, 16, 16, 384)", style=solid]; +"2251 view_94" -> "2252 permute_78" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2252 permute_78" -> "2253 reshape_76" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2253 reshape_76" -> "2255 reshape_76_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"2253 reshape_76" -> "2286 new_zeros_8" [label="(4, 64, 384)", style=solid]; +"2254 clone_17" -> "2262 linear_106" [label="(1152,)", style=solid]; +"2255 reshape_76_0_0_nncf_smooth_quant_0" -> "2256 quantize_per_tensor_default_105" [label="(4, 64, 384)", style=solid]; +"2256 quantize_per_tensor_default_105" -> "2257 dequantize_per_tensor_default_105" [label="(4, 64, 384)", style=solid]; +"2257 dequantize_per_tensor_default_105" -> "2262 linear_106" [label="(4, 64, 384)", style=solid]; +"2258 linear_106_scale_0" -> "2261 dequantize_per_channel_default_107" [label="(1152,)", style=solid]; +"2259 linear_106_zero_point_0" -> "2261 dequantize_per_channel_default_107" [label="(1152,)", style=solid]; +"2260 _frozen_param107" -> "2261 dequantize_per_channel_default_107" [label="(1152, 384)", style=solid]; +"2261 dequantize_per_channel_default_107" -> "2262 linear_106" [label="(1152, 384)", style=solid]; +"2262 linear_106" -> "2263 reshape_77" [label="(4, 64, 1152)", style=solid]; +"2263 reshape_77" -> "2264 permute_79" [label="(4, 64, 3, 12, 32)", style=solid]; +"2264 permute_79" -> "2265 select_51" [label="(3, 4, 12, 64, 32)", style=solid]; +"2264 permute_79" -> "2266 select_52" [label="(3, 4, 12, 64, 32)", style=solid]; +"2264 permute_79" -> "2267 select_53" [label="(3, 4, 12, 64, 32)", style=solid]; +"2265 select_51" -> "2268 linalg_vector_norm_34" [label="(4, 12, 64, 32)", style=solid]; +"2265 select_51" -> "2270 expand_as_34" [label="(4, 12, 64, 32)", style=solid]; +"2265 select_51" -> "2271 div_34" [label="(4, 12, 64, 32)", style=solid]; +"2266 select_52" -> "2274 linalg_vector_norm_35" [label="(4, 12, 64, 32)", style=solid]; +"2266 select_52" -> "2276 expand_as_35" [label="(4, 12, 64, 32)", style=solid]; +"2266 select_52" -> "2277 div_35" [label="(4, 12, 64, 32)", style=solid]; +"2267 select_53" -> "2304 matmul_35" [label="(4, 12, 64, 32)", style=solid]; +"2268 linalg_vector_norm_34" -> "2269 clamp_min_34" [label="(4, 12, 64, 1)", style=solid]; +"2269 clamp_min_34" -> "2270 expand_as_34" [label="(4, 12, 64, 1)", style=solid]; +"2270 expand_as_34" -> "2271 div_34" [label="(4, 12, 64, 32)", style=solid]; +"2271 div_34" -> "2272 quantize_per_tensor_default_106" [label="(4, 12, 64, 32)", style=solid]; +"2272 quantize_per_tensor_default_106" -> "2273 dequantize_per_tensor_default_106" [label="(4, 12, 64, 32)", style=solid]; +"2273 dequantize_per_tensor_default_106" -> "2281 matmul_34" [label="(4, 12, 64, 32)", style=solid]; +"2274 linalg_vector_norm_35" -> "2275 clamp_min_35" [label="(4, 12, 64, 1)", style=solid]; +"2275 clamp_min_35" -> "2276 expand_as_35" [label="(4, 12, 64, 1)", style=solid]; +"2276 expand_as_35" -> "2277 div_35" [label="(4, 12, 64, 32)", style=solid]; +"2277 div_35" -> "2278 quantize_per_tensor_default_107" [label="(4, 12, 64, 32)", style=solid]; +"2278 quantize_per_tensor_default_107" -> "2279 dequantize_per_tensor_default_107" [label="(4, 12, 64, 32)", style=solid]; +"2279 dequantize_per_tensor_default_107" -> "2280 transpose_34" [label="(4, 12, 64, 32)", style=solid]; +"2280 transpose_34" -> "2281 matmul_34" [label="(4, 12, 32, 64)", style=solid]; +"2281 matmul_34" -> "2284 mul_35" [label="(4, 12, 64, 64)", style=solid]; +"2282 clamp_17" -> "2283 exp_17" [label="(12, 1, 1)", style=solid]; +"2283 exp_17" -> "2284 mul_35" [label="(12, 1, 1)", style=solid]; +"2284 mul_35" -> "2285 add_59" [label="(4, 12, 64, 64)", style=solid]; +"2285 add_59" -> "2297 view_96" [label="(4, 12, 64, 64)", style=solid]; +"2286 new_zeros_8" -> "2287 view_95" [label="(16, 16)", style=solid]; +"2287 view_95" -> "2288 permute_80" [label="(2, 8, 2, 8)", style=solid]; +"2288 permute_80" -> "2289 reshape_78" [label="(2, 2, 8, 8)", style=solid]; +"2289 reshape_78" -> "2290 unsqueeze_50" [label="(4, 64)", style=solid]; +"2289 reshape_78" -> "2291 unsqueeze_51" [label="(4, 64)", style=solid]; +"2290 unsqueeze_50" -> "2292 sub_8" [label="(4, 1, 64)", style=solid]; +"2291 unsqueeze_51" -> "2292 sub_8" [label="(4, 64, 1)", style=solid]; +"2292 sub_8" -> "2293 ne_8" [label="(4, 64, 64)", style=solid]; +"2292 sub_8" -> "2294 masked_fill_16" [label="(4, 64, 64)", style=solid]; +"2292 sub_8" -> "2295 eq_8" [label="(4, 64, 64)", style=solid]; +"2293 ne_8" -> "2294 masked_fill_16" [label="(4, 64, 64)", style=solid]; +"2294 masked_fill_16" -> "2296 masked_fill_17" [label="(4, 64, 64)", style=solid]; +"2295 eq_8" -> "2296 masked_fill_17" [label="(4, 64, 64)", style=solid]; +"2296 masked_fill_17" -> "2298 unsqueeze_52" [label="(4, 64, 64)", style=solid]; +"2297 view_96" -> "2300 add_60" [label="(1, 4, 12, 64, 64)", style=solid]; +"2298 unsqueeze_52" -> "2299 unsqueeze_53" [label="(4, 1, 64, 64)", style=solid]; +"2299 unsqueeze_53" -> "2300 add_60" [label="(1, 4, 1, 64, 64)", style=solid]; +"2300 add_60" -> "2301 view_97" [label="(1, 4, 12, 64, 64)", style=solid]; +"2301 view_97" -> "2302 softmax_17" [label="(4, 12, 64, 64)", style=solid]; +"2302 softmax_17" -> "2303 dropout_68" [label="(4, 12, 64, 64)", style=solid]; +"2303 dropout_68" -> "2304 matmul_35" [label="(4, 12, 64, 64)", style=solid]; +"2304 matmul_35" -> "2305 transpose_35" [label="(4, 12, 64, 32)", style=solid]; +"2305 transpose_35" -> "2306 reshape_79" [label="(4, 64, 12, 32)", style=solid]; +"2306 reshape_79" -> "2307 reshape_79_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"2307 reshape_79_0_0_nncf_smooth_quant_0" -> "2308 quantize_per_tensor_default_108" [label="(4, 64, 384)", style=solid]; +"2308 quantize_per_tensor_default_108" -> "2309 dequantize_per_tensor_default_108" [label="(4, 64, 384)", style=solid]; +"2309 dequantize_per_tensor_default_108" -> "2314 linear_107" [label="(4, 64, 384)", style=solid]; +"2310 linear_107_scale_0" -> "2313 dequantize_per_channel_default_108" [label="(384,)", style=solid]; +"2311 linear_107_zero_point_0" -> "2313 dequantize_per_channel_default_108" [label="(384,)", style=solid]; +"2312 _frozen_param108" -> "2313 dequantize_per_channel_default_108" [label="(384, 384)", style=solid]; +"2313 dequantize_per_channel_default_108" -> "2314 linear_107" [label="(384, 384)", style=solid]; +"2314 linear_107" -> "2315 dropout_69" [label="(4, 64, 384)", style=solid]; +"2315 dropout_69" -> "2316 view_98" [label="(4, 64, 384)", style=solid]; +"2316 view_98" -> "2317 permute_81" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2317 permute_81" -> "2318 reshape_80" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2318 reshape_80" -> "2319 roll_17" [label="(1, 16, 16, 384)", style=solid]; +"2319 roll_17" -> "2320 slice_269" [label="(1, 16, 16, 384)", style=solid]; +"2320 slice_269" -> "2321 slice_270" [label="(1, 16, 16, 384)", style=solid]; +"2321 slice_270" -> "2322 slice_271" [label="(1, 14, 16, 384)", style=solid]; +"2322 slice_271" -> "2323 slice_272" [label="(1, 14, 14, 384)", style=solid]; +"2323 slice_272" -> "2324 contiguous_33" [label="(1, 14, 14, 384)", style=solid]; +"2324 contiguous_33" -> "2325 layer_norm_37" [label="(1, 14, 14, 384)", style=solid]; +"2325 layer_norm_37" -> "2326 add_61" [label="(1, 14, 14, 384)", style=solid]; +"2326 add_61" -> "2327 add_61_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"2326 add_61" -> "2347 add_62" [label="(1, 14, 14, 384)", style=solid]; +"2327 add_61_0_0_nncf_smooth_quant_0" -> "2328 quantize_per_tensor_default_109" [label="(1, 14, 14, 384)", style=solid]; +"2328 quantize_per_tensor_default_109" -> "2329 dequantize_per_tensor_default_109" [label="(1, 14, 14, 384)", style=solid]; +"2329 dequantize_per_tensor_default_109" -> "2334 linear_108" [label="(1, 14, 14, 384)", style=solid]; +"2330 linear_108_scale_0" -> "2333 dequantize_per_channel_default_109" [label="(1536,)", style=solid]; +"2331 linear_108_zero_point_0" -> "2333 dequantize_per_channel_default_109" [label="(1536,)", style=solid]; +"2332 _frozen_param109" -> "2333 dequantize_per_channel_default_109" [label="(1536, 384)", style=solid]; +"2333 dequantize_per_channel_default_109" -> "2334 linear_108" [label="(1536, 384)", style=solid]; +"2334 linear_108" -> "2335 gelu_17" [label="(1, 14, 14, 1536)", style=solid]; +"2335 gelu_17" -> "2336 dropout_70" [label="(1, 14, 14, 1536)", style=solid]; +"2336 dropout_70" -> "2337 dropout_70_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"2337 dropout_70_0_0_nncf_smooth_quant_0" -> "2338 quantize_per_tensor_default_110" [label="(1, 14, 14, 1536)", style=solid]; +"2338 quantize_per_tensor_default_110" -> "2339 dequantize_per_tensor_default_110" [label="(1, 14, 14, 1536)", style=solid]; +"2339 dequantize_per_tensor_default_110" -> "2344 linear_109" [label="(1, 14, 14, 1536)", style=solid]; +"2340 linear_109_scale_0" -> "2343 dequantize_per_channel_default_110" [label="(384,)", style=solid]; +"2341 linear_109_zero_point_0" -> "2343 dequantize_per_channel_default_110" [label="(384,)", style=solid]; +"2342 _frozen_param110" -> "2343 dequantize_per_channel_default_110" [label="(384, 1536)", style=solid]; +"2343 dequantize_per_channel_default_110" -> "2344 linear_109" [label="(384, 1536)", style=solid]; +"2344 linear_109" -> "2345 dropout_71" [label="(1, 14, 14, 384)", style=solid]; +"2345 dropout_71" -> "2346 layer_norm_38" [label="(1, 14, 14, 384)", style=solid]; +"2346 layer_norm_38" -> "2347 add_62" [label="(1, 14, 14, 384)", style=solid]; +"2347 add_62" -> "2369 pad_20" [label="(1, 14, 14, 384)", style=solid]; +"2347 add_62" -> "2428 add_64" [label="(1, 14, 14, 384)", style=solid]; +"2348 features_5_14_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "2353 linear_110" [label="(1, 15, 15, 2)", style=solid]; +"2349 linear_110_scale_0" -> "2352 dequantize_per_channel_default_111" [label="(512,)", style=solid]; +"2350 linear_110_zero_point_0" -> "2352 dequantize_per_channel_default_111" [label="(512,)", style=solid]; +"2351 _frozen_param111" -> "2352 dequantize_per_channel_default_111" [label="(512, 2)", style=solid]; +"2352 dequantize_per_channel_default_111" -> "2353 linear_110" [label="(512, 2)", style=solid]; +"2353 linear_110" -> "2354 relu__18" [label="(1, 15, 15, 512)", style=solid]; +"2354 relu__18" -> "2355 relu__18_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"2355 relu__18_0_0_nncf_smooth_quant_0" -> "2360 linear_111" [label="(1, 15, 15, 512)", style=solid]; +"2356 linear_111_scale_0" -> "2359 dequantize_per_channel_default_112" [label="(12,)", style=solid]; +"2357 linear_111_zero_point_0" -> "2359 dequantize_per_channel_default_112" [label="(12,)", style=solid]; +"2358 _frozen_param112" -> "2359 dequantize_per_channel_default_112" [label="(12, 512)", style=solid]; +"2359 dequantize_per_channel_default_112" -> "2360 linear_111" [label="(12, 512)", style=solid]; +"2360 linear_111" -> "2361 view_99" [label="(1, 15, 15, 12)", style=solid]; +"2361 view_99" -> "2362 index_18" [label="(225, 12)", style=solid]; +"2362 index_18" -> "2363 view_100" [label="(4096, 12)", style=solid]; +"2363 view_100" -> "2364 permute_82" [label="(64, 64, 12)", style=solid]; +"2364 permute_82" -> "2365 contiguous_34" [label="(12, 64, 64)", style=solid]; +"2365 contiguous_34" -> "2366 unsqueeze_54" [label="(12, 64, 64)", style=solid]; +"2366 unsqueeze_54" -> "2367 sigmoid_18" [label="(1, 12, 64, 64)", style=solid]; +"2367 sigmoid_18" -> "2368 mul_36" [label="(1, 12, 64, 64)", style=solid]; +"2368 mul_36" -> "2404 add_63" [label="(1, 12, 64, 64)", style=solid]; +"2369 pad_20" -> "2370 view_101" [label="(1, 16, 16, 384)", style=solid]; +"2370 view_101" -> "2371 permute_83" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2371 permute_83" -> "2372 reshape_81" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2372 reshape_81" -> "2374 reshape_81_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"2373 clone_18" -> "2381 linear_112" [label="(1152,)", style=solid]; +"2374 reshape_81_0_0_nncf_smooth_quant_0" -> "2375 quantize_per_tensor_default_111" [label="(4, 64, 384)", style=solid]; +"2375 quantize_per_tensor_default_111" -> "2376 dequantize_per_tensor_default_111" [label="(4, 64, 384)", style=solid]; +"2376 dequantize_per_tensor_default_111" -> "2381 linear_112" [label="(4, 64, 384)", style=solid]; +"2377 linear_112_scale_0" -> "2380 dequantize_per_channel_default_113" [label="(1152,)", style=solid]; +"2378 linear_112_zero_point_0" -> "2380 dequantize_per_channel_default_113" [label="(1152,)", style=solid]; +"2379 _frozen_param113" -> "2380 dequantize_per_channel_default_113" [label="(1152, 384)", style=solid]; +"2380 dequantize_per_channel_default_113" -> "2381 linear_112" [label="(1152, 384)", style=solid]; +"2381 linear_112" -> "2382 reshape_82" [label="(4, 64, 1152)", style=solid]; +"2382 reshape_82" -> "2383 permute_84" [label="(4, 64, 3, 12, 32)", style=solid]; +"2383 permute_84" -> "2384 select_54" [label="(3, 4, 12, 64, 32)", style=solid]; +"2383 permute_84" -> "2385 select_55" [label="(3, 4, 12, 64, 32)", style=solid]; +"2383 permute_84" -> "2386 select_56" [label="(3, 4, 12, 64, 32)", style=solid]; +"2384 select_54" -> "2387 linalg_vector_norm_36" [label="(4, 12, 64, 32)", style=solid]; +"2384 select_54" -> "2389 expand_as_36" [label="(4, 12, 64, 32)", style=solid]; +"2384 select_54" -> "2390 div_36" [label="(4, 12, 64, 32)", style=solid]; +"2385 select_55" -> "2393 linalg_vector_norm_37" [label="(4, 12, 64, 32)", style=solid]; +"2385 select_55" -> "2395 expand_as_37" [label="(4, 12, 64, 32)", style=solid]; +"2385 select_55" -> "2396 div_37" [label="(4, 12, 64, 32)", style=solid]; +"2386 select_56" -> "2407 matmul_37" [label="(4, 12, 64, 32)", style=solid]; +"2387 linalg_vector_norm_36" -> "2388 clamp_min_36" [label="(4, 12, 64, 1)", style=solid]; +"2388 clamp_min_36" -> "2389 expand_as_36" [label="(4, 12, 64, 1)", style=solid]; +"2389 expand_as_36" -> "2390 div_36" [label="(4, 12, 64, 32)", style=solid]; +"2390 div_36" -> "2391 quantize_per_tensor_default_112" [label="(4, 12, 64, 32)", style=solid]; +"2391 quantize_per_tensor_default_112" -> "2392 dequantize_per_tensor_default_112" [label="(4, 12, 64, 32)", style=solid]; +"2392 dequantize_per_tensor_default_112" -> "2400 matmul_36" [label="(4, 12, 64, 32)", style=solid]; +"2393 linalg_vector_norm_37" -> "2394 clamp_min_37" [label="(4, 12, 64, 1)", style=solid]; +"2394 clamp_min_37" -> "2395 expand_as_37" [label="(4, 12, 64, 1)", style=solid]; +"2395 expand_as_37" -> "2396 div_37" [label="(4, 12, 64, 32)", style=solid]; +"2396 div_37" -> "2397 quantize_per_tensor_default_113" [label="(4, 12, 64, 32)", style=solid]; +"2397 quantize_per_tensor_default_113" -> "2398 dequantize_per_tensor_default_113" [label="(4, 12, 64, 32)", style=solid]; +"2398 dequantize_per_tensor_default_113" -> "2399 transpose_36" [label="(4, 12, 64, 32)", style=solid]; +"2399 transpose_36" -> "2400 matmul_36" [label="(4, 12, 32, 64)", style=solid]; +"2400 matmul_36" -> "2403 mul_37" [label="(4, 12, 64, 64)", style=solid]; +"2401 clamp_18" -> "2402 exp_18" [label="(12, 1, 1)", style=solid]; +"2402 exp_18" -> "2403 mul_37" [label="(12, 1, 1)", style=solid]; +"2403 mul_37" -> "2404 add_63" [label="(4, 12, 64, 64)", style=solid]; +"2404 add_63" -> "2405 softmax_18" [label="(4, 12, 64, 64)", style=solid]; +"2405 softmax_18" -> "2406 dropout_72" [label="(4, 12, 64, 64)", style=solid]; +"2406 dropout_72" -> "2407 matmul_37" [label="(4, 12, 64, 64)", style=solid]; +"2407 matmul_37" -> "2408 transpose_37" [label="(4, 12, 64, 32)", style=solid]; +"2408 transpose_37" -> "2409 reshape_83" [label="(4, 64, 12, 32)", style=solid]; +"2409 reshape_83" -> "2410 reshape_83_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"2410 reshape_83_0_0_nncf_smooth_quant_0" -> "2411 quantize_per_tensor_default_114" [label="(4, 64, 384)", style=solid]; +"2411 quantize_per_tensor_default_114" -> "2412 dequantize_per_tensor_default_114" [label="(4, 64, 384)", style=solid]; +"2412 dequantize_per_tensor_default_114" -> "2417 linear_113" [label="(4, 64, 384)", style=solid]; +"2413 linear_113_scale_0" -> "2416 dequantize_per_channel_default_114" [label="(384,)", style=solid]; +"2414 linear_113_zero_point_0" -> "2416 dequantize_per_channel_default_114" [label="(384,)", style=solid]; +"2415 _frozen_param114" -> "2416 dequantize_per_channel_default_114" [label="(384, 384)", style=solid]; +"2416 dequantize_per_channel_default_114" -> "2417 linear_113" [label="(384, 384)", style=solid]; +"2417 linear_113" -> "2418 dropout_73" [label="(4, 64, 384)", style=solid]; +"2418 dropout_73" -> "2419 view_102" [label="(4, 64, 384)", style=solid]; +"2419 view_102" -> "2420 permute_85" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2420 permute_85" -> "2421 reshape_84" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2421 reshape_84" -> "2422 slice_274" [label="(1, 16, 16, 384)", style=solid]; +"2422 slice_274" -> "2423 slice_275" [label="(1, 16, 16, 384)", style=solid]; +"2423 slice_275" -> "2424 slice_276" [label="(1, 14, 16, 384)", style=solid]; +"2424 slice_276" -> "2425 slice_277" [label="(1, 14, 14, 384)", style=solid]; +"2425 slice_277" -> "2426 contiguous_35" [label="(1, 14, 14, 384)", style=solid]; +"2426 contiguous_35" -> "2427 layer_norm_39" [label="(1, 14, 14, 384)", style=solid]; +"2427 layer_norm_39" -> "2428 add_64" [label="(1, 14, 14, 384)", style=solid]; +"2428 add_64" -> "2429 add_64_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"2428 add_64" -> "2449 add_65" [label="(1, 14, 14, 384)", style=solid]; +"2429 add_64_0_0_nncf_smooth_quant_0" -> "2430 quantize_per_tensor_default_115" [label="(1, 14, 14, 384)", style=solid]; +"2430 quantize_per_tensor_default_115" -> "2431 dequantize_per_tensor_default_115" [label="(1, 14, 14, 384)", style=solid]; +"2431 dequantize_per_tensor_default_115" -> "2436 linear_114" [label="(1, 14, 14, 384)", style=solid]; +"2432 linear_114_scale_0" -> "2435 dequantize_per_channel_default_115" [label="(1536,)", style=solid]; +"2433 linear_114_zero_point_0" -> "2435 dequantize_per_channel_default_115" [label="(1536,)", style=solid]; +"2434 _frozen_param115" -> "2435 dequantize_per_channel_default_115" [label="(1536, 384)", style=solid]; +"2435 dequantize_per_channel_default_115" -> "2436 linear_114" [label="(1536, 384)", style=solid]; +"2436 linear_114" -> "2437 gelu_18" [label="(1, 14, 14, 1536)", style=solid]; +"2437 gelu_18" -> "2438 dropout_74" [label="(1, 14, 14, 1536)", style=solid]; +"2438 dropout_74" -> "2439 dropout_74_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"2439 dropout_74_0_0_nncf_smooth_quant_0" -> "2440 quantize_per_tensor_default_116" [label="(1, 14, 14, 1536)", style=solid]; +"2440 quantize_per_tensor_default_116" -> "2441 dequantize_per_tensor_default_116" [label="(1, 14, 14, 1536)", style=solid]; +"2441 dequantize_per_tensor_default_116" -> "2446 linear_115" [label="(1, 14, 14, 1536)", style=solid]; +"2442 linear_115_scale_0" -> "2445 dequantize_per_channel_default_116" [label="(384,)", style=solid]; +"2443 linear_115_zero_point_0" -> "2445 dequantize_per_channel_default_116" [label="(384,)", style=solid]; +"2444 _frozen_param116" -> "2445 dequantize_per_channel_default_116" [label="(384, 1536)", style=solid]; +"2445 dequantize_per_channel_default_116" -> "2446 linear_115" [label="(384, 1536)", style=solid]; +"2446 linear_115" -> "2447 dropout_75" [label="(1, 14, 14, 384)", style=solid]; +"2447 dropout_75" -> "2448 layer_norm_40" [label="(1, 14, 14, 384)", style=solid]; +"2448 layer_norm_40" -> "2449 add_65" [label="(1, 14, 14, 384)", style=solid]; +"2449 add_65" -> "2471 pad_21" [label="(1, 14, 14, 384)", style=solid]; +"2449 add_65" -> "2548 add_68" [label="(1, 14, 14, 384)", style=solid]; +"2450 features_5_15_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "2455 linear_116" [label="(1, 15, 15, 2)", style=solid]; +"2451 linear_116_scale_0" -> "2454 dequantize_per_channel_default_117" [label="(512,)", style=solid]; +"2452 linear_116_zero_point_0" -> "2454 dequantize_per_channel_default_117" [label="(512,)", style=solid]; +"2453 _frozen_param117" -> "2454 dequantize_per_channel_default_117" [label="(512, 2)", style=solid]; +"2454 dequantize_per_channel_default_117" -> "2455 linear_116" [label="(512, 2)", style=solid]; +"2455 linear_116" -> "2456 relu__19" [label="(1, 15, 15, 512)", style=solid]; +"2456 relu__19" -> "2457 relu__19_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"2457 relu__19_0_0_nncf_smooth_quant_0" -> "2462 linear_117" [label="(1, 15, 15, 512)", style=solid]; +"2458 linear_117_scale_0" -> "2461 dequantize_per_channel_default_118" [label="(12,)", style=solid]; +"2459 linear_117_zero_point_0" -> "2461 dequantize_per_channel_default_118" [label="(12,)", style=solid]; +"2460 _frozen_param118" -> "2461 dequantize_per_channel_default_118" [label="(12, 512)", style=solid]; +"2461 dequantize_per_channel_default_118" -> "2462 linear_117" [label="(12, 512)", style=solid]; +"2462 linear_117" -> "2463 view_103" [label="(1, 15, 15, 12)", style=solid]; +"2463 view_103" -> "2464 index_19" [label="(225, 12)", style=solid]; +"2464 index_19" -> "2465 view_104" [label="(4096, 12)", style=solid]; +"2465 view_104" -> "2466 permute_86" [label="(64, 64, 12)", style=solid]; +"2466 permute_86" -> "2467 contiguous_36" [label="(12, 64, 64)", style=solid]; +"2467 contiguous_36" -> "2468 unsqueeze_55" [label="(12, 64, 64)", style=solid]; +"2468 unsqueeze_55" -> "2469 sigmoid_19" [label="(1, 12, 64, 64)", style=solid]; +"2469 sigmoid_19" -> "2470 mul_38" [label="(1, 12, 64, 64)", style=solid]; +"2470 mul_38" -> "2507 add_66" [label="(1, 12, 64, 64)", style=solid]; +"2471 pad_21" -> "2472 roll_18" [label="(1, 16, 16, 384)", style=solid]; +"2472 roll_18" -> "2473 view_105" [label="(1, 16, 16, 384)", style=solid]; +"2473 view_105" -> "2474 permute_87" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2474 permute_87" -> "2475 reshape_85" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2475 reshape_85" -> "2477 reshape_85_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"2475 reshape_85" -> "2508 new_zeros_9" [label="(4, 64, 384)", style=solid]; +"2476 clone_19" -> "2484 linear_118" [label="(1152,)", style=solid]; +"2477 reshape_85_0_0_nncf_smooth_quant_0" -> "2478 quantize_per_tensor_default_117" [label="(4, 64, 384)", style=solid]; +"2478 quantize_per_tensor_default_117" -> "2479 dequantize_per_tensor_default_117" [label="(4, 64, 384)", style=solid]; +"2479 dequantize_per_tensor_default_117" -> "2484 linear_118" [label="(4, 64, 384)", style=solid]; +"2480 linear_118_scale_0" -> "2483 dequantize_per_channel_default_119" [label="(1152,)", style=solid]; +"2481 linear_118_zero_point_0" -> "2483 dequantize_per_channel_default_119" [label="(1152,)", style=solid]; +"2482 _frozen_param119" -> "2483 dequantize_per_channel_default_119" [label="(1152, 384)", style=solid]; +"2483 dequantize_per_channel_default_119" -> "2484 linear_118" [label="(1152, 384)", style=solid]; +"2484 linear_118" -> "2485 reshape_86" [label="(4, 64, 1152)", style=solid]; +"2485 reshape_86" -> "2486 permute_88" [label="(4, 64, 3, 12, 32)", style=solid]; +"2486 permute_88" -> "2487 select_57" [label="(3, 4, 12, 64, 32)", style=solid]; +"2486 permute_88" -> "2488 select_58" [label="(3, 4, 12, 64, 32)", style=solid]; +"2486 permute_88" -> "2489 select_59" [label="(3, 4, 12, 64, 32)", style=solid]; +"2487 select_57" -> "2490 linalg_vector_norm_38" [label="(4, 12, 64, 32)", style=solid]; +"2487 select_57" -> "2492 expand_as_38" [label="(4, 12, 64, 32)", style=solid]; +"2487 select_57" -> "2493 div_38" [label="(4, 12, 64, 32)", style=solid]; +"2488 select_58" -> "2496 linalg_vector_norm_39" [label="(4, 12, 64, 32)", style=solid]; +"2488 select_58" -> "2498 expand_as_39" [label="(4, 12, 64, 32)", style=solid]; +"2488 select_58" -> "2499 div_39" [label="(4, 12, 64, 32)", style=solid]; +"2489 select_59" -> "2526 matmul_39" [label="(4, 12, 64, 32)", style=solid]; +"2490 linalg_vector_norm_38" -> "2491 clamp_min_38" [label="(4, 12, 64, 1)", style=solid]; +"2491 clamp_min_38" -> "2492 expand_as_38" [label="(4, 12, 64, 1)", style=solid]; +"2492 expand_as_38" -> "2493 div_38" [label="(4, 12, 64, 32)", style=solid]; +"2493 div_38" -> "2494 quantize_per_tensor_default_118" [label="(4, 12, 64, 32)", style=solid]; +"2494 quantize_per_tensor_default_118" -> "2495 dequantize_per_tensor_default_118" [label="(4, 12, 64, 32)", style=solid]; +"2495 dequantize_per_tensor_default_118" -> "2503 matmul_38" [label="(4, 12, 64, 32)", style=solid]; +"2496 linalg_vector_norm_39" -> "2497 clamp_min_39" [label="(4, 12, 64, 1)", style=solid]; +"2497 clamp_min_39" -> "2498 expand_as_39" [label="(4, 12, 64, 1)", style=solid]; +"2498 expand_as_39" -> "2499 div_39" [label="(4, 12, 64, 32)", style=solid]; +"2499 div_39" -> "2500 quantize_per_tensor_default_119" [label="(4, 12, 64, 32)", style=solid]; +"2500 quantize_per_tensor_default_119" -> "2501 dequantize_per_tensor_default_119" [label="(4, 12, 64, 32)", style=solid]; +"2501 dequantize_per_tensor_default_119" -> "2502 transpose_38" [label="(4, 12, 64, 32)", style=solid]; +"2502 transpose_38" -> "2503 matmul_38" [label="(4, 12, 32, 64)", style=solid]; +"2503 matmul_38" -> "2506 mul_39" [label="(4, 12, 64, 64)", style=solid]; +"2504 clamp_19" -> "2505 exp_19" [label="(12, 1, 1)", style=solid]; +"2505 exp_19" -> "2506 mul_39" [label="(12, 1, 1)", style=solid]; +"2506 mul_39" -> "2507 add_66" [label="(4, 12, 64, 64)", style=solid]; +"2507 add_66" -> "2519 view_107" [label="(4, 12, 64, 64)", style=solid]; +"2508 new_zeros_9" -> "2509 view_106" [label="(16, 16)", style=solid]; +"2509 view_106" -> "2510 permute_89" [label="(2, 8, 2, 8)", style=solid]; +"2510 permute_89" -> "2511 reshape_87" [label="(2, 2, 8, 8)", style=solid]; +"2511 reshape_87" -> "2512 unsqueeze_56" [label="(4, 64)", style=solid]; +"2511 reshape_87" -> "2513 unsqueeze_57" [label="(4, 64)", style=solid]; +"2512 unsqueeze_56" -> "2514 sub_9" [label="(4, 1, 64)", style=solid]; +"2513 unsqueeze_57" -> "2514 sub_9" [label="(4, 64, 1)", style=solid]; +"2514 sub_9" -> "2515 ne_9" [label="(4, 64, 64)", style=solid]; +"2514 sub_9" -> "2516 masked_fill_18" [label="(4, 64, 64)", style=solid]; +"2514 sub_9" -> "2517 eq_9" [label="(4, 64, 64)", style=solid]; +"2515 ne_9" -> "2516 masked_fill_18" [label="(4, 64, 64)", style=solid]; +"2516 masked_fill_18" -> "2518 masked_fill_19" [label="(4, 64, 64)", style=solid]; +"2517 eq_9" -> "2518 masked_fill_19" [label="(4, 64, 64)", style=solid]; +"2518 masked_fill_19" -> "2520 unsqueeze_58" [label="(4, 64, 64)", style=solid]; +"2519 view_107" -> "2522 add_67" [label="(1, 4, 12, 64, 64)", style=solid]; +"2520 unsqueeze_58" -> "2521 unsqueeze_59" [label="(4, 1, 64, 64)", style=solid]; +"2521 unsqueeze_59" -> "2522 add_67" [label="(1, 4, 1, 64, 64)", style=solid]; +"2522 add_67" -> "2523 view_108" [label="(1, 4, 12, 64, 64)", style=solid]; +"2523 view_108" -> "2524 softmax_19" [label="(4, 12, 64, 64)", style=solid]; +"2524 softmax_19" -> "2525 dropout_76" [label="(4, 12, 64, 64)", style=solid]; +"2525 dropout_76" -> "2526 matmul_39" [label="(4, 12, 64, 64)", style=solid]; +"2526 matmul_39" -> "2527 transpose_39" [label="(4, 12, 64, 32)", style=solid]; +"2527 transpose_39" -> "2528 reshape_88" [label="(4, 64, 12, 32)", style=solid]; +"2528 reshape_88" -> "2529 reshape_88_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"2529 reshape_88_0_0_nncf_smooth_quant_0" -> "2530 quantize_per_tensor_default_120" [label="(4, 64, 384)", style=solid]; +"2530 quantize_per_tensor_default_120" -> "2531 dequantize_per_tensor_default_120" [label="(4, 64, 384)", style=solid]; +"2531 dequantize_per_tensor_default_120" -> "2536 linear_119" [label="(4, 64, 384)", style=solid]; +"2532 linear_119_scale_0" -> "2535 dequantize_per_channel_default_120" [label="(384,)", style=solid]; +"2533 linear_119_zero_point_0" -> "2535 dequantize_per_channel_default_120" [label="(384,)", style=solid]; +"2534 _frozen_param120" -> "2535 dequantize_per_channel_default_120" [label="(384, 384)", style=solid]; +"2535 dequantize_per_channel_default_120" -> "2536 linear_119" [label="(384, 384)", style=solid]; +"2536 linear_119" -> "2537 dropout_77" [label="(4, 64, 384)", style=solid]; +"2537 dropout_77" -> "2538 view_109" [label="(4, 64, 384)", style=solid]; +"2538 view_109" -> "2539 permute_90" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2539 permute_90" -> "2540 reshape_89" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2540 reshape_89" -> "2541 roll_19" [label="(1, 16, 16, 384)", style=solid]; +"2541 roll_19" -> "2542 slice_297" [label="(1, 16, 16, 384)", style=solid]; +"2542 slice_297" -> "2543 slice_298" [label="(1, 16, 16, 384)", style=solid]; +"2543 slice_298" -> "2544 slice_299" [label="(1, 14, 16, 384)", style=solid]; +"2544 slice_299" -> "2545 slice_300" [label="(1, 14, 14, 384)", style=solid]; +"2545 slice_300" -> "2546 contiguous_37" [label="(1, 14, 14, 384)", style=solid]; +"2546 contiguous_37" -> "2547 layer_norm_41" [label="(1, 14, 14, 384)", style=solid]; +"2547 layer_norm_41" -> "2548 add_68" [label="(1, 14, 14, 384)", style=solid]; +"2548 add_68" -> "2549 add_68_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"2548 add_68" -> "2569 add_69" [label="(1, 14, 14, 384)", style=solid]; +"2549 add_68_0_0_nncf_smooth_quant_0" -> "2550 quantize_per_tensor_default_121" [label="(1, 14, 14, 384)", style=solid]; +"2550 quantize_per_tensor_default_121" -> "2551 dequantize_per_tensor_default_121" [label="(1, 14, 14, 384)", style=solid]; +"2551 dequantize_per_tensor_default_121" -> "2556 linear_120" [label="(1, 14, 14, 384)", style=solid]; +"2552 linear_120_scale_0" -> "2555 dequantize_per_channel_default_121" [label="(1536,)", style=solid]; +"2553 linear_120_zero_point_0" -> "2555 dequantize_per_channel_default_121" [label="(1536,)", style=solid]; +"2554 _frozen_param121" -> "2555 dequantize_per_channel_default_121" [label="(1536, 384)", style=solid]; +"2555 dequantize_per_channel_default_121" -> "2556 linear_120" [label="(1536, 384)", style=solid]; +"2556 linear_120" -> "2557 gelu_19" [label="(1, 14, 14, 1536)", style=solid]; +"2557 gelu_19" -> "2558 dropout_78" [label="(1, 14, 14, 1536)", style=solid]; +"2558 dropout_78" -> "2559 dropout_78_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"2559 dropout_78_0_0_nncf_smooth_quant_0" -> "2560 quantize_per_tensor_default_122" [label="(1, 14, 14, 1536)", style=solid]; +"2560 quantize_per_tensor_default_122" -> "2561 dequantize_per_tensor_default_122" [label="(1, 14, 14, 1536)", style=solid]; +"2561 dequantize_per_tensor_default_122" -> "2566 linear_121" [label="(1, 14, 14, 1536)", style=solid]; +"2562 linear_121_scale_0" -> "2565 dequantize_per_channel_default_122" [label="(384,)", style=solid]; +"2563 linear_121_zero_point_0" -> "2565 dequantize_per_channel_default_122" [label="(384,)", style=solid]; +"2564 _frozen_param122" -> "2565 dequantize_per_channel_default_122" [label="(384, 1536)", style=solid]; +"2565 dequantize_per_channel_default_122" -> "2566 linear_121" [label="(384, 1536)", style=solid]; +"2566 linear_121" -> "2567 dropout_79" [label="(1, 14, 14, 384)", style=solid]; +"2567 dropout_79" -> "2568 layer_norm_42" [label="(1, 14, 14, 384)", style=solid]; +"2568 layer_norm_42" -> "2569 add_69" [label="(1, 14, 14, 384)", style=solid]; +"2569 add_69" -> "2591 pad_22" [label="(1, 14, 14, 384)", style=solid]; +"2569 add_69" -> "2650 add_71" [label="(1, 14, 14, 384)", style=solid]; +"2570 features_5_16_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "2575 linear_122" [label="(1, 15, 15, 2)", style=solid]; +"2571 linear_122_scale_0" -> "2574 dequantize_per_channel_default_123" [label="(512,)", style=solid]; +"2572 linear_122_zero_point_0" -> "2574 dequantize_per_channel_default_123" [label="(512,)", style=solid]; +"2573 _frozen_param123" -> "2574 dequantize_per_channel_default_123" [label="(512, 2)", style=solid]; +"2574 dequantize_per_channel_default_123" -> "2575 linear_122" [label="(512, 2)", style=solid]; +"2575 linear_122" -> "2576 relu__20" [label="(1, 15, 15, 512)", style=solid]; +"2576 relu__20" -> "2577 relu__20_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"2577 relu__20_0_0_nncf_smooth_quant_0" -> "2582 linear_123" [label="(1, 15, 15, 512)", style=solid]; +"2578 linear_123_scale_0" -> "2581 dequantize_per_channel_default_124" [label="(12,)", style=solid]; +"2579 linear_123_zero_point_0" -> "2581 dequantize_per_channel_default_124" [label="(12,)", style=solid]; +"2580 _frozen_param124" -> "2581 dequantize_per_channel_default_124" [label="(12, 512)", style=solid]; +"2581 dequantize_per_channel_default_124" -> "2582 linear_123" [label="(12, 512)", style=solid]; +"2582 linear_123" -> "2583 view_110" [label="(1, 15, 15, 12)", style=solid]; +"2583 view_110" -> "2584 index_20" [label="(225, 12)", style=solid]; +"2584 index_20" -> "2585 view_111" [label="(4096, 12)", style=solid]; +"2585 view_111" -> "2586 permute_91" [label="(64, 64, 12)", style=solid]; +"2586 permute_91" -> "2587 contiguous_38" [label="(12, 64, 64)", style=solid]; +"2587 contiguous_38" -> "2588 unsqueeze_60" [label="(12, 64, 64)", style=solid]; +"2588 unsqueeze_60" -> "2589 sigmoid_20" [label="(1, 12, 64, 64)", style=solid]; +"2589 sigmoid_20" -> "2590 mul_40" [label="(1, 12, 64, 64)", style=solid]; +"2590 mul_40" -> "2626 add_70" [label="(1, 12, 64, 64)", style=solid]; +"2591 pad_22" -> "2592 view_112" [label="(1, 16, 16, 384)", style=solid]; +"2592 view_112" -> "2593 permute_92" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2593 permute_92" -> "2594 reshape_90" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2594 reshape_90" -> "2596 reshape_90_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"2595 clone_20" -> "2603 linear_124" [label="(1152,)", style=solid]; +"2596 reshape_90_0_0_nncf_smooth_quant_0" -> "2597 quantize_per_tensor_default_123" [label="(4, 64, 384)", style=solid]; +"2597 quantize_per_tensor_default_123" -> "2598 dequantize_per_tensor_default_123" [label="(4, 64, 384)", style=solid]; +"2598 dequantize_per_tensor_default_123" -> "2603 linear_124" [label="(4, 64, 384)", style=solid]; +"2599 linear_124_scale_0" -> "2602 dequantize_per_channel_default_125" [label="(1152,)", style=solid]; +"2600 linear_124_zero_point_0" -> "2602 dequantize_per_channel_default_125" [label="(1152,)", style=solid]; +"2601 _frozen_param125" -> "2602 dequantize_per_channel_default_125" [label="(1152, 384)", style=solid]; +"2602 dequantize_per_channel_default_125" -> "2603 linear_124" [label="(1152, 384)", style=solid]; +"2603 linear_124" -> "2604 reshape_91" [label="(4, 64, 1152)", style=solid]; +"2604 reshape_91" -> "2605 permute_93" [label="(4, 64, 3, 12, 32)", style=solid]; +"2605 permute_93" -> "2606 select_60" [label="(3, 4, 12, 64, 32)", style=solid]; +"2605 permute_93" -> "2607 select_61" [label="(3, 4, 12, 64, 32)", style=solid]; +"2605 permute_93" -> "2608 select_62" [label="(3, 4, 12, 64, 32)", style=solid]; +"2606 select_60" -> "2609 linalg_vector_norm_40" [label="(4, 12, 64, 32)", style=solid]; +"2606 select_60" -> "2611 expand_as_40" [label="(4, 12, 64, 32)", style=solid]; +"2606 select_60" -> "2612 div_40" [label="(4, 12, 64, 32)", style=solid]; +"2607 select_61" -> "2615 linalg_vector_norm_41" [label="(4, 12, 64, 32)", style=solid]; +"2607 select_61" -> "2617 expand_as_41" [label="(4, 12, 64, 32)", style=solid]; +"2607 select_61" -> "2618 div_41" [label="(4, 12, 64, 32)", style=solid]; +"2608 select_62" -> "2629 matmul_41" [label="(4, 12, 64, 32)", style=solid]; +"2609 linalg_vector_norm_40" -> "2610 clamp_min_40" [label="(4, 12, 64, 1)", style=solid]; +"2610 clamp_min_40" -> "2611 expand_as_40" [label="(4, 12, 64, 1)", style=solid]; +"2611 expand_as_40" -> "2612 div_40" [label="(4, 12, 64, 32)", style=solid]; +"2612 div_40" -> "2613 quantize_per_tensor_default_124" [label="(4, 12, 64, 32)", style=solid]; +"2613 quantize_per_tensor_default_124" -> "2614 dequantize_per_tensor_default_124" [label="(4, 12, 64, 32)", style=solid]; +"2614 dequantize_per_tensor_default_124" -> "2622 matmul_40" [label="(4, 12, 64, 32)", style=solid]; +"2615 linalg_vector_norm_41" -> "2616 clamp_min_41" [label="(4, 12, 64, 1)", style=solid]; +"2616 clamp_min_41" -> "2617 expand_as_41" [label="(4, 12, 64, 1)", style=solid]; +"2617 expand_as_41" -> "2618 div_41" [label="(4, 12, 64, 32)", style=solid]; +"2618 div_41" -> "2619 quantize_per_tensor_default_125" [label="(4, 12, 64, 32)", style=solid]; +"2619 quantize_per_tensor_default_125" -> "2620 dequantize_per_tensor_default_125" [label="(4, 12, 64, 32)", style=solid]; +"2620 dequantize_per_tensor_default_125" -> "2621 transpose_40" [label="(4, 12, 64, 32)", style=solid]; +"2621 transpose_40" -> "2622 matmul_40" [label="(4, 12, 32, 64)", style=solid]; +"2622 matmul_40" -> "2625 mul_41" [label="(4, 12, 64, 64)", style=solid]; +"2623 clamp_20" -> "2624 exp_20" [label="(12, 1, 1)", style=solid]; +"2624 exp_20" -> "2625 mul_41" [label="(12, 1, 1)", style=solid]; +"2625 mul_41" -> "2626 add_70" [label="(4, 12, 64, 64)", style=solid]; +"2626 add_70" -> "2627 softmax_20" [label="(4, 12, 64, 64)", style=solid]; +"2627 softmax_20" -> "2628 dropout_80" [label="(4, 12, 64, 64)", style=solid]; +"2628 dropout_80" -> "2629 matmul_41" [label="(4, 12, 64, 64)", style=solid]; +"2629 matmul_41" -> "2630 transpose_41" [label="(4, 12, 64, 32)", style=solid]; +"2630 transpose_41" -> "2631 reshape_92" [label="(4, 64, 12, 32)", style=solid]; +"2631 reshape_92" -> "2632 reshape_92_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"2632 reshape_92_0_0_nncf_smooth_quant_0" -> "2633 quantize_per_tensor_default_126" [label="(4, 64, 384)", style=solid]; +"2633 quantize_per_tensor_default_126" -> "2634 dequantize_per_tensor_default_126" [label="(4, 64, 384)", style=solid]; +"2634 dequantize_per_tensor_default_126" -> "2639 linear_125" [label="(4, 64, 384)", style=solid]; +"2635 linear_125_scale_0" -> "2638 dequantize_per_channel_default_126" [label="(384,)", style=solid]; +"2636 linear_125_zero_point_0" -> "2638 dequantize_per_channel_default_126" [label="(384,)", style=solid]; +"2637 _frozen_param126" -> "2638 dequantize_per_channel_default_126" [label="(384, 384)", style=solid]; +"2638 dequantize_per_channel_default_126" -> "2639 linear_125" [label="(384, 384)", style=solid]; +"2639 linear_125" -> "2640 dropout_81" [label="(4, 64, 384)", style=solid]; +"2640 dropout_81" -> "2641 view_113" [label="(4, 64, 384)", style=solid]; +"2641 view_113" -> "2642 permute_94" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2642 permute_94" -> "2643 reshape_93" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2643 reshape_93" -> "2644 slice_302" [label="(1, 16, 16, 384)", style=solid]; +"2644 slice_302" -> "2645 slice_303" [label="(1, 16, 16, 384)", style=solid]; +"2645 slice_303" -> "2646 slice_304" [label="(1, 14, 16, 384)", style=solid]; +"2646 slice_304" -> "2647 slice_305" [label="(1, 14, 14, 384)", style=solid]; +"2647 slice_305" -> "2648 contiguous_39" [label="(1, 14, 14, 384)", style=solid]; +"2648 contiguous_39" -> "2649 layer_norm_43" [label="(1, 14, 14, 384)", style=solid]; +"2649 layer_norm_43" -> "2650 add_71" [label="(1, 14, 14, 384)", style=solid]; +"2650 add_71" -> "2651 add_71_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"2650 add_71" -> "2671 add_72" [label="(1, 14, 14, 384)", style=solid]; +"2651 add_71_0_0_nncf_smooth_quant_0" -> "2652 quantize_per_tensor_default_127" [label="(1, 14, 14, 384)", style=solid]; +"2652 quantize_per_tensor_default_127" -> "2653 dequantize_per_tensor_default_127" [label="(1, 14, 14, 384)", style=solid]; +"2653 dequantize_per_tensor_default_127" -> "2658 linear_126" [label="(1, 14, 14, 384)", style=solid]; +"2654 linear_126_scale_0" -> "2657 dequantize_per_channel_default_127" [label="(1536,)", style=solid]; +"2655 linear_126_zero_point_0" -> "2657 dequantize_per_channel_default_127" [label="(1536,)", style=solid]; +"2656 _frozen_param127" -> "2657 dequantize_per_channel_default_127" [label="(1536, 384)", style=solid]; +"2657 dequantize_per_channel_default_127" -> "2658 linear_126" [label="(1536, 384)", style=solid]; +"2658 linear_126" -> "2659 gelu_20" [label="(1, 14, 14, 1536)", style=solid]; +"2659 gelu_20" -> "2660 dropout_82" [label="(1, 14, 14, 1536)", style=solid]; +"2660 dropout_82" -> "2661 dropout_82_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"2661 dropout_82_0_0_nncf_smooth_quant_0" -> "2662 quantize_per_tensor_default_128" [label="(1, 14, 14, 1536)", style=solid]; +"2662 quantize_per_tensor_default_128" -> "2663 dequantize_per_tensor_default_128" [label="(1, 14, 14, 1536)", style=solid]; +"2663 dequantize_per_tensor_default_128" -> "2668 linear_127" [label="(1, 14, 14, 1536)", style=solid]; +"2664 linear_127_scale_0" -> "2667 dequantize_per_channel_default_128" [label="(384,)", style=solid]; +"2665 linear_127_zero_point_0" -> "2667 dequantize_per_channel_default_128" [label="(384,)", style=solid]; +"2666 _frozen_param128" -> "2667 dequantize_per_channel_default_128" [label="(384, 1536)", style=solid]; +"2667 dequantize_per_channel_default_128" -> "2668 linear_127" [label="(384, 1536)", style=solid]; +"2668 linear_127" -> "2669 dropout_83" [label="(1, 14, 14, 384)", style=solid]; +"2669 dropout_83" -> "2670 layer_norm_44" [label="(1, 14, 14, 384)", style=solid]; +"2670 layer_norm_44" -> "2671 add_72" [label="(1, 14, 14, 384)", style=solid]; +"2671 add_72" -> "2693 pad_23" [label="(1, 14, 14, 384)", style=solid]; +"2671 add_72" -> "2770 add_75" [label="(1, 14, 14, 384)", style=solid]; +"2672 features_5_17_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "2677 linear_128" [label="(1, 15, 15, 2)", style=solid]; +"2673 linear_128_scale_0" -> "2676 dequantize_per_channel_default_129" [label="(512,)", style=solid]; +"2674 linear_128_zero_point_0" -> "2676 dequantize_per_channel_default_129" [label="(512,)", style=solid]; +"2675 _frozen_param129" -> "2676 dequantize_per_channel_default_129" [label="(512, 2)", style=solid]; +"2676 dequantize_per_channel_default_129" -> "2677 linear_128" [label="(512, 2)", style=solid]; +"2677 linear_128" -> "2678 relu__21" [label="(1, 15, 15, 512)", style=solid]; +"2678 relu__21" -> "2679 relu__21_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"2679 relu__21_0_0_nncf_smooth_quant_0" -> "2684 linear_129" [label="(1, 15, 15, 512)", style=solid]; +"2680 linear_129_scale_0" -> "2683 dequantize_per_channel_default_130" [label="(12,)", style=solid]; +"2681 linear_129_zero_point_0" -> "2683 dequantize_per_channel_default_130" [label="(12,)", style=solid]; +"2682 _frozen_param130" -> "2683 dequantize_per_channel_default_130" [label="(12, 512)", style=solid]; +"2683 dequantize_per_channel_default_130" -> "2684 linear_129" [label="(12, 512)", style=solid]; +"2684 linear_129" -> "2685 view_114" [label="(1, 15, 15, 12)", style=solid]; +"2685 view_114" -> "2686 index_21" [label="(225, 12)", style=solid]; +"2686 index_21" -> "2687 view_115" [label="(4096, 12)", style=solid]; +"2687 view_115" -> "2688 permute_95" [label="(64, 64, 12)", style=solid]; +"2688 permute_95" -> "2689 contiguous_40" [label="(12, 64, 64)", style=solid]; +"2689 contiguous_40" -> "2690 unsqueeze_61" [label="(12, 64, 64)", style=solid]; +"2690 unsqueeze_61" -> "2691 sigmoid_21" [label="(1, 12, 64, 64)", style=solid]; +"2691 sigmoid_21" -> "2692 mul_42" [label="(1, 12, 64, 64)", style=solid]; +"2692 mul_42" -> "2729 add_73" [label="(1, 12, 64, 64)", style=solid]; +"2693 pad_23" -> "2694 roll_20" [label="(1, 16, 16, 384)", style=solid]; +"2694 roll_20" -> "2695 view_116" [label="(1, 16, 16, 384)", style=solid]; +"2695 view_116" -> "2696 permute_96" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2696 permute_96" -> "2697 reshape_94" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2697 reshape_94" -> "2699 reshape_94_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"2697 reshape_94" -> "2730 new_zeros_10" [label="(4, 64, 384)", style=solid]; +"2698 clone_21" -> "2706 linear_130" [label="(1152,)", style=solid]; +"2699 reshape_94_0_0_nncf_smooth_quant_0" -> "2700 quantize_per_tensor_default_129" [label="(4, 64, 384)", style=solid]; +"2700 quantize_per_tensor_default_129" -> "2701 dequantize_per_tensor_default_129" [label="(4, 64, 384)", style=solid]; +"2701 dequantize_per_tensor_default_129" -> "2706 linear_130" [label="(4, 64, 384)", style=solid]; +"2702 linear_130_scale_0" -> "2705 dequantize_per_channel_default_131" [label="(1152,)", style=solid]; +"2703 linear_130_zero_point_0" -> "2705 dequantize_per_channel_default_131" [label="(1152,)", style=solid]; +"2704 _frozen_param131" -> "2705 dequantize_per_channel_default_131" [label="(1152, 384)", style=solid]; +"2705 dequantize_per_channel_default_131" -> "2706 linear_130" [label="(1152, 384)", style=solid]; +"2706 linear_130" -> "2707 reshape_95" [label="(4, 64, 1152)", style=solid]; +"2707 reshape_95" -> "2708 permute_97" [label="(4, 64, 3, 12, 32)", style=solid]; +"2708 permute_97" -> "2709 select_63" [label="(3, 4, 12, 64, 32)", style=solid]; +"2708 permute_97" -> "2710 select_64" [label="(3, 4, 12, 64, 32)", style=solid]; +"2708 permute_97" -> "2711 select_65" [label="(3, 4, 12, 64, 32)", style=solid]; +"2709 select_63" -> "2712 linalg_vector_norm_42" [label="(4, 12, 64, 32)", style=solid]; +"2709 select_63" -> "2714 expand_as_42" [label="(4, 12, 64, 32)", style=solid]; +"2709 select_63" -> "2715 div_42" [label="(4, 12, 64, 32)", style=solid]; +"2710 select_64" -> "2718 linalg_vector_norm_43" [label="(4, 12, 64, 32)", style=solid]; +"2710 select_64" -> "2720 expand_as_43" [label="(4, 12, 64, 32)", style=solid]; +"2710 select_64" -> "2721 div_43" [label="(4, 12, 64, 32)", style=solid]; +"2711 select_65" -> "2748 matmul_43" [label="(4, 12, 64, 32)", style=solid]; +"2712 linalg_vector_norm_42" -> "2713 clamp_min_42" [label="(4, 12, 64, 1)", style=solid]; +"2713 clamp_min_42" -> "2714 expand_as_42" [label="(4, 12, 64, 1)", style=solid]; +"2714 expand_as_42" -> "2715 div_42" [label="(4, 12, 64, 32)", style=solid]; +"2715 div_42" -> "2716 quantize_per_tensor_default_130" [label="(4, 12, 64, 32)", style=solid]; +"2716 quantize_per_tensor_default_130" -> "2717 dequantize_per_tensor_default_130" [label="(4, 12, 64, 32)", style=solid]; +"2717 dequantize_per_tensor_default_130" -> "2725 matmul_42" [label="(4, 12, 64, 32)", style=solid]; +"2718 linalg_vector_norm_43" -> "2719 clamp_min_43" [label="(4, 12, 64, 1)", style=solid]; +"2719 clamp_min_43" -> "2720 expand_as_43" [label="(4, 12, 64, 1)", style=solid]; +"2720 expand_as_43" -> "2721 div_43" [label="(4, 12, 64, 32)", style=solid]; +"2721 div_43" -> "2722 quantize_per_tensor_default_131" [label="(4, 12, 64, 32)", style=solid]; +"2722 quantize_per_tensor_default_131" -> "2723 dequantize_per_tensor_default_131" [label="(4, 12, 64, 32)", style=solid]; +"2723 dequantize_per_tensor_default_131" -> "2724 transpose_42" [label="(4, 12, 64, 32)", style=solid]; +"2724 transpose_42" -> "2725 matmul_42" [label="(4, 12, 32, 64)", style=solid]; +"2725 matmul_42" -> "2728 mul_43" [label="(4, 12, 64, 64)", style=solid]; +"2726 clamp_21" -> "2727 exp_21" [label="(12, 1, 1)", style=solid]; +"2727 exp_21" -> "2728 mul_43" [label="(12, 1, 1)", style=solid]; +"2728 mul_43" -> "2729 add_73" [label="(4, 12, 64, 64)", style=solid]; +"2729 add_73" -> "2741 view_118" [label="(4, 12, 64, 64)", style=solid]; +"2730 new_zeros_10" -> "2731 view_117" [label="(16, 16)", style=solid]; +"2731 view_117" -> "2732 permute_98" [label="(2, 8, 2, 8)", style=solid]; +"2732 permute_98" -> "2733 reshape_96" [label="(2, 2, 8, 8)", style=solid]; +"2733 reshape_96" -> "2734 unsqueeze_62" [label="(4, 64)", style=solid]; +"2733 reshape_96" -> "2735 unsqueeze_63" [label="(4, 64)", style=solid]; +"2734 unsqueeze_62" -> "2736 sub_10" [label="(4, 1, 64)", style=solid]; +"2735 unsqueeze_63" -> "2736 sub_10" [label="(4, 64, 1)", style=solid]; +"2736 sub_10" -> "2737 ne_10" [label="(4, 64, 64)", style=solid]; +"2736 sub_10" -> "2738 masked_fill_20" [label="(4, 64, 64)", style=solid]; +"2736 sub_10" -> "2739 eq_10" [label="(4, 64, 64)", style=solid]; +"2737 ne_10" -> "2738 masked_fill_20" [label="(4, 64, 64)", style=solid]; +"2738 masked_fill_20" -> "2740 masked_fill_21" [label="(4, 64, 64)", style=solid]; +"2739 eq_10" -> "2740 masked_fill_21" [label="(4, 64, 64)", style=solid]; +"2740 masked_fill_21" -> "2742 unsqueeze_64" [label="(4, 64, 64)", style=solid]; +"2741 view_118" -> "2744 add_74" [label="(1, 4, 12, 64, 64)", style=solid]; +"2742 unsqueeze_64" -> "2743 unsqueeze_65" [label="(4, 1, 64, 64)", style=solid]; +"2743 unsqueeze_65" -> "2744 add_74" [label="(1, 4, 1, 64, 64)", style=solid]; +"2744 add_74" -> "2745 view_119" [label="(1, 4, 12, 64, 64)", style=solid]; +"2745 view_119" -> "2746 softmax_21" [label="(4, 12, 64, 64)", style=solid]; +"2746 softmax_21" -> "2747 dropout_84" [label="(4, 12, 64, 64)", style=solid]; +"2747 dropout_84" -> "2748 matmul_43" [label="(4, 12, 64, 64)", style=solid]; +"2748 matmul_43" -> "2749 transpose_43" [label="(4, 12, 64, 32)", style=solid]; +"2749 transpose_43" -> "2750 reshape_97" [label="(4, 64, 12, 32)", style=solid]; +"2750 reshape_97" -> "2751 reshape_97_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"2751 reshape_97_0_0_nncf_smooth_quant_0" -> "2752 quantize_per_tensor_default_132" [label="(4, 64, 384)", style=solid]; +"2752 quantize_per_tensor_default_132" -> "2753 dequantize_per_tensor_default_132" [label="(4, 64, 384)", style=solid]; +"2753 dequantize_per_tensor_default_132" -> "2758 linear_131" [label="(4, 64, 384)", style=solid]; +"2754 linear_131_scale_0" -> "2757 dequantize_per_channel_default_132" [label="(384,)", style=solid]; +"2755 linear_131_zero_point_0" -> "2757 dequantize_per_channel_default_132" [label="(384,)", style=solid]; +"2756 _frozen_param132" -> "2757 dequantize_per_channel_default_132" [label="(384, 384)", style=solid]; +"2757 dequantize_per_channel_default_132" -> "2758 linear_131" [label="(384, 384)", style=solid]; +"2758 linear_131" -> "2759 dropout_85" [label="(4, 64, 384)", style=solid]; +"2759 dropout_85" -> "2760 view_120" [label="(4, 64, 384)", style=solid]; +"2760 view_120" -> "2761 permute_99" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2761 permute_99" -> "2762 reshape_98" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2762 reshape_98" -> "2763 roll_21" [label="(1, 16, 16, 384)", style=solid]; +"2763 roll_21" -> "2764 slice_325" [label="(1, 16, 16, 384)", style=solid]; +"2764 slice_325" -> "2765 slice_326" [label="(1, 16, 16, 384)", style=solid]; +"2765 slice_326" -> "2766 slice_327" [label="(1, 14, 16, 384)", style=solid]; +"2766 slice_327" -> "2767 slice_328" [label="(1, 14, 14, 384)", style=solid]; +"2767 slice_328" -> "2768 contiguous_41" [label="(1, 14, 14, 384)", style=solid]; +"2768 contiguous_41" -> "2769 layer_norm_45" [label="(1, 14, 14, 384)", style=solid]; +"2769 layer_norm_45" -> "2770 add_75" [label="(1, 14, 14, 384)", style=solid]; +"2770 add_75" -> "2771 add_75_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"2770 add_75" -> "2791 add_76" [label="(1, 14, 14, 384)", style=solid]; +"2771 add_75_0_0_nncf_smooth_quant_0" -> "2772 quantize_per_tensor_default_133" [label="(1, 14, 14, 384)", style=solid]; +"2772 quantize_per_tensor_default_133" -> "2773 dequantize_per_tensor_default_133" [label="(1, 14, 14, 384)", style=solid]; +"2773 dequantize_per_tensor_default_133" -> "2778 linear_132" [label="(1, 14, 14, 384)", style=solid]; +"2774 linear_132_scale_0" -> "2777 dequantize_per_channel_default_133" [label="(1536,)", style=solid]; +"2775 linear_132_zero_point_0" -> "2777 dequantize_per_channel_default_133" [label="(1536,)", style=solid]; +"2776 _frozen_param133" -> "2777 dequantize_per_channel_default_133" [label="(1536, 384)", style=solid]; +"2777 dequantize_per_channel_default_133" -> "2778 linear_132" [label="(1536, 384)", style=solid]; +"2778 linear_132" -> "2779 gelu_21" [label="(1, 14, 14, 1536)", style=solid]; +"2779 gelu_21" -> "2780 dropout_86" [label="(1, 14, 14, 1536)", style=solid]; +"2780 dropout_86" -> "2781 dropout_86_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"2781 dropout_86_0_0_nncf_smooth_quant_0" -> "2782 quantize_per_tensor_default_134" [label="(1, 14, 14, 1536)", style=solid]; +"2782 quantize_per_tensor_default_134" -> "2783 dequantize_per_tensor_default_134" [label="(1, 14, 14, 1536)", style=solid]; +"2783 dequantize_per_tensor_default_134" -> "2788 linear_133" [label="(1, 14, 14, 1536)", style=solid]; +"2784 linear_133_scale_0" -> "2787 dequantize_per_channel_default_134" [label="(384,)", style=solid]; +"2785 linear_133_zero_point_0" -> "2787 dequantize_per_channel_default_134" [label="(384,)", style=solid]; +"2786 _frozen_param134" -> "2787 dequantize_per_channel_default_134" [label="(384, 1536)", style=solid]; +"2787 dequantize_per_channel_default_134" -> "2788 linear_133" [label="(384, 1536)", style=solid]; +"2788 linear_133" -> "2789 dropout_87" [label="(1, 14, 14, 384)", style=solid]; +"2789 dropout_87" -> "2790 layer_norm_46" [label="(1, 14, 14, 384)", style=solid]; +"2790 layer_norm_46" -> "2791 add_76" [label="(1, 14, 14, 384)", style=solid]; +"2791 add_76" -> "2792 pad_24" [label="(1, 14, 14, 384)", style=solid]; +"2792 pad_24" -> "2793 slice_329" [label="(1, 14, 14, 384)", style=solid]; +"2792 pad_24" -> "2796 slice_332" [label="(1, 14, 14, 384)", style=solid]; +"2792 pad_24" -> "2799 slice_335" [label="(1, 14, 14, 384)", style=solid]; +"2792 pad_24" -> "2802 slice_338" [label="(1, 14, 14, 384)", style=solid]; +"2793 slice_329" -> "2794 slice_330" [label="(1, 7, 14, 384)", style=solid]; +"2794 slice_330" -> "2795 slice_331" [label="(1, 7, 7, 384)", style=solid]; +"2795 slice_331" -> "2805 cat_2" [label="(1, 7, 7, 384)", style=solid]; +"2796 slice_332" -> "2797 slice_333" [label="(1, 7, 14, 384)", style=solid]; +"2797 slice_333" -> "2798 slice_334" [label="(1, 7, 7, 384)", style=solid]; +"2798 slice_334" -> "2805 cat_2" [label="(1, 7, 7, 384)", style=solid]; +"2799 slice_335" -> "2800 slice_336" [label="(1, 7, 14, 384)", style=solid]; +"2800 slice_336" -> "2801 slice_337" [label="(1, 7, 7, 384)", style=solid]; +"2801 slice_337" -> "2805 cat_2" [label="(1, 7, 7, 384)", style=solid]; +"2802 slice_338" -> "2803 slice_339" [label="(1, 7, 14, 384)", style=solid]; +"2803 slice_339" -> "2804 slice_340" [label="(1, 7, 7, 384)", style=solid]; +"2804 slice_340" -> "2805 cat_2" [label="(1, 7, 7, 384)", style=solid]; +"2805 cat_2" -> "2806 cat_2_0_0_nncf_smooth_quant_0" [label="(1, 7, 7, 1536)", style=solid]; +"2806 cat_2_0_0_nncf_smooth_quant_0" -> "2807 quantize_per_tensor_default_135" [label="(1, 7, 7, 1536)", style=solid]; +"2807 quantize_per_tensor_default_135" -> "2808 dequantize_per_tensor_default_135" [label="(1, 7, 7, 1536)", style=solid]; +"2808 dequantize_per_tensor_default_135" -> "2813 linear_134" [label="(1, 7, 7, 1536)", style=solid]; +"2809 linear_134_scale_0" -> "2812 dequantize_per_channel_default_135" [label="(768,)", style=solid]; +"2810 linear_134_zero_point_0" -> "2812 dequantize_per_channel_default_135" [label="(768,)", style=solid]; +"2811 _frozen_param135" -> "2812 dequantize_per_channel_default_135" [label="(768, 1536)", style=solid]; +"2812 dequantize_per_channel_default_135" -> "2813 linear_134" [label="(768, 1536)", style=solid]; +"2813 linear_134" -> "2814 layer_norm_47" [label="(1, 7, 7, 768)", style=solid]; +"2814 layer_norm_47" -> "2836 pad_25" [label="(1, 7, 7, 768)", style=solid]; +"2814 layer_norm_47" -> "2895 add_78" [label="(1, 7, 7, 768)", style=solid]; +"2815 features_7_0_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "2820 linear_135" [label="(1, 15, 15, 2)", style=solid]; +"2816 linear_135_scale_0" -> "2819 dequantize_per_channel_default_136" [label="(512,)", style=solid]; +"2817 linear_135_zero_point_0" -> "2819 dequantize_per_channel_default_136" [label="(512,)", style=solid]; +"2818 _frozen_param136" -> "2819 dequantize_per_channel_default_136" [label="(512, 2)", style=solid]; +"2819 dequantize_per_channel_default_136" -> "2820 linear_135" [label="(512, 2)", style=solid]; +"2820 linear_135" -> "2821 relu__22" [label="(1, 15, 15, 512)", style=solid]; +"2821 relu__22" -> "2822 relu__22_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"2822 relu__22_0_0_nncf_smooth_quant_0" -> "2827 linear_136" [label="(1, 15, 15, 512)", style=solid]; +"2823 linear_136_scale_0" -> "2826 dequantize_per_channel_default_137" [label="(24,)", style=solid]; +"2824 linear_136_zero_point_0" -> "2826 dequantize_per_channel_default_137" [label="(24,)", style=solid]; +"2825 _frozen_param137" -> "2826 dequantize_per_channel_default_137" [label="(24, 512)", style=solid]; +"2826 dequantize_per_channel_default_137" -> "2827 linear_136" [label="(24, 512)", style=solid]; +"2827 linear_136" -> "2828 view_121" [label="(1, 15, 15, 24)", style=solid]; +"2828 view_121" -> "2829 index_22" [label="(225, 24)", style=solid]; +"2829 index_22" -> "2830 view_122" [label="(4096, 24)", style=solid]; +"2830 view_122" -> "2831 permute_100" [label="(64, 64, 24)", style=solid]; +"2831 permute_100" -> "2832 contiguous_42" [label="(24, 64, 64)", style=solid]; +"2832 contiguous_42" -> "2833 unsqueeze_66" [label="(24, 64, 64)", style=solid]; +"2833 unsqueeze_66" -> "2834 sigmoid_22" [label="(1, 24, 64, 64)", style=solid]; +"2834 sigmoid_22" -> "2835 mul_44" [label="(1, 24, 64, 64)", style=solid]; +"2835 mul_44" -> "2871 add_77" [label="(1, 24, 64, 64)", style=solid]; +"2836 pad_25" -> "2837 view_123" [label="(1, 8, 8, 768)", style=solid]; +"2837 view_123" -> "2838 permute_101" [label="(1, 1, 8, 1, 8, 768)", style=solid]; +"2838 permute_101" -> "2839 reshape_99" [label="(1, 1, 1, 8, 8, 768)", style=solid]; +"2839 reshape_99" -> "2841 reshape_99_0_0_nncf_smooth_quant_0" [label="(1, 64, 768)", style=solid]; +"2840 clone_22" -> "2848 linear_137" [label="(2304,)", style=solid]; +"2841 reshape_99_0_0_nncf_smooth_quant_0" -> "2842 quantize_per_tensor_default_136" [label="(1, 64, 768)", style=solid]; +"2842 quantize_per_tensor_default_136" -> "2843 dequantize_per_tensor_default_136" [label="(1, 64, 768)", style=solid]; +"2843 dequantize_per_tensor_default_136" -> "2848 linear_137" [label="(1, 64, 768)", style=solid]; +"2844 linear_137_scale_0" -> "2847 dequantize_per_channel_default_138" [label="(2304,)", style=solid]; +"2845 linear_137_zero_point_0" -> "2847 dequantize_per_channel_default_138" [label="(2304,)", style=solid]; +"2846 _frozen_param138" -> "2847 dequantize_per_channel_default_138" [label="(2304, 768)", style=solid]; +"2847 dequantize_per_channel_default_138" -> "2848 linear_137" [label="(2304, 768)", style=solid]; +"2848 linear_137" -> "2849 reshape_100" [label="(1, 64, 2304)", style=solid]; +"2849 reshape_100" -> "2850 permute_102" [label="(1, 64, 3, 24, 32)", style=solid]; +"2850 permute_102" -> "2851 select_66" [label="(3, 1, 24, 64, 32)", style=solid]; +"2850 permute_102" -> "2852 select_67" [label="(3, 1, 24, 64, 32)", style=solid]; +"2850 permute_102" -> "2853 select_68" [label="(3, 1, 24, 64, 32)", style=solid]; +"2851 select_66" -> "2854 linalg_vector_norm_44" [label="(1, 24, 64, 32)", style=solid]; +"2851 select_66" -> "2856 expand_as_44" [label="(1, 24, 64, 32)", style=solid]; +"2851 select_66" -> "2857 div_44" [label="(1, 24, 64, 32)", style=solid]; +"2852 select_67" -> "2860 linalg_vector_norm_45" [label="(1, 24, 64, 32)", style=solid]; +"2852 select_67" -> "2862 expand_as_45" [label="(1, 24, 64, 32)", style=solid]; +"2852 select_67" -> "2863 div_45" [label="(1, 24, 64, 32)", style=solid]; +"2853 select_68" -> "2874 matmul_45" [label="(1, 24, 64, 32)", style=solid]; +"2854 linalg_vector_norm_44" -> "2855 clamp_min_44" [label="(1, 24, 64, 1)", style=solid]; +"2855 clamp_min_44" -> "2856 expand_as_44" [label="(1, 24, 64, 1)", style=solid]; +"2856 expand_as_44" -> "2857 div_44" [label="(1, 24, 64, 32)", style=solid]; +"2857 div_44" -> "2858 quantize_per_tensor_default_137" [label="(1, 24, 64, 32)", style=solid]; +"2858 quantize_per_tensor_default_137" -> "2859 dequantize_per_tensor_default_137" [label="(1, 24, 64, 32)", style=solid]; +"2859 dequantize_per_tensor_default_137" -> "2867 matmul_44" [label="(1, 24, 64, 32)", style=solid]; +"2860 linalg_vector_norm_45" -> "2861 clamp_min_45" [label="(1, 24, 64, 1)", style=solid]; +"2861 clamp_min_45" -> "2862 expand_as_45" [label="(1, 24, 64, 1)", style=solid]; +"2862 expand_as_45" -> "2863 div_45" [label="(1, 24, 64, 32)", style=solid]; +"2863 div_45" -> "2864 quantize_per_tensor_default_138" [label="(1, 24, 64, 32)", style=solid]; +"2864 quantize_per_tensor_default_138" -> "2865 dequantize_per_tensor_default_138" [label="(1, 24, 64, 32)", style=solid]; +"2865 dequantize_per_tensor_default_138" -> "2866 transpose_44" [label="(1, 24, 64, 32)", style=solid]; +"2866 transpose_44" -> "2867 matmul_44" [label="(1, 24, 32, 64)", style=solid]; +"2867 matmul_44" -> "2870 mul_45" [label="(1, 24, 64, 64)", style=solid]; +"2868 clamp_22" -> "2869 exp_22" [label="(24, 1, 1)", style=solid]; +"2869 exp_22" -> "2870 mul_45" [label="(24, 1, 1)", style=solid]; +"2870 mul_45" -> "2871 add_77" [label="(1, 24, 64, 64)", style=solid]; +"2871 add_77" -> "2872 softmax_22" [label="(1, 24, 64, 64)", style=solid]; +"2872 softmax_22" -> "2873 dropout_88" [label="(1, 24, 64, 64)", style=solid]; +"2873 dropout_88" -> "2874 matmul_45" [label="(1, 24, 64, 64)", style=solid]; +"2874 matmul_45" -> "2875 transpose_45" [label="(1, 24, 64, 32)", style=solid]; +"2875 transpose_45" -> "2876 reshape_101" [label="(1, 64, 24, 32)", style=solid]; +"2876 reshape_101" -> "2877 reshape_101_0_0_nncf_smooth_quant_0" [label="(1, 64, 768)", style=solid]; +"2877 reshape_101_0_0_nncf_smooth_quant_0" -> "2878 quantize_per_tensor_default_139" [label="(1, 64, 768)", style=solid]; +"2878 quantize_per_tensor_default_139" -> "2879 dequantize_per_tensor_default_139" [label="(1, 64, 768)", style=solid]; +"2879 dequantize_per_tensor_default_139" -> "2884 linear_138" [label="(1, 64, 768)", style=solid]; +"2880 linear_138_scale_0" -> "2883 dequantize_per_channel_default_139" [label="(768,)", style=solid]; +"2881 linear_138_zero_point_0" -> "2883 dequantize_per_channel_default_139" [label="(768,)", style=solid]; +"2882 _frozen_param139" -> "2883 dequantize_per_channel_default_139" [label="(768, 768)", style=solid]; +"2883 dequantize_per_channel_default_139" -> "2884 linear_138" [label="(768, 768)", style=solid]; +"2884 linear_138" -> "2885 dropout_89" [label="(1, 64, 768)", style=solid]; +"2885 dropout_89" -> "2886 view_124" [label="(1, 64, 768)", style=solid]; +"2886 view_124" -> "2887 permute_103" [label="(1, 1, 1, 8, 8, 768)", style=solid]; +"2887 permute_103" -> "2888 reshape_102" [label="(1, 1, 8, 1, 8, 768)", style=solid]; +"2888 reshape_102" -> "2889 slice_342" [label="(1, 8, 8, 768)", style=solid]; +"2889 slice_342" -> "2890 slice_343" [label="(1, 8, 8, 768)", style=solid]; +"2890 slice_343" -> "2891 slice_344" [label="(1, 7, 8, 768)", style=solid]; +"2891 slice_344" -> "2892 slice_345" [label="(1, 7, 7, 768)", style=solid]; +"2892 slice_345" -> "2893 contiguous_43" [label="(1, 7, 7, 768)", style=solid]; +"2893 contiguous_43" -> "2894 layer_norm_48" [label="(1, 7, 7, 768)", style=solid]; +"2894 layer_norm_48" -> "2895 add_78" [label="(1, 7, 7, 768)", style=solid]; +"2895 add_78" -> "2896 add_78_0_0_nncf_smooth_quant_0" [label="(1, 7, 7, 768)", style=solid]; +"2895 add_78" -> "2916 add_79" [label="(1, 7, 7, 768)", style=solid]; +"2896 add_78_0_0_nncf_smooth_quant_0" -> "2897 quantize_per_tensor_default_140" [label="(1, 7, 7, 768)", style=solid]; +"2897 quantize_per_tensor_default_140" -> "2898 dequantize_per_tensor_default_140" [label="(1, 7, 7, 768)", style=solid]; +"2898 dequantize_per_tensor_default_140" -> "2903 linear_139" [label="(1, 7, 7, 768)", style=solid]; +"2899 linear_139_scale_0" -> "2902 dequantize_per_channel_default_140" [label="(3072,)", style=solid]; +"2900 linear_139_zero_point_0" -> "2902 dequantize_per_channel_default_140" [label="(3072,)", style=solid]; +"2901 _frozen_param140" -> "2902 dequantize_per_channel_default_140" [label="(3072, 768)", style=solid]; +"2902 dequantize_per_channel_default_140" -> "2903 linear_139" [label="(3072, 768)", style=solid]; +"2903 linear_139" -> "2904 gelu_22" [label="(1, 7, 7, 3072)", style=solid]; +"2904 gelu_22" -> "2905 dropout_90" [label="(1, 7, 7, 3072)", style=solid]; +"2905 dropout_90" -> "2906 dropout_90_0_0_nncf_smooth_quant_0" [label="(1, 7, 7, 3072)", style=solid]; +"2906 dropout_90_0_0_nncf_smooth_quant_0" -> "2907 quantize_per_tensor_default_141" [label="(1, 7, 7, 3072)", style=solid]; +"2907 quantize_per_tensor_default_141" -> "2908 dequantize_per_tensor_default_141" [label="(1, 7, 7, 3072)", style=solid]; +"2908 dequantize_per_tensor_default_141" -> "2913 linear_140" [label="(1, 7, 7, 3072)", style=solid]; +"2909 linear_140_scale_0" -> "2912 dequantize_per_channel_default_141" [label="(768,)", style=solid]; +"2910 linear_140_zero_point_0" -> "2912 dequantize_per_channel_default_141" [label="(768,)", style=solid]; +"2911 _frozen_param141" -> "2912 dequantize_per_channel_default_141" [label="(768, 3072)", style=solid]; +"2912 dequantize_per_channel_default_141" -> "2913 linear_140" [label="(768, 3072)", style=solid]; +"2913 linear_140" -> "2914 dropout_91" [label="(1, 7, 7, 768)", style=solid]; +"2914 dropout_91" -> "2915 layer_norm_49" [label="(1, 7, 7, 768)", style=solid]; +"2915 layer_norm_49" -> "2916 add_79" [label="(1, 7, 7, 768)", style=solid]; +"2916 add_79" -> "2938 pad_26" [label="(1, 7, 7, 768)", style=solid]; +"2916 add_79" -> "2997 add_81" [label="(1, 7, 7, 768)", style=solid]; +"2917 features_7_1_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "2922 linear_141" [label="(1, 15, 15, 2)", style=solid]; +"2918 linear_141_scale_0" -> "2921 dequantize_per_channel_default_142" [label="(512,)", style=solid]; +"2919 linear_141_zero_point_0" -> "2921 dequantize_per_channel_default_142" [label="(512,)", style=solid]; +"2920 _frozen_param142" -> "2921 dequantize_per_channel_default_142" [label="(512, 2)", style=solid]; +"2921 dequantize_per_channel_default_142" -> "2922 linear_141" [label="(512, 2)", style=solid]; +"2922 linear_141" -> "2923 relu__23" [label="(1, 15, 15, 512)", style=solid]; +"2923 relu__23" -> "2924 relu__23_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; +"2924 relu__23_0_0_nncf_smooth_quant_0" -> "2929 linear_142" [label="(1, 15, 15, 512)", style=solid]; +"2925 linear_142_scale_0" -> "2928 dequantize_per_channel_default_143" [label="(24,)", style=solid]; +"2926 linear_142_zero_point_0" -> "2928 dequantize_per_channel_default_143" [label="(24,)", style=solid]; +"2927 _frozen_param143" -> "2928 dequantize_per_channel_default_143" [label="(24, 512)", style=solid]; +"2928 dequantize_per_channel_default_143" -> "2929 linear_142" [label="(24, 512)", style=solid]; +"2929 linear_142" -> "2930 view_125" [label="(1, 15, 15, 24)", style=solid]; +"2930 view_125" -> "2931 index_23" [label="(225, 24)", style=solid]; +"2931 index_23" -> "2932 view_126" [label="(4096, 24)", style=solid]; +"2932 view_126" -> "2933 permute_104" [label="(64, 64, 24)", style=solid]; +"2933 permute_104" -> "2934 contiguous_44" [label="(24, 64, 64)", style=solid]; +"2934 contiguous_44" -> "2935 unsqueeze_67" [label="(24, 64, 64)", style=solid]; +"2935 unsqueeze_67" -> "2936 sigmoid_23" [label="(1, 24, 64, 64)", style=solid]; +"2936 sigmoid_23" -> "2937 mul_46" [label="(1, 24, 64, 64)", style=solid]; +"2937 mul_46" -> "2973 add_80" [label="(1, 24, 64, 64)", style=solid]; +"2938 pad_26" -> "2939 view_127" [label="(1, 8, 8, 768)", style=solid]; +"2939 view_127" -> "2940 permute_105" [label="(1, 1, 8, 1, 8, 768)", style=solid]; +"2940 permute_105" -> "2941 reshape_103" [label="(1, 1, 1, 8, 8, 768)", style=solid]; +"2941 reshape_103" -> "2943 reshape_103_0_0_nncf_smooth_quant_0" [label="(1, 64, 768)", style=solid]; +"2942 clone_23" -> "2950 linear_143" [label="(2304,)", style=solid]; +"2943 reshape_103_0_0_nncf_smooth_quant_0" -> "2944 quantize_per_tensor_default_142" [label="(1, 64, 768)", style=solid]; +"2944 quantize_per_tensor_default_142" -> "2945 dequantize_per_tensor_default_142" [label="(1, 64, 768)", style=solid]; +"2945 dequantize_per_tensor_default_142" -> "2950 linear_143" [label="(1, 64, 768)", style=solid]; +"2946 linear_143_scale_0" -> "2949 dequantize_per_channel_default_144" [label="(2304,)", style=solid]; +"2947 linear_143_zero_point_0" -> "2949 dequantize_per_channel_default_144" [label="(2304,)", style=solid]; +"2948 _frozen_param144" -> "2949 dequantize_per_channel_default_144" [label="(2304, 768)", style=solid]; +"2949 dequantize_per_channel_default_144" -> "2950 linear_143" [label="(2304, 768)", style=solid]; +"2950 linear_143" -> "2951 reshape_104" [label="(1, 64, 2304)", style=solid]; +"2951 reshape_104" -> "2952 permute_106" [label="(1, 64, 3, 24, 32)", style=solid]; +"2952 permute_106" -> "2953 select_69" [label="(3, 1, 24, 64, 32)", style=solid]; +"2952 permute_106" -> "2954 select_70" [label="(3, 1, 24, 64, 32)", style=solid]; +"2952 permute_106" -> "2955 select_71" [label="(3, 1, 24, 64, 32)", style=solid]; +"2953 select_69" -> "2956 linalg_vector_norm_46" [label="(1, 24, 64, 32)", style=solid]; +"2953 select_69" -> "2958 expand_as_46" [label="(1, 24, 64, 32)", style=solid]; +"2953 select_69" -> "2959 div_46" [label="(1, 24, 64, 32)", style=solid]; +"2954 select_70" -> "2962 linalg_vector_norm_47" [label="(1, 24, 64, 32)", style=solid]; +"2954 select_70" -> "2964 expand_as_47" [label="(1, 24, 64, 32)", style=solid]; +"2954 select_70" -> "2965 div_47" [label="(1, 24, 64, 32)", style=solid]; +"2955 select_71" -> "2976 matmul_47" [label="(1, 24, 64, 32)", style=solid]; +"2956 linalg_vector_norm_46" -> "2957 clamp_min_46" [label="(1, 24, 64, 1)", style=solid]; +"2957 clamp_min_46" -> "2958 expand_as_46" [label="(1, 24, 64, 1)", style=solid]; +"2958 expand_as_46" -> "2959 div_46" [label="(1, 24, 64, 32)", style=solid]; +"2959 div_46" -> "2960 quantize_per_tensor_default_143" [label="(1, 24, 64, 32)", style=solid]; +"2960 quantize_per_tensor_default_143" -> "2961 dequantize_per_tensor_default_143" [label="(1, 24, 64, 32)", style=solid]; +"2961 dequantize_per_tensor_default_143" -> "2969 matmul_46" [label="(1, 24, 64, 32)", style=solid]; +"2962 linalg_vector_norm_47" -> "2963 clamp_min_47" [label="(1, 24, 64, 1)", style=solid]; +"2963 clamp_min_47" -> "2964 expand_as_47" [label="(1, 24, 64, 1)", style=solid]; +"2964 expand_as_47" -> "2965 div_47" [label="(1, 24, 64, 32)", style=solid]; +"2965 div_47" -> "2966 quantize_per_tensor_default_144" [label="(1, 24, 64, 32)", style=solid]; +"2966 quantize_per_tensor_default_144" -> "2967 dequantize_per_tensor_default_144" [label="(1, 24, 64, 32)", style=solid]; +"2967 dequantize_per_tensor_default_144" -> "2968 transpose_46" [label="(1, 24, 64, 32)", style=solid]; +"2968 transpose_46" -> "2969 matmul_46" [label="(1, 24, 32, 64)", style=solid]; +"2969 matmul_46" -> "2972 mul_47" [label="(1, 24, 64, 64)", style=solid]; +"2970 clamp_23" -> "2971 exp_23" [label="(24, 1, 1)", style=solid]; +"2971 exp_23" -> "2972 mul_47" [label="(24, 1, 1)", style=solid]; +"2972 mul_47" -> "2973 add_80" [label="(1, 24, 64, 64)", style=solid]; +"2973 add_80" -> "2974 softmax_23" [label="(1, 24, 64, 64)", style=solid]; +"2974 softmax_23" -> "2975 dropout_92" [label="(1, 24, 64, 64)", style=solid]; +"2975 dropout_92" -> "2976 matmul_47" [label="(1, 24, 64, 64)", style=solid]; +"2976 matmul_47" -> "2977 transpose_47" [label="(1, 24, 64, 32)", style=solid]; +"2977 transpose_47" -> "2978 reshape_105" [label="(1, 64, 24, 32)", style=solid]; +"2978 reshape_105" -> "2979 reshape_105_0_0_nncf_smooth_quant_0" [label="(1, 64, 768)", style=solid]; +"2979 reshape_105_0_0_nncf_smooth_quant_0" -> "2980 quantize_per_tensor_default_145" [label="(1, 64, 768)", style=solid]; +"2980 quantize_per_tensor_default_145" -> "2981 dequantize_per_tensor_default_145" [label="(1, 64, 768)", style=solid]; +"2981 dequantize_per_tensor_default_145" -> "2986 linear_144" [label="(1, 64, 768)", style=solid]; +"2982 linear_144_scale_0" -> "2985 dequantize_per_channel_default_145" [label="(768,)", style=solid]; +"2983 linear_144_zero_point_0" -> "2985 dequantize_per_channel_default_145" [label="(768,)", style=solid]; +"2984 _frozen_param145" -> "2985 dequantize_per_channel_default_145" [label="(768, 768)", style=solid]; +"2985 dequantize_per_channel_default_145" -> "2986 linear_144" [label="(768, 768)", style=solid]; +"2986 linear_144" -> "2987 dropout_93" [label="(1, 64, 768)", style=solid]; +"2987 dropout_93" -> "2988 view_128" [label="(1, 64, 768)", style=solid]; +"2988 view_128" -> "2989 permute_107" [label="(1, 1, 1, 8, 8, 768)", style=solid]; +"2989 permute_107" -> "2990 reshape_106" [label="(1, 1, 8, 1, 8, 768)", style=solid]; +"2990 reshape_106" -> "2991 slice_347" [label="(1, 8, 8, 768)", style=solid]; +"2991 slice_347" -> "2992 slice_348" [label="(1, 8, 8, 768)", style=solid]; +"2992 slice_348" -> "2993 slice_349" [label="(1, 7, 8, 768)", style=solid]; +"2993 slice_349" -> "2994 slice_350" [label="(1, 7, 7, 768)", style=solid]; +"2994 slice_350" -> "2995 contiguous_45" [label="(1, 7, 7, 768)", style=solid]; +"2995 contiguous_45" -> "2996 layer_norm_50" [label="(1, 7, 7, 768)", style=solid]; +"2996 layer_norm_50" -> "2997 add_81" [label="(1, 7, 7, 768)", style=solid]; +"2997 add_81" -> "2998 add_81_0_0_nncf_smooth_quant_0" [label="(1, 7, 7, 768)", style=solid]; +"2997 add_81" -> "3018 add_82" [label="(1, 7, 7, 768)", style=solid]; +"2998 add_81_0_0_nncf_smooth_quant_0" -> "2999 quantize_per_tensor_default_146" [label="(1, 7, 7, 768)", style=solid]; +"2999 quantize_per_tensor_default_146" -> "3000 dequantize_per_tensor_default_146" [label="(1, 7, 7, 768)", style=solid]; +"3000 dequantize_per_tensor_default_146" -> "3005 linear_145" [label="(1, 7, 7, 768)", style=solid]; +"3001 linear_145_scale_0" -> "3004 dequantize_per_channel_default_146" [label="(3072,)", style=solid]; +"3002 linear_145_zero_point_0" -> "3004 dequantize_per_channel_default_146" [label="(3072,)", style=solid]; +"3003 _frozen_param146" -> "3004 dequantize_per_channel_default_146" [label="(3072, 768)", style=solid]; +"3004 dequantize_per_channel_default_146" -> "3005 linear_145" [label="(3072, 768)", style=solid]; +"3005 linear_145" -> "3006 gelu_23" [label="(1, 7, 7, 3072)", style=solid]; +"3006 gelu_23" -> "3007 dropout_94" [label="(1, 7, 7, 3072)", style=solid]; +"3007 dropout_94" -> "3008 dropout_94_0_0_nncf_smooth_quant_0" [label="(1, 7, 7, 3072)", style=solid]; +"3008 dropout_94_0_0_nncf_smooth_quant_0" -> "3009 quantize_per_tensor_default_147" [label="(1, 7, 7, 3072)", style=solid]; +"3009 quantize_per_tensor_default_147" -> "3010 dequantize_per_tensor_default_147" [label="(1, 7, 7, 3072)", style=solid]; +"3010 dequantize_per_tensor_default_147" -> "3015 linear_146" [label="(1, 7, 7, 3072)", style=solid]; +"3011 linear_146_scale_0" -> "3014 dequantize_per_channel_default_147" [label="(768,)", style=solid]; +"3012 linear_146_zero_point_0" -> "3014 dequantize_per_channel_default_147" [label="(768,)", style=solid]; +"3013 _frozen_param147" -> "3014 dequantize_per_channel_default_147" [label="(768, 3072)", style=solid]; +"3014 dequantize_per_channel_default_147" -> "3015 linear_146" [label="(768, 3072)", style=solid]; +"3015 linear_146" -> "3016 dropout_95" [label="(1, 7, 7, 768)", style=solid]; +"3016 dropout_95" -> "3017 layer_norm_51" [label="(1, 7, 7, 768)", style=solid]; +"3017 layer_norm_51" -> "3018 add_82" [label="(1, 7, 7, 768)", style=solid]; +"3018 add_82" -> "3019 layer_norm_52" [label="(1, 7, 7, 768)", style=solid]; +"3019 layer_norm_52" -> "3020 permute_108" [label="(1, 7, 7, 768)", style=solid]; +"3020 permute_108" -> "3021 adaptive_avg_pool2d" [label="(1, 768, 7, 7)", style=solid]; +"3021 adaptive_avg_pool2d" -> "3022 flatten" [label="(1, 768, 1, 1)", style=solid]; +"3022 flatten" -> "3023 flatten_0_0_nncf_smooth_quant_0" [label="(1, 768)", style=solid]; +"3023 flatten_0_0_nncf_smooth_quant_0" -> "3024 quantize_per_tensor_default_148" [label="(1, 768)", style=solid]; +"3024 quantize_per_tensor_default_148" -> "3025 dequantize_per_tensor_default_148" [label="(1, 768)", style=solid]; +"3025 dequantize_per_tensor_default_148" -> "3030 linear_147" [label="(1, 768)", style=solid]; +"3026 linear_147_scale_0" -> "3029 dequantize_per_channel_default_148" [label="(1000,)", style=solid]; +"3027 linear_147_zero_point_0" -> "3029 dequantize_per_channel_default_148" [label="(1000,)", style=solid]; +"3028 _frozen_param148" -> "3029 dequantize_per_channel_default_148" [label="(1000, 768)", style=solid]; +"3029 dequantize_per_channel_default_148" -> "3030 linear_147" [label="(1000, 768)", style=solid]; +"3030 linear_147" -> "3031 output" [label="(1, 1000)", style=solid]; +} diff --git a/tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/synthetic_transformer.dot b/tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/synthetic_transformer.dot new file mode 100644 index 00000000000..462a2e45e41 --- /dev/null +++ b/tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/synthetic_transformer.dot @@ -0,0 +1,45 @@ +strict digraph { +"0 wte_weight" [id=0, type=get_attr]; +"1 linear_bias" [id=1, type=get_attr]; +"2 lm_head_bias" [id=2, type=get_attr]; +"3 input_ids" [id=3, type=input]; +"4 embedding" [id=4, type=embedding]; +"5 embedding_0_0_nncf_smooth_quant_0" [id=5, type=call_module]; +"6 quantize_per_tensor_default" [id=6, type=quantize_per_tensor]; +"7 dequantize_per_tensor_default" [id=7, type=dequantize_per_tensor]; +"8 linear_scale_0" [id=8, type=get_attr]; +"9 linear_zero_point_0" [id=9, type=get_attr]; +"10 _frozen_param0" [id=10, type=get_attr]; +"11 dequantize_per_channel_default" [id=11, type=dequantize_per_channel]; +"12 linear" [id=12, type=linear]; +"13 linear_0_0_nncf_smooth_quant_0" [id=13, type=call_module]; +"14 quantize_per_tensor_default_1" [id=14, type=quantize_per_tensor]; +"15 dequantize_per_tensor_default_1" [id=15, type=dequantize_per_tensor]; +"16 linear_1_scale_0" [id=16, type=get_attr]; +"17 linear_1_zero_point_0" [id=17, type=get_attr]; +"18 _frozen_param1" [id=18, type=get_attr]; +"19 dequantize_per_channel_default_1" [id=19, type=dequantize_per_channel]; +"20 linear_1" [id=20, type=linear]; +"21 output" [id=21, type=output]; +"0 wte_weight" -> "4 embedding" [label="(10, 5)", style=solid]; +"1 linear_bias" -> "12 linear" [label="(5,)", style=solid]; +"2 lm_head_bias" -> "20 linear_1" [label="(10,)", style=solid]; +"3 input_ids" -> "4 embedding" [label="(5,)", style=solid]; +"4 embedding" -> "5 embedding_0_0_nncf_smooth_quant_0" [label="(5, 5)", style=solid]; +"5 embedding_0_0_nncf_smooth_quant_0" -> "6 quantize_per_tensor_default" [label="(5, 5)", style=solid]; +"6 quantize_per_tensor_default" -> "7 dequantize_per_tensor_default" [label="(5, 5)", style=solid]; +"7 dequantize_per_tensor_default" -> "12 linear" [label="(5, 5)", style=solid]; +"8 linear_scale_0" -> "11 dequantize_per_channel_default" [label="(5,)", style=solid]; +"9 linear_zero_point_0" -> "11 dequantize_per_channel_default" [label="(5,)", style=solid]; +"10 _frozen_param0" -> "11 dequantize_per_channel_default" [label="(5, 5)", style=solid]; +"11 dequantize_per_channel_default" -> "12 linear" [label="(5, 5)", style=solid]; +"12 linear" -> "13 linear_0_0_nncf_smooth_quant_0" [label="(5, 5)", style=solid]; +"13 linear_0_0_nncf_smooth_quant_0" -> "14 quantize_per_tensor_default_1" [label="(5, 5)", style=solid]; +"14 quantize_per_tensor_default_1" -> "15 dequantize_per_tensor_default_1" [label="(5, 5)", style=solid]; +"15 dequantize_per_tensor_default_1" -> "20 linear_1" [label="(5, 5)", style=solid]; +"16 linear_1_scale_0" -> "19 dequantize_per_channel_default_1" [label="(10,)", style=solid]; +"17 linear_1_zero_point_0" -> "19 dequantize_per_channel_default_1" [label="(10,)", style=solid]; +"18 _frozen_param1" -> "19 dequantize_per_channel_default_1" [label="(10, 5)", style=solid]; +"19 dequantize_per_channel_default_1" -> "20 linear_1" [label="(10, 5)", style=solid]; +"20 linear_1" -> "21 output" [label="(5, 10)", style=solid]; +} diff --git a/tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/unet.dot b/tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/unet.dot new file mode 100644 index 00000000000..c1ab3d80c09 --- /dev/null +++ b/tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/unet.dot @@ -0,0 +1,469 @@ +strict digraph { +"0 down_path_0_block_0_bias" [id=0, type=get_attr]; +"1 down_path_0_block_3_bias" [id=1, type=get_attr]; +"2 down_path_1_block_0_bias" [id=2, type=get_attr]; +"3 down_path_1_block_3_bias" [id=3, type=get_attr]; +"4 down_path_2_block_0_bias" [id=4, type=get_attr]; +"5 down_path_2_block_3_bias" [id=5, type=get_attr]; +"6 down_path_3_block_0_bias" [id=6, type=get_attr]; +"7 down_path_3_block_3_bias" [id=7, type=get_attr]; +"8 down_path_4_block_0_bias" [id=8, type=get_attr]; +"9 down_path_4_block_3_bias" [id=9, type=get_attr]; +"10 up_path_0_up_bias" [id=10, type=get_attr]; +"11 up_path_0_conv_block_block_0_bias" [id=11, type=get_attr]; +"12 up_path_0_conv_block_block_3_bias" [id=12, type=get_attr]; +"13 up_path_1_up_bias" [id=13, type=get_attr]; +"14 up_path_1_conv_block_block_0_bias" [id=14, type=get_attr]; +"15 up_path_1_conv_block_block_3_bias" [id=15, type=get_attr]; +"16 up_path_2_up_bias" [id=16, type=get_attr]; +"17 up_path_2_conv_block_block_0_bias" [id=17, type=get_attr]; +"18 up_path_2_conv_block_block_3_bias" [id=18, type=get_attr]; +"19 up_path_3_up_bias" [id=19, type=get_attr]; +"20 up_path_3_conv_block_block_0_bias" [id=20, type=get_attr]; +"21 up_path_3_conv_block_block_3_bias" [id=21, type=get_attr]; +"22 last_bias" [id=22, type=get_attr]; +"23 x" [id=23, type=input]; +"24 quantize_per_tensor_default_8" [id=24, type=quantize_per_tensor]; +"25 dequantize_per_tensor_default_12" [id=25, type=dequantize_per_tensor]; +"26 conv2d_scale_0" [id=26, type=get_attr]; +"27 conv2d_zero_point_0" [id=27, type=get_attr]; +"28 _frozen_param0" [id=28, type=get_attr]; +"29 dequantize_per_channel_default" [id=29, type=dequantize_per_channel]; +"30 conv2d" [id=30, type=conv2d]; +"31 relu" [id=31, type=relu]; +"32 quantize_per_tensor_default_9" [id=32, type=quantize_per_tensor]; +"33 dequantize_per_tensor_default_13" [id=33, type=dequantize_per_tensor]; +"34 conv2d_1_scale_0" [id=34, type=get_attr]; +"35 conv2d_1_zero_point_0" [id=35, type=get_attr]; +"36 _frozen_param1" [id=36, type=get_attr]; +"37 dequantize_per_channel_default_1" [id=37, type=dequantize_per_channel]; +"38 conv2d_1" [id=38, type=conv2d]; +"39 relu_1" [id=39, type=relu]; +"40 quantize_per_tensor_default" [id=40, type=quantize_per_tensor]; +"41 dequantize_per_tensor_default_1" [id=41, type=dequantize_per_tensor]; +"42 dequantize_per_tensor_default" [id=42, type=dequantize_per_tensor]; +"43 max_pool2d" [id=43, type=max_pool2d]; +"44 conv2d_2_scale_0" [id=44, type=get_attr]; +"45 conv2d_2_zero_point_0" [id=45, type=get_attr]; +"46 _frozen_param2" [id=46, type=get_attr]; +"47 dequantize_per_channel_default_2" [id=47, type=dequantize_per_channel]; +"48 conv2d_2" [id=48, type=conv2d]; +"49 relu_2" [id=49, type=relu]; +"50 quantize_per_tensor_default_10" [id=50, type=quantize_per_tensor]; +"51 dequantize_per_tensor_default_14" [id=51, type=dequantize_per_tensor]; +"52 conv2d_3_scale_0" [id=52, type=get_attr]; +"53 conv2d_3_zero_point_0" [id=53, type=get_attr]; +"54 _frozen_param3" [id=54, type=get_attr]; +"55 dequantize_per_channel_default_3" [id=55, type=dequantize_per_channel]; +"56 conv2d_3" [id=56, type=conv2d]; +"57 relu_3" [id=57, type=relu]; +"58 quantize_per_tensor_default_3" [id=58, type=quantize_per_tensor]; +"59 dequantize_per_tensor_default_5" [id=59, type=dequantize_per_tensor]; +"60 dequantize_per_tensor_default_4" [id=60, type=dequantize_per_tensor]; +"61 max_pool2d_1" [id=61, type=max_pool2d]; +"62 conv2d_4_scale_0" [id=62, type=get_attr]; +"63 conv2d_4_zero_point_0" [id=63, type=get_attr]; +"64 _frozen_param4" [id=64, type=get_attr]; +"65 dequantize_per_channel_default_4" [id=65, type=dequantize_per_channel]; +"66 conv2d_4" [id=66, type=conv2d]; +"67 relu_4" [id=67, type=relu]; +"68 quantize_per_tensor_default_11" [id=68, type=quantize_per_tensor]; +"69 dequantize_per_tensor_default_15" [id=69, type=dequantize_per_tensor]; +"70 conv2d_5_scale_0" [id=70, type=get_attr]; +"71 conv2d_5_zero_point_0" [id=71, type=get_attr]; +"72 _frozen_param5" [id=72, type=get_attr]; +"73 dequantize_per_channel_default_5" [id=73, type=dequantize_per_channel]; +"74 conv2d_5" [id=74, type=conv2d]; +"75 relu_5" [id=75, type=relu]; +"76 quantize_per_tensor_default_4" [id=76, type=quantize_per_tensor]; +"77 dequantize_per_tensor_default_7" [id=77, type=dequantize_per_tensor]; +"78 dequantize_per_tensor_default_6" [id=78, type=dequantize_per_tensor]; +"79 max_pool2d_2" [id=79, type=max_pool2d]; +"80 conv2d_6_scale_0" [id=80, type=get_attr]; +"81 conv2d_6_zero_point_0" [id=81, type=get_attr]; +"82 _frozen_param6" [id=82, type=get_attr]; +"83 dequantize_per_channel_default_6" [id=83, type=dequantize_per_channel]; +"84 conv2d_6" [id=84, type=conv2d]; +"85 relu_6" [id=85, type=relu]; +"86 quantize_per_tensor_default_12" [id=86, type=quantize_per_tensor]; +"87 dequantize_per_tensor_default_16" [id=87, type=dequantize_per_tensor]; +"88 conv2d_7_scale_0" [id=88, type=get_attr]; +"89 conv2d_7_zero_point_0" [id=89, type=get_attr]; +"90 _frozen_param7" [id=90, type=get_attr]; +"91 dequantize_per_channel_default_7" [id=91, type=dequantize_per_channel]; +"92 conv2d_7" [id=92, type=conv2d]; +"93 relu_7" [id=93, type=relu]; +"94 quantize_per_tensor_default_7" [id=94, type=quantize_per_tensor]; +"95 dequantize_per_tensor_default_11" [id=95, type=dequantize_per_tensor]; +"96 dequantize_per_tensor_default_10" [id=96, type=dequantize_per_tensor]; +"97 max_pool2d_3" [id=97, type=max_pool2d]; +"98 conv2d_8_scale_0" [id=98, type=get_attr]; +"99 conv2d_8_zero_point_0" [id=99, type=get_attr]; +"100 _frozen_param8" [id=100, type=get_attr]; +"101 dequantize_per_channel_default_8" [id=101, type=dequantize_per_channel]; +"102 conv2d_8" [id=102, type=conv2d]; +"103 relu_8" [id=103, type=relu]; +"104 quantize_per_tensor_default_13" [id=104, type=quantize_per_tensor]; +"105 dequantize_per_tensor_default_17" [id=105, type=dequantize_per_tensor]; +"106 conv2d_9_scale_0" [id=106, type=get_attr]; +"107 conv2d_9_zero_point_0" [id=107, type=get_attr]; +"108 _frozen_param9" [id=108, type=get_attr]; +"109 dequantize_per_channel_default_9" [id=109, type=dequantize_per_channel]; +"110 conv2d_9" [id=110, type=conv2d]; +"111 relu_9" [id=111, type=relu]; +"112 quantize_per_tensor_default_14" [id=112, type=quantize_per_tensor]; +"113 dequantize_per_tensor_default_18" [id=113, type=dequantize_per_tensor]; +"114 conv_transpose2d_scale_0" [id=114, type=get_attr]; +"115 conv_transpose2d_zero_point_0" [id=115, type=get_attr]; +"116 _frozen_param10" [id=116, type=get_attr]; +"117 dequantize_per_channel_default_10" [id=117, type=dequantize_per_channel]; +"118 conv_transpose2d" [id=118, type=conv_transpose2d]; +"119 quantize_per_tensor_default_6" [id=119, type=quantize_per_tensor]; +"120 dequantize_per_tensor_default_9" [id=120, type=dequantize_per_tensor]; +"121 slice_1" [id=121, type=slice]; +"122 slice_2" [id=122, type=slice]; +"123 slice_3" [id=123, type=slice]; +"124 slice_4" [id=124, type=slice]; +"125 cat" [id=125, type=cat]; +"126 conv2d_10_scale_0" [id=126, type=get_attr]; +"127 conv2d_10_zero_point_0" [id=127, type=get_attr]; +"128 _frozen_param11" [id=128, type=get_attr]; +"129 dequantize_per_channel_default_11" [id=129, type=dequantize_per_channel]; +"130 conv2d_10" [id=130, type=conv2d]; +"131 relu_10" [id=131, type=relu]; +"132 quantize_per_tensor_default_15" [id=132, type=quantize_per_tensor]; +"133 dequantize_per_tensor_default_19" [id=133, type=dequantize_per_tensor]; +"134 conv2d_11_scale_0" [id=134, type=get_attr]; +"135 conv2d_11_zero_point_0" [id=135, type=get_attr]; +"136 _frozen_param12" [id=136, type=get_attr]; +"137 dequantize_per_channel_default_12" [id=137, type=dequantize_per_channel]; +"138 conv2d_11" [id=138, type=conv2d]; +"139 relu_11" [id=139, type=relu]; +"140 quantize_per_tensor_default_16" [id=140, type=quantize_per_tensor]; +"141 dequantize_per_tensor_default_20" [id=141, type=dequantize_per_tensor]; +"142 conv_transpose2d_1_scale_0" [id=142, type=get_attr]; +"143 conv_transpose2d_1_zero_point_0" [id=143, type=get_attr]; +"144 _frozen_param13" [id=144, type=get_attr]; +"145 dequantize_per_channel_default_13" [id=145, type=dequantize_per_channel]; +"146 conv_transpose2d_1" [id=146, type=conv_transpose2d]; +"147 quantize_per_tensor_default_5" [id=147, type=quantize_per_tensor]; +"148 dequantize_per_tensor_default_8" [id=148, type=dequantize_per_tensor]; +"149 slice_5" [id=149, type=slice]; +"150 slice_6" [id=150, type=slice]; +"151 slice_7" [id=151, type=slice]; +"152 slice_8" [id=152, type=slice]; +"153 cat_1" [id=153, type=cat]; +"154 conv2d_12_scale_0" [id=154, type=get_attr]; +"155 conv2d_12_zero_point_0" [id=155, type=get_attr]; +"156 _frozen_param14" [id=156, type=get_attr]; +"157 dequantize_per_channel_default_14" [id=157, type=dequantize_per_channel]; +"158 conv2d_12" [id=158, type=conv2d]; +"159 relu_12" [id=159, type=relu]; +"160 quantize_per_tensor_default_17" [id=160, type=quantize_per_tensor]; +"161 dequantize_per_tensor_default_21" [id=161, type=dequantize_per_tensor]; +"162 conv2d_13_scale_0" [id=162, type=get_attr]; +"163 conv2d_13_zero_point_0" [id=163, type=get_attr]; +"164 _frozen_param15" [id=164, type=get_attr]; +"165 dequantize_per_channel_default_15" [id=165, type=dequantize_per_channel]; +"166 conv2d_13" [id=166, type=conv2d]; +"167 relu_13" [id=167, type=relu]; +"168 quantize_per_tensor_default_18" [id=168, type=quantize_per_tensor]; +"169 dequantize_per_tensor_default_22" [id=169, type=dequantize_per_tensor]; +"170 conv_transpose2d_2_scale_0" [id=170, type=get_attr]; +"171 conv_transpose2d_2_zero_point_0" [id=171, type=get_attr]; +"172 _frozen_param16" [id=172, type=get_attr]; +"173 dequantize_per_channel_default_16" [id=173, type=dequantize_per_channel]; +"174 conv_transpose2d_2" [id=174, type=conv_transpose2d]; +"175 quantize_per_tensor_default_2" [id=175, type=quantize_per_tensor]; +"176 dequantize_per_tensor_default_3" [id=176, type=dequantize_per_tensor]; +"177 slice_9" [id=177, type=slice]; +"178 slice_10" [id=178, type=slice]; +"179 slice_11" [id=179, type=slice]; +"180 slice_12" [id=180, type=slice]; +"181 cat_2" [id=181, type=cat]; +"182 conv2d_14_scale_0" [id=182, type=get_attr]; +"183 conv2d_14_zero_point_0" [id=183, type=get_attr]; +"184 _frozen_param17" [id=184, type=get_attr]; +"185 dequantize_per_channel_default_17" [id=185, type=dequantize_per_channel]; +"186 conv2d_14" [id=186, type=conv2d]; +"187 relu_14" [id=187, type=relu]; +"188 quantize_per_tensor_default_19" [id=188, type=quantize_per_tensor]; +"189 dequantize_per_tensor_default_23" [id=189, type=dequantize_per_tensor]; +"190 conv2d_15_scale_0" [id=190, type=get_attr]; +"191 conv2d_15_zero_point_0" [id=191, type=get_attr]; +"192 _frozen_param18" [id=192, type=get_attr]; +"193 dequantize_per_channel_default_18" [id=193, type=dequantize_per_channel]; +"194 conv2d_15" [id=194, type=conv2d]; +"195 relu_15" [id=195, type=relu]; +"196 quantize_per_tensor_default_20" [id=196, type=quantize_per_tensor]; +"197 dequantize_per_tensor_default_24" [id=197, type=dequantize_per_tensor]; +"198 conv_transpose2d_3_scale_0" [id=198, type=get_attr]; +"199 conv_transpose2d_3_zero_point_0" [id=199, type=get_attr]; +"200 _frozen_param19" [id=200, type=get_attr]; +"201 dequantize_per_channel_default_19" [id=201, type=dequantize_per_channel]; +"202 conv_transpose2d_3" [id=202, type=conv_transpose2d]; +"203 quantize_per_tensor_default_1" [id=203, type=quantize_per_tensor]; +"204 dequantize_per_tensor_default_2" [id=204, type=dequantize_per_tensor]; +"205 slice_13" [id=205, type=slice]; +"206 slice_14" [id=206, type=slice]; +"207 slice_15" [id=207, type=slice]; +"208 slice_16" [id=208, type=slice]; +"209 cat_3" [id=209, type=cat]; +"210 conv2d_16_scale_0" [id=210, type=get_attr]; +"211 conv2d_16_zero_point_0" [id=211, type=get_attr]; +"212 _frozen_param20" [id=212, type=get_attr]; +"213 dequantize_per_channel_default_20" [id=213, type=dequantize_per_channel]; +"214 conv2d_16" [id=214, type=conv2d]; +"215 relu_16" [id=215, type=relu]; +"216 quantize_per_tensor_default_21" [id=216, type=quantize_per_tensor]; +"217 dequantize_per_tensor_default_25" [id=217, type=dequantize_per_tensor]; +"218 conv2d_17_scale_0" [id=218, type=get_attr]; +"219 conv2d_17_zero_point_0" [id=219, type=get_attr]; +"220 _frozen_param21" [id=220, type=get_attr]; +"221 dequantize_per_channel_default_21" [id=221, type=dequantize_per_channel]; +"222 conv2d_17" [id=222, type=conv2d]; +"223 relu_17" [id=223, type=relu]; +"224 quantize_per_tensor_default_22" [id=224, type=quantize_per_tensor]; +"225 dequantize_per_tensor_default_26" [id=225, type=dequantize_per_tensor]; +"226 conv2d_18_scale_0" [id=226, type=get_attr]; +"227 conv2d_18_zero_point_0" [id=227, type=get_attr]; +"228 _frozen_param22" [id=228, type=get_attr]; +"229 dequantize_per_channel_default_22" [id=229, type=dequantize_per_channel]; +"230 conv2d_18" [id=230, type=conv2d]; +"231 output" [id=231, type=output]; +"0 down_path_0_block_0_bias" -> "30 conv2d" [label="(64,)", style=solid]; +"1 down_path_0_block_3_bias" -> "38 conv2d_1" [label="(64,)", style=solid]; +"2 down_path_1_block_0_bias" -> "48 conv2d_2" [label="(128,)", style=solid]; +"3 down_path_1_block_3_bias" -> "56 conv2d_3" [label="(128,)", style=solid]; +"4 down_path_2_block_0_bias" -> "66 conv2d_4" [label="(256,)", style=solid]; +"5 down_path_2_block_3_bias" -> "74 conv2d_5" [label="(256,)", style=solid]; +"6 down_path_3_block_0_bias" -> "84 conv2d_6" [label="(512,)", style=solid]; +"7 down_path_3_block_3_bias" -> "92 conv2d_7" [label="(512,)", style=solid]; +"8 down_path_4_block_0_bias" -> "102 conv2d_8" [label="(1024,)", style=solid]; +"9 down_path_4_block_3_bias" -> "110 conv2d_9" [label="(1024,)", style=solid]; +"10 up_path_0_up_bias" -> "118 conv_transpose2d" [label="(512,)", style=solid]; +"11 up_path_0_conv_block_block_0_bias" -> "130 conv2d_10" [label="(512,)", style=solid]; +"12 up_path_0_conv_block_block_3_bias" -> "138 conv2d_11" [label="(512,)", style=solid]; +"13 up_path_1_up_bias" -> "146 conv_transpose2d_1" [label="(256,)", style=solid]; +"14 up_path_1_conv_block_block_0_bias" -> "158 conv2d_12" [label="(256,)", style=solid]; +"15 up_path_1_conv_block_block_3_bias" -> "166 conv2d_13" [label="(256,)", style=solid]; +"16 up_path_2_up_bias" -> "174 conv_transpose2d_2" [label="(128,)", style=solid]; +"17 up_path_2_conv_block_block_0_bias" -> "186 conv2d_14" [label="(128,)", style=solid]; +"18 up_path_2_conv_block_block_3_bias" -> "194 conv2d_15" [label="(128,)", style=solid]; +"19 up_path_3_up_bias" -> "202 conv_transpose2d_3" [label="(64,)", style=solid]; +"20 up_path_3_conv_block_block_0_bias" -> "214 conv2d_16" [label="(64,)", style=solid]; +"21 up_path_3_conv_block_block_3_bias" -> "222 conv2d_17" [label="(64,)", style=solid]; +"22 last_bias" -> "230 conv2d_18" [label="(12,)", style=solid]; +"23 x" -> "24 quantize_per_tensor_default_8" [label="(1, 3, 224, 224)", style=solid]; +"24 quantize_per_tensor_default_8" -> "25 dequantize_per_tensor_default_12" [label="(1, 3, 224, 224)", style=solid]; +"25 dequantize_per_tensor_default_12" -> "30 conv2d" [label="(1, 3, 224, 224)", style=solid]; +"26 conv2d_scale_0" -> "29 dequantize_per_channel_default" [label="(64,)", style=solid]; +"27 conv2d_zero_point_0" -> "29 dequantize_per_channel_default" [label="(64,)", style=solid]; +"28 _frozen_param0" -> "29 dequantize_per_channel_default" [label="(64, 3, 3, 3)", style=solid]; +"29 dequantize_per_channel_default" -> "30 conv2d" [label="(64, 3, 3, 3)", style=solid]; +"30 conv2d" -> "31 relu" [label="(1, 64, 222, 222)", style=solid]; +"31 relu" -> "32 quantize_per_tensor_default_9" [label="(1, 64, 222, 222)", style=solid]; +"32 quantize_per_tensor_default_9" -> "33 dequantize_per_tensor_default_13" [label="(1, 64, 222, 222)", style=solid]; +"33 dequantize_per_tensor_default_13" -> "38 conv2d_1" [label="(1, 64, 222, 222)", style=solid]; +"34 conv2d_1_scale_0" -> "37 dequantize_per_channel_default_1" [label="(64,)", style=solid]; +"35 conv2d_1_zero_point_0" -> "37 dequantize_per_channel_default_1" [label="(64,)", style=solid]; +"36 _frozen_param1" -> "37 dequantize_per_channel_default_1" [label="(64, 64, 3, 3)", style=solid]; +"37 dequantize_per_channel_default_1" -> "38 conv2d_1" [label="(64, 64, 3, 3)", style=solid]; +"38 conv2d_1" -> "39 relu_1" [label="(1, 64, 220, 220)", style=solid]; +"39 relu_1" -> "40 quantize_per_tensor_default" [label="(1, 64, 220, 220)", style=solid]; +"40 quantize_per_tensor_default" -> "41 dequantize_per_tensor_default_1" [label="(1, 64, 220, 220)", style=solid]; +"40 quantize_per_tensor_default" -> "42 dequantize_per_tensor_default" [label="(1, 64, 220, 220)", style=solid]; +"41 dequantize_per_tensor_default_1" -> "205 slice_13" [label="(1, 64, 220, 220)", style=solid]; +"42 dequantize_per_tensor_default" -> "43 max_pool2d" [label="(1, 64, 220, 220)", style=solid]; +"43 max_pool2d" -> "48 conv2d_2" [label="(1, 64, 110, 110)", style=solid]; +"44 conv2d_2_scale_0" -> "47 dequantize_per_channel_default_2" [label="(128,)", style=solid]; +"45 conv2d_2_zero_point_0" -> "47 dequantize_per_channel_default_2" [label="(128,)", style=solid]; +"46 _frozen_param2" -> "47 dequantize_per_channel_default_2" [label="(128, 64, 3, 3)", style=solid]; +"47 dequantize_per_channel_default_2" -> "48 conv2d_2" [label="(128, 64, 3, 3)", style=solid]; +"48 conv2d_2" -> "49 relu_2" [label="(1, 128, 108, 108)", style=solid]; +"49 relu_2" -> "50 quantize_per_tensor_default_10" [label="(1, 128, 108, 108)", style=solid]; +"50 quantize_per_tensor_default_10" -> "51 dequantize_per_tensor_default_14" [label="(1, 128, 108, 108)", style=solid]; +"51 dequantize_per_tensor_default_14" -> "56 conv2d_3" [label="(1, 128, 108, 108)", style=solid]; +"52 conv2d_3_scale_0" -> "55 dequantize_per_channel_default_3" [label="(128,)", style=solid]; +"53 conv2d_3_zero_point_0" -> "55 dequantize_per_channel_default_3" [label="(128,)", style=solid]; +"54 _frozen_param3" -> "55 dequantize_per_channel_default_3" [label="(128, 128, 3, 3)", style=solid]; +"55 dequantize_per_channel_default_3" -> "56 conv2d_3" [label="(128, 128, 3, 3)", style=solid]; +"56 conv2d_3" -> "57 relu_3" [label="(1, 128, 106, 106)", style=solid]; +"57 relu_3" -> "58 quantize_per_tensor_default_3" [label="(1, 128, 106, 106)", style=solid]; +"58 quantize_per_tensor_default_3" -> "59 dequantize_per_tensor_default_5" [label="(1, 128, 106, 106)", style=solid]; +"58 quantize_per_tensor_default_3" -> "60 dequantize_per_tensor_default_4" [label="(1, 128, 106, 106)", style=solid]; +"59 dequantize_per_tensor_default_5" -> "177 slice_9" [label="(1, 128, 106, 106)", style=solid]; +"60 dequantize_per_tensor_default_4" -> "61 max_pool2d_1" [label="(1, 128, 106, 106)", style=solid]; +"61 max_pool2d_1" -> "66 conv2d_4" [label="(1, 128, 53, 53)", style=solid]; +"62 conv2d_4_scale_0" -> "65 dequantize_per_channel_default_4" [label="(256,)", style=solid]; +"63 conv2d_4_zero_point_0" -> "65 dequantize_per_channel_default_4" [label="(256,)", style=solid]; +"64 _frozen_param4" -> "65 dequantize_per_channel_default_4" [label="(256, 128, 3, 3)", style=solid]; +"65 dequantize_per_channel_default_4" -> "66 conv2d_4" [label="(256, 128, 3, 3)", style=solid]; +"66 conv2d_4" -> "67 relu_4" [label="(1, 256, 51, 51)", style=solid]; +"67 relu_4" -> "68 quantize_per_tensor_default_11" [label="(1, 256, 51, 51)", style=solid]; +"68 quantize_per_tensor_default_11" -> "69 dequantize_per_tensor_default_15" [label="(1, 256, 51, 51)", style=solid]; +"69 dequantize_per_tensor_default_15" -> "74 conv2d_5" [label="(1, 256, 51, 51)", style=solid]; +"70 conv2d_5_scale_0" -> "73 dequantize_per_channel_default_5" [label="(256,)", style=solid]; +"71 conv2d_5_zero_point_0" -> "73 dequantize_per_channel_default_5" [label="(256,)", style=solid]; +"72 _frozen_param5" -> "73 dequantize_per_channel_default_5" [label="(256, 256, 3, 3)", style=solid]; +"73 dequantize_per_channel_default_5" -> "74 conv2d_5" [label="(256, 256, 3, 3)", style=solid]; +"74 conv2d_5" -> "75 relu_5" [label="(1, 256, 49, 49)", style=solid]; +"75 relu_5" -> "76 quantize_per_tensor_default_4" [label="(1, 256, 49, 49)", style=solid]; +"76 quantize_per_tensor_default_4" -> "77 dequantize_per_tensor_default_7" [label="(1, 256, 49, 49)", style=solid]; +"76 quantize_per_tensor_default_4" -> "78 dequantize_per_tensor_default_6" [label="(1, 256, 49, 49)", style=solid]; +"77 dequantize_per_tensor_default_7" -> "149 slice_5" [label="(1, 256, 49, 49)", style=solid]; +"78 dequantize_per_tensor_default_6" -> "79 max_pool2d_2" [label="(1, 256, 49, 49)", style=solid]; +"79 max_pool2d_2" -> "84 conv2d_6" [label="(1, 256, 24, 24)", style=solid]; +"80 conv2d_6_scale_0" -> "83 dequantize_per_channel_default_6" [label="(512,)", style=solid]; +"81 conv2d_6_zero_point_0" -> "83 dequantize_per_channel_default_6" [label="(512,)", style=solid]; +"82 _frozen_param6" -> "83 dequantize_per_channel_default_6" [label="(512, 256, 3, 3)", style=solid]; +"83 dequantize_per_channel_default_6" -> "84 conv2d_6" [label="(512, 256, 3, 3)", style=solid]; +"84 conv2d_6" -> "85 relu_6" [label="(1, 512, 22, 22)", style=solid]; +"85 relu_6" -> "86 quantize_per_tensor_default_12" [label="(1, 512, 22, 22)", style=solid]; +"86 quantize_per_tensor_default_12" -> "87 dequantize_per_tensor_default_16" [label="(1, 512, 22, 22)", style=solid]; +"87 dequantize_per_tensor_default_16" -> "92 conv2d_7" [label="(1, 512, 22, 22)", style=solid]; +"88 conv2d_7_scale_0" -> "91 dequantize_per_channel_default_7" [label="(512,)", style=solid]; +"89 conv2d_7_zero_point_0" -> "91 dequantize_per_channel_default_7" [label="(512,)", style=solid]; +"90 _frozen_param7" -> "91 dequantize_per_channel_default_7" [label="(512, 512, 3, 3)", style=solid]; +"91 dequantize_per_channel_default_7" -> "92 conv2d_7" [label="(512, 512, 3, 3)", style=solid]; +"92 conv2d_7" -> "93 relu_7" [label="(1, 512, 20, 20)", style=solid]; +"93 relu_7" -> "94 quantize_per_tensor_default_7" [label="(1, 512, 20, 20)", style=solid]; +"94 quantize_per_tensor_default_7" -> "95 dequantize_per_tensor_default_11" [label="(1, 512, 20, 20)", style=solid]; +"94 quantize_per_tensor_default_7" -> "96 dequantize_per_tensor_default_10" [label="(1, 512, 20, 20)", style=solid]; +"95 dequantize_per_tensor_default_11" -> "121 slice_1" [label="(1, 512, 20, 20)", style=solid]; +"96 dequantize_per_tensor_default_10" -> "97 max_pool2d_3" [label="(1, 512, 20, 20)", style=solid]; +"97 max_pool2d_3" -> "102 conv2d_8" [label="(1, 512, 10, 10)", style=solid]; +"98 conv2d_8_scale_0" -> "101 dequantize_per_channel_default_8" [label="(1024,)", style=solid]; +"99 conv2d_8_zero_point_0" -> "101 dequantize_per_channel_default_8" [label="(1024,)", style=solid]; +"100 _frozen_param8" -> "101 dequantize_per_channel_default_8" [label="(1024, 512, 3, 3)", style=solid]; +"101 dequantize_per_channel_default_8" -> "102 conv2d_8" [label="(1024, 512, 3, 3)", style=solid]; +"102 conv2d_8" -> "103 relu_8" [label="(1, 1024, 8, 8)", style=solid]; +"103 relu_8" -> "104 quantize_per_tensor_default_13" [label="(1, 1024, 8, 8)", style=solid]; +"104 quantize_per_tensor_default_13" -> "105 dequantize_per_tensor_default_17" [label="(1, 1024, 8, 8)", style=solid]; +"105 dequantize_per_tensor_default_17" -> "110 conv2d_9" [label="(1, 1024, 8, 8)", style=solid]; +"106 conv2d_9_scale_0" -> "109 dequantize_per_channel_default_9" [label="(1024,)", style=solid]; +"107 conv2d_9_zero_point_0" -> "109 dequantize_per_channel_default_9" [label="(1024,)", style=solid]; +"108 _frozen_param9" -> "109 dequantize_per_channel_default_9" [label="(1024, 1024, 3, 3)", style=solid]; +"109 dequantize_per_channel_default_9" -> "110 conv2d_9" [label="(1024, 1024, 3, 3)", style=solid]; +"110 conv2d_9" -> "111 relu_9" [label="(1, 1024, 6, 6)", style=solid]; +"111 relu_9" -> "112 quantize_per_tensor_default_14" [label="(1, 1024, 6, 6)", style=solid]; +"112 quantize_per_tensor_default_14" -> "113 dequantize_per_tensor_default_18" [label="(1, 1024, 6, 6)", style=solid]; +"113 dequantize_per_tensor_default_18" -> "118 conv_transpose2d" [label="(1, 1024, 6, 6)", style=solid]; +"114 conv_transpose2d_scale_0" -> "117 dequantize_per_channel_default_10" [label="(1024,)", style=solid]; +"115 conv_transpose2d_zero_point_0" -> "117 dequantize_per_channel_default_10" [label="(1024,)", style=solid]; +"116 _frozen_param10" -> "117 dequantize_per_channel_default_10" [label="(1024, 512, 2, 2)", style=solid]; +"117 dequantize_per_channel_default_10" -> "118 conv_transpose2d" [label="(1024, 512, 2, 2)", style=solid]; +"118 conv_transpose2d" -> "119 quantize_per_tensor_default_6" [label="(1, 512, 12, 12)", style=solid]; +"119 quantize_per_tensor_default_6" -> "120 dequantize_per_tensor_default_9" [label="(1, 512, 12, 12)", style=solid]; +"120 dequantize_per_tensor_default_9" -> "125 cat" [label="(1, 512, 12, 12)", style=solid]; +"121 slice_1" -> "122 slice_2" [label="(1, 512, 20, 20)", style=solid]; +"122 slice_2" -> "123 slice_3" [label="(1, 512, 20, 20)", style=solid]; +"123 slice_3" -> "124 slice_4" [label="(1, 512, 12, 20)", style=solid]; +"124 slice_4" -> "125 cat" [label="(1, 512, 12, 12)", style=solid]; +"125 cat" -> "130 conv2d_10" [label="(1, 1024, 12, 12)", style=solid]; +"126 conv2d_10_scale_0" -> "129 dequantize_per_channel_default_11" [label="(512,)", style=solid]; +"127 conv2d_10_zero_point_0" -> "129 dequantize_per_channel_default_11" [label="(512,)", style=solid]; +"128 _frozen_param11" -> "129 dequantize_per_channel_default_11" [label="(512, 1024, 3, 3)", style=solid]; +"129 dequantize_per_channel_default_11" -> "130 conv2d_10" [label="(512, 1024, 3, 3)", style=solid]; +"130 conv2d_10" -> "131 relu_10" [label="(1, 512, 10, 10)", style=solid]; +"131 relu_10" -> "132 quantize_per_tensor_default_15" [label="(1, 512, 10, 10)", style=solid]; +"132 quantize_per_tensor_default_15" -> "133 dequantize_per_tensor_default_19" [label="(1, 512, 10, 10)", style=solid]; +"133 dequantize_per_tensor_default_19" -> "138 conv2d_11" [label="(1, 512, 10, 10)", style=solid]; +"134 conv2d_11_scale_0" -> "137 dequantize_per_channel_default_12" [label="(512,)", style=solid]; +"135 conv2d_11_zero_point_0" -> "137 dequantize_per_channel_default_12" [label="(512,)", style=solid]; +"136 _frozen_param12" -> "137 dequantize_per_channel_default_12" [label="(512, 512, 3, 3)", style=solid]; +"137 dequantize_per_channel_default_12" -> "138 conv2d_11" [label="(512, 512, 3, 3)", style=solid]; +"138 conv2d_11" -> "139 relu_11" [label="(1, 512, 8, 8)", style=solid]; +"139 relu_11" -> "140 quantize_per_tensor_default_16" [label="(1, 512, 8, 8)", style=solid]; +"140 quantize_per_tensor_default_16" -> "141 dequantize_per_tensor_default_20" [label="(1, 512, 8, 8)", style=solid]; +"141 dequantize_per_tensor_default_20" -> "146 conv_transpose2d_1" [label="(1, 512, 8, 8)", style=solid]; +"142 conv_transpose2d_1_scale_0" -> "145 dequantize_per_channel_default_13" [label="(512,)", style=solid]; +"143 conv_transpose2d_1_zero_point_0" -> "145 dequantize_per_channel_default_13" [label="(512,)", style=solid]; +"144 _frozen_param13" -> "145 dequantize_per_channel_default_13" [label="(512, 256, 2, 2)", style=solid]; +"145 dequantize_per_channel_default_13" -> "146 conv_transpose2d_1" [label="(512, 256, 2, 2)", style=solid]; +"146 conv_transpose2d_1" -> "147 quantize_per_tensor_default_5" [label="(1, 256, 16, 16)", style=solid]; +"147 quantize_per_tensor_default_5" -> "148 dequantize_per_tensor_default_8" [label="(1, 256, 16, 16)", style=solid]; +"148 dequantize_per_tensor_default_8" -> "153 cat_1" [label="(1, 256, 16, 16)", style=solid]; +"149 slice_5" -> "150 slice_6" [label="(1, 256, 49, 49)", style=solid]; +"150 slice_6" -> "151 slice_7" [label="(1, 256, 49, 49)", style=solid]; +"151 slice_7" -> "152 slice_8" [label="(1, 256, 16, 49)", style=solid]; +"152 slice_8" -> "153 cat_1" [label="(1, 256, 16, 16)", style=solid]; +"153 cat_1" -> "158 conv2d_12" [label="(1, 512, 16, 16)", style=solid]; +"154 conv2d_12_scale_0" -> "157 dequantize_per_channel_default_14" [label="(256,)", style=solid]; +"155 conv2d_12_zero_point_0" -> "157 dequantize_per_channel_default_14" [label="(256,)", style=solid]; +"156 _frozen_param14" -> "157 dequantize_per_channel_default_14" [label="(256, 512, 3, 3)", style=solid]; +"157 dequantize_per_channel_default_14" -> "158 conv2d_12" [label="(256, 512, 3, 3)", style=solid]; +"158 conv2d_12" -> "159 relu_12" [label="(1, 256, 14, 14)", style=solid]; +"159 relu_12" -> "160 quantize_per_tensor_default_17" [label="(1, 256, 14, 14)", style=solid]; +"160 quantize_per_tensor_default_17" -> "161 dequantize_per_tensor_default_21" [label="(1, 256, 14, 14)", style=solid]; +"161 dequantize_per_tensor_default_21" -> "166 conv2d_13" [label="(1, 256, 14, 14)", style=solid]; +"162 conv2d_13_scale_0" -> "165 dequantize_per_channel_default_15" [label="(256,)", style=solid]; +"163 conv2d_13_zero_point_0" -> "165 dequantize_per_channel_default_15" [label="(256,)", style=solid]; +"164 _frozen_param15" -> "165 dequantize_per_channel_default_15" [label="(256, 256, 3, 3)", style=solid]; +"165 dequantize_per_channel_default_15" -> "166 conv2d_13" [label="(256, 256, 3, 3)", style=solid]; +"166 conv2d_13" -> "167 relu_13" [label="(1, 256, 12, 12)", style=solid]; +"167 relu_13" -> "168 quantize_per_tensor_default_18" [label="(1, 256, 12, 12)", style=solid]; +"168 quantize_per_tensor_default_18" -> "169 dequantize_per_tensor_default_22" [label="(1, 256, 12, 12)", style=solid]; +"169 dequantize_per_tensor_default_22" -> "174 conv_transpose2d_2" [label="(1, 256, 12, 12)", style=solid]; +"170 conv_transpose2d_2_scale_0" -> "173 dequantize_per_channel_default_16" [label="(256,)", style=solid]; +"171 conv_transpose2d_2_zero_point_0" -> "173 dequantize_per_channel_default_16" [label="(256,)", style=solid]; +"172 _frozen_param16" -> "173 dequantize_per_channel_default_16" [label="(256, 128, 2, 2)", style=solid]; +"173 dequantize_per_channel_default_16" -> "174 conv_transpose2d_2" [label="(256, 128, 2, 2)", style=solid]; +"174 conv_transpose2d_2" -> "175 quantize_per_tensor_default_2" [label="(1, 128, 24, 24)", style=solid]; +"175 quantize_per_tensor_default_2" -> "176 dequantize_per_tensor_default_3" [label="(1, 128, 24, 24)", style=solid]; +"176 dequantize_per_tensor_default_3" -> "181 cat_2" [label="(1, 128, 24, 24)", style=solid]; +"177 slice_9" -> "178 slice_10" [label="(1, 128, 106, 106)", style=solid]; +"178 slice_10" -> "179 slice_11" [label="(1, 128, 106, 106)", style=solid]; +"179 slice_11" -> "180 slice_12" [label="(1, 128, 24, 106)", style=solid]; +"180 slice_12" -> "181 cat_2" [label="(1, 128, 24, 24)", style=solid]; +"181 cat_2" -> "186 conv2d_14" [label="(1, 256, 24, 24)", style=solid]; +"182 conv2d_14_scale_0" -> "185 dequantize_per_channel_default_17" [label="(128,)", style=solid]; +"183 conv2d_14_zero_point_0" -> "185 dequantize_per_channel_default_17" [label="(128,)", style=solid]; +"184 _frozen_param17" -> "185 dequantize_per_channel_default_17" [label="(128, 256, 3, 3)", style=solid]; +"185 dequantize_per_channel_default_17" -> "186 conv2d_14" [label="(128, 256, 3, 3)", style=solid]; +"186 conv2d_14" -> "187 relu_14" [label="(1, 128, 22, 22)", style=solid]; +"187 relu_14" -> "188 quantize_per_tensor_default_19" [label="(1, 128, 22, 22)", style=solid]; +"188 quantize_per_tensor_default_19" -> "189 dequantize_per_tensor_default_23" [label="(1, 128, 22, 22)", style=solid]; +"189 dequantize_per_tensor_default_23" -> "194 conv2d_15" [label="(1, 128, 22, 22)", style=solid]; +"190 conv2d_15_scale_0" -> "193 dequantize_per_channel_default_18" [label="(128,)", style=solid]; +"191 conv2d_15_zero_point_0" -> "193 dequantize_per_channel_default_18" [label="(128,)", style=solid]; +"192 _frozen_param18" -> "193 dequantize_per_channel_default_18" [label="(128, 128, 3, 3)", style=solid]; +"193 dequantize_per_channel_default_18" -> "194 conv2d_15" [label="(128, 128, 3, 3)", style=solid]; +"194 conv2d_15" -> "195 relu_15" [label="(1, 128, 20, 20)", style=solid]; +"195 relu_15" -> "196 quantize_per_tensor_default_20" [label="(1, 128, 20, 20)", style=solid]; +"196 quantize_per_tensor_default_20" -> "197 dequantize_per_tensor_default_24" [label="(1, 128, 20, 20)", style=solid]; +"197 dequantize_per_tensor_default_24" -> "202 conv_transpose2d_3" [label="(1, 128, 20, 20)", style=solid]; +"198 conv_transpose2d_3_scale_0" -> "201 dequantize_per_channel_default_19" [label="(128,)", style=solid]; +"199 conv_transpose2d_3_zero_point_0" -> "201 dequantize_per_channel_default_19" [label="(128,)", style=solid]; +"200 _frozen_param19" -> "201 dequantize_per_channel_default_19" [label="(128, 64, 2, 2)", style=solid]; +"201 dequantize_per_channel_default_19" -> "202 conv_transpose2d_3" [label="(128, 64, 2, 2)", style=solid]; +"202 conv_transpose2d_3" -> "203 quantize_per_tensor_default_1" [label="(1, 64, 40, 40)", style=solid]; +"203 quantize_per_tensor_default_1" -> "204 dequantize_per_tensor_default_2" [label="(1, 64, 40, 40)", style=solid]; +"204 dequantize_per_tensor_default_2" -> "209 cat_3" [label="(1, 64, 40, 40)", style=solid]; +"205 slice_13" -> "206 slice_14" [label="(1, 64, 220, 220)", style=solid]; +"206 slice_14" -> "207 slice_15" [label="(1, 64, 220, 220)", style=solid]; +"207 slice_15" -> "208 slice_16" [label="(1, 64, 40, 220)", style=solid]; +"208 slice_16" -> "209 cat_3" [label="(1, 64, 40, 40)", style=solid]; +"209 cat_3" -> "214 conv2d_16" [label="(1, 128, 40, 40)", style=solid]; +"210 conv2d_16_scale_0" -> "213 dequantize_per_channel_default_20" [label="(64,)", style=solid]; +"211 conv2d_16_zero_point_0" -> "213 dequantize_per_channel_default_20" [label="(64,)", style=solid]; +"212 _frozen_param20" -> "213 dequantize_per_channel_default_20" [label="(64, 128, 3, 3)", style=solid]; +"213 dequantize_per_channel_default_20" -> "214 conv2d_16" [label="(64, 128, 3, 3)", style=solid]; +"214 conv2d_16" -> "215 relu_16" [label="(1, 64, 38, 38)", style=solid]; +"215 relu_16" -> "216 quantize_per_tensor_default_21" [label="(1, 64, 38, 38)", style=solid]; +"216 quantize_per_tensor_default_21" -> "217 dequantize_per_tensor_default_25" [label="(1, 64, 38, 38)", style=solid]; +"217 dequantize_per_tensor_default_25" -> "222 conv2d_17" [label="(1, 64, 38, 38)", style=solid]; +"218 conv2d_17_scale_0" -> "221 dequantize_per_channel_default_21" [label="(64,)", style=solid]; +"219 conv2d_17_zero_point_0" -> "221 dequantize_per_channel_default_21" [label="(64,)", style=solid]; +"220 _frozen_param21" -> "221 dequantize_per_channel_default_21" [label="(64, 64, 3, 3)", style=solid]; +"221 dequantize_per_channel_default_21" -> "222 conv2d_17" [label="(64, 64, 3, 3)", style=solid]; +"222 conv2d_17" -> "223 relu_17" [label="(1, 64, 36, 36)", style=solid]; +"223 relu_17" -> "224 quantize_per_tensor_default_22" [label="(1, 64, 36, 36)", style=solid]; +"224 quantize_per_tensor_default_22" -> "225 dequantize_per_tensor_default_26" [label="(1, 64, 36, 36)", style=solid]; +"225 dequantize_per_tensor_default_26" -> "230 conv2d_18" [label="(1, 64, 36, 36)", style=solid]; +"226 conv2d_18_scale_0" -> "229 dequantize_per_channel_default_22" [label="(12,)", style=solid]; +"227 conv2d_18_zero_point_0" -> "229 dequantize_per_channel_default_22" [label="(12,)", style=solid]; +"228 _frozen_param22" -> "229 dequantize_per_channel_default_22" [label="(12, 64, 1, 1)", style=solid]; +"229 dequantize_per_channel_default_22" -> "230 conv2d_18" [label="(12, 64, 1, 1)", style=solid]; +"230 conv2d_18" -> "231 output" [label="(1, 12, 36, 36)", style=solid]; +} diff --git a/tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/vit_b_16.dot b/tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/vit_b_16.dot new file mode 100644 index 00000000000..6209b772176 --- /dev/null +++ b/tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/vit_b_16.dot @@ -0,0 +1,1913 @@ +strict digraph { +"0 class_token" [id=0, type=get_attr]; +"1 encoder_pos_embedding" [id=1, type=get_attr]; +"2 conv_proj_bias" [id=2, type=get_attr]; +"3 encoder_layers_encoder_layer_0_ln_1_weight" [id=3, type=get_attr]; +"4 encoder_layers_encoder_layer_0_ln_1_bias" [id=4, type=get_attr]; +"5 encoder_layers_encoder_layer_0_self_attention_in_proj_bias" [id=5, type=get_attr]; +"6 encoder_layers_encoder_layer_0_self_attention_out_proj_bias" [id=6, type=get_attr]; +"7 encoder_layers_encoder_layer_0_ln_2_weight" [id=7, type=get_attr]; +"8 encoder_layers_encoder_layer_0_ln_2_bias" [id=8, type=get_attr]; +"9 encoder_layers_encoder_layer_0_mlp_0_bias" [id=9, type=get_attr]; +"10 encoder_layers_encoder_layer_0_mlp_3_bias" [id=10, type=get_attr]; +"11 encoder_layers_encoder_layer_1_ln_1_weight" [id=11, type=get_attr]; +"12 encoder_layers_encoder_layer_1_ln_1_bias" [id=12, type=get_attr]; +"13 encoder_layers_encoder_layer_1_self_attention_in_proj_bias" [id=13, type=get_attr]; +"14 encoder_layers_encoder_layer_1_self_attention_out_proj_bias" [id=14, type=get_attr]; +"15 encoder_layers_encoder_layer_1_ln_2_weight" [id=15, type=get_attr]; +"16 encoder_layers_encoder_layer_1_ln_2_bias" [id=16, type=get_attr]; +"17 encoder_layers_encoder_layer_1_mlp_0_bias" [id=17, type=get_attr]; +"18 encoder_layers_encoder_layer_1_mlp_3_bias" [id=18, type=get_attr]; +"19 encoder_layers_encoder_layer_2_ln_1_weight" [id=19, type=get_attr]; +"20 encoder_layers_encoder_layer_2_ln_1_bias" [id=20, type=get_attr]; +"21 encoder_layers_encoder_layer_2_self_attention_in_proj_bias" [id=21, type=get_attr]; +"22 encoder_layers_encoder_layer_2_self_attention_out_proj_bias" [id=22, type=get_attr]; +"23 encoder_layers_encoder_layer_2_ln_2_weight" [id=23, type=get_attr]; +"24 encoder_layers_encoder_layer_2_ln_2_bias" [id=24, type=get_attr]; +"25 encoder_layers_encoder_layer_2_mlp_0_bias" [id=25, type=get_attr]; +"26 encoder_layers_encoder_layer_2_mlp_3_bias" [id=26, type=get_attr]; +"27 encoder_layers_encoder_layer_3_ln_1_weight" [id=27, type=get_attr]; +"28 encoder_layers_encoder_layer_3_ln_1_bias" [id=28, type=get_attr]; +"29 encoder_layers_encoder_layer_3_self_attention_in_proj_bias" [id=29, type=get_attr]; +"30 encoder_layers_encoder_layer_3_self_attention_out_proj_bias" [id=30, type=get_attr]; +"31 encoder_layers_encoder_layer_3_ln_2_weight" [id=31, type=get_attr]; +"32 encoder_layers_encoder_layer_3_ln_2_bias" [id=32, type=get_attr]; +"33 encoder_layers_encoder_layer_3_mlp_0_bias" [id=33, type=get_attr]; +"34 encoder_layers_encoder_layer_3_mlp_3_bias" [id=34, type=get_attr]; +"35 encoder_layers_encoder_layer_4_ln_1_weight" [id=35, type=get_attr]; +"36 encoder_layers_encoder_layer_4_ln_1_bias" [id=36, type=get_attr]; +"37 encoder_layers_encoder_layer_4_self_attention_in_proj_bias" [id=37, type=get_attr]; +"38 encoder_layers_encoder_layer_4_self_attention_out_proj_bias" [id=38, type=get_attr]; +"39 encoder_layers_encoder_layer_4_ln_2_weight" [id=39, type=get_attr]; +"40 encoder_layers_encoder_layer_4_ln_2_bias" [id=40, type=get_attr]; +"41 encoder_layers_encoder_layer_4_mlp_0_bias" [id=41, type=get_attr]; +"42 encoder_layers_encoder_layer_4_mlp_3_bias" [id=42, type=get_attr]; +"43 encoder_layers_encoder_layer_5_ln_1_weight" [id=43, type=get_attr]; +"44 encoder_layers_encoder_layer_5_ln_1_bias" [id=44, type=get_attr]; +"45 encoder_layers_encoder_layer_5_self_attention_in_proj_bias" [id=45, type=get_attr]; +"46 encoder_layers_encoder_layer_5_self_attention_out_proj_bias" [id=46, type=get_attr]; +"47 encoder_layers_encoder_layer_5_ln_2_weight" [id=47, type=get_attr]; +"48 encoder_layers_encoder_layer_5_ln_2_bias" [id=48, type=get_attr]; +"49 encoder_layers_encoder_layer_5_mlp_0_bias" [id=49, type=get_attr]; +"50 encoder_layers_encoder_layer_5_mlp_3_bias" [id=50, type=get_attr]; +"51 encoder_layers_encoder_layer_6_ln_1_weight" [id=51, type=get_attr]; +"52 encoder_layers_encoder_layer_6_ln_1_bias" [id=52, type=get_attr]; +"53 encoder_layers_encoder_layer_6_self_attention_in_proj_bias" [id=53, type=get_attr]; +"54 encoder_layers_encoder_layer_6_self_attention_out_proj_bias" [id=54, type=get_attr]; +"55 encoder_layers_encoder_layer_6_ln_2_weight" [id=55, type=get_attr]; +"56 encoder_layers_encoder_layer_6_ln_2_bias" [id=56, type=get_attr]; +"57 encoder_layers_encoder_layer_6_mlp_0_bias" [id=57, type=get_attr]; +"58 encoder_layers_encoder_layer_6_mlp_3_bias" [id=58, type=get_attr]; +"59 encoder_layers_encoder_layer_7_ln_1_weight" [id=59, type=get_attr]; +"60 encoder_layers_encoder_layer_7_ln_1_bias" [id=60, type=get_attr]; +"61 encoder_layers_encoder_layer_7_self_attention_in_proj_bias" [id=61, type=get_attr]; +"62 encoder_layers_encoder_layer_7_self_attention_out_proj_bias" [id=62, type=get_attr]; +"63 encoder_layers_encoder_layer_7_ln_2_weight" [id=63, type=get_attr]; +"64 encoder_layers_encoder_layer_7_ln_2_bias" [id=64, type=get_attr]; +"65 encoder_layers_encoder_layer_7_mlp_0_bias" [id=65, type=get_attr]; +"66 encoder_layers_encoder_layer_7_mlp_3_bias" [id=66, type=get_attr]; +"67 encoder_layers_encoder_layer_8_ln_1_weight" [id=67, type=get_attr]; +"68 encoder_layers_encoder_layer_8_ln_1_bias" [id=68, type=get_attr]; +"69 encoder_layers_encoder_layer_8_self_attention_in_proj_bias" [id=69, type=get_attr]; +"70 encoder_layers_encoder_layer_8_self_attention_out_proj_bias" [id=70, type=get_attr]; +"71 encoder_layers_encoder_layer_8_ln_2_weight" [id=71, type=get_attr]; +"72 encoder_layers_encoder_layer_8_ln_2_bias" [id=72, type=get_attr]; +"73 encoder_layers_encoder_layer_8_mlp_0_bias" [id=73, type=get_attr]; +"74 encoder_layers_encoder_layer_8_mlp_3_bias" [id=74, type=get_attr]; +"75 encoder_layers_encoder_layer_9_ln_1_weight" [id=75, type=get_attr]; +"76 encoder_layers_encoder_layer_9_ln_1_bias" [id=76, type=get_attr]; +"77 encoder_layers_encoder_layer_9_self_attention_in_proj_bias" [id=77, type=get_attr]; +"78 encoder_layers_encoder_layer_9_self_attention_out_proj_bias" [id=78, type=get_attr]; +"79 encoder_layers_encoder_layer_9_ln_2_weight" [id=79, type=get_attr]; +"80 encoder_layers_encoder_layer_9_ln_2_bias" [id=80, type=get_attr]; +"81 encoder_layers_encoder_layer_9_mlp_0_bias" [id=81, type=get_attr]; +"82 encoder_layers_encoder_layer_9_mlp_3_bias" [id=82, type=get_attr]; +"83 encoder_layers_encoder_layer_10_ln_1_weight" [id=83, type=get_attr]; +"84 encoder_layers_encoder_layer_10_ln_1_bias" [id=84, type=get_attr]; +"85 encoder_layers_encoder_layer_10_self_attention_in_proj_bias" [id=85, type=get_attr]; +"86 encoder_layers_encoder_layer_10_self_attention_out_proj_bias" [id=86, type=get_attr]; +"87 encoder_layers_encoder_layer_10_ln_2_weight" [id=87, type=get_attr]; +"88 encoder_layers_encoder_layer_10_ln_2_bias" [id=88, type=get_attr]; +"89 encoder_layers_encoder_layer_10_mlp_0_bias" [id=89, type=get_attr]; +"90 encoder_layers_encoder_layer_10_mlp_3_bias" [id=90, type=get_attr]; +"91 encoder_layers_encoder_layer_11_ln_1_weight" [id=91, type=get_attr]; +"92 encoder_layers_encoder_layer_11_ln_1_bias" [id=92, type=get_attr]; +"93 encoder_layers_encoder_layer_11_self_attention_in_proj_bias" [id=93, type=get_attr]; +"94 encoder_layers_encoder_layer_11_self_attention_out_proj_bias" [id=94, type=get_attr]; +"95 encoder_layers_encoder_layer_11_ln_2_weight" [id=95, type=get_attr]; +"96 encoder_layers_encoder_layer_11_ln_2_bias" [id=96, type=get_attr]; +"97 encoder_layers_encoder_layer_11_mlp_0_bias" [id=97, type=get_attr]; +"98 encoder_layers_encoder_layer_11_mlp_3_bias" [id=98, type=get_attr]; +"99 encoder_ln_weight" [id=99, type=get_attr]; +"100 encoder_ln_bias" [id=100, type=get_attr]; +"101 heads_head_bias" [id=101, type=get_attr]; +"102 x" [id=102, type=input]; +"103 quantize_per_tensor_default" [id=103, type=quantize_per_tensor]; +"104 dequantize_per_tensor_default" [id=104, type=dequantize_per_tensor]; +"105 conv2d_scale_0" [id=105, type=get_attr]; +"106 conv2d_zero_point_0" [id=106, type=get_attr]; +"107 _frozen_param0" [id=107, type=get_attr]; +"108 dequantize_per_channel_default" [id=108, type=dequantize_per_channel]; +"109 conv2d" [id=109, type=conv2d]; +"110 reshape" [id=110, type=reshape]; +"111 permute" [id=111, type=permute]; +"112 expand" [id=112, type=expand]; +"113 cat" [id=113, type=cat]; +"114 add" [id=114, type=add]; +"115 dropout" [id=115, type=dropout]; +"116 layer_norm" [id=116, type=layer_norm]; +"117 transpose" [id=117, type=transpose]; +"118 transpose_0_0_nncf_smooth_quant_0" [id=118, type=call_module]; +"119 quantize_per_tensor_default_1" [id=119, type=quantize_per_tensor]; +"120 dequantize_per_tensor_default_1" [id=120, type=dequantize_per_tensor]; +"121 linear_scale_0" [id=121, type=get_attr]; +"122 linear_zero_point_0" [id=122, type=get_attr]; +"123 _frozen_param1" [id=123, type=get_attr]; +"124 dequantize_per_channel_default_1" [id=124, type=dequantize_per_channel]; +"125 linear" [id=125, type=linear]; +"126 unflatten" [id=126, type=unflatten]; +"127 unsqueeze" [id=127, type=unsqueeze]; +"128 transpose_1" [id=128, type=transpose]; +"129 squeeze" [id=129, type=squeeze]; +"130 contiguous" [id=130, type=contiguous]; +"131 quantize_per_tensor_default_2" [id=131, type=quantize_per_tensor]; +"132 dequantize_per_tensor_default_2" [id=132, type=dequantize_per_tensor]; +"133 select" [id=133, type=select]; +"134 quantize_per_tensor_default_3" [id=134, type=quantize_per_tensor]; +"135 dequantize_per_tensor_default_3" [id=135, type=dequantize_per_tensor]; +"136 select_1" [id=136, type=select]; +"137 select_2" [id=137, type=select]; +"138 view" [id=138, type=view]; +"139 transpose_2" [id=139, type=transpose]; +"140 view_1" [id=140, type=view]; +"141 transpose_3" [id=141, type=transpose]; +"142 view_2" [id=142, type=view]; +"143 transpose_4" [id=143, type=transpose]; +"144 view_3" [id=144, type=view]; +"145 view_4" [id=145, type=view]; +"146 view_5" [id=146, type=view]; +"147 scaled_dot_product_attention" [id=147, type=scaled_dot_product_attention]; +"148 permute_1" [id=148, type=permute]; +"149 view_6" [id=149, type=view]; +"150 view_6_0_0_nncf_smooth_quant_0" [id=150, type=call_module]; +"151 quantize_per_tensor_default_4" [id=151, type=quantize_per_tensor]; +"152 dequantize_per_tensor_default_4" [id=152, type=dequantize_per_tensor]; +"153 linear_1_scale_0" [id=153, type=get_attr]; +"154 linear_1_zero_point_0" [id=154, type=get_attr]; +"155 _frozen_param2" [id=155, type=get_attr]; +"156 dequantize_per_channel_default_2" [id=156, type=dequantize_per_channel]; +"157 linear_1" [id=157, type=linear]; +"158 view_7" [id=158, type=view]; +"159 transpose_5" [id=159, type=transpose]; +"160 dropout_1" [id=160, type=dropout]; +"161 add_1" [id=161, type=add]; +"162 layer_norm_1" [id=162, type=layer_norm]; +"163 layer_norm_1_0_0_nncf_smooth_quant_0" [id=163, type=call_module]; +"164 quantize_per_tensor_default_5" [id=164, type=quantize_per_tensor]; +"165 dequantize_per_tensor_default_5" [id=165, type=dequantize_per_tensor]; +"166 linear_2_scale_0" [id=166, type=get_attr]; +"167 linear_2_zero_point_0" [id=167, type=get_attr]; +"168 _frozen_param3" [id=168, type=get_attr]; +"169 dequantize_per_channel_default_3" [id=169, type=dequantize_per_channel]; +"170 linear_2" [id=170, type=linear]; +"171 gelu" [id=171, type=gelu]; +"172 dropout_2" [id=172, type=dropout]; +"173 dropout_2_0_0_nncf_smooth_quant_0" [id=173, type=call_module]; +"174 quantize_per_tensor_default_6" [id=174, type=quantize_per_tensor]; +"175 dequantize_per_tensor_default_6" [id=175, type=dequantize_per_tensor]; +"176 linear_3_scale_0" [id=176, type=get_attr]; +"177 linear_3_zero_point_0" [id=177, type=get_attr]; +"178 _frozen_param4" [id=178, type=get_attr]; +"179 dequantize_per_channel_default_4" [id=179, type=dequantize_per_channel]; +"180 linear_3" [id=180, type=linear]; +"181 dropout_3" [id=181, type=dropout]; +"182 add_2" [id=182, type=add]; +"183 layer_norm_2" [id=183, type=layer_norm]; +"184 transpose_6" [id=184, type=transpose]; +"185 transpose_6_0_0_nncf_smooth_quant_0" [id=185, type=call_module]; +"186 quantize_per_tensor_default_7" [id=186, type=quantize_per_tensor]; +"187 dequantize_per_tensor_default_7" [id=187, type=dequantize_per_tensor]; +"188 linear_4_scale_0" [id=188, type=get_attr]; +"189 linear_4_zero_point_0" [id=189, type=get_attr]; +"190 _frozen_param5" [id=190, type=get_attr]; +"191 dequantize_per_channel_default_5" [id=191, type=dequantize_per_channel]; +"192 linear_4" [id=192, type=linear]; +"193 unflatten_1" [id=193, type=unflatten]; +"194 unsqueeze_1" [id=194, type=unsqueeze]; +"195 transpose_7" [id=195, type=transpose]; +"196 squeeze_1" [id=196, type=squeeze]; +"197 contiguous_1" [id=197, type=contiguous]; +"198 quantize_per_tensor_default_8" [id=198, type=quantize_per_tensor]; +"199 dequantize_per_tensor_default_8" [id=199, type=dequantize_per_tensor]; +"200 select_3" [id=200, type=select]; +"201 quantize_per_tensor_default_9" [id=201, type=quantize_per_tensor]; +"202 dequantize_per_tensor_default_9" [id=202, type=dequantize_per_tensor]; +"203 select_4" [id=203, type=select]; +"204 select_5" [id=204, type=select]; +"205 view_8" [id=205, type=view]; +"206 transpose_8" [id=206, type=transpose]; +"207 view_9" [id=207, type=view]; +"208 transpose_9" [id=208, type=transpose]; +"209 view_10" [id=209, type=view]; +"210 transpose_10" [id=210, type=transpose]; +"211 view_11" [id=211, type=view]; +"212 view_12" [id=212, type=view]; +"213 view_13" [id=213, type=view]; +"214 scaled_dot_product_attention_1" [id=214, type=scaled_dot_product_attention]; +"215 permute_2" [id=215, type=permute]; +"216 view_14" [id=216, type=view]; +"217 view_14_0_0_nncf_smooth_quant_0" [id=217, type=call_module]; +"218 quantize_per_tensor_default_10" [id=218, type=quantize_per_tensor]; +"219 dequantize_per_tensor_default_10" [id=219, type=dequantize_per_tensor]; +"220 linear_5_scale_0" [id=220, type=get_attr]; +"221 linear_5_zero_point_0" [id=221, type=get_attr]; +"222 _frozen_param6" [id=222, type=get_attr]; +"223 dequantize_per_channel_default_6" [id=223, type=dequantize_per_channel]; +"224 linear_5" [id=224, type=linear]; +"225 view_15" [id=225, type=view]; +"226 transpose_11" [id=226, type=transpose]; +"227 dropout_4" [id=227, type=dropout]; +"228 add_3" [id=228, type=add]; +"229 layer_norm_3" [id=229, type=layer_norm]; +"230 layer_norm_3_0_0_nncf_smooth_quant_0" [id=230, type=call_module]; +"231 quantize_per_tensor_default_11" [id=231, type=quantize_per_tensor]; +"232 dequantize_per_tensor_default_11" [id=232, type=dequantize_per_tensor]; +"233 linear_6_scale_0" [id=233, type=get_attr]; +"234 linear_6_zero_point_0" [id=234, type=get_attr]; +"235 _frozen_param7" [id=235, type=get_attr]; +"236 dequantize_per_channel_default_7" [id=236, type=dequantize_per_channel]; +"237 linear_6" [id=237, type=linear]; +"238 gelu_1" [id=238, type=gelu]; +"239 dropout_5" [id=239, type=dropout]; +"240 dropout_5_0_0_nncf_smooth_quant_0" [id=240, type=call_module]; +"241 quantize_per_tensor_default_12" [id=241, type=quantize_per_tensor]; +"242 dequantize_per_tensor_default_12" [id=242, type=dequantize_per_tensor]; +"243 linear_7_scale_0" [id=243, type=get_attr]; +"244 linear_7_zero_point_0" [id=244, type=get_attr]; +"245 _frozen_param8" [id=245, type=get_attr]; +"246 dequantize_per_channel_default_8" [id=246, type=dequantize_per_channel]; +"247 linear_7" [id=247, type=linear]; +"248 dropout_6" [id=248, type=dropout]; +"249 add_4" [id=249, type=add]; +"250 layer_norm_4" [id=250, type=layer_norm]; +"251 transpose_12" [id=251, type=transpose]; +"252 transpose_12_0_0_nncf_smooth_quant_0" [id=252, type=call_module]; +"253 quantize_per_tensor_default_13" [id=253, type=quantize_per_tensor]; +"254 dequantize_per_tensor_default_13" [id=254, type=dequantize_per_tensor]; +"255 linear_8_scale_0" [id=255, type=get_attr]; +"256 linear_8_zero_point_0" [id=256, type=get_attr]; +"257 _frozen_param9" [id=257, type=get_attr]; +"258 dequantize_per_channel_default_9" [id=258, type=dequantize_per_channel]; +"259 linear_8" [id=259, type=linear]; +"260 unflatten_2" [id=260, type=unflatten]; +"261 unsqueeze_2" [id=261, type=unsqueeze]; +"262 transpose_13" [id=262, type=transpose]; +"263 squeeze_2" [id=263, type=squeeze]; +"264 contiguous_2" [id=264, type=contiguous]; +"265 quantize_per_tensor_default_14" [id=265, type=quantize_per_tensor]; +"266 dequantize_per_tensor_default_14" [id=266, type=dequantize_per_tensor]; +"267 select_6" [id=267, type=select]; +"268 quantize_per_tensor_default_15" [id=268, type=quantize_per_tensor]; +"269 dequantize_per_tensor_default_15" [id=269, type=dequantize_per_tensor]; +"270 select_7" [id=270, type=select]; +"271 select_8" [id=271, type=select]; +"272 view_16" [id=272, type=view]; +"273 transpose_14" [id=273, type=transpose]; +"274 view_17" [id=274, type=view]; +"275 transpose_15" [id=275, type=transpose]; +"276 view_18" [id=276, type=view]; +"277 transpose_16" [id=277, type=transpose]; +"278 view_19" [id=278, type=view]; +"279 view_20" [id=279, type=view]; +"280 view_21" [id=280, type=view]; +"281 scaled_dot_product_attention_2" [id=281, type=scaled_dot_product_attention]; +"282 permute_3" [id=282, type=permute]; +"283 view_22" [id=283, type=view]; +"284 view_22_0_0_nncf_smooth_quant_0" [id=284, type=call_module]; +"285 quantize_per_tensor_default_16" [id=285, type=quantize_per_tensor]; +"286 dequantize_per_tensor_default_16" [id=286, type=dequantize_per_tensor]; +"287 linear_9_scale_0" [id=287, type=get_attr]; +"288 linear_9_zero_point_0" [id=288, type=get_attr]; +"289 _frozen_param10" [id=289, type=get_attr]; +"290 dequantize_per_channel_default_10" [id=290, type=dequantize_per_channel]; +"291 linear_9" [id=291, type=linear]; +"292 view_23" [id=292, type=view]; +"293 transpose_17" [id=293, type=transpose]; +"294 dropout_7" [id=294, type=dropout]; +"295 add_5" [id=295, type=add]; +"296 layer_norm_5" [id=296, type=layer_norm]; +"297 layer_norm_5_0_0_nncf_smooth_quant_0" [id=297, type=call_module]; +"298 quantize_per_tensor_default_17" [id=298, type=quantize_per_tensor]; +"299 dequantize_per_tensor_default_17" [id=299, type=dequantize_per_tensor]; +"300 linear_10_scale_0" [id=300, type=get_attr]; +"301 linear_10_zero_point_0" [id=301, type=get_attr]; +"302 _frozen_param11" [id=302, type=get_attr]; +"303 dequantize_per_channel_default_11" [id=303, type=dequantize_per_channel]; +"304 linear_10" [id=304, type=linear]; +"305 gelu_2" [id=305, type=gelu]; +"306 dropout_8" [id=306, type=dropout]; +"307 dropout_8_0_0_nncf_smooth_quant_0" [id=307, type=call_module]; +"308 quantize_per_tensor_default_18" [id=308, type=quantize_per_tensor]; +"309 dequantize_per_tensor_default_18" [id=309, type=dequantize_per_tensor]; +"310 linear_11_scale_0" [id=310, type=get_attr]; +"311 linear_11_zero_point_0" [id=311, type=get_attr]; +"312 _frozen_param12" [id=312, type=get_attr]; +"313 dequantize_per_channel_default_12" [id=313, type=dequantize_per_channel]; +"314 linear_11" [id=314, type=linear]; +"315 dropout_9" [id=315, type=dropout]; +"316 add_6" [id=316, type=add]; +"317 layer_norm_6" [id=317, type=layer_norm]; +"318 transpose_18" [id=318, type=transpose]; +"319 transpose_18_0_0_nncf_smooth_quant_0" [id=319, type=call_module]; +"320 quantize_per_tensor_default_19" [id=320, type=quantize_per_tensor]; +"321 dequantize_per_tensor_default_19" [id=321, type=dequantize_per_tensor]; +"322 linear_12_scale_0" [id=322, type=get_attr]; +"323 linear_12_zero_point_0" [id=323, type=get_attr]; +"324 _frozen_param13" [id=324, type=get_attr]; +"325 dequantize_per_channel_default_13" [id=325, type=dequantize_per_channel]; +"326 linear_12" [id=326, type=linear]; +"327 unflatten_3" [id=327, type=unflatten]; +"328 unsqueeze_3" [id=328, type=unsqueeze]; +"329 transpose_19" [id=329, type=transpose]; +"330 squeeze_3" [id=330, type=squeeze]; +"331 contiguous_3" [id=331, type=contiguous]; +"332 quantize_per_tensor_default_20" [id=332, type=quantize_per_tensor]; +"333 dequantize_per_tensor_default_20" [id=333, type=dequantize_per_tensor]; +"334 select_9" [id=334, type=select]; +"335 quantize_per_tensor_default_21" [id=335, type=quantize_per_tensor]; +"336 dequantize_per_tensor_default_21" [id=336, type=dequantize_per_tensor]; +"337 select_10" [id=337, type=select]; +"338 select_11" [id=338, type=select]; +"339 view_24" [id=339, type=view]; +"340 transpose_20" [id=340, type=transpose]; +"341 view_25" [id=341, type=view]; +"342 transpose_21" [id=342, type=transpose]; +"343 view_26" [id=343, type=view]; +"344 transpose_22" [id=344, type=transpose]; +"345 view_27" [id=345, type=view]; +"346 view_28" [id=346, type=view]; +"347 view_29" [id=347, type=view]; +"348 scaled_dot_product_attention_3" [id=348, type=scaled_dot_product_attention]; +"349 permute_4" [id=349, type=permute]; +"350 view_30" [id=350, type=view]; +"351 view_30_0_0_nncf_smooth_quant_0" [id=351, type=call_module]; +"352 quantize_per_tensor_default_22" [id=352, type=quantize_per_tensor]; +"353 dequantize_per_tensor_default_22" [id=353, type=dequantize_per_tensor]; +"354 linear_13_scale_0" [id=354, type=get_attr]; +"355 linear_13_zero_point_0" [id=355, type=get_attr]; +"356 _frozen_param14" [id=356, type=get_attr]; +"357 dequantize_per_channel_default_14" [id=357, type=dequantize_per_channel]; +"358 linear_13" [id=358, type=linear]; +"359 view_31" [id=359, type=view]; +"360 transpose_23" [id=360, type=transpose]; +"361 dropout_10" [id=361, type=dropout]; +"362 add_7" [id=362, type=add]; +"363 layer_norm_7" [id=363, type=layer_norm]; +"364 layer_norm_7_0_0_nncf_smooth_quant_0" [id=364, type=call_module]; +"365 quantize_per_tensor_default_23" [id=365, type=quantize_per_tensor]; +"366 dequantize_per_tensor_default_23" [id=366, type=dequantize_per_tensor]; +"367 linear_14_scale_0" [id=367, type=get_attr]; +"368 linear_14_zero_point_0" [id=368, type=get_attr]; +"369 _frozen_param15" [id=369, type=get_attr]; +"370 dequantize_per_channel_default_15" [id=370, type=dequantize_per_channel]; +"371 linear_14" [id=371, type=linear]; +"372 gelu_3" [id=372, type=gelu]; +"373 dropout_11" [id=373, type=dropout]; +"374 dropout_11_0_0_nncf_smooth_quant_0" [id=374, type=call_module]; +"375 quantize_per_tensor_default_24" [id=375, type=quantize_per_tensor]; +"376 dequantize_per_tensor_default_24" [id=376, type=dequantize_per_tensor]; +"377 linear_15_scale_0" [id=377, type=get_attr]; +"378 linear_15_zero_point_0" [id=378, type=get_attr]; +"379 _frozen_param16" [id=379, type=get_attr]; +"380 dequantize_per_channel_default_16" [id=380, type=dequantize_per_channel]; +"381 linear_15" [id=381, type=linear]; +"382 dropout_12" [id=382, type=dropout]; +"383 add_8" [id=383, type=add]; +"384 layer_norm_8" [id=384, type=layer_norm]; +"385 transpose_24" [id=385, type=transpose]; +"386 transpose_24_0_0_nncf_smooth_quant_0" [id=386, type=call_module]; +"387 quantize_per_tensor_default_25" [id=387, type=quantize_per_tensor]; +"388 dequantize_per_tensor_default_25" [id=388, type=dequantize_per_tensor]; +"389 linear_16_scale_0" [id=389, type=get_attr]; +"390 linear_16_zero_point_0" [id=390, type=get_attr]; +"391 _frozen_param17" [id=391, type=get_attr]; +"392 dequantize_per_channel_default_17" [id=392, type=dequantize_per_channel]; +"393 linear_16" [id=393, type=linear]; +"394 unflatten_4" [id=394, type=unflatten]; +"395 unsqueeze_4" [id=395, type=unsqueeze]; +"396 transpose_25" [id=396, type=transpose]; +"397 squeeze_4" [id=397, type=squeeze]; +"398 contiguous_4" [id=398, type=contiguous]; +"399 quantize_per_tensor_default_26" [id=399, type=quantize_per_tensor]; +"400 dequantize_per_tensor_default_26" [id=400, type=dequantize_per_tensor]; +"401 select_12" [id=401, type=select]; +"402 quantize_per_tensor_default_27" [id=402, type=quantize_per_tensor]; +"403 dequantize_per_tensor_default_27" [id=403, type=dequantize_per_tensor]; +"404 select_13" [id=404, type=select]; +"405 select_14" [id=405, type=select]; +"406 view_32" [id=406, type=view]; +"407 transpose_26" [id=407, type=transpose]; +"408 view_33" [id=408, type=view]; +"409 transpose_27" [id=409, type=transpose]; +"410 view_34" [id=410, type=view]; +"411 transpose_28" [id=411, type=transpose]; +"412 view_35" [id=412, type=view]; +"413 view_36" [id=413, type=view]; +"414 view_37" [id=414, type=view]; +"415 scaled_dot_product_attention_4" [id=415, type=scaled_dot_product_attention]; +"416 permute_5" [id=416, type=permute]; +"417 view_38" [id=417, type=view]; +"418 view_38_0_0_nncf_smooth_quant_0" [id=418, type=call_module]; +"419 quantize_per_tensor_default_28" [id=419, type=quantize_per_tensor]; +"420 dequantize_per_tensor_default_28" [id=420, type=dequantize_per_tensor]; +"421 linear_17_scale_0" [id=421, type=get_attr]; +"422 linear_17_zero_point_0" [id=422, type=get_attr]; +"423 _frozen_param18" [id=423, type=get_attr]; +"424 dequantize_per_channel_default_18" [id=424, type=dequantize_per_channel]; +"425 linear_17" [id=425, type=linear]; +"426 view_39" [id=426, type=view]; +"427 transpose_29" [id=427, type=transpose]; +"428 dropout_13" [id=428, type=dropout]; +"429 add_9" [id=429, type=add]; +"430 layer_norm_9" [id=430, type=layer_norm]; +"431 layer_norm_9_0_0_nncf_smooth_quant_0" [id=431, type=call_module]; +"432 quantize_per_tensor_default_29" [id=432, type=quantize_per_tensor]; +"433 dequantize_per_tensor_default_29" [id=433, type=dequantize_per_tensor]; +"434 linear_18_scale_0" [id=434, type=get_attr]; +"435 linear_18_zero_point_0" [id=435, type=get_attr]; +"436 _frozen_param19" [id=436, type=get_attr]; +"437 dequantize_per_channel_default_19" [id=437, type=dequantize_per_channel]; +"438 linear_18" [id=438, type=linear]; +"439 gelu_4" [id=439, type=gelu]; +"440 dropout_14" [id=440, type=dropout]; +"441 dropout_14_0_0_nncf_smooth_quant_0" [id=441, type=call_module]; +"442 quantize_per_tensor_default_30" [id=442, type=quantize_per_tensor]; +"443 dequantize_per_tensor_default_30" [id=443, type=dequantize_per_tensor]; +"444 linear_19_scale_0" [id=444, type=get_attr]; +"445 linear_19_zero_point_0" [id=445, type=get_attr]; +"446 _frozen_param20" [id=446, type=get_attr]; +"447 dequantize_per_channel_default_20" [id=447, type=dequantize_per_channel]; +"448 linear_19" [id=448, type=linear]; +"449 dropout_15" [id=449, type=dropout]; +"450 add_10" [id=450, type=add]; +"451 layer_norm_10" [id=451, type=layer_norm]; +"452 transpose_30" [id=452, type=transpose]; +"453 transpose_30_0_0_nncf_smooth_quant_0" [id=453, type=call_module]; +"454 quantize_per_tensor_default_31" [id=454, type=quantize_per_tensor]; +"455 dequantize_per_tensor_default_31" [id=455, type=dequantize_per_tensor]; +"456 linear_20_scale_0" [id=456, type=get_attr]; +"457 linear_20_zero_point_0" [id=457, type=get_attr]; +"458 _frozen_param21" [id=458, type=get_attr]; +"459 dequantize_per_channel_default_21" [id=459, type=dequantize_per_channel]; +"460 linear_20" [id=460, type=linear]; +"461 unflatten_5" [id=461, type=unflatten]; +"462 unsqueeze_5" [id=462, type=unsqueeze]; +"463 transpose_31" [id=463, type=transpose]; +"464 squeeze_5" [id=464, type=squeeze]; +"465 contiguous_5" [id=465, type=contiguous]; +"466 quantize_per_tensor_default_32" [id=466, type=quantize_per_tensor]; +"467 dequantize_per_tensor_default_32" [id=467, type=dequantize_per_tensor]; +"468 select_15" [id=468, type=select]; +"469 quantize_per_tensor_default_33" [id=469, type=quantize_per_tensor]; +"470 dequantize_per_tensor_default_33" [id=470, type=dequantize_per_tensor]; +"471 select_16" [id=471, type=select]; +"472 select_17" [id=472, type=select]; +"473 view_40" [id=473, type=view]; +"474 transpose_32" [id=474, type=transpose]; +"475 view_41" [id=475, type=view]; +"476 transpose_33" [id=476, type=transpose]; +"477 view_42" [id=477, type=view]; +"478 transpose_34" [id=478, type=transpose]; +"479 view_43" [id=479, type=view]; +"480 view_44" [id=480, type=view]; +"481 view_45" [id=481, type=view]; +"482 scaled_dot_product_attention_5" [id=482, type=scaled_dot_product_attention]; +"483 permute_6" [id=483, type=permute]; +"484 view_46" [id=484, type=view]; +"485 view_46_0_0_nncf_smooth_quant_0" [id=485, type=call_module]; +"486 quantize_per_tensor_default_34" [id=486, type=quantize_per_tensor]; +"487 dequantize_per_tensor_default_34" [id=487, type=dequantize_per_tensor]; +"488 linear_21_scale_0" [id=488, type=get_attr]; +"489 linear_21_zero_point_0" [id=489, type=get_attr]; +"490 _frozen_param22" [id=490, type=get_attr]; +"491 dequantize_per_channel_default_22" [id=491, type=dequantize_per_channel]; +"492 linear_21" [id=492, type=linear]; +"493 view_47" [id=493, type=view]; +"494 transpose_35" [id=494, type=transpose]; +"495 dropout_16" [id=495, type=dropout]; +"496 add_11" [id=496, type=add]; +"497 layer_norm_11" [id=497, type=layer_norm]; +"498 layer_norm_11_0_0_nncf_smooth_quant_0" [id=498, type=call_module]; +"499 quantize_per_tensor_default_35" [id=499, type=quantize_per_tensor]; +"500 dequantize_per_tensor_default_35" [id=500, type=dequantize_per_tensor]; +"501 linear_22_scale_0" [id=501, type=get_attr]; +"502 linear_22_zero_point_0" [id=502, type=get_attr]; +"503 _frozen_param23" [id=503, type=get_attr]; +"504 dequantize_per_channel_default_23" [id=504, type=dequantize_per_channel]; +"505 linear_22" [id=505, type=linear]; +"506 gelu_5" [id=506, type=gelu]; +"507 dropout_17" [id=507, type=dropout]; +"508 dropout_17_0_0_nncf_smooth_quant_0" [id=508, type=call_module]; +"509 quantize_per_tensor_default_36" [id=509, type=quantize_per_tensor]; +"510 dequantize_per_tensor_default_36" [id=510, type=dequantize_per_tensor]; +"511 linear_23_scale_0" [id=511, type=get_attr]; +"512 linear_23_zero_point_0" [id=512, type=get_attr]; +"513 _frozen_param24" [id=513, type=get_attr]; +"514 dequantize_per_channel_default_24" [id=514, type=dequantize_per_channel]; +"515 linear_23" [id=515, type=linear]; +"516 dropout_18" [id=516, type=dropout]; +"517 add_12" [id=517, type=add]; +"518 layer_norm_12" [id=518, type=layer_norm]; +"519 transpose_36" [id=519, type=transpose]; +"520 transpose_36_0_0_nncf_smooth_quant_0" [id=520, type=call_module]; +"521 quantize_per_tensor_default_37" [id=521, type=quantize_per_tensor]; +"522 dequantize_per_tensor_default_37" [id=522, type=dequantize_per_tensor]; +"523 linear_24_scale_0" [id=523, type=get_attr]; +"524 linear_24_zero_point_0" [id=524, type=get_attr]; +"525 _frozen_param25" [id=525, type=get_attr]; +"526 dequantize_per_channel_default_25" [id=526, type=dequantize_per_channel]; +"527 linear_24" [id=527, type=linear]; +"528 unflatten_6" [id=528, type=unflatten]; +"529 unsqueeze_6" [id=529, type=unsqueeze]; +"530 transpose_37" [id=530, type=transpose]; +"531 squeeze_6" [id=531, type=squeeze]; +"532 contiguous_6" [id=532, type=contiguous]; +"533 quantize_per_tensor_default_38" [id=533, type=quantize_per_tensor]; +"534 dequantize_per_tensor_default_38" [id=534, type=dequantize_per_tensor]; +"535 select_18" [id=535, type=select]; +"536 quantize_per_tensor_default_39" [id=536, type=quantize_per_tensor]; +"537 dequantize_per_tensor_default_39" [id=537, type=dequantize_per_tensor]; +"538 select_19" [id=538, type=select]; +"539 select_20" [id=539, type=select]; +"540 view_48" [id=540, type=view]; +"541 transpose_38" [id=541, type=transpose]; +"542 view_49" [id=542, type=view]; +"543 transpose_39" [id=543, type=transpose]; +"544 view_50" [id=544, type=view]; +"545 transpose_40" [id=545, type=transpose]; +"546 view_51" [id=546, type=view]; +"547 view_52" [id=547, type=view]; +"548 view_53" [id=548, type=view]; +"549 scaled_dot_product_attention_6" [id=549, type=scaled_dot_product_attention]; +"550 permute_7" [id=550, type=permute]; +"551 view_54" [id=551, type=view]; +"552 view_54_0_0_nncf_smooth_quant_0" [id=552, type=call_module]; +"553 quantize_per_tensor_default_40" [id=553, type=quantize_per_tensor]; +"554 dequantize_per_tensor_default_40" [id=554, type=dequantize_per_tensor]; +"555 linear_25_scale_0" [id=555, type=get_attr]; +"556 linear_25_zero_point_0" [id=556, type=get_attr]; +"557 _frozen_param26" [id=557, type=get_attr]; +"558 dequantize_per_channel_default_26" [id=558, type=dequantize_per_channel]; +"559 linear_25" [id=559, type=linear]; +"560 view_55" [id=560, type=view]; +"561 transpose_41" [id=561, type=transpose]; +"562 dropout_19" [id=562, type=dropout]; +"563 add_13" [id=563, type=add]; +"564 layer_norm_13" [id=564, type=layer_norm]; +"565 layer_norm_13_0_0_nncf_smooth_quant_0" [id=565, type=call_module]; +"566 quantize_per_tensor_default_41" [id=566, type=quantize_per_tensor]; +"567 dequantize_per_tensor_default_41" [id=567, type=dequantize_per_tensor]; +"568 linear_26_scale_0" [id=568, type=get_attr]; +"569 linear_26_zero_point_0" [id=569, type=get_attr]; +"570 _frozen_param27" [id=570, type=get_attr]; +"571 dequantize_per_channel_default_27" [id=571, type=dequantize_per_channel]; +"572 linear_26" [id=572, type=linear]; +"573 gelu_6" [id=573, type=gelu]; +"574 dropout_20" [id=574, type=dropout]; +"575 dropout_20_0_0_nncf_smooth_quant_0" [id=575, type=call_module]; +"576 quantize_per_tensor_default_42" [id=576, type=quantize_per_tensor]; +"577 dequantize_per_tensor_default_42" [id=577, type=dequantize_per_tensor]; +"578 linear_27_scale_0" [id=578, type=get_attr]; +"579 linear_27_zero_point_0" [id=579, type=get_attr]; +"580 _frozen_param28" [id=580, type=get_attr]; +"581 dequantize_per_channel_default_28" [id=581, type=dequantize_per_channel]; +"582 linear_27" [id=582, type=linear]; +"583 dropout_21" [id=583, type=dropout]; +"584 add_14" [id=584, type=add]; +"585 layer_norm_14" [id=585, type=layer_norm]; +"586 transpose_42" [id=586, type=transpose]; +"587 transpose_42_0_0_nncf_smooth_quant_0" [id=587, type=call_module]; +"588 quantize_per_tensor_default_43" [id=588, type=quantize_per_tensor]; +"589 dequantize_per_tensor_default_43" [id=589, type=dequantize_per_tensor]; +"590 linear_28_scale_0" [id=590, type=get_attr]; +"591 linear_28_zero_point_0" [id=591, type=get_attr]; +"592 _frozen_param29" [id=592, type=get_attr]; +"593 dequantize_per_channel_default_29" [id=593, type=dequantize_per_channel]; +"594 linear_28" [id=594, type=linear]; +"595 unflatten_7" [id=595, type=unflatten]; +"596 unsqueeze_7" [id=596, type=unsqueeze]; +"597 transpose_43" [id=597, type=transpose]; +"598 squeeze_7" [id=598, type=squeeze]; +"599 contiguous_7" [id=599, type=contiguous]; +"600 quantize_per_tensor_default_44" [id=600, type=quantize_per_tensor]; +"601 dequantize_per_tensor_default_44" [id=601, type=dequantize_per_tensor]; +"602 select_21" [id=602, type=select]; +"603 quantize_per_tensor_default_45" [id=603, type=quantize_per_tensor]; +"604 dequantize_per_tensor_default_45" [id=604, type=dequantize_per_tensor]; +"605 select_22" [id=605, type=select]; +"606 select_23" [id=606, type=select]; +"607 view_56" [id=607, type=view]; +"608 transpose_44" [id=608, type=transpose]; +"609 view_57" [id=609, type=view]; +"610 transpose_45" [id=610, type=transpose]; +"611 view_58" [id=611, type=view]; +"612 transpose_46" [id=612, type=transpose]; +"613 view_59" [id=613, type=view]; +"614 view_60" [id=614, type=view]; +"615 view_61" [id=615, type=view]; +"616 scaled_dot_product_attention_7" [id=616, type=scaled_dot_product_attention]; +"617 permute_8" [id=617, type=permute]; +"618 view_62" [id=618, type=view]; +"619 view_62_0_0_nncf_smooth_quant_0" [id=619, type=call_module]; +"620 quantize_per_tensor_default_46" [id=620, type=quantize_per_tensor]; +"621 dequantize_per_tensor_default_46" [id=621, type=dequantize_per_tensor]; +"622 linear_29_scale_0" [id=622, type=get_attr]; +"623 linear_29_zero_point_0" [id=623, type=get_attr]; +"624 _frozen_param30" [id=624, type=get_attr]; +"625 dequantize_per_channel_default_30" [id=625, type=dequantize_per_channel]; +"626 linear_29" [id=626, type=linear]; +"627 view_63" [id=627, type=view]; +"628 transpose_47" [id=628, type=transpose]; +"629 dropout_22" [id=629, type=dropout]; +"630 add_15" [id=630, type=add]; +"631 layer_norm_15" [id=631, type=layer_norm]; +"632 layer_norm_15_0_0_nncf_smooth_quant_0" [id=632, type=call_module]; +"633 quantize_per_tensor_default_47" [id=633, type=quantize_per_tensor]; +"634 dequantize_per_tensor_default_47" [id=634, type=dequantize_per_tensor]; +"635 linear_30_scale_0" [id=635, type=get_attr]; +"636 linear_30_zero_point_0" [id=636, type=get_attr]; +"637 _frozen_param31" [id=637, type=get_attr]; +"638 dequantize_per_channel_default_31" [id=638, type=dequantize_per_channel]; +"639 linear_30" [id=639, type=linear]; +"640 gelu_7" [id=640, type=gelu]; +"641 dropout_23" [id=641, type=dropout]; +"642 dropout_23_0_0_nncf_smooth_quant_0" [id=642, type=call_module]; +"643 quantize_per_tensor_default_48" [id=643, type=quantize_per_tensor]; +"644 dequantize_per_tensor_default_48" [id=644, type=dequantize_per_tensor]; +"645 linear_31_scale_0" [id=645, type=get_attr]; +"646 linear_31_zero_point_0" [id=646, type=get_attr]; +"647 _frozen_param32" [id=647, type=get_attr]; +"648 dequantize_per_channel_default_32" [id=648, type=dequantize_per_channel]; +"649 linear_31" [id=649, type=linear]; +"650 dropout_24" [id=650, type=dropout]; +"651 add_16" [id=651, type=add]; +"652 layer_norm_16" [id=652, type=layer_norm]; +"653 transpose_48" [id=653, type=transpose]; +"654 transpose_48_0_0_nncf_smooth_quant_0" [id=654, type=call_module]; +"655 quantize_per_tensor_default_49" [id=655, type=quantize_per_tensor]; +"656 dequantize_per_tensor_default_49" [id=656, type=dequantize_per_tensor]; +"657 linear_32_scale_0" [id=657, type=get_attr]; +"658 linear_32_zero_point_0" [id=658, type=get_attr]; +"659 _frozen_param33" [id=659, type=get_attr]; +"660 dequantize_per_channel_default_33" [id=660, type=dequantize_per_channel]; +"661 linear_32" [id=661, type=linear]; +"662 unflatten_8" [id=662, type=unflatten]; +"663 unsqueeze_8" [id=663, type=unsqueeze]; +"664 transpose_49" [id=664, type=transpose]; +"665 squeeze_8" [id=665, type=squeeze]; +"666 contiguous_8" [id=666, type=contiguous]; +"667 quantize_per_tensor_default_50" [id=667, type=quantize_per_tensor]; +"668 dequantize_per_tensor_default_50" [id=668, type=dequantize_per_tensor]; +"669 select_24" [id=669, type=select]; +"670 quantize_per_tensor_default_51" [id=670, type=quantize_per_tensor]; +"671 dequantize_per_tensor_default_51" [id=671, type=dequantize_per_tensor]; +"672 select_25" [id=672, type=select]; +"673 select_26" [id=673, type=select]; +"674 view_64" [id=674, type=view]; +"675 transpose_50" [id=675, type=transpose]; +"676 view_65" [id=676, type=view]; +"677 transpose_51" [id=677, type=transpose]; +"678 view_66" [id=678, type=view]; +"679 transpose_52" [id=679, type=transpose]; +"680 view_67" [id=680, type=view]; +"681 view_68" [id=681, type=view]; +"682 view_69" [id=682, type=view]; +"683 scaled_dot_product_attention_8" [id=683, type=scaled_dot_product_attention]; +"684 permute_9" [id=684, type=permute]; +"685 view_70" [id=685, type=view]; +"686 view_70_0_0_nncf_smooth_quant_0" [id=686, type=call_module]; +"687 quantize_per_tensor_default_52" [id=687, type=quantize_per_tensor]; +"688 dequantize_per_tensor_default_52" [id=688, type=dequantize_per_tensor]; +"689 linear_33_scale_0" [id=689, type=get_attr]; +"690 linear_33_zero_point_0" [id=690, type=get_attr]; +"691 _frozen_param34" [id=691, type=get_attr]; +"692 dequantize_per_channel_default_34" [id=692, type=dequantize_per_channel]; +"693 linear_33" [id=693, type=linear]; +"694 view_71" [id=694, type=view]; +"695 transpose_53" [id=695, type=transpose]; +"696 dropout_25" [id=696, type=dropout]; +"697 add_17" [id=697, type=add]; +"698 layer_norm_17" [id=698, type=layer_norm]; +"699 layer_norm_17_0_0_nncf_smooth_quant_0" [id=699, type=call_module]; +"700 quantize_per_tensor_default_53" [id=700, type=quantize_per_tensor]; +"701 dequantize_per_tensor_default_53" [id=701, type=dequantize_per_tensor]; +"702 linear_34_scale_0" [id=702, type=get_attr]; +"703 linear_34_zero_point_0" [id=703, type=get_attr]; +"704 _frozen_param35" [id=704, type=get_attr]; +"705 dequantize_per_channel_default_35" [id=705, type=dequantize_per_channel]; +"706 linear_34" [id=706, type=linear]; +"707 gelu_8" [id=707, type=gelu]; +"708 dropout_26" [id=708, type=dropout]; +"709 dropout_26_0_0_nncf_smooth_quant_0" [id=709, type=call_module]; +"710 quantize_per_tensor_default_54" [id=710, type=quantize_per_tensor]; +"711 dequantize_per_tensor_default_54" [id=711, type=dequantize_per_tensor]; +"712 linear_35_scale_0" [id=712, type=get_attr]; +"713 linear_35_zero_point_0" [id=713, type=get_attr]; +"714 _frozen_param36" [id=714, type=get_attr]; +"715 dequantize_per_channel_default_36" [id=715, type=dequantize_per_channel]; +"716 linear_35" [id=716, type=linear]; +"717 dropout_27" [id=717, type=dropout]; +"718 add_18" [id=718, type=add]; +"719 layer_norm_18" [id=719, type=layer_norm]; +"720 transpose_54" [id=720, type=transpose]; +"721 transpose_54_0_0_nncf_smooth_quant_0" [id=721, type=call_module]; +"722 quantize_per_tensor_default_55" [id=722, type=quantize_per_tensor]; +"723 dequantize_per_tensor_default_55" [id=723, type=dequantize_per_tensor]; +"724 linear_36_scale_0" [id=724, type=get_attr]; +"725 linear_36_zero_point_0" [id=725, type=get_attr]; +"726 _frozen_param37" [id=726, type=get_attr]; +"727 dequantize_per_channel_default_37" [id=727, type=dequantize_per_channel]; +"728 linear_36" [id=728, type=linear]; +"729 unflatten_9" [id=729, type=unflatten]; +"730 unsqueeze_9" [id=730, type=unsqueeze]; +"731 transpose_55" [id=731, type=transpose]; +"732 squeeze_9" [id=732, type=squeeze]; +"733 contiguous_9" [id=733, type=contiguous]; +"734 quantize_per_tensor_default_56" [id=734, type=quantize_per_tensor]; +"735 dequantize_per_tensor_default_56" [id=735, type=dequantize_per_tensor]; +"736 select_27" [id=736, type=select]; +"737 quantize_per_tensor_default_57" [id=737, type=quantize_per_tensor]; +"738 dequantize_per_tensor_default_57" [id=738, type=dequantize_per_tensor]; +"739 select_28" [id=739, type=select]; +"740 select_29" [id=740, type=select]; +"741 view_72" [id=741, type=view]; +"742 transpose_56" [id=742, type=transpose]; +"743 view_73" [id=743, type=view]; +"744 transpose_57" [id=744, type=transpose]; +"745 view_74" [id=745, type=view]; +"746 transpose_58" [id=746, type=transpose]; +"747 view_75" [id=747, type=view]; +"748 view_76" [id=748, type=view]; +"749 view_77" [id=749, type=view]; +"750 scaled_dot_product_attention_9" [id=750, type=scaled_dot_product_attention]; +"751 permute_10" [id=751, type=permute]; +"752 view_78" [id=752, type=view]; +"753 view_78_0_0_nncf_smooth_quant_0" [id=753, type=call_module]; +"754 quantize_per_tensor_default_58" [id=754, type=quantize_per_tensor]; +"755 dequantize_per_tensor_default_58" [id=755, type=dequantize_per_tensor]; +"756 linear_37_scale_0" [id=756, type=get_attr]; +"757 linear_37_zero_point_0" [id=757, type=get_attr]; +"758 _frozen_param38" [id=758, type=get_attr]; +"759 dequantize_per_channel_default_38" [id=759, type=dequantize_per_channel]; +"760 linear_37" [id=760, type=linear]; +"761 view_79" [id=761, type=view]; +"762 transpose_59" [id=762, type=transpose]; +"763 dropout_28" [id=763, type=dropout]; +"764 add_19" [id=764, type=add]; +"765 layer_norm_19" [id=765, type=layer_norm]; +"766 layer_norm_19_0_0_nncf_smooth_quant_0" [id=766, type=call_module]; +"767 quantize_per_tensor_default_59" [id=767, type=quantize_per_tensor]; +"768 dequantize_per_tensor_default_59" [id=768, type=dequantize_per_tensor]; +"769 linear_38_scale_0" [id=769, type=get_attr]; +"770 linear_38_zero_point_0" [id=770, type=get_attr]; +"771 _frozen_param39" [id=771, type=get_attr]; +"772 dequantize_per_channel_default_39" [id=772, type=dequantize_per_channel]; +"773 linear_38" [id=773, type=linear]; +"774 gelu_9" [id=774, type=gelu]; +"775 dropout_29" [id=775, type=dropout]; +"776 dropout_29_0_0_nncf_smooth_quant_0" [id=776, type=call_module]; +"777 quantize_per_tensor_default_60" [id=777, type=quantize_per_tensor]; +"778 dequantize_per_tensor_default_60" [id=778, type=dequantize_per_tensor]; +"779 linear_39_scale_0" [id=779, type=get_attr]; +"780 linear_39_zero_point_0" [id=780, type=get_attr]; +"781 _frozen_param40" [id=781, type=get_attr]; +"782 dequantize_per_channel_default_40" [id=782, type=dequantize_per_channel]; +"783 linear_39" [id=783, type=linear]; +"784 dropout_30" [id=784, type=dropout]; +"785 add_20" [id=785, type=add]; +"786 layer_norm_20" [id=786, type=layer_norm]; +"787 transpose_60" [id=787, type=transpose]; +"788 transpose_60_0_0_nncf_smooth_quant_0" [id=788, type=call_module]; +"789 quantize_per_tensor_default_61" [id=789, type=quantize_per_tensor]; +"790 dequantize_per_tensor_default_61" [id=790, type=dequantize_per_tensor]; +"791 linear_40_scale_0" [id=791, type=get_attr]; +"792 linear_40_zero_point_0" [id=792, type=get_attr]; +"793 _frozen_param41" [id=793, type=get_attr]; +"794 dequantize_per_channel_default_41" [id=794, type=dequantize_per_channel]; +"795 linear_40" [id=795, type=linear]; +"796 unflatten_10" [id=796, type=unflatten]; +"797 unsqueeze_10" [id=797, type=unsqueeze]; +"798 transpose_61" [id=798, type=transpose]; +"799 squeeze_10" [id=799, type=squeeze]; +"800 contiguous_10" [id=800, type=contiguous]; +"801 quantize_per_tensor_default_62" [id=801, type=quantize_per_tensor]; +"802 dequantize_per_tensor_default_62" [id=802, type=dequantize_per_tensor]; +"803 select_30" [id=803, type=select]; +"804 quantize_per_tensor_default_63" [id=804, type=quantize_per_tensor]; +"805 dequantize_per_tensor_default_63" [id=805, type=dequantize_per_tensor]; +"806 select_31" [id=806, type=select]; +"807 select_32" [id=807, type=select]; +"808 view_80" [id=808, type=view]; +"809 transpose_62" [id=809, type=transpose]; +"810 view_81" [id=810, type=view]; +"811 transpose_63" [id=811, type=transpose]; +"812 view_82" [id=812, type=view]; +"813 transpose_64" [id=813, type=transpose]; +"814 view_83" [id=814, type=view]; +"815 view_84" [id=815, type=view]; +"816 view_85" [id=816, type=view]; +"817 scaled_dot_product_attention_10" [id=817, type=scaled_dot_product_attention]; +"818 permute_11" [id=818, type=permute]; +"819 view_86" [id=819, type=view]; +"820 view_86_0_0_nncf_smooth_quant_0" [id=820, type=call_module]; +"821 quantize_per_tensor_default_64" [id=821, type=quantize_per_tensor]; +"822 dequantize_per_tensor_default_64" [id=822, type=dequantize_per_tensor]; +"823 linear_41_scale_0" [id=823, type=get_attr]; +"824 linear_41_zero_point_0" [id=824, type=get_attr]; +"825 _frozen_param42" [id=825, type=get_attr]; +"826 dequantize_per_channel_default_42" [id=826, type=dequantize_per_channel]; +"827 linear_41" [id=827, type=linear]; +"828 view_87" [id=828, type=view]; +"829 transpose_65" [id=829, type=transpose]; +"830 dropout_31" [id=830, type=dropout]; +"831 add_21" [id=831, type=add]; +"832 layer_norm_21" [id=832, type=layer_norm]; +"833 layer_norm_21_0_0_nncf_smooth_quant_0" [id=833, type=call_module]; +"834 quantize_per_tensor_default_65" [id=834, type=quantize_per_tensor]; +"835 dequantize_per_tensor_default_65" [id=835, type=dequantize_per_tensor]; +"836 linear_42_scale_0" [id=836, type=get_attr]; +"837 linear_42_zero_point_0" [id=837, type=get_attr]; +"838 _frozen_param43" [id=838, type=get_attr]; +"839 dequantize_per_channel_default_43" [id=839, type=dequantize_per_channel]; +"840 linear_42" [id=840, type=linear]; +"841 gelu_10" [id=841, type=gelu]; +"842 dropout_32" [id=842, type=dropout]; +"843 dropout_32_0_0_nncf_smooth_quant_0" [id=843, type=call_module]; +"844 quantize_per_tensor_default_66" [id=844, type=quantize_per_tensor]; +"845 dequantize_per_tensor_default_66" [id=845, type=dequantize_per_tensor]; +"846 linear_43_scale_0" [id=846, type=get_attr]; +"847 linear_43_zero_point_0" [id=847, type=get_attr]; +"848 _frozen_param44" [id=848, type=get_attr]; +"849 dequantize_per_channel_default_44" [id=849, type=dequantize_per_channel]; +"850 linear_43" [id=850, type=linear]; +"851 dropout_33" [id=851, type=dropout]; +"852 add_22" [id=852, type=add]; +"853 layer_norm_22" [id=853, type=layer_norm]; +"854 transpose_66" [id=854, type=transpose]; +"855 transpose_66_0_0_nncf_smooth_quant_0" [id=855, type=call_module]; +"856 quantize_per_tensor_default_67" [id=856, type=quantize_per_tensor]; +"857 dequantize_per_tensor_default_67" [id=857, type=dequantize_per_tensor]; +"858 linear_44_scale_0" [id=858, type=get_attr]; +"859 linear_44_zero_point_0" [id=859, type=get_attr]; +"860 _frozen_param45" [id=860, type=get_attr]; +"861 dequantize_per_channel_default_45" [id=861, type=dequantize_per_channel]; +"862 linear_44" [id=862, type=linear]; +"863 unflatten_11" [id=863, type=unflatten]; +"864 unsqueeze_11" [id=864, type=unsqueeze]; +"865 transpose_67" [id=865, type=transpose]; +"866 squeeze_11" [id=866, type=squeeze]; +"867 contiguous_11" [id=867, type=contiguous]; +"868 quantize_per_tensor_default_68" [id=868, type=quantize_per_tensor]; +"869 dequantize_per_tensor_default_68" [id=869, type=dequantize_per_tensor]; +"870 select_33" [id=870, type=select]; +"871 quantize_per_tensor_default_69" [id=871, type=quantize_per_tensor]; +"872 dequantize_per_tensor_default_69" [id=872, type=dequantize_per_tensor]; +"873 select_34" [id=873, type=select]; +"874 select_35" [id=874, type=select]; +"875 view_88" [id=875, type=view]; +"876 transpose_68" [id=876, type=transpose]; +"877 view_89" [id=877, type=view]; +"878 transpose_69" [id=878, type=transpose]; +"879 view_90" [id=879, type=view]; +"880 transpose_70" [id=880, type=transpose]; +"881 view_91" [id=881, type=view]; +"882 view_92" [id=882, type=view]; +"883 view_93" [id=883, type=view]; +"884 scaled_dot_product_attention_11" [id=884, type=scaled_dot_product_attention]; +"885 permute_12" [id=885, type=permute]; +"886 view_94" [id=886, type=view]; +"887 view_94_0_0_nncf_smooth_quant_0" [id=887, type=call_module]; +"888 quantize_per_tensor_default_70" [id=888, type=quantize_per_tensor]; +"889 dequantize_per_tensor_default_70" [id=889, type=dequantize_per_tensor]; +"890 linear_45_scale_0" [id=890, type=get_attr]; +"891 linear_45_zero_point_0" [id=891, type=get_attr]; +"892 _frozen_param46" [id=892, type=get_attr]; +"893 dequantize_per_channel_default_46" [id=893, type=dequantize_per_channel]; +"894 linear_45" [id=894, type=linear]; +"895 view_95" [id=895, type=view]; +"896 transpose_71" [id=896, type=transpose]; +"897 dropout_34" [id=897, type=dropout]; +"898 add_23" [id=898, type=add]; +"899 layer_norm_23" [id=899, type=layer_norm]; +"900 layer_norm_23_0_0_nncf_smooth_quant_0" [id=900, type=call_module]; +"901 quantize_per_tensor_default_71" [id=901, type=quantize_per_tensor]; +"902 dequantize_per_tensor_default_71" [id=902, type=dequantize_per_tensor]; +"903 linear_46_scale_0" [id=903, type=get_attr]; +"904 linear_46_zero_point_0" [id=904, type=get_attr]; +"905 _frozen_param47" [id=905, type=get_attr]; +"906 dequantize_per_channel_default_47" [id=906, type=dequantize_per_channel]; +"907 linear_46" [id=907, type=linear]; +"908 gelu_11" [id=908, type=gelu]; +"909 dropout_35" [id=909, type=dropout]; +"910 dropout_35_0_0_nncf_smooth_quant_0" [id=910, type=call_module]; +"911 quantize_per_tensor_default_72" [id=911, type=quantize_per_tensor]; +"912 dequantize_per_tensor_default_72" [id=912, type=dequantize_per_tensor]; +"913 linear_47_scale_0" [id=913, type=get_attr]; +"914 linear_47_zero_point_0" [id=914, type=get_attr]; +"915 _frozen_param48" [id=915, type=get_attr]; +"916 dequantize_per_channel_default_48" [id=916, type=dequantize_per_channel]; +"917 linear_47" [id=917, type=linear]; +"918 dropout_36" [id=918, type=dropout]; +"919 add_24" [id=919, type=add]; +"920 layer_norm_24" [id=920, type=layer_norm]; +"921 slice_1" [id=921, type=slice]; +"922 select_36" [id=922, type=select]; +"923 select_36_0_0_nncf_smooth_quant_0" [id=923, type=call_module]; +"924 quantize_per_tensor_default_73" [id=924, type=quantize_per_tensor]; +"925 dequantize_per_tensor_default_73" [id=925, type=dequantize_per_tensor]; +"926 linear_48_scale_0" [id=926, type=get_attr]; +"927 linear_48_zero_point_0" [id=927, type=get_attr]; +"928 _frozen_param49" [id=928, type=get_attr]; +"929 dequantize_per_channel_default_49" [id=929, type=dequantize_per_channel]; +"930 linear_48" [id=930, type=linear]; +"931 output" [id=931, type=output]; +"0 class_token" -> "112 expand" [label="(1, 1, 768)", style=solid]; +"1 encoder_pos_embedding" -> "114 add" [label="(1, 197, 768)", style=solid]; +"2 conv_proj_bias" -> "109 conv2d" [label="(768,)", style=solid]; +"3 encoder_layers_encoder_layer_0_ln_1_weight" -> "116 layer_norm" [label="(768,)", style=solid]; +"4 encoder_layers_encoder_layer_0_ln_1_bias" -> "116 layer_norm" [label="(768,)", style=solid]; +"5 encoder_layers_encoder_layer_0_self_attention_in_proj_bias" -> "125 linear" [label="(2304,)", style=solid]; +"6 encoder_layers_encoder_layer_0_self_attention_out_proj_bias" -> "157 linear_1" [label="(768,)", style=solid]; +"7 encoder_layers_encoder_layer_0_ln_2_weight" -> "162 layer_norm_1" [label="(768,)", style=solid]; +"8 encoder_layers_encoder_layer_0_ln_2_bias" -> "162 layer_norm_1" [label="(768,)", style=solid]; +"9 encoder_layers_encoder_layer_0_mlp_0_bias" -> "170 linear_2" [label="(3072,)", style=solid]; +"10 encoder_layers_encoder_layer_0_mlp_3_bias" -> "180 linear_3" [label="(768,)", style=solid]; +"11 encoder_layers_encoder_layer_1_ln_1_weight" -> "183 layer_norm_2" [label="(768,)", style=solid]; +"12 encoder_layers_encoder_layer_1_ln_1_bias" -> "183 layer_norm_2" [label="(768,)", style=solid]; +"13 encoder_layers_encoder_layer_1_self_attention_in_proj_bias" -> "192 linear_4" [label="(2304,)", style=solid]; +"14 encoder_layers_encoder_layer_1_self_attention_out_proj_bias" -> "224 linear_5" [label="(768,)", style=solid]; +"15 encoder_layers_encoder_layer_1_ln_2_weight" -> "229 layer_norm_3" [label="(768,)", style=solid]; +"16 encoder_layers_encoder_layer_1_ln_2_bias" -> "229 layer_norm_3" [label="(768,)", style=solid]; +"17 encoder_layers_encoder_layer_1_mlp_0_bias" -> "237 linear_6" [label="(3072,)", style=solid]; +"18 encoder_layers_encoder_layer_1_mlp_3_bias" -> "247 linear_7" [label="(768,)", style=solid]; +"19 encoder_layers_encoder_layer_2_ln_1_weight" -> "250 layer_norm_4" [label="(768,)", style=solid]; +"20 encoder_layers_encoder_layer_2_ln_1_bias" -> "250 layer_norm_4" [label="(768,)", style=solid]; +"21 encoder_layers_encoder_layer_2_self_attention_in_proj_bias" -> "259 linear_8" [label="(2304,)", style=solid]; +"22 encoder_layers_encoder_layer_2_self_attention_out_proj_bias" -> "291 linear_9" [label="(768,)", style=solid]; +"23 encoder_layers_encoder_layer_2_ln_2_weight" -> "296 layer_norm_5" [label="(768,)", style=solid]; +"24 encoder_layers_encoder_layer_2_ln_2_bias" -> "296 layer_norm_5" [label="(768,)", style=solid]; +"25 encoder_layers_encoder_layer_2_mlp_0_bias" -> "304 linear_10" [label="(3072,)", style=solid]; +"26 encoder_layers_encoder_layer_2_mlp_3_bias" -> "314 linear_11" [label="(768,)", style=solid]; +"27 encoder_layers_encoder_layer_3_ln_1_weight" -> "317 layer_norm_6" [label="(768,)", style=solid]; +"28 encoder_layers_encoder_layer_3_ln_1_bias" -> "317 layer_norm_6" [label="(768,)", style=solid]; +"29 encoder_layers_encoder_layer_3_self_attention_in_proj_bias" -> "326 linear_12" [label="(2304,)", style=solid]; +"30 encoder_layers_encoder_layer_3_self_attention_out_proj_bias" -> "358 linear_13" [label="(768,)", style=solid]; +"31 encoder_layers_encoder_layer_3_ln_2_weight" -> "363 layer_norm_7" [label="(768,)", style=solid]; +"32 encoder_layers_encoder_layer_3_ln_2_bias" -> "363 layer_norm_7" [label="(768,)", style=solid]; +"33 encoder_layers_encoder_layer_3_mlp_0_bias" -> "371 linear_14" [label="(3072,)", style=solid]; +"34 encoder_layers_encoder_layer_3_mlp_3_bias" -> "381 linear_15" [label="(768,)", style=solid]; +"35 encoder_layers_encoder_layer_4_ln_1_weight" -> "384 layer_norm_8" [label="(768,)", style=solid]; +"36 encoder_layers_encoder_layer_4_ln_1_bias" -> "384 layer_norm_8" [label="(768,)", style=solid]; +"37 encoder_layers_encoder_layer_4_self_attention_in_proj_bias" -> "393 linear_16" [label="(2304,)", style=solid]; +"38 encoder_layers_encoder_layer_4_self_attention_out_proj_bias" -> "425 linear_17" [label="(768,)", style=solid]; +"39 encoder_layers_encoder_layer_4_ln_2_weight" -> "430 layer_norm_9" [label="(768,)", style=solid]; +"40 encoder_layers_encoder_layer_4_ln_2_bias" -> "430 layer_norm_9" [label="(768,)", style=solid]; +"41 encoder_layers_encoder_layer_4_mlp_0_bias" -> "438 linear_18" [label="(3072,)", style=solid]; +"42 encoder_layers_encoder_layer_4_mlp_3_bias" -> "448 linear_19" [label="(768,)", style=solid]; +"43 encoder_layers_encoder_layer_5_ln_1_weight" -> "451 layer_norm_10" [label="(768,)", style=solid]; +"44 encoder_layers_encoder_layer_5_ln_1_bias" -> "451 layer_norm_10" [label="(768,)", style=solid]; +"45 encoder_layers_encoder_layer_5_self_attention_in_proj_bias" -> "460 linear_20" [label="(2304,)", style=solid]; +"46 encoder_layers_encoder_layer_5_self_attention_out_proj_bias" -> "492 linear_21" [label="(768,)", style=solid]; +"47 encoder_layers_encoder_layer_5_ln_2_weight" -> "497 layer_norm_11" [label="(768,)", style=solid]; +"48 encoder_layers_encoder_layer_5_ln_2_bias" -> "497 layer_norm_11" [label="(768,)", style=solid]; +"49 encoder_layers_encoder_layer_5_mlp_0_bias" -> "505 linear_22" [label="(3072,)", style=solid]; +"50 encoder_layers_encoder_layer_5_mlp_3_bias" -> "515 linear_23" [label="(768,)", style=solid]; +"51 encoder_layers_encoder_layer_6_ln_1_weight" -> "518 layer_norm_12" [label="(768,)", style=solid]; +"52 encoder_layers_encoder_layer_6_ln_1_bias" -> "518 layer_norm_12" [label="(768,)", style=solid]; +"53 encoder_layers_encoder_layer_6_self_attention_in_proj_bias" -> "527 linear_24" [label="(2304,)", style=solid]; +"54 encoder_layers_encoder_layer_6_self_attention_out_proj_bias" -> "559 linear_25" [label="(768,)", style=solid]; +"55 encoder_layers_encoder_layer_6_ln_2_weight" -> "564 layer_norm_13" [label="(768,)", style=solid]; +"56 encoder_layers_encoder_layer_6_ln_2_bias" -> "564 layer_norm_13" [label="(768,)", style=solid]; +"57 encoder_layers_encoder_layer_6_mlp_0_bias" -> "572 linear_26" [label="(3072,)", style=solid]; +"58 encoder_layers_encoder_layer_6_mlp_3_bias" -> "582 linear_27" [label="(768,)", style=solid]; +"59 encoder_layers_encoder_layer_7_ln_1_weight" -> "585 layer_norm_14" [label="(768,)", style=solid]; +"60 encoder_layers_encoder_layer_7_ln_1_bias" -> "585 layer_norm_14" [label="(768,)", style=solid]; +"61 encoder_layers_encoder_layer_7_self_attention_in_proj_bias" -> "594 linear_28" [label="(2304,)", style=solid]; +"62 encoder_layers_encoder_layer_7_self_attention_out_proj_bias" -> "626 linear_29" [label="(768,)", style=solid]; +"63 encoder_layers_encoder_layer_7_ln_2_weight" -> "631 layer_norm_15" [label="(768,)", style=solid]; +"64 encoder_layers_encoder_layer_7_ln_2_bias" -> "631 layer_norm_15" [label="(768,)", style=solid]; +"65 encoder_layers_encoder_layer_7_mlp_0_bias" -> "639 linear_30" [label="(3072,)", style=solid]; +"66 encoder_layers_encoder_layer_7_mlp_3_bias" -> "649 linear_31" [label="(768,)", style=solid]; +"67 encoder_layers_encoder_layer_8_ln_1_weight" -> "652 layer_norm_16" [label="(768,)", style=solid]; +"68 encoder_layers_encoder_layer_8_ln_1_bias" -> "652 layer_norm_16" [label="(768,)", style=solid]; +"69 encoder_layers_encoder_layer_8_self_attention_in_proj_bias" -> "661 linear_32" [label="(2304,)", style=solid]; +"70 encoder_layers_encoder_layer_8_self_attention_out_proj_bias" -> "693 linear_33" [label="(768,)", style=solid]; +"71 encoder_layers_encoder_layer_8_ln_2_weight" -> "698 layer_norm_17" [label="(768,)", style=solid]; +"72 encoder_layers_encoder_layer_8_ln_2_bias" -> "698 layer_norm_17" [label="(768,)", style=solid]; +"73 encoder_layers_encoder_layer_8_mlp_0_bias" -> "706 linear_34" [label="(3072,)", style=solid]; +"74 encoder_layers_encoder_layer_8_mlp_3_bias" -> "716 linear_35" [label="(768,)", style=solid]; +"75 encoder_layers_encoder_layer_9_ln_1_weight" -> "719 layer_norm_18" [label="(768,)", style=solid]; +"76 encoder_layers_encoder_layer_9_ln_1_bias" -> "719 layer_norm_18" [label="(768,)", style=solid]; +"77 encoder_layers_encoder_layer_9_self_attention_in_proj_bias" -> "728 linear_36" [label="(2304,)", style=solid]; +"78 encoder_layers_encoder_layer_9_self_attention_out_proj_bias" -> "760 linear_37" [label="(768,)", style=solid]; +"79 encoder_layers_encoder_layer_9_ln_2_weight" -> "765 layer_norm_19" [label="(768,)", style=solid]; +"80 encoder_layers_encoder_layer_9_ln_2_bias" -> "765 layer_norm_19" [label="(768,)", style=solid]; +"81 encoder_layers_encoder_layer_9_mlp_0_bias" -> "773 linear_38" [label="(3072,)", style=solid]; +"82 encoder_layers_encoder_layer_9_mlp_3_bias" -> "783 linear_39" [label="(768,)", style=solid]; +"83 encoder_layers_encoder_layer_10_ln_1_weight" -> "786 layer_norm_20" [label="(768,)", style=solid]; +"84 encoder_layers_encoder_layer_10_ln_1_bias" -> "786 layer_norm_20" [label="(768,)", style=solid]; +"85 encoder_layers_encoder_layer_10_self_attention_in_proj_bias" -> "795 linear_40" [label="(2304,)", style=solid]; +"86 encoder_layers_encoder_layer_10_self_attention_out_proj_bias" -> "827 linear_41" [label="(768,)", style=solid]; +"87 encoder_layers_encoder_layer_10_ln_2_weight" -> "832 layer_norm_21" [label="(768,)", style=solid]; +"88 encoder_layers_encoder_layer_10_ln_2_bias" -> "832 layer_norm_21" [label="(768,)", style=solid]; +"89 encoder_layers_encoder_layer_10_mlp_0_bias" -> "840 linear_42" [label="(3072,)", style=solid]; +"90 encoder_layers_encoder_layer_10_mlp_3_bias" -> "850 linear_43" [label="(768,)", style=solid]; +"91 encoder_layers_encoder_layer_11_ln_1_weight" -> "853 layer_norm_22" [label="(768,)", style=solid]; +"92 encoder_layers_encoder_layer_11_ln_1_bias" -> "853 layer_norm_22" [label="(768,)", style=solid]; +"93 encoder_layers_encoder_layer_11_self_attention_in_proj_bias" -> "862 linear_44" [label="(2304,)", style=solid]; +"94 encoder_layers_encoder_layer_11_self_attention_out_proj_bias" -> "894 linear_45" [label="(768,)", style=solid]; +"95 encoder_layers_encoder_layer_11_ln_2_weight" -> "899 layer_norm_23" [label="(768,)", style=solid]; +"96 encoder_layers_encoder_layer_11_ln_2_bias" -> "899 layer_norm_23" [label="(768,)", style=solid]; +"97 encoder_layers_encoder_layer_11_mlp_0_bias" -> "907 linear_46" [label="(3072,)", style=solid]; +"98 encoder_layers_encoder_layer_11_mlp_3_bias" -> "917 linear_47" [label="(768,)", style=solid]; +"99 encoder_ln_weight" -> "920 layer_norm_24" [label="(768,)", style=solid]; +"100 encoder_ln_bias" -> "920 layer_norm_24" [label="(768,)", style=solid]; +"101 heads_head_bias" -> "930 linear_48" [label="(1000,)", style=solid]; +"102 x" -> "103 quantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; +"103 quantize_per_tensor_default" -> "104 dequantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; +"104 dequantize_per_tensor_default" -> "109 conv2d" [label="(1, 3, 224, 224)", style=solid]; +"105 conv2d_scale_0" -> "108 dequantize_per_channel_default" [label="(768,)", style=solid]; +"106 conv2d_zero_point_0" -> "108 dequantize_per_channel_default" [label="(768,)", style=solid]; +"107 _frozen_param0" -> "108 dequantize_per_channel_default" [label="(768, 3, 16, 16)", style=solid]; +"108 dequantize_per_channel_default" -> "109 conv2d" [label="(768, 3, 16, 16)", style=solid]; +"109 conv2d" -> "110 reshape" [label="(1, 768, 14, 14)", style=solid]; +"110 reshape" -> "111 permute" [label="(1, 768, 196)", style=solid]; +"111 permute" -> "113 cat" [label="(1, 196, 768)", style=solid]; +"112 expand" -> "113 cat" [label="(1, 1, 768)", style=solid]; +"113 cat" -> "114 add" [label="(1, 197, 768)", style=solid]; +"114 add" -> "115 dropout" [label="(1, 197, 768)", style=solid]; +"115 dropout" -> "116 layer_norm" [label="(1, 197, 768)", style=solid]; +"115 dropout" -> "161 add_1" [label="(1, 197, 768)", style=solid]; +"116 layer_norm" -> "117 transpose" [label="(1, 197, 768)", style=solid]; +"117 transpose" -> "118 transpose_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"118 transpose_0_0_nncf_smooth_quant_0" -> "119 quantize_per_tensor_default_1" [label="(197, 1, 768)", style=solid]; +"119 quantize_per_tensor_default_1" -> "120 dequantize_per_tensor_default_1" [label="(197, 1, 768)", style=solid]; +"120 dequantize_per_tensor_default_1" -> "125 linear" [label="(197, 1, 768)", style=solid]; +"121 linear_scale_0" -> "124 dequantize_per_channel_default_1" [label="(2304,)", style=solid]; +"122 linear_zero_point_0" -> "124 dequantize_per_channel_default_1" [label="(2304,)", style=solid]; +"123 _frozen_param1" -> "124 dequantize_per_channel_default_1" [label="(2304, 768)", style=solid]; +"124 dequantize_per_channel_default_1" -> "125 linear" [label="(2304, 768)", style=solid]; +"125 linear" -> "126 unflatten" [label="(197, 1, 2304)", style=solid]; +"126 unflatten" -> "127 unsqueeze" [label="(197, 1, 3, 768)", style=solid]; +"127 unsqueeze" -> "128 transpose_1" [label="(1, 197, 1, 3, 768)", style=solid]; +"128 transpose_1" -> "129 squeeze" [label="(3, 197, 1, 1, 768)", style=solid]; +"129 squeeze" -> "130 contiguous" [label="(3, 197, 1, 768)", style=solid]; +"130 contiguous" -> "131 quantize_per_tensor_default_2" [label="(3, 197, 1, 768)", style=solid]; +"130 contiguous" -> "134 quantize_per_tensor_default_3" [label="(3, 197, 1, 768)", style=solid]; +"130 contiguous" -> "137 select_2" [label="(3, 197, 1, 768)", style=solid]; +"131 quantize_per_tensor_default_2" -> "132 dequantize_per_tensor_default_2" [label="(3, 197, 1, 768)", style=solid]; +"132 dequantize_per_tensor_default_2" -> "133 select" [label="(3, 197, 1, 768)", style=solid]; +"133 select" -> "138 view" [label="(197, 1, 768)", style=solid]; +"134 quantize_per_tensor_default_3" -> "135 dequantize_per_tensor_default_3" [label="(3, 197, 1, 768)", style=solid]; +"135 dequantize_per_tensor_default_3" -> "136 select_1" [label="(3, 197, 1, 768)", style=solid]; +"136 select_1" -> "140 view_1" [label="(197, 1, 768)", style=solid]; +"137 select_2" -> "142 view_2" [label="(197, 1, 768)", style=solid]; +"138 view" -> "139 transpose_2" [label="(197, 12, 64)", style=solid]; +"139 transpose_2" -> "144 view_3" [label="(12, 197, 64)", style=solid]; +"140 view_1" -> "141 transpose_3" [label="(197, 12, 64)", style=solid]; +"141 transpose_3" -> "145 view_4" [label="(12, 197, 64)", style=solid]; +"142 view_2" -> "143 transpose_4" [label="(197, 12, 64)", style=solid]; +"143 transpose_4" -> "146 view_5" [label="(12, 197, 64)", style=solid]; +"144 view_3" -> "147 scaled_dot_product_attention" [label="(1, 12, 197, 64)", style=solid]; +"145 view_4" -> "147 scaled_dot_product_attention" [label="(1, 12, 197, 64)", style=solid]; +"146 view_5" -> "147 scaled_dot_product_attention" [label="(1, 12, 197, 64)", style=solid]; +"147 scaled_dot_product_attention" -> "148 permute_1" [label="(1, 12, 197, 64)", style=solid]; +"148 permute_1" -> "149 view_6" [label="(197, 1, 12, 64)", style=solid]; +"149 view_6" -> "150 view_6_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"150 view_6_0_0_nncf_smooth_quant_0" -> "151 quantize_per_tensor_default_4" [label="(197, 768)", style=solid]; +"151 quantize_per_tensor_default_4" -> "152 dequantize_per_tensor_default_4" [label="(197, 768)", style=solid]; +"152 dequantize_per_tensor_default_4" -> "157 linear_1" [label="(197, 768)", style=solid]; +"153 linear_1_scale_0" -> "156 dequantize_per_channel_default_2" [label="(768,)", style=solid]; +"154 linear_1_zero_point_0" -> "156 dequantize_per_channel_default_2" [label="(768,)", style=solid]; +"155 _frozen_param2" -> "156 dequantize_per_channel_default_2" [label="(768, 768)", style=solid]; +"156 dequantize_per_channel_default_2" -> "157 linear_1" [label="(768, 768)", style=solid]; +"157 linear_1" -> "158 view_7" [label="(197, 768)", style=solid]; +"158 view_7" -> "159 transpose_5" [label="(197, 1, 768)", style=solid]; +"159 transpose_5" -> "160 dropout_1" [label="(1, 197, 768)", style=solid]; +"160 dropout_1" -> "161 add_1" [label="(1, 197, 768)", style=solid]; +"161 add_1" -> "162 layer_norm_1" [label="(1, 197, 768)", style=solid]; +"161 add_1" -> "182 add_2" [label="(1, 197, 768)", style=solid]; +"162 layer_norm_1" -> "163 layer_norm_1_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"163 layer_norm_1_0_0_nncf_smooth_quant_0" -> "164 quantize_per_tensor_default_5" [label="(1, 197, 768)", style=solid]; +"164 quantize_per_tensor_default_5" -> "165 dequantize_per_tensor_default_5" [label="(1, 197, 768)", style=solid]; +"165 dequantize_per_tensor_default_5" -> "170 linear_2" [label="(1, 197, 768)", style=solid]; +"166 linear_2_scale_0" -> "169 dequantize_per_channel_default_3" [label="(3072,)", style=solid]; +"167 linear_2_zero_point_0" -> "169 dequantize_per_channel_default_3" [label="(3072,)", style=solid]; +"168 _frozen_param3" -> "169 dequantize_per_channel_default_3" [label="(3072, 768)", style=solid]; +"169 dequantize_per_channel_default_3" -> "170 linear_2" [label="(3072, 768)", style=solid]; +"170 linear_2" -> "171 gelu" [label="(1, 197, 3072)", style=solid]; +"171 gelu" -> "172 dropout_2" [label="(1, 197, 3072)", style=solid]; +"172 dropout_2" -> "173 dropout_2_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"173 dropout_2_0_0_nncf_smooth_quant_0" -> "174 quantize_per_tensor_default_6" [label="(1, 197, 3072)", style=solid]; +"174 quantize_per_tensor_default_6" -> "175 dequantize_per_tensor_default_6" [label="(1, 197, 3072)", style=solid]; +"175 dequantize_per_tensor_default_6" -> "180 linear_3" [label="(1, 197, 3072)", style=solid]; +"176 linear_3_scale_0" -> "179 dequantize_per_channel_default_4" [label="(768,)", style=solid]; +"177 linear_3_zero_point_0" -> "179 dequantize_per_channel_default_4" [label="(768,)", style=solid]; +"178 _frozen_param4" -> "179 dequantize_per_channel_default_4" [label="(768, 3072)", style=solid]; +"179 dequantize_per_channel_default_4" -> "180 linear_3" [label="(768, 3072)", style=solid]; +"180 linear_3" -> "181 dropout_3" [label="(1, 197, 768)", style=solid]; +"181 dropout_3" -> "182 add_2" [label="(1, 197, 768)", style=solid]; +"182 add_2" -> "183 layer_norm_2" [label="(1, 197, 768)", style=solid]; +"182 add_2" -> "228 add_3" [label="(1, 197, 768)", style=solid]; +"183 layer_norm_2" -> "184 transpose_6" [label="(1, 197, 768)", style=solid]; +"184 transpose_6" -> "185 transpose_6_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"185 transpose_6_0_0_nncf_smooth_quant_0" -> "186 quantize_per_tensor_default_7" [label="(197, 1, 768)", style=solid]; +"186 quantize_per_tensor_default_7" -> "187 dequantize_per_tensor_default_7" [label="(197, 1, 768)", style=solid]; +"187 dequantize_per_tensor_default_7" -> "192 linear_4" [label="(197, 1, 768)", style=solid]; +"188 linear_4_scale_0" -> "191 dequantize_per_channel_default_5" [label="(2304,)", style=solid]; +"189 linear_4_zero_point_0" -> "191 dequantize_per_channel_default_5" [label="(2304,)", style=solid]; +"190 _frozen_param5" -> "191 dequantize_per_channel_default_5" [label="(2304, 768)", style=solid]; +"191 dequantize_per_channel_default_5" -> "192 linear_4" [label="(2304, 768)", style=solid]; +"192 linear_4" -> "193 unflatten_1" [label="(197, 1, 2304)", style=solid]; +"193 unflatten_1" -> "194 unsqueeze_1" [label="(197, 1, 3, 768)", style=solid]; +"194 unsqueeze_1" -> "195 transpose_7" [label="(1, 197, 1, 3, 768)", style=solid]; +"195 transpose_7" -> "196 squeeze_1" [label="(3, 197, 1, 1, 768)", style=solid]; +"196 squeeze_1" -> "197 contiguous_1" [label="(3, 197, 1, 768)", style=solid]; +"197 contiguous_1" -> "198 quantize_per_tensor_default_8" [label="(3, 197, 1, 768)", style=solid]; +"197 contiguous_1" -> "201 quantize_per_tensor_default_9" [label="(3, 197, 1, 768)", style=solid]; +"197 contiguous_1" -> "204 select_5" [label="(3, 197, 1, 768)", style=solid]; +"198 quantize_per_tensor_default_8" -> "199 dequantize_per_tensor_default_8" [label="(3, 197, 1, 768)", style=solid]; +"199 dequantize_per_tensor_default_8" -> "200 select_3" [label="(3, 197, 1, 768)", style=solid]; +"200 select_3" -> "205 view_8" [label="(197, 1, 768)", style=solid]; +"201 quantize_per_tensor_default_9" -> "202 dequantize_per_tensor_default_9" [label="(3, 197, 1, 768)", style=solid]; +"202 dequantize_per_tensor_default_9" -> "203 select_4" [label="(3, 197, 1, 768)", style=solid]; +"203 select_4" -> "207 view_9" [label="(197, 1, 768)", style=solid]; +"204 select_5" -> "209 view_10" [label="(197, 1, 768)", style=solid]; +"205 view_8" -> "206 transpose_8" [label="(197, 12, 64)", style=solid]; +"206 transpose_8" -> "211 view_11" [label="(12, 197, 64)", style=solid]; +"207 view_9" -> "208 transpose_9" [label="(197, 12, 64)", style=solid]; +"208 transpose_9" -> "212 view_12" [label="(12, 197, 64)", style=solid]; +"209 view_10" -> "210 transpose_10" [label="(197, 12, 64)", style=solid]; +"210 transpose_10" -> "213 view_13" [label="(12, 197, 64)", style=solid]; +"211 view_11" -> "214 scaled_dot_product_attention_1" [label="(1, 12, 197, 64)", style=solid]; +"212 view_12" -> "214 scaled_dot_product_attention_1" [label="(1, 12, 197, 64)", style=solid]; +"213 view_13" -> "214 scaled_dot_product_attention_1" [label="(1, 12, 197, 64)", style=solid]; +"214 scaled_dot_product_attention_1" -> "215 permute_2" [label="(1, 12, 197, 64)", style=solid]; +"215 permute_2" -> "216 view_14" [label="(197, 1, 12, 64)", style=solid]; +"216 view_14" -> "217 view_14_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"217 view_14_0_0_nncf_smooth_quant_0" -> "218 quantize_per_tensor_default_10" [label="(197, 768)", style=solid]; +"218 quantize_per_tensor_default_10" -> "219 dequantize_per_tensor_default_10" [label="(197, 768)", style=solid]; +"219 dequantize_per_tensor_default_10" -> "224 linear_5" [label="(197, 768)", style=solid]; +"220 linear_5_scale_0" -> "223 dequantize_per_channel_default_6" [label="(768,)", style=solid]; +"221 linear_5_zero_point_0" -> "223 dequantize_per_channel_default_6" [label="(768,)", style=solid]; +"222 _frozen_param6" -> "223 dequantize_per_channel_default_6" [label="(768, 768)", style=solid]; +"223 dequantize_per_channel_default_6" -> "224 linear_5" [label="(768, 768)", style=solid]; +"224 linear_5" -> "225 view_15" [label="(197, 768)", style=solid]; +"225 view_15" -> "226 transpose_11" [label="(197, 1, 768)", style=solid]; +"226 transpose_11" -> "227 dropout_4" [label="(1, 197, 768)", style=solid]; +"227 dropout_4" -> "228 add_3" [label="(1, 197, 768)", style=solid]; +"228 add_3" -> "229 layer_norm_3" [label="(1, 197, 768)", style=solid]; +"228 add_3" -> "249 add_4" [label="(1, 197, 768)", style=solid]; +"229 layer_norm_3" -> "230 layer_norm_3_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"230 layer_norm_3_0_0_nncf_smooth_quant_0" -> "231 quantize_per_tensor_default_11" [label="(1, 197, 768)", style=solid]; +"231 quantize_per_tensor_default_11" -> "232 dequantize_per_tensor_default_11" [label="(1, 197, 768)", style=solid]; +"232 dequantize_per_tensor_default_11" -> "237 linear_6" [label="(1, 197, 768)", style=solid]; +"233 linear_6_scale_0" -> "236 dequantize_per_channel_default_7" [label="(3072,)", style=solid]; +"234 linear_6_zero_point_0" -> "236 dequantize_per_channel_default_7" [label="(3072,)", style=solid]; +"235 _frozen_param7" -> "236 dequantize_per_channel_default_7" [label="(3072, 768)", style=solid]; +"236 dequantize_per_channel_default_7" -> "237 linear_6" [label="(3072, 768)", style=solid]; +"237 linear_6" -> "238 gelu_1" [label="(1, 197, 3072)", style=solid]; +"238 gelu_1" -> "239 dropout_5" [label="(1, 197, 3072)", style=solid]; +"239 dropout_5" -> "240 dropout_5_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"240 dropout_5_0_0_nncf_smooth_quant_0" -> "241 quantize_per_tensor_default_12" [label="(1, 197, 3072)", style=solid]; +"241 quantize_per_tensor_default_12" -> "242 dequantize_per_tensor_default_12" [label="(1, 197, 3072)", style=solid]; +"242 dequantize_per_tensor_default_12" -> "247 linear_7" [label="(1, 197, 3072)", style=solid]; +"243 linear_7_scale_0" -> "246 dequantize_per_channel_default_8" [label="(768,)", style=solid]; +"244 linear_7_zero_point_0" -> "246 dequantize_per_channel_default_8" [label="(768,)", style=solid]; +"245 _frozen_param8" -> "246 dequantize_per_channel_default_8" [label="(768, 3072)", style=solid]; +"246 dequantize_per_channel_default_8" -> "247 linear_7" [label="(768, 3072)", style=solid]; +"247 linear_7" -> "248 dropout_6" [label="(1, 197, 768)", style=solid]; +"248 dropout_6" -> "249 add_4" [label="(1, 197, 768)", style=solid]; +"249 add_4" -> "250 layer_norm_4" [label="(1, 197, 768)", style=solid]; +"249 add_4" -> "295 add_5" [label="(1, 197, 768)", style=solid]; +"250 layer_norm_4" -> "251 transpose_12" [label="(1, 197, 768)", style=solid]; +"251 transpose_12" -> "252 transpose_12_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"252 transpose_12_0_0_nncf_smooth_quant_0" -> "253 quantize_per_tensor_default_13" [label="(197, 1, 768)", style=solid]; +"253 quantize_per_tensor_default_13" -> "254 dequantize_per_tensor_default_13" [label="(197, 1, 768)", style=solid]; +"254 dequantize_per_tensor_default_13" -> "259 linear_8" [label="(197, 1, 768)", style=solid]; +"255 linear_8_scale_0" -> "258 dequantize_per_channel_default_9" [label="(2304,)", style=solid]; +"256 linear_8_zero_point_0" -> "258 dequantize_per_channel_default_9" [label="(2304,)", style=solid]; +"257 _frozen_param9" -> "258 dequantize_per_channel_default_9" [label="(2304, 768)", style=solid]; +"258 dequantize_per_channel_default_9" -> "259 linear_8" [label="(2304, 768)", style=solid]; +"259 linear_8" -> "260 unflatten_2" [label="(197, 1, 2304)", style=solid]; +"260 unflatten_2" -> "261 unsqueeze_2" [label="(197, 1, 3, 768)", style=solid]; +"261 unsqueeze_2" -> "262 transpose_13" [label="(1, 197, 1, 3, 768)", style=solid]; +"262 transpose_13" -> "263 squeeze_2" [label="(3, 197, 1, 1, 768)", style=solid]; +"263 squeeze_2" -> "264 contiguous_2" [label="(3, 197, 1, 768)", style=solid]; +"264 contiguous_2" -> "265 quantize_per_tensor_default_14" [label="(3, 197, 1, 768)", style=solid]; +"264 contiguous_2" -> "268 quantize_per_tensor_default_15" [label="(3, 197, 1, 768)", style=solid]; +"264 contiguous_2" -> "271 select_8" [label="(3, 197, 1, 768)", style=solid]; +"265 quantize_per_tensor_default_14" -> "266 dequantize_per_tensor_default_14" [label="(3, 197, 1, 768)", style=solid]; +"266 dequantize_per_tensor_default_14" -> "267 select_6" [label="(3, 197, 1, 768)", style=solid]; +"267 select_6" -> "272 view_16" [label="(197, 1, 768)", style=solid]; +"268 quantize_per_tensor_default_15" -> "269 dequantize_per_tensor_default_15" [label="(3, 197, 1, 768)", style=solid]; +"269 dequantize_per_tensor_default_15" -> "270 select_7" [label="(3, 197, 1, 768)", style=solid]; +"270 select_7" -> "274 view_17" [label="(197, 1, 768)", style=solid]; +"271 select_8" -> "276 view_18" [label="(197, 1, 768)", style=solid]; +"272 view_16" -> "273 transpose_14" [label="(197, 12, 64)", style=solid]; +"273 transpose_14" -> "278 view_19" [label="(12, 197, 64)", style=solid]; +"274 view_17" -> "275 transpose_15" [label="(197, 12, 64)", style=solid]; +"275 transpose_15" -> "279 view_20" [label="(12, 197, 64)", style=solid]; +"276 view_18" -> "277 transpose_16" [label="(197, 12, 64)", style=solid]; +"277 transpose_16" -> "280 view_21" [label="(12, 197, 64)", style=solid]; +"278 view_19" -> "281 scaled_dot_product_attention_2" [label="(1, 12, 197, 64)", style=solid]; +"279 view_20" -> "281 scaled_dot_product_attention_2" [label="(1, 12, 197, 64)", style=solid]; +"280 view_21" -> "281 scaled_dot_product_attention_2" [label="(1, 12, 197, 64)", style=solid]; +"281 scaled_dot_product_attention_2" -> "282 permute_3" [label="(1, 12, 197, 64)", style=solid]; +"282 permute_3" -> "283 view_22" [label="(197, 1, 12, 64)", style=solid]; +"283 view_22" -> "284 view_22_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"284 view_22_0_0_nncf_smooth_quant_0" -> "285 quantize_per_tensor_default_16" [label="(197, 768)", style=solid]; +"285 quantize_per_tensor_default_16" -> "286 dequantize_per_tensor_default_16" [label="(197, 768)", style=solid]; +"286 dequantize_per_tensor_default_16" -> "291 linear_9" [label="(197, 768)", style=solid]; +"287 linear_9_scale_0" -> "290 dequantize_per_channel_default_10" [label="(768,)", style=solid]; +"288 linear_9_zero_point_0" -> "290 dequantize_per_channel_default_10" [label="(768,)", style=solid]; +"289 _frozen_param10" -> "290 dequantize_per_channel_default_10" [label="(768, 768)", style=solid]; +"290 dequantize_per_channel_default_10" -> "291 linear_9" [label="(768, 768)", style=solid]; +"291 linear_9" -> "292 view_23" [label="(197, 768)", style=solid]; +"292 view_23" -> "293 transpose_17" [label="(197, 1, 768)", style=solid]; +"293 transpose_17" -> "294 dropout_7" [label="(1, 197, 768)", style=solid]; +"294 dropout_7" -> "295 add_5" [label="(1, 197, 768)", style=solid]; +"295 add_5" -> "296 layer_norm_5" [label="(1, 197, 768)", style=solid]; +"295 add_5" -> "316 add_6" [label="(1, 197, 768)", style=solid]; +"296 layer_norm_5" -> "297 layer_norm_5_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"297 layer_norm_5_0_0_nncf_smooth_quant_0" -> "298 quantize_per_tensor_default_17" [label="(1, 197, 768)", style=solid]; +"298 quantize_per_tensor_default_17" -> "299 dequantize_per_tensor_default_17" [label="(1, 197, 768)", style=solid]; +"299 dequantize_per_tensor_default_17" -> "304 linear_10" [label="(1, 197, 768)", style=solid]; +"300 linear_10_scale_0" -> "303 dequantize_per_channel_default_11" [label="(3072,)", style=solid]; +"301 linear_10_zero_point_0" -> "303 dequantize_per_channel_default_11" [label="(3072,)", style=solid]; +"302 _frozen_param11" -> "303 dequantize_per_channel_default_11" [label="(3072, 768)", style=solid]; +"303 dequantize_per_channel_default_11" -> "304 linear_10" [label="(3072, 768)", style=solid]; +"304 linear_10" -> "305 gelu_2" [label="(1, 197, 3072)", style=solid]; +"305 gelu_2" -> "306 dropout_8" [label="(1, 197, 3072)", style=solid]; +"306 dropout_8" -> "307 dropout_8_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"307 dropout_8_0_0_nncf_smooth_quant_0" -> "308 quantize_per_tensor_default_18" [label="(1, 197, 3072)", style=solid]; +"308 quantize_per_tensor_default_18" -> "309 dequantize_per_tensor_default_18" [label="(1, 197, 3072)", style=solid]; +"309 dequantize_per_tensor_default_18" -> "314 linear_11" [label="(1, 197, 3072)", style=solid]; +"310 linear_11_scale_0" -> "313 dequantize_per_channel_default_12" [label="(768,)", style=solid]; +"311 linear_11_zero_point_0" -> "313 dequantize_per_channel_default_12" [label="(768,)", style=solid]; +"312 _frozen_param12" -> "313 dequantize_per_channel_default_12" [label="(768, 3072)", style=solid]; +"313 dequantize_per_channel_default_12" -> "314 linear_11" [label="(768, 3072)", style=solid]; +"314 linear_11" -> "315 dropout_9" [label="(1, 197, 768)", style=solid]; +"315 dropout_9" -> "316 add_6" [label="(1, 197, 768)", style=solid]; +"316 add_6" -> "317 layer_norm_6" [label="(1, 197, 768)", style=solid]; +"316 add_6" -> "362 add_7" [label="(1, 197, 768)", style=solid]; +"317 layer_norm_6" -> "318 transpose_18" [label="(1, 197, 768)", style=solid]; +"318 transpose_18" -> "319 transpose_18_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"319 transpose_18_0_0_nncf_smooth_quant_0" -> "320 quantize_per_tensor_default_19" [label="(197, 1, 768)", style=solid]; +"320 quantize_per_tensor_default_19" -> "321 dequantize_per_tensor_default_19" [label="(197, 1, 768)", style=solid]; +"321 dequantize_per_tensor_default_19" -> "326 linear_12" [label="(197, 1, 768)", style=solid]; +"322 linear_12_scale_0" -> "325 dequantize_per_channel_default_13" [label="(2304,)", style=solid]; +"323 linear_12_zero_point_0" -> "325 dequantize_per_channel_default_13" [label="(2304,)", style=solid]; +"324 _frozen_param13" -> "325 dequantize_per_channel_default_13" [label="(2304, 768)", style=solid]; +"325 dequantize_per_channel_default_13" -> "326 linear_12" [label="(2304, 768)", style=solid]; +"326 linear_12" -> "327 unflatten_3" [label="(197, 1, 2304)", style=solid]; +"327 unflatten_3" -> "328 unsqueeze_3" [label="(197, 1, 3, 768)", style=solid]; +"328 unsqueeze_3" -> "329 transpose_19" [label="(1, 197, 1, 3, 768)", style=solid]; +"329 transpose_19" -> "330 squeeze_3" [label="(3, 197, 1, 1, 768)", style=solid]; +"330 squeeze_3" -> "331 contiguous_3" [label="(3, 197, 1, 768)", style=solid]; +"331 contiguous_3" -> "332 quantize_per_tensor_default_20" [label="(3, 197, 1, 768)", style=solid]; +"331 contiguous_3" -> "335 quantize_per_tensor_default_21" [label="(3, 197, 1, 768)", style=solid]; +"331 contiguous_3" -> "338 select_11" [label="(3, 197, 1, 768)", style=solid]; +"332 quantize_per_tensor_default_20" -> "333 dequantize_per_tensor_default_20" [label="(3, 197, 1, 768)", style=solid]; +"333 dequantize_per_tensor_default_20" -> "334 select_9" [label="(3, 197, 1, 768)", style=solid]; +"334 select_9" -> "339 view_24" [label="(197, 1, 768)", style=solid]; +"335 quantize_per_tensor_default_21" -> "336 dequantize_per_tensor_default_21" [label="(3, 197, 1, 768)", style=solid]; +"336 dequantize_per_tensor_default_21" -> "337 select_10" [label="(3, 197, 1, 768)", style=solid]; +"337 select_10" -> "341 view_25" [label="(197, 1, 768)", style=solid]; +"338 select_11" -> "343 view_26" [label="(197, 1, 768)", style=solid]; +"339 view_24" -> "340 transpose_20" [label="(197, 12, 64)", style=solid]; +"340 transpose_20" -> "345 view_27" [label="(12, 197, 64)", style=solid]; +"341 view_25" -> "342 transpose_21" [label="(197, 12, 64)", style=solid]; +"342 transpose_21" -> "346 view_28" [label="(12, 197, 64)", style=solid]; +"343 view_26" -> "344 transpose_22" [label="(197, 12, 64)", style=solid]; +"344 transpose_22" -> "347 view_29" [label="(12, 197, 64)", style=solid]; +"345 view_27" -> "348 scaled_dot_product_attention_3" [label="(1, 12, 197, 64)", style=solid]; +"346 view_28" -> "348 scaled_dot_product_attention_3" [label="(1, 12, 197, 64)", style=solid]; +"347 view_29" -> "348 scaled_dot_product_attention_3" [label="(1, 12, 197, 64)", style=solid]; +"348 scaled_dot_product_attention_3" -> "349 permute_4" [label="(1, 12, 197, 64)", style=solid]; +"349 permute_4" -> "350 view_30" [label="(197, 1, 12, 64)", style=solid]; +"350 view_30" -> "351 view_30_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"351 view_30_0_0_nncf_smooth_quant_0" -> "352 quantize_per_tensor_default_22" [label="(197, 768)", style=solid]; +"352 quantize_per_tensor_default_22" -> "353 dequantize_per_tensor_default_22" [label="(197, 768)", style=solid]; +"353 dequantize_per_tensor_default_22" -> "358 linear_13" [label="(197, 768)", style=solid]; +"354 linear_13_scale_0" -> "357 dequantize_per_channel_default_14" [label="(768,)", style=solid]; +"355 linear_13_zero_point_0" -> "357 dequantize_per_channel_default_14" [label="(768,)", style=solid]; +"356 _frozen_param14" -> "357 dequantize_per_channel_default_14" [label="(768, 768)", style=solid]; +"357 dequantize_per_channel_default_14" -> "358 linear_13" [label="(768, 768)", style=solid]; +"358 linear_13" -> "359 view_31" [label="(197, 768)", style=solid]; +"359 view_31" -> "360 transpose_23" [label="(197, 1, 768)", style=solid]; +"360 transpose_23" -> "361 dropout_10" [label="(1, 197, 768)", style=solid]; +"361 dropout_10" -> "362 add_7" [label="(1, 197, 768)", style=solid]; +"362 add_7" -> "363 layer_norm_7" [label="(1, 197, 768)", style=solid]; +"362 add_7" -> "383 add_8" [label="(1, 197, 768)", style=solid]; +"363 layer_norm_7" -> "364 layer_norm_7_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"364 layer_norm_7_0_0_nncf_smooth_quant_0" -> "365 quantize_per_tensor_default_23" [label="(1, 197, 768)", style=solid]; +"365 quantize_per_tensor_default_23" -> "366 dequantize_per_tensor_default_23" [label="(1, 197, 768)", style=solid]; +"366 dequantize_per_tensor_default_23" -> "371 linear_14" [label="(1, 197, 768)", style=solid]; +"367 linear_14_scale_0" -> "370 dequantize_per_channel_default_15" [label="(3072,)", style=solid]; +"368 linear_14_zero_point_0" -> "370 dequantize_per_channel_default_15" [label="(3072,)", style=solid]; +"369 _frozen_param15" -> "370 dequantize_per_channel_default_15" [label="(3072, 768)", style=solid]; +"370 dequantize_per_channel_default_15" -> "371 linear_14" [label="(3072, 768)", style=solid]; +"371 linear_14" -> "372 gelu_3" [label="(1, 197, 3072)", style=solid]; +"372 gelu_3" -> "373 dropout_11" [label="(1, 197, 3072)", style=solid]; +"373 dropout_11" -> "374 dropout_11_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"374 dropout_11_0_0_nncf_smooth_quant_0" -> "375 quantize_per_tensor_default_24" [label="(1, 197, 3072)", style=solid]; +"375 quantize_per_tensor_default_24" -> "376 dequantize_per_tensor_default_24" [label="(1, 197, 3072)", style=solid]; +"376 dequantize_per_tensor_default_24" -> "381 linear_15" [label="(1, 197, 3072)", style=solid]; +"377 linear_15_scale_0" -> "380 dequantize_per_channel_default_16" [label="(768,)", style=solid]; +"378 linear_15_zero_point_0" -> "380 dequantize_per_channel_default_16" [label="(768,)", style=solid]; +"379 _frozen_param16" -> "380 dequantize_per_channel_default_16" [label="(768, 3072)", style=solid]; +"380 dequantize_per_channel_default_16" -> "381 linear_15" [label="(768, 3072)", style=solid]; +"381 linear_15" -> "382 dropout_12" [label="(1, 197, 768)", style=solid]; +"382 dropout_12" -> "383 add_8" [label="(1, 197, 768)", style=solid]; +"383 add_8" -> "384 layer_norm_8" [label="(1, 197, 768)", style=solid]; +"383 add_8" -> "429 add_9" [label="(1, 197, 768)", style=solid]; +"384 layer_norm_8" -> "385 transpose_24" [label="(1, 197, 768)", style=solid]; +"385 transpose_24" -> "386 transpose_24_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"386 transpose_24_0_0_nncf_smooth_quant_0" -> "387 quantize_per_tensor_default_25" [label="(197, 1, 768)", style=solid]; +"387 quantize_per_tensor_default_25" -> "388 dequantize_per_tensor_default_25" [label="(197, 1, 768)", style=solid]; +"388 dequantize_per_tensor_default_25" -> "393 linear_16" [label="(197, 1, 768)", style=solid]; +"389 linear_16_scale_0" -> "392 dequantize_per_channel_default_17" [label="(2304,)", style=solid]; +"390 linear_16_zero_point_0" -> "392 dequantize_per_channel_default_17" [label="(2304,)", style=solid]; +"391 _frozen_param17" -> "392 dequantize_per_channel_default_17" [label="(2304, 768)", style=solid]; +"392 dequantize_per_channel_default_17" -> "393 linear_16" [label="(2304, 768)", style=solid]; +"393 linear_16" -> "394 unflatten_4" [label="(197, 1, 2304)", style=solid]; +"394 unflatten_4" -> "395 unsqueeze_4" [label="(197, 1, 3, 768)", style=solid]; +"395 unsqueeze_4" -> "396 transpose_25" [label="(1, 197, 1, 3, 768)", style=solid]; +"396 transpose_25" -> "397 squeeze_4" [label="(3, 197, 1, 1, 768)", style=solid]; +"397 squeeze_4" -> "398 contiguous_4" [label="(3, 197, 1, 768)", style=solid]; +"398 contiguous_4" -> "399 quantize_per_tensor_default_26" [label="(3, 197, 1, 768)", style=solid]; +"398 contiguous_4" -> "402 quantize_per_tensor_default_27" [label="(3, 197, 1, 768)", style=solid]; +"398 contiguous_4" -> "405 select_14" [label="(3, 197, 1, 768)", style=solid]; +"399 quantize_per_tensor_default_26" -> "400 dequantize_per_tensor_default_26" [label="(3, 197, 1, 768)", style=solid]; +"400 dequantize_per_tensor_default_26" -> "401 select_12" [label="(3, 197, 1, 768)", style=solid]; +"401 select_12" -> "406 view_32" [label="(197, 1, 768)", style=solid]; +"402 quantize_per_tensor_default_27" -> "403 dequantize_per_tensor_default_27" [label="(3, 197, 1, 768)", style=solid]; +"403 dequantize_per_tensor_default_27" -> "404 select_13" [label="(3, 197, 1, 768)", style=solid]; +"404 select_13" -> "408 view_33" [label="(197, 1, 768)", style=solid]; +"405 select_14" -> "410 view_34" [label="(197, 1, 768)", style=solid]; +"406 view_32" -> "407 transpose_26" [label="(197, 12, 64)", style=solid]; +"407 transpose_26" -> "412 view_35" [label="(12, 197, 64)", style=solid]; +"408 view_33" -> "409 transpose_27" [label="(197, 12, 64)", style=solid]; +"409 transpose_27" -> "413 view_36" [label="(12, 197, 64)", style=solid]; +"410 view_34" -> "411 transpose_28" [label="(197, 12, 64)", style=solid]; +"411 transpose_28" -> "414 view_37" [label="(12, 197, 64)", style=solid]; +"412 view_35" -> "415 scaled_dot_product_attention_4" [label="(1, 12, 197, 64)", style=solid]; +"413 view_36" -> "415 scaled_dot_product_attention_4" [label="(1, 12, 197, 64)", style=solid]; +"414 view_37" -> "415 scaled_dot_product_attention_4" [label="(1, 12, 197, 64)", style=solid]; +"415 scaled_dot_product_attention_4" -> "416 permute_5" [label="(1, 12, 197, 64)", style=solid]; +"416 permute_5" -> "417 view_38" [label="(197, 1, 12, 64)", style=solid]; +"417 view_38" -> "418 view_38_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"418 view_38_0_0_nncf_smooth_quant_0" -> "419 quantize_per_tensor_default_28" [label="(197, 768)", style=solid]; +"419 quantize_per_tensor_default_28" -> "420 dequantize_per_tensor_default_28" [label="(197, 768)", style=solid]; +"420 dequantize_per_tensor_default_28" -> "425 linear_17" [label="(197, 768)", style=solid]; +"421 linear_17_scale_0" -> "424 dequantize_per_channel_default_18" [label="(768,)", style=solid]; +"422 linear_17_zero_point_0" -> "424 dequantize_per_channel_default_18" [label="(768,)", style=solid]; +"423 _frozen_param18" -> "424 dequantize_per_channel_default_18" [label="(768, 768)", style=solid]; +"424 dequantize_per_channel_default_18" -> "425 linear_17" [label="(768, 768)", style=solid]; +"425 linear_17" -> "426 view_39" [label="(197, 768)", style=solid]; +"426 view_39" -> "427 transpose_29" [label="(197, 1, 768)", style=solid]; +"427 transpose_29" -> "428 dropout_13" [label="(1, 197, 768)", style=solid]; +"428 dropout_13" -> "429 add_9" [label="(1, 197, 768)", style=solid]; +"429 add_9" -> "430 layer_norm_9" [label="(1, 197, 768)", style=solid]; +"429 add_9" -> "450 add_10" [label="(1, 197, 768)", style=solid]; +"430 layer_norm_9" -> "431 layer_norm_9_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"431 layer_norm_9_0_0_nncf_smooth_quant_0" -> "432 quantize_per_tensor_default_29" [label="(1, 197, 768)", style=solid]; +"432 quantize_per_tensor_default_29" -> "433 dequantize_per_tensor_default_29" [label="(1, 197, 768)", style=solid]; +"433 dequantize_per_tensor_default_29" -> "438 linear_18" [label="(1, 197, 768)", style=solid]; +"434 linear_18_scale_0" -> "437 dequantize_per_channel_default_19" [label="(3072,)", style=solid]; +"435 linear_18_zero_point_0" -> "437 dequantize_per_channel_default_19" [label="(3072,)", style=solid]; +"436 _frozen_param19" -> "437 dequantize_per_channel_default_19" [label="(3072, 768)", style=solid]; +"437 dequantize_per_channel_default_19" -> "438 linear_18" [label="(3072, 768)", style=solid]; +"438 linear_18" -> "439 gelu_4" [label="(1, 197, 3072)", style=solid]; +"439 gelu_4" -> "440 dropout_14" [label="(1, 197, 3072)", style=solid]; +"440 dropout_14" -> "441 dropout_14_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"441 dropout_14_0_0_nncf_smooth_quant_0" -> "442 quantize_per_tensor_default_30" [label="(1, 197, 3072)", style=solid]; +"442 quantize_per_tensor_default_30" -> "443 dequantize_per_tensor_default_30" [label="(1, 197, 3072)", style=solid]; +"443 dequantize_per_tensor_default_30" -> "448 linear_19" [label="(1, 197, 3072)", style=solid]; +"444 linear_19_scale_0" -> "447 dequantize_per_channel_default_20" [label="(768,)", style=solid]; +"445 linear_19_zero_point_0" -> "447 dequantize_per_channel_default_20" [label="(768,)", style=solid]; +"446 _frozen_param20" -> "447 dequantize_per_channel_default_20" [label="(768, 3072)", style=solid]; +"447 dequantize_per_channel_default_20" -> "448 linear_19" [label="(768, 3072)", style=solid]; +"448 linear_19" -> "449 dropout_15" [label="(1, 197, 768)", style=solid]; +"449 dropout_15" -> "450 add_10" [label="(1, 197, 768)", style=solid]; +"450 add_10" -> "451 layer_norm_10" [label="(1, 197, 768)", style=solid]; +"450 add_10" -> "496 add_11" [label="(1, 197, 768)", style=solid]; +"451 layer_norm_10" -> "452 transpose_30" [label="(1, 197, 768)", style=solid]; +"452 transpose_30" -> "453 transpose_30_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"453 transpose_30_0_0_nncf_smooth_quant_0" -> "454 quantize_per_tensor_default_31" [label="(197, 1, 768)", style=solid]; +"454 quantize_per_tensor_default_31" -> "455 dequantize_per_tensor_default_31" [label="(197, 1, 768)", style=solid]; +"455 dequantize_per_tensor_default_31" -> "460 linear_20" [label="(197, 1, 768)", style=solid]; +"456 linear_20_scale_0" -> "459 dequantize_per_channel_default_21" [label="(2304,)", style=solid]; +"457 linear_20_zero_point_0" -> "459 dequantize_per_channel_default_21" [label="(2304,)", style=solid]; +"458 _frozen_param21" -> "459 dequantize_per_channel_default_21" [label="(2304, 768)", style=solid]; +"459 dequantize_per_channel_default_21" -> "460 linear_20" [label="(2304, 768)", style=solid]; +"460 linear_20" -> "461 unflatten_5" [label="(197, 1, 2304)", style=solid]; +"461 unflatten_5" -> "462 unsqueeze_5" [label="(197, 1, 3, 768)", style=solid]; +"462 unsqueeze_5" -> "463 transpose_31" [label="(1, 197, 1, 3, 768)", style=solid]; +"463 transpose_31" -> "464 squeeze_5" [label="(3, 197, 1, 1, 768)", style=solid]; +"464 squeeze_5" -> "465 contiguous_5" [label="(3, 197, 1, 768)", style=solid]; +"465 contiguous_5" -> "466 quantize_per_tensor_default_32" [label="(3, 197, 1, 768)", style=solid]; +"465 contiguous_5" -> "469 quantize_per_tensor_default_33" [label="(3, 197, 1, 768)", style=solid]; +"465 contiguous_5" -> "472 select_17" [label="(3, 197, 1, 768)", style=solid]; +"466 quantize_per_tensor_default_32" -> "467 dequantize_per_tensor_default_32" [label="(3, 197, 1, 768)", style=solid]; +"467 dequantize_per_tensor_default_32" -> "468 select_15" [label="(3, 197, 1, 768)", style=solid]; +"468 select_15" -> "473 view_40" [label="(197, 1, 768)", style=solid]; +"469 quantize_per_tensor_default_33" -> "470 dequantize_per_tensor_default_33" [label="(3, 197, 1, 768)", style=solid]; +"470 dequantize_per_tensor_default_33" -> "471 select_16" [label="(3, 197, 1, 768)", style=solid]; +"471 select_16" -> "475 view_41" [label="(197, 1, 768)", style=solid]; +"472 select_17" -> "477 view_42" [label="(197, 1, 768)", style=solid]; +"473 view_40" -> "474 transpose_32" [label="(197, 12, 64)", style=solid]; +"474 transpose_32" -> "479 view_43" [label="(12, 197, 64)", style=solid]; +"475 view_41" -> "476 transpose_33" [label="(197, 12, 64)", style=solid]; +"476 transpose_33" -> "480 view_44" [label="(12, 197, 64)", style=solid]; +"477 view_42" -> "478 transpose_34" [label="(197, 12, 64)", style=solid]; +"478 transpose_34" -> "481 view_45" [label="(12, 197, 64)", style=solid]; +"479 view_43" -> "482 scaled_dot_product_attention_5" [label="(1, 12, 197, 64)", style=solid]; +"480 view_44" -> "482 scaled_dot_product_attention_5" [label="(1, 12, 197, 64)", style=solid]; +"481 view_45" -> "482 scaled_dot_product_attention_5" [label="(1, 12, 197, 64)", style=solid]; +"482 scaled_dot_product_attention_5" -> "483 permute_6" [label="(1, 12, 197, 64)", style=solid]; +"483 permute_6" -> "484 view_46" [label="(197, 1, 12, 64)", style=solid]; +"484 view_46" -> "485 view_46_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"485 view_46_0_0_nncf_smooth_quant_0" -> "486 quantize_per_tensor_default_34" [label="(197, 768)", style=solid]; +"486 quantize_per_tensor_default_34" -> "487 dequantize_per_tensor_default_34" [label="(197, 768)", style=solid]; +"487 dequantize_per_tensor_default_34" -> "492 linear_21" [label="(197, 768)", style=solid]; +"488 linear_21_scale_0" -> "491 dequantize_per_channel_default_22" [label="(768,)", style=solid]; +"489 linear_21_zero_point_0" -> "491 dequantize_per_channel_default_22" [label="(768,)", style=solid]; +"490 _frozen_param22" -> "491 dequantize_per_channel_default_22" [label="(768, 768)", style=solid]; +"491 dequantize_per_channel_default_22" -> "492 linear_21" [label="(768, 768)", style=solid]; +"492 linear_21" -> "493 view_47" [label="(197, 768)", style=solid]; +"493 view_47" -> "494 transpose_35" [label="(197, 1, 768)", style=solid]; +"494 transpose_35" -> "495 dropout_16" [label="(1, 197, 768)", style=solid]; +"495 dropout_16" -> "496 add_11" [label="(1, 197, 768)", style=solid]; +"496 add_11" -> "497 layer_norm_11" [label="(1, 197, 768)", style=solid]; +"496 add_11" -> "517 add_12" [label="(1, 197, 768)", style=solid]; +"497 layer_norm_11" -> "498 layer_norm_11_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"498 layer_norm_11_0_0_nncf_smooth_quant_0" -> "499 quantize_per_tensor_default_35" [label="(1, 197, 768)", style=solid]; +"499 quantize_per_tensor_default_35" -> "500 dequantize_per_tensor_default_35" [label="(1, 197, 768)", style=solid]; +"500 dequantize_per_tensor_default_35" -> "505 linear_22" [label="(1, 197, 768)", style=solid]; +"501 linear_22_scale_0" -> "504 dequantize_per_channel_default_23" [label="(3072,)", style=solid]; +"502 linear_22_zero_point_0" -> "504 dequantize_per_channel_default_23" [label="(3072,)", style=solid]; +"503 _frozen_param23" -> "504 dequantize_per_channel_default_23" [label="(3072, 768)", style=solid]; +"504 dequantize_per_channel_default_23" -> "505 linear_22" [label="(3072, 768)", style=solid]; +"505 linear_22" -> "506 gelu_5" [label="(1, 197, 3072)", style=solid]; +"506 gelu_5" -> "507 dropout_17" [label="(1, 197, 3072)", style=solid]; +"507 dropout_17" -> "508 dropout_17_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"508 dropout_17_0_0_nncf_smooth_quant_0" -> "509 quantize_per_tensor_default_36" [label="(1, 197, 3072)", style=solid]; +"509 quantize_per_tensor_default_36" -> "510 dequantize_per_tensor_default_36" [label="(1, 197, 3072)", style=solid]; +"510 dequantize_per_tensor_default_36" -> "515 linear_23" [label="(1, 197, 3072)", style=solid]; +"511 linear_23_scale_0" -> "514 dequantize_per_channel_default_24" [label="(768,)", style=solid]; +"512 linear_23_zero_point_0" -> "514 dequantize_per_channel_default_24" [label="(768,)", style=solid]; +"513 _frozen_param24" -> "514 dequantize_per_channel_default_24" [label="(768, 3072)", style=solid]; +"514 dequantize_per_channel_default_24" -> "515 linear_23" [label="(768, 3072)", style=solid]; +"515 linear_23" -> "516 dropout_18" [label="(1, 197, 768)", style=solid]; +"516 dropout_18" -> "517 add_12" [label="(1, 197, 768)", style=solid]; +"517 add_12" -> "518 layer_norm_12" [label="(1, 197, 768)", style=solid]; +"517 add_12" -> "563 add_13" [label="(1, 197, 768)", style=solid]; +"518 layer_norm_12" -> "519 transpose_36" [label="(1, 197, 768)", style=solid]; +"519 transpose_36" -> "520 transpose_36_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"520 transpose_36_0_0_nncf_smooth_quant_0" -> "521 quantize_per_tensor_default_37" [label="(197, 1, 768)", style=solid]; +"521 quantize_per_tensor_default_37" -> "522 dequantize_per_tensor_default_37" [label="(197, 1, 768)", style=solid]; +"522 dequantize_per_tensor_default_37" -> "527 linear_24" [label="(197, 1, 768)", style=solid]; +"523 linear_24_scale_0" -> "526 dequantize_per_channel_default_25" [label="(2304,)", style=solid]; +"524 linear_24_zero_point_0" -> "526 dequantize_per_channel_default_25" [label="(2304,)", style=solid]; +"525 _frozen_param25" -> "526 dequantize_per_channel_default_25" [label="(2304, 768)", style=solid]; +"526 dequantize_per_channel_default_25" -> "527 linear_24" [label="(2304, 768)", style=solid]; +"527 linear_24" -> "528 unflatten_6" [label="(197, 1, 2304)", style=solid]; +"528 unflatten_6" -> "529 unsqueeze_6" [label="(197, 1, 3, 768)", style=solid]; +"529 unsqueeze_6" -> "530 transpose_37" [label="(1, 197, 1, 3, 768)", style=solid]; +"530 transpose_37" -> "531 squeeze_6" [label="(3, 197, 1, 1, 768)", style=solid]; +"531 squeeze_6" -> "532 contiguous_6" [label="(3, 197, 1, 768)", style=solid]; +"532 contiguous_6" -> "533 quantize_per_tensor_default_38" [label="(3, 197, 1, 768)", style=solid]; +"532 contiguous_6" -> "536 quantize_per_tensor_default_39" [label="(3, 197, 1, 768)", style=solid]; +"532 contiguous_6" -> "539 select_20" [label="(3, 197, 1, 768)", style=solid]; +"533 quantize_per_tensor_default_38" -> "534 dequantize_per_tensor_default_38" [label="(3, 197, 1, 768)", style=solid]; +"534 dequantize_per_tensor_default_38" -> "535 select_18" [label="(3, 197, 1, 768)", style=solid]; +"535 select_18" -> "540 view_48" [label="(197, 1, 768)", style=solid]; +"536 quantize_per_tensor_default_39" -> "537 dequantize_per_tensor_default_39" [label="(3, 197, 1, 768)", style=solid]; +"537 dequantize_per_tensor_default_39" -> "538 select_19" [label="(3, 197, 1, 768)", style=solid]; +"538 select_19" -> "542 view_49" [label="(197, 1, 768)", style=solid]; +"539 select_20" -> "544 view_50" [label="(197, 1, 768)", style=solid]; +"540 view_48" -> "541 transpose_38" [label="(197, 12, 64)", style=solid]; +"541 transpose_38" -> "546 view_51" [label="(12, 197, 64)", style=solid]; +"542 view_49" -> "543 transpose_39" [label="(197, 12, 64)", style=solid]; +"543 transpose_39" -> "547 view_52" [label="(12, 197, 64)", style=solid]; +"544 view_50" -> "545 transpose_40" [label="(197, 12, 64)", style=solid]; +"545 transpose_40" -> "548 view_53" [label="(12, 197, 64)", style=solid]; +"546 view_51" -> "549 scaled_dot_product_attention_6" [label="(1, 12, 197, 64)", style=solid]; +"547 view_52" -> "549 scaled_dot_product_attention_6" [label="(1, 12, 197, 64)", style=solid]; +"548 view_53" -> "549 scaled_dot_product_attention_6" [label="(1, 12, 197, 64)", style=solid]; +"549 scaled_dot_product_attention_6" -> "550 permute_7" [label="(1, 12, 197, 64)", style=solid]; +"550 permute_7" -> "551 view_54" [label="(197, 1, 12, 64)", style=solid]; +"551 view_54" -> "552 view_54_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"552 view_54_0_0_nncf_smooth_quant_0" -> "553 quantize_per_tensor_default_40" [label="(197, 768)", style=solid]; +"553 quantize_per_tensor_default_40" -> "554 dequantize_per_tensor_default_40" [label="(197, 768)", style=solid]; +"554 dequantize_per_tensor_default_40" -> "559 linear_25" [label="(197, 768)", style=solid]; +"555 linear_25_scale_0" -> "558 dequantize_per_channel_default_26" [label="(768,)", style=solid]; +"556 linear_25_zero_point_0" -> "558 dequantize_per_channel_default_26" [label="(768,)", style=solid]; +"557 _frozen_param26" -> "558 dequantize_per_channel_default_26" [label="(768, 768)", style=solid]; +"558 dequantize_per_channel_default_26" -> "559 linear_25" [label="(768, 768)", style=solid]; +"559 linear_25" -> "560 view_55" [label="(197, 768)", style=solid]; +"560 view_55" -> "561 transpose_41" [label="(197, 1, 768)", style=solid]; +"561 transpose_41" -> "562 dropout_19" [label="(1, 197, 768)", style=solid]; +"562 dropout_19" -> "563 add_13" [label="(1, 197, 768)", style=solid]; +"563 add_13" -> "564 layer_norm_13" [label="(1, 197, 768)", style=solid]; +"563 add_13" -> "584 add_14" [label="(1, 197, 768)", style=solid]; +"564 layer_norm_13" -> "565 layer_norm_13_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"565 layer_norm_13_0_0_nncf_smooth_quant_0" -> "566 quantize_per_tensor_default_41" [label="(1, 197, 768)", style=solid]; +"566 quantize_per_tensor_default_41" -> "567 dequantize_per_tensor_default_41" [label="(1, 197, 768)", style=solid]; +"567 dequantize_per_tensor_default_41" -> "572 linear_26" [label="(1, 197, 768)", style=solid]; +"568 linear_26_scale_0" -> "571 dequantize_per_channel_default_27" [label="(3072,)", style=solid]; +"569 linear_26_zero_point_0" -> "571 dequantize_per_channel_default_27" [label="(3072,)", style=solid]; +"570 _frozen_param27" -> "571 dequantize_per_channel_default_27" [label="(3072, 768)", style=solid]; +"571 dequantize_per_channel_default_27" -> "572 linear_26" [label="(3072, 768)", style=solid]; +"572 linear_26" -> "573 gelu_6" [label="(1, 197, 3072)", style=solid]; +"573 gelu_6" -> "574 dropout_20" [label="(1, 197, 3072)", style=solid]; +"574 dropout_20" -> "575 dropout_20_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"575 dropout_20_0_0_nncf_smooth_quant_0" -> "576 quantize_per_tensor_default_42" [label="(1, 197, 3072)", style=solid]; +"576 quantize_per_tensor_default_42" -> "577 dequantize_per_tensor_default_42" [label="(1, 197, 3072)", style=solid]; +"577 dequantize_per_tensor_default_42" -> "582 linear_27" [label="(1, 197, 3072)", style=solid]; +"578 linear_27_scale_0" -> "581 dequantize_per_channel_default_28" [label="(768,)", style=solid]; +"579 linear_27_zero_point_0" -> "581 dequantize_per_channel_default_28" [label="(768,)", style=solid]; +"580 _frozen_param28" -> "581 dequantize_per_channel_default_28" [label="(768, 3072)", style=solid]; +"581 dequantize_per_channel_default_28" -> "582 linear_27" [label="(768, 3072)", style=solid]; +"582 linear_27" -> "583 dropout_21" [label="(1, 197, 768)", style=solid]; +"583 dropout_21" -> "584 add_14" [label="(1, 197, 768)", style=solid]; +"584 add_14" -> "585 layer_norm_14" [label="(1, 197, 768)", style=solid]; +"584 add_14" -> "630 add_15" [label="(1, 197, 768)", style=solid]; +"585 layer_norm_14" -> "586 transpose_42" [label="(1, 197, 768)", style=solid]; +"586 transpose_42" -> "587 transpose_42_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"587 transpose_42_0_0_nncf_smooth_quant_0" -> "588 quantize_per_tensor_default_43" [label="(197, 1, 768)", style=solid]; +"588 quantize_per_tensor_default_43" -> "589 dequantize_per_tensor_default_43" [label="(197, 1, 768)", style=solid]; +"589 dequantize_per_tensor_default_43" -> "594 linear_28" [label="(197, 1, 768)", style=solid]; +"590 linear_28_scale_0" -> "593 dequantize_per_channel_default_29" [label="(2304,)", style=solid]; +"591 linear_28_zero_point_0" -> "593 dequantize_per_channel_default_29" [label="(2304,)", style=solid]; +"592 _frozen_param29" -> "593 dequantize_per_channel_default_29" [label="(2304, 768)", style=solid]; +"593 dequantize_per_channel_default_29" -> "594 linear_28" [label="(2304, 768)", style=solid]; +"594 linear_28" -> "595 unflatten_7" [label="(197, 1, 2304)", style=solid]; +"595 unflatten_7" -> "596 unsqueeze_7" [label="(197, 1, 3, 768)", style=solid]; +"596 unsqueeze_7" -> "597 transpose_43" [label="(1, 197, 1, 3, 768)", style=solid]; +"597 transpose_43" -> "598 squeeze_7" [label="(3, 197, 1, 1, 768)", style=solid]; +"598 squeeze_7" -> "599 contiguous_7" [label="(3, 197, 1, 768)", style=solid]; +"599 contiguous_7" -> "600 quantize_per_tensor_default_44" [label="(3, 197, 1, 768)", style=solid]; +"599 contiguous_7" -> "603 quantize_per_tensor_default_45" [label="(3, 197, 1, 768)", style=solid]; +"599 contiguous_7" -> "606 select_23" [label="(3, 197, 1, 768)", style=solid]; +"600 quantize_per_tensor_default_44" -> "601 dequantize_per_tensor_default_44" [label="(3, 197, 1, 768)", style=solid]; +"601 dequantize_per_tensor_default_44" -> "602 select_21" [label="(3, 197, 1, 768)", style=solid]; +"602 select_21" -> "607 view_56" [label="(197, 1, 768)", style=solid]; +"603 quantize_per_tensor_default_45" -> "604 dequantize_per_tensor_default_45" [label="(3, 197, 1, 768)", style=solid]; +"604 dequantize_per_tensor_default_45" -> "605 select_22" [label="(3, 197, 1, 768)", style=solid]; +"605 select_22" -> "609 view_57" [label="(197, 1, 768)", style=solid]; +"606 select_23" -> "611 view_58" [label="(197, 1, 768)", style=solid]; +"607 view_56" -> "608 transpose_44" [label="(197, 12, 64)", style=solid]; +"608 transpose_44" -> "613 view_59" [label="(12, 197, 64)", style=solid]; +"609 view_57" -> "610 transpose_45" [label="(197, 12, 64)", style=solid]; +"610 transpose_45" -> "614 view_60" [label="(12, 197, 64)", style=solid]; +"611 view_58" -> "612 transpose_46" [label="(197, 12, 64)", style=solid]; +"612 transpose_46" -> "615 view_61" [label="(12, 197, 64)", style=solid]; +"613 view_59" -> "616 scaled_dot_product_attention_7" [label="(1, 12, 197, 64)", style=solid]; +"614 view_60" -> "616 scaled_dot_product_attention_7" [label="(1, 12, 197, 64)", style=solid]; +"615 view_61" -> "616 scaled_dot_product_attention_7" [label="(1, 12, 197, 64)", style=solid]; +"616 scaled_dot_product_attention_7" -> "617 permute_8" [label="(1, 12, 197, 64)", style=solid]; +"617 permute_8" -> "618 view_62" [label="(197, 1, 12, 64)", style=solid]; +"618 view_62" -> "619 view_62_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"619 view_62_0_0_nncf_smooth_quant_0" -> "620 quantize_per_tensor_default_46" [label="(197, 768)", style=solid]; +"620 quantize_per_tensor_default_46" -> "621 dequantize_per_tensor_default_46" [label="(197, 768)", style=solid]; +"621 dequantize_per_tensor_default_46" -> "626 linear_29" [label="(197, 768)", style=solid]; +"622 linear_29_scale_0" -> "625 dequantize_per_channel_default_30" [label="(768,)", style=solid]; +"623 linear_29_zero_point_0" -> "625 dequantize_per_channel_default_30" [label="(768,)", style=solid]; +"624 _frozen_param30" -> "625 dequantize_per_channel_default_30" [label="(768, 768)", style=solid]; +"625 dequantize_per_channel_default_30" -> "626 linear_29" [label="(768, 768)", style=solid]; +"626 linear_29" -> "627 view_63" [label="(197, 768)", style=solid]; +"627 view_63" -> "628 transpose_47" [label="(197, 1, 768)", style=solid]; +"628 transpose_47" -> "629 dropout_22" [label="(1, 197, 768)", style=solid]; +"629 dropout_22" -> "630 add_15" [label="(1, 197, 768)", style=solid]; +"630 add_15" -> "631 layer_norm_15" [label="(1, 197, 768)", style=solid]; +"630 add_15" -> "651 add_16" [label="(1, 197, 768)", style=solid]; +"631 layer_norm_15" -> "632 layer_norm_15_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"632 layer_norm_15_0_0_nncf_smooth_quant_0" -> "633 quantize_per_tensor_default_47" [label="(1, 197, 768)", style=solid]; +"633 quantize_per_tensor_default_47" -> "634 dequantize_per_tensor_default_47" [label="(1, 197, 768)", style=solid]; +"634 dequantize_per_tensor_default_47" -> "639 linear_30" [label="(1, 197, 768)", style=solid]; +"635 linear_30_scale_0" -> "638 dequantize_per_channel_default_31" [label="(3072,)", style=solid]; +"636 linear_30_zero_point_0" -> "638 dequantize_per_channel_default_31" [label="(3072,)", style=solid]; +"637 _frozen_param31" -> "638 dequantize_per_channel_default_31" [label="(3072, 768)", style=solid]; +"638 dequantize_per_channel_default_31" -> "639 linear_30" [label="(3072, 768)", style=solid]; +"639 linear_30" -> "640 gelu_7" [label="(1, 197, 3072)", style=solid]; +"640 gelu_7" -> "641 dropout_23" [label="(1, 197, 3072)", style=solid]; +"641 dropout_23" -> "642 dropout_23_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"642 dropout_23_0_0_nncf_smooth_quant_0" -> "643 quantize_per_tensor_default_48" [label="(1, 197, 3072)", style=solid]; +"643 quantize_per_tensor_default_48" -> "644 dequantize_per_tensor_default_48" [label="(1, 197, 3072)", style=solid]; +"644 dequantize_per_tensor_default_48" -> "649 linear_31" [label="(1, 197, 3072)", style=solid]; +"645 linear_31_scale_0" -> "648 dequantize_per_channel_default_32" [label="(768,)", style=solid]; +"646 linear_31_zero_point_0" -> "648 dequantize_per_channel_default_32" [label="(768,)", style=solid]; +"647 _frozen_param32" -> "648 dequantize_per_channel_default_32" [label="(768, 3072)", style=solid]; +"648 dequantize_per_channel_default_32" -> "649 linear_31" [label="(768, 3072)", style=solid]; +"649 linear_31" -> "650 dropout_24" [label="(1, 197, 768)", style=solid]; +"650 dropout_24" -> "651 add_16" [label="(1, 197, 768)", style=solid]; +"651 add_16" -> "652 layer_norm_16" [label="(1, 197, 768)", style=solid]; +"651 add_16" -> "697 add_17" [label="(1, 197, 768)", style=solid]; +"652 layer_norm_16" -> "653 transpose_48" [label="(1, 197, 768)", style=solid]; +"653 transpose_48" -> "654 transpose_48_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"654 transpose_48_0_0_nncf_smooth_quant_0" -> "655 quantize_per_tensor_default_49" [label="(197, 1, 768)", style=solid]; +"655 quantize_per_tensor_default_49" -> "656 dequantize_per_tensor_default_49" [label="(197, 1, 768)", style=solid]; +"656 dequantize_per_tensor_default_49" -> "661 linear_32" [label="(197, 1, 768)", style=solid]; +"657 linear_32_scale_0" -> "660 dequantize_per_channel_default_33" [label="(2304,)", style=solid]; +"658 linear_32_zero_point_0" -> "660 dequantize_per_channel_default_33" [label="(2304,)", style=solid]; +"659 _frozen_param33" -> "660 dequantize_per_channel_default_33" [label="(2304, 768)", style=solid]; +"660 dequantize_per_channel_default_33" -> "661 linear_32" [label="(2304, 768)", style=solid]; +"661 linear_32" -> "662 unflatten_8" [label="(197, 1, 2304)", style=solid]; +"662 unflatten_8" -> "663 unsqueeze_8" [label="(197, 1, 3, 768)", style=solid]; +"663 unsqueeze_8" -> "664 transpose_49" [label="(1, 197, 1, 3, 768)", style=solid]; +"664 transpose_49" -> "665 squeeze_8" [label="(3, 197, 1, 1, 768)", style=solid]; +"665 squeeze_8" -> "666 contiguous_8" [label="(3, 197, 1, 768)", style=solid]; +"666 contiguous_8" -> "667 quantize_per_tensor_default_50" [label="(3, 197, 1, 768)", style=solid]; +"666 contiguous_8" -> "670 quantize_per_tensor_default_51" [label="(3, 197, 1, 768)", style=solid]; +"666 contiguous_8" -> "673 select_26" [label="(3, 197, 1, 768)", style=solid]; +"667 quantize_per_tensor_default_50" -> "668 dequantize_per_tensor_default_50" [label="(3, 197, 1, 768)", style=solid]; +"668 dequantize_per_tensor_default_50" -> "669 select_24" [label="(3, 197, 1, 768)", style=solid]; +"669 select_24" -> "674 view_64" [label="(197, 1, 768)", style=solid]; +"670 quantize_per_tensor_default_51" -> "671 dequantize_per_tensor_default_51" [label="(3, 197, 1, 768)", style=solid]; +"671 dequantize_per_tensor_default_51" -> "672 select_25" [label="(3, 197, 1, 768)", style=solid]; +"672 select_25" -> "676 view_65" [label="(197, 1, 768)", style=solid]; +"673 select_26" -> "678 view_66" [label="(197, 1, 768)", style=solid]; +"674 view_64" -> "675 transpose_50" [label="(197, 12, 64)", style=solid]; +"675 transpose_50" -> "680 view_67" [label="(12, 197, 64)", style=solid]; +"676 view_65" -> "677 transpose_51" [label="(197, 12, 64)", style=solid]; +"677 transpose_51" -> "681 view_68" [label="(12, 197, 64)", style=solid]; +"678 view_66" -> "679 transpose_52" [label="(197, 12, 64)", style=solid]; +"679 transpose_52" -> "682 view_69" [label="(12, 197, 64)", style=solid]; +"680 view_67" -> "683 scaled_dot_product_attention_8" [label="(1, 12, 197, 64)", style=solid]; +"681 view_68" -> "683 scaled_dot_product_attention_8" [label="(1, 12, 197, 64)", style=solid]; +"682 view_69" -> "683 scaled_dot_product_attention_8" [label="(1, 12, 197, 64)", style=solid]; +"683 scaled_dot_product_attention_8" -> "684 permute_9" [label="(1, 12, 197, 64)", style=solid]; +"684 permute_9" -> "685 view_70" [label="(197, 1, 12, 64)", style=solid]; +"685 view_70" -> "686 view_70_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"686 view_70_0_0_nncf_smooth_quant_0" -> "687 quantize_per_tensor_default_52" [label="(197, 768)", style=solid]; +"687 quantize_per_tensor_default_52" -> "688 dequantize_per_tensor_default_52" [label="(197, 768)", style=solid]; +"688 dequantize_per_tensor_default_52" -> "693 linear_33" [label="(197, 768)", style=solid]; +"689 linear_33_scale_0" -> "692 dequantize_per_channel_default_34" [label="(768,)", style=solid]; +"690 linear_33_zero_point_0" -> "692 dequantize_per_channel_default_34" [label="(768,)", style=solid]; +"691 _frozen_param34" -> "692 dequantize_per_channel_default_34" [label="(768, 768)", style=solid]; +"692 dequantize_per_channel_default_34" -> "693 linear_33" [label="(768, 768)", style=solid]; +"693 linear_33" -> "694 view_71" [label="(197, 768)", style=solid]; +"694 view_71" -> "695 transpose_53" [label="(197, 1, 768)", style=solid]; +"695 transpose_53" -> "696 dropout_25" [label="(1, 197, 768)", style=solid]; +"696 dropout_25" -> "697 add_17" [label="(1, 197, 768)", style=solid]; +"697 add_17" -> "698 layer_norm_17" [label="(1, 197, 768)", style=solid]; +"697 add_17" -> "718 add_18" [label="(1, 197, 768)", style=solid]; +"698 layer_norm_17" -> "699 layer_norm_17_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"699 layer_norm_17_0_0_nncf_smooth_quant_0" -> "700 quantize_per_tensor_default_53" [label="(1, 197, 768)", style=solid]; +"700 quantize_per_tensor_default_53" -> "701 dequantize_per_tensor_default_53" [label="(1, 197, 768)", style=solid]; +"701 dequantize_per_tensor_default_53" -> "706 linear_34" [label="(1, 197, 768)", style=solid]; +"702 linear_34_scale_0" -> "705 dequantize_per_channel_default_35" [label="(3072,)", style=solid]; +"703 linear_34_zero_point_0" -> "705 dequantize_per_channel_default_35" [label="(3072,)", style=solid]; +"704 _frozen_param35" -> "705 dequantize_per_channel_default_35" [label="(3072, 768)", style=solid]; +"705 dequantize_per_channel_default_35" -> "706 linear_34" [label="(3072, 768)", style=solid]; +"706 linear_34" -> "707 gelu_8" [label="(1, 197, 3072)", style=solid]; +"707 gelu_8" -> "708 dropout_26" [label="(1, 197, 3072)", style=solid]; +"708 dropout_26" -> "709 dropout_26_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"709 dropout_26_0_0_nncf_smooth_quant_0" -> "710 quantize_per_tensor_default_54" [label="(1, 197, 3072)", style=solid]; +"710 quantize_per_tensor_default_54" -> "711 dequantize_per_tensor_default_54" [label="(1, 197, 3072)", style=solid]; +"711 dequantize_per_tensor_default_54" -> "716 linear_35" [label="(1, 197, 3072)", style=solid]; +"712 linear_35_scale_0" -> "715 dequantize_per_channel_default_36" [label="(768,)", style=solid]; +"713 linear_35_zero_point_0" -> "715 dequantize_per_channel_default_36" [label="(768,)", style=solid]; +"714 _frozen_param36" -> "715 dequantize_per_channel_default_36" [label="(768, 3072)", style=solid]; +"715 dequantize_per_channel_default_36" -> "716 linear_35" [label="(768, 3072)", style=solid]; +"716 linear_35" -> "717 dropout_27" [label="(1, 197, 768)", style=solid]; +"717 dropout_27" -> "718 add_18" [label="(1, 197, 768)", style=solid]; +"718 add_18" -> "719 layer_norm_18" [label="(1, 197, 768)", style=solid]; +"718 add_18" -> "764 add_19" [label="(1, 197, 768)", style=solid]; +"719 layer_norm_18" -> "720 transpose_54" [label="(1, 197, 768)", style=solid]; +"720 transpose_54" -> "721 transpose_54_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"721 transpose_54_0_0_nncf_smooth_quant_0" -> "722 quantize_per_tensor_default_55" [label="(197, 1, 768)", style=solid]; +"722 quantize_per_tensor_default_55" -> "723 dequantize_per_tensor_default_55" [label="(197, 1, 768)", style=solid]; +"723 dequantize_per_tensor_default_55" -> "728 linear_36" [label="(197, 1, 768)", style=solid]; +"724 linear_36_scale_0" -> "727 dequantize_per_channel_default_37" [label="(2304,)", style=solid]; +"725 linear_36_zero_point_0" -> "727 dequantize_per_channel_default_37" [label="(2304,)", style=solid]; +"726 _frozen_param37" -> "727 dequantize_per_channel_default_37" [label="(2304, 768)", style=solid]; +"727 dequantize_per_channel_default_37" -> "728 linear_36" [label="(2304, 768)", style=solid]; +"728 linear_36" -> "729 unflatten_9" [label="(197, 1, 2304)", style=solid]; +"729 unflatten_9" -> "730 unsqueeze_9" [label="(197, 1, 3, 768)", style=solid]; +"730 unsqueeze_9" -> "731 transpose_55" [label="(1, 197, 1, 3, 768)", style=solid]; +"731 transpose_55" -> "732 squeeze_9" [label="(3, 197, 1, 1, 768)", style=solid]; +"732 squeeze_9" -> "733 contiguous_9" [label="(3, 197, 1, 768)", style=solid]; +"733 contiguous_9" -> "734 quantize_per_tensor_default_56" [label="(3, 197, 1, 768)", style=solid]; +"733 contiguous_9" -> "737 quantize_per_tensor_default_57" [label="(3, 197, 1, 768)", style=solid]; +"733 contiguous_9" -> "740 select_29" [label="(3, 197, 1, 768)", style=solid]; +"734 quantize_per_tensor_default_56" -> "735 dequantize_per_tensor_default_56" [label="(3, 197, 1, 768)", style=solid]; +"735 dequantize_per_tensor_default_56" -> "736 select_27" [label="(3, 197, 1, 768)", style=solid]; +"736 select_27" -> "741 view_72" [label="(197, 1, 768)", style=solid]; +"737 quantize_per_tensor_default_57" -> "738 dequantize_per_tensor_default_57" [label="(3, 197, 1, 768)", style=solid]; +"738 dequantize_per_tensor_default_57" -> "739 select_28" [label="(3, 197, 1, 768)", style=solid]; +"739 select_28" -> "743 view_73" [label="(197, 1, 768)", style=solid]; +"740 select_29" -> "745 view_74" [label="(197, 1, 768)", style=solid]; +"741 view_72" -> "742 transpose_56" [label="(197, 12, 64)", style=solid]; +"742 transpose_56" -> "747 view_75" [label="(12, 197, 64)", style=solid]; +"743 view_73" -> "744 transpose_57" [label="(197, 12, 64)", style=solid]; +"744 transpose_57" -> "748 view_76" [label="(12, 197, 64)", style=solid]; +"745 view_74" -> "746 transpose_58" [label="(197, 12, 64)", style=solid]; +"746 transpose_58" -> "749 view_77" [label="(12, 197, 64)", style=solid]; +"747 view_75" -> "750 scaled_dot_product_attention_9" [label="(1, 12, 197, 64)", style=solid]; +"748 view_76" -> "750 scaled_dot_product_attention_9" [label="(1, 12, 197, 64)", style=solid]; +"749 view_77" -> "750 scaled_dot_product_attention_9" [label="(1, 12, 197, 64)", style=solid]; +"750 scaled_dot_product_attention_9" -> "751 permute_10" [label="(1, 12, 197, 64)", style=solid]; +"751 permute_10" -> "752 view_78" [label="(197, 1, 12, 64)", style=solid]; +"752 view_78" -> "753 view_78_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"753 view_78_0_0_nncf_smooth_quant_0" -> "754 quantize_per_tensor_default_58" [label="(197, 768)", style=solid]; +"754 quantize_per_tensor_default_58" -> "755 dequantize_per_tensor_default_58" [label="(197, 768)", style=solid]; +"755 dequantize_per_tensor_default_58" -> "760 linear_37" [label="(197, 768)", style=solid]; +"756 linear_37_scale_0" -> "759 dequantize_per_channel_default_38" [label="(768,)", style=solid]; +"757 linear_37_zero_point_0" -> "759 dequantize_per_channel_default_38" [label="(768,)", style=solid]; +"758 _frozen_param38" -> "759 dequantize_per_channel_default_38" [label="(768, 768)", style=solid]; +"759 dequantize_per_channel_default_38" -> "760 linear_37" [label="(768, 768)", style=solid]; +"760 linear_37" -> "761 view_79" [label="(197, 768)", style=solid]; +"761 view_79" -> "762 transpose_59" [label="(197, 1, 768)", style=solid]; +"762 transpose_59" -> "763 dropout_28" [label="(1, 197, 768)", style=solid]; +"763 dropout_28" -> "764 add_19" [label="(1, 197, 768)", style=solid]; +"764 add_19" -> "765 layer_norm_19" [label="(1, 197, 768)", style=solid]; +"764 add_19" -> "785 add_20" [label="(1, 197, 768)", style=solid]; +"765 layer_norm_19" -> "766 layer_norm_19_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"766 layer_norm_19_0_0_nncf_smooth_quant_0" -> "767 quantize_per_tensor_default_59" [label="(1, 197, 768)", style=solid]; +"767 quantize_per_tensor_default_59" -> "768 dequantize_per_tensor_default_59" [label="(1, 197, 768)", style=solid]; +"768 dequantize_per_tensor_default_59" -> "773 linear_38" [label="(1, 197, 768)", style=solid]; +"769 linear_38_scale_0" -> "772 dequantize_per_channel_default_39" [label="(3072,)", style=solid]; +"770 linear_38_zero_point_0" -> "772 dequantize_per_channel_default_39" [label="(3072,)", style=solid]; +"771 _frozen_param39" -> "772 dequantize_per_channel_default_39" [label="(3072, 768)", style=solid]; +"772 dequantize_per_channel_default_39" -> "773 linear_38" [label="(3072, 768)", style=solid]; +"773 linear_38" -> "774 gelu_9" [label="(1, 197, 3072)", style=solid]; +"774 gelu_9" -> "775 dropout_29" [label="(1, 197, 3072)", style=solid]; +"775 dropout_29" -> "776 dropout_29_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"776 dropout_29_0_0_nncf_smooth_quant_0" -> "777 quantize_per_tensor_default_60" [label="(1, 197, 3072)", style=solid]; +"777 quantize_per_tensor_default_60" -> "778 dequantize_per_tensor_default_60" [label="(1, 197, 3072)", style=solid]; +"778 dequantize_per_tensor_default_60" -> "783 linear_39" [label="(1, 197, 3072)", style=solid]; +"779 linear_39_scale_0" -> "782 dequantize_per_channel_default_40" [label="(768,)", style=solid]; +"780 linear_39_zero_point_0" -> "782 dequantize_per_channel_default_40" [label="(768,)", style=solid]; +"781 _frozen_param40" -> "782 dequantize_per_channel_default_40" [label="(768, 3072)", style=solid]; +"782 dequantize_per_channel_default_40" -> "783 linear_39" [label="(768, 3072)", style=solid]; +"783 linear_39" -> "784 dropout_30" [label="(1, 197, 768)", style=solid]; +"784 dropout_30" -> "785 add_20" [label="(1, 197, 768)", style=solid]; +"785 add_20" -> "786 layer_norm_20" [label="(1, 197, 768)", style=solid]; +"785 add_20" -> "831 add_21" [label="(1, 197, 768)", style=solid]; +"786 layer_norm_20" -> "787 transpose_60" [label="(1, 197, 768)", style=solid]; +"787 transpose_60" -> "788 transpose_60_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"788 transpose_60_0_0_nncf_smooth_quant_0" -> "789 quantize_per_tensor_default_61" [label="(197, 1, 768)", style=solid]; +"789 quantize_per_tensor_default_61" -> "790 dequantize_per_tensor_default_61" [label="(197, 1, 768)", style=solid]; +"790 dequantize_per_tensor_default_61" -> "795 linear_40" [label="(197, 1, 768)", style=solid]; +"791 linear_40_scale_0" -> "794 dequantize_per_channel_default_41" [label="(2304,)", style=solid]; +"792 linear_40_zero_point_0" -> "794 dequantize_per_channel_default_41" [label="(2304,)", style=solid]; +"793 _frozen_param41" -> "794 dequantize_per_channel_default_41" [label="(2304, 768)", style=solid]; +"794 dequantize_per_channel_default_41" -> "795 linear_40" [label="(2304, 768)", style=solid]; +"795 linear_40" -> "796 unflatten_10" [label="(197, 1, 2304)", style=solid]; +"796 unflatten_10" -> "797 unsqueeze_10" [label="(197, 1, 3, 768)", style=solid]; +"797 unsqueeze_10" -> "798 transpose_61" [label="(1, 197, 1, 3, 768)", style=solid]; +"798 transpose_61" -> "799 squeeze_10" [label="(3, 197, 1, 1, 768)", style=solid]; +"799 squeeze_10" -> "800 contiguous_10" [label="(3, 197, 1, 768)", style=solid]; +"800 contiguous_10" -> "801 quantize_per_tensor_default_62" [label="(3, 197, 1, 768)", style=solid]; +"800 contiguous_10" -> "804 quantize_per_tensor_default_63" [label="(3, 197, 1, 768)", style=solid]; +"800 contiguous_10" -> "807 select_32" [label="(3, 197, 1, 768)", style=solid]; +"801 quantize_per_tensor_default_62" -> "802 dequantize_per_tensor_default_62" [label="(3, 197, 1, 768)", style=solid]; +"802 dequantize_per_tensor_default_62" -> "803 select_30" [label="(3, 197, 1, 768)", style=solid]; +"803 select_30" -> "808 view_80" [label="(197, 1, 768)", style=solid]; +"804 quantize_per_tensor_default_63" -> "805 dequantize_per_tensor_default_63" [label="(3, 197, 1, 768)", style=solid]; +"805 dequantize_per_tensor_default_63" -> "806 select_31" [label="(3, 197, 1, 768)", style=solid]; +"806 select_31" -> "810 view_81" [label="(197, 1, 768)", style=solid]; +"807 select_32" -> "812 view_82" [label="(197, 1, 768)", style=solid]; +"808 view_80" -> "809 transpose_62" [label="(197, 12, 64)", style=solid]; +"809 transpose_62" -> "814 view_83" [label="(12, 197, 64)", style=solid]; +"810 view_81" -> "811 transpose_63" [label="(197, 12, 64)", style=solid]; +"811 transpose_63" -> "815 view_84" [label="(12, 197, 64)", style=solid]; +"812 view_82" -> "813 transpose_64" [label="(197, 12, 64)", style=solid]; +"813 transpose_64" -> "816 view_85" [label="(12, 197, 64)", style=solid]; +"814 view_83" -> "817 scaled_dot_product_attention_10" [label="(1, 12, 197, 64)", style=solid]; +"815 view_84" -> "817 scaled_dot_product_attention_10" [label="(1, 12, 197, 64)", style=solid]; +"816 view_85" -> "817 scaled_dot_product_attention_10" [label="(1, 12, 197, 64)", style=solid]; +"817 scaled_dot_product_attention_10" -> "818 permute_11" [label="(1, 12, 197, 64)", style=solid]; +"818 permute_11" -> "819 view_86" [label="(197, 1, 12, 64)", style=solid]; +"819 view_86" -> "820 view_86_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"820 view_86_0_0_nncf_smooth_quant_0" -> "821 quantize_per_tensor_default_64" [label="(197, 768)", style=solid]; +"821 quantize_per_tensor_default_64" -> "822 dequantize_per_tensor_default_64" [label="(197, 768)", style=solid]; +"822 dequantize_per_tensor_default_64" -> "827 linear_41" [label="(197, 768)", style=solid]; +"823 linear_41_scale_0" -> "826 dequantize_per_channel_default_42" [label="(768,)", style=solid]; +"824 linear_41_zero_point_0" -> "826 dequantize_per_channel_default_42" [label="(768,)", style=solid]; +"825 _frozen_param42" -> "826 dequantize_per_channel_default_42" [label="(768, 768)", style=solid]; +"826 dequantize_per_channel_default_42" -> "827 linear_41" [label="(768, 768)", style=solid]; +"827 linear_41" -> "828 view_87" [label="(197, 768)", style=solid]; +"828 view_87" -> "829 transpose_65" [label="(197, 1, 768)", style=solid]; +"829 transpose_65" -> "830 dropout_31" [label="(1, 197, 768)", style=solid]; +"830 dropout_31" -> "831 add_21" [label="(1, 197, 768)", style=solid]; +"831 add_21" -> "832 layer_norm_21" [label="(1, 197, 768)", style=solid]; +"831 add_21" -> "852 add_22" [label="(1, 197, 768)", style=solid]; +"832 layer_norm_21" -> "833 layer_norm_21_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"833 layer_norm_21_0_0_nncf_smooth_quant_0" -> "834 quantize_per_tensor_default_65" [label="(1, 197, 768)", style=solid]; +"834 quantize_per_tensor_default_65" -> "835 dequantize_per_tensor_default_65" [label="(1, 197, 768)", style=solid]; +"835 dequantize_per_tensor_default_65" -> "840 linear_42" [label="(1, 197, 768)", style=solid]; +"836 linear_42_scale_0" -> "839 dequantize_per_channel_default_43" [label="(3072,)", style=solid]; +"837 linear_42_zero_point_0" -> "839 dequantize_per_channel_default_43" [label="(3072,)", style=solid]; +"838 _frozen_param43" -> "839 dequantize_per_channel_default_43" [label="(3072, 768)", style=solid]; +"839 dequantize_per_channel_default_43" -> "840 linear_42" [label="(3072, 768)", style=solid]; +"840 linear_42" -> "841 gelu_10" [label="(1, 197, 3072)", style=solid]; +"841 gelu_10" -> "842 dropout_32" [label="(1, 197, 3072)", style=solid]; +"842 dropout_32" -> "843 dropout_32_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"843 dropout_32_0_0_nncf_smooth_quant_0" -> "844 quantize_per_tensor_default_66" [label="(1, 197, 3072)", style=solid]; +"844 quantize_per_tensor_default_66" -> "845 dequantize_per_tensor_default_66" [label="(1, 197, 3072)", style=solid]; +"845 dequantize_per_tensor_default_66" -> "850 linear_43" [label="(1, 197, 3072)", style=solid]; +"846 linear_43_scale_0" -> "849 dequantize_per_channel_default_44" [label="(768,)", style=solid]; +"847 linear_43_zero_point_0" -> "849 dequantize_per_channel_default_44" [label="(768,)", style=solid]; +"848 _frozen_param44" -> "849 dequantize_per_channel_default_44" [label="(768, 3072)", style=solid]; +"849 dequantize_per_channel_default_44" -> "850 linear_43" [label="(768, 3072)", style=solid]; +"850 linear_43" -> "851 dropout_33" [label="(1, 197, 768)", style=solid]; +"851 dropout_33" -> "852 add_22" [label="(1, 197, 768)", style=solid]; +"852 add_22" -> "853 layer_norm_22" [label="(1, 197, 768)", style=solid]; +"852 add_22" -> "898 add_23" [label="(1, 197, 768)", style=solid]; +"853 layer_norm_22" -> "854 transpose_66" [label="(1, 197, 768)", style=solid]; +"854 transpose_66" -> "855 transpose_66_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"855 transpose_66_0_0_nncf_smooth_quant_0" -> "856 quantize_per_tensor_default_67" [label="(197, 1, 768)", style=solid]; +"856 quantize_per_tensor_default_67" -> "857 dequantize_per_tensor_default_67" [label="(197, 1, 768)", style=solid]; +"857 dequantize_per_tensor_default_67" -> "862 linear_44" [label="(197, 1, 768)", style=solid]; +"858 linear_44_scale_0" -> "861 dequantize_per_channel_default_45" [label="(2304,)", style=solid]; +"859 linear_44_zero_point_0" -> "861 dequantize_per_channel_default_45" [label="(2304,)", style=solid]; +"860 _frozen_param45" -> "861 dequantize_per_channel_default_45" [label="(2304, 768)", style=solid]; +"861 dequantize_per_channel_default_45" -> "862 linear_44" [label="(2304, 768)", style=solid]; +"862 linear_44" -> "863 unflatten_11" [label="(197, 1, 2304)", style=solid]; +"863 unflatten_11" -> "864 unsqueeze_11" [label="(197, 1, 3, 768)", style=solid]; +"864 unsqueeze_11" -> "865 transpose_67" [label="(1, 197, 1, 3, 768)", style=solid]; +"865 transpose_67" -> "866 squeeze_11" [label="(3, 197, 1, 1, 768)", style=solid]; +"866 squeeze_11" -> "867 contiguous_11" [label="(3, 197, 1, 768)", style=solid]; +"867 contiguous_11" -> "868 quantize_per_tensor_default_68" [label="(3, 197, 1, 768)", style=solid]; +"867 contiguous_11" -> "871 quantize_per_tensor_default_69" [label="(3, 197, 1, 768)", style=solid]; +"867 contiguous_11" -> "874 select_35" [label="(3, 197, 1, 768)", style=solid]; +"868 quantize_per_tensor_default_68" -> "869 dequantize_per_tensor_default_68" [label="(3, 197, 1, 768)", style=solid]; +"869 dequantize_per_tensor_default_68" -> "870 select_33" [label="(3, 197, 1, 768)", style=solid]; +"870 select_33" -> "875 view_88" [label="(197, 1, 768)", style=solid]; +"871 quantize_per_tensor_default_69" -> "872 dequantize_per_tensor_default_69" [label="(3, 197, 1, 768)", style=solid]; +"872 dequantize_per_tensor_default_69" -> "873 select_34" [label="(3, 197, 1, 768)", style=solid]; +"873 select_34" -> "877 view_89" [label="(197, 1, 768)", style=solid]; +"874 select_35" -> "879 view_90" [label="(197, 1, 768)", style=solid]; +"875 view_88" -> "876 transpose_68" [label="(197, 12, 64)", style=solid]; +"876 transpose_68" -> "881 view_91" [label="(12, 197, 64)", style=solid]; +"877 view_89" -> "878 transpose_69" [label="(197, 12, 64)", style=solid]; +"878 transpose_69" -> "882 view_92" [label="(12, 197, 64)", style=solid]; +"879 view_90" -> "880 transpose_70" [label="(197, 12, 64)", style=solid]; +"880 transpose_70" -> "883 view_93" [label="(12, 197, 64)", style=solid]; +"881 view_91" -> "884 scaled_dot_product_attention_11" [label="(1, 12, 197, 64)", style=solid]; +"882 view_92" -> "884 scaled_dot_product_attention_11" [label="(1, 12, 197, 64)", style=solid]; +"883 view_93" -> "884 scaled_dot_product_attention_11" [label="(1, 12, 197, 64)", style=solid]; +"884 scaled_dot_product_attention_11" -> "885 permute_12" [label="(1, 12, 197, 64)", style=solid]; +"885 permute_12" -> "886 view_94" [label="(197, 1, 12, 64)", style=solid]; +"886 view_94" -> "887 view_94_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"887 view_94_0_0_nncf_smooth_quant_0" -> "888 quantize_per_tensor_default_70" [label="(197, 768)", style=solid]; +"888 quantize_per_tensor_default_70" -> "889 dequantize_per_tensor_default_70" [label="(197, 768)", style=solid]; +"889 dequantize_per_tensor_default_70" -> "894 linear_45" [label="(197, 768)", style=solid]; +"890 linear_45_scale_0" -> "893 dequantize_per_channel_default_46" [label="(768,)", style=solid]; +"891 linear_45_zero_point_0" -> "893 dequantize_per_channel_default_46" [label="(768,)", style=solid]; +"892 _frozen_param46" -> "893 dequantize_per_channel_default_46" [label="(768, 768)", style=solid]; +"893 dequantize_per_channel_default_46" -> "894 linear_45" [label="(768, 768)", style=solid]; +"894 linear_45" -> "895 view_95" [label="(197, 768)", style=solid]; +"895 view_95" -> "896 transpose_71" [label="(197, 1, 768)", style=solid]; +"896 transpose_71" -> "897 dropout_34" [label="(1, 197, 768)", style=solid]; +"897 dropout_34" -> "898 add_23" [label="(1, 197, 768)", style=solid]; +"898 add_23" -> "899 layer_norm_23" [label="(1, 197, 768)", style=solid]; +"898 add_23" -> "919 add_24" [label="(1, 197, 768)", style=solid]; +"899 layer_norm_23" -> "900 layer_norm_23_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"900 layer_norm_23_0_0_nncf_smooth_quant_0" -> "901 quantize_per_tensor_default_71" [label="(1, 197, 768)", style=solid]; +"901 quantize_per_tensor_default_71" -> "902 dequantize_per_tensor_default_71" [label="(1, 197, 768)", style=solid]; +"902 dequantize_per_tensor_default_71" -> "907 linear_46" [label="(1, 197, 768)", style=solid]; +"903 linear_46_scale_0" -> "906 dequantize_per_channel_default_47" [label="(3072,)", style=solid]; +"904 linear_46_zero_point_0" -> "906 dequantize_per_channel_default_47" [label="(3072,)", style=solid]; +"905 _frozen_param47" -> "906 dequantize_per_channel_default_47" [label="(3072, 768)", style=solid]; +"906 dequantize_per_channel_default_47" -> "907 linear_46" [label="(3072, 768)", style=solid]; +"907 linear_46" -> "908 gelu_11" [label="(1, 197, 3072)", style=solid]; +"908 gelu_11" -> "909 dropout_35" [label="(1, 197, 3072)", style=solid]; +"909 dropout_35" -> "910 dropout_35_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"910 dropout_35_0_0_nncf_smooth_quant_0" -> "911 quantize_per_tensor_default_72" [label="(1, 197, 3072)", style=solid]; +"911 quantize_per_tensor_default_72" -> "912 dequantize_per_tensor_default_72" [label="(1, 197, 3072)", style=solid]; +"912 dequantize_per_tensor_default_72" -> "917 linear_47" [label="(1, 197, 3072)", style=solid]; +"913 linear_47_scale_0" -> "916 dequantize_per_channel_default_48" [label="(768,)", style=solid]; +"914 linear_47_zero_point_0" -> "916 dequantize_per_channel_default_48" [label="(768,)", style=solid]; +"915 _frozen_param48" -> "916 dequantize_per_channel_default_48" [label="(768, 3072)", style=solid]; +"916 dequantize_per_channel_default_48" -> "917 linear_47" [label="(768, 3072)", style=solid]; +"917 linear_47" -> "918 dropout_36" [label="(1, 197, 768)", style=solid]; +"918 dropout_36" -> "919 add_24" [label="(1, 197, 768)", style=solid]; +"919 add_24" -> "920 layer_norm_24" [label="(1, 197, 768)", style=solid]; +"920 layer_norm_24" -> "921 slice_1" [label="(1, 197, 768)", style=solid]; +"921 slice_1" -> "922 select_36" [label="(1, 197, 768)", style=solid]; +"922 select_36" -> "923 select_36_0_0_nncf_smooth_quant_0" [label="(1, 768)", style=solid]; +"923 select_36_0_0_nncf_smooth_quant_0" -> "924 quantize_per_tensor_default_73" [label="(1, 768)", style=solid]; +"924 quantize_per_tensor_default_73" -> "925 dequantize_per_tensor_default_73" [label="(1, 768)", style=solid]; +"925 dequantize_per_tensor_default_73" -> "930 linear_48" [label="(1, 768)", style=solid]; +"926 linear_48_scale_0" -> "929 dequantize_per_channel_default_49" [label="(1000,)", style=solid]; +"927 linear_48_zero_point_0" -> "929 dequantize_per_channel_default_49" [label="(1000,)", style=solid]; +"928 _frozen_param49" -> "929 dequantize_per_channel_default_49" [label="(1000, 768)", style=solid]; +"929 dequantize_per_channel_default_49" -> "930 linear_48" [label="(1000, 768)", style=solid]; +"930 linear_48" -> "931 output" [label="(1, 1000)", style=solid]; +} diff --git a/tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/yolo11n_sdpa_block.dot b/tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/yolo11n_sdpa_block.dot new file mode 100644 index 00000000000..63ad8ac0b09 --- /dev/null +++ b/tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/yolo11n_sdpa_block.dot @@ -0,0 +1,61 @@ +strict digraph { +"0 x" [id=0, type=input]; +"1 x_0_0_nncf_smooth_quant_0" [id=1, type=call_module]; +"2 quantize_per_tensor_default" [id=2, type=quantize_per_tensor]; +"3 dequantize_per_tensor_default" [id=3, type=dequantize_per_tensor]; +"4 linear_scale_0" [id=4, type=get_attr]; +"5 linear_zero_point_0" [id=5, type=get_attr]; +"6 _frozen_param0" [id=6, type=get_attr]; +"7 dequantize_per_channel_default" [id=7, type=dequantize_per_channel]; +"8 linear" [id=8, type=linear]; +"9 quantize_per_tensor_default_1" [id=9, type=quantize_per_tensor]; +"10 dequantize_per_tensor_default_1" [id=10, type=dequantize_per_tensor]; +"11 slice_1" [id=11, type=slice]; +"12 slice_2" [id=12, type=slice]; +"13 slice_3" [id=13, type=slice]; +"14 quantize_per_tensor_default_2" [id=14, type=quantize_per_tensor]; +"15 dequantize_per_tensor_default_2" [id=15, type=dequantize_per_tensor]; +"16 slice_4" [id=16, type=slice]; +"17 slice_5" [id=17, type=slice]; +"18 slice_6" [id=18, type=slice]; +"19 slice_7" [id=19, type=slice]; +"20 slice_8" [id=20, type=slice]; +"21 slice_9" [id=21, type=slice]; +"22 transpose" [id=22, type=transpose]; +"23 matmul" [id=23, type=matmul]; +"24 div_" [id=24, type=div_]; +"25 softmax" [id=25, type=softmax]; +"26 transpose_1" [id=26, type=transpose]; +"27 matmul_1" [id=27, type=matmul]; +"28 output" [id=28, type=output]; +"0 x" -> "1 x_0_0_nncf_smooth_quant_0" [label="(1, 2, 4)", style=solid]; +"1 x_0_0_nncf_smooth_quant_0" -> "2 quantize_per_tensor_default" [label="(1, 2, 4)", style=solid]; +"2 quantize_per_tensor_default" -> "3 dequantize_per_tensor_default" [label="(1, 2, 4)", style=solid]; +"3 dequantize_per_tensor_default" -> "8 linear" [label="(1, 2, 4)", style=solid]; +"4 linear_scale_0" -> "7 dequantize_per_channel_default" [label="(12,)", style=solid]; +"5 linear_zero_point_0" -> "7 dequantize_per_channel_default" [label="(12,)", style=solid]; +"6 _frozen_param0" -> "7 dequantize_per_channel_default" [label="(12, 4)", style=solid]; +"7 dequantize_per_channel_default" -> "8 linear" [label="(12, 4)", style=solid]; +"8 linear" -> "9 quantize_per_tensor_default_1" [label="(1, 2, 12)", style=solid]; +"8 linear" -> "14 quantize_per_tensor_default_2" [label="(1, 2, 12)", style=solid]; +"8 linear" -> "19 slice_7" [label="(1, 2, 12)", style=solid]; +"9 quantize_per_tensor_default_1" -> "10 dequantize_per_tensor_default_1" [label="(1, 2, 12)", style=solid]; +"10 dequantize_per_tensor_default_1" -> "11 slice_1" [label="(1, 2, 12)", style=solid]; +"11 slice_1" -> "12 slice_2" [label="(1, 2, 12)", style=solid]; +"12 slice_2" -> "13 slice_3" [label="(1, 2, 12)", style=solid]; +"13 slice_3" -> "23 matmul" [label="(1, 2, 4)", style=solid]; +"14 quantize_per_tensor_default_2" -> "15 dequantize_per_tensor_default_2" [label="(1, 2, 12)", style=solid]; +"15 dequantize_per_tensor_default_2" -> "16 slice_4" [label="(1, 2, 12)", style=solid]; +"16 slice_4" -> "17 slice_5" [label="(1, 2, 12)", style=solid]; +"17 slice_5" -> "18 slice_6" [label="(1, 2, 12)", style=solid]; +"18 slice_6" -> "22 transpose" [label="(1, 2, 4)", style=solid]; +"19 slice_7" -> "20 slice_8" [label="(1, 2, 12)", style=solid]; +"20 slice_8" -> "21 slice_9" [label="(1, 2, 12)", style=solid]; +"21 slice_9" -> "27 matmul_1" [label="(1, 2, 4)", style=solid]; +"22 transpose" -> "23 matmul" [label="(1, 4, 2)", style=solid]; +"23 matmul" -> "24 div_" [label="(1, 2, 2)", style=solid]; +"24 div_" -> "25 softmax" [label="(1, 2, 2)", style=solid]; +"25 softmax" -> "26 transpose_1" [label="(1, 2, 2)", style=solid]; +"26 transpose_1" -> "27 matmul_1" [label="(1, 2, 2)", style=solid]; +"27 matmul_1" -> "28 output" [label="(1, 2, 4)", style=solid]; +} From 429c10f100afe34bab39f16051ab1bb68e9cd4e6 Mon Sep 17 00:00:00 2001 From: dlyakhov Date: Fri, 29 Nov 2024 12:06:01 +0100 Subject: [PATCH 08/15] MinMaxAlgoritm is reused for OpenVINOQuantizer --- .../quantizer/openvino_quantizer.py | 91 +++ .../quantizer/openvino_quantizer/backend.py | 323 ---------- .../openvino_quantizer/openvino_quantizer.py | 593 ------------------ .../openvino_quantizer/torch_fx_backend.py | 366 ----------- .../algorithms/min_max/algorithm.py | 5 +- .../algorithms/min_max/quantizer/backend.py | 323 ---------- .../min_max/quantizer/torch_fx_backend.py | 366 ----------- tests/torch/fx/test_quantizer.py | 121 ++-- 8 files changed, 156 insertions(+), 2032 deletions(-) create mode 100644 nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer.py delete mode 100644 nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer/backend.py delete mode 100644 nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer/openvino_quantizer.py delete mode 100644 nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer/torch_fx_backend.py delete mode 100644 nncf/quantization/algorithms/min_max/quantizer/backend.py delete mode 100644 nncf/quantization/algorithms/min_max/quantizer/torch_fx_backend.py diff --git a/nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer.py b/nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer.py new file mode 100644 index 00000000000..36f1f1fd3e3 --- /dev/null +++ b/nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer.py @@ -0,0 +1,91 @@ +# Copyright (c) 2024 Intel Corporation +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# http://www.apache.org/licenses/LICENSE-2.0 +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from typing import Optional, Union + +import torch.fx + +from nncf.common.graph.graph import NNCFGraph +from nncf.common.quantization.quantizer_propagation.solver import QuantizerPropagationRule +from nncf.common.quantization.quantizer_setup import SingleConfigQuantizerSetup +from nncf.common.quantization.structs import QuantizationPreset +from nncf.experimental.common.quantization.algorithms.quantizer.base_quantizer import NNCFQuantizer +from nncf.parameters import ModelType +from nncf.parameters import QuantizationMode +from nncf.parameters import TargetDevice +from nncf.quantization.advanced_parameters import FP8QuantizationParameters +from nncf.quantization.advanced_parameters import OverflowFix +from nncf.quantization.advanced_parameters import QuantizationParameters +from nncf.quantization.algorithms.min_max.algorithm import MinMaxQuantization +from nncf.scopes import IgnoredScope + + +class OpenVINOQuantizer(NNCFQuantizer): + def __init__( + self, + mode: Optional[QuantizationMode] = None, + preset: Optional[QuantizationPreset] = None, + target_device: TargetDevice = TargetDevice.ANY, + model_type: Optional[ModelType] = None, + ignored_scope: Optional[IgnoredScope] = None, + overflow_fix: Optional[OverflowFix] = None, + quantize_outputs: bool = False, + activations_quantization_params: Union[QuantizationParameters, FP8QuantizationParameters] = None, + weights_quantization_params: Union[QuantizationParameters, FP8QuantizationParameters] = None, + quantizer_propagation_rule: Optional[QuantizerPropagationRule] = None, + ): + """ + :param mode: Defines optimization mode for the algorithm. None by default. + :param preset: A preset controls the quantization mode (symmetric and asymmetric). + It can take the following values: + - `performance`: Symmetric quantization of weights and activations. + - `mixed`: Symmetric quantization of weights and asymmetric quantization of activations. + Default value is None. In this case, `mixed` preset is used for `transformer` + model type otherwise `performance`. + :param target_device: A target device the specificity of which will be taken + into account while compressing in order to obtain the best performance + for this type of device, defaults to TargetDevice.ANY. + :param model_type: Model type is needed to specify additional patterns + in the model. Supported only `transformer` now. + :param ignored_scope: An ignored scope that defined the list of model control + flow graph nodes to be ignored during quantization. + :param overflow_fix: This option controls whether to apply the overflow issue + fix for the 8-bit quantization. + :param quantize_outputs: Whether to insert additional quantizers right before + each of the model outputs. + :param activations_quantization_params: Quantization parameters for model + activations. + :param weights_quantization_params: Quantization parameters for model weights. + :param quantizer_propagation_rule: The strategy to be used while propagating and merging quantizers. + """ + self._min_max_algo = MinMaxQuantization( + mode=mode, + preset=preset, + target_device=target_device, + model_type=model_type, + ignored_scope=ignored_scope, + overflow_fix=overflow_fix, + quantize_outputs=quantize_outputs, + activations_quantization_params=activations_quantization_params, + weights_quantization_params=weights_quantization_params, + quantizer_propagation_rule=quantizer_propagation_rule, + ) + + def get_quantization_setup(self, model: torch.fx.GraphModule, nncf_graph: NNCFGraph) -> SingleConfigQuantizerSetup: + """ + Builds SingleConfigQuantizerSetup for the given model. + + :param model: Backend-specific model, for which Quantization Target Points are being seek. + :param nncf_graph: NNCFGraph instance. + :return: SingleConfigQuantizerSetup for the given model. + """ + self._min_max_algo._set_backend_entity(model) + return self._min_max_algo._find_quantization_setup(model, nncf_graph) diff --git a/nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer/backend.py b/nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer/backend.py deleted file mode 100644 index 9f7ea5441e1..00000000000 --- a/nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer/backend.py +++ /dev/null @@ -1,323 +0,0 @@ -# Copyright (c) 2024 Intel Corporation -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# http://www.apache.org/licenses/LICENSE-2.0 -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. - -from abc import ABC -from abc import abstractmethod -from typing import Dict, List, Optional, Set, Tuple, TypeVar - -from nncf.common.graph.graph import NNCFGraph -from nncf.common.graph.graph import NNCFNode -from nncf.common.graph.operator_metatypes import OperatorMetatype -from nncf.common.graph.transformations.commands import TargetPoint -from nncf.common.graph.transformations.commands import TargetType -from nncf.common.graph.transformations.commands import TransformationCommand -from nncf.common.hardware.config import HWConfig -from nncf.common.quantization.structs import QuantizerConfig -from nncf.common.tensor_statistics.collectors import TensorStatisticCollectorBase -from nncf.parameters import ModelType -from nncf.parameters import TargetDevice -from nncf.quantization.fake_quantize import FakeConvertParameters -from nncf.quantization.fake_quantize import FakeQuantizeParameters -from nncf.quantization.range_estimator import RangeEstimatorParameters - -TModel = TypeVar("TModel") - - -class OpenVINOQuantizerBackend(ABC): - @property - @abstractmethod - def preserved_metatypes(self) -> List[OperatorMetatype]: - """ - Property for backend-specific metatypes that require preserving float subgraphs - when removing the ShapeOf subgraph. - """ - - @property - @abstractmethod - def mat_mul_metatypes(self) -> List[OperatorMetatype]: - """ - Property for the backend-specific MatMul metatypes. - """ - - @property - @abstractmethod - def post_processing_metatypes(self) -> List[OperatorMetatype]: - """ - Property for the backend-specific post-processing metatypes (NonMaximumSupression, TopK, etc.). - """ - - @property - @abstractmethod - def conv_metatypes(self) -> List[OperatorMetatype]: - """ - Property for the backend-specific Convolution metatypes. - """ - - @property - @abstractmethod - def shapeof_metatypes(self) -> List[OperatorMetatype]: - """ - Property for the backend-specific ShapeOf metatypes. - """ - - @property - @abstractmethod - def dropout_metatypes(self) -> List[OperatorMetatype]: - """ - Property for the backend-specific Dropout metatypes. - """ - - @property - @abstractmethod - def elementwise_metatypes(self) -> List[OperatorMetatype]: - """ - Property for the backend-specific Elementwises metatypes. - """ - - @property - @abstractmethod - def overflow_fix_metatypes(self) -> List[OperatorMetatype]: - """ - Property for the backend-specific metatypes for which overflow_fix is applicable. - """ - - @property - @abstractmethod - def add_metatypes(self) -> List[OperatorMetatype]: - """ - Property for the backend-specific metatypes that also can be interpreted as Add layer. - """ - - @property - @abstractmethod - def group_conv_metatypes(self) -> List[OperatorMetatype]: - """ - Property for the backend-specific Grouped Convolution metatypes. - """ - - @property - @abstractmethod - def scaled_dot_product_attention_metatypes(self) -> List[OperatorMetatype]: - """ - Property for the backend-specific Scaled Dot Product Attention metatypes. - """ - - @property - @abstractmethod - def scales_unification_map(self) -> Dict[OperatorMetatype, OperatorMetatype]: - """ - Property for the backend-specific metatypes that produces quantizers that might be unified. - """ - - @property - @abstractmethod - def hw_config(self) -> HWConfig: - """ - Property for the hardware backend-specific configuration. - """ - - @property - @abstractmethod - def quant_trait_op_dict(self) -> Dict[int, OperatorMetatype]: - """ - Property for the backend-specific dictionary that contains QuantizationTrait-specific metatypes. - """ - - @staticmethod - @abstractmethod - def target_point(target_type: TargetType, target_node_name: str, port_id: int) -> TargetPoint: - """ - Returns backend-specific target point. - - :param target_type: Type of the location that should be modified. - :param target_node_name: Name of the located node. - :param port_id: Port ID of the tensor for the statistics distribution. - :return: Backend-specific TargetPoint. - """ - - @staticmethod - @abstractmethod - def create_quantizer_insertion_command( - nncf_graph: NNCFGraph, - target_point: TargetPoint, - quantizer_config: QuantizerConfig, - parameters: FakeQuantizeParameters, - ) -> TransformationCommand: - """ - Returns backend-specific quantizer insertion command. - - :param nncf_graph: NNCFGraph to get input/output shapes for the target point. - :param target_point: Target location for the quantizer insertion. - :param quantizer_config: QuantizerConfig instance for the current layer. - :param parameters: FakeQuantizeParameters to calculate activation quantization parameters. - :return: Backend-specific TransformationCommand for the quantizer insertion operation. - """ - - @staticmethod - @abstractmethod - def create_unified_scales_quantizers_insertion_commands( - nncf_graph: NNCFGraph, - target_points: List[TargetPoint], - quantizer_config: QuantizerConfig, - parameters: FakeQuantizeParameters, - ) -> List[TransformationCommand]: - """ - Returns backend-specific unified scales quantizers insertion commands. - - :param nncf_graph: NNCFGraph to get input/output shapes for the target point. - :param target_points: List of target locations for the quantizers insertion. - :param quantizer_config: QuantizerConfig instance for the current layer. - :param parameters: FakeQuantizeParameters to calculate activation quantization parameters. - :return: List of backend-specific TransformationCommands - for the quantizers with unified scales insertion operations. - """ - - @staticmethod - @abstractmethod - def create_convert_insertion_command( - target_point: TargetPoint, - parameters: FakeConvertParameters, - ) -> TransformationCommand: - """ - Returns backend-specific convert insertion command. - - :param target_point: Target location for the correction. - :param parameters: FakeConvertParameters to calculate activation quantization parameters. - :return: Backend-specific TransformationCommand for the quantizer insertion operation. - """ - - @staticmethod - @abstractmethod - def get_start_nodes_for_activation_path_tracing(nncf_graph: NNCFGraph) -> List[NNCFNode]: - """ - Returns a list of NNCFNodes to use as start nodes for activation path tracing. - - :param nncf_graph: NNCFGraph to get the start nodes. - :return: List of NNCFNodes to use as start nodes for activation path tracing. - - """ - - @staticmethod - @abstractmethod - def get_target_point_shape(nncf_graph: NNCFGraph, node: NNCFNode, target_point: TargetPoint) -> Tuple[int, ...]: - """ - Returns shape of a target point tensor. - - :param nncf_graph: NNCFGraph instance. - :param node: NNCFNode. - :param target_point: Target point of which tensor shape is seeked. - :return: Shape of target point tensor. - """ - - @staticmethod - @abstractmethod - def get_weight_quantization_axes(node: NNCFNode, target_point: TargetPoint, ndims: int) -> Tuple[int, ...]: - """ - Returns axes for per-channel quantization of weights of the node placed on a input port_id. - - :param node: Quantized node with the weight. - :param target_point: Corresponding target point. - :param ndims: Number of dimensions of weight. - :return: Axes for per-channel quantization of weights. - """ - - @staticmethod - @abstractmethod - def get_statistic_collector( - range_estimator_params: RangeEstimatorParameters, - use_abs_max: bool, - reduction_axes: Optional[Tuple[int, ...]], - aggregation_axes: Optional[Tuple[int, ...]], - inplace: bool, - num_samples: Optional[int] = None, - ) -> TensorStatisticCollectorBase: - """ - Returns backend-specific statistic collector. - - :param range_estimator_params: Parameters that specify estimators types. - :param use_abs_max: Wheather reduce absolute values of input tensors or not. - :param reduction_axes: Axes for reducer. - :param aggregation_axes: Axes for aggregator. - :param inplace: Whether to calculate statistic inplace or not. - :param num_samples: Maximum number of samples to collect. - :return: Backend-specific TensorStatisticCollectorBase for the statistics calculation. - """ - - @staticmethod - @abstractmethod - def get_weight_tensor_port_ids(node: NNCFNode, graph: NNCFGraph) -> List[Optional[int]]: - """ - Returns node's input port indices with weight tensors. - - :param node: NNCFNode to find its weight input port indices. - :param graph: NNCFGraph instance. - :return: Weights input port indices. - """ - - @staticmethod - def get_weight_name(nncf_graph: NNCFGraph, target_point: TargetPoint) -> str: - """ - Returns node's weight name corresponding to port ID. - - :param nncf_graph: NNCFGraph instance. - :param target_point: The TargetPoint instance that contains layer's information. - :return: Weight name. - """ - - @staticmethod - def should_quantize_weight(weight_name: str, quantized_weight_names: Set[str]) -> bool: - """ - Return True if weight should be quantized. - - :param weight_name: Weight name. - :param quantized_weight_names: Set containing already quantized weight names. - :return: A boolean value specifying whether a weight should be quantized. - """ - - @staticmethod - @abstractmethod - def get_ignored_metatypes(model_type: ModelType, device: TargetDevice) -> List[OperatorMetatype]: - """ - Returns ignored metatypes based on a model type and device parameters. - - :param model_type: Model type parameter. - :param device: Target device. - :return: List of ignored metatypes. - """ - - @staticmethod - @abstractmethod - def get_ignored_names_by_layer_attributes(nncf_graph: NNCFGraph) -> Set[str]: - """ - Returns names of ignored nodes based on layer_attributes. - - :param nncf_graph: NNCFGraph instance. - :return: List of ignored names. - """ - - @abstractmethod - def get_weight_nodes(self, nncf_graph: NNCFGraph) -> List[NNCFNode]: - """ - Returns nodes that have weights. - - :param nncf_graph: Instance of NNCFGraph. - :return: All nodes with weights. - """ - - @abstractmethod - def is_matmul_with_constant(self, node: NNCFNode, nncf_graph: NNCFGraph) -> bool: - """ - Returns true if given nncf matmul node is a matmul with a constant, False otherwise. - - :param Node: Instance of NNCFNode. - :param nncf_graph: Instance of NNCFGraph. - :return: True if given nncf matmul node is a matmul with a constant, False otherwise. - """ diff --git a/nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer/openvino_quantizer.py b/nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer/openvino_quantizer.py deleted file mode 100644 index 11ce83637c2..00000000000 --- a/nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer/openvino_quantizer.py +++ /dev/null @@ -1,593 +0,0 @@ -# Copyright (c) 2024 Intel Corporation -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# http://www.apache.org/licenses/LICENSE-2.0 -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. - -import dataclasses -from copy import deepcopy -from typing import Dict, List, Optional, Set, TypeVar, Union - -import numpy as np - -import nncf -from nncf.common.graph.graph import NNCFGraph -from nncf.common.graph.patterns import GraphPattern -from nncf.common.graph.patterns.manager import PatternsManager -from nncf.common.hardware.config import get_hw_config_type -from nncf.common.insertion_point_graph import InsertionPointGraph -from nncf.common.logging import nncf_logger -from nncf.common.quantization.config_assignment import assign_qconfig_lists_to_modules -from nncf.common.quantization.quantizer_propagation.solver import QuantizerPropagationRule -from nncf.common.quantization.quantizer_propagation.solver import QuantizerPropagationSolver -from nncf.common.quantization.quantizer_propagation.structs import IgnoreReason -from nncf.common.quantization.quantizer_setup import SingleConfigQuantizerSetup -from nncf.common.quantization.structs import QuantizableWeightedLayerNode -from nncf.common.quantization.structs import QuantizationConstraints -from nncf.common.quantization.structs import QuantizationPreset -from nncf.common.quantization.structs import QuantizationScheme -from nncf.common.quantization.structs import QuantizerConfig -from nncf.common.quantization.structs import QuantizerGroup -from nncf.common.utils.backend import BackendType -from nncf.common.utils.backend import get_backend -from nncf.parameters import ModelType -from nncf.parameters import QuantizationMode -from nncf.parameters import TargetDevice -from nncf.quantization.advanced_parameters import FP8QuantizationParameters -from nncf.quantization.advanced_parameters import FP8Type -from nncf.quantization.advanced_parameters import OverflowFix -from nncf.quantization.advanced_parameters import QuantizationParameters -from nncf.experimental.common.quantization.algorithms.quantizer.quantizer import NNCFQuantizer -from nncf.quantization.passes import transform_to_inference_graph -from nncf.scopes import IgnoredScope -from nncf.scopes import get_ignored_node_names_from_ignored_scope - -TModel = TypeVar("TModel") - -DEFAULT_QCONFIG = QuantizerConfig( - num_bits=8, mode=QuantizationScheme.SYMMETRIC, signedness_to_force=None, per_channel=False -) - - -@dataclasses.dataclass -class ModeBasedDefaults: - """ - Contains default values that should be set in case of abscense. - """ - - overflow_fix: OverflowFix = OverflowFix.FIRST_LAYER - activations_quantization_params: Union[QuantizationParameters, FP8QuantizationParameters] = dataclasses.field( - default_factory=QuantizationParameters - ) - weights_quantization_params: Union[QuantizationParameters, FP8QuantizationParameters] = dataclasses.field( - default_factory=QuantizationParameters - ) - - -MODE_BASED_DEFAULTS = { - None: ModeBasedDefaults(), - QuantizationMode.FP8_E4M3: ModeBasedDefaults( - overflow_fix=OverflowFix.DISABLE, - activations_quantization_params=FP8QuantizationParameters(FP8Type.E4M3), - weights_quantization_params=FP8QuantizationParameters(FP8Type.E4M3), - ), - QuantizationMode.FP8_E5M2: ModeBasedDefaults( - overflow_fix=OverflowFix.DISABLE, - activations_quantization_params=FP8QuantizationParameters(FP8Type.E5M2), - weights_quantization_params=FP8QuantizationParameters(FP8Type.E5M2), - ), -} - - -class NNCFOVQuantizer(NNCFQuantizer): - def __init__( - self, - mode: Optional[QuantizationMode] = None, - preset: Optional[QuantizationPreset] = None, - target_device: TargetDevice = TargetDevice.ANY, - model_type: Optional[ModelType] = None, - ignored_scope: Optional[IgnoredScope] = None, - overflow_fix: Optional[OverflowFix] = None, - quantize_outputs: bool = False, - activations_quantization_params: Union[QuantizationParameters, FP8QuantizationParameters] = None, - weights_quantization_params: Union[QuantizationParameters, FP8QuantizationParameters] = None, - quantizer_propagation_rule: Optional[QuantizerPropagationRule] = None, - ): - """ - :param mode: Defines optimization mode for the algorithm. None by default. - :param preset: A preset controls the quantization mode (symmetric and asymmetric). - It can take the following values: - - `performance`: Symmetric quantization of weights and activations. - - `mixed`: Symmetric quantization of weights and asymmetric quantization of activations. - Default value is None. In this case, `mixed` preset is used for `transformer` - model type otherwise `performance`. - :param target_device: A target device the specificity of which will be taken - into account while compressing in order to obtain the best performance - for this type of device, defaults to TargetDevice.ANY. - :param model_type: Model type is needed to specify additional patterns - in the model. Supported only `transformer` now. - :param ignored_scope: An ignored scope that defined the list of model control - flow graph nodes to be ignored during quantization. - :param overflow_fix: This option controls whether to apply the overflow issue - fix for the 8-bit quantization. - :param quantize_outputs: Whether to insert additional quantizers right before - each of the model outputs. - :param activations_quantization_params: Quantization parameters for model - activations. - :param weights_quantization_params: Quantization parameters for model weights. - :param quantizer_propagation_rule: The strategy to be used while propagating and merging quantizers. - """ - self._target_device = target_device - self._mode = mode - self._model_type = model_type - self._overflow_fix = overflow_fix - self._quantize_outputs = quantize_outputs - self._activations_quantization_params = activations_quantization_params - self._weights_quantization_params = weights_quantization_params - self._preset = preset - self._ignored_scope = IgnoredScope() if ignored_scope is None else ignored_scope - self.quantizer_propagation_rule = quantizer_propagation_rule - - # preset definition - if self._preset is None: - if model_type == ModelType.TRANSFORMER: - self._preset = QuantizationPreset.MIXED - else: - self._preset = QuantizationPreset.PERFORMANCE - - self._override_device() - self._set_mode_based_defaults() - self._review_mode_based_defaults() - - self._quantization_params = { - QuantizerGroup.WEIGHTS: self._weights_quantization_params, - QuantizerGroup.ACTIVATIONS: self._activations_quantization_params, - } - - # Calculates global quantizer constraints - self._global_quantizer_constraints = {} - for quantizer_group in QuantizerGroup: - self._global_quantizer_constraints[quantizer_group] = self._get_quantizer_constraints( - quantizer_group, self._preset, self._quantization_params[quantizer_group] - ) - - self._algorithm_key = f"MMQ_{hash(self)}" - - def _override_device(self) -> None: - """ - Overrides NPU device to use CPU quantization scheme. - """ - if self._target_device == TargetDevice.NPU: - act_bits, weight_bits = 8, 8 - if self._activations_quantization_params and self._activations_quantization_params.num_bits: - act_bits = self._activations_quantization_params.num_bits - if self._weights_quantization_params and self._weights_quantization_params.num_bits: - weight_bits = self._weights_quantization_params.num_bits - - if act_bits == 8 and weight_bits == 8: - self._target_device == TargetDevice.CPU - nncf_logger.debug("Target device NPU was changed to CPU!") - - def _set_mode_based_defaults(self) -> None: - """ - Sets defaults for the algorithms based on the provided mode. - """ - mode_based_defaults = MODE_BASED_DEFAULTS[self._mode] - for field in dataclasses.fields(mode_based_defaults): - self_name = "_" + field.name - default_value = getattr(mode_based_defaults, field.name) - if getattr(self, self_name) is None: - setattr(self, self_name, default_value) - - def _review_mode_based_defaults(self): - """ - Reviews default values because mode option doesn't support them. - """ - if self._mode in (QuantizationMode.FP8_E4M3, QuantizationMode.FP8_E5M2): - nncf_logger.warning(f"You're using experimental option mode with {self._mode} value.") - - if self._preset != QuantizationPreset.PERFORMANCE: - raise nncf.ParameterNotSupportedError( - f"preset option with {self._preset} value is not supported with the mode option!" - ) - - if self._target_device not in [TargetDevice.CPU, TargetDevice.ANY]: - raise nncf.ParameterNotSupportedError( - f"target_device option with {self._target_device} value is not supported with the mode option!" - ) - - if self._overflow_fix != OverflowFix.DISABLE: - raise nncf.ParameterNotSupportedError( - f"overflow_fix option with {self._overflow_fix} value is not supported with the mode option!" - ) - - if self._quantize_outputs: - raise nncf.ParameterNotSupportedError("quantize_outputs option is not supported with the mode option!") - - if isinstance(self._weights_quantization_params, QuantizationParameters): - raise nncf.ParameterNotSupportedError( - "quantization_params option for weights with " - f"{self._weights_quantization_params} " - "value is not supported with the mode option!" - ) - - if isinstance(self._activations_quantization_params, QuantizationParameters): - raise nncf.ParameterNotSupportedError( - "quantization_params option for activations with " - f"{self._activations_quantization_params} " - "value is not supported with the mode option!" - ) - elif self._mode is None: - if isinstance(self._weights_quantization_params, FP8QuantizationParameters): - raise nncf.ParameterNotSupportedError( - "quantization_params option for weights with " - f"{self._weights_quantization_params} " - "value is not supported with the mode: None option!" - ) - - if isinstance(self._activations_quantization_params, FP8QuantizationParameters): - raise nncf.ParameterNotSupportedError( - "quantization_params option for activations with " - f"{self._activations_quantization_params} " - "value is not supported with the mode: None option!" - ) - - @property - def available_backends(self) -> List[BackendType]: - return [BackendType.ONNX, BackendType.OPENVINO, BackendType.TORCH, BackendType.TORCH_FX] - - def _get_quantizer_constraints( - self, - group: QuantizerGroup, - preset: QuantizationPreset, - quantization_params: Union[QuantizationParameters, FP8QuantizationParameters], - ) -> QuantizationConstraints: - """ - Returns QuantizationConstraints for the provided quantizer group. - - :param group: Quantizer group. - :param preset: Quantization preset. - :param quantization_params: Quantization parameters. - :return: QuantizationConstraints. - """ - constraints = {"mode": preset.get_params_configured_by_preset(group)["mode"]} - if quantization_params is None: - return QuantizationConstraints(**constraints) - - if isinstance(quantization_params, FP8QuantizationParameters): - if self._mode is None: - raise nncf.InternalError( - f"FP8QuantizationParameters for {group.value} can not be used without QuantizationMode option!" - ) - return QuantizationConstraints(**constraints) - - if quantization_params.mode is not None: - constraints["mode"] = quantization_params.mode - if quantization_params.num_bits is not None: - constraints["num_bits"] = quantization_params.num_bits - if quantization_params.per_channel is not None: - constraints["per_channel"] = quantization_params.per_channel - if quantization_params.signedness_to_force is not None: - constraints["signedness_to_force"] = quantization_params.signedness_to_force - - return QuantizationConstraints(**constraints) - - def _set_backend_entity(self, model: TModel) -> None: - """ - Creates a helper class with a backed-specific logic of the algorithm - - :param model: backend-specific input model - """ - model_backend = get_backend(model) - if model_backend == BackendType.TORCH_FX: - from nncf.experimental.common.quantization.algorithms.quantizer.openvino_quantizer.torch_fx_backend import OpenVINOQuantizerBackend - - self._backend_entity = OpenVINOQuantizerBackend() - else: - raise nncf.UnsupportedBackendError( - "Cannot return backend-specific entity because {} is not supported!".format(model_backend.value) - ) - - def _get_default_qconfig(self, constraints: QuantizationConstraints = None) -> QuantizerConfig: - """ - Returns default quantizer configuration, based on the provided constraints. - - :param constraints: Quantization constraints. - :return: Quantizer config. - """ - qconfig = deepcopy(DEFAULT_QCONFIG) - if constraints is not None: - qconfig = constraints.apply_constraints_to(qconfig) - return qconfig - - def _get_ignored_names( - self, nncf_graph: NNCFGraph, inference_nncf_graph: NNCFGraph, ignored_patterns: GraphPattern - ) -> Dict[str, IgnoreReason]: - """ - Returns all node names that are ignored for quantization: - Firstly, the ignored names are obtained from user-defined ignored the scope. - Secondly, the ignored names are updated from model_type parameter. - Lastly, the ignored names are updated from ignored_patterns. - - :param nncf_graph: NNCFGraph instance. - :param inference_nncf_graph: Inference graph without constant flows. - :param ignored_patterns: Ignored patterns. - :return: Ignored node names and ignore reason for quantization. - """ - user_ignored_names = get_ignored_node_names_from_ignored_scope( - self._ignored_scope, nncf_graph, strict=self._ignored_scope.validate - ) - autogenerated_ignored_names = self._get_ignored_names_by_ignored_patterns( - inference_nncf_graph, ignored_patterns - ) - autogenerated_ignored_names |= self._backend_entity.get_ignored_names_by_layer_attributes(inference_nncf_graph) - autogenerated_ignored_names |= self._get_ignored_names_by_algorithm(inference_nncf_graph) - ignored_names = {name: IgnoreReason.AUTOGENERATED for name in autogenerated_ignored_names} - # User ignored scope has higher priority - ignored_names.update({name: IgnoreReason.USER_REQUESTED for name in user_ignored_names}) - return ignored_names - - def _get_ignored_names_by_ignored_patterns( - self, inference_nncf_graph: NNCFGraph, ignored_patterns: GraphPattern - ) -> Set[str]: - """ - Returns node names matched ignored_patterns. - - :param nncf_graph: Inference graph without constant flows. - :param ignored_patterns: Ignored patterns. - :return: IgnoredScope with all node names matched ignored_patterns. - """ - nncf_node_names = set() - for subgraph in inference_nncf_graph.find_matching_subgraphs(ignored_patterns, strict=False): - for nncf_node in subgraph: - nncf_node_names.add(nncf_node.node_name) - return nncf_node_names - - def _get_ignored_names_by_algorithm(self, inference_nncf_graph: NNCFGraph) -> Set[str]: - """ - Returns node names for ignored_algorithms matched `quantization`. - - :param inference_nncf_graph: Inference NNCFGraph instance. - :return: IgnoredScope with corresponded nodes. - """ - nncf_node_names = set() - for nncf_node in inference_nncf_graph.get_all_nodes(): - if "ptq_quantization" in nncf_node.ignored_algorithms: - nncf_node_names.add(nncf_node.node_name) - return nncf_node_names - - def _get_scope_overrides(self, inference_nncf_graph: NNCFGraph) -> Dict: - """ - Returns a dictionary of quantization configuration overrides for inputs to matching operation nodes. - - :param inference_nncf_graph: Inference NNCFGraph instance. - :return: A dictionary of quantization configuration overrides for inputs to matching operation nodes. - """ - scaled_dot_product_attention_node_names = [ - node.node_name - for node in inference_nncf_graph.get_nodes_by_metatypes( - self._backend_entity.scaled_dot_product_attention_metatypes - ) - ] - - scope_overrides_activations = {} - for node_name in scaled_dot_product_attention_node_names: - scope_overrides_activations[node_name] = {"mode": "symmetric"} - return {"activations": scope_overrides_activations} - - def _get_quantizer_setup( - self, - nncf_graph: NNCFGraph, - inference_nncf_graph: NNCFGraph, - hw_patterns: GraphPattern, - ignored_patterns: GraphPattern, - ) -> SingleConfigQuantizerSetup: - """ - Returns SingleConfigQuantizerSetup instance based on the input NNCFGraph. - - :param nncf_graph: NNCFGraph instance. - :param hw_patterns: Hardware patterns. - :param ignored_patterns: Ignored patterns. - :return: SingleConfigQuantizerSetup for the current NNCFGraph entity. - """ - hw_config_type = get_hw_config_type(self._target_device.value) - hw_config_path = self._backend_entity.hw_config.get_path_to_hw_config(hw_config_type) - hw_config = self._backend_entity.hw_config.from_json(hw_config_path) - - ignored_names = self._get_ignored_names(nncf_graph, inference_nncf_graph, ignored_patterns) - weight_nodes = self._backend_entity.get_weight_nodes(nncf_graph) - - default_weight_qconfig = self._get_default_qconfig(self._global_quantizer_constraints[QuantizerGroup.WEIGHTS]) - weighted_node_and_qconf_lists = assign_qconfig_lists_to_modules( - nodes_with_weights=weight_nodes, - default_weight_qconfig=default_weight_qconfig, - global_weight_constraints=self._global_quantizer_constraints[QuantizerGroup.WEIGHTS], - scope_overrides_dict=None, - hw_config=hw_config, - ) - quantizable_layer_nodes = [ - QuantizableWeightedLayerNode(node, qconf_list) for node, qconf_list in weighted_node_and_qconf_lists.items() - ] - - scope_overrides = self._get_scope_overrides(inference_nncf_graph) - - ip_graph = InsertionPointGraph(inference_nncf_graph) - ip_graph = ip_graph.get_ip_graph_with_merged_hw_optimized_operations(hw_patterns) - post_processing_types = self._backend_entity.post_processing_metatypes - metatypes_to_ignore = self._backend_entity.get_ignored_metatypes(self._model_type, self._target_device) - solver = QuantizerPropagationSolver( - activation_ignored_scopes=ignored_names, - weight_ignored_scopes=list(ignored_names.keys()), - hw_config=hw_config, - default_trait_to_metatype_map=self._backend_entity.quant_trait_op_dict, - propagation_strategy=self.quantizer_propagation_rule, - default_qconfig_list=[ - self._get_default_qconfig(self._global_quantizer_constraints[QuantizerGroup.ACTIVATIONS]) - ], - quantizable_layer_nodes=quantizable_layer_nodes, - quantize_outputs=self._quantize_outputs, - global_constraints=self._global_quantizer_constraints, - post_processing_marker_metatypes=post_processing_types, - metatypes_to_ignore=metatypes_to_ignore, - scales_unification_map=self._backend_entity.scales_unification_map, - scope_overrides=scope_overrides, - ) - - quantization_proposal = solver.run_on_ip_graph(ip_graph, self._backend_entity.elementwise_metatypes) - multi_config_setup = quantization_proposal.quantizer_setup - single_config_setup = multi_config_setup.select_first_qconfig_for_each_point() - finalized_proposal = quantization_proposal.finalize(single_config_setup) - final_setup = solver.get_final_quantizer_setup(finalized_proposal) - return final_setup - - def get_quantization_setup(self, model: TModel, nncf_graph: NNCFGraph) -> SingleConfigQuantizerSetup: - """ - Initializes a cache, finds quantization target points and them puts in the cache. - - :param model: Backend-specific model, for which Quantization Target Points are being seek. - :param nncf_graph: NNCFGraph instance. - :return: Mapping of quantization target points with associated quantization configuration, - along with target points for scale unification. - """ - self._set_backend_entity(model) - backend = get_backend(model) - device = self._target_device - model_type = self._model_type - ignored_patterns = PatternsManager.get_full_ignored_pattern_graph( - backend=backend, device=device, model_type=model_type - ) - hw_patterns = PatternsManager.get_full_hw_pattern_graph(backend=backend, device=device, model_type=model_type) - - inference_nncf_graph = transform_to_inference_graph( - deepcopy(nncf_graph), - self._backend_entity.get_start_nodes_for_activation_path_tracing(nncf_graph), - self._backend_entity.shapeof_metatypes, - self._backend_entity.dropout_metatypes, - self._backend_entity.preserved_metatypes, - ) - - quantizer_setup = self._get_quantizer_setup(nncf_graph, inference_nncf_graph, hw_patterns, ignored_patterns) - self._apply_model_type_pass(self._model_type, quantizer_setup, nncf_graph) - self._apply_device_pass(self._target_device, quantizer_setup, inference_nncf_graph) - return quantizer_setup - - def _apply_model_type_pass( - self, model_type: Optional[ModelType], quantizer_setup: SingleConfigQuantizerSetup, nncf_graph: NNCFGraph - ) -> None: - """ - Applies changes in-place into quantizer setup based on model_type and device parameters. - - :param model_type: Model type parameter. - :param quantizer_setup: Quantizer setup which considered to update. - :param nncf_graph: Instance of NNCFGraph. - :return: None - """ - if model_type == ModelType.TRANSFORMER: - for quantization_point in quantizer_setup.quantization_points.values(): - if quantization_point.is_activation_quantization_point(): - for node_name in quantization_point.directly_quantized_operator_node_names: - node = nncf_graph.get_node_by_name(node_name) - if node.metatype not in self._backend_entity.mat_mul_metatypes: - continue - if ( - quantization_point.qconfig.mode != QuantizationScheme.SYMMETRIC - and not self._backend_entity.is_matmul_with_constant(node, nncf_graph) - ): - quantization_point.qconfig.mode = QuantizationScheme.SYMMETRIC - nncf_logger.debug( - f"Update quantization mode for the node {node_name}" - f" to the symmetric due to ModelType parameter." - ) - - def _apply_device_pass( - self, target_device: TargetDevice, quantizer_setup: SingleConfigQuantizerSetup, nncf_graph: NNCFGraph - ) -> None: - """ - This method applies model post-processing device passes to SingleConfigQuantizerSetup in-place. - - :param target_device: TargetDevice instance. - :param quantizer_setup: SingleConfigQuantizerSetup instance to update. - :param nncf_graph: NNCFGraph. - :return: None. - """ - - passes_map = {TargetDevice.CPU_SPR: self._apply_spr_pass} - - if target_device not in passes_map: - return - - passes_map[target_device](quantizer_setup, nncf_graph) - - def _apply_spr_pass( - self, quantizer_setup: SingleConfigQuantizerSetup, nncf_graph: NNCFGraph - ) -> SingleConfigQuantizerSetup: - """ - Applies CPU_SPR-related pass. - The main action is to remove one of the quantizers before elementwise layer (e.g. Add). - This action allows to get performance boost on SPR devices. - - :param quantizer_setup: SingleConfigQuantizerSetup instance to update. - :param nncf_graph: NNCFGraph instance to update. - :return: Modified SingleConfigQuantizerSetup. - """ - - def _is_node_after_producers(node): - input_node = node - while True: - input_node = nncf_graph.get_previous_nodes(input_node) - if len(input_node) > 1: - return False - input_node = input_node[0] - if input_node.metatype in producer_metatypes: - return True - - producer_metatypes = ( - self._backend_entity.conv_metatypes - + self._backend_entity.mat_mul_metatypes - + self._backend_entity.group_conv_metatypes - ) - - quantizer_setup_map = { - p.insertion_point.target_node_name: q_key for q_key, p in quantizer_setup.quantization_points.items() - } - - # Walking through all Add layers. - for add_node in nncf_graph.get_nodes_by_metatypes(self._backend_entity.add_metatypes): - add_inputs = nncf_graph.get_previous_nodes(add_node) - - # Filtering Add based on it's input. - # Need to find Add layer only with two activations as input. - if len(add_inputs) == 2 and all(n.node_name in quantizer_setup_map for n in add_inputs): - # Sorting of the inputs based on length of input's consumer in descending order. - add_inputs.sort(key=lambda n: len(nncf_graph.get_next_nodes(n)), reverse=True) - fq_1_producer, fq_2_producer = add_inputs - fq_1_q_key = quantizer_setup_map[fq_1_producer.node_name] - fq_2_q_key = quantizer_setup_map[fq_2_producer.node_name] - - # In the case of the two quantizers where one of them produces data into branching, - # it needs to remove the quantizer without branching after it. - if ( - len(nncf_graph.get_next_nodes(fq_1_producer)) > 1 - and len(nncf_graph.get_next_nodes(fq_2_producer)) == 1 - ): - quantizer_setup.discard(fq_2_q_key, True) - continue - - # In the case of the two quantizers without the branching after them, - # it needs to check that all quantizers follows after producer nodes. - if _is_node_after_producers(fq_1_producer) and _is_node_after_producers(fq_2_producer): - fq_1_prod_shape = np.prod(nncf_graph.get_output_edges_by_port_id(fq_1_producer, 0)[0].tensor_shape) - fq_2_prod_shape = np.prod(nncf_graph.get_output_edges_by_port_id(fq_2_producer, 0)[0].tensor_shape) - - # Then it needs to remove quantizer with the smallest shape. - if fq_1_prod_shape >= fq_2_prod_shape: - quantizer_setup.discard(fq_1_q_key, True) - else: - quantizer_setup.discard(fq_2_q_key, True) - - return quantizer_setup diff --git a/nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer/torch_fx_backend.py b/nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer/torch_fx_backend.py deleted file mode 100644 index 4115bec611e..00000000000 --- a/nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer/torch_fx_backend.py +++ /dev/null @@ -1,366 +0,0 @@ -# Copyright (c) 2024 Intel Corporation -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# http://www.apache.org/licenses/LICENSE-2.0 -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. - -from typing import Dict, List, Optional, Set, Tuple - -import torch -from torch.quantization.fake_quantize import FakeQuantize - -import nncf -import nncf.torch.graph.operator_metatypes as om -from nncf.common.graph.graph import NNCFGraph -from nncf.common.graph.graph import NNCFNode -from nncf.common.graph.operator_metatypes import OperatorMetatype -from nncf.common.graph.transformations.commands import TargetType -from nncf.common.graph.transformations.commands import TransformationCommand -from nncf.common.hardware.config import HWConfig -from nncf.common.quantization.structs import QuantizationScheme as QuantizationMode -from nncf.common.quantization.structs import QuantizerConfig -from nncf.experimental.common.tensor_statistics.collectors import AGGREGATORS_MAP -from nncf.experimental.common.tensor_statistics.collectors import TensorCollector -from nncf.experimental.common.tensor_statistics.statistics import MinMaxTensorStatistic -from nncf.experimental.torch.fx.commands import FXApplyTransformationCommand -from nncf.experimental.torch.fx.model_utils import get_target_point -from nncf.experimental.torch.fx.transformations import qdq_insertion_transformation_builder -from nncf.parameters import ModelType -from nncf.parameters import TargetDevice -from nncf.quantization.advanced_parameters import StatisticsType -from nncf.experimental.common.quantization.algorithms.quantizer.openvino_quantizer.backend import OpenVINOQuantizerBackend -from nncf.quantization.fake_quantize import FakeConvertParameters -from nncf.quantization.fake_quantize import FakeQuantizeParameters -from nncf.quantization.range_estimator import AggregatorType -from nncf.quantization.range_estimator import RangeEstimatorParameters -from nncf.torch.graph.graph import PTNNCFGraph -from nncf.torch.graph.graph import PTTargetPoint -from nncf.torch.graph.operator_metatypes import ELEMENTWISE_OPERATIONS -from nncf.torch.graph.transformations.commands import PTSharedFnInsertionCommand -from nncf.torch.hardware.config import PTHWConfig -from nncf.torch.model_graph_manager import get_weight_tensor_port_ids -from nncf.torch.nncf_network import NNCFNetwork -from nncf.torch.quantization.default_quantization import DEFAULT_PT_QUANT_TRAIT_TO_OP_DICT -from nncf.torch.quantization.layers import QUANTIZATION_MODULES -from nncf.torch.quantization.layers import AsymmetricQuantizer -from nncf.torch.quantization.layers import BaseQuantizer -from nncf.torch.quantization.layers import PTQuantizerSpec -from nncf.torch.quantization.layers import get_scale_shape -from nncf.torch.quantization.strip import convert_to_torch_fakequantizer -from nncf.torch.tensor_statistics.collectors import PT_REDUCERS_MAP - -FXMinMaxAlgoBackend = None - - -class FXNNCFQuantizerBackend(OpenVINOQuantizerBackend): - - @property - def preserved_metatypes(self) -> List[OperatorMetatype]: - return [] - - @property - def mat_mul_metatypes(self) -> List[OperatorMetatype]: - return [om.PTLinearMetatype, om.PTMatMulMetatype] - - @property - def post_processing_metatypes(self) -> List[OperatorMetatype]: - return [] - - @property - def shapeof_metatypes(self) -> List[OperatorMetatype]: - return [] - - @property - def dropout_metatypes(self) -> List[OperatorMetatype]: - return [om.PTDropoutMetatype] - - @property - def read_variable_metatypes(self) -> List[OperatorMetatype]: - return [] - - @property - def conv_metatypes(self) -> List[OperatorMetatype]: - return [om.PTConv1dMetatype, om.PTConv2dMetatype, om.PTConv3dMetatype] - - @property - def elementwise_metatypes(self) -> List[OperatorMetatype]: - return ELEMENTWISE_OPERATIONS - - @property - def overflow_fix_metatypes(self) -> List[OperatorMetatype]: - return [ - om.PTConv1dMetatype, - om.PTConv2dMetatype, - om.PTConv3dMetatype, - om.PTLinearMetatype, - om.PTConvTranspose1dMetatype, - om.PTConvTranspose2dMetatype, - om.PTConvTranspose3dMetatype, - ] - - @property - def add_metatypes(self) -> List[OperatorMetatype]: - return [om.PTAddMetatype] - - @property - def group_conv_metatypes(self) -> List[OperatorMetatype]: - return self.conv_metatypes - - @property - def scaled_dot_product_attention_metatypes(self) -> List[OperatorMetatype]: - return [om.PTScaledDotProductAttentionMetatype] - - @property - def scales_unification_map(self) -> Dict[OperatorMetatype, OperatorMetatype]: - return {om.PTCatMetatype: self.overflow_fix_metatypes} - - @property - def hw_config(self) -> HWConfig: - return PTHWConfig - - @property - def quant_trait_op_dict(self) -> Dict[int, OperatorMetatype]: - return DEFAULT_PT_QUANT_TRAIT_TO_OP_DICT - - @staticmethod - def get_start_nodes_for_activation_path_tracing(nncf_graph: PTNNCFGraph) -> List[NNCFNode]: - return nncf_graph.get_input_nodes() - - @staticmethod - def target_point(target_type: TargetType, target_node_name: str, port_id: int) -> PTTargetPoint: - return get_target_point(target_type, target_node_name, port_id) - - @staticmethod - def create_convert_insertion_command( - target_point: PTTargetPoint, - parameters: FakeConvertParameters, - ) -> TransformationCommand: - raise nncf.InternalError("FakeConvert insertion not implemented in PyTorch backend!") - - @staticmethod - def get_target_point_shape(nncf_graph: PTNNCFGraph, node: NNCFNode, target_point: PTTargetPoint) -> Tuple[int, ...]: - return nncf_graph.get_input_shape_for_insertion_point(target_point) - - @staticmethod - def get_weight_quantization_axes(node: NNCFNode, target_point: PTTargetPoint, ndims: int) -> Tuple[int]: - # TODO(dlyakhov): support transpose conv and other cases - return (0,) - - @staticmethod - def get_statistic_collector( - range_estimator_params: RangeEstimatorParameters, - use_abs_max: bool, - reduction_axes: Optional[Tuple[int, ...]], - aggregation_axes: Optional[Tuple[int, ...]], - inplace: bool, - num_samples: Optional[int] = None, - ) -> TensorCollector: - collector = TensorCollector(MinMaxTensorStatistic) - for params, container_key in zip( - [range_estimator_params.min, range_estimator_params.max], - [MinMaxTensorStatistic.MIN_STAT, MinMaxTensorStatistic.MAX_STAT], - ): - if params.statistics_type not in PT_REDUCERS_MAP: - raise nncf.InternalError( - f"Statistic type: {params.statistics_type} is not supported for Torch PTQ backend yet." - ) - - if params.aggregator_type not in AGGREGATORS_MAP: - raise nncf.InternalError( - f"Aggregator type: {params.aggregator_type} is not supported for Torch PTQ backend yet." - ) - - statistic_type = params.statistics_type - if statistic_type in [StatisticsType.QUANTILE, StatisticsType.ABS_QUANTILE]: - # TODO(dlyakhov): merge two quantile aggregators in one - if container_key == MinMaxTensorStatistic.MIN_STAT: - quantile = params.quantile_outlier_prob - else: - quantile = 1 - params.quantile_outlier_prob - reducer = PT_REDUCERS_MAP[statistic_type](reduction_axes=reduction_axes, quantile=[quantile]) - else: - if use_abs_max and statistic_type == StatisticsType.MAX: - statistic_type = StatisticsType.ABS_MAX - reducer = PT_REDUCERS_MAP[statistic_type](reduction_axes=reduction_axes) - - kwargs = { - "num_samples": num_samples, - "aggregation_axes": aggregation_axes, - } - if params.aggregator_type in [AggregatorType.MEAN_NO_OUTLIERS, AggregatorType.MEDIAN_NO_OUTLIERS]: - kwargs.update({"quantile": params.quantile_outlier_prob}) - aggregator = AGGREGATORS_MAP[params.aggregator_type](**kwargs) - - collector.register_statistic_branch(container_key, reducer, aggregator) - return collector - - @staticmethod - def get_weight_tensor_port_ids(node: NNCFNode, graph: NNCFGraph) -> List[Optional[int]]: - return get_weight_tensor_port_ids(node, graph) - - @staticmethod - def get_weight_name(nncf_graph: NNCFGraph, target_point: PTTargetPoint) -> str: - weighted_node = nncf_graph.get_node_by_name(target_point.target_node_name) - weight_edge = nncf_graph.get_input_edge_by_port_id(weighted_node, target_point.input_port_id) - weight = weight_edge.from_node - return weight.node_name - - @staticmethod - def should_quantize_weight(weight_name: str, quantized_weight_names: Set[str]) -> bool: - # If the nodes share one weight tensor, we should have only one quantizer on that - return weight_name not in quantized_weight_names - - @staticmethod - def get_weight_config(config: QuantizerConfig, model: NNCFNetwork) -> QuantizerConfig: - return config - - @staticmethod - def _get_input_scale_shape( - nncf_graph: NNCFGraph, target_point: PTTargetPoint, per_channel: bool - ) -> Tuple[Tuple[int, ...], Tuple[int, ...], int]: - is_weights = target_point.is_weight_target_point() - if is_weights: - # TODO(dlyakhov): support transpose conv/ make channel_idx common - channel_idx = 0 - else: - channel_idx = 1 # channel dim for activations - - input_shape = nncf_graph.get_input_shape_for_insertion_point(target_point) - scale_shape = tuple( - get_scale_shape(input_shape, is_weights=is_weights, per_channel=per_channel, channel_idx=channel_idx) - ) - - return input_shape, scale_shape, channel_idx - - @staticmethod - def _create_quantizer( - quantizer_config: QuantizerConfig, - scale_shape: Tuple, - parameters: FakeQuantizeParameters, - target_type: TargetType, - ) -> FakeQuantize: - mode = quantizer_config.mode - quantizer_cls = QUANTIZATION_MODULES.get(mode) - narrow_range = target_type == TargetType.OPERATION_WITH_WEIGHTS and mode == QuantizationMode.SYMMETRIC - quantizer_spec = PTQuantizerSpec.from_config( - quantizer_config, - narrow_range=narrow_range, - scale_shape=scale_shape, - half_range=False, - logarithm_scale=False, - is_quantized_on_export=False, - compression_lr_multiplier=None, - ) - quantizer = quantizer_cls(quantizer_spec) - - # Fill it with minmax - # TODO(dlyakhov) Prevent creation of intermediate objects like nncf quantizer. - FXMinMaxAlgoBackend._fill_quantizer_parameters(quantizer, parameters, quantizer_spec.scale_shape) - # Convert to the torch fake quantizer - torch_fq = convert_to_torch_fakequantizer(quantizer) - return torch_fq - - @staticmethod - def _fill_quantizer_parameters(quantizer: BaseQuantizer, parameters: FakeQuantizeParameters, scale_shape) -> None: - if isinstance(quantizer, AsymmetricQuantizer): - quantizer.input_low = torch.nn.Parameter(parameters.input_low.data.reshape(scale_shape)) - input_range = parameters.input_high - parameters.input_low - # Subtract eps from the input_range to make quantizer parameters equal to - # original parameters on the forward call. - quantizer.input_range = torch.nn.Parameter((input_range.data - quantizer.eps).reshape(scale_shape)) - else: - quantizer.signed = bool(torch.any(parameters.input_low.data < 0)) - # Subtract eps from the scale to make quantizer parameters equal to - # original parameters on the forward call. - quantizer.scale = torch.nn.Parameter((parameters.input_high.data - quantizer.eps).reshape(scale_shape)) - - @staticmethod - def create_quantizer_insertion_command( - nncf_graph: NNCFGraph, - target_point: PTTargetPoint, - quantizer_config: QuantizerConfig, - parameters: FakeQuantizeParameters, - ) -> FXApplyTransformationCommand: - _, scale_shape, _ = FXMinMaxAlgoBackend._get_input_scale_shape( - nncf_graph, target_point, quantizer_config.per_channel - ) - - quantizer = FXMinMaxAlgoBackend._create_quantizer( - quantizer_config, scale_shape, parameters, target_point.target_type - ) - transformation = qdq_insertion_transformation_builder(quantizer, [target_point]) - return FXApplyTransformationCommand(transformation) - - @staticmethod - def create_unified_scales_quantizers_insertion_commands( - nncf_graph: NNCFGraph, - target_points: List[PTTargetPoint], - quantizer_config: QuantizerConfig, - parameters: FakeQuantizeParameters, - ) -> List[PTSharedFnInsertionCommand]: - _, scale_shape, _ = FXMinMaxAlgoBackend._get_input_scale_shape( - nncf_graph, target_points[0], quantizer_config.per_channel - ) - - quantizer = FXMinMaxAlgoBackend._create_quantizer( - quantizer_config, scale_shape, parameters, target_points[0].target_type - ) - - transformations = [] - for tp in target_points: - transformation = qdq_insertion_transformation_builder(quantizer, [tp]) - transformations.append(FXApplyTransformationCommand(transformation)) - return transformations - - @staticmethod - def get_ignored_metatypes(model_type: ModelType, device: TargetDevice) -> List[OperatorMetatype]: - types = [] - if model_type == ModelType.TRANSFORMER: - types = [ - om.PTAddMetatype, - om.PTPowerMetatype, - om.PTSubMetatype, - om.PTAvgPool2dMetatype, - om.PTAvgPool3dMetatype, - om.PTMeanMetatype, - om.PTSumMetatype, - om.PTReduceL2, - om.PTDivMetatype, - om.PTMaxMetatype, - om.PTSqueezeMetatype, - om.PTLayerNormMetatype, - om.PTModuleLayerNormMetatype, - om.PTGroupNormMetatype, - om.PTModuleGroupNormMetatype, - # Batchnorm - om.PTBatchNormMetatype, - om.PTModuleBatchNormMetatype, - ] - if device != TargetDevice.CPU_SPR: - types.append(om.PTMulMetatype) - return types - - @staticmethod - def get_ignored_names_by_layer_attributes(nncf_graph: NNCFGraph) -> Set[str]: - return set() - - def get_weight_nodes(self, nncf_graph: NNCFGraph) -> List[NNCFNode]: - weight_nodes_candidates = [ - node - for node in nncf_graph.get_all_nodes() - if issubclass(node.metatype, om.PTOperatorMetatype) and node.metatype.weight_port_ids - ] - weight_nodes = [] - for node in weight_nodes_candidates: - if node.metatype in self.mat_mul_metatypes and not self.is_matmul_with_constant(node, nncf_graph): - continue - weight_nodes.append(node) - return weight_nodes - - def is_matmul_with_constant(self, node: NNCFNode, nncf_graph: NNCFGraph) -> bool: - return node.metatype in self.mat_mul_metatypes and len(get_weight_tensor_port_ids(node, nncf_graph)) > 0 diff --git a/nncf/quantization/algorithms/min_max/algorithm.py b/nncf/quantization/algorithms/min_max/algorithm.py index b6728b292bf..4388cae02c6 100644 --- a/nncf/quantization/algorithms/min_max/algorithm.py +++ b/nncf/quantization/algorithms/min_max/algorithm.py @@ -725,12 +725,11 @@ def _get_activation_quantization_target_point( def find_quantization_setup(self, model: TModel, nncf_graph: NNCFGraph) -> SingleConfigQuantizerSetup: """ - Initializes a cache, finds quantization target points and them puts in the cache. + Builds SingleConfigQuantizerSetup for the given model. :param model: Backend-specific model, for which Quantization Target Points are being seek. :param nncf_graph: NNCFGraph instance. - :return: Mapping of quantization target points with associated quantization configuration, - along with target points for scale unification. + :return: SingleConfigQuantizerSetup for the given model. """ backend = get_backend(model) device = self._target_device diff --git a/nncf/quantization/algorithms/min_max/quantizer/backend.py b/nncf/quantization/algorithms/min_max/quantizer/backend.py deleted file mode 100644 index 821d57a82ea..00000000000 --- a/nncf/quantization/algorithms/min_max/quantizer/backend.py +++ /dev/null @@ -1,323 +0,0 @@ -# Copyright (c) 2024 Intel Corporation -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# http://www.apache.org/licenses/LICENSE-2.0 -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. - -from abc import ABC -from abc import abstractmethod -from typing import Dict, List, Optional, Set, Tuple, TypeVar - -from nncf.common.graph.graph import NNCFGraph -from nncf.common.graph.graph import NNCFNode -from nncf.common.graph.operator_metatypes import OperatorMetatype -from nncf.common.graph.transformations.commands import TargetPoint -from nncf.common.graph.transformations.commands import TargetType -from nncf.common.graph.transformations.commands import TransformationCommand -from nncf.common.hardware.config import HWConfig -from nncf.common.quantization.structs import QuantizerConfig -from nncf.common.tensor_statistics.collectors import TensorStatisticCollectorBase -from nncf.parameters import ModelType -from nncf.parameters import TargetDevice -from nncf.quantization.fake_quantize import FakeConvertParameters -from nncf.quantization.fake_quantize import FakeQuantizeParameters -from nncf.quantization.range_estimator import RangeEstimatorParameters - -TModel = TypeVar("TModel") - - -class NNCFQuantizerBackend(ABC): - @property - @abstractmethod - def preserved_metatypes(self) -> List[OperatorMetatype]: - """ - Property for backend-specific metatypes that require preserving float subgraphs - when removing the ShapeOf subgraph. - """ - - @property - @abstractmethod - def mat_mul_metatypes(self) -> List[OperatorMetatype]: - """ - Property for the backend-specific MatMul metatypes. - """ - - @property - @abstractmethod - def post_processing_metatypes(self) -> List[OperatorMetatype]: - """ - Property for the backend-specific post-processing metatypes (NonMaximumSupression, TopK, etc.). - """ - - @property - @abstractmethod - def conv_metatypes(self) -> List[OperatorMetatype]: - """ - Property for the backend-specific Convolution metatypes. - """ - - @property - @abstractmethod - def shapeof_metatypes(self) -> List[OperatorMetatype]: - """ - Property for the backend-specific ShapeOf metatypes. - """ - - @property - @abstractmethod - def dropout_metatypes(self) -> List[OperatorMetatype]: - """ - Property for the backend-specific Dropout metatypes. - """ - - @property - @abstractmethod - def elementwise_metatypes(self) -> List[OperatorMetatype]: - """ - Property for the backend-specific Elementwises metatypes. - """ - - @property - @abstractmethod - def overflow_fix_metatypes(self) -> List[OperatorMetatype]: - """ - Property for the backend-specific metatypes for which overflow_fix is applicable. - """ - - @property - @abstractmethod - def add_metatypes(self) -> List[OperatorMetatype]: - """ - Property for the backend-specific metatypes that also can be interpreted as Add layer. - """ - - @property - @abstractmethod - def group_conv_metatypes(self) -> List[OperatorMetatype]: - """ - Property for the backend-specific Grouped Convolution metatypes. - """ - - @property - @abstractmethod - def scaled_dot_product_attention_metatypes(self) -> List[OperatorMetatype]: - """ - Property for the backend-specific Scaled Dot Product Attention metatypes. - """ - - @property - @abstractmethod - def scales_unification_map(self) -> Dict[OperatorMetatype, OperatorMetatype]: - """ - Property for the backend-specific metatypes that produces quantizers that might be unified. - """ - - @property - @abstractmethod - def hw_config(self) -> HWConfig: - """ - Property for the hardware backend-specific configuration. - """ - - @property - @abstractmethod - def quant_trait_op_dict(self) -> Dict[int, OperatorMetatype]: - """ - Property for the backend-specific dictionary that contains QuantizationTrait-specific metatypes. - """ - - @staticmethod - @abstractmethod - def target_point(target_type: TargetType, target_node_name: str, port_id: int) -> TargetPoint: - """ - Returns backend-specific target point. - - :param target_type: Type of the location that should be modified. - :param target_node_name: Name of the located node. - :param port_id: Port ID of the tensor for the statistics distribution. - :return: Backend-specific TargetPoint. - """ - - @staticmethod - @abstractmethod - def create_quantizer_insertion_command( - nncf_graph: NNCFGraph, - target_point: TargetPoint, - quantizer_config: QuantizerConfig, - parameters: FakeQuantizeParameters, - ) -> TransformationCommand: - """ - Returns backend-specific quantizer insertion command. - - :param nncf_graph: NNCFGraph to get input/output shapes for the target point. - :param target_point: Target location for the quantizer insertion. - :param quantizer_config: QuantizerConfig instance for the current layer. - :param parameters: FakeQuantizeParameters to calculate activation quantization parameters. - :return: Backend-specific TransformationCommand for the quantizer insertion operation. - """ - - @staticmethod - @abstractmethod - def create_unified_scales_quantizers_insertion_commands( - nncf_graph: NNCFGraph, - target_points: List[TargetPoint], - quantizer_config: QuantizerConfig, - parameters: FakeQuantizeParameters, - ) -> List[TransformationCommand]: - """ - Returns backend-specific unified scales quantizers insertion commands. - - :param nncf_graph: NNCFGraph to get input/output shapes for the target point. - :param target_points: List of target locations for the quantizers insertion. - :param quantizer_config: QuantizerConfig instance for the current layer. - :param parameters: FakeQuantizeParameters to calculate activation quantization parameters. - :return: List of backend-specific TransformationCommands - for the quantizers with unified scales insertion operations. - """ - - @staticmethod - @abstractmethod - def create_convert_insertion_command( - target_point: TargetPoint, - parameters: FakeConvertParameters, - ) -> TransformationCommand: - """ - Returns backend-specific convert insertion command. - - :param target_point: Target location for the correction. - :param parameters: FakeConvertParameters to calculate activation quantization parameters. - :return: Backend-specific TransformationCommand for the quantizer insertion operation. - """ - - @staticmethod - @abstractmethod - def get_start_nodes_for_activation_path_tracing(nncf_graph: NNCFGraph) -> List[NNCFNode]: - """ - Returns a list of NNCFNodes to use as start nodes for activation path tracing. - - :param nncf_graph: NNCFGraph to get the start nodes. - :return: List of NNCFNodes to use as start nodes for activation path tracing. - - """ - - @staticmethod - @abstractmethod - def get_target_point_shape(nncf_graph: NNCFGraph, node: NNCFNode, target_point: TargetPoint) -> Tuple[int, ...]: - """ - Returns shape of a target point tensor. - - :param nncf_graph: NNCFGraph instance. - :param node: NNCFNode. - :param target_point: Target point of which tensor shape is seeked. - :return: Shape of target point tensor. - """ - - @staticmethod - @abstractmethod - def get_weight_quantization_axes(node: NNCFNode, target_point: TargetPoint, ndims: int) -> Tuple[int, ...]: - """ - Returns axes for per-channel quantization of weights of the node placed on a input port_id. - - :param node: Quantized node with the weight. - :param target_point: Corresponding target point. - :param ndims: Number of dimensions of weight. - :return: Axes for per-channel quantization of weights. - """ - - @staticmethod - @abstractmethod - def get_statistic_collector( - range_estimator_params: RangeEstimatorParameters, - use_abs_max: bool, - reduction_axes: Optional[Tuple[int, ...]], - aggregation_axes: Optional[Tuple[int, ...]], - inplace: bool, - num_samples: Optional[int] = None, - ) -> TensorStatisticCollectorBase: - """ - Returns backend-specific statistic collector. - - :param range_estimator_params: Parameters that specify estimators types. - :param use_abs_max: Wheather reduce absolute values of input tensors or not. - :param reduction_axes: Axes for reducer. - :param aggregation_axes: Axes for aggregator. - :param inplace: Whether to calculate statistic inplace or not. - :param num_samples: Maximum number of samples to collect. - :return: Backend-specific TensorStatisticCollectorBase for the statistics calculation. - """ - - @staticmethod - @abstractmethod - def get_weight_tensor_port_ids(node: NNCFNode, graph: NNCFGraph) -> List[Optional[int]]: - """ - Returns node's input port indices with weight tensors. - - :param node: NNCFNode to find its weight input port indices. - :param graph: NNCFGraph instance. - :return: Weights input port indices. - """ - - @staticmethod - def get_weight_name(nncf_graph: NNCFGraph, target_point: TargetPoint) -> str: - """ - Returns node's weight name corresponding to port ID. - - :param nncf_graph: NNCFGraph instance. - :param target_point: The TargetPoint instance that contains layer's information. - :return: Weight name. - """ - - @staticmethod - def should_quantize_weight(weight_name: str, quantized_weight_names: Set[str]) -> bool: - """ - Return True if weight should be quantized. - - :param weight_name: Weight name. - :param quantized_weight_names: Set containing already quantized weight names. - :return: A boolean value specifying whether a weight should be quantized. - """ - - @staticmethod - @abstractmethod - def get_ignored_metatypes(model_type: ModelType, device: TargetDevice) -> List[OperatorMetatype]: - """ - Returns ignored metatypes based on a model type and device parameters. - - :param model_type: Model type parameter. - :param device: Target device. - :return: List of ignored metatypes. - """ - - @staticmethod - @abstractmethod - def get_ignored_names_by_layer_attributes(nncf_graph: NNCFGraph) -> Set[str]: - """ - Returns names of ignored nodes based on layer_attributes. - - :param nncf_graph: NNCFGraph instance. - :return: List of ignored names. - """ - - @abstractmethod - def get_weight_nodes(self, nncf_graph: NNCFGraph) -> List[NNCFNode]: - """ - Returns nodes that have weights. - - :param nncf_graph: Instance of NNCFGraph. - :return: All nodes with weights. - """ - - @abstractmethod - def is_matmul_with_constant(self, node: NNCFNode, nncf_graph: NNCFGraph) -> bool: - """ - Returns true if given nncf matmul node is a matmul with a constant, False otherwise. - - :param Node: Instance of NNCFNode. - :param nncf_graph: Instance of NNCFGraph. - :return: True if given nncf matmul node is a matmul with a constant, False otherwise. - """ diff --git a/nncf/quantization/algorithms/min_max/quantizer/torch_fx_backend.py b/nncf/quantization/algorithms/min_max/quantizer/torch_fx_backend.py deleted file mode 100644 index 9ed869b543c..00000000000 --- a/nncf/quantization/algorithms/min_max/quantizer/torch_fx_backend.py +++ /dev/null @@ -1,366 +0,0 @@ -# Copyright (c) 2024 Intel Corporation -# Licensed under the Apache License, Version 2.0 (the "License"); -# you may not use this file except in compliance with the License. -# You may obtain a copy of the License at -# http://www.apache.org/licenses/LICENSE-2.0 -# Unless required by applicable law or agreed to in writing, software -# distributed under the License is distributed on an "AS IS" BASIS, -# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. -# See the License for the specific language governing permissions and -# limitations under the License. - -from typing import Dict, List, Optional, Set, Tuple - -import torch -from torch.quantization.fake_quantize import FakeQuantize - -import nncf -import nncf.torch.graph.operator_metatypes as om -from nncf.common.graph.graph import NNCFGraph -from nncf.common.graph.graph import NNCFNode -from nncf.common.graph.operator_metatypes import OperatorMetatype -from nncf.common.graph.transformations.commands import TargetType -from nncf.common.graph.transformations.commands import TransformationCommand -from nncf.common.hardware.config import HWConfig -from nncf.common.quantization.structs import QuantizationScheme as QuantizationMode -from nncf.common.quantization.structs import QuantizerConfig -from nncf.experimental.common.tensor_statistics.collectors import AGGREGATORS_MAP -from nncf.experimental.common.tensor_statistics.collectors import TensorCollector -from nncf.experimental.common.tensor_statistics.statistics import MinMaxTensorStatistic -from nncf.experimental.torch.fx.commands import FXApplyTransformationCommand -from nncf.experimental.torch.fx.model_utils import get_target_point -from nncf.experimental.torch.fx.transformations import qdq_insertion_transformation_builder -from nncf.parameters import ModelType -from nncf.parameters import TargetDevice -from nncf.quantization.advanced_parameters import StatisticsType -from nncf.quantization.algorithms.min_max.quantizer.backend import NNCFQuantizerBackend -from nncf.quantization.fake_quantize import FakeConvertParameters -from nncf.quantization.fake_quantize import FakeQuantizeParameters -from nncf.quantization.range_estimator import AggregatorType -from nncf.quantization.range_estimator import RangeEstimatorParameters -from nncf.torch.graph.graph import PTNNCFGraph -from nncf.torch.graph.graph import PTTargetPoint -from nncf.torch.graph.operator_metatypes import ELEMENTWISE_OPERATIONS -from nncf.torch.graph.transformations.commands import PTSharedFnInsertionCommand -from nncf.torch.hardware.config import PTHWConfig -from nncf.torch.model_graph_manager import get_weight_tensor_port_ids -from nncf.torch.nncf_network import NNCFNetwork -from nncf.torch.quantization.default_quantization import DEFAULT_PT_QUANT_TRAIT_TO_OP_DICT -from nncf.torch.quantization.layers import QUANTIZATION_MODULES -from nncf.torch.quantization.layers import AsymmetricQuantizer -from nncf.torch.quantization.layers import BaseQuantizer -from nncf.torch.quantization.layers import PTQuantizerSpec -from nncf.torch.quantization.layers import get_scale_shape -from nncf.torch.quantization.strip import convert_to_torch_fakequantizer -from nncf.torch.tensor_statistics.collectors import PT_REDUCERS_MAP - -FXMinMaxAlgoBackend = None - - -class FXNNCFQuantizerBackend(NNCFQuantizerBackend): - - @property - def preserved_metatypes(self) -> List[OperatorMetatype]: - return [] - - @property - def mat_mul_metatypes(self) -> List[OperatorMetatype]: - return [om.PTLinearMetatype, om.PTMatMulMetatype] - - @property - def post_processing_metatypes(self) -> List[OperatorMetatype]: - return [] - - @property - def shapeof_metatypes(self) -> List[OperatorMetatype]: - return [] - - @property - def dropout_metatypes(self) -> List[OperatorMetatype]: - return [om.PTDropoutMetatype] - - @property - def read_variable_metatypes(self) -> List[OperatorMetatype]: - return [] - - @property - def conv_metatypes(self) -> List[OperatorMetatype]: - return [om.PTConv1dMetatype, om.PTConv2dMetatype, om.PTConv3dMetatype] - - @property - def elementwise_metatypes(self) -> List[OperatorMetatype]: - return ELEMENTWISE_OPERATIONS - - @property - def overflow_fix_metatypes(self) -> List[OperatorMetatype]: - return [ - om.PTConv1dMetatype, - om.PTConv2dMetatype, - om.PTConv3dMetatype, - om.PTLinearMetatype, - om.PTConvTranspose1dMetatype, - om.PTConvTranspose2dMetatype, - om.PTConvTranspose3dMetatype, - ] - - @property - def add_metatypes(self) -> List[OperatorMetatype]: - return [om.PTAddMetatype] - - @property - def group_conv_metatypes(self) -> List[OperatorMetatype]: - return self.conv_metatypes - - @property - def scaled_dot_product_attention_metatypes(self) -> List[OperatorMetatype]: - return [om.PTScaledDotProductAttentionMetatype] - - @property - def scales_unification_map(self) -> Dict[OperatorMetatype, OperatorMetatype]: - return {om.PTCatMetatype: self.overflow_fix_metatypes} - - @property - def hw_config(self) -> HWConfig: - return PTHWConfig - - @property - def quant_trait_op_dict(self) -> Dict[int, OperatorMetatype]: - return DEFAULT_PT_QUANT_TRAIT_TO_OP_DICT - - @staticmethod - def get_start_nodes_for_activation_path_tracing(nncf_graph: PTNNCFGraph) -> List[NNCFNode]: - return nncf_graph.get_input_nodes() - - @staticmethod - def target_point(target_type: TargetType, target_node_name: str, port_id: int) -> PTTargetPoint: - return get_target_point(target_type, target_node_name, port_id) - - @staticmethod - def create_convert_insertion_command( - target_point: PTTargetPoint, - parameters: FakeConvertParameters, - ) -> TransformationCommand: - raise nncf.InternalError("FakeConvert insertion not implemented in PyTorch backend!") - - @staticmethod - def get_target_point_shape(nncf_graph: PTNNCFGraph, node: NNCFNode, target_point: PTTargetPoint) -> Tuple[int, ...]: - return nncf_graph.get_input_shape_for_insertion_point(target_point) - - @staticmethod - def get_weight_quantization_axes(node: NNCFNode, target_point: PTTargetPoint, ndims: int) -> Tuple[int]: - # TODO(dlyakhov): support transpose conv and other cases - return (0,) - - @staticmethod - def get_statistic_collector( - range_estimator_params: RangeEstimatorParameters, - use_abs_max: bool, - reduction_axes: Optional[Tuple[int, ...]], - aggregation_axes: Optional[Tuple[int, ...]], - inplace: bool, - num_samples: Optional[int] = None, - ) -> TensorCollector: - collector = TensorCollector(MinMaxTensorStatistic) - for params, container_key in zip( - [range_estimator_params.min, range_estimator_params.max], - [MinMaxTensorStatistic.MIN_STAT, MinMaxTensorStatistic.MAX_STAT], - ): - if params.statistics_type not in PT_REDUCERS_MAP: - raise nncf.InternalError( - f"Statistic type: {params.statistics_type} is not supported for Torch PTQ backend yet." - ) - - if params.aggregator_type not in AGGREGATORS_MAP: - raise nncf.InternalError( - f"Aggregator type: {params.aggregator_type} is not supported for Torch PTQ backend yet." - ) - - statistic_type = params.statistics_type - if statistic_type in [StatisticsType.QUANTILE, StatisticsType.ABS_QUANTILE]: - # TODO(dlyakhov): merge two quantile aggregators in one - if container_key == MinMaxTensorStatistic.MIN_STAT: - quantile = params.quantile_outlier_prob - else: - quantile = 1 - params.quantile_outlier_prob - reducer = PT_REDUCERS_MAP[statistic_type](reduction_axes=reduction_axes, quantile=[quantile]) - else: - if use_abs_max and statistic_type == StatisticsType.MAX: - statistic_type = StatisticsType.ABS_MAX - reducer = PT_REDUCERS_MAP[statistic_type](reduction_axes=reduction_axes) - - kwargs = { - "num_samples": num_samples, - "aggregation_axes": aggregation_axes, - } - if params.aggregator_type in [AggregatorType.MEAN_NO_OUTLIERS, AggregatorType.MEDIAN_NO_OUTLIERS]: - kwargs.update({"quantile": params.quantile_outlier_prob}) - aggregator = AGGREGATORS_MAP[params.aggregator_type](**kwargs) - - collector.register_statistic_branch(container_key, reducer, aggregator) - return collector - - @staticmethod - def get_weight_tensor_port_ids(node: NNCFNode, graph: NNCFGraph) -> List[Optional[int]]: - return get_weight_tensor_port_ids(node, graph) - - @staticmethod - def get_weight_name(nncf_graph: NNCFGraph, target_point: PTTargetPoint) -> str: - weighted_node = nncf_graph.get_node_by_name(target_point.target_node_name) - weight_edge = nncf_graph.get_input_edge_by_port_id(weighted_node, target_point.input_port_id) - weight = weight_edge.from_node - return weight.node_name - - @staticmethod - def should_quantize_weight(weight_name: str, quantized_weight_names: Set[str]) -> bool: - # If the nodes share one weight tensor, we should have only one quantizer on that - return weight_name not in quantized_weight_names - - @staticmethod - def get_weight_config(config: QuantizerConfig, model: NNCFNetwork) -> QuantizerConfig: - return config - - @staticmethod - def _get_input_scale_shape( - nncf_graph: NNCFGraph, target_point: PTTargetPoint, per_channel: bool - ) -> Tuple[Tuple[int, ...], Tuple[int, ...], int]: - is_weights = target_point.is_weight_target_point() - if is_weights: - # TODO(dlyakhov): support transpose conv/ make channel_idx common - channel_idx = 0 - else: - channel_idx = 1 # channel dim for activations - - input_shape = nncf_graph.get_input_shape_for_insertion_point(target_point) - scale_shape = tuple( - get_scale_shape(input_shape, is_weights=is_weights, per_channel=per_channel, channel_idx=channel_idx) - ) - - return input_shape, scale_shape, channel_idx - - @staticmethod - def _create_quantizer( - quantizer_config: QuantizerConfig, - scale_shape: Tuple, - parameters: FakeQuantizeParameters, - target_type: TargetType, - ) -> FakeQuantize: - mode = quantizer_config.mode - quantizer_cls = QUANTIZATION_MODULES.get(mode) - narrow_range = target_type == TargetType.OPERATION_WITH_WEIGHTS and mode == QuantizationMode.SYMMETRIC - quantizer_spec = PTQuantizerSpec.from_config( - quantizer_config, - narrow_range=narrow_range, - scale_shape=scale_shape, - half_range=False, - logarithm_scale=False, - is_quantized_on_export=False, - compression_lr_multiplier=None, - ) - quantizer = quantizer_cls(quantizer_spec) - - # Fill it with minmax - # TODO(dlyakhov) Prevent creation of intermediate objects like nncf quantizer. - FXMinMaxAlgoBackend._fill_quantizer_parameters(quantizer, parameters, quantizer_spec.scale_shape) - # Convert to the torch fake quantizer - torch_fq = convert_to_torch_fakequantizer(quantizer) - return torch_fq - - @staticmethod - def _fill_quantizer_parameters(quantizer: BaseQuantizer, parameters: FakeQuantizeParameters, scale_shape) -> None: - if isinstance(quantizer, AsymmetricQuantizer): - quantizer.input_low = torch.nn.Parameter(parameters.input_low.data.reshape(scale_shape)) - input_range = parameters.input_high - parameters.input_low - # Subtract eps from the input_range to make quantizer parameters equal to - # original parameters on the forward call. - quantizer.input_range = torch.nn.Parameter((input_range.data - quantizer.eps).reshape(scale_shape)) - else: - quantizer.signed = bool(torch.any(parameters.input_low.data < 0)) - # Subtract eps from the scale to make quantizer parameters equal to - # original parameters on the forward call. - quantizer.scale = torch.nn.Parameter((parameters.input_high.data - quantizer.eps).reshape(scale_shape)) - - @staticmethod - def create_quantizer_insertion_command( - nncf_graph: NNCFGraph, - target_point: PTTargetPoint, - quantizer_config: QuantizerConfig, - parameters: FakeQuantizeParameters, - ) -> FXApplyTransformationCommand: - _, scale_shape, _ = FXMinMaxAlgoBackend._get_input_scale_shape( - nncf_graph, target_point, quantizer_config.per_channel - ) - - quantizer = FXMinMaxAlgoBackend._create_quantizer( - quantizer_config, scale_shape, parameters, target_point.target_type - ) - transformation = qdq_insertion_transformation_builder(quantizer, [target_point]) - return FXApplyTransformationCommand(transformation) - - @staticmethod - def create_unified_scales_quantizers_insertion_commands( - nncf_graph: NNCFGraph, - target_points: List[PTTargetPoint], - quantizer_config: QuantizerConfig, - parameters: FakeQuantizeParameters, - ) -> List[PTSharedFnInsertionCommand]: - _, scale_shape, _ = FXMinMaxAlgoBackend._get_input_scale_shape( - nncf_graph, target_points[0], quantizer_config.per_channel - ) - - quantizer = FXMinMaxAlgoBackend._create_quantizer( - quantizer_config, scale_shape, parameters, target_points[0].target_type - ) - - transformations = [] - for tp in target_points: - transformation = qdq_insertion_transformation_builder(quantizer, [tp]) - transformations.append(FXApplyTransformationCommand(transformation)) - return transformations - - @staticmethod - def get_ignored_metatypes(model_type: ModelType, device: TargetDevice) -> List[OperatorMetatype]: - types = [] - if model_type == ModelType.TRANSFORMER: - types = [ - om.PTAddMetatype, - om.PTPowerMetatype, - om.PTSubMetatype, - om.PTAvgPool2dMetatype, - om.PTAvgPool3dMetatype, - om.PTMeanMetatype, - om.PTSumMetatype, - om.PTReduceL2, - om.PTDivMetatype, - om.PTMaxMetatype, - om.PTSqueezeMetatype, - om.PTLayerNormMetatype, - om.PTModuleLayerNormMetatype, - om.PTGroupNormMetatype, - om.PTModuleGroupNormMetatype, - # Batchnorm - om.PTBatchNormMetatype, - om.PTModuleBatchNormMetatype, - ] - if device != TargetDevice.CPU_SPR: - types.append(om.PTMulMetatype) - return types - - @staticmethod - def get_ignored_names_by_layer_attributes(nncf_graph: NNCFGraph) -> Set[str]: - return set() - - def get_weight_nodes(self, nncf_graph: NNCFGraph) -> List[NNCFNode]: - weight_nodes_candidates = [ - node - for node in nncf_graph.get_all_nodes() - if issubclass(node.metatype, om.PTOperatorMetatype) and node.metatype.weight_port_ids - ] - weight_nodes = [] - for node in weight_nodes_candidates: - if node.metatype in self.mat_mul_metatypes and not self.is_matmul_with_constant(node, nncf_graph): - continue - weight_nodes.append(node) - return weight_nodes - - def is_matmul_with_constant(self, node: NNCFNode, nncf_graph: NNCFGraph) -> bool: - return node.metatype in self.mat_mul_metatypes and len(get_weight_tensor_port_ids(node, nncf_graph)) > 0 diff --git a/tests/torch/fx/test_quantizer.py b/tests/torch/fx/test_quantizer.py index f95f59a4341..9c6fbabd2a9 100644 --- a/tests/torch/fx/test_quantizer.py +++ b/tests/torch/fx/test_quantizer.py @@ -27,6 +27,7 @@ from torch.ao.quantization.quantizer.x86_inductor_quantizer import get_default_x86_inductor_quantization_config import nncf +from nncf.experimental.common.quantization.algorithms.quantizer.openvino_quantizer import OpenVINOQuantizer from nncf.experimental.torch.fx.nncf_graph_builder import GraphConverter from nncf.experimental.torch.fx.quantization.quantize_pt2e import quantize_pt2e from tests.torch import test_models @@ -71,6 +72,10 @@ def get_x86_quantizer(*args, **kwarsg) -> X86InductorQuantizer: return quantizer +def get_openvino_quantizer(*args, **kwargs) -> OpenVINOQuantizer: + return OpenVINOQuantizer(*args, **kwargs) + + TEST_MODELS_QUANIZED = ( (ModelCase(test_models.UNet, "unet", [1, 3, 224, 224]), {}, {}), (torchvision_model_case("resnet18", (1, 3, 224, 224)), {}, {}), @@ -98,65 +103,65 @@ def get_x86_quantizer(*args, **kwarsg) -> X86InductorQuantizer: ) -@pytest.mark.parametrize("quantizer_builder", [get_x86_quantizer]) -class TestModelQuantization: - @staticmethod - def _build_torch_fx_model(model_case: ModelCase) -> Tuple[torch.fx.GraphModule, torch.Tensor]: - model = model_case.model_builder() - dtype = torch.int32 if model_case.model_id == "synthetic_transformer" else torch.float32 - example_input = torch.ones(model_case.input_shape, dtype=dtype) - fx_model = get_torch_fx_model(model, example_input) - return fx_model, example_input +def _build_torch_fx_model(model_case: ModelCase) -> Tuple[torch.fx.GraphModule, torch.Tensor]: + model = model_case.model_builder() + dtype = torch.int32 if model_case.model_id == "synthetic_transformer" else torch.float32 + example_input = torch.ones(model_case.input_shape, dtype=dtype) + fx_model = get_torch_fx_model(model, example_input) + return fx_model, example_input + + +def _get_calibration_dataset(example_input: torch.Tensor) -> nncf.Dataset: + def transform_fn(data_item): + return data_item.to("cpu") + + return nncf.Dataset([example_input], transform_fn) - @staticmethod - def _get_calibration_dataset(example_input: torch.Tensor) -> nncf.Dataset: - def transform_fn(data_item): - return data_item.to("cpu") - return nncf.Dataset([example_input], transform_fn) +@pytest.mark.parametrize( + ("model_case", "quantizer_params", "pt2e_params"), + TEST_MODELS_QUANIZED, + ids=[m[0].model_id for m in TEST_MODELS_QUANIZED], +) +@pytest.mark.parametrize( + "quantizer_builder", [get_x86_quantizer, get_openvino_quantizer], ids=["X86InductorQuantizer", "OpenVINOQuantizer"] +) +def test_quantized_model( + quantizer_builder: Callable[[Tuple[Any, ...]], Quantizer], + model_case: ModelCase, + quantizer_params, + pt2e_params, +): + fx_model, example_input = _build_torch_fx_model(model_case) + calibration_dataset = _get_calibration_dataset(example_input) + + quantizer = quantizer_builder(**quantizer_params) + quantized_model = quantize_pt2e( + fx_model, + quantizer, + calibration_dataset=calibration_dataset, + fast_bias_correction=None, # BC is disabled + fold_quantize=True, + **pt2e_params, + ) - @pytest.mark.parametrize( - ("model_case", "quantizer_params", "pt2e_params"), - TEST_MODELS_QUANIZED, - ids=[m[0].model_id for m in TEST_MODELS_QUANIZED], + # Uncomment to visualize torch fx graph + # from tests.torch.fx.helpers import visualize_fx_model + # visualize_fx_model(quantized_model, f"{model_case.model_id}_int8.svg") + + nncf_graph = GraphConverter.create_nncf_graph(quantized_model) + check_graph( + nncf_graph, + get_dot_filename(model_case.model_id), + FX_QUANTIZED_DIR_NAME / quantizer.__class__.__name__, + extended=True, ) - def test_quantized_model( - self, - quantizer_builder: Callable[[Tuple[Any, ...]], Quantizer], - model_case: ModelCase, - quantizer_params, - pt2e_params, - ): - fx_model, example_input = self._build_torch_fx_model(model_case) - calibration_dataset = self._get_calibration_dataset(example_input) - - quantizer = quantizer_builder(**quantizer_params) - quantized_model = quantize_pt2e( - fx_model, - quantizer, - calibration_dataset=calibration_dataset, - fast_bias_correction=None, # BC is disabled - fold_quantize=True, - **pt2e_params, - ) - - # Uncomment to visualize torch fx graph - # from tests.torch.fx.helpers import visualize_fx_model - # visualize_fx_model(quantized_model, f"{model_case.model_id}_int8.svg") - - nncf_graph = GraphConverter.create_nncf_graph(quantized_model) - check_graph( - nncf_graph, - get_dot_filename(model_case.model_id), - FX_QUANTIZED_DIR_NAME / quantizer.__class__.__name__, - extended=True, - ) - - # Uncomment to visualize reference graphs - # from torch.ao.quantization.quantize_pt2e import convert_pt2e - # from torch.ao.quantization.quantize_pt2e import prepare_pt2e - # prepared_model = prepare_pt2e(fx_model, quantizer) - # prepared_model(example_input) - # ao_quantized_model = convert_pt2e(prepared_model) - # ao_nncf_graph = GraphConverter.create_nncf_graph(ao_quantized_model) - # ao_nncf_graph.visualize_graph("ao_" + get_dot_filename(model_case.model_id)) + + # Uncomment to visualize reference graphs + # from torch.ao.quantization.quantize_pt2e import convert_pt2e + # from torch.ao.quantization.quantize_pt2e import prepare_pt2e + # prepared_model = prepare_pt2e(fx_model, quantizer) + # prepared_model(example_input) + # ao_quantized_model = convert_pt2e(prepared_model) + # ao_nncf_graph = GraphConverter.create_nncf_graph(ao_quantized_model) + # ao_nncf_graph.visualize_graph("ao_" + get_dot_filename(model_case.model_id)) From bba9ee53498a48dfbb344b8899eab334031967a2 Mon Sep 17 00:00:00 2001 From: dlyakhov Date: Fri, 29 Nov 2024 17:33:47 +0100 Subject: [PATCH 09/15] OpenVINOQuantizer as torch.ao Quantizer --- .../quantizer/openvino_quantizer.py | 93 +- .../mobilenet_v3_small.dot | 1038 +++ .../resnet18.dot | 455 ++ .../swin_v2_s.dot | 7100 +++++++++++++++++ .../synthetic_transformer.dot | 41 + .../unet.dot | 461 ++ .../vit_b_16.dot | 1791 +++++ .../yolo11n_sdpa_block.dot | 57 + tests/torch/fx/test_quantizer.py | 24 + 9 files changed, 11051 insertions(+), 9 deletions(-) create mode 100644 tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/mobilenet_v3_small.dot create mode 100644 tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/resnet18.dot create mode 100644 tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/swin_v2_s.dot create mode 100644 tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/synthetic_transformer.dot create mode 100644 tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/unet.dot create mode 100644 tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/vit_b_16.dot create mode 100644 tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/yolo11n_sdpa_block.dot diff --git a/nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer.py b/nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer.py index 36f1f1fd3e3..45a1c69be24 100644 --- a/nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer.py +++ b/nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer.py @@ -9,15 +9,26 @@ # See the License for the specific language governing permissions and # limitations under the License. +from collections import defaultdict from typing import Optional, Union import torch.fx +from torch.ao.quantization.observer import HistogramObserver +from torch.ao.quantization.observer import PerChannelMinMaxObserver +from torch.ao.quantization.quantizer.quantizer import QuantizationAnnotation as InductorQAnotation +from torch.ao.quantization.quantizer.quantizer import QuantizationSpec as InductorQuantizationSpec +from torch.ao.quantization.quantizer.quantizer import Quantizer as InductorQuantizer from nncf.common.graph.graph import NNCFGraph from nncf.common.quantization.quantizer_propagation.solver import QuantizerPropagationRule +from nncf.common.quantization.quantizer_setup import ActivationQuantizationInsertionPoint from nncf.common.quantization.quantizer_setup import SingleConfigQuantizerSetup from nncf.common.quantization.structs import QuantizationPreset +from nncf.common.quantization.structs import QuantizationScheme +from nncf.common.quantization.structs import QuantizerConfig as NNCFQuantizerConfig from nncf.experimental.common.quantization.algorithms.quantizer.base_quantizer import NNCFQuantizer +from nncf.experimental.torch.fx.nncf_graph_builder import GraphConverter +from nncf.experimental.torch.fx.node_utils import get_graph_node_by_name from nncf.parameters import ModelType from nncf.parameters import QuantizationMode from nncf.parameters import TargetDevice @@ -27,8 +38,10 @@ from nncf.quantization.algorithms.min_max.algorithm import MinMaxQuantization from nncf.scopes import IgnoredScope +QUANT_ANNOTATION_KEY = "quantization_annotation" -class OpenVINOQuantizer(NNCFQuantizer): + +class OpenVINOQuantizer(InductorQuantizer, NNCFQuantizer): def __init__( self, mode: Optional[QuantizationMode] = None, @@ -80,12 +93,74 @@ def __init__( ) def get_quantization_setup(self, model: torch.fx.GraphModule, nncf_graph: NNCFGraph) -> SingleConfigQuantizerSetup: - """ - Builds SingleConfigQuantizerSetup for the given model. - - :param model: Backend-specific model, for which Quantization Target Points are being seek. - :param nncf_graph: NNCFGraph instance. - :return: SingleConfigQuantizerSetup for the given model. - """ self._min_max_algo._set_backend_entity(model) - return self._min_max_algo._find_quantization_setup(model, nncf_graph) + return self._min_max_algo.find_quantization_setup(model, nncf_graph) + + def annotate(self, model: torch.fx.GraphModule) -> torch.fx.GraphModule: + nncf_grpah = GraphConverter.create_nncf_graph(model) + quantization_setup = self.get_quantization_setup(model, nncf_grpah) + target_node_vs_qp = defaultdict(list) + graph = model.graph + for qp in quantization_setup.quantization_points.values(): + target_node_vs_qp[qp.insertion_point.target_node_name].append(qp) + + for target_node_name, qps in target_node_vs_qp.items(): + input_qspec_map = dict() + output_qspec = None + target_node = get_graph_node_by_name(graph, target_node_name) + for qp in qps: + ip = qp.insertion_point + if isinstance(ip, ActivationQuantizationInsertionPoint): + inductor_qspec = self._convert_nncf_qspec_to_inductor_qspec(qp.qconfig, is_weight=False) + if ip.input_port_id is None: + output_qspec = inductor_qspec + else: + node = target_node.all_input_nodes[ip.input_port_id] + input_qspec_map[node] = inductor_qspec + else: + inductor_qspec = self._convert_nncf_qspec_to_inductor_qspec(qp.qconfig, is_weight=True) + weight_node = target_node.all_input_nodes[1] + input_qspec_map[weight_node] = inductor_qspec + + annotation = InductorQAnotation(input_qspec_map=input_qspec_map, output_qspec=output_qspec) + assert QUANT_ANNOTATION_KEY not in target_node.meta + target_node.meta[QUANT_ANNOTATION_KEY] = annotation + + def _convert_nncf_qspec_to_inductor_qspec(self, qspec: NNCFQuantizerConfig, is_weight: bool): + extra_args = {"eps": 2**-12} + if qspec.per_channel: + torch_qscheme = ( + torch.per_channel_symmetric if qspec.mode is QuantizationScheme.SYMMETRIC else torch.per_channel_affine + ) + else: + torch_qscheme = ( + torch.per_tensor_symmetric if qspec.mode is QuantizationScheme.SYMMETRIC else torch.per_tensor_affine + ) + if is_weight: + observer = PerChannelMinMaxObserver + quant_min = -128 + quant_max = 127 + dtype = torch.int8 + channel_axis = 0 + else: + observer = ( + HistogramObserver + if torch_qscheme in [torch.per_tensor_symmetric, torch.per_tensor_affine] + else PerChannelMinMaxObserver + ) + quant_min = 0 + quant_max = 255 + dtype = torch.int8 if qspec.signedness_to_force else torch.uint8 + channel_axis = 1 # channel dim for activations + return InductorQuantizationSpec( + dtype=dtype, + observer_or_fake_quant_ctr=observer.with_args(**extra_args), + quant_min=quant_min, + quant_max=quant_max, + qscheme=torch_qscheme, + ch_axis=channel_axis, + is_dynamic=False, + ) + + def validate(self, model: torch.fx.GraphModule) -> None: + pass diff --git a/tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/mobilenet_v3_small.dot b/tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/mobilenet_v3_small.dot new file mode 100644 index 00000000000..8ae7faa470d --- /dev/null +++ b/tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/mobilenet_v3_small.dot @@ -0,0 +1,1038 @@ +strict digraph { +"0 _scale_0" [id=0, type=get_attr]; +"1 _zero_point_0" [id=1, type=get_attr]; +"2 _frozen_param0" [id=2, type=get_attr]; +"3 dequantize_per_channel_default" [id=3, type=dequantize_per_channel]; +"4 _scale_1" [id=4, type=get_attr]; +"5 _zero_point_1" [id=5, type=get_attr]; +"6 _frozen_param1" [id=6, type=get_attr]; +"7 dequantize_per_channel_default_1" [id=7, type=dequantize_per_channel]; +"8 features_1_block_1_fc1_weight" [id=8, type=get_attr]; +"9 features_1_block_1_fc1_bias" [id=9, type=get_attr]; +"10 features_1_block_1_fc2_weight" [id=10, type=get_attr]; +"11 features_1_block_1_fc2_bias" [id=11, type=get_attr]; +"12 _scale_2" [id=12, type=get_attr]; +"13 _zero_point_2" [id=13, type=get_attr]; +"14 _frozen_param2" [id=14, type=get_attr]; +"15 dequantize_per_channel_default_2" [id=15, type=dequantize_per_channel]; +"16 _scale_3" [id=16, type=get_attr]; +"17 _zero_point_3" [id=17, type=get_attr]; +"18 _frozen_param3" [id=18, type=get_attr]; +"19 dequantize_per_channel_default_3" [id=19, type=dequantize_per_channel]; +"20 _scale_4" [id=20, type=get_attr]; +"21 _zero_point_4" [id=21, type=get_attr]; +"22 _frozen_param4" [id=22, type=get_attr]; +"23 dequantize_per_channel_default_4" [id=23, type=dequantize_per_channel]; +"24 _scale_5" [id=24, type=get_attr]; +"25 _zero_point_5" [id=25, type=get_attr]; +"26 _frozen_param5" [id=26, type=get_attr]; +"27 dequantize_per_channel_default_5" [id=27, type=dequantize_per_channel]; +"28 _scale_6" [id=28, type=get_attr]; +"29 _zero_point_6" [id=29, type=get_attr]; +"30 _frozen_param6" [id=30, type=get_attr]; +"31 dequantize_per_channel_default_6" [id=31, type=dequantize_per_channel]; +"32 _scale_7" [id=32, type=get_attr]; +"33 _zero_point_7" [id=33, type=get_attr]; +"34 _frozen_param7" [id=34, type=get_attr]; +"35 dequantize_per_channel_default_7" [id=35, type=dequantize_per_channel]; +"36 _scale_8" [id=36, type=get_attr]; +"37 _zero_point_8" [id=37, type=get_attr]; +"38 _frozen_param8" [id=38, type=get_attr]; +"39 dequantize_per_channel_default_8" [id=39, type=dequantize_per_channel]; +"40 _scale_9" [id=40, type=get_attr]; +"41 _zero_point_9" [id=41, type=get_attr]; +"42 _frozen_param9" [id=42, type=get_attr]; +"43 dequantize_per_channel_default_9" [id=43, type=dequantize_per_channel]; +"44 _scale_10" [id=44, type=get_attr]; +"45 _zero_point_10" [id=45, type=get_attr]; +"46 _frozen_param10" [id=46, type=get_attr]; +"47 dequantize_per_channel_default_10" [id=47, type=dequantize_per_channel]; +"48 features_4_block_2_fc1_weight" [id=48, type=get_attr]; +"49 features_4_block_2_fc1_bias" [id=49, type=get_attr]; +"50 features_4_block_2_fc2_weight" [id=50, type=get_attr]; +"51 features_4_block_2_fc2_bias" [id=51, type=get_attr]; +"52 _scale_11" [id=52, type=get_attr]; +"53 _zero_point_11" [id=53, type=get_attr]; +"54 _frozen_param11" [id=54, type=get_attr]; +"55 dequantize_per_channel_default_11" [id=55, type=dequantize_per_channel]; +"56 _scale_12" [id=56, type=get_attr]; +"57 _zero_point_12" [id=57, type=get_attr]; +"58 _frozen_param12" [id=58, type=get_attr]; +"59 dequantize_per_channel_default_12" [id=59, type=dequantize_per_channel]; +"60 _scale_13" [id=60, type=get_attr]; +"61 _zero_point_13" [id=61, type=get_attr]; +"62 _frozen_param13" [id=62, type=get_attr]; +"63 dequantize_per_channel_default_13" [id=63, type=dequantize_per_channel]; +"64 features_5_block_2_fc1_weight" [id=64, type=get_attr]; +"65 features_5_block_2_fc1_bias" [id=65, type=get_attr]; +"66 features_5_block_2_fc2_weight" [id=66, type=get_attr]; +"67 features_5_block_2_fc2_bias" [id=67, type=get_attr]; +"68 _scale_14" [id=68, type=get_attr]; +"69 _zero_point_14" [id=69, type=get_attr]; +"70 _frozen_param14" [id=70, type=get_attr]; +"71 dequantize_per_channel_default_14" [id=71, type=dequantize_per_channel]; +"72 _scale_15" [id=72, type=get_attr]; +"73 _zero_point_15" [id=73, type=get_attr]; +"74 _frozen_param15" [id=74, type=get_attr]; +"75 dequantize_per_channel_default_15" [id=75, type=dequantize_per_channel]; +"76 _scale_16" [id=76, type=get_attr]; +"77 _zero_point_16" [id=77, type=get_attr]; +"78 _frozen_param16" [id=78, type=get_attr]; +"79 dequantize_per_channel_default_16" [id=79, type=dequantize_per_channel]; +"80 features_6_block_2_fc1_weight" [id=80, type=get_attr]; +"81 features_6_block_2_fc1_bias" [id=81, type=get_attr]; +"82 features_6_block_2_fc2_weight" [id=82, type=get_attr]; +"83 features_6_block_2_fc2_bias" [id=83, type=get_attr]; +"84 _scale_17" [id=84, type=get_attr]; +"85 _zero_point_17" [id=85, type=get_attr]; +"86 _frozen_param17" [id=86, type=get_attr]; +"87 dequantize_per_channel_default_17" [id=87, type=dequantize_per_channel]; +"88 _scale_18" [id=88, type=get_attr]; +"89 _zero_point_18" [id=89, type=get_attr]; +"90 _frozen_param18" [id=90, type=get_attr]; +"91 dequantize_per_channel_default_18" [id=91, type=dequantize_per_channel]; +"92 _scale_19" [id=92, type=get_attr]; +"93 _zero_point_19" [id=93, type=get_attr]; +"94 _frozen_param19" [id=94, type=get_attr]; +"95 dequantize_per_channel_default_19" [id=95, type=dequantize_per_channel]; +"96 features_7_block_2_fc1_weight" [id=96, type=get_attr]; +"97 features_7_block_2_fc1_bias" [id=97, type=get_attr]; +"98 features_7_block_2_fc2_weight" [id=98, type=get_attr]; +"99 features_7_block_2_fc2_bias" [id=99, type=get_attr]; +"100 _scale_20" [id=100, type=get_attr]; +"101 _zero_point_20" [id=101, type=get_attr]; +"102 _frozen_param20" [id=102, type=get_attr]; +"103 dequantize_per_channel_default_20" [id=103, type=dequantize_per_channel]; +"104 _scale_21" [id=104, type=get_attr]; +"105 _zero_point_21" [id=105, type=get_attr]; +"106 _frozen_param21" [id=106, type=get_attr]; +"107 dequantize_per_channel_default_21" [id=107, type=dequantize_per_channel]; +"108 _scale_22" [id=108, type=get_attr]; +"109 _zero_point_22" [id=109, type=get_attr]; +"110 _frozen_param22" [id=110, type=get_attr]; +"111 dequantize_per_channel_default_22" [id=111, type=dequantize_per_channel]; +"112 features_8_block_2_fc1_weight" [id=112, type=get_attr]; +"113 features_8_block_2_fc1_bias" [id=113, type=get_attr]; +"114 features_8_block_2_fc2_weight" [id=114, type=get_attr]; +"115 features_8_block_2_fc2_bias" [id=115, type=get_attr]; +"116 _scale_23" [id=116, type=get_attr]; +"117 _zero_point_23" [id=117, type=get_attr]; +"118 _frozen_param23" [id=118, type=get_attr]; +"119 dequantize_per_channel_default_23" [id=119, type=dequantize_per_channel]; +"120 _scale_24" [id=120, type=get_attr]; +"121 _zero_point_24" [id=121, type=get_attr]; +"122 _frozen_param24" [id=122, type=get_attr]; +"123 dequantize_per_channel_default_24" [id=123, type=dequantize_per_channel]; +"124 _scale_25" [id=124, type=get_attr]; +"125 _zero_point_25" [id=125, type=get_attr]; +"126 _frozen_param25" [id=126, type=get_attr]; +"127 dequantize_per_channel_default_25" [id=127, type=dequantize_per_channel]; +"128 features_9_block_2_fc1_weight" [id=128, type=get_attr]; +"129 features_9_block_2_fc1_bias" [id=129, type=get_attr]; +"130 features_9_block_2_fc2_weight" [id=130, type=get_attr]; +"131 features_9_block_2_fc2_bias" [id=131, type=get_attr]; +"132 _scale_26" [id=132, type=get_attr]; +"133 _zero_point_26" [id=133, type=get_attr]; +"134 _frozen_param26" [id=134, type=get_attr]; +"135 dequantize_per_channel_default_26" [id=135, type=dequantize_per_channel]; +"136 _scale_27" [id=136, type=get_attr]; +"137 _zero_point_27" [id=137, type=get_attr]; +"138 _frozen_param27" [id=138, type=get_attr]; +"139 dequantize_per_channel_default_27" [id=139, type=dequantize_per_channel]; +"140 _scale_28" [id=140, type=get_attr]; +"141 _zero_point_28" [id=141, type=get_attr]; +"142 _frozen_param28" [id=142, type=get_attr]; +"143 dequantize_per_channel_default_28" [id=143, type=dequantize_per_channel]; +"144 features_10_block_2_fc1_weight" [id=144, type=get_attr]; +"145 features_10_block_2_fc1_bias" [id=145, type=get_attr]; +"146 features_10_block_2_fc2_weight" [id=146, type=get_attr]; +"147 features_10_block_2_fc2_bias" [id=147, type=get_attr]; +"148 _scale_29" [id=148, type=get_attr]; +"149 _zero_point_29" [id=149, type=get_attr]; +"150 _frozen_param29" [id=150, type=get_attr]; +"151 dequantize_per_channel_default_29" [id=151, type=dequantize_per_channel]; +"152 _scale_30" [id=152, type=get_attr]; +"153 _zero_point_30" [id=153, type=get_attr]; +"154 _frozen_param30" [id=154, type=get_attr]; +"155 dequantize_per_channel_default_30" [id=155, type=dequantize_per_channel]; +"156 _scale_31" [id=156, type=get_attr]; +"157 _zero_point_31" [id=157, type=get_attr]; +"158 _frozen_param31" [id=158, type=get_attr]; +"159 dequantize_per_channel_default_31" [id=159, type=dequantize_per_channel]; +"160 features_11_block_2_fc1_weight" [id=160, type=get_attr]; +"161 features_11_block_2_fc1_bias" [id=161, type=get_attr]; +"162 features_11_block_2_fc2_weight" [id=162, type=get_attr]; +"163 features_11_block_2_fc2_bias" [id=163, type=get_attr]; +"164 _scale_32" [id=164, type=get_attr]; +"165 _zero_point_32" [id=165, type=get_attr]; +"166 _frozen_param32" [id=166, type=get_attr]; +"167 dequantize_per_channel_default_32" [id=167, type=dequantize_per_channel]; +"168 _scale_33" [id=168, type=get_attr]; +"169 _zero_point_33" [id=169, type=get_attr]; +"170 _frozen_param33" [id=170, type=get_attr]; +"171 dequantize_per_channel_default_33" [id=171, type=dequantize_per_channel]; +"172 _scale_34" [id=172, type=get_attr]; +"173 _zero_point_34" [id=173, type=get_attr]; +"174 _frozen_param34" [id=174, type=get_attr]; +"175 dequantize_per_channel_default_34" [id=175, type=dequantize_per_channel]; +"176 classifier_0_bias" [id=176, type=get_attr]; +"177 _scale_35" [id=177, type=get_attr]; +"178 _zero_point_35" [id=178, type=get_attr]; +"179 _frozen_param35" [id=179, type=get_attr]; +"180 dequantize_per_channel_default_35" [id=180, type=dequantize_per_channel]; +"181 classifier_3_bias" [id=181, type=get_attr]; +"182 x" [id=182, type=input]; +"183 quantize_per_tensor_default" [id=183, type=quantize_per_tensor]; +"184 dequantize_per_tensor_default" [id=184, type=dequantize_per_tensor]; +"185 features_0_0_weight_bias" [id=185, type=get_attr]; +"186 conv2d" [id=186, type=conv2d]; +"187 hardswish_" [id=187, type=hardswish_]; +"188 _2_scale_0" [id=188, type=get_attr]; +"189 _2_zero_point_0" [id=189, type=get_attr]; +"190 quantize_per_channel_default_36" [id=190, type=quantize_per_channel]; +"191 dequantize_per_channel_default_36" [id=191, type=dequantize_per_channel]; +"192 features_1_block_0_0_weight_bias" [id=192, type=get_attr]; +"193 conv2d_1" [id=193, type=conv2d]; +"194 relu_" [id=194, type=relu_]; +"195 quantize_per_tensor_default_1" [id=195, type=quantize_per_tensor]; +"196 dequantize_per_tensor_default_1" [id=196, type=dequantize_per_tensor]; +"197 dequantize_per_tensor_default_50" [id=197, type=dequantize_per_tensor]; +"198 adaptive_avg_pool2d" [id=198, type=adaptive_avg_pool2d]; +"199 conv2d_2" [id=199, type=conv2d]; +"200 relu" [id=200, type=relu]; +"201 conv2d_3" [id=201, type=conv2d]; +"202 hardsigmoid" [id=202, type=hardsigmoid]; +"203 quantize_per_tensor_default_2" [id=203, type=quantize_per_tensor]; +"204 dequantize_per_tensor_default_2" [id=204, type=dequantize_per_tensor]; +"205 mul" [id=205, type=mul]; +"206 quantize_per_tensor_default_3" [id=206, type=quantize_per_tensor]; +"207 dequantize_per_tensor_default_3" [id=207, type=dequantize_per_tensor]; +"208 features_1_block_2_0_weight_bias" [id=208, type=get_attr]; +"209 conv2d_4" [id=209, type=conv2d]; +"210 quantize_per_tensor_default_4" [id=210, type=quantize_per_tensor]; +"211 dequantize_per_tensor_default_4" [id=211, type=dequantize_per_tensor]; +"212 features_2_block_0_0_weight_bias" [id=212, type=get_attr]; +"213 conv2d_5" [id=213, type=conv2d]; +"214 relu__1" [id=214, type=relu_]; +"215 _2_scale_1" [id=215, type=get_attr]; +"216 _2_zero_point_1" [id=216, type=get_attr]; +"217 quantize_per_channel_default_37" [id=217, type=quantize_per_channel]; +"218 dequantize_per_channel_default_37" [id=218, type=dequantize_per_channel]; +"219 features_2_block_1_0_weight_bias" [id=219, type=get_attr]; +"220 conv2d_6" [id=220, type=conv2d]; +"221 relu__2" [id=221, type=relu_]; +"222 quantize_per_tensor_default_5" [id=222, type=quantize_per_tensor]; +"223 dequantize_per_tensor_default_5" [id=223, type=dequantize_per_tensor]; +"224 features_2_block_2_0_weight_bias" [id=224, type=get_attr]; +"225 conv2d_7" [id=225, type=conv2d]; +"226 quantize_per_tensor_default_6" [id=226, type=quantize_per_tensor]; +"227 dequantize_per_tensor_default_52" [id=227, type=dequantize_per_tensor]; +"228 dequantize_per_tensor_default_51" [id=228, type=dequantize_per_tensor]; +"229 features_3_block_0_0_weight_bias" [id=229, type=get_attr]; +"230 conv2d_8" [id=230, type=conv2d]; +"231 relu__3" [id=231, type=relu_]; +"232 _2_scale_2" [id=232, type=get_attr]; +"233 _2_zero_point_2" [id=233, type=get_attr]; +"234 quantize_per_channel_default_38" [id=234, type=quantize_per_channel]; +"235 dequantize_per_channel_default_38" [id=235, type=dequantize_per_channel]; +"236 features_3_block_1_0_weight_bias" [id=236, type=get_attr]; +"237 conv2d_9" [id=237, type=conv2d]; +"238 relu__4" [id=238, type=relu_]; +"239 quantize_per_tensor_default_7" [id=239, type=quantize_per_tensor]; +"240 dequantize_per_tensor_default_7" [id=240, type=dequantize_per_tensor]; +"241 features_3_block_2_0_weight_bias" [id=241, type=get_attr]; +"242 conv2d_10" [id=242, type=conv2d]; +"243 quantize_per_tensor_default_8" [id=243, type=quantize_per_tensor]; +"244 dequantize_per_tensor_default_8" [id=244, type=dequantize_per_tensor]; +"245 add_" [id=245, type=add_]; +"246 quantize_per_tensor_default_9" [id=246, type=quantize_per_tensor]; +"247 dequantize_per_tensor_default_9" [id=247, type=dequantize_per_tensor]; +"248 features_4_block_0_0_weight_bias" [id=248, type=get_attr]; +"249 conv2d_11" [id=249, type=conv2d]; +"250 hardswish__1" [id=250, type=hardswish_]; +"251 _2_scale_3" [id=251, type=get_attr]; +"252 _2_zero_point_3" [id=252, type=get_attr]; +"253 quantize_per_channel_default_39" [id=253, type=quantize_per_channel]; +"254 dequantize_per_channel_default_39" [id=254, type=dequantize_per_channel]; +"255 features_4_block_1_0_weight_bias" [id=255, type=get_attr]; +"256 conv2d_12" [id=256, type=conv2d]; +"257 hardswish__2" [id=257, type=hardswish_]; +"258 quantize_per_tensor_default_10" [id=258, type=quantize_per_tensor]; +"259 dequantize_per_tensor_default_10" [id=259, type=dequantize_per_tensor]; +"260 dequantize_per_tensor_default_53" [id=260, type=dequantize_per_tensor]; +"261 adaptive_avg_pool2d_1" [id=261, type=adaptive_avg_pool2d]; +"262 conv2d_13" [id=262, type=conv2d]; +"263 relu_1" [id=263, type=relu]; +"264 conv2d_14" [id=264, type=conv2d]; +"265 hardsigmoid_1" [id=265, type=hardsigmoid]; +"266 quantize_per_tensor_default_11" [id=266, type=quantize_per_tensor]; +"267 dequantize_per_tensor_default_11" [id=267, type=dequantize_per_tensor]; +"268 mul_1" [id=268, type=mul]; +"269 quantize_per_tensor_default_12" [id=269, type=quantize_per_tensor]; +"270 dequantize_per_tensor_default_12" [id=270, type=dequantize_per_tensor]; +"271 features_4_block_3_0_weight_bias" [id=271, type=get_attr]; +"272 conv2d_15" [id=272, type=conv2d]; +"273 quantize_per_tensor_default_13" [id=273, type=quantize_per_tensor]; +"274 dequantize_per_tensor_default_55" [id=274, type=dequantize_per_tensor]; +"275 dequantize_per_tensor_default_54" [id=275, type=dequantize_per_tensor]; +"276 features_5_block_0_0_weight_bias" [id=276, type=get_attr]; +"277 conv2d_16" [id=277, type=conv2d]; +"278 hardswish__3" [id=278, type=hardswish_]; +"279 _2_scale_4" [id=279, type=get_attr]; +"280 _2_zero_point_4" [id=280, type=get_attr]; +"281 quantize_per_channel_default_40" [id=281, type=quantize_per_channel]; +"282 dequantize_per_channel_default_40" [id=282, type=dequantize_per_channel]; +"283 features_5_block_1_0_weight_bias" [id=283, type=get_attr]; +"284 conv2d_17" [id=284, type=conv2d]; +"285 hardswish__4" [id=285, type=hardswish_]; +"286 quantize_per_tensor_default_14" [id=286, type=quantize_per_tensor]; +"287 dequantize_per_tensor_default_14" [id=287, type=dequantize_per_tensor]; +"288 dequantize_per_tensor_default_56" [id=288, type=dequantize_per_tensor]; +"289 adaptive_avg_pool2d_2" [id=289, type=adaptive_avg_pool2d]; +"290 conv2d_18" [id=290, type=conv2d]; +"291 relu_2" [id=291, type=relu]; +"292 conv2d_19" [id=292, type=conv2d]; +"293 hardsigmoid_2" [id=293, type=hardsigmoid]; +"294 quantize_per_tensor_default_15" [id=294, type=quantize_per_tensor]; +"295 dequantize_per_tensor_default_15" [id=295, type=dequantize_per_tensor]; +"296 mul_2" [id=296, type=mul]; +"297 quantize_per_tensor_default_16" [id=297, type=quantize_per_tensor]; +"298 dequantize_per_tensor_default_16" [id=298, type=dequantize_per_tensor]; +"299 features_5_block_3_0_weight_bias" [id=299, type=get_attr]; +"300 conv2d_20" [id=300, type=conv2d]; +"301 quantize_per_tensor_default_17" [id=301, type=quantize_per_tensor]; +"302 dequantize_per_tensor_default_17" [id=302, type=dequantize_per_tensor]; +"303 add__1" [id=303, type=add_]; +"304 quantize_per_tensor_default_18" [id=304, type=quantize_per_tensor]; +"305 dequantize_per_tensor_default_58" [id=305, type=dequantize_per_tensor]; +"306 dequantize_per_tensor_default_57" [id=306, type=dequantize_per_tensor]; +"307 features_6_block_0_0_weight_bias" [id=307, type=get_attr]; +"308 conv2d_21" [id=308, type=conv2d]; +"309 hardswish__5" [id=309, type=hardswish_]; +"310 _2_scale_5" [id=310, type=get_attr]; +"311 _2_zero_point_5" [id=311, type=get_attr]; +"312 quantize_per_channel_default_41" [id=312, type=quantize_per_channel]; +"313 dequantize_per_channel_default_41" [id=313, type=dequantize_per_channel]; +"314 features_6_block_1_0_weight_bias" [id=314, type=get_attr]; +"315 conv2d_22" [id=315, type=conv2d]; +"316 hardswish__6" [id=316, type=hardswish_]; +"317 quantize_per_tensor_default_19" [id=317, type=quantize_per_tensor]; +"318 dequantize_per_tensor_default_19" [id=318, type=dequantize_per_tensor]; +"319 dequantize_per_tensor_default_59" [id=319, type=dequantize_per_tensor]; +"320 adaptive_avg_pool2d_3" [id=320, type=adaptive_avg_pool2d]; +"321 conv2d_23" [id=321, type=conv2d]; +"322 relu_3" [id=322, type=relu]; +"323 conv2d_24" [id=323, type=conv2d]; +"324 hardsigmoid_3" [id=324, type=hardsigmoid]; +"325 quantize_per_tensor_default_20" [id=325, type=quantize_per_tensor]; +"326 dequantize_per_tensor_default_20" [id=326, type=dequantize_per_tensor]; +"327 mul_3" [id=327, type=mul]; +"328 quantize_per_tensor_default_21" [id=328, type=quantize_per_tensor]; +"329 dequantize_per_tensor_default_21" [id=329, type=dequantize_per_tensor]; +"330 features_6_block_3_0_weight_bias" [id=330, type=get_attr]; +"331 conv2d_25" [id=331, type=conv2d]; +"332 quantize_per_tensor_default_22" [id=332, type=quantize_per_tensor]; +"333 dequantize_per_tensor_default_22" [id=333, type=dequantize_per_tensor]; +"334 add__2" [id=334, type=add_]; +"335 quantize_per_tensor_default_23" [id=335, type=quantize_per_tensor]; +"336 dequantize_per_tensor_default_23" [id=336, type=dequantize_per_tensor]; +"337 features_7_block_0_0_weight_bias" [id=337, type=get_attr]; +"338 conv2d_26" [id=338, type=conv2d]; +"339 hardswish__7" [id=339, type=hardswish_]; +"340 _2_scale_6" [id=340, type=get_attr]; +"341 _2_zero_point_6" [id=341, type=get_attr]; +"342 quantize_per_channel_default_42" [id=342, type=quantize_per_channel]; +"343 dequantize_per_channel_default_42" [id=343, type=dequantize_per_channel]; +"344 features_7_block_1_0_weight_bias" [id=344, type=get_attr]; +"345 conv2d_27" [id=345, type=conv2d]; +"346 hardswish__8" [id=346, type=hardswish_]; +"347 quantize_per_tensor_default_24" [id=347, type=quantize_per_tensor]; +"348 dequantize_per_tensor_default_24" [id=348, type=dequantize_per_tensor]; +"349 dequantize_per_tensor_default_60" [id=349, type=dequantize_per_tensor]; +"350 adaptive_avg_pool2d_4" [id=350, type=adaptive_avg_pool2d]; +"351 conv2d_28" [id=351, type=conv2d]; +"352 relu_4" [id=352, type=relu]; +"353 conv2d_29" [id=353, type=conv2d]; +"354 hardsigmoid_4" [id=354, type=hardsigmoid]; +"355 quantize_per_tensor_default_25" [id=355, type=quantize_per_tensor]; +"356 dequantize_per_tensor_default_25" [id=356, type=dequantize_per_tensor]; +"357 mul_4" [id=357, type=mul]; +"358 quantize_per_tensor_default_26" [id=358, type=quantize_per_tensor]; +"359 dequantize_per_tensor_default_26" [id=359, type=dequantize_per_tensor]; +"360 features_7_block_3_0_weight_bias" [id=360, type=get_attr]; +"361 conv2d_30" [id=361, type=conv2d]; +"362 quantize_per_tensor_default_27" [id=362, type=quantize_per_tensor]; +"363 dequantize_per_tensor_default_62" [id=363, type=dequantize_per_tensor]; +"364 dequantize_per_tensor_default_61" [id=364, type=dequantize_per_tensor]; +"365 features_8_block_0_0_weight_bias" [id=365, type=get_attr]; +"366 conv2d_31" [id=366, type=conv2d]; +"367 hardswish__9" [id=367, type=hardswish_]; +"368 _2_scale_7" [id=368, type=get_attr]; +"369 _2_zero_point_7" [id=369, type=get_attr]; +"370 quantize_per_channel_default_43" [id=370, type=quantize_per_channel]; +"371 dequantize_per_channel_default_43" [id=371, type=dequantize_per_channel]; +"372 features_8_block_1_0_weight_bias" [id=372, type=get_attr]; +"373 conv2d_32" [id=373, type=conv2d]; +"374 hardswish__10" [id=374, type=hardswish_]; +"375 quantize_per_tensor_default_28" [id=375, type=quantize_per_tensor]; +"376 dequantize_per_tensor_default_28" [id=376, type=dequantize_per_tensor]; +"377 dequantize_per_tensor_default_63" [id=377, type=dequantize_per_tensor]; +"378 adaptive_avg_pool2d_5" [id=378, type=adaptive_avg_pool2d]; +"379 conv2d_33" [id=379, type=conv2d]; +"380 relu_5" [id=380, type=relu]; +"381 conv2d_34" [id=381, type=conv2d]; +"382 hardsigmoid_5" [id=382, type=hardsigmoid]; +"383 quantize_per_tensor_default_29" [id=383, type=quantize_per_tensor]; +"384 dequantize_per_tensor_default_29" [id=384, type=dequantize_per_tensor]; +"385 mul_5" [id=385, type=mul]; +"386 quantize_per_tensor_default_30" [id=386, type=quantize_per_tensor]; +"387 dequantize_per_tensor_default_30" [id=387, type=dequantize_per_tensor]; +"388 features_8_block_3_0_weight_bias" [id=388, type=get_attr]; +"389 conv2d_35" [id=389, type=conv2d]; +"390 quantize_per_tensor_default_31" [id=390, type=quantize_per_tensor]; +"391 dequantize_per_tensor_default_31" [id=391, type=dequantize_per_tensor]; +"392 add__3" [id=392, type=add_]; +"393 quantize_per_tensor_default_32" [id=393, type=quantize_per_tensor]; +"394 dequantize_per_tensor_default_32" [id=394, type=dequantize_per_tensor]; +"395 features_9_block_0_0_weight_bias" [id=395, type=get_attr]; +"396 conv2d_36" [id=396, type=conv2d]; +"397 hardswish__11" [id=397, type=hardswish_]; +"398 _2_scale_8" [id=398, type=get_attr]; +"399 _2_zero_point_8" [id=399, type=get_attr]; +"400 quantize_per_channel_default_44" [id=400, type=quantize_per_channel]; +"401 dequantize_per_channel_default_44" [id=401, type=dequantize_per_channel]; +"402 features_9_block_1_0_weight_bias" [id=402, type=get_attr]; +"403 conv2d_37" [id=403, type=conv2d]; +"404 hardswish__12" [id=404, type=hardswish_]; +"405 quantize_per_tensor_default_33" [id=405, type=quantize_per_tensor]; +"406 dequantize_per_tensor_default_33" [id=406, type=dequantize_per_tensor]; +"407 dequantize_per_tensor_default_64" [id=407, type=dequantize_per_tensor]; +"408 adaptive_avg_pool2d_6" [id=408, type=adaptive_avg_pool2d]; +"409 conv2d_38" [id=409, type=conv2d]; +"410 relu_6" [id=410, type=relu]; +"411 conv2d_39" [id=411, type=conv2d]; +"412 hardsigmoid_6" [id=412, type=hardsigmoid]; +"413 quantize_per_tensor_default_34" [id=413, type=quantize_per_tensor]; +"414 dequantize_per_tensor_default_34" [id=414, type=dequantize_per_tensor]; +"415 mul_6" [id=415, type=mul]; +"416 quantize_per_tensor_default_35" [id=416, type=quantize_per_tensor]; +"417 dequantize_per_tensor_default_35" [id=417, type=dequantize_per_tensor]; +"418 features_9_block_3_0_weight_bias" [id=418, type=get_attr]; +"419 conv2d_40" [id=419, type=conv2d]; +"420 quantize_per_tensor_default_36" [id=420, type=quantize_per_tensor]; +"421 dequantize_per_tensor_default_66" [id=421, type=dequantize_per_tensor]; +"422 dequantize_per_tensor_default_65" [id=422, type=dequantize_per_tensor]; +"423 features_10_block_0_0_weight_bias" [id=423, type=get_attr]; +"424 conv2d_41" [id=424, type=conv2d]; +"425 hardswish__13" [id=425, type=hardswish_]; +"426 _2_scale_9" [id=426, type=get_attr]; +"427 _2_zero_point_9" [id=427, type=get_attr]; +"428 quantize_per_channel_default_45" [id=428, type=quantize_per_channel]; +"429 dequantize_per_channel_default_45" [id=429, type=dequantize_per_channel]; +"430 features_10_block_1_0_weight_bias" [id=430, type=get_attr]; +"431 conv2d_42" [id=431, type=conv2d]; +"432 hardswish__14" [id=432, type=hardswish_]; +"433 quantize_per_tensor_default_37" [id=433, type=quantize_per_tensor]; +"434 dequantize_per_tensor_default_37" [id=434, type=dequantize_per_tensor]; +"435 dequantize_per_tensor_default_67" [id=435, type=dequantize_per_tensor]; +"436 adaptive_avg_pool2d_7" [id=436, type=adaptive_avg_pool2d]; +"437 conv2d_43" [id=437, type=conv2d]; +"438 relu_7" [id=438, type=relu]; +"439 conv2d_44" [id=439, type=conv2d]; +"440 hardsigmoid_7" [id=440, type=hardsigmoid]; +"441 quantize_per_tensor_default_38" [id=441, type=quantize_per_tensor]; +"442 dequantize_per_tensor_default_38" [id=442, type=dequantize_per_tensor]; +"443 mul_7" [id=443, type=mul]; +"444 quantize_per_tensor_default_39" [id=444, type=quantize_per_tensor]; +"445 dequantize_per_tensor_default_39" [id=445, type=dequantize_per_tensor]; +"446 features_10_block_3_0_weight_bias" [id=446, type=get_attr]; +"447 conv2d_45" [id=447, type=conv2d]; +"448 quantize_per_tensor_default_40" [id=448, type=quantize_per_tensor]; +"449 dequantize_per_tensor_default_40" [id=449, type=dequantize_per_tensor]; +"450 add__4" [id=450, type=add_]; +"451 quantize_per_tensor_default_41" [id=451, type=quantize_per_tensor]; +"452 dequantize_per_tensor_default_69" [id=452, type=dequantize_per_tensor]; +"453 dequantize_per_tensor_default_68" [id=453, type=dequantize_per_tensor]; +"454 features_11_block_0_0_weight_bias" [id=454, type=get_attr]; +"455 conv2d_46" [id=455, type=conv2d]; +"456 hardswish__15" [id=456, type=hardswish_]; +"457 _2_scale_10" [id=457, type=get_attr]; +"458 _2_zero_point_10" [id=458, type=get_attr]; +"459 quantize_per_channel_default_46" [id=459, type=quantize_per_channel]; +"460 dequantize_per_channel_default_46" [id=460, type=dequantize_per_channel]; +"461 features_11_block_1_0_weight_bias" [id=461, type=get_attr]; +"462 conv2d_47" [id=462, type=conv2d]; +"463 hardswish__16" [id=463, type=hardswish_]; +"464 quantize_per_tensor_default_42" [id=464, type=quantize_per_tensor]; +"465 dequantize_per_tensor_default_42" [id=465, type=dequantize_per_tensor]; +"466 dequantize_per_tensor_default_70" [id=466, type=dequantize_per_tensor]; +"467 adaptive_avg_pool2d_8" [id=467, type=adaptive_avg_pool2d]; +"468 conv2d_48" [id=468, type=conv2d]; +"469 relu_8" [id=469, type=relu]; +"470 conv2d_49" [id=470, type=conv2d]; +"471 hardsigmoid_8" [id=471, type=hardsigmoid]; +"472 quantize_per_tensor_default_43" [id=472, type=quantize_per_tensor]; +"473 dequantize_per_tensor_default_43" [id=473, type=dequantize_per_tensor]; +"474 mul_8" [id=474, type=mul]; +"475 quantize_per_tensor_default_44" [id=475, type=quantize_per_tensor]; +"476 dequantize_per_tensor_default_44" [id=476, type=dequantize_per_tensor]; +"477 features_11_block_3_0_weight_bias" [id=477, type=get_attr]; +"478 conv2d_50" [id=478, type=conv2d]; +"479 quantize_per_tensor_default_45" [id=479, type=quantize_per_tensor]; +"480 dequantize_per_tensor_default_45" [id=480, type=dequantize_per_tensor]; +"481 add__5" [id=481, type=add_]; +"482 quantize_per_tensor_default_46" [id=482, type=quantize_per_tensor]; +"483 dequantize_per_tensor_default_46" [id=483, type=dequantize_per_tensor]; +"484 features_12_0_weight_bias" [id=484, type=get_attr]; +"485 conv2d_51" [id=485, type=conv2d]; +"486 hardswish__17" [id=486, type=hardswish_]; +"487 quantize_per_tensor_default_47" [id=487, type=quantize_per_tensor]; +"488 dequantize_per_tensor_default_47" [id=488, type=dequantize_per_tensor]; +"489 adaptive_avg_pool2d_9" [id=489, type=adaptive_avg_pool2d]; +"490 quantize_per_tensor_default_48" [id=490, type=quantize_per_tensor]; +"491 dequantize_per_tensor_default_48" [id=491, type=dequantize_per_tensor]; +"492 flatten" [id=492, type=flatten]; +"493 linear" [id=493, type=linear]; +"494 hardswish__18" [id=494, type=hardswish_]; +"495 quantize_per_tensor_default_49" [id=495, type=quantize_per_tensor]; +"496 dequantize_per_tensor_default_49" [id=496, type=dequantize_per_tensor]; +"497 dropout_" [id=497, type=dropout_]; +"498 linear_1" [id=498, type=linear]; +"499 output_1" [id=499, type=output]; +"0 _scale_0" -> "3 dequantize_per_channel_default" [label="(16,)", style=solid]; +"1 _zero_point_0" -> "3 dequantize_per_channel_default" [label="(16,)", style=solid]; +"2 _frozen_param0" -> "3 dequantize_per_channel_default" [label="(16, 3, 3, 3)", style=solid]; +"3 dequantize_per_channel_default" -> "186 conv2d" [label=None, style=solid]; +"4 _scale_1" -> "7 dequantize_per_channel_default_1" [label="(16,)", style=solid]; +"5 _zero_point_1" -> "7 dequantize_per_channel_default_1" [label="(16,)", style=solid]; +"6 _frozen_param1" -> "7 dequantize_per_channel_default_1" [label="(16, 1, 3, 3)", style=solid]; +"7 dequantize_per_channel_default_1" -> "193 conv2d_1" [label=None, style=solid]; +"8 features_1_block_1_fc1_weight" -> "199 conv2d_2" [label="(8, 16, 1, 1)", style=solid]; +"9 features_1_block_1_fc1_bias" -> "199 conv2d_2" [label="(8,)", style=solid]; +"10 features_1_block_1_fc2_weight" -> "201 conv2d_3" [label="(16, 8, 1, 1)", style=solid]; +"11 features_1_block_1_fc2_bias" -> "201 conv2d_3" [label="(16,)", style=solid]; +"12 _scale_2" -> "15 dequantize_per_channel_default_2" [label="(16,)", style=solid]; +"13 _zero_point_2" -> "15 dequantize_per_channel_default_2" [label="(16,)", style=solid]; +"14 _frozen_param2" -> "15 dequantize_per_channel_default_2" [label="(16, 16, 1, 1)", style=solid]; +"15 dequantize_per_channel_default_2" -> "209 conv2d_4" [label=None, style=solid]; +"16 _scale_3" -> "19 dequantize_per_channel_default_3" [label="(72,)", style=solid]; +"17 _zero_point_3" -> "19 dequantize_per_channel_default_3" [label="(72,)", style=solid]; +"18 _frozen_param3" -> "19 dequantize_per_channel_default_3" [label="(72, 16, 1, 1)", style=solid]; +"19 dequantize_per_channel_default_3" -> "213 conv2d_5" [label=None, style=solid]; +"20 _scale_4" -> "23 dequantize_per_channel_default_4" [label="(72,)", style=solid]; +"21 _zero_point_4" -> "23 dequantize_per_channel_default_4" [label="(72,)", style=solid]; +"22 _frozen_param4" -> "23 dequantize_per_channel_default_4" [label="(72, 1, 3, 3)", style=solid]; +"23 dequantize_per_channel_default_4" -> "220 conv2d_6" [label=None, style=solid]; +"24 _scale_5" -> "27 dequantize_per_channel_default_5" [label="(24,)", style=solid]; +"25 _zero_point_5" -> "27 dequantize_per_channel_default_5" [label="(24,)", style=solid]; +"26 _frozen_param5" -> "27 dequantize_per_channel_default_5" [label="(24, 72, 1, 1)", style=solid]; +"27 dequantize_per_channel_default_5" -> "225 conv2d_7" [label=None, style=solid]; +"28 _scale_6" -> "31 dequantize_per_channel_default_6" [label="(88,)", style=solid]; +"29 _zero_point_6" -> "31 dequantize_per_channel_default_6" [label="(88,)", style=solid]; +"30 _frozen_param6" -> "31 dequantize_per_channel_default_6" [label="(88, 24, 1, 1)", style=solid]; +"31 dequantize_per_channel_default_6" -> "230 conv2d_8" [label=None, style=solid]; +"32 _scale_7" -> "35 dequantize_per_channel_default_7" [label="(88,)", style=solid]; +"33 _zero_point_7" -> "35 dequantize_per_channel_default_7" [label="(88,)", style=solid]; +"34 _frozen_param7" -> "35 dequantize_per_channel_default_7" [label="(88, 1, 3, 3)", style=solid]; +"35 dequantize_per_channel_default_7" -> "237 conv2d_9" [label=None, style=solid]; +"36 _scale_8" -> "39 dequantize_per_channel_default_8" [label="(24,)", style=solid]; +"37 _zero_point_8" -> "39 dequantize_per_channel_default_8" [label="(24,)", style=solid]; +"38 _frozen_param8" -> "39 dequantize_per_channel_default_8" [label="(24, 88, 1, 1)", style=solid]; +"39 dequantize_per_channel_default_8" -> "242 conv2d_10" [label=None, style=solid]; +"40 _scale_9" -> "43 dequantize_per_channel_default_9" [label="(96,)", style=solid]; +"41 _zero_point_9" -> "43 dequantize_per_channel_default_9" [label="(96,)", style=solid]; +"42 _frozen_param9" -> "43 dequantize_per_channel_default_9" [label="(96, 24, 1, 1)", style=solid]; +"43 dequantize_per_channel_default_9" -> "249 conv2d_11" [label=None, style=solid]; +"44 _scale_10" -> "47 dequantize_per_channel_default_10" [label="(96,)", style=solid]; +"45 _zero_point_10" -> "47 dequantize_per_channel_default_10" [label="(96,)", style=solid]; +"46 _frozen_param10" -> "47 dequantize_per_channel_default_10" [label="(96, 1, 5, 5)", style=solid]; +"47 dequantize_per_channel_default_10" -> "256 conv2d_12" [label=None, style=solid]; +"48 features_4_block_2_fc1_weight" -> "262 conv2d_13" [label="(24, 96, 1, 1)", style=solid]; +"49 features_4_block_2_fc1_bias" -> "262 conv2d_13" [label="(24,)", style=solid]; +"50 features_4_block_2_fc2_weight" -> "264 conv2d_14" [label="(96, 24, 1, 1)", style=solid]; +"51 features_4_block_2_fc2_bias" -> "264 conv2d_14" [label="(96,)", style=solid]; +"52 _scale_11" -> "55 dequantize_per_channel_default_11" [label="(40,)", style=solid]; +"53 _zero_point_11" -> "55 dequantize_per_channel_default_11" [label="(40,)", style=solid]; +"54 _frozen_param11" -> "55 dequantize_per_channel_default_11" [label="(40, 96, 1, 1)", style=solid]; +"55 dequantize_per_channel_default_11" -> "272 conv2d_15" [label=None, style=solid]; +"56 _scale_12" -> "59 dequantize_per_channel_default_12" [label="(240,)", style=solid]; +"57 _zero_point_12" -> "59 dequantize_per_channel_default_12" [label="(240,)", style=solid]; +"58 _frozen_param12" -> "59 dequantize_per_channel_default_12" [label="(240, 40, 1, 1)", style=solid]; +"59 dequantize_per_channel_default_12" -> "277 conv2d_16" [label=None, style=solid]; +"60 _scale_13" -> "63 dequantize_per_channel_default_13" [label="(240,)", style=solid]; +"61 _zero_point_13" -> "63 dequantize_per_channel_default_13" [label="(240,)", style=solid]; +"62 _frozen_param13" -> "63 dequantize_per_channel_default_13" [label="(240, 1, 5, 5)", style=solid]; +"63 dequantize_per_channel_default_13" -> "284 conv2d_17" [label=None, style=solid]; +"64 features_5_block_2_fc1_weight" -> "290 conv2d_18" [label="(64, 240, 1, 1)", style=solid]; +"65 features_5_block_2_fc1_bias" -> "290 conv2d_18" [label="(64,)", style=solid]; +"66 features_5_block_2_fc2_weight" -> "292 conv2d_19" [label="(240, 64, 1, 1)", style=solid]; +"67 features_5_block_2_fc2_bias" -> "292 conv2d_19" [label="(240,)", style=solid]; +"68 _scale_14" -> "71 dequantize_per_channel_default_14" [label="(40,)", style=solid]; +"69 _zero_point_14" -> "71 dequantize_per_channel_default_14" [label="(40,)", style=solid]; +"70 _frozen_param14" -> "71 dequantize_per_channel_default_14" [label="(40, 240, 1, 1)", style=solid]; +"71 dequantize_per_channel_default_14" -> "300 conv2d_20" [label=None, style=solid]; +"72 _scale_15" -> "75 dequantize_per_channel_default_15" [label="(240,)", style=solid]; +"73 _zero_point_15" -> "75 dequantize_per_channel_default_15" [label="(240,)", style=solid]; +"74 _frozen_param15" -> "75 dequantize_per_channel_default_15" [label="(240, 40, 1, 1)", style=solid]; +"75 dequantize_per_channel_default_15" -> "308 conv2d_21" [label=None, style=solid]; +"76 _scale_16" -> "79 dequantize_per_channel_default_16" [label="(240,)", style=solid]; +"77 _zero_point_16" -> "79 dequantize_per_channel_default_16" [label="(240,)", style=solid]; +"78 _frozen_param16" -> "79 dequantize_per_channel_default_16" [label="(240, 1, 5, 5)", style=solid]; +"79 dequantize_per_channel_default_16" -> "315 conv2d_22" [label=None, style=solid]; +"80 features_6_block_2_fc1_weight" -> "321 conv2d_23" [label="(64, 240, 1, 1)", style=solid]; +"81 features_6_block_2_fc1_bias" -> "321 conv2d_23" [label="(64,)", style=solid]; +"82 features_6_block_2_fc2_weight" -> "323 conv2d_24" [label="(240, 64, 1, 1)", style=solid]; +"83 features_6_block_2_fc2_bias" -> "323 conv2d_24" [label="(240,)", style=solid]; +"84 _scale_17" -> "87 dequantize_per_channel_default_17" [label="(40,)", style=solid]; +"85 _zero_point_17" -> "87 dequantize_per_channel_default_17" [label="(40,)", style=solid]; +"86 _frozen_param17" -> "87 dequantize_per_channel_default_17" [label="(40, 240, 1, 1)", style=solid]; +"87 dequantize_per_channel_default_17" -> "331 conv2d_25" [label=None, style=solid]; +"88 _scale_18" -> "91 dequantize_per_channel_default_18" [label="(120,)", style=solid]; +"89 _zero_point_18" -> "91 dequantize_per_channel_default_18" [label="(120,)", style=solid]; +"90 _frozen_param18" -> "91 dequantize_per_channel_default_18" [label="(120, 40, 1, 1)", style=solid]; +"91 dequantize_per_channel_default_18" -> "338 conv2d_26" [label=None, style=solid]; +"92 _scale_19" -> "95 dequantize_per_channel_default_19" [label="(120,)", style=solid]; +"93 _zero_point_19" -> "95 dequantize_per_channel_default_19" [label="(120,)", style=solid]; +"94 _frozen_param19" -> "95 dequantize_per_channel_default_19" [label="(120, 1, 5, 5)", style=solid]; +"95 dequantize_per_channel_default_19" -> "345 conv2d_27" [label=None, style=solid]; +"96 features_7_block_2_fc1_weight" -> "351 conv2d_28" [label="(32, 120, 1, 1)", style=solid]; +"97 features_7_block_2_fc1_bias" -> "351 conv2d_28" [label="(32,)", style=solid]; +"98 features_7_block_2_fc2_weight" -> "353 conv2d_29" [label="(120, 32, 1, 1)", style=solid]; +"99 features_7_block_2_fc2_bias" -> "353 conv2d_29" [label="(120,)", style=solid]; +"100 _scale_20" -> "103 dequantize_per_channel_default_20" [label="(48,)", style=solid]; +"101 _zero_point_20" -> "103 dequantize_per_channel_default_20" [label="(48,)", style=solid]; +"102 _frozen_param20" -> "103 dequantize_per_channel_default_20" [label="(48, 120, 1, 1)", style=solid]; +"103 dequantize_per_channel_default_20" -> "361 conv2d_30" [label=None, style=solid]; +"104 _scale_21" -> "107 dequantize_per_channel_default_21" [label="(144,)", style=solid]; +"105 _zero_point_21" -> "107 dequantize_per_channel_default_21" [label="(144,)", style=solid]; +"106 _frozen_param21" -> "107 dequantize_per_channel_default_21" [label="(144, 48, 1, 1)", style=solid]; +"107 dequantize_per_channel_default_21" -> "366 conv2d_31" [label=None, style=solid]; +"108 _scale_22" -> "111 dequantize_per_channel_default_22" [label="(144,)", style=solid]; +"109 _zero_point_22" -> "111 dequantize_per_channel_default_22" [label="(144,)", style=solid]; +"110 _frozen_param22" -> "111 dequantize_per_channel_default_22" [label="(144, 1, 5, 5)", style=solid]; +"111 dequantize_per_channel_default_22" -> "373 conv2d_32" [label=None, style=solid]; +"112 features_8_block_2_fc1_weight" -> "379 conv2d_33" [label="(40, 144, 1, 1)", style=solid]; +"113 features_8_block_2_fc1_bias" -> "379 conv2d_33" [label="(40,)", style=solid]; +"114 features_8_block_2_fc2_weight" -> "381 conv2d_34" [label="(144, 40, 1, 1)", style=solid]; +"115 features_8_block_2_fc2_bias" -> "381 conv2d_34" [label="(144,)", style=solid]; +"116 _scale_23" -> "119 dequantize_per_channel_default_23" [label="(48,)", style=solid]; +"117 _zero_point_23" -> "119 dequantize_per_channel_default_23" [label="(48,)", style=solid]; +"118 _frozen_param23" -> "119 dequantize_per_channel_default_23" [label="(48, 144, 1, 1)", style=solid]; +"119 dequantize_per_channel_default_23" -> "389 conv2d_35" [label=None, style=solid]; +"120 _scale_24" -> "123 dequantize_per_channel_default_24" [label="(288,)", style=solid]; +"121 _zero_point_24" -> "123 dequantize_per_channel_default_24" [label="(288,)", style=solid]; +"122 _frozen_param24" -> "123 dequantize_per_channel_default_24" [label="(288, 48, 1, 1)", style=solid]; +"123 dequantize_per_channel_default_24" -> "396 conv2d_36" [label=None, style=solid]; +"124 _scale_25" -> "127 dequantize_per_channel_default_25" [label="(288,)", style=solid]; +"125 _zero_point_25" -> "127 dequantize_per_channel_default_25" [label="(288,)", style=solid]; +"126 _frozen_param25" -> "127 dequantize_per_channel_default_25" [label="(288, 1, 5, 5)", style=solid]; +"127 dequantize_per_channel_default_25" -> "403 conv2d_37" [label=None, style=solid]; +"128 features_9_block_2_fc1_weight" -> "409 conv2d_38" [label="(72, 288, 1, 1)", style=solid]; +"129 features_9_block_2_fc1_bias" -> "409 conv2d_38" [label="(72,)", style=solid]; +"130 features_9_block_2_fc2_weight" -> "411 conv2d_39" [label="(288, 72, 1, 1)", style=solid]; +"131 features_9_block_2_fc2_bias" -> "411 conv2d_39" [label="(288,)", style=solid]; +"132 _scale_26" -> "135 dequantize_per_channel_default_26" [label="(96,)", style=solid]; +"133 _zero_point_26" -> "135 dequantize_per_channel_default_26" [label="(96,)", style=solid]; +"134 _frozen_param26" -> "135 dequantize_per_channel_default_26" [label="(96, 288, 1, 1)", style=solid]; +"135 dequantize_per_channel_default_26" -> "419 conv2d_40" [label=None, style=solid]; +"136 _scale_27" -> "139 dequantize_per_channel_default_27" [label="(576,)", style=solid]; +"137 _zero_point_27" -> "139 dequantize_per_channel_default_27" [label="(576,)", style=solid]; +"138 _frozen_param27" -> "139 dequantize_per_channel_default_27" [label="(576, 96, 1, 1)", style=solid]; +"139 dequantize_per_channel_default_27" -> "424 conv2d_41" [label=None, style=solid]; +"140 _scale_28" -> "143 dequantize_per_channel_default_28" [label="(576,)", style=solid]; +"141 _zero_point_28" -> "143 dequantize_per_channel_default_28" [label="(576,)", style=solid]; +"142 _frozen_param28" -> "143 dequantize_per_channel_default_28" [label="(576, 1, 5, 5)", style=solid]; +"143 dequantize_per_channel_default_28" -> "431 conv2d_42" [label=None, style=solid]; +"144 features_10_block_2_fc1_weight" -> "437 conv2d_43" [label="(144, 576, 1, 1)", style=solid]; +"145 features_10_block_2_fc1_bias" -> "437 conv2d_43" [label="(144,)", style=solid]; +"146 features_10_block_2_fc2_weight" -> "439 conv2d_44" [label="(576, 144, 1, 1)", style=solid]; +"147 features_10_block_2_fc2_bias" -> "439 conv2d_44" [label="(576,)", style=solid]; +"148 _scale_29" -> "151 dequantize_per_channel_default_29" [label="(96,)", style=solid]; +"149 _zero_point_29" -> "151 dequantize_per_channel_default_29" [label="(96,)", style=solid]; +"150 _frozen_param29" -> "151 dequantize_per_channel_default_29" [label="(96, 576, 1, 1)", style=solid]; +"151 dequantize_per_channel_default_29" -> "447 conv2d_45" [label=None, style=solid]; +"152 _scale_30" -> "155 dequantize_per_channel_default_30" [label="(576,)", style=solid]; +"153 _zero_point_30" -> "155 dequantize_per_channel_default_30" [label="(576,)", style=solid]; +"154 _frozen_param30" -> "155 dequantize_per_channel_default_30" [label="(576, 96, 1, 1)", style=solid]; +"155 dequantize_per_channel_default_30" -> "455 conv2d_46" [label=None, style=solid]; +"156 _scale_31" -> "159 dequantize_per_channel_default_31" [label="(576,)", style=solid]; +"157 _zero_point_31" -> "159 dequantize_per_channel_default_31" [label="(576,)", style=solid]; +"158 _frozen_param31" -> "159 dequantize_per_channel_default_31" [label="(576, 1, 5, 5)", style=solid]; +"159 dequantize_per_channel_default_31" -> "462 conv2d_47" [label=None, style=solid]; +"160 features_11_block_2_fc1_weight" -> "468 conv2d_48" [label="(144, 576, 1, 1)", style=solid]; +"161 features_11_block_2_fc1_bias" -> "468 conv2d_48" [label="(144,)", style=solid]; +"162 features_11_block_2_fc2_weight" -> "470 conv2d_49" [label="(576, 144, 1, 1)", style=solid]; +"163 features_11_block_2_fc2_bias" -> "470 conv2d_49" [label="(576,)", style=solid]; +"164 _scale_32" -> "167 dequantize_per_channel_default_32" [label="(96,)", style=solid]; +"165 _zero_point_32" -> "167 dequantize_per_channel_default_32" [label="(96,)", style=solid]; +"166 _frozen_param32" -> "167 dequantize_per_channel_default_32" [label="(96, 576, 1, 1)", style=solid]; +"167 dequantize_per_channel_default_32" -> "478 conv2d_50" [label=None, style=solid]; +"168 _scale_33" -> "171 dequantize_per_channel_default_33" [label="(576,)", style=solid]; +"169 _zero_point_33" -> "171 dequantize_per_channel_default_33" [label="(576,)", style=solid]; +"170 _frozen_param33" -> "171 dequantize_per_channel_default_33" [label="(576, 96, 1, 1)", style=solid]; +"171 dequantize_per_channel_default_33" -> "485 conv2d_51" [label=None, style=solid]; +"172 _scale_34" -> "175 dequantize_per_channel_default_34" [label="(1024,)", style=solid]; +"173 _zero_point_34" -> "175 dequantize_per_channel_default_34" [label="(1024,)", style=solid]; +"174 _frozen_param34" -> "175 dequantize_per_channel_default_34" [label="(1024, 576)", style=solid]; +"175 dequantize_per_channel_default_34" -> "493 linear" [label=None, style=solid]; +"176 classifier_0_bias" -> "493 linear" [label="(1024,)", style=solid]; +"177 _scale_35" -> "180 dequantize_per_channel_default_35" [label="(1000,)", style=solid]; +"178 _zero_point_35" -> "180 dequantize_per_channel_default_35" [label="(1000,)", style=solid]; +"179 _frozen_param35" -> "180 dequantize_per_channel_default_35" [label="(1000, 1024)", style=solid]; +"180 dequantize_per_channel_default_35" -> "498 linear_1" [label=None, style=solid]; +"181 classifier_3_bias" -> "498 linear_1" [label="(1000,)", style=solid]; +"182 x" -> "183 quantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; +"183 quantize_per_tensor_default" -> "184 dequantize_per_tensor_default" [label=None, style=solid]; +"184 dequantize_per_tensor_default" -> "186 conv2d" [label=None, style=solid]; +"185 features_0_0_weight_bias" -> "186 conv2d" [label="(16,)", style=solid]; +"186 conv2d" -> "187 hardswish_" [label="(1, 16, 112, 112)", style=solid]; +"187 hardswish_" -> "190 quantize_per_channel_default_36" [label="(1, 16, 112, 112)", style=solid]; +"188 _2_scale_0" -> "190 quantize_per_channel_default_36" [label="(16,)", style=solid]; +"188 _2_scale_0" -> "191 dequantize_per_channel_default_36" [label="(16,)", style=solid]; +"189 _2_zero_point_0" -> "190 quantize_per_channel_default_36" [label="(16,)", style=solid]; +"189 _2_zero_point_0" -> "191 dequantize_per_channel_default_36" [label="(16,)", style=solid]; +"190 quantize_per_channel_default_36" -> "191 dequantize_per_channel_default_36" [label=None, style=solid]; +"191 dequantize_per_channel_default_36" -> "193 conv2d_1" [label=None, style=solid]; +"192 features_1_block_0_0_weight_bias" -> "193 conv2d_1" [label="(16,)", style=solid]; +"193 conv2d_1" -> "194 relu_" [label="(1, 16, 56, 56)", style=solid]; +"194 relu_" -> "195 quantize_per_tensor_default_1" [label="(1, 16, 56, 56)", style=solid]; +"195 quantize_per_tensor_default_1" -> "196 dequantize_per_tensor_default_1" [label=None, style=solid]; +"195 quantize_per_tensor_default_1" -> "197 dequantize_per_tensor_default_50" [label=None, style=solid]; +"196 dequantize_per_tensor_default_1" -> "198 adaptive_avg_pool2d" [label=None, style=solid]; +"197 dequantize_per_tensor_default_50" -> "205 mul" [label=None, style=solid]; +"198 adaptive_avg_pool2d" -> "199 conv2d_2" [label="(1, 16, 1, 1)", style=solid]; +"199 conv2d_2" -> "200 relu" [label="(1, 8, 1, 1)", style=solid]; +"200 relu" -> "201 conv2d_3" [label="(1, 8, 1, 1)", style=solid]; +"201 conv2d_3" -> "202 hardsigmoid" [label="(1, 16, 1, 1)", style=solid]; +"202 hardsigmoid" -> "203 quantize_per_tensor_default_2" [label="(1, 16, 1, 1)", style=solid]; +"203 quantize_per_tensor_default_2" -> "204 dequantize_per_tensor_default_2" [label=None, style=solid]; +"204 dequantize_per_tensor_default_2" -> "205 mul" [label=None, style=solid]; +"205 mul" -> "206 quantize_per_tensor_default_3" [label="(1, 16, 56, 56)", style=solid]; +"206 quantize_per_tensor_default_3" -> "207 dequantize_per_tensor_default_3" [label=None, style=solid]; +"207 dequantize_per_tensor_default_3" -> "209 conv2d_4" [label=None, style=solid]; +"208 features_1_block_2_0_weight_bias" -> "209 conv2d_4" [label="(16,)", style=solid]; +"209 conv2d_4" -> "210 quantize_per_tensor_default_4" [label="(1, 16, 56, 56)", style=solid]; +"210 quantize_per_tensor_default_4" -> "211 dequantize_per_tensor_default_4" [label=None, style=solid]; +"211 dequantize_per_tensor_default_4" -> "213 conv2d_5" [label=None, style=solid]; +"212 features_2_block_0_0_weight_bias" -> "213 conv2d_5" [label="(72,)", style=solid]; +"213 conv2d_5" -> "214 relu__1" [label="(1, 72, 56, 56)", style=solid]; +"214 relu__1" -> "217 quantize_per_channel_default_37" [label="(1, 72, 56, 56)", style=solid]; +"215 _2_scale_1" -> "217 quantize_per_channel_default_37" [label="(72,)", style=solid]; +"215 _2_scale_1" -> "218 dequantize_per_channel_default_37" [label="(72,)", style=solid]; +"216 _2_zero_point_1" -> "217 quantize_per_channel_default_37" [label="(72,)", style=solid]; +"216 _2_zero_point_1" -> "218 dequantize_per_channel_default_37" [label="(72,)", style=solid]; +"217 quantize_per_channel_default_37" -> "218 dequantize_per_channel_default_37" [label=None, style=solid]; +"218 dequantize_per_channel_default_37" -> "220 conv2d_6" [label=None, style=solid]; +"219 features_2_block_1_0_weight_bias" -> "220 conv2d_6" [label="(72,)", style=solid]; +"220 conv2d_6" -> "221 relu__2" [label="(1, 72, 28, 28)", style=solid]; +"221 relu__2" -> "222 quantize_per_tensor_default_5" [label="(1, 72, 28, 28)", style=solid]; +"222 quantize_per_tensor_default_5" -> "223 dequantize_per_tensor_default_5" [label=None, style=solid]; +"223 dequantize_per_tensor_default_5" -> "225 conv2d_7" [label=None, style=solid]; +"224 features_2_block_2_0_weight_bias" -> "225 conv2d_7" [label="(24,)", style=solid]; +"225 conv2d_7" -> "226 quantize_per_tensor_default_6" [label="(1, 24, 28, 28)", style=solid]; +"226 quantize_per_tensor_default_6" -> "227 dequantize_per_tensor_default_52" [label=None, style=solid]; +"226 quantize_per_tensor_default_6" -> "228 dequantize_per_tensor_default_51" [label=None, style=solid]; +"227 dequantize_per_tensor_default_52" -> "245 add_" [label=None, style=solid]; +"228 dequantize_per_tensor_default_51" -> "230 conv2d_8" [label=None, style=solid]; +"229 features_3_block_0_0_weight_bias" -> "230 conv2d_8" [label="(88,)", style=solid]; +"230 conv2d_8" -> "231 relu__3" [label="(1, 88, 28, 28)", style=solid]; +"231 relu__3" -> "234 quantize_per_channel_default_38" [label="(1, 88, 28, 28)", style=solid]; +"232 _2_scale_2" -> "234 quantize_per_channel_default_38" [label="(88,)", style=solid]; +"232 _2_scale_2" -> "235 dequantize_per_channel_default_38" [label="(88,)", style=solid]; +"233 _2_zero_point_2" -> "234 quantize_per_channel_default_38" [label="(88,)", style=solid]; +"233 _2_zero_point_2" -> "235 dequantize_per_channel_default_38" [label="(88,)", style=solid]; +"234 quantize_per_channel_default_38" -> "235 dequantize_per_channel_default_38" [label=None, style=solid]; +"235 dequantize_per_channel_default_38" -> "237 conv2d_9" [label=None, style=solid]; +"236 features_3_block_1_0_weight_bias" -> "237 conv2d_9" [label="(88,)", style=solid]; +"237 conv2d_9" -> "238 relu__4" [label="(1, 88, 28, 28)", style=solid]; +"238 relu__4" -> "239 quantize_per_tensor_default_7" [label="(1, 88, 28, 28)", style=solid]; +"239 quantize_per_tensor_default_7" -> "240 dequantize_per_tensor_default_7" [label=None, style=solid]; +"240 dequantize_per_tensor_default_7" -> "242 conv2d_10" [label=None, style=solid]; +"241 features_3_block_2_0_weight_bias" -> "242 conv2d_10" [label="(24,)", style=solid]; +"242 conv2d_10" -> "243 quantize_per_tensor_default_8" [label="(1, 24, 28, 28)", style=solid]; +"243 quantize_per_tensor_default_8" -> "244 dequantize_per_tensor_default_8" [label=None, style=solid]; +"244 dequantize_per_tensor_default_8" -> "245 add_" [label=None, style=solid]; +"245 add_" -> "246 quantize_per_tensor_default_9" [label="(1, 24, 28, 28)", style=solid]; +"246 quantize_per_tensor_default_9" -> "247 dequantize_per_tensor_default_9" [label=None, style=solid]; +"247 dequantize_per_tensor_default_9" -> "249 conv2d_11" [label=None, style=solid]; +"248 features_4_block_0_0_weight_bias" -> "249 conv2d_11" [label="(96,)", style=solid]; +"249 conv2d_11" -> "250 hardswish__1" [label="(1, 96, 28, 28)", style=solid]; +"250 hardswish__1" -> "253 quantize_per_channel_default_39" [label="(1, 96, 28, 28)", style=solid]; +"251 _2_scale_3" -> "253 quantize_per_channel_default_39" [label="(96,)", style=solid]; +"251 _2_scale_3" -> "254 dequantize_per_channel_default_39" [label="(96,)", style=solid]; +"252 _2_zero_point_3" -> "253 quantize_per_channel_default_39" [label="(96,)", style=solid]; +"252 _2_zero_point_3" -> "254 dequantize_per_channel_default_39" [label="(96,)", style=solid]; +"253 quantize_per_channel_default_39" -> "254 dequantize_per_channel_default_39" [label=None, style=solid]; +"254 dequantize_per_channel_default_39" -> "256 conv2d_12" [label=None, style=solid]; +"255 features_4_block_1_0_weight_bias" -> "256 conv2d_12" [label="(96,)", style=solid]; +"256 conv2d_12" -> "257 hardswish__2" [label="(1, 96, 14, 14)", style=solid]; +"257 hardswish__2" -> "258 quantize_per_tensor_default_10" [label="(1, 96, 14, 14)", style=solid]; +"258 quantize_per_tensor_default_10" -> "259 dequantize_per_tensor_default_10" [label=None, style=solid]; +"258 quantize_per_tensor_default_10" -> "260 dequantize_per_tensor_default_53" [label=None, style=solid]; +"259 dequantize_per_tensor_default_10" -> "261 adaptive_avg_pool2d_1" [label=None, style=solid]; +"260 dequantize_per_tensor_default_53" -> "268 mul_1" [label=None, style=solid]; +"261 adaptive_avg_pool2d_1" -> "262 conv2d_13" [label="(1, 96, 1, 1)", style=solid]; +"262 conv2d_13" -> "263 relu_1" [label="(1, 24, 1, 1)", style=solid]; +"263 relu_1" -> "264 conv2d_14" [label="(1, 24, 1, 1)", style=solid]; +"264 conv2d_14" -> "265 hardsigmoid_1" [label="(1, 96, 1, 1)", style=solid]; +"265 hardsigmoid_1" -> "266 quantize_per_tensor_default_11" [label="(1, 96, 1, 1)", style=solid]; +"266 quantize_per_tensor_default_11" -> "267 dequantize_per_tensor_default_11" [label=None, style=solid]; +"267 dequantize_per_tensor_default_11" -> "268 mul_1" [label=None, style=solid]; +"268 mul_1" -> "269 quantize_per_tensor_default_12" [label="(1, 96, 14, 14)", style=solid]; +"269 quantize_per_tensor_default_12" -> "270 dequantize_per_tensor_default_12" [label=None, style=solid]; +"270 dequantize_per_tensor_default_12" -> "272 conv2d_15" [label=None, style=solid]; +"271 features_4_block_3_0_weight_bias" -> "272 conv2d_15" [label="(40,)", style=solid]; +"272 conv2d_15" -> "273 quantize_per_tensor_default_13" [label="(1, 40, 14, 14)", style=solid]; +"273 quantize_per_tensor_default_13" -> "274 dequantize_per_tensor_default_55" [label=None, style=solid]; +"273 quantize_per_tensor_default_13" -> "275 dequantize_per_tensor_default_54" [label=None, style=solid]; +"274 dequantize_per_tensor_default_55" -> "303 add__1" [label=None, style=solid]; +"275 dequantize_per_tensor_default_54" -> "277 conv2d_16" [label=None, style=solid]; +"276 features_5_block_0_0_weight_bias" -> "277 conv2d_16" [label="(240,)", style=solid]; +"277 conv2d_16" -> "278 hardswish__3" [label="(1, 240, 14, 14)", style=solid]; +"278 hardswish__3" -> "281 quantize_per_channel_default_40" [label="(1, 240, 14, 14)", style=solid]; +"279 _2_scale_4" -> "281 quantize_per_channel_default_40" [label="(240,)", style=solid]; +"279 _2_scale_4" -> "282 dequantize_per_channel_default_40" [label="(240,)", style=solid]; +"280 _2_zero_point_4" -> "281 quantize_per_channel_default_40" [label="(240,)", style=solid]; +"280 _2_zero_point_4" -> "282 dequantize_per_channel_default_40" [label="(240,)", style=solid]; +"281 quantize_per_channel_default_40" -> "282 dequantize_per_channel_default_40" [label=None, style=solid]; +"282 dequantize_per_channel_default_40" -> "284 conv2d_17" [label=None, style=solid]; +"283 features_5_block_1_0_weight_bias" -> "284 conv2d_17" [label="(240,)", style=solid]; +"284 conv2d_17" -> "285 hardswish__4" [label="(1, 240, 14, 14)", style=solid]; +"285 hardswish__4" -> "286 quantize_per_tensor_default_14" [label="(1, 240, 14, 14)", style=solid]; +"286 quantize_per_tensor_default_14" -> "287 dequantize_per_tensor_default_14" [label=None, style=solid]; +"286 quantize_per_tensor_default_14" -> "288 dequantize_per_tensor_default_56" [label=None, style=solid]; +"287 dequantize_per_tensor_default_14" -> "289 adaptive_avg_pool2d_2" [label=None, style=solid]; +"288 dequantize_per_tensor_default_56" -> "296 mul_2" [label=None, style=solid]; +"289 adaptive_avg_pool2d_2" -> "290 conv2d_18" [label="(1, 240, 1, 1)", style=solid]; +"290 conv2d_18" -> "291 relu_2" [label="(1, 64, 1, 1)", style=solid]; +"291 relu_2" -> "292 conv2d_19" [label="(1, 64, 1, 1)", style=solid]; +"292 conv2d_19" -> "293 hardsigmoid_2" [label="(1, 240, 1, 1)", style=solid]; +"293 hardsigmoid_2" -> "294 quantize_per_tensor_default_15" [label="(1, 240, 1, 1)", style=solid]; +"294 quantize_per_tensor_default_15" -> "295 dequantize_per_tensor_default_15" [label=None, style=solid]; +"295 dequantize_per_tensor_default_15" -> "296 mul_2" [label=None, style=solid]; +"296 mul_2" -> "297 quantize_per_tensor_default_16" [label="(1, 240, 14, 14)", style=solid]; +"297 quantize_per_tensor_default_16" -> "298 dequantize_per_tensor_default_16" [label=None, style=solid]; +"298 dequantize_per_tensor_default_16" -> "300 conv2d_20" [label=None, style=solid]; +"299 features_5_block_3_0_weight_bias" -> "300 conv2d_20" [label="(40,)", style=solid]; +"300 conv2d_20" -> "301 quantize_per_tensor_default_17" [label="(1, 40, 14, 14)", style=solid]; +"301 quantize_per_tensor_default_17" -> "302 dequantize_per_tensor_default_17" [label=None, style=solid]; +"302 dequantize_per_tensor_default_17" -> "303 add__1" [label=None, style=solid]; +"303 add__1" -> "304 quantize_per_tensor_default_18" [label="(1, 40, 14, 14)", style=solid]; +"304 quantize_per_tensor_default_18" -> "305 dequantize_per_tensor_default_58" [label=None, style=solid]; +"304 quantize_per_tensor_default_18" -> "306 dequantize_per_tensor_default_57" [label=None, style=solid]; +"305 dequantize_per_tensor_default_58" -> "334 add__2" [label=None, style=solid]; +"306 dequantize_per_tensor_default_57" -> "308 conv2d_21" [label=None, style=solid]; +"307 features_6_block_0_0_weight_bias" -> "308 conv2d_21" [label="(240,)", style=solid]; +"308 conv2d_21" -> "309 hardswish__5" [label="(1, 240, 14, 14)", style=solid]; +"309 hardswish__5" -> "312 quantize_per_channel_default_41" [label="(1, 240, 14, 14)", style=solid]; +"310 _2_scale_5" -> "312 quantize_per_channel_default_41" [label="(240,)", style=solid]; +"310 _2_scale_5" -> "313 dequantize_per_channel_default_41" [label="(240,)", style=solid]; +"311 _2_zero_point_5" -> "312 quantize_per_channel_default_41" [label="(240,)", style=solid]; +"311 _2_zero_point_5" -> "313 dequantize_per_channel_default_41" [label="(240,)", style=solid]; +"312 quantize_per_channel_default_41" -> "313 dequantize_per_channel_default_41" [label=None, style=solid]; +"313 dequantize_per_channel_default_41" -> "315 conv2d_22" [label=None, style=solid]; +"314 features_6_block_1_0_weight_bias" -> "315 conv2d_22" [label="(240,)", style=solid]; +"315 conv2d_22" -> "316 hardswish__6" [label="(1, 240, 14, 14)", style=solid]; +"316 hardswish__6" -> "317 quantize_per_tensor_default_19" [label="(1, 240, 14, 14)", style=solid]; +"317 quantize_per_tensor_default_19" -> "318 dequantize_per_tensor_default_19" [label=None, style=solid]; +"317 quantize_per_tensor_default_19" -> "319 dequantize_per_tensor_default_59" [label=None, style=solid]; +"318 dequantize_per_tensor_default_19" -> "320 adaptive_avg_pool2d_3" [label=None, style=solid]; +"319 dequantize_per_tensor_default_59" -> "327 mul_3" [label=None, style=solid]; +"320 adaptive_avg_pool2d_3" -> "321 conv2d_23" [label="(1, 240, 1, 1)", style=solid]; +"321 conv2d_23" -> "322 relu_3" [label="(1, 64, 1, 1)", style=solid]; +"322 relu_3" -> "323 conv2d_24" [label="(1, 64, 1, 1)", style=solid]; +"323 conv2d_24" -> "324 hardsigmoid_3" [label="(1, 240, 1, 1)", style=solid]; +"324 hardsigmoid_3" -> "325 quantize_per_tensor_default_20" [label="(1, 240, 1, 1)", style=solid]; +"325 quantize_per_tensor_default_20" -> "326 dequantize_per_tensor_default_20" [label=None, style=solid]; +"326 dequantize_per_tensor_default_20" -> "327 mul_3" [label=None, style=solid]; +"327 mul_3" -> "328 quantize_per_tensor_default_21" [label="(1, 240, 14, 14)", style=solid]; +"328 quantize_per_tensor_default_21" -> "329 dequantize_per_tensor_default_21" [label=None, style=solid]; +"329 dequantize_per_tensor_default_21" -> "331 conv2d_25" [label=None, style=solid]; +"330 features_6_block_3_0_weight_bias" -> "331 conv2d_25" [label="(40,)", style=solid]; +"331 conv2d_25" -> "332 quantize_per_tensor_default_22" [label="(1, 40, 14, 14)", style=solid]; +"332 quantize_per_tensor_default_22" -> "333 dequantize_per_tensor_default_22" [label=None, style=solid]; +"333 dequantize_per_tensor_default_22" -> "334 add__2" [label=None, style=solid]; +"334 add__2" -> "335 quantize_per_tensor_default_23" [label="(1, 40, 14, 14)", style=solid]; +"335 quantize_per_tensor_default_23" -> "336 dequantize_per_tensor_default_23" [label=None, style=solid]; +"336 dequantize_per_tensor_default_23" -> "338 conv2d_26" [label=None, style=solid]; +"337 features_7_block_0_0_weight_bias" -> "338 conv2d_26" [label="(120,)", style=solid]; +"338 conv2d_26" -> "339 hardswish__7" [label="(1, 120, 14, 14)", style=solid]; +"339 hardswish__7" -> "342 quantize_per_channel_default_42" [label="(1, 120, 14, 14)", style=solid]; +"340 _2_scale_6" -> "342 quantize_per_channel_default_42" [label="(120,)", style=solid]; +"340 _2_scale_6" -> "343 dequantize_per_channel_default_42" [label="(120,)", style=solid]; +"341 _2_zero_point_6" -> "342 quantize_per_channel_default_42" [label="(120,)", style=solid]; +"341 _2_zero_point_6" -> "343 dequantize_per_channel_default_42" [label="(120,)", style=solid]; +"342 quantize_per_channel_default_42" -> "343 dequantize_per_channel_default_42" [label=None, style=solid]; +"343 dequantize_per_channel_default_42" -> "345 conv2d_27" [label=None, style=solid]; +"344 features_7_block_1_0_weight_bias" -> "345 conv2d_27" [label="(120,)", style=solid]; +"345 conv2d_27" -> "346 hardswish__8" [label="(1, 120, 14, 14)", style=solid]; +"346 hardswish__8" -> "347 quantize_per_tensor_default_24" [label="(1, 120, 14, 14)", style=solid]; +"347 quantize_per_tensor_default_24" -> "348 dequantize_per_tensor_default_24" [label=None, style=solid]; +"347 quantize_per_tensor_default_24" -> "349 dequantize_per_tensor_default_60" [label=None, style=solid]; +"348 dequantize_per_tensor_default_24" -> "350 adaptive_avg_pool2d_4" [label=None, style=solid]; +"349 dequantize_per_tensor_default_60" -> "357 mul_4" [label=None, style=solid]; +"350 adaptive_avg_pool2d_4" -> "351 conv2d_28" [label="(1, 120, 1, 1)", style=solid]; +"351 conv2d_28" -> "352 relu_4" [label="(1, 32, 1, 1)", style=solid]; +"352 relu_4" -> "353 conv2d_29" [label="(1, 32, 1, 1)", style=solid]; +"353 conv2d_29" -> "354 hardsigmoid_4" [label="(1, 120, 1, 1)", style=solid]; +"354 hardsigmoid_4" -> "355 quantize_per_tensor_default_25" [label="(1, 120, 1, 1)", style=solid]; +"355 quantize_per_tensor_default_25" -> "356 dequantize_per_tensor_default_25" [label=None, style=solid]; +"356 dequantize_per_tensor_default_25" -> "357 mul_4" [label=None, style=solid]; +"357 mul_4" -> "358 quantize_per_tensor_default_26" [label="(1, 120, 14, 14)", style=solid]; +"358 quantize_per_tensor_default_26" -> "359 dequantize_per_tensor_default_26" [label=None, style=solid]; +"359 dequantize_per_tensor_default_26" -> "361 conv2d_30" [label=None, style=solid]; +"360 features_7_block_3_0_weight_bias" -> "361 conv2d_30" [label="(48,)", style=solid]; +"361 conv2d_30" -> "362 quantize_per_tensor_default_27" [label="(1, 48, 14, 14)", style=solid]; +"362 quantize_per_tensor_default_27" -> "363 dequantize_per_tensor_default_62" [label=None, style=solid]; +"362 quantize_per_tensor_default_27" -> "364 dequantize_per_tensor_default_61" [label=None, style=solid]; +"363 dequantize_per_tensor_default_62" -> "392 add__3" [label=None, style=solid]; +"364 dequantize_per_tensor_default_61" -> "366 conv2d_31" [label=None, style=solid]; +"365 features_8_block_0_0_weight_bias" -> "366 conv2d_31" [label="(144,)", style=solid]; +"366 conv2d_31" -> "367 hardswish__9" [label="(1, 144, 14, 14)", style=solid]; +"367 hardswish__9" -> "370 quantize_per_channel_default_43" [label="(1, 144, 14, 14)", style=solid]; +"368 _2_scale_7" -> "370 quantize_per_channel_default_43" [label="(144,)", style=solid]; +"368 _2_scale_7" -> "371 dequantize_per_channel_default_43" [label="(144,)", style=solid]; +"369 _2_zero_point_7" -> "370 quantize_per_channel_default_43" [label="(144,)", style=solid]; +"369 _2_zero_point_7" -> "371 dequantize_per_channel_default_43" [label="(144,)", style=solid]; +"370 quantize_per_channel_default_43" -> "371 dequantize_per_channel_default_43" [label=None, style=solid]; +"371 dequantize_per_channel_default_43" -> "373 conv2d_32" [label=None, style=solid]; +"372 features_8_block_1_0_weight_bias" -> "373 conv2d_32" [label="(144,)", style=solid]; +"373 conv2d_32" -> "374 hardswish__10" [label="(1, 144, 14, 14)", style=solid]; +"374 hardswish__10" -> "375 quantize_per_tensor_default_28" [label="(1, 144, 14, 14)", style=solid]; +"375 quantize_per_tensor_default_28" -> "376 dequantize_per_tensor_default_28" [label=None, style=solid]; +"375 quantize_per_tensor_default_28" -> "377 dequantize_per_tensor_default_63" [label=None, style=solid]; +"376 dequantize_per_tensor_default_28" -> "378 adaptive_avg_pool2d_5" [label=None, style=solid]; +"377 dequantize_per_tensor_default_63" -> "385 mul_5" [label=None, style=solid]; +"378 adaptive_avg_pool2d_5" -> "379 conv2d_33" [label="(1, 144, 1, 1)", style=solid]; +"379 conv2d_33" -> "380 relu_5" [label="(1, 40, 1, 1)", style=solid]; +"380 relu_5" -> "381 conv2d_34" [label="(1, 40, 1, 1)", style=solid]; +"381 conv2d_34" -> "382 hardsigmoid_5" [label="(1, 144, 1, 1)", style=solid]; +"382 hardsigmoid_5" -> "383 quantize_per_tensor_default_29" [label="(1, 144, 1, 1)", style=solid]; +"383 quantize_per_tensor_default_29" -> "384 dequantize_per_tensor_default_29" [label=None, style=solid]; +"384 dequantize_per_tensor_default_29" -> "385 mul_5" [label=None, style=solid]; +"385 mul_5" -> "386 quantize_per_tensor_default_30" [label="(1, 144, 14, 14)", style=solid]; +"386 quantize_per_tensor_default_30" -> "387 dequantize_per_tensor_default_30" [label=None, style=solid]; +"387 dequantize_per_tensor_default_30" -> "389 conv2d_35" [label=None, style=solid]; +"388 features_8_block_3_0_weight_bias" -> "389 conv2d_35" [label="(48,)", style=solid]; +"389 conv2d_35" -> "390 quantize_per_tensor_default_31" [label="(1, 48, 14, 14)", style=solid]; +"390 quantize_per_tensor_default_31" -> "391 dequantize_per_tensor_default_31" [label=None, style=solid]; +"391 dequantize_per_tensor_default_31" -> "392 add__3" [label=None, style=solid]; +"392 add__3" -> "393 quantize_per_tensor_default_32" [label="(1, 48, 14, 14)", style=solid]; +"393 quantize_per_tensor_default_32" -> "394 dequantize_per_tensor_default_32" [label=None, style=solid]; +"394 dequantize_per_tensor_default_32" -> "396 conv2d_36" [label=None, style=solid]; +"395 features_9_block_0_0_weight_bias" -> "396 conv2d_36" [label="(288,)", style=solid]; +"396 conv2d_36" -> "397 hardswish__11" [label="(1, 288, 14, 14)", style=solid]; +"397 hardswish__11" -> "400 quantize_per_channel_default_44" [label="(1, 288, 14, 14)", style=solid]; +"398 _2_scale_8" -> "400 quantize_per_channel_default_44" [label="(288,)", style=solid]; +"398 _2_scale_8" -> "401 dequantize_per_channel_default_44" [label="(288,)", style=solid]; +"399 _2_zero_point_8" -> "400 quantize_per_channel_default_44" [label="(288,)", style=solid]; +"399 _2_zero_point_8" -> "401 dequantize_per_channel_default_44" [label="(288,)", style=solid]; +"400 quantize_per_channel_default_44" -> "401 dequantize_per_channel_default_44" [label=None, style=solid]; +"401 dequantize_per_channel_default_44" -> "403 conv2d_37" [label=None, style=solid]; +"402 features_9_block_1_0_weight_bias" -> "403 conv2d_37" [label="(288,)", style=solid]; +"403 conv2d_37" -> "404 hardswish__12" [label="(1, 288, 7, 7)", style=solid]; +"404 hardswish__12" -> "405 quantize_per_tensor_default_33" [label="(1, 288, 7, 7)", style=solid]; +"405 quantize_per_tensor_default_33" -> "406 dequantize_per_tensor_default_33" [label=None, style=solid]; +"405 quantize_per_tensor_default_33" -> "407 dequantize_per_tensor_default_64" [label=None, style=solid]; +"406 dequantize_per_tensor_default_33" -> "408 adaptive_avg_pool2d_6" [label=None, style=solid]; +"407 dequantize_per_tensor_default_64" -> "415 mul_6" [label=None, style=solid]; +"408 adaptive_avg_pool2d_6" -> "409 conv2d_38" [label="(1, 288, 1, 1)", style=solid]; +"409 conv2d_38" -> "410 relu_6" [label="(1, 72, 1, 1)", style=solid]; +"410 relu_6" -> "411 conv2d_39" [label="(1, 72, 1, 1)", style=solid]; +"411 conv2d_39" -> "412 hardsigmoid_6" [label="(1, 288, 1, 1)", style=solid]; +"412 hardsigmoid_6" -> "413 quantize_per_tensor_default_34" [label="(1, 288, 1, 1)", style=solid]; +"413 quantize_per_tensor_default_34" -> "414 dequantize_per_tensor_default_34" [label=None, style=solid]; +"414 dequantize_per_tensor_default_34" -> "415 mul_6" [label=None, style=solid]; +"415 mul_6" -> "416 quantize_per_tensor_default_35" [label="(1, 288, 7, 7)", style=solid]; +"416 quantize_per_tensor_default_35" -> "417 dequantize_per_tensor_default_35" [label=None, style=solid]; +"417 dequantize_per_tensor_default_35" -> "419 conv2d_40" [label=None, style=solid]; +"418 features_9_block_3_0_weight_bias" -> "419 conv2d_40" [label="(96,)", style=solid]; +"419 conv2d_40" -> "420 quantize_per_tensor_default_36" [label="(1, 96, 7, 7)", style=solid]; +"420 quantize_per_tensor_default_36" -> "421 dequantize_per_tensor_default_66" [label=None, style=solid]; +"420 quantize_per_tensor_default_36" -> "422 dequantize_per_tensor_default_65" [label=None, style=solid]; +"421 dequantize_per_tensor_default_66" -> "450 add__4" [label=None, style=solid]; +"422 dequantize_per_tensor_default_65" -> "424 conv2d_41" [label=None, style=solid]; +"423 features_10_block_0_0_weight_bias" -> "424 conv2d_41" [label="(576,)", style=solid]; +"424 conv2d_41" -> "425 hardswish__13" [label="(1, 576, 7, 7)", style=solid]; +"425 hardswish__13" -> "428 quantize_per_channel_default_45" [label="(1, 576, 7, 7)", style=solid]; +"426 _2_scale_9" -> "428 quantize_per_channel_default_45" [label="(576,)", style=solid]; +"426 _2_scale_9" -> "429 dequantize_per_channel_default_45" [label="(576,)", style=solid]; +"427 _2_zero_point_9" -> "428 quantize_per_channel_default_45" [label="(576,)", style=solid]; +"427 _2_zero_point_9" -> "429 dequantize_per_channel_default_45" [label="(576,)", style=solid]; +"428 quantize_per_channel_default_45" -> "429 dequantize_per_channel_default_45" [label=None, style=solid]; +"429 dequantize_per_channel_default_45" -> "431 conv2d_42" [label=None, style=solid]; +"430 features_10_block_1_0_weight_bias" -> "431 conv2d_42" [label="(576,)", style=solid]; +"431 conv2d_42" -> "432 hardswish__14" [label="(1, 576, 7, 7)", style=solid]; +"432 hardswish__14" -> "433 quantize_per_tensor_default_37" [label="(1, 576, 7, 7)", style=solid]; +"433 quantize_per_tensor_default_37" -> "434 dequantize_per_tensor_default_37" [label=None, style=solid]; +"433 quantize_per_tensor_default_37" -> "435 dequantize_per_tensor_default_67" [label=None, style=solid]; +"434 dequantize_per_tensor_default_37" -> "436 adaptive_avg_pool2d_7" [label=None, style=solid]; +"435 dequantize_per_tensor_default_67" -> "443 mul_7" [label=None, style=solid]; +"436 adaptive_avg_pool2d_7" -> "437 conv2d_43" [label="(1, 576, 1, 1)", style=solid]; +"437 conv2d_43" -> "438 relu_7" [label="(1, 144, 1, 1)", style=solid]; +"438 relu_7" -> "439 conv2d_44" [label="(1, 144, 1, 1)", style=solid]; +"439 conv2d_44" -> "440 hardsigmoid_7" [label="(1, 576, 1, 1)", style=solid]; +"440 hardsigmoid_7" -> "441 quantize_per_tensor_default_38" [label="(1, 576, 1, 1)", style=solid]; +"441 quantize_per_tensor_default_38" -> "442 dequantize_per_tensor_default_38" [label=None, style=solid]; +"442 dequantize_per_tensor_default_38" -> "443 mul_7" [label=None, style=solid]; +"443 mul_7" -> "444 quantize_per_tensor_default_39" [label="(1, 576, 7, 7)", style=solid]; +"444 quantize_per_tensor_default_39" -> "445 dequantize_per_tensor_default_39" [label=None, style=solid]; +"445 dequantize_per_tensor_default_39" -> "447 conv2d_45" [label=None, style=solid]; +"446 features_10_block_3_0_weight_bias" -> "447 conv2d_45" [label="(96,)", style=solid]; +"447 conv2d_45" -> "448 quantize_per_tensor_default_40" [label="(1, 96, 7, 7)", style=solid]; +"448 quantize_per_tensor_default_40" -> "449 dequantize_per_tensor_default_40" [label=None, style=solid]; +"449 dequantize_per_tensor_default_40" -> "450 add__4" [label=None, style=solid]; +"450 add__4" -> "451 quantize_per_tensor_default_41" [label="(1, 96, 7, 7)", style=solid]; +"451 quantize_per_tensor_default_41" -> "452 dequantize_per_tensor_default_69" [label=None, style=solid]; +"451 quantize_per_tensor_default_41" -> "453 dequantize_per_tensor_default_68" [label=None, style=solid]; +"452 dequantize_per_tensor_default_69" -> "481 add__5" [label=None, style=solid]; +"453 dequantize_per_tensor_default_68" -> "455 conv2d_46" [label=None, style=solid]; +"454 features_11_block_0_0_weight_bias" -> "455 conv2d_46" [label="(576,)", style=solid]; +"455 conv2d_46" -> "456 hardswish__15" [label="(1, 576, 7, 7)", style=solid]; +"456 hardswish__15" -> "459 quantize_per_channel_default_46" [label="(1, 576, 7, 7)", style=solid]; +"457 _2_scale_10" -> "459 quantize_per_channel_default_46" [label="(576,)", style=solid]; +"457 _2_scale_10" -> "460 dequantize_per_channel_default_46" [label="(576,)", style=solid]; +"458 _2_zero_point_10" -> "459 quantize_per_channel_default_46" [label="(576,)", style=solid]; +"458 _2_zero_point_10" -> "460 dequantize_per_channel_default_46" [label="(576,)", style=solid]; +"459 quantize_per_channel_default_46" -> "460 dequantize_per_channel_default_46" [label=None, style=solid]; +"460 dequantize_per_channel_default_46" -> "462 conv2d_47" [label=None, style=solid]; +"461 features_11_block_1_0_weight_bias" -> "462 conv2d_47" [label="(576,)", style=solid]; +"462 conv2d_47" -> "463 hardswish__16" [label="(1, 576, 7, 7)", style=solid]; +"463 hardswish__16" -> "464 quantize_per_tensor_default_42" [label="(1, 576, 7, 7)", style=solid]; +"464 quantize_per_tensor_default_42" -> "465 dequantize_per_tensor_default_42" [label=None, style=solid]; +"464 quantize_per_tensor_default_42" -> "466 dequantize_per_tensor_default_70" [label=None, style=solid]; +"465 dequantize_per_tensor_default_42" -> "467 adaptive_avg_pool2d_8" [label=None, style=solid]; +"466 dequantize_per_tensor_default_70" -> "474 mul_8" [label=None, style=solid]; +"467 adaptive_avg_pool2d_8" -> "468 conv2d_48" [label="(1, 576, 1, 1)", style=solid]; +"468 conv2d_48" -> "469 relu_8" [label="(1, 144, 1, 1)", style=solid]; +"469 relu_8" -> "470 conv2d_49" [label="(1, 144, 1, 1)", style=solid]; +"470 conv2d_49" -> "471 hardsigmoid_8" [label="(1, 576, 1, 1)", style=solid]; +"471 hardsigmoid_8" -> "472 quantize_per_tensor_default_43" [label="(1, 576, 1, 1)", style=solid]; +"472 quantize_per_tensor_default_43" -> "473 dequantize_per_tensor_default_43" [label=None, style=solid]; +"473 dequantize_per_tensor_default_43" -> "474 mul_8" [label=None, style=solid]; +"474 mul_8" -> "475 quantize_per_tensor_default_44" [label="(1, 576, 7, 7)", style=solid]; +"475 quantize_per_tensor_default_44" -> "476 dequantize_per_tensor_default_44" [label=None, style=solid]; +"476 dequantize_per_tensor_default_44" -> "478 conv2d_50" [label=None, style=solid]; +"477 features_11_block_3_0_weight_bias" -> "478 conv2d_50" [label="(96,)", style=solid]; +"478 conv2d_50" -> "479 quantize_per_tensor_default_45" [label="(1, 96, 7, 7)", style=solid]; +"479 quantize_per_tensor_default_45" -> "480 dequantize_per_tensor_default_45" [label=None, style=solid]; +"480 dequantize_per_tensor_default_45" -> "481 add__5" [label=None, style=solid]; +"481 add__5" -> "482 quantize_per_tensor_default_46" [label="(1, 96, 7, 7)", style=solid]; +"482 quantize_per_tensor_default_46" -> "483 dequantize_per_tensor_default_46" [label=None, style=solid]; +"483 dequantize_per_tensor_default_46" -> "485 conv2d_51" [label=None, style=solid]; +"484 features_12_0_weight_bias" -> "485 conv2d_51" [label="(576,)", style=solid]; +"485 conv2d_51" -> "486 hardswish__17" [label="(1, 576, 7, 7)", style=solid]; +"486 hardswish__17" -> "487 quantize_per_tensor_default_47" [label="(1, 576, 7, 7)", style=solid]; +"487 quantize_per_tensor_default_47" -> "488 dequantize_per_tensor_default_47" [label=None, style=solid]; +"488 dequantize_per_tensor_default_47" -> "489 adaptive_avg_pool2d_9" [label=None, style=solid]; +"489 adaptive_avg_pool2d_9" -> "490 quantize_per_tensor_default_48" [label="(1, 576, 1, 1)", style=solid]; +"490 quantize_per_tensor_default_48" -> "491 dequantize_per_tensor_default_48" [label=None, style=solid]; +"491 dequantize_per_tensor_default_48" -> "492 flatten" [label=None, style=solid]; +"492 flatten" -> "493 linear" [label="(1, 576)", style=solid]; +"493 linear" -> "494 hardswish__18" [label="(1, 1024)", style=solid]; +"494 hardswish__18" -> "495 quantize_per_tensor_default_49" [label="(1, 1024)", style=solid]; +"495 quantize_per_tensor_default_49" -> "496 dequantize_per_tensor_default_49" [label=None, style=solid]; +"496 dequantize_per_tensor_default_49" -> "497 dropout_" [label=None, style=solid]; +"497 dropout_" -> "498 linear_1" [label="(1, 1024)", style=solid]; +"498 linear_1" -> "499 output_1" [label="(1, 1000)", style=solid]; +} diff --git a/tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/resnet18.dot b/tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/resnet18.dot new file mode 100644 index 00000000000..8b57fcd27eb --- /dev/null +++ b/tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/resnet18.dot @@ -0,0 +1,455 @@ +strict digraph { +"0 _scale_0" [id=0, type=get_attr]; +"1 _zero_point_0" [id=1, type=get_attr]; +"2 _frozen_param0" [id=2, type=get_attr]; +"3 dequantize_per_channel_default" [id=3, type=dequantize_per_channel]; +"4 _scale_1" [id=4, type=get_attr]; +"5 _zero_point_1" [id=5, type=get_attr]; +"6 _frozen_param1" [id=6, type=get_attr]; +"7 dequantize_per_channel_default_1" [id=7, type=dequantize_per_channel]; +"8 _scale_2" [id=8, type=get_attr]; +"9 _zero_point_2" [id=9, type=get_attr]; +"10 _frozen_param2" [id=10, type=get_attr]; +"11 dequantize_per_channel_default_2" [id=11, type=dequantize_per_channel]; +"12 _scale_3" [id=12, type=get_attr]; +"13 _zero_point_3" [id=13, type=get_attr]; +"14 _frozen_param3" [id=14, type=get_attr]; +"15 dequantize_per_channel_default_3" [id=15, type=dequantize_per_channel]; +"16 _scale_4" [id=16, type=get_attr]; +"17 _zero_point_4" [id=17, type=get_attr]; +"18 _frozen_param4" [id=18, type=get_attr]; +"19 dequantize_per_channel_default_4" [id=19, type=dequantize_per_channel]; +"20 _scale_5" [id=20, type=get_attr]; +"21 _zero_point_5" [id=21, type=get_attr]; +"22 _frozen_param5" [id=22, type=get_attr]; +"23 dequantize_per_channel_default_5" [id=23, type=dequantize_per_channel]; +"24 _scale_6" [id=24, type=get_attr]; +"25 _zero_point_6" [id=25, type=get_attr]; +"26 _frozen_param6" [id=26, type=get_attr]; +"27 dequantize_per_channel_default_6" [id=27, type=dequantize_per_channel]; +"28 _scale_7" [id=28, type=get_attr]; +"29 _zero_point_7" [id=29, type=get_attr]; +"30 _frozen_param7" [id=30, type=get_attr]; +"31 dequantize_per_channel_default_7" [id=31, type=dequantize_per_channel]; +"32 _scale_8" [id=32, type=get_attr]; +"33 _zero_point_8" [id=33, type=get_attr]; +"34 _frozen_param8" [id=34, type=get_attr]; +"35 dequantize_per_channel_default_8" [id=35, type=dequantize_per_channel]; +"36 _scale_9" [id=36, type=get_attr]; +"37 _zero_point_9" [id=37, type=get_attr]; +"38 _frozen_param9" [id=38, type=get_attr]; +"39 dequantize_per_channel_default_9" [id=39, type=dequantize_per_channel]; +"40 _scale_10" [id=40, type=get_attr]; +"41 _zero_point_10" [id=41, type=get_attr]; +"42 _frozen_param10" [id=42, type=get_attr]; +"43 dequantize_per_channel_default_10" [id=43, type=dequantize_per_channel]; +"44 _scale_11" [id=44, type=get_attr]; +"45 _zero_point_11" [id=45, type=get_attr]; +"46 _frozen_param11" [id=46, type=get_attr]; +"47 dequantize_per_channel_default_11" [id=47, type=dequantize_per_channel]; +"48 _scale_12" [id=48, type=get_attr]; +"49 _zero_point_12" [id=49, type=get_attr]; +"50 _frozen_param12" [id=50, type=get_attr]; +"51 dequantize_per_channel_default_12" [id=51, type=dequantize_per_channel]; +"52 _scale_13" [id=52, type=get_attr]; +"53 _zero_point_13" [id=53, type=get_attr]; +"54 _frozen_param13" [id=54, type=get_attr]; +"55 dequantize_per_channel_default_13" [id=55, type=dequantize_per_channel]; +"56 _scale_14" [id=56, type=get_attr]; +"57 _zero_point_14" [id=57, type=get_attr]; +"58 _frozen_param14" [id=58, type=get_attr]; +"59 dequantize_per_channel_default_14" [id=59, type=dequantize_per_channel]; +"60 _scale_15" [id=60, type=get_attr]; +"61 _zero_point_15" [id=61, type=get_attr]; +"62 _frozen_param15" [id=62, type=get_attr]; +"63 dequantize_per_channel_default_15" [id=63, type=dequantize_per_channel]; +"64 _scale_16" [id=64, type=get_attr]; +"65 _zero_point_16" [id=65, type=get_attr]; +"66 _frozen_param16" [id=66, type=get_attr]; +"67 dequantize_per_channel_default_16" [id=67, type=dequantize_per_channel]; +"68 _scale_17" [id=68, type=get_attr]; +"69 _zero_point_17" [id=69, type=get_attr]; +"70 _frozen_param17" [id=70, type=get_attr]; +"71 dequantize_per_channel_default_17" [id=71, type=dequantize_per_channel]; +"72 _scale_18" [id=72, type=get_attr]; +"73 _zero_point_18" [id=73, type=get_attr]; +"74 _frozen_param18" [id=74, type=get_attr]; +"75 dequantize_per_channel_default_18" [id=75, type=dequantize_per_channel]; +"76 _scale_19" [id=76, type=get_attr]; +"77 _zero_point_19" [id=77, type=get_attr]; +"78 _frozen_param19" [id=78, type=get_attr]; +"79 dequantize_per_channel_default_19" [id=79, type=dequantize_per_channel]; +"80 _scale_20" [id=80, type=get_attr]; +"81 _zero_point_20" [id=81, type=get_attr]; +"82 _frozen_param20" [id=82, type=get_attr]; +"83 dequantize_per_channel_default_20" [id=83, type=dequantize_per_channel]; +"84 fc_bias" [id=84, type=get_attr]; +"85 x" [id=85, type=input]; +"86 quantize_per_tensor_default" [id=86, type=quantize_per_tensor]; +"87 dequantize_per_tensor_default" [id=87, type=dequantize_per_tensor]; +"88 conv1_weight_bias" [id=88, type=get_attr]; +"89 conv2d" [id=89, type=conv2d]; +"90 relu_" [id=90, type=relu_]; +"91 quantize_per_tensor_default_1" [id=91, type=quantize_per_tensor]; +"92 dequantize_per_tensor_default_1" [id=92, type=dequantize_per_tensor]; +"93 max_pool2d" [id=93, type=max_pool2d]; +"94 layer1_0_conv1_weight_bias" [id=94, type=get_attr]; +"95 conv2d_1" [id=95, type=conv2d]; +"96 relu__1" [id=96, type=relu_]; +"97 quantize_per_tensor_default_2" [id=97, type=quantize_per_tensor]; +"98 dequantize_per_tensor_default_2" [id=98, type=dequantize_per_tensor]; +"99 layer1_0_conv2_weight_bias" [id=99, type=get_attr]; +"100 conv2d_2" [id=100, type=conv2d]; +"101 quantize_per_tensor_default_3" [id=101, type=quantize_per_tensor]; +"102 dequantize_per_tensor_default_3" [id=102, type=dequantize_per_tensor]; +"103 add_" [id=103, type=add_]; +"104 relu__2" [id=104, type=relu_]; +"105 quantize_per_tensor_default_4" [id=105, type=quantize_per_tensor]; +"106 dequantize_per_tensor_default_4" [id=106, type=dequantize_per_tensor]; +"107 dequantize_per_tensor_default_30" [id=107, type=dequantize_per_tensor]; +"108 layer1_1_conv1_weight_bias" [id=108, type=get_attr]; +"109 conv2d_3" [id=109, type=conv2d]; +"110 relu__3" [id=110, type=relu_]; +"111 quantize_per_tensor_default_5" [id=111, type=quantize_per_tensor]; +"112 dequantize_per_tensor_default_5" [id=112, type=dequantize_per_tensor]; +"113 layer1_1_conv2_weight_bias" [id=113, type=get_attr]; +"114 conv2d_4" [id=114, type=conv2d]; +"115 quantize_per_tensor_default_6" [id=115, type=quantize_per_tensor]; +"116 dequantize_per_tensor_default_6" [id=116, type=dequantize_per_tensor]; +"117 add__1" [id=117, type=add_]; +"118 relu__4" [id=118, type=relu_]; +"119 quantize_per_tensor_default_7" [id=119, type=quantize_per_tensor]; +"120 dequantize_per_tensor_default_32" [id=120, type=dequantize_per_tensor]; +"121 dequantize_per_tensor_default_31" [id=121, type=dequantize_per_tensor]; +"122 layer2_0_conv1_weight_bias" [id=122, type=get_attr]; +"123 conv2d_5" [id=123, type=conv2d]; +"124 relu__5" [id=124, type=relu_]; +"125 quantize_per_tensor_default_8" [id=125, type=quantize_per_tensor]; +"126 dequantize_per_tensor_default_8" [id=126, type=dequantize_per_tensor]; +"127 layer2_0_conv2_weight_bias" [id=127, type=get_attr]; +"128 conv2d_6" [id=128, type=conv2d]; +"129 quantize_per_tensor_default_9" [id=129, type=quantize_per_tensor]; +"130 dequantize_per_tensor_default_9" [id=130, type=dequantize_per_tensor]; +"131 layer2_0_downsample_0_weight_bias" [id=131, type=get_attr]; +"132 conv2d_7" [id=132, type=conv2d]; +"133 quantize_per_tensor_default_10" [id=133, type=quantize_per_tensor]; +"134 dequantize_per_tensor_default_10" [id=134, type=dequantize_per_tensor]; +"135 add__2" [id=135, type=add_]; +"136 relu__6" [id=136, type=relu_]; +"137 quantize_per_tensor_default_11" [id=137, type=quantize_per_tensor]; +"138 dequantize_per_tensor_default_11" [id=138, type=dequantize_per_tensor]; +"139 dequantize_per_tensor_default_33" [id=139, type=dequantize_per_tensor]; +"140 layer2_1_conv1_weight_bias" [id=140, type=get_attr]; +"141 conv2d_8" [id=141, type=conv2d]; +"142 relu__7" [id=142, type=relu_]; +"143 quantize_per_tensor_default_12" [id=143, type=quantize_per_tensor]; +"144 dequantize_per_tensor_default_12" [id=144, type=dequantize_per_tensor]; +"145 layer2_1_conv2_weight_bias" [id=145, type=get_attr]; +"146 conv2d_9" [id=146, type=conv2d]; +"147 quantize_per_tensor_default_13" [id=147, type=quantize_per_tensor]; +"148 dequantize_per_tensor_default_13" [id=148, type=dequantize_per_tensor]; +"149 add__3" [id=149, type=add_]; +"150 relu__8" [id=150, type=relu_]; +"151 quantize_per_tensor_default_14" [id=151, type=quantize_per_tensor]; +"152 dequantize_per_tensor_default_35" [id=152, type=dequantize_per_tensor]; +"153 dequantize_per_tensor_default_34" [id=153, type=dequantize_per_tensor]; +"154 layer3_0_conv1_weight_bias" [id=154, type=get_attr]; +"155 conv2d_10" [id=155, type=conv2d]; +"156 relu__9" [id=156, type=relu_]; +"157 quantize_per_tensor_default_15" [id=157, type=quantize_per_tensor]; +"158 dequantize_per_tensor_default_15" [id=158, type=dequantize_per_tensor]; +"159 layer3_0_conv2_weight_bias" [id=159, type=get_attr]; +"160 conv2d_11" [id=160, type=conv2d]; +"161 quantize_per_tensor_default_16" [id=161, type=quantize_per_tensor]; +"162 dequantize_per_tensor_default_16" [id=162, type=dequantize_per_tensor]; +"163 layer3_0_downsample_0_weight_bias" [id=163, type=get_attr]; +"164 conv2d_12" [id=164, type=conv2d]; +"165 quantize_per_tensor_default_17" [id=165, type=quantize_per_tensor]; +"166 dequantize_per_tensor_default_17" [id=166, type=dequantize_per_tensor]; +"167 add__4" [id=167, type=add_]; +"168 relu__10" [id=168, type=relu_]; +"169 quantize_per_tensor_default_18" [id=169, type=quantize_per_tensor]; +"170 dequantize_per_tensor_default_18" [id=170, type=dequantize_per_tensor]; +"171 dequantize_per_tensor_default_36" [id=171, type=dequantize_per_tensor]; +"172 layer3_1_conv1_weight_bias" [id=172, type=get_attr]; +"173 conv2d_13" [id=173, type=conv2d]; +"174 relu__11" [id=174, type=relu_]; +"175 quantize_per_tensor_default_19" [id=175, type=quantize_per_tensor]; +"176 dequantize_per_tensor_default_19" [id=176, type=dequantize_per_tensor]; +"177 layer3_1_conv2_weight_bias" [id=177, type=get_attr]; +"178 conv2d_14" [id=178, type=conv2d]; +"179 quantize_per_tensor_default_20" [id=179, type=quantize_per_tensor]; +"180 dequantize_per_tensor_default_20" [id=180, type=dequantize_per_tensor]; +"181 add__5" [id=181, type=add_]; +"182 relu__12" [id=182, type=relu_]; +"183 quantize_per_tensor_default_21" [id=183, type=quantize_per_tensor]; +"184 dequantize_per_tensor_default_38" [id=184, type=dequantize_per_tensor]; +"185 dequantize_per_tensor_default_37" [id=185, type=dequantize_per_tensor]; +"186 layer4_0_conv1_weight_bias" [id=186, type=get_attr]; +"187 conv2d_15" [id=187, type=conv2d]; +"188 relu__13" [id=188, type=relu_]; +"189 quantize_per_tensor_default_22" [id=189, type=quantize_per_tensor]; +"190 dequantize_per_tensor_default_22" [id=190, type=dequantize_per_tensor]; +"191 layer4_0_conv2_weight_bias" [id=191, type=get_attr]; +"192 conv2d_16" [id=192, type=conv2d]; +"193 quantize_per_tensor_default_23" [id=193, type=quantize_per_tensor]; +"194 dequantize_per_tensor_default_23" [id=194, type=dequantize_per_tensor]; +"195 layer4_0_downsample_0_weight_bias" [id=195, type=get_attr]; +"196 conv2d_17" [id=196, type=conv2d]; +"197 quantize_per_tensor_default_24" [id=197, type=quantize_per_tensor]; +"198 dequantize_per_tensor_default_24" [id=198, type=dequantize_per_tensor]; +"199 add__6" [id=199, type=add_]; +"200 relu__14" [id=200, type=relu_]; +"201 quantize_per_tensor_default_25" [id=201, type=quantize_per_tensor]; +"202 dequantize_per_tensor_default_25" [id=202, type=dequantize_per_tensor]; +"203 dequantize_per_tensor_default_39" [id=203, type=dequantize_per_tensor]; +"204 layer4_1_conv1_weight_bias" [id=204, type=get_attr]; +"205 conv2d_18" [id=205, type=conv2d]; +"206 relu__15" [id=206, type=relu_]; +"207 quantize_per_tensor_default_26" [id=207, type=quantize_per_tensor]; +"208 dequantize_per_tensor_default_26" [id=208, type=dequantize_per_tensor]; +"209 layer4_1_conv2_weight_bias" [id=209, type=get_attr]; +"210 conv2d_19" [id=210, type=conv2d]; +"211 quantize_per_tensor_default_27" [id=211, type=quantize_per_tensor]; +"212 dequantize_per_tensor_default_27" [id=212, type=dequantize_per_tensor]; +"213 add__7" [id=213, type=add_]; +"214 relu__16" [id=214, type=relu_]; +"215 quantize_per_tensor_default_28" [id=215, type=quantize_per_tensor]; +"216 dequantize_per_tensor_default_28" [id=216, type=dequantize_per_tensor]; +"217 adaptive_avg_pool2d" [id=217, type=adaptive_avg_pool2d]; +"218 quantize_per_tensor_default_29" [id=218, type=quantize_per_tensor]; +"219 dequantize_per_tensor_default_29" [id=219, type=dequantize_per_tensor]; +"220 flatten" [id=220, type=flatten]; +"221 linear" [id=221, type=linear]; +"222 output_1" [id=222, type=output]; +"0 _scale_0" -> "3 dequantize_per_channel_default" [label="(64,)", style=solid]; +"1 _zero_point_0" -> "3 dequantize_per_channel_default" [label="(64,)", style=solid]; +"2 _frozen_param0" -> "3 dequantize_per_channel_default" [label="(64, 3, 7, 7)", style=solid]; +"3 dequantize_per_channel_default" -> "89 conv2d" [label=None, style=solid]; +"4 _scale_1" -> "7 dequantize_per_channel_default_1" [label="(64,)", style=solid]; +"5 _zero_point_1" -> "7 dequantize_per_channel_default_1" [label="(64,)", style=solid]; +"6 _frozen_param1" -> "7 dequantize_per_channel_default_1" [label="(64, 64, 3, 3)", style=solid]; +"7 dequantize_per_channel_default_1" -> "95 conv2d_1" [label=None, style=solid]; +"8 _scale_2" -> "11 dequantize_per_channel_default_2" [label="(64,)", style=solid]; +"9 _zero_point_2" -> "11 dequantize_per_channel_default_2" [label="(64,)", style=solid]; +"10 _frozen_param2" -> "11 dequantize_per_channel_default_2" [label="(64, 64, 3, 3)", style=solid]; +"11 dequantize_per_channel_default_2" -> "100 conv2d_2" [label=None, style=solid]; +"12 _scale_3" -> "15 dequantize_per_channel_default_3" [label="(64,)", style=solid]; +"13 _zero_point_3" -> "15 dequantize_per_channel_default_3" [label="(64,)", style=solid]; +"14 _frozen_param3" -> "15 dequantize_per_channel_default_3" [label="(64, 64, 3, 3)", style=solid]; +"15 dequantize_per_channel_default_3" -> "109 conv2d_3" [label=None, style=solid]; +"16 _scale_4" -> "19 dequantize_per_channel_default_4" [label="(64,)", style=solid]; +"17 _zero_point_4" -> "19 dequantize_per_channel_default_4" [label="(64,)", style=solid]; +"18 _frozen_param4" -> "19 dequantize_per_channel_default_4" [label="(64, 64, 3, 3)", style=solid]; +"19 dequantize_per_channel_default_4" -> "114 conv2d_4" [label=None, style=solid]; +"20 _scale_5" -> "23 dequantize_per_channel_default_5" [label="(128,)", style=solid]; +"21 _zero_point_5" -> "23 dequantize_per_channel_default_5" [label="(128,)", style=solid]; +"22 _frozen_param5" -> "23 dequantize_per_channel_default_5" [label="(128, 64, 3, 3)", style=solid]; +"23 dequantize_per_channel_default_5" -> "123 conv2d_5" [label=None, style=solid]; +"24 _scale_6" -> "27 dequantize_per_channel_default_6" [label="(128,)", style=solid]; +"25 _zero_point_6" -> "27 dequantize_per_channel_default_6" [label="(128,)", style=solid]; +"26 _frozen_param6" -> "27 dequantize_per_channel_default_6" [label="(128, 128, 3, 3)", style=solid]; +"27 dequantize_per_channel_default_6" -> "128 conv2d_6" [label=None, style=solid]; +"28 _scale_7" -> "31 dequantize_per_channel_default_7" [label="(128,)", style=solid]; +"29 _zero_point_7" -> "31 dequantize_per_channel_default_7" [label="(128,)", style=solid]; +"30 _frozen_param7" -> "31 dequantize_per_channel_default_7" [label="(128, 64, 1, 1)", style=solid]; +"31 dequantize_per_channel_default_7" -> "132 conv2d_7" [label=None, style=solid]; +"32 _scale_8" -> "35 dequantize_per_channel_default_8" [label="(128,)", style=solid]; +"33 _zero_point_8" -> "35 dequantize_per_channel_default_8" [label="(128,)", style=solid]; +"34 _frozen_param8" -> "35 dequantize_per_channel_default_8" [label="(128, 128, 3, 3)", style=solid]; +"35 dequantize_per_channel_default_8" -> "141 conv2d_8" [label=None, style=solid]; +"36 _scale_9" -> "39 dequantize_per_channel_default_9" [label="(128,)", style=solid]; +"37 _zero_point_9" -> "39 dequantize_per_channel_default_9" [label="(128,)", style=solid]; +"38 _frozen_param9" -> "39 dequantize_per_channel_default_9" [label="(128, 128, 3, 3)", style=solid]; +"39 dequantize_per_channel_default_9" -> "146 conv2d_9" [label=None, style=solid]; +"40 _scale_10" -> "43 dequantize_per_channel_default_10" [label="(256,)", style=solid]; +"41 _zero_point_10" -> "43 dequantize_per_channel_default_10" [label="(256,)", style=solid]; +"42 _frozen_param10" -> "43 dequantize_per_channel_default_10" [label="(256, 128, 3, 3)", style=solid]; +"43 dequantize_per_channel_default_10" -> "155 conv2d_10" [label=None, style=solid]; +"44 _scale_11" -> "47 dequantize_per_channel_default_11" [label="(256,)", style=solid]; +"45 _zero_point_11" -> "47 dequantize_per_channel_default_11" [label="(256,)", style=solid]; +"46 _frozen_param11" -> "47 dequantize_per_channel_default_11" [label="(256, 256, 3, 3)", style=solid]; +"47 dequantize_per_channel_default_11" -> "160 conv2d_11" [label=None, style=solid]; +"48 _scale_12" -> "51 dequantize_per_channel_default_12" [label="(256,)", style=solid]; +"49 _zero_point_12" -> "51 dequantize_per_channel_default_12" [label="(256,)", style=solid]; +"50 _frozen_param12" -> "51 dequantize_per_channel_default_12" [label="(256, 128, 1, 1)", style=solid]; +"51 dequantize_per_channel_default_12" -> "164 conv2d_12" [label=None, style=solid]; +"52 _scale_13" -> "55 dequantize_per_channel_default_13" [label="(256,)", style=solid]; +"53 _zero_point_13" -> "55 dequantize_per_channel_default_13" [label="(256,)", style=solid]; +"54 _frozen_param13" -> "55 dequantize_per_channel_default_13" [label="(256, 256, 3, 3)", style=solid]; +"55 dequantize_per_channel_default_13" -> "173 conv2d_13" [label=None, style=solid]; +"56 _scale_14" -> "59 dequantize_per_channel_default_14" [label="(256,)", style=solid]; +"57 _zero_point_14" -> "59 dequantize_per_channel_default_14" [label="(256,)", style=solid]; +"58 _frozen_param14" -> "59 dequantize_per_channel_default_14" [label="(256, 256, 3, 3)", style=solid]; +"59 dequantize_per_channel_default_14" -> "178 conv2d_14" [label=None, style=solid]; +"60 _scale_15" -> "63 dequantize_per_channel_default_15" [label="(512,)", style=solid]; +"61 _zero_point_15" -> "63 dequantize_per_channel_default_15" [label="(512,)", style=solid]; +"62 _frozen_param15" -> "63 dequantize_per_channel_default_15" [label="(512, 256, 3, 3)", style=solid]; +"63 dequantize_per_channel_default_15" -> "187 conv2d_15" [label=None, style=solid]; +"64 _scale_16" -> "67 dequantize_per_channel_default_16" [label="(512,)", style=solid]; +"65 _zero_point_16" -> "67 dequantize_per_channel_default_16" [label="(512,)", style=solid]; +"66 _frozen_param16" -> "67 dequantize_per_channel_default_16" [label="(512, 512, 3, 3)", style=solid]; +"67 dequantize_per_channel_default_16" -> "192 conv2d_16" [label=None, style=solid]; +"68 _scale_17" -> "71 dequantize_per_channel_default_17" [label="(512,)", style=solid]; +"69 _zero_point_17" -> "71 dequantize_per_channel_default_17" [label="(512,)", style=solid]; +"70 _frozen_param17" -> "71 dequantize_per_channel_default_17" [label="(512, 256, 1, 1)", style=solid]; +"71 dequantize_per_channel_default_17" -> "196 conv2d_17" [label=None, style=solid]; +"72 _scale_18" -> "75 dequantize_per_channel_default_18" [label="(512,)", style=solid]; +"73 _zero_point_18" -> "75 dequantize_per_channel_default_18" [label="(512,)", style=solid]; +"74 _frozen_param18" -> "75 dequantize_per_channel_default_18" [label="(512, 512, 3, 3)", style=solid]; +"75 dequantize_per_channel_default_18" -> "205 conv2d_18" [label=None, style=solid]; +"76 _scale_19" -> "79 dequantize_per_channel_default_19" [label="(512,)", style=solid]; +"77 _zero_point_19" -> "79 dequantize_per_channel_default_19" [label="(512,)", style=solid]; +"78 _frozen_param19" -> "79 dequantize_per_channel_default_19" [label="(512, 512, 3, 3)", style=solid]; +"79 dequantize_per_channel_default_19" -> "210 conv2d_19" [label=None, style=solid]; +"80 _scale_20" -> "83 dequantize_per_channel_default_20" [label="(1000,)", style=solid]; +"81 _zero_point_20" -> "83 dequantize_per_channel_default_20" [label="(1000,)", style=solid]; +"82 _frozen_param20" -> "83 dequantize_per_channel_default_20" [label="(1000, 512)", style=solid]; +"83 dequantize_per_channel_default_20" -> "221 linear" [label=None, style=solid]; +"84 fc_bias" -> "221 linear" [label="(1000,)", style=solid]; +"85 x" -> "86 quantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; +"86 quantize_per_tensor_default" -> "87 dequantize_per_tensor_default" [label=None, style=solid]; +"87 dequantize_per_tensor_default" -> "89 conv2d" [label=None, style=solid]; +"88 conv1_weight_bias" -> "89 conv2d" [label="(64,)", style=solid]; +"89 conv2d" -> "90 relu_" [label="(1, 64, 112, 112)", style=solid]; +"90 relu_" -> "91 quantize_per_tensor_default_1" [label="(1, 64, 112, 112)", style=solid]; +"91 quantize_per_tensor_default_1" -> "92 dequantize_per_tensor_default_1" [label=None, style=solid]; +"92 dequantize_per_tensor_default_1" -> "93 max_pool2d" [label=None, style=solid]; +"93 max_pool2d" -> "95 conv2d_1" [label="(1, 64, 56, 56)", style=solid]; +"93 max_pool2d" -> "103 add_" [label="(1, 64, 56, 56)", style=solid]; +"94 layer1_0_conv1_weight_bias" -> "95 conv2d_1" [label="(64,)", style=solid]; +"95 conv2d_1" -> "96 relu__1" [label="(1, 64, 56, 56)", style=solid]; +"96 relu__1" -> "97 quantize_per_tensor_default_2" [label="(1, 64, 56, 56)", style=solid]; +"97 quantize_per_tensor_default_2" -> "98 dequantize_per_tensor_default_2" [label=None, style=solid]; +"98 dequantize_per_tensor_default_2" -> "100 conv2d_2" [label=None, style=solid]; +"99 layer1_0_conv2_weight_bias" -> "100 conv2d_2" [label="(64,)", style=solid]; +"100 conv2d_2" -> "101 quantize_per_tensor_default_3" [label="(1, 64, 56, 56)", style=solid]; +"101 quantize_per_tensor_default_3" -> "102 dequantize_per_tensor_default_3" [label=None, style=solid]; +"102 dequantize_per_tensor_default_3" -> "103 add_" [label=None, style=solid]; +"103 add_" -> "104 relu__2" [label="(1, 64, 56, 56)", style=solid]; +"104 relu__2" -> "105 quantize_per_tensor_default_4" [label="(1, 64, 56, 56)", style=solid]; +"105 quantize_per_tensor_default_4" -> "106 dequantize_per_tensor_default_4" [label=None, style=solid]; +"105 quantize_per_tensor_default_4" -> "107 dequantize_per_tensor_default_30" [label=None, style=solid]; +"106 dequantize_per_tensor_default_4" -> "117 add__1" [label=None, style=solid]; +"107 dequantize_per_tensor_default_30" -> "109 conv2d_3" [label=None, style=solid]; +"108 layer1_1_conv1_weight_bias" -> "109 conv2d_3" [label="(64,)", style=solid]; +"109 conv2d_3" -> "110 relu__3" [label="(1, 64, 56, 56)", style=solid]; +"110 relu__3" -> "111 quantize_per_tensor_default_5" [label="(1, 64, 56, 56)", style=solid]; +"111 quantize_per_tensor_default_5" -> "112 dequantize_per_tensor_default_5" [label=None, style=solid]; +"112 dequantize_per_tensor_default_5" -> "114 conv2d_4" [label=None, style=solid]; +"113 layer1_1_conv2_weight_bias" -> "114 conv2d_4" [label="(64,)", style=solid]; +"114 conv2d_4" -> "115 quantize_per_tensor_default_6" [label="(1, 64, 56, 56)", style=solid]; +"115 quantize_per_tensor_default_6" -> "116 dequantize_per_tensor_default_6" [label=None, style=solid]; +"116 dequantize_per_tensor_default_6" -> "117 add__1" [label=None, style=solid]; +"117 add__1" -> "118 relu__4" [label="(1, 64, 56, 56)", style=solid]; +"118 relu__4" -> "119 quantize_per_tensor_default_7" [label="(1, 64, 56, 56)", style=solid]; +"119 quantize_per_tensor_default_7" -> "120 dequantize_per_tensor_default_32" [label=None, style=solid]; +"119 quantize_per_tensor_default_7" -> "121 dequantize_per_tensor_default_31" [label=None, style=solid]; +"120 dequantize_per_tensor_default_32" -> "132 conv2d_7" [label=None, style=solid]; +"121 dequantize_per_tensor_default_31" -> "123 conv2d_5" [label=None, style=solid]; +"122 layer2_0_conv1_weight_bias" -> "123 conv2d_5" [label="(128,)", style=solid]; +"123 conv2d_5" -> "124 relu__5" [label="(1, 128, 28, 28)", style=solid]; +"124 relu__5" -> "125 quantize_per_tensor_default_8" [label="(1, 128, 28, 28)", style=solid]; +"125 quantize_per_tensor_default_8" -> "126 dequantize_per_tensor_default_8" [label=None, style=solid]; +"126 dequantize_per_tensor_default_8" -> "128 conv2d_6" [label=None, style=solid]; +"127 layer2_0_conv2_weight_bias" -> "128 conv2d_6" [label="(128,)", style=solid]; +"128 conv2d_6" -> "129 quantize_per_tensor_default_9" [label="(1, 128, 28, 28)", style=solid]; +"129 quantize_per_tensor_default_9" -> "130 dequantize_per_tensor_default_9" [label=None, style=solid]; +"130 dequantize_per_tensor_default_9" -> "135 add__2" [label=None, style=solid]; +"131 layer2_0_downsample_0_weight_bias" -> "132 conv2d_7" [label="(128,)", style=solid]; +"132 conv2d_7" -> "133 quantize_per_tensor_default_10" [label="(1, 128, 28, 28)", style=solid]; +"133 quantize_per_tensor_default_10" -> "134 dequantize_per_tensor_default_10" [label=None, style=solid]; +"134 dequantize_per_tensor_default_10" -> "135 add__2" [label=None, style=solid]; +"135 add__2" -> "136 relu__6" [label="(1, 128, 28, 28)", style=solid]; +"136 relu__6" -> "137 quantize_per_tensor_default_11" [label="(1, 128, 28, 28)", style=solid]; +"137 quantize_per_tensor_default_11" -> "138 dequantize_per_tensor_default_11" [label=None, style=solid]; +"137 quantize_per_tensor_default_11" -> "139 dequantize_per_tensor_default_33" [label=None, style=solid]; +"138 dequantize_per_tensor_default_11" -> "149 add__3" [label=None, style=solid]; +"139 dequantize_per_tensor_default_33" -> "141 conv2d_8" [label=None, style=solid]; +"140 layer2_1_conv1_weight_bias" -> "141 conv2d_8" [label="(128,)", style=solid]; +"141 conv2d_8" -> "142 relu__7" [label="(1, 128, 28, 28)", style=solid]; +"142 relu__7" -> "143 quantize_per_tensor_default_12" [label="(1, 128, 28, 28)", style=solid]; +"143 quantize_per_tensor_default_12" -> "144 dequantize_per_tensor_default_12" [label=None, style=solid]; +"144 dequantize_per_tensor_default_12" -> "146 conv2d_9" [label=None, style=solid]; +"145 layer2_1_conv2_weight_bias" -> "146 conv2d_9" [label="(128,)", style=solid]; +"146 conv2d_9" -> "147 quantize_per_tensor_default_13" [label="(1, 128, 28, 28)", style=solid]; +"147 quantize_per_tensor_default_13" -> "148 dequantize_per_tensor_default_13" [label=None, style=solid]; +"148 dequantize_per_tensor_default_13" -> "149 add__3" [label=None, style=solid]; +"149 add__3" -> "150 relu__8" [label="(1, 128, 28, 28)", style=solid]; +"150 relu__8" -> "151 quantize_per_tensor_default_14" [label="(1, 128, 28, 28)", style=solid]; +"151 quantize_per_tensor_default_14" -> "152 dequantize_per_tensor_default_35" [label=None, style=solid]; +"151 quantize_per_tensor_default_14" -> "153 dequantize_per_tensor_default_34" [label=None, style=solid]; +"152 dequantize_per_tensor_default_35" -> "164 conv2d_12" [label=None, style=solid]; +"153 dequantize_per_tensor_default_34" -> "155 conv2d_10" [label=None, style=solid]; +"154 layer3_0_conv1_weight_bias" -> "155 conv2d_10" [label="(256,)", style=solid]; +"155 conv2d_10" -> "156 relu__9" [label="(1, 256, 14, 14)", style=solid]; +"156 relu__9" -> "157 quantize_per_tensor_default_15" [label="(1, 256, 14, 14)", style=solid]; +"157 quantize_per_tensor_default_15" -> "158 dequantize_per_tensor_default_15" [label=None, style=solid]; +"158 dequantize_per_tensor_default_15" -> "160 conv2d_11" [label=None, style=solid]; +"159 layer3_0_conv2_weight_bias" -> "160 conv2d_11" [label="(256,)", style=solid]; +"160 conv2d_11" -> "161 quantize_per_tensor_default_16" [label="(1, 256, 14, 14)", style=solid]; +"161 quantize_per_tensor_default_16" -> "162 dequantize_per_tensor_default_16" [label=None, style=solid]; +"162 dequantize_per_tensor_default_16" -> "167 add__4" [label=None, style=solid]; +"163 layer3_0_downsample_0_weight_bias" -> "164 conv2d_12" [label="(256,)", style=solid]; +"164 conv2d_12" -> "165 quantize_per_tensor_default_17" [label="(1, 256, 14, 14)", style=solid]; +"165 quantize_per_tensor_default_17" -> "166 dequantize_per_tensor_default_17" [label=None, style=solid]; +"166 dequantize_per_tensor_default_17" -> "167 add__4" [label=None, style=solid]; +"167 add__4" -> "168 relu__10" [label="(1, 256, 14, 14)", style=solid]; +"168 relu__10" -> "169 quantize_per_tensor_default_18" [label="(1, 256, 14, 14)", style=solid]; +"169 quantize_per_tensor_default_18" -> "170 dequantize_per_tensor_default_18" [label=None, style=solid]; +"169 quantize_per_tensor_default_18" -> "171 dequantize_per_tensor_default_36" [label=None, style=solid]; +"170 dequantize_per_tensor_default_18" -> "181 add__5" [label=None, style=solid]; +"171 dequantize_per_tensor_default_36" -> "173 conv2d_13" [label=None, style=solid]; +"172 layer3_1_conv1_weight_bias" -> "173 conv2d_13" [label="(256,)", style=solid]; +"173 conv2d_13" -> "174 relu__11" [label="(1, 256, 14, 14)", style=solid]; +"174 relu__11" -> "175 quantize_per_tensor_default_19" [label="(1, 256, 14, 14)", style=solid]; +"175 quantize_per_tensor_default_19" -> "176 dequantize_per_tensor_default_19" [label=None, style=solid]; +"176 dequantize_per_tensor_default_19" -> "178 conv2d_14" [label=None, style=solid]; +"177 layer3_1_conv2_weight_bias" -> "178 conv2d_14" [label="(256,)", style=solid]; +"178 conv2d_14" -> "179 quantize_per_tensor_default_20" [label="(1, 256, 14, 14)", style=solid]; +"179 quantize_per_tensor_default_20" -> "180 dequantize_per_tensor_default_20" [label=None, style=solid]; +"180 dequantize_per_tensor_default_20" -> "181 add__5" [label=None, style=solid]; +"181 add__5" -> "182 relu__12" [label="(1, 256, 14, 14)", style=solid]; +"182 relu__12" -> "183 quantize_per_tensor_default_21" [label="(1, 256, 14, 14)", style=solid]; +"183 quantize_per_tensor_default_21" -> "184 dequantize_per_tensor_default_38" [label=None, style=solid]; +"183 quantize_per_tensor_default_21" -> "185 dequantize_per_tensor_default_37" [label=None, style=solid]; +"184 dequantize_per_tensor_default_38" -> "196 conv2d_17" [label=None, style=solid]; +"185 dequantize_per_tensor_default_37" -> "187 conv2d_15" [label=None, style=solid]; +"186 layer4_0_conv1_weight_bias" -> "187 conv2d_15" [label="(512,)", style=solid]; +"187 conv2d_15" -> "188 relu__13" [label="(1, 512, 7, 7)", style=solid]; +"188 relu__13" -> "189 quantize_per_tensor_default_22" [label="(1, 512, 7, 7)", style=solid]; +"189 quantize_per_tensor_default_22" -> "190 dequantize_per_tensor_default_22" [label=None, style=solid]; +"190 dequantize_per_tensor_default_22" -> "192 conv2d_16" [label=None, style=solid]; +"191 layer4_0_conv2_weight_bias" -> "192 conv2d_16" [label="(512,)", style=solid]; +"192 conv2d_16" -> "193 quantize_per_tensor_default_23" [label="(1, 512, 7, 7)", style=solid]; +"193 quantize_per_tensor_default_23" -> "194 dequantize_per_tensor_default_23" [label=None, style=solid]; +"194 dequantize_per_tensor_default_23" -> "199 add__6" [label=None, style=solid]; +"195 layer4_0_downsample_0_weight_bias" -> "196 conv2d_17" [label="(512,)", style=solid]; +"196 conv2d_17" -> "197 quantize_per_tensor_default_24" [label="(1, 512, 7, 7)", style=solid]; +"197 quantize_per_tensor_default_24" -> "198 dequantize_per_tensor_default_24" [label=None, style=solid]; +"198 dequantize_per_tensor_default_24" -> "199 add__6" [label=None, style=solid]; +"199 add__6" -> "200 relu__14" [label="(1, 512, 7, 7)", style=solid]; +"200 relu__14" -> "201 quantize_per_tensor_default_25" [label="(1, 512, 7, 7)", style=solid]; +"201 quantize_per_tensor_default_25" -> "202 dequantize_per_tensor_default_25" [label=None, style=solid]; +"201 quantize_per_tensor_default_25" -> "203 dequantize_per_tensor_default_39" [label=None, style=solid]; +"202 dequantize_per_tensor_default_25" -> "213 add__7" [label=None, style=solid]; +"203 dequantize_per_tensor_default_39" -> "205 conv2d_18" [label=None, style=solid]; +"204 layer4_1_conv1_weight_bias" -> "205 conv2d_18" [label="(512,)", style=solid]; +"205 conv2d_18" -> "206 relu__15" [label="(1, 512, 7, 7)", style=solid]; +"206 relu__15" -> "207 quantize_per_tensor_default_26" [label="(1, 512, 7, 7)", style=solid]; +"207 quantize_per_tensor_default_26" -> "208 dequantize_per_tensor_default_26" [label=None, style=solid]; +"208 dequantize_per_tensor_default_26" -> "210 conv2d_19" [label=None, style=solid]; +"209 layer4_1_conv2_weight_bias" -> "210 conv2d_19" [label="(512,)", style=solid]; +"210 conv2d_19" -> "211 quantize_per_tensor_default_27" [label="(1, 512, 7, 7)", style=solid]; +"211 quantize_per_tensor_default_27" -> "212 dequantize_per_tensor_default_27" [label=None, style=solid]; +"212 dequantize_per_tensor_default_27" -> "213 add__7" [label=None, style=solid]; +"213 add__7" -> "214 relu__16" [label="(1, 512, 7, 7)", style=solid]; +"214 relu__16" -> "215 quantize_per_tensor_default_28" [label="(1, 512, 7, 7)", style=solid]; +"215 quantize_per_tensor_default_28" -> "216 dequantize_per_tensor_default_28" [label=None, style=solid]; +"216 dequantize_per_tensor_default_28" -> "217 adaptive_avg_pool2d" [label=None, style=solid]; +"217 adaptive_avg_pool2d" -> "218 quantize_per_tensor_default_29" [label="(1, 512, 1, 1)", style=solid]; +"218 quantize_per_tensor_default_29" -> "219 dequantize_per_tensor_default_29" [label=None, style=solid]; +"219 dequantize_per_tensor_default_29" -> "220 flatten" [label=None, style=solid]; +"220 flatten" -> "221 linear" [label="(1, 512)", style=solid]; +"221 linear" -> "222 output_1" [label="(1, 1000)", style=solid]; +} diff --git a/tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/swin_v2_s.dot b/tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/swin_v2_s.dot new file mode 100644 index 00000000000..a3e4f7384fa --- /dev/null +++ b/tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/swin_v2_s.dot @@ -0,0 +1,7100 @@ +strict digraph { +"0 _scale_0" [id=0, type=get_attr]; +"1 _zero_point_0" [id=1, type=get_attr]; +"2 _frozen_param0" [id=2, type=get_attr]; +"3 dequantize_per_channel_default" [id=3, type=dequantize_per_channel]; +"4 _scale_1" [id=4, type=get_attr]; +"5 _zero_point_1" [id=5, type=get_attr]; +"6 _frozen_param1" [id=6, type=get_attr]; +"7 dequantize_per_channel_default_1" [id=7, type=dequantize_per_channel]; +"8 features_1_0_attn_qkv_bias" [id=8, type=get_attr]; +"9 features_1_0_attn_proj_bias" [id=9, type=get_attr]; +"10 features_1_0_attn_logit_scale" [id=10, type=get_attr]; +"11 _scale_2" [id=11, type=get_attr]; +"12 _zero_point_2" [id=12, type=get_attr]; +"13 _frozen_param2" [id=13, type=get_attr]; +"14 dequantize_per_channel_default_2" [id=14, type=dequantize_per_channel]; +"15 _scale_3" [id=15, type=get_attr]; +"16 _zero_point_3" [id=16, type=get_attr]; +"17 _frozen_param3" [id=17, type=get_attr]; +"18 dequantize_per_channel_default_3" [id=18, type=dequantize_per_channel]; +"19 features_1_1_attn_qkv_bias" [id=19, type=get_attr]; +"20 features_1_1_attn_proj_bias" [id=20, type=get_attr]; +"21 features_1_1_attn_logit_scale" [id=21, type=get_attr]; +"22 _scale_4" [id=22, type=get_attr]; +"23 _zero_point_4" [id=23, type=get_attr]; +"24 _frozen_param4" [id=24, type=get_attr]; +"25 dequantize_per_channel_default_4" [id=25, type=dequantize_per_channel]; +"26 _scale_5" [id=26, type=get_attr]; +"27 _zero_point_5" [id=27, type=get_attr]; +"28 _frozen_param5" [id=28, type=get_attr]; +"29 dequantize_per_channel_default_5" [id=29, type=dequantize_per_channel]; +"30 features_3_0_attn_qkv_bias" [id=30, type=get_attr]; +"31 features_3_0_attn_proj_bias" [id=31, type=get_attr]; +"32 features_3_0_attn_logit_scale" [id=32, type=get_attr]; +"33 _scale_6" [id=33, type=get_attr]; +"34 _zero_point_6" [id=34, type=get_attr]; +"35 _frozen_param6" [id=35, type=get_attr]; +"36 dequantize_per_channel_default_6" [id=36, type=dequantize_per_channel]; +"37 _scale_7" [id=37, type=get_attr]; +"38 _zero_point_7" [id=38, type=get_attr]; +"39 _frozen_param7" [id=39, type=get_attr]; +"40 dequantize_per_channel_default_7" [id=40, type=dequantize_per_channel]; +"41 features_3_1_attn_qkv_bias" [id=41, type=get_attr]; +"42 features_3_1_attn_proj_bias" [id=42, type=get_attr]; +"43 features_3_1_attn_logit_scale" [id=43, type=get_attr]; +"44 _scale_8" [id=44, type=get_attr]; +"45 _zero_point_8" [id=45, type=get_attr]; +"46 _frozen_param8" [id=46, type=get_attr]; +"47 dequantize_per_channel_default_8" [id=47, type=dequantize_per_channel]; +"48 _scale_9" [id=48, type=get_attr]; +"49 _zero_point_9" [id=49, type=get_attr]; +"50 _frozen_param9" [id=50, type=get_attr]; +"51 dequantize_per_channel_default_9" [id=51, type=dequantize_per_channel]; +"52 features_5_0_attn_qkv_bias" [id=52, type=get_attr]; +"53 features_5_0_attn_proj_bias" [id=53, type=get_attr]; +"54 features_5_0_attn_logit_scale" [id=54, type=get_attr]; +"55 _scale_10" [id=55, type=get_attr]; +"56 _zero_point_10" [id=56, type=get_attr]; +"57 _frozen_param10" [id=57, type=get_attr]; +"58 dequantize_per_channel_default_10" [id=58, type=dequantize_per_channel]; +"59 _scale_11" [id=59, type=get_attr]; +"60 _zero_point_11" [id=60, type=get_attr]; +"61 _frozen_param11" [id=61, type=get_attr]; +"62 dequantize_per_channel_default_11" [id=62, type=dequantize_per_channel]; +"63 features_5_1_attn_qkv_bias" [id=63, type=get_attr]; +"64 features_5_1_attn_proj_bias" [id=64, type=get_attr]; +"65 features_5_1_attn_logit_scale" [id=65, type=get_attr]; +"66 _scale_12" [id=66, type=get_attr]; +"67 _zero_point_12" [id=67, type=get_attr]; +"68 _frozen_param12" [id=68, type=get_attr]; +"69 dequantize_per_channel_default_12" [id=69, type=dequantize_per_channel]; +"70 _scale_13" [id=70, type=get_attr]; +"71 _zero_point_13" [id=71, type=get_attr]; +"72 _frozen_param13" [id=72, type=get_attr]; +"73 dequantize_per_channel_default_13" [id=73, type=dequantize_per_channel]; +"74 features_5_2_attn_qkv_bias" [id=74, type=get_attr]; +"75 features_5_2_attn_proj_bias" [id=75, type=get_attr]; +"76 features_5_2_attn_logit_scale" [id=76, type=get_attr]; +"77 _scale_14" [id=77, type=get_attr]; +"78 _zero_point_14" [id=78, type=get_attr]; +"79 _frozen_param14" [id=79, type=get_attr]; +"80 dequantize_per_channel_default_14" [id=80, type=dequantize_per_channel]; +"81 _scale_15" [id=81, type=get_attr]; +"82 _zero_point_15" [id=82, type=get_attr]; +"83 _frozen_param15" [id=83, type=get_attr]; +"84 dequantize_per_channel_default_15" [id=84, type=dequantize_per_channel]; +"85 features_5_3_attn_qkv_bias" [id=85, type=get_attr]; +"86 features_5_3_attn_proj_bias" [id=86, type=get_attr]; +"87 features_5_3_attn_logit_scale" [id=87, type=get_attr]; +"88 _scale_16" [id=88, type=get_attr]; +"89 _zero_point_16" [id=89, type=get_attr]; +"90 _frozen_param16" [id=90, type=get_attr]; +"91 dequantize_per_channel_default_16" [id=91, type=dequantize_per_channel]; +"92 _scale_17" [id=92, type=get_attr]; +"93 _zero_point_17" [id=93, type=get_attr]; +"94 _frozen_param17" [id=94, type=get_attr]; +"95 dequantize_per_channel_default_17" [id=95, type=dequantize_per_channel]; +"96 features_5_4_attn_qkv_bias" [id=96, type=get_attr]; +"97 features_5_4_attn_proj_bias" [id=97, type=get_attr]; +"98 features_5_4_attn_logit_scale" [id=98, type=get_attr]; +"99 _scale_18" [id=99, type=get_attr]; +"100 _zero_point_18" [id=100, type=get_attr]; +"101 _frozen_param18" [id=101, type=get_attr]; +"102 dequantize_per_channel_default_18" [id=102, type=dequantize_per_channel]; +"103 _scale_19" [id=103, type=get_attr]; +"104 _zero_point_19" [id=104, type=get_attr]; +"105 _frozen_param19" [id=105, type=get_attr]; +"106 dequantize_per_channel_default_19" [id=106, type=dequantize_per_channel]; +"107 features_5_5_attn_qkv_bias" [id=107, type=get_attr]; +"108 features_5_5_attn_proj_bias" [id=108, type=get_attr]; +"109 features_5_5_attn_logit_scale" [id=109, type=get_attr]; +"110 _scale_20" [id=110, type=get_attr]; +"111 _zero_point_20" [id=111, type=get_attr]; +"112 _frozen_param20" [id=112, type=get_attr]; +"113 dequantize_per_channel_default_20" [id=113, type=dequantize_per_channel]; +"114 _scale_21" [id=114, type=get_attr]; +"115 _zero_point_21" [id=115, type=get_attr]; +"116 _frozen_param21" [id=116, type=get_attr]; +"117 dequantize_per_channel_default_21" [id=117, type=dequantize_per_channel]; +"118 features_5_6_attn_qkv_bias" [id=118, type=get_attr]; +"119 features_5_6_attn_proj_bias" [id=119, type=get_attr]; +"120 features_5_6_attn_logit_scale" [id=120, type=get_attr]; +"121 _scale_22" [id=121, type=get_attr]; +"122 _zero_point_22" [id=122, type=get_attr]; +"123 _frozen_param22" [id=123, type=get_attr]; +"124 dequantize_per_channel_default_22" [id=124, type=dequantize_per_channel]; +"125 _scale_23" [id=125, type=get_attr]; +"126 _zero_point_23" [id=126, type=get_attr]; +"127 _frozen_param23" [id=127, type=get_attr]; +"128 dequantize_per_channel_default_23" [id=128, type=dequantize_per_channel]; +"129 features_5_7_attn_qkv_bias" [id=129, type=get_attr]; +"130 features_5_7_attn_proj_bias" [id=130, type=get_attr]; +"131 features_5_7_attn_logit_scale" [id=131, type=get_attr]; +"132 _scale_24" [id=132, type=get_attr]; +"133 _zero_point_24" [id=133, type=get_attr]; +"134 _frozen_param24" [id=134, type=get_attr]; +"135 dequantize_per_channel_default_24" [id=135, type=dequantize_per_channel]; +"136 _scale_25" [id=136, type=get_attr]; +"137 _zero_point_25" [id=137, type=get_attr]; +"138 _frozen_param25" [id=138, type=get_attr]; +"139 dequantize_per_channel_default_25" [id=139, type=dequantize_per_channel]; +"140 features_5_8_attn_qkv_bias" [id=140, type=get_attr]; +"141 features_5_8_attn_proj_bias" [id=141, type=get_attr]; +"142 features_5_8_attn_logit_scale" [id=142, type=get_attr]; +"143 _scale_26" [id=143, type=get_attr]; +"144 _zero_point_26" [id=144, type=get_attr]; +"145 _frozen_param26" [id=145, type=get_attr]; +"146 dequantize_per_channel_default_26" [id=146, type=dequantize_per_channel]; +"147 _scale_27" [id=147, type=get_attr]; +"148 _zero_point_27" [id=148, type=get_attr]; +"149 _frozen_param27" [id=149, type=get_attr]; +"150 dequantize_per_channel_default_27" [id=150, type=dequantize_per_channel]; +"151 features_5_9_attn_qkv_bias" [id=151, type=get_attr]; +"152 features_5_9_attn_proj_bias" [id=152, type=get_attr]; +"153 features_5_9_attn_logit_scale" [id=153, type=get_attr]; +"154 _scale_28" [id=154, type=get_attr]; +"155 _zero_point_28" [id=155, type=get_attr]; +"156 _frozen_param28" [id=156, type=get_attr]; +"157 dequantize_per_channel_default_28" [id=157, type=dequantize_per_channel]; +"158 _scale_29" [id=158, type=get_attr]; +"159 _zero_point_29" [id=159, type=get_attr]; +"160 _frozen_param29" [id=160, type=get_attr]; +"161 dequantize_per_channel_default_29" [id=161, type=dequantize_per_channel]; +"162 features_5_10_attn_qkv_bias" [id=162, type=get_attr]; +"163 features_5_10_attn_proj_bias" [id=163, type=get_attr]; +"164 features_5_10_attn_logit_scale" [id=164, type=get_attr]; +"165 _scale_30" [id=165, type=get_attr]; +"166 _zero_point_30" [id=166, type=get_attr]; +"167 _frozen_param30" [id=167, type=get_attr]; +"168 dequantize_per_channel_default_30" [id=168, type=dequantize_per_channel]; +"169 _scale_31" [id=169, type=get_attr]; +"170 _zero_point_31" [id=170, type=get_attr]; +"171 _frozen_param31" [id=171, type=get_attr]; +"172 dequantize_per_channel_default_31" [id=172, type=dequantize_per_channel]; +"173 features_5_11_attn_qkv_bias" [id=173, type=get_attr]; +"174 features_5_11_attn_proj_bias" [id=174, type=get_attr]; +"175 features_5_11_attn_logit_scale" [id=175, type=get_attr]; +"176 _scale_32" [id=176, type=get_attr]; +"177 _zero_point_32" [id=177, type=get_attr]; +"178 _frozen_param32" [id=178, type=get_attr]; +"179 dequantize_per_channel_default_32" [id=179, type=dequantize_per_channel]; +"180 _scale_33" [id=180, type=get_attr]; +"181 _zero_point_33" [id=181, type=get_attr]; +"182 _frozen_param33" [id=182, type=get_attr]; +"183 dequantize_per_channel_default_33" [id=183, type=dequantize_per_channel]; +"184 features_5_12_attn_qkv_bias" [id=184, type=get_attr]; +"185 features_5_12_attn_proj_bias" [id=185, type=get_attr]; +"186 features_5_12_attn_logit_scale" [id=186, type=get_attr]; +"187 _scale_34" [id=187, type=get_attr]; +"188 _zero_point_34" [id=188, type=get_attr]; +"189 _frozen_param34" [id=189, type=get_attr]; +"190 dequantize_per_channel_default_34" [id=190, type=dequantize_per_channel]; +"191 _scale_35" [id=191, type=get_attr]; +"192 _zero_point_35" [id=192, type=get_attr]; +"193 _frozen_param35" [id=193, type=get_attr]; +"194 dequantize_per_channel_default_35" [id=194, type=dequantize_per_channel]; +"195 features_5_13_attn_qkv_bias" [id=195, type=get_attr]; +"196 features_5_13_attn_proj_bias" [id=196, type=get_attr]; +"197 features_5_13_attn_logit_scale" [id=197, type=get_attr]; +"198 _scale_36" [id=198, type=get_attr]; +"199 _zero_point_36" [id=199, type=get_attr]; +"200 _frozen_param36" [id=200, type=get_attr]; +"201 dequantize_per_channel_default_36" [id=201, type=dequantize_per_channel]; +"202 _scale_37" [id=202, type=get_attr]; +"203 _zero_point_37" [id=203, type=get_attr]; +"204 _frozen_param37" [id=204, type=get_attr]; +"205 dequantize_per_channel_default_37" [id=205, type=dequantize_per_channel]; +"206 features_5_14_attn_qkv_bias" [id=206, type=get_attr]; +"207 features_5_14_attn_proj_bias" [id=207, type=get_attr]; +"208 features_5_14_attn_logit_scale" [id=208, type=get_attr]; +"209 _scale_38" [id=209, type=get_attr]; +"210 _zero_point_38" [id=210, type=get_attr]; +"211 _frozen_param38" [id=211, type=get_attr]; +"212 dequantize_per_channel_default_38" [id=212, type=dequantize_per_channel]; +"213 _scale_39" [id=213, type=get_attr]; +"214 _zero_point_39" [id=214, type=get_attr]; +"215 _frozen_param39" [id=215, type=get_attr]; +"216 dequantize_per_channel_default_39" [id=216, type=dequantize_per_channel]; +"217 features_5_15_attn_qkv_bias" [id=217, type=get_attr]; +"218 features_5_15_attn_proj_bias" [id=218, type=get_attr]; +"219 features_5_15_attn_logit_scale" [id=219, type=get_attr]; +"220 _scale_40" [id=220, type=get_attr]; +"221 _zero_point_40" [id=221, type=get_attr]; +"222 _frozen_param40" [id=222, type=get_attr]; +"223 dequantize_per_channel_default_40" [id=223, type=dequantize_per_channel]; +"224 _scale_41" [id=224, type=get_attr]; +"225 _zero_point_41" [id=225, type=get_attr]; +"226 _frozen_param41" [id=226, type=get_attr]; +"227 dequantize_per_channel_default_41" [id=227, type=dequantize_per_channel]; +"228 features_5_16_attn_qkv_bias" [id=228, type=get_attr]; +"229 features_5_16_attn_proj_bias" [id=229, type=get_attr]; +"230 features_5_16_attn_logit_scale" [id=230, type=get_attr]; +"231 _scale_42" [id=231, type=get_attr]; +"232 _zero_point_42" [id=232, type=get_attr]; +"233 _frozen_param42" [id=233, type=get_attr]; +"234 dequantize_per_channel_default_42" [id=234, type=dequantize_per_channel]; +"235 _scale_43" [id=235, type=get_attr]; +"236 _zero_point_43" [id=236, type=get_attr]; +"237 _frozen_param43" [id=237, type=get_attr]; +"238 dequantize_per_channel_default_43" [id=238, type=dequantize_per_channel]; +"239 features_5_17_attn_qkv_bias" [id=239, type=get_attr]; +"240 features_5_17_attn_proj_bias" [id=240, type=get_attr]; +"241 features_5_17_attn_logit_scale" [id=241, type=get_attr]; +"242 _scale_44" [id=242, type=get_attr]; +"243 _zero_point_44" [id=243, type=get_attr]; +"244 _frozen_param44" [id=244, type=get_attr]; +"245 dequantize_per_channel_default_44" [id=245, type=dequantize_per_channel]; +"246 _scale_45" [id=246, type=get_attr]; +"247 _zero_point_45" [id=247, type=get_attr]; +"248 _frozen_param45" [id=248, type=get_attr]; +"249 dequantize_per_channel_default_45" [id=249, type=dequantize_per_channel]; +"250 features_7_0_attn_qkv_bias" [id=250, type=get_attr]; +"251 features_7_0_attn_proj_bias" [id=251, type=get_attr]; +"252 features_7_0_attn_logit_scale" [id=252, type=get_attr]; +"253 _scale_46" [id=253, type=get_attr]; +"254 _zero_point_46" [id=254, type=get_attr]; +"255 _frozen_param46" [id=255, type=get_attr]; +"256 dequantize_per_channel_default_46" [id=256, type=dequantize_per_channel]; +"257 _scale_47" [id=257, type=get_attr]; +"258 _zero_point_47" [id=258, type=get_attr]; +"259 _frozen_param47" [id=259, type=get_attr]; +"260 dequantize_per_channel_default_47" [id=260, type=dequantize_per_channel]; +"261 features_7_1_attn_qkv_bias" [id=261, type=get_attr]; +"262 features_7_1_attn_proj_bias" [id=262, type=get_attr]; +"263 features_7_1_attn_logit_scale" [id=263, type=get_attr]; +"264 _scale_48" [id=264, type=get_attr]; +"265 _zero_point_48" [id=265, type=get_attr]; +"266 _frozen_param48" [id=266, type=get_attr]; +"267 dequantize_per_channel_default_48" [id=267, type=dequantize_per_channel]; +"268 features_0_0_bias" [id=268, type=get_attr]; +"269 features_0_2_weight" [id=269, type=get_attr]; +"270 features_0_2_bias" [id=270, type=get_attr]; +"271 _scale_49" [id=271, type=get_attr]; +"272 _zero_point_49" [id=272, type=get_attr]; +"273 _frozen_param49" [id=273, type=get_attr]; +"274 dequantize_per_channel_default_49" [id=274, type=dequantize_per_channel]; +"275 features_1_0_attn_cpb_mlp_0_bias" [id=275, type=get_attr]; +"276 _scale_50" [id=276, type=get_attr]; +"277 _zero_point_50" [id=277, type=get_attr]; +"278 _frozen_param50" [id=278, type=get_attr]; +"279 dequantize_per_channel_default_50" [id=279, type=dequantize_per_channel]; +"280 features_1_0_norm1_weight" [id=280, type=get_attr]; +"281 features_1_0_norm1_bias" [id=281, type=get_attr]; +"282 _scale_51" [id=282, type=get_attr]; +"283 _zero_point_51" [id=283, type=get_attr]; +"284 _frozen_param51" [id=284, type=get_attr]; +"285 dequantize_per_channel_default_51" [id=285, type=dequantize_per_channel]; +"286 features_1_0_mlp_0_bias" [id=286, type=get_attr]; +"287 _scale_52" [id=287, type=get_attr]; +"288 _zero_point_52" [id=288, type=get_attr]; +"289 _frozen_param52" [id=289, type=get_attr]; +"290 dequantize_per_channel_default_52" [id=290, type=dequantize_per_channel]; +"291 features_1_0_mlp_3_bias" [id=291, type=get_attr]; +"292 features_1_0_norm2_weight" [id=292, type=get_attr]; +"293 features_1_0_norm2_bias" [id=293, type=get_attr]; +"294 _scale_53" [id=294, type=get_attr]; +"295 _zero_point_53" [id=295, type=get_attr]; +"296 _frozen_param53" [id=296, type=get_attr]; +"297 dequantize_per_channel_default_53" [id=297, type=dequantize_per_channel]; +"298 features_1_1_attn_cpb_mlp_0_bias" [id=298, type=get_attr]; +"299 _scale_54" [id=299, type=get_attr]; +"300 _zero_point_54" [id=300, type=get_attr]; +"301 _frozen_param54" [id=301, type=get_attr]; +"302 dequantize_per_channel_default_54" [id=302, type=dequantize_per_channel]; +"303 features_1_1_norm1_weight" [id=303, type=get_attr]; +"304 features_1_1_norm1_bias" [id=304, type=get_attr]; +"305 _scale_55" [id=305, type=get_attr]; +"306 _zero_point_55" [id=306, type=get_attr]; +"307 _frozen_param55" [id=307, type=get_attr]; +"308 dequantize_per_channel_default_55" [id=308, type=dequantize_per_channel]; +"309 features_1_1_mlp_0_bias" [id=309, type=get_attr]; +"310 _scale_56" [id=310, type=get_attr]; +"311 _zero_point_56" [id=311, type=get_attr]; +"312 _frozen_param56" [id=312, type=get_attr]; +"313 dequantize_per_channel_default_56" [id=313, type=dequantize_per_channel]; +"314 features_1_1_mlp_3_bias" [id=314, type=get_attr]; +"315 features_1_1_norm2_weight" [id=315, type=get_attr]; +"316 features_1_1_norm2_bias" [id=316, type=get_attr]; +"317 _scale_57" [id=317, type=get_attr]; +"318 _zero_point_57" [id=318, type=get_attr]; +"319 _frozen_param57" [id=319, type=get_attr]; +"320 dequantize_per_channel_default_57" [id=320, type=dequantize_per_channel]; +"321 features_2_norm_weight" [id=321, type=get_attr]; +"322 features_2_norm_bias" [id=322, type=get_attr]; +"323 _scale_58" [id=323, type=get_attr]; +"324 _zero_point_58" [id=324, type=get_attr]; +"325 _frozen_param58" [id=325, type=get_attr]; +"326 dequantize_per_channel_default_58" [id=326, type=dequantize_per_channel]; +"327 features_3_0_attn_cpb_mlp_0_bias" [id=327, type=get_attr]; +"328 _scale_59" [id=328, type=get_attr]; +"329 _zero_point_59" [id=329, type=get_attr]; +"330 _frozen_param59" [id=330, type=get_attr]; +"331 dequantize_per_channel_default_59" [id=331, type=dequantize_per_channel]; +"332 features_3_0_norm1_weight" [id=332, type=get_attr]; +"333 features_3_0_norm1_bias" [id=333, type=get_attr]; +"334 _scale_60" [id=334, type=get_attr]; +"335 _zero_point_60" [id=335, type=get_attr]; +"336 _frozen_param60" [id=336, type=get_attr]; +"337 dequantize_per_channel_default_60" [id=337, type=dequantize_per_channel]; +"338 features_3_0_mlp_0_bias" [id=338, type=get_attr]; +"339 _scale_61" [id=339, type=get_attr]; +"340 _zero_point_61" [id=340, type=get_attr]; +"341 _frozen_param61" [id=341, type=get_attr]; +"342 dequantize_per_channel_default_61" [id=342, type=dequantize_per_channel]; +"343 features_3_0_mlp_3_bias" [id=343, type=get_attr]; +"344 features_3_0_norm2_weight" [id=344, type=get_attr]; +"345 features_3_0_norm2_bias" [id=345, type=get_attr]; +"346 _scale_62" [id=346, type=get_attr]; +"347 _zero_point_62" [id=347, type=get_attr]; +"348 _frozen_param62" [id=348, type=get_attr]; +"349 dequantize_per_channel_default_62" [id=349, type=dequantize_per_channel]; +"350 features_3_1_attn_cpb_mlp_0_bias" [id=350, type=get_attr]; +"351 _scale_63" [id=351, type=get_attr]; +"352 _zero_point_63" [id=352, type=get_attr]; +"353 _frozen_param63" [id=353, type=get_attr]; +"354 dequantize_per_channel_default_63" [id=354, type=dequantize_per_channel]; +"355 features_3_1_norm1_weight" [id=355, type=get_attr]; +"356 features_3_1_norm1_bias" [id=356, type=get_attr]; +"357 _scale_64" [id=357, type=get_attr]; +"358 _zero_point_64" [id=358, type=get_attr]; +"359 _frozen_param64" [id=359, type=get_attr]; +"360 dequantize_per_channel_default_64" [id=360, type=dequantize_per_channel]; +"361 features_3_1_mlp_0_bias" [id=361, type=get_attr]; +"362 _scale_65" [id=362, type=get_attr]; +"363 _zero_point_65" [id=363, type=get_attr]; +"364 _frozen_param65" [id=364, type=get_attr]; +"365 dequantize_per_channel_default_65" [id=365, type=dequantize_per_channel]; +"366 features_3_1_mlp_3_bias" [id=366, type=get_attr]; +"367 features_3_1_norm2_weight" [id=367, type=get_attr]; +"368 features_3_1_norm2_bias" [id=368, type=get_attr]; +"369 _scale_66" [id=369, type=get_attr]; +"370 _zero_point_66" [id=370, type=get_attr]; +"371 _frozen_param66" [id=371, type=get_attr]; +"372 dequantize_per_channel_default_66" [id=372, type=dequantize_per_channel]; +"373 features_4_norm_weight" [id=373, type=get_attr]; +"374 features_4_norm_bias" [id=374, type=get_attr]; +"375 _scale_67" [id=375, type=get_attr]; +"376 _zero_point_67" [id=376, type=get_attr]; +"377 _frozen_param67" [id=377, type=get_attr]; +"378 dequantize_per_channel_default_67" [id=378, type=dequantize_per_channel]; +"379 features_5_0_attn_cpb_mlp_0_bias" [id=379, type=get_attr]; +"380 _scale_68" [id=380, type=get_attr]; +"381 _zero_point_68" [id=381, type=get_attr]; +"382 _frozen_param68" [id=382, type=get_attr]; +"383 dequantize_per_channel_default_68" [id=383, type=dequantize_per_channel]; +"384 features_5_0_norm1_weight" [id=384, type=get_attr]; +"385 features_5_0_norm1_bias" [id=385, type=get_attr]; +"386 _scale_69" [id=386, type=get_attr]; +"387 _zero_point_69" [id=387, type=get_attr]; +"388 _frozen_param69" [id=388, type=get_attr]; +"389 dequantize_per_channel_default_69" [id=389, type=dequantize_per_channel]; +"390 features_5_0_mlp_0_bias" [id=390, type=get_attr]; +"391 _scale_70" [id=391, type=get_attr]; +"392 _zero_point_70" [id=392, type=get_attr]; +"393 _frozen_param70" [id=393, type=get_attr]; +"394 dequantize_per_channel_default_70" [id=394, type=dequantize_per_channel]; +"395 features_5_0_mlp_3_bias" [id=395, type=get_attr]; +"396 features_5_0_norm2_weight" [id=396, type=get_attr]; +"397 features_5_0_norm2_bias" [id=397, type=get_attr]; +"398 _scale_71" [id=398, type=get_attr]; +"399 _zero_point_71" [id=399, type=get_attr]; +"400 _frozen_param71" [id=400, type=get_attr]; +"401 dequantize_per_channel_default_71" [id=401, type=dequantize_per_channel]; +"402 features_5_1_attn_cpb_mlp_0_bias" [id=402, type=get_attr]; +"403 _scale_72" [id=403, type=get_attr]; +"404 _zero_point_72" [id=404, type=get_attr]; +"405 _frozen_param72" [id=405, type=get_attr]; +"406 dequantize_per_channel_default_72" [id=406, type=dequantize_per_channel]; +"407 features_5_1_norm1_weight" [id=407, type=get_attr]; +"408 features_5_1_norm1_bias" [id=408, type=get_attr]; +"409 _scale_73" [id=409, type=get_attr]; +"410 _zero_point_73" [id=410, type=get_attr]; +"411 _frozen_param73" [id=411, type=get_attr]; +"412 dequantize_per_channel_default_73" [id=412, type=dequantize_per_channel]; +"413 features_5_1_mlp_0_bias" [id=413, type=get_attr]; +"414 _scale_74" [id=414, type=get_attr]; +"415 _zero_point_74" [id=415, type=get_attr]; +"416 _frozen_param74" [id=416, type=get_attr]; +"417 dequantize_per_channel_default_74" [id=417, type=dequantize_per_channel]; +"418 features_5_1_mlp_3_bias" [id=418, type=get_attr]; +"419 features_5_1_norm2_weight" [id=419, type=get_attr]; +"420 features_5_1_norm2_bias" [id=420, type=get_attr]; +"421 _scale_75" [id=421, type=get_attr]; +"422 _zero_point_75" [id=422, type=get_attr]; +"423 _frozen_param75" [id=423, type=get_attr]; +"424 dequantize_per_channel_default_75" [id=424, type=dequantize_per_channel]; +"425 features_5_2_attn_cpb_mlp_0_bias" [id=425, type=get_attr]; +"426 _scale_76" [id=426, type=get_attr]; +"427 _zero_point_76" [id=427, type=get_attr]; +"428 _frozen_param76" [id=428, type=get_attr]; +"429 dequantize_per_channel_default_76" [id=429, type=dequantize_per_channel]; +"430 features_5_2_norm1_weight" [id=430, type=get_attr]; +"431 features_5_2_norm1_bias" [id=431, type=get_attr]; +"432 _scale_77" [id=432, type=get_attr]; +"433 _zero_point_77" [id=433, type=get_attr]; +"434 _frozen_param77" [id=434, type=get_attr]; +"435 dequantize_per_channel_default_77" [id=435, type=dequantize_per_channel]; +"436 features_5_2_mlp_0_bias" [id=436, type=get_attr]; +"437 _scale_78" [id=437, type=get_attr]; +"438 _zero_point_78" [id=438, type=get_attr]; +"439 _frozen_param78" [id=439, type=get_attr]; +"440 dequantize_per_channel_default_78" [id=440, type=dequantize_per_channel]; +"441 features_5_2_mlp_3_bias" [id=441, type=get_attr]; +"442 features_5_2_norm2_weight" [id=442, type=get_attr]; +"443 features_5_2_norm2_bias" [id=443, type=get_attr]; +"444 _scale_79" [id=444, type=get_attr]; +"445 _zero_point_79" [id=445, type=get_attr]; +"446 _frozen_param79" [id=446, type=get_attr]; +"447 dequantize_per_channel_default_79" [id=447, type=dequantize_per_channel]; +"448 features_5_3_attn_cpb_mlp_0_bias" [id=448, type=get_attr]; +"449 _scale_80" [id=449, type=get_attr]; +"450 _zero_point_80" [id=450, type=get_attr]; +"451 _frozen_param80" [id=451, type=get_attr]; +"452 dequantize_per_channel_default_80" [id=452, type=dequantize_per_channel]; +"453 features_5_3_norm1_weight" [id=453, type=get_attr]; +"454 features_5_3_norm1_bias" [id=454, type=get_attr]; +"455 _scale_81" [id=455, type=get_attr]; +"456 _zero_point_81" [id=456, type=get_attr]; +"457 _frozen_param81" [id=457, type=get_attr]; +"458 dequantize_per_channel_default_81" [id=458, type=dequantize_per_channel]; +"459 features_5_3_mlp_0_bias" [id=459, type=get_attr]; +"460 _scale_82" [id=460, type=get_attr]; +"461 _zero_point_82" [id=461, type=get_attr]; +"462 _frozen_param82" [id=462, type=get_attr]; +"463 dequantize_per_channel_default_82" [id=463, type=dequantize_per_channel]; +"464 features_5_3_mlp_3_bias" [id=464, type=get_attr]; +"465 features_5_3_norm2_weight" [id=465, type=get_attr]; +"466 features_5_3_norm2_bias" [id=466, type=get_attr]; +"467 _scale_83" [id=467, type=get_attr]; +"468 _zero_point_83" [id=468, type=get_attr]; +"469 _frozen_param83" [id=469, type=get_attr]; +"470 dequantize_per_channel_default_83" [id=470, type=dequantize_per_channel]; +"471 features_5_4_attn_cpb_mlp_0_bias" [id=471, type=get_attr]; +"472 _scale_84" [id=472, type=get_attr]; +"473 _zero_point_84" [id=473, type=get_attr]; +"474 _frozen_param84" [id=474, type=get_attr]; +"475 dequantize_per_channel_default_84" [id=475, type=dequantize_per_channel]; +"476 features_5_4_norm1_weight" [id=476, type=get_attr]; +"477 features_5_4_norm1_bias" [id=477, type=get_attr]; +"478 _scale_85" [id=478, type=get_attr]; +"479 _zero_point_85" [id=479, type=get_attr]; +"480 _frozen_param85" [id=480, type=get_attr]; +"481 dequantize_per_channel_default_85" [id=481, type=dequantize_per_channel]; +"482 features_5_4_mlp_0_bias" [id=482, type=get_attr]; +"483 _scale_86" [id=483, type=get_attr]; +"484 _zero_point_86" [id=484, type=get_attr]; +"485 _frozen_param86" [id=485, type=get_attr]; +"486 dequantize_per_channel_default_86" [id=486, type=dequantize_per_channel]; +"487 features_5_4_mlp_3_bias" [id=487, type=get_attr]; +"488 features_5_4_norm2_weight" [id=488, type=get_attr]; +"489 features_5_4_norm2_bias" [id=489, type=get_attr]; +"490 _scale_87" [id=490, type=get_attr]; +"491 _zero_point_87" [id=491, type=get_attr]; +"492 _frozen_param87" [id=492, type=get_attr]; +"493 dequantize_per_channel_default_87" [id=493, type=dequantize_per_channel]; +"494 features_5_5_attn_cpb_mlp_0_bias" [id=494, type=get_attr]; +"495 _scale_88" [id=495, type=get_attr]; +"496 _zero_point_88" [id=496, type=get_attr]; +"497 _frozen_param88" [id=497, type=get_attr]; +"498 dequantize_per_channel_default_88" [id=498, type=dequantize_per_channel]; +"499 features_5_5_norm1_weight" [id=499, type=get_attr]; +"500 features_5_5_norm1_bias" [id=500, type=get_attr]; +"501 _scale_89" [id=501, type=get_attr]; +"502 _zero_point_89" [id=502, type=get_attr]; +"503 _frozen_param89" [id=503, type=get_attr]; +"504 dequantize_per_channel_default_89" [id=504, type=dequantize_per_channel]; +"505 features_5_5_mlp_0_bias" [id=505, type=get_attr]; +"506 _scale_90" [id=506, type=get_attr]; +"507 _zero_point_90" [id=507, type=get_attr]; +"508 _frozen_param90" [id=508, type=get_attr]; +"509 dequantize_per_channel_default_90" [id=509, type=dequantize_per_channel]; +"510 features_5_5_mlp_3_bias" [id=510, type=get_attr]; +"511 features_5_5_norm2_weight" [id=511, type=get_attr]; +"512 features_5_5_norm2_bias" [id=512, type=get_attr]; +"513 _scale_91" [id=513, type=get_attr]; +"514 _zero_point_91" [id=514, type=get_attr]; +"515 _frozen_param91" [id=515, type=get_attr]; +"516 dequantize_per_channel_default_91" [id=516, type=dequantize_per_channel]; +"517 features_5_6_attn_cpb_mlp_0_bias" [id=517, type=get_attr]; +"518 _scale_92" [id=518, type=get_attr]; +"519 _zero_point_92" [id=519, type=get_attr]; +"520 _frozen_param92" [id=520, type=get_attr]; +"521 dequantize_per_channel_default_92" [id=521, type=dequantize_per_channel]; +"522 features_5_6_norm1_weight" [id=522, type=get_attr]; +"523 features_5_6_norm1_bias" [id=523, type=get_attr]; +"524 _scale_93" [id=524, type=get_attr]; +"525 _zero_point_93" [id=525, type=get_attr]; +"526 _frozen_param93" [id=526, type=get_attr]; +"527 dequantize_per_channel_default_93" [id=527, type=dequantize_per_channel]; +"528 features_5_6_mlp_0_bias" [id=528, type=get_attr]; +"529 _scale_94" [id=529, type=get_attr]; +"530 _zero_point_94" [id=530, type=get_attr]; +"531 _frozen_param94" [id=531, type=get_attr]; +"532 dequantize_per_channel_default_94" [id=532, type=dequantize_per_channel]; +"533 features_5_6_mlp_3_bias" [id=533, type=get_attr]; +"534 features_5_6_norm2_weight" [id=534, type=get_attr]; +"535 features_5_6_norm2_bias" [id=535, type=get_attr]; +"536 _scale_95" [id=536, type=get_attr]; +"537 _zero_point_95" [id=537, type=get_attr]; +"538 _frozen_param95" [id=538, type=get_attr]; +"539 dequantize_per_channel_default_95" [id=539, type=dequantize_per_channel]; +"540 features_5_7_attn_cpb_mlp_0_bias" [id=540, type=get_attr]; +"541 _scale_96" [id=541, type=get_attr]; +"542 _zero_point_96" [id=542, type=get_attr]; +"543 _frozen_param96" [id=543, type=get_attr]; +"544 dequantize_per_channel_default_96" [id=544, type=dequantize_per_channel]; +"545 features_5_7_norm1_weight" [id=545, type=get_attr]; +"546 features_5_7_norm1_bias" [id=546, type=get_attr]; +"547 _scale_97" [id=547, type=get_attr]; +"548 _zero_point_97" [id=548, type=get_attr]; +"549 _frozen_param97" [id=549, type=get_attr]; +"550 dequantize_per_channel_default_97" [id=550, type=dequantize_per_channel]; +"551 features_5_7_mlp_0_bias" [id=551, type=get_attr]; +"552 _scale_98" [id=552, type=get_attr]; +"553 _zero_point_98" [id=553, type=get_attr]; +"554 _frozen_param98" [id=554, type=get_attr]; +"555 dequantize_per_channel_default_98" [id=555, type=dequantize_per_channel]; +"556 features_5_7_mlp_3_bias" [id=556, type=get_attr]; +"557 features_5_7_norm2_weight" [id=557, type=get_attr]; +"558 features_5_7_norm2_bias" [id=558, type=get_attr]; +"559 _scale_99" [id=559, type=get_attr]; +"560 _zero_point_99" [id=560, type=get_attr]; +"561 _frozen_param99" [id=561, type=get_attr]; +"562 dequantize_per_channel_default_99" [id=562, type=dequantize_per_channel]; +"563 features_5_8_attn_cpb_mlp_0_bias" [id=563, type=get_attr]; +"564 _scale_100" [id=564, type=get_attr]; +"565 _zero_point_100" [id=565, type=get_attr]; +"566 _frozen_param100" [id=566, type=get_attr]; +"567 dequantize_per_channel_default_100" [id=567, type=dequantize_per_channel]; +"568 features_5_8_norm1_weight" [id=568, type=get_attr]; +"569 features_5_8_norm1_bias" [id=569, type=get_attr]; +"570 _scale_101" [id=570, type=get_attr]; +"571 _zero_point_101" [id=571, type=get_attr]; +"572 _frozen_param101" [id=572, type=get_attr]; +"573 dequantize_per_channel_default_101" [id=573, type=dequantize_per_channel]; +"574 features_5_8_mlp_0_bias" [id=574, type=get_attr]; +"575 _scale_102" [id=575, type=get_attr]; +"576 _zero_point_102" [id=576, type=get_attr]; +"577 _frozen_param102" [id=577, type=get_attr]; +"578 dequantize_per_channel_default_102" [id=578, type=dequantize_per_channel]; +"579 features_5_8_mlp_3_bias" [id=579, type=get_attr]; +"580 features_5_8_norm2_weight" [id=580, type=get_attr]; +"581 features_5_8_norm2_bias" [id=581, type=get_attr]; +"582 _scale_103" [id=582, type=get_attr]; +"583 _zero_point_103" [id=583, type=get_attr]; +"584 _frozen_param103" [id=584, type=get_attr]; +"585 dequantize_per_channel_default_103" [id=585, type=dequantize_per_channel]; +"586 features_5_9_attn_cpb_mlp_0_bias" [id=586, type=get_attr]; +"587 _scale_104" [id=587, type=get_attr]; +"588 _zero_point_104" [id=588, type=get_attr]; +"589 _frozen_param104" [id=589, type=get_attr]; +"590 dequantize_per_channel_default_104" [id=590, type=dequantize_per_channel]; +"591 features_5_9_norm1_weight" [id=591, type=get_attr]; +"592 features_5_9_norm1_bias" [id=592, type=get_attr]; +"593 _scale_105" [id=593, type=get_attr]; +"594 _zero_point_105" [id=594, type=get_attr]; +"595 _frozen_param105" [id=595, type=get_attr]; +"596 dequantize_per_channel_default_105" [id=596, type=dequantize_per_channel]; +"597 features_5_9_mlp_0_bias" [id=597, type=get_attr]; +"598 _scale_106" [id=598, type=get_attr]; +"599 _zero_point_106" [id=599, type=get_attr]; +"600 _frozen_param106" [id=600, type=get_attr]; +"601 dequantize_per_channel_default_106" [id=601, type=dequantize_per_channel]; +"602 features_5_9_mlp_3_bias" [id=602, type=get_attr]; +"603 features_5_9_norm2_weight" [id=603, type=get_attr]; +"604 features_5_9_norm2_bias" [id=604, type=get_attr]; +"605 _scale_107" [id=605, type=get_attr]; +"606 _zero_point_107" [id=606, type=get_attr]; +"607 _frozen_param107" [id=607, type=get_attr]; +"608 dequantize_per_channel_default_107" [id=608, type=dequantize_per_channel]; +"609 features_5_10_attn_cpb_mlp_0_bias" [id=609, type=get_attr]; +"610 _scale_108" [id=610, type=get_attr]; +"611 _zero_point_108" [id=611, type=get_attr]; +"612 _frozen_param108" [id=612, type=get_attr]; +"613 dequantize_per_channel_default_108" [id=613, type=dequantize_per_channel]; +"614 features_5_10_norm1_weight" [id=614, type=get_attr]; +"615 features_5_10_norm1_bias" [id=615, type=get_attr]; +"616 _scale_109" [id=616, type=get_attr]; +"617 _zero_point_109" [id=617, type=get_attr]; +"618 _frozen_param109" [id=618, type=get_attr]; +"619 dequantize_per_channel_default_109" [id=619, type=dequantize_per_channel]; +"620 features_5_10_mlp_0_bias" [id=620, type=get_attr]; +"621 _scale_110" [id=621, type=get_attr]; +"622 _zero_point_110" [id=622, type=get_attr]; +"623 _frozen_param110" [id=623, type=get_attr]; +"624 dequantize_per_channel_default_110" [id=624, type=dequantize_per_channel]; +"625 features_5_10_mlp_3_bias" [id=625, type=get_attr]; +"626 features_5_10_norm2_weight" [id=626, type=get_attr]; +"627 features_5_10_norm2_bias" [id=627, type=get_attr]; +"628 _scale_111" [id=628, type=get_attr]; +"629 _zero_point_111" [id=629, type=get_attr]; +"630 _frozen_param111" [id=630, type=get_attr]; +"631 dequantize_per_channel_default_111" [id=631, type=dequantize_per_channel]; +"632 features_5_11_attn_cpb_mlp_0_bias" [id=632, type=get_attr]; +"633 _scale_112" [id=633, type=get_attr]; +"634 _zero_point_112" [id=634, type=get_attr]; +"635 _frozen_param112" [id=635, type=get_attr]; +"636 dequantize_per_channel_default_112" [id=636, type=dequantize_per_channel]; +"637 features_5_11_norm1_weight" [id=637, type=get_attr]; +"638 features_5_11_norm1_bias" [id=638, type=get_attr]; +"639 _scale_113" [id=639, type=get_attr]; +"640 _zero_point_113" [id=640, type=get_attr]; +"641 _frozen_param113" [id=641, type=get_attr]; +"642 dequantize_per_channel_default_113" [id=642, type=dequantize_per_channel]; +"643 features_5_11_mlp_0_bias" [id=643, type=get_attr]; +"644 _scale_114" [id=644, type=get_attr]; +"645 _zero_point_114" [id=645, type=get_attr]; +"646 _frozen_param114" [id=646, type=get_attr]; +"647 dequantize_per_channel_default_114" [id=647, type=dequantize_per_channel]; +"648 features_5_11_mlp_3_bias" [id=648, type=get_attr]; +"649 features_5_11_norm2_weight" [id=649, type=get_attr]; +"650 features_5_11_norm2_bias" [id=650, type=get_attr]; +"651 _scale_115" [id=651, type=get_attr]; +"652 _zero_point_115" [id=652, type=get_attr]; +"653 _frozen_param115" [id=653, type=get_attr]; +"654 dequantize_per_channel_default_115" [id=654, type=dequantize_per_channel]; +"655 features_5_12_attn_cpb_mlp_0_bias" [id=655, type=get_attr]; +"656 _scale_116" [id=656, type=get_attr]; +"657 _zero_point_116" [id=657, type=get_attr]; +"658 _frozen_param116" [id=658, type=get_attr]; +"659 dequantize_per_channel_default_116" [id=659, type=dequantize_per_channel]; +"660 features_5_12_norm1_weight" [id=660, type=get_attr]; +"661 features_5_12_norm1_bias" [id=661, type=get_attr]; +"662 _scale_117" [id=662, type=get_attr]; +"663 _zero_point_117" [id=663, type=get_attr]; +"664 _frozen_param117" [id=664, type=get_attr]; +"665 dequantize_per_channel_default_117" [id=665, type=dequantize_per_channel]; +"666 features_5_12_mlp_0_bias" [id=666, type=get_attr]; +"667 _scale_118" [id=667, type=get_attr]; +"668 _zero_point_118" [id=668, type=get_attr]; +"669 _frozen_param118" [id=669, type=get_attr]; +"670 dequantize_per_channel_default_118" [id=670, type=dequantize_per_channel]; +"671 features_5_12_mlp_3_bias" [id=671, type=get_attr]; +"672 features_5_12_norm2_weight" [id=672, type=get_attr]; +"673 features_5_12_norm2_bias" [id=673, type=get_attr]; +"674 _scale_119" [id=674, type=get_attr]; +"675 _zero_point_119" [id=675, type=get_attr]; +"676 _frozen_param119" [id=676, type=get_attr]; +"677 dequantize_per_channel_default_119" [id=677, type=dequantize_per_channel]; +"678 features_5_13_attn_cpb_mlp_0_bias" [id=678, type=get_attr]; +"679 _scale_120" [id=679, type=get_attr]; +"680 _zero_point_120" [id=680, type=get_attr]; +"681 _frozen_param120" [id=681, type=get_attr]; +"682 dequantize_per_channel_default_120" [id=682, type=dequantize_per_channel]; +"683 features_5_13_norm1_weight" [id=683, type=get_attr]; +"684 features_5_13_norm1_bias" [id=684, type=get_attr]; +"685 _scale_121" [id=685, type=get_attr]; +"686 _zero_point_121" [id=686, type=get_attr]; +"687 _frozen_param121" [id=687, type=get_attr]; +"688 dequantize_per_channel_default_121" [id=688, type=dequantize_per_channel]; +"689 features_5_13_mlp_0_bias" [id=689, type=get_attr]; +"690 _scale_122" [id=690, type=get_attr]; +"691 _zero_point_122" [id=691, type=get_attr]; +"692 _frozen_param122" [id=692, type=get_attr]; +"693 dequantize_per_channel_default_122" [id=693, type=dequantize_per_channel]; +"694 features_5_13_mlp_3_bias" [id=694, type=get_attr]; +"695 features_5_13_norm2_weight" [id=695, type=get_attr]; +"696 features_5_13_norm2_bias" [id=696, type=get_attr]; +"697 _scale_123" [id=697, type=get_attr]; +"698 _zero_point_123" [id=698, type=get_attr]; +"699 _frozen_param123" [id=699, type=get_attr]; +"700 dequantize_per_channel_default_123" [id=700, type=dequantize_per_channel]; +"701 features_5_14_attn_cpb_mlp_0_bias" [id=701, type=get_attr]; +"702 _scale_124" [id=702, type=get_attr]; +"703 _zero_point_124" [id=703, type=get_attr]; +"704 _frozen_param124" [id=704, type=get_attr]; +"705 dequantize_per_channel_default_124" [id=705, type=dequantize_per_channel]; +"706 features_5_14_norm1_weight" [id=706, type=get_attr]; +"707 features_5_14_norm1_bias" [id=707, type=get_attr]; +"708 _scale_125" [id=708, type=get_attr]; +"709 _zero_point_125" [id=709, type=get_attr]; +"710 _frozen_param125" [id=710, type=get_attr]; +"711 dequantize_per_channel_default_125" [id=711, type=dequantize_per_channel]; +"712 features_5_14_mlp_0_bias" [id=712, type=get_attr]; +"713 _scale_126" [id=713, type=get_attr]; +"714 _zero_point_126" [id=714, type=get_attr]; +"715 _frozen_param126" [id=715, type=get_attr]; +"716 dequantize_per_channel_default_126" [id=716, type=dequantize_per_channel]; +"717 features_5_14_mlp_3_bias" [id=717, type=get_attr]; +"718 features_5_14_norm2_weight" [id=718, type=get_attr]; +"719 features_5_14_norm2_bias" [id=719, type=get_attr]; +"720 _scale_127" [id=720, type=get_attr]; +"721 _zero_point_127" [id=721, type=get_attr]; +"722 _frozen_param127" [id=722, type=get_attr]; +"723 dequantize_per_channel_default_127" [id=723, type=dequantize_per_channel]; +"724 features_5_15_attn_cpb_mlp_0_bias" [id=724, type=get_attr]; +"725 _scale_128" [id=725, type=get_attr]; +"726 _zero_point_128" [id=726, type=get_attr]; +"727 _frozen_param128" [id=727, type=get_attr]; +"728 dequantize_per_channel_default_128" [id=728, type=dequantize_per_channel]; +"729 features_5_15_norm1_weight" [id=729, type=get_attr]; +"730 features_5_15_norm1_bias" [id=730, type=get_attr]; +"731 _scale_129" [id=731, type=get_attr]; +"732 _zero_point_129" [id=732, type=get_attr]; +"733 _frozen_param129" [id=733, type=get_attr]; +"734 dequantize_per_channel_default_129" [id=734, type=dequantize_per_channel]; +"735 features_5_15_mlp_0_bias" [id=735, type=get_attr]; +"736 _scale_130" [id=736, type=get_attr]; +"737 _zero_point_130" [id=737, type=get_attr]; +"738 _frozen_param130" [id=738, type=get_attr]; +"739 dequantize_per_channel_default_130" [id=739, type=dequantize_per_channel]; +"740 features_5_15_mlp_3_bias" [id=740, type=get_attr]; +"741 features_5_15_norm2_weight" [id=741, type=get_attr]; +"742 features_5_15_norm2_bias" [id=742, type=get_attr]; +"743 _scale_131" [id=743, type=get_attr]; +"744 _zero_point_131" [id=744, type=get_attr]; +"745 _frozen_param131" [id=745, type=get_attr]; +"746 dequantize_per_channel_default_131" [id=746, type=dequantize_per_channel]; +"747 features_5_16_attn_cpb_mlp_0_bias" [id=747, type=get_attr]; +"748 _scale_132" [id=748, type=get_attr]; +"749 _zero_point_132" [id=749, type=get_attr]; +"750 _frozen_param132" [id=750, type=get_attr]; +"751 dequantize_per_channel_default_132" [id=751, type=dequantize_per_channel]; +"752 features_5_16_norm1_weight" [id=752, type=get_attr]; +"753 features_5_16_norm1_bias" [id=753, type=get_attr]; +"754 _scale_133" [id=754, type=get_attr]; +"755 _zero_point_133" [id=755, type=get_attr]; +"756 _frozen_param133" [id=756, type=get_attr]; +"757 dequantize_per_channel_default_133" [id=757, type=dequantize_per_channel]; +"758 features_5_16_mlp_0_bias" [id=758, type=get_attr]; +"759 _scale_134" [id=759, type=get_attr]; +"760 _zero_point_134" [id=760, type=get_attr]; +"761 _frozen_param134" [id=761, type=get_attr]; +"762 dequantize_per_channel_default_134" [id=762, type=dequantize_per_channel]; +"763 features_5_16_mlp_3_bias" [id=763, type=get_attr]; +"764 features_5_16_norm2_weight" [id=764, type=get_attr]; +"765 features_5_16_norm2_bias" [id=765, type=get_attr]; +"766 _scale_135" [id=766, type=get_attr]; +"767 _zero_point_135" [id=767, type=get_attr]; +"768 _frozen_param135" [id=768, type=get_attr]; +"769 dequantize_per_channel_default_135" [id=769, type=dequantize_per_channel]; +"770 features_5_17_attn_cpb_mlp_0_bias" [id=770, type=get_attr]; +"771 _scale_136" [id=771, type=get_attr]; +"772 _zero_point_136" [id=772, type=get_attr]; +"773 _frozen_param136" [id=773, type=get_attr]; +"774 dequantize_per_channel_default_136" [id=774, type=dequantize_per_channel]; +"775 features_5_17_norm1_weight" [id=775, type=get_attr]; +"776 features_5_17_norm1_bias" [id=776, type=get_attr]; +"777 _scale_137" [id=777, type=get_attr]; +"778 _zero_point_137" [id=778, type=get_attr]; +"779 _frozen_param137" [id=779, type=get_attr]; +"780 dequantize_per_channel_default_137" [id=780, type=dequantize_per_channel]; +"781 features_5_17_mlp_0_bias" [id=781, type=get_attr]; +"782 _scale_138" [id=782, type=get_attr]; +"783 _zero_point_138" [id=783, type=get_attr]; +"784 _frozen_param138" [id=784, type=get_attr]; +"785 dequantize_per_channel_default_138" [id=785, type=dequantize_per_channel]; +"786 features_5_17_mlp_3_bias" [id=786, type=get_attr]; +"787 features_5_17_norm2_weight" [id=787, type=get_attr]; +"788 features_5_17_norm2_bias" [id=788, type=get_attr]; +"789 _scale_139" [id=789, type=get_attr]; +"790 _zero_point_139" [id=790, type=get_attr]; +"791 _frozen_param139" [id=791, type=get_attr]; +"792 dequantize_per_channel_default_139" [id=792, type=dequantize_per_channel]; +"793 features_6_norm_weight" [id=793, type=get_attr]; +"794 features_6_norm_bias" [id=794, type=get_attr]; +"795 _scale_140" [id=795, type=get_attr]; +"796 _zero_point_140" [id=796, type=get_attr]; +"797 _frozen_param140" [id=797, type=get_attr]; +"798 dequantize_per_channel_default_140" [id=798, type=dequantize_per_channel]; +"799 features_7_0_attn_cpb_mlp_0_bias" [id=799, type=get_attr]; +"800 _scale_141" [id=800, type=get_attr]; +"801 _zero_point_141" [id=801, type=get_attr]; +"802 _frozen_param141" [id=802, type=get_attr]; +"803 dequantize_per_channel_default_141" [id=803, type=dequantize_per_channel]; +"804 features_7_0_norm1_weight" [id=804, type=get_attr]; +"805 features_7_0_norm1_bias" [id=805, type=get_attr]; +"806 _scale_142" [id=806, type=get_attr]; +"807 _zero_point_142" [id=807, type=get_attr]; +"808 _frozen_param142" [id=808, type=get_attr]; +"809 dequantize_per_channel_default_142" [id=809, type=dequantize_per_channel]; +"810 features_7_0_mlp_0_bias" [id=810, type=get_attr]; +"811 _scale_143" [id=811, type=get_attr]; +"812 _zero_point_143" [id=812, type=get_attr]; +"813 _frozen_param143" [id=813, type=get_attr]; +"814 dequantize_per_channel_default_143" [id=814, type=dequantize_per_channel]; +"815 features_7_0_mlp_3_bias" [id=815, type=get_attr]; +"816 features_7_0_norm2_weight" [id=816, type=get_attr]; +"817 features_7_0_norm2_bias" [id=817, type=get_attr]; +"818 _scale_144" [id=818, type=get_attr]; +"819 _zero_point_144" [id=819, type=get_attr]; +"820 _frozen_param144" [id=820, type=get_attr]; +"821 dequantize_per_channel_default_144" [id=821, type=dequantize_per_channel]; +"822 features_7_1_attn_cpb_mlp_0_bias" [id=822, type=get_attr]; +"823 _scale_145" [id=823, type=get_attr]; +"824 _zero_point_145" [id=824, type=get_attr]; +"825 _frozen_param145" [id=825, type=get_attr]; +"826 dequantize_per_channel_default_145" [id=826, type=dequantize_per_channel]; +"827 features_7_1_norm1_weight" [id=827, type=get_attr]; +"828 features_7_1_norm1_bias" [id=828, type=get_attr]; +"829 _scale_146" [id=829, type=get_attr]; +"830 _zero_point_146" [id=830, type=get_attr]; +"831 _frozen_param146" [id=831, type=get_attr]; +"832 dequantize_per_channel_default_146" [id=832, type=dequantize_per_channel]; +"833 features_7_1_mlp_0_bias" [id=833, type=get_attr]; +"834 _scale_147" [id=834, type=get_attr]; +"835 _zero_point_147" [id=835, type=get_attr]; +"836 _frozen_param147" [id=836, type=get_attr]; +"837 dequantize_per_channel_default_147" [id=837, type=dequantize_per_channel]; +"838 features_7_1_mlp_3_bias" [id=838, type=get_attr]; +"839 features_7_1_norm2_weight" [id=839, type=get_attr]; +"840 features_7_1_norm2_bias" [id=840, type=get_attr]; +"841 norm_weight" [id=841, type=get_attr]; +"842 norm_bias" [id=842, type=get_attr]; +"843 _scale_148" [id=843, type=get_attr]; +"844 _zero_point_148" [id=844, type=get_attr]; +"845 _frozen_param148" [id=845, type=get_attr]; +"846 dequantize_per_channel_default_148" [id=846, type=dequantize_per_channel]; +"847 head_bias" [id=847, type=get_attr]; +"848 features_1_0_attn_relative_coords_table" [id=848, type=get_attr]; +"849 features_1_0_attn_relative_position_index" [id=849, type=get_attr]; +"850 features_1_1_attn_relative_coords_table" [id=850, type=get_attr]; +"851 features_1_1_attn_relative_position_index" [id=851, type=get_attr]; +"852 features_3_0_attn_relative_coords_table" [id=852, type=get_attr]; +"853 features_3_0_attn_relative_position_index" [id=853, type=get_attr]; +"854 features_3_1_attn_relative_coords_table" [id=854, type=get_attr]; +"855 features_3_1_attn_relative_position_index" [id=855, type=get_attr]; +"856 features_5_0_attn_relative_coords_table" [id=856, type=get_attr]; +"857 features_5_0_attn_relative_position_index" [id=857, type=get_attr]; +"858 features_5_1_attn_relative_coords_table" [id=858, type=get_attr]; +"859 features_5_1_attn_relative_position_index" [id=859, type=get_attr]; +"860 features_5_2_attn_relative_coords_table" [id=860, type=get_attr]; +"861 features_5_2_attn_relative_position_index" [id=861, type=get_attr]; +"862 features_5_3_attn_relative_coords_table" [id=862, type=get_attr]; +"863 features_5_3_attn_relative_position_index" [id=863, type=get_attr]; +"864 features_5_4_attn_relative_coords_table" [id=864, type=get_attr]; +"865 features_5_4_attn_relative_position_index" [id=865, type=get_attr]; +"866 features_5_5_attn_relative_coords_table" [id=866, type=get_attr]; +"867 features_5_5_attn_relative_position_index" [id=867, type=get_attr]; +"868 features_5_6_attn_relative_coords_table" [id=868, type=get_attr]; +"869 features_5_6_attn_relative_position_index" [id=869, type=get_attr]; +"870 features_5_7_attn_relative_coords_table" [id=870, type=get_attr]; +"871 features_5_7_attn_relative_position_index" [id=871, type=get_attr]; +"872 features_5_8_attn_relative_coords_table" [id=872, type=get_attr]; +"873 features_5_8_attn_relative_position_index" [id=873, type=get_attr]; +"874 features_5_9_attn_relative_coords_table" [id=874, type=get_attr]; +"875 features_5_9_attn_relative_position_index" [id=875, type=get_attr]; +"876 features_5_10_attn_relative_coords_table" [id=876, type=get_attr]; +"877 features_5_10_attn_relative_position_index" [id=877, type=get_attr]; +"878 features_5_11_attn_relative_coords_table" [id=878, type=get_attr]; +"879 features_5_11_attn_relative_position_index" [id=879, type=get_attr]; +"880 features_5_12_attn_relative_coords_table" [id=880, type=get_attr]; +"881 features_5_12_attn_relative_position_index" [id=881, type=get_attr]; +"882 features_5_13_attn_relative_coords_table" [id=882, type=get_attr]; +"883 features_5_13_attn_relative_position_index" [id=883, type=get_attr]; +"884 features_5_14_attn_relative_coords_table" [id=884, type=get_attr]; +"885 features_5_14_attn_relative_position_index" [id=885, type=get_attr]; +"886 features_5_15_attn_relative_coords_table" [id=886, type=get_attr]; +"887 features_5_15_attn_relative_position_index" [id=887, type=get_attr]; +"888 features_5_16_attn_relative_coords_table" [id=888, type=get_attr]; +"889 features_5_16_attn_relative_position_index" [id=889, type=get_attr]; +"890 features_5_17_attn_relative_coords_table" [id=890, type=get_attr]; +"891 features_5_17_attn_relative_position_index" [id=891, type=get_attr]; +"892 features_7_0_attn_relative_coords_table" [id=892, type=get_attr]; +"893 features_7_0_attn_relative_position_index" [id=893, type=get_attr]; +"894 features_7_1_attn_relative_coords_table" [id=894, type=get_attr]; +"895 features_7_1_attn_relative_position_index" [id=895, type=get_attr]; +"896 features_1_1_attn_lifted_tensor_0" [id=896, type=get_attr]; +"897 features_1_1_attn_lifted_tensor_1" [id=897, type=get_attr]; +"898 features_1_1_attn_lifted_tensor_2" [id=898, type=get_attr]; +"899 features_1_1_attn_lifted_tensor_3" [id=899, type=get_attr]; +"900 features_1_1_attn_lifted_tensor_4" [id=900, type=get_attr]; +"901 features_1_1_attn_lifted_tensor_5" [id=901, type=get_attr]; +"902 features_1_1_attn_lifted_tensor_6" [id=902, type=get_attr]; +"903 features_1_1_attn_lifted_tensor_7" [id=903, type=get_attr]; +"904 features_1_1_attn_lifted_tensor_8" [id=904, type=get_attr]; +"905 features_3_1_attn_lifted_tensor_9" [id=905, type=get_attr]; +"906 features_3_1_attn_lifted_tensor_10" [id=906, type=get_attr]; +"907 features_3_1_attn_lifted_tensor_11" [id=907, type=get_attr]; +"908 features_3_1_attn_lifted_tensor_12" [id=908, type=get_attr]; +"909 features_3_1_attn_lifted_tensor_13" [id=909, type=get_attr]; +"910 features_3_1_attn_lifted_tensor_14" [id=910, type=get_attr]; +"911 features_3_1_attn_lifted_tensor_15" [id=911, type=get_attr]; +"912 features_3_1_attn_lifted_tensor_16" [id=912, type=get_attr]; +"913 features_3_1_attn_lifted_tensor_17" [id=913, type=get_attr]; +"914 features_5_1_attn_lifted_tensor_18" [id=914, type=get_attr]; +"915 features_5_1_attn_lifted_tensor_19" [id=915, type=get_attr]; +"916 features_5_1_attn_lifted_tensor_20" [id=916, type=get_attr]; +"917 features_5_1_attn_lifted_tensor_21" [id=917, type=get_attr]; +"918 features_5_1_attn_lifted_tensor_22" [id=918, type=get_attr]; +"919 features_5_1_attn_lifted_tensor_23" [id=919, type=get_attr]; +"920 features_5_1_attn_lifted_tensor_24" [id=920, type=get_attr]; +"921 features_5_1_attn_lifted_tensor_25" [id=921, type=get_attr]; +"922 features_5_1_attn_lifted_tensor_26" [id=922, type=get_attr]; +"923 features_5_3_attn_lifted_tensor_27" [id=923, type=get_attr]; +"924 features_5_3_attn_lifted_tensor_28" [id=924, type=get_attr]; +"925 features_5_3_attn_lifted_tensor_29" [id=925, type=get_attr]; +"926 features_5_3_attn_lifted_tensor_30" [id=926, type=get_attr]; +"927 features_5_3_attn_lifted_tensor_31" [id=927, type=get_attr]; +"928 features_5_3_attn_lifted_tensor_32" [id=928, type=get_attr]; +"929 features_5_3_attn_lifted_tensor_33" [id=929, type=get_attr]; +"930 features_5_3_attn_lifted_tensor_34" [id=930, type=get_attr]; +"931 features_5_3_attn_lifted_tensor_35" [id=931, type=get_attr]; +"932 features_5_5_attn_lifted_tensor_36" [id=932, type=get_attr]; +"933 features_5_5_attn_lifted_tensor_37" [id=933, type=get_attr]; +"934 features_5_5_attn_lifted_tensor_38" [id=934, type=get_attr]; +"935 features_5_5_attn_lifted_tensor_39" [id=935, type=get_attr]; +"936 features_5_5_attn_lifted_tensor_40" [id=936, type=get_attr]; +"937 features_5_5_attn_lifted_tensor_41" [id=937, type=get_attr]; +"938 features_5_5_attn_lifted_tensor_42" [id=938, type=get_attr]; +"939 features_5_5_attn_lifted_tensor_43" [id=939, type=get_attr]; +"940 features_5_5_attn_lifted_tensor_44" [id=940, type=get_attr]; +"941 features_5_7_attn_lifted_tensor_45" [id=941, type=get_attr]; +"942 features_5_7_attn_lifted_tensor_46" [id=942, type=get_attr]; +"943 features_5_7_attn_lifted_tensor_47" [id=943, type=get_attr]; +"944 features_5_7_attn_lifted_tensor_48" [id=944, type=get_attr]; +"945 features_5_7_attn_lifted_tensor_49" [id=945, type=get_attr]; +"946 features_5_7_attn_lifted_tensor_50" [id=946, type=get_attr]; +"947 features_5_7_attn_lifted_tensor_51" [id=947, type=get_attr]; +"948 features_5_7_attn_lifted_tensor_52" [id=948, type=get_attr]; +"949 features_5_7_attn_lifted_tensor_53" [id=949, type=get_attr]; +"950 features_5_9_attn_lifted_tensor_54" [id=950, type=get_attr]; +"951 features_5_9_attn_lifted_tensor_55" [id=951, type=get_attr]; +"952 features_5_9_attn_lifted_tensor_56" [id=952, type=get_attr]; +"953 features_5_9_attn_lifted_tensor_57" [id=953, type=get_attr]; +"954 features_5_9_attn_lifted_tensor_58" [id=954, type=get_attr]; +"955 features_5_9_attn_lifted_tensor_59" [id=955, type=get_attr]; +"956 features_5_9_attn_lifted_tensor_60" [id=956, type=get_attr]; +"957 features_5_9_attn_lifted_tensor_61" [id=957, type=get_attr]; +"958 features_5_9_attn_lifted_tensor_62" [id=958, type=get_attr]; +"959 features_5_11_attn_lifted_tensor_63" [id=959, type=get_attr]; +"960 features_5_11_attn_lifted_tensor_64" [id=960, type=get_attr]; +"961 features_5_11_attn_lifted_tensor_65" [id=961, type=get_attr]; +"962 features_5_11_attn_lifted_tensor_66" [id=962, type=get_attr]; +"963 features_5_11_attn_lifted_tensor_67" [id=963, type=get_attr]; +"964 features_5_11_attn_lifted_tensor_68" [id=964, type=get_attr]; +"965 features_5_11_attn_lifted_tensor_69" [id=965, type=get_attr]; +"966 features_5_11_attn_lifted_tensor_70" [id=966, type=get_attr]; +"967 features_5_11_attn_lifted_tensor_71" [id=967, type=get_attr]; +"968 features_5_13_attn_lifted_tensor_72" [id=968, type=get_attr]; +"969 features_5_13_attn_lifted_tensor_73" [id=969, type=get_attr]; +"970 features_5_13_attn_lifted_tensor_74" [id=970, type=get_attr]; +"971 features_5_13_attn_lifted_tensor_75" [id=971, type=get_attr]; +"972 features_5_13_attn_lifted_tensor_76" [id=972, type=get_attr]; +"973 features_5_13_attn_lifted_tensor_77" [id=973, type=get_attr]; +"974 features_5_13_attn_lifted_tensor_78" [id=974, type=get_attr]; +"975 features_5_13_attn_lifted_tensor_79" [id=975, type=get_attr]; +"976 features_5_13_attn_lifted_tensor_80" [id=976, type=get_attr]; +"977 features_5_15_attn_lifted_tensor_81" [id=977, type=get_attr]; +"978 features_5_15_attn_lifted_tensor_82" [id=978, type=get_attr]; +"979 features_5_15_attn_lifted_tensor_83" [id=979, type=get_attr]; +"980 features_5_15_attn_lifted_tensor_84" [id=980, type=get_attr]; +"981 features_5_15_attn_lifted_tensor_85" [id=981, type=get_attr]; +"982 features_5_15_attn_lifted_tensor_86" [id=982, type=get_attr]; +"983 features_5_15_attn_lifted_tensor_87" [id=983, type=get_attr]; +"984 features_5_15_attn_lifted_tensor_88" [id=984, type=get_attr]; +"985 features_5_15_attn_lifted_tensor_89" [id=985, type=get_attr]; +"986 features_5_17_attn_lifted_tensor_90" [id=986, type=get_attr]; +"987 features_5_17_attn_lifted_tensor_91" [id=987, type=get_attr]; +"988 features_5_17_attn_lifted_tensor_92" [id=988, type=get_attr]; +"989 features_5_17_attn_lifted_tensor_93" [id=989, type=get_attr]; +"990 features_5_17_attn_lifted_tensor_94" [id=990, type=get_attr]; +"991 features_5_17_attn_lifted_tensor_95" [id=991, type=get_attr]; +"992 features_5_17_attn_lifted_tensor_96" [id=992, type=get_attr]; +"993 features_5_17_attn_lifted_tensor_97" [id=993, type=get_attr]; +"994 features_5_17_attn_lifted_tensor_98" [id=994, type=get_attr]; +"995 x" [id=995, type=input]; +"996 quantize_per_tensor_default" [id=996, type=quantize_per_tensor]; +"997 dequantize_per_tensor_default" [id=997, type=dequantize_per_tensor]; +"998 conv2d" [id=998, type=conv2d]; +"999 permute" [id=999, type=permute]; +"1000 layer_norm" [id=1000, type=layer_norm]; +"1001 quantize_per_tensor_default_1" [id=1001, type=quantize_per_tensor]; +"1002 dequantize_per_tensor_default_1" [id=1002, type=dequantize_per_tensor]; +"1003 linear" [id=1003, type=linear]; +"1004 relu_" [id=1004, type=relu_]; +"1005 linear_1" [id=1005, type=linear]; +"1006 view" [id=1006, type=view]; +"1007 index" [id=1007, type=index]; +"1008 view_1" [id=1008, type=view]; +"1009 permute_1" [id=1009, type=permute]; +"1010 contiguous" [id=1010, type=contiguous]; +"1011 unsqueeze" [id=1011, type=unsqueeze]; +"1012 sigmoid" [id=1012, type=sigmoid]; +"1013 mul" [id=1013, type=mul]; +"1014 pad" [id=1014, type=pad]; +"1015 view_2" [id=1015, type=view]; +"1016 permute_2" [id=1016, type=permute]; +"1017 reshape" [id=1017, type=reshape]; +"1018 clone" [id=1018, type=clone]; +"1019 slice_1" [id=1019, type=slice]; +"1020 zero_" [id=1020, type=zero_]; +"1021 linear_2" [id=1021, type=linear]; +"1022 reshape_1" [id=1022, type=reshape]; +"1023 permute_3" [id=1023, type=permute]; +"1024 select" [id=1024, type=select]; +"1025 select_1" [id=1025, type=select]; +"1026 select_2" [id=1026, type=select]; +"1027 linalg_vector_norm" [id=1027, type=linalg_vector_norm]; +"1028 clamp_min" [id=1028, type=clamp_min]; +"1029 expand_as" [id=1029, type=expand_as]; +"1030 div" [id=1030, type=div]; +"1031 quantize_per_tensor_default_2" [id=1031, type=quantize_per_tensor]; +"1032 dequantize_per_tensor_default_2" [id=1032, type=dequantize_per_tensor]; +"1033 linalg_vector_norm_1" [id=1033, type=linalg_vector_norm]; +"1034 clamp_min_1" [id=1034, type=clamp_min]; +"1035 expand_as_1" [id=1035, type=expand_as]; +"1036 div_1" [id=1036, type=div]; +"1037 quantize_per_tensor_default_3" [id=1037, type=quantize_per_tensor]; +"1038 dequantize_per_tensor_default_3" [id=1038, type=dequantize_per_tensor]; +"1039 transpose" [id=1039, type=transpose]; +"1040 matmul" [id=1040, type=matmul]; +"1041 clamp" [id=1041, type=clamp]; +"1042 exp" [id=1042, type=exp]; +"1043 mul_1" [id=1043, type=mul]; +"1044 add" [id=1044, type=add]; +"1045 softmax" [id=1045, type=softmax]; +"1046 dropout" [id=1046, type=dropout]; +"1047 matmul_1" [id=1047, type=matmul]; +"1048 quantize_per_tensor_default_4" [id=1048, type=quantize_per_tensor]; +"1049 dequantize_per_tensor_default_4" [id=1049, type=dequantize_per_tensor]; +"1050 transpose_1" [id=1050, type=transpose]; +"1051 reshape_2" [id=1051, type=reshape]; +"1052 linear_3" [id=1052, type=linear]; +"1053 dropout_1" [id=1053, type=dropout]; +"1054 view_3" [id=1054, type=view]; +"1055 permute_4" [id=1055, type=permute]; +"1056 reshape_3" [id=1056, type=reshape]; +"1057 slice_2" [id=1057, type=slice]; +"1058 slice_3" [id=1058, type=slice]; +"1059 layer_norm_1" [id=1059, type=layer_norm]; +"1060 add_1" [id=1060, type=add]; +"1061 quantize_per_tensor_default_5" [id=1061, type=quantize_per_tensor]; +"1062 dequantize_per_tensor_default_5" [id=1062, type=dequantize_per_tensor]; +"1063 linear_4" [id=1063, type=linear]; +"1064 gelu" [id=1064, type=gelu]; +"1065 quantize_per_tensor_default_6" [id=1065, type=quantize_per_tensor]; +"1066 dequantize_per_tensor_default_6" [id=1066, type=dequantize_per_tensor]; +"1067 dropout_2" [id=1067, type=dropout]; +"1068 linear_5" [id=1068, type=linear]; +"1069 dropout_3" [id=1069, type=dropout]; +"1070 layer_norm_2" [id=1070, type=layer_norm]; +"1071 add_2" [id=1071, type=add]; +"1072 linear_6" [id=1072, type=linear]; +"1073 relu__1" [id=1073, type=relu_]; +"1074 linear_7" [id=1074, type=linear]; +"1075 view_4" [id=1075, type=view]; +"1076 index_1" [id=1076, type=index]; +"1077 view_5" [id=1077, type=view]; +"1078 permute_5" [id=1078, type=permute]; +"1079 contiguous_1" [id=1079, type=contiguous]; +"1080 unsqueeze_1" [id=1080, type=unsqueeze]; +"1081 sigmoid_1" [id=1081, type=sigmoid]; +"1082 mul_2" [id=1082, type=mul]; +"1083 pad_1" [id=1083, type=pad]; +"1084 roll" [id=1084, type=roll]; +"1085 view_6" [id=1085, type=view]; +"1086 permute_6" [id=1086, type=permute]; +"1087 reshape_4" [id=1087, type=reshape]; +"1088 quantize_per_tensor_default_7" [id=1088, type=quantize_per_tensor]; +"1089 dequantize_per_tensor_default_7" [id=1089, type=dequantize_per_tensor]; +"1090 clone_1" [id=1090, type=clone]; +"1091 slice_4" [id=1091, type=slice]; +"1092 zero__1" [id=1092, type=zero_]; +"1093 linear_8" [id=1093, type=linear]; +"1094 reshape_5" [id=1094, type=reshape]; +"1095 permute_7" [id=1095, type=permute]; +"1096 select_3" [id=1096, type=select]; +"1097 select_4" [id=1097, type=select]; +"1098 select_5" [id=1098, type=select]; +"1099 linalg_vector_norm_2" [id=1099, type=linalg_vector_norm]; +"1100 clamp_min_2" [id=1100, type=clamp_min]; +"1101 expand_as_2" [id=1101, type=expand_as]; +"1102 div_2" [id=1102, type=div]; +"1103 quantize_per_tensor_default_8" [id=1103, type=quantize_per_tensor]; +"1104 dequantize_per_tensor_default_8" [id=1104, type=dequantize_per_tensor]; +"1105 linalg_vector_norm_3" [id=1105, type=linalg_vector_norm]; +"1106 clamp_min_3" [id=1106, type=clamp_min]; +"1107 expand_as_3" [id=1107, type=expand_as]; +"1108 div_3" [id=1108, type=div]; +"1109 quantize_per_tensor_default_9" [id=1109, type=quantize_per_tensor]; +"1110 dequantize_per_tensor_default_9" [id=1110, type=dequantize_per_tensor]; +"1111 transpose_2" [id=1111, type=transpose]; +"1112 matmul_2" [id=1112, type=matmul]; +"1113 clamp_1" [id=1113, type=clamp]; +"1114 exp_1" [id=1114, type=exp]; +"1115 mul_3" [id=1115, type=mul]; +"1116 add_3" [id=1116, type=add]; +"1117 new_zeros" [id=1117, type=new_zeros]; +"1118 lift_fresh_copy" [id=1118, type=lift_fresh_copy]; +"1119 slice_5" [id=1119, type=slice]; +"1120 slice_6" [id=1120, type=slice]; +"1121 fill_" [id=1121, type=fill_]; +"1122 lift_fresh_copy_1" [id=1122, type=lift_fresh_copy]; +"1123 slice_7" [id=1123, type=slice]; +"1124 slice_8" [id=1124, type=slice]; +"1125 fill__1" [id=1125, type=fill_]; +"1126 lift_fresh_copy_2" [id=1126, type=lift_fresh_copy]; +"1127 slice_9" [id=1127, type=slice]; +"1128 slice_10" [id=1128, type=slice]; +"1129 fill__2" [id=1129, type=fill_]; +"1130 lift_fresh_copy_3" [id=1130, type=lift_fresh_copy]; +"1131 slice_11" [id=1131, type=slice]; +"1132 slice_12" [id=1132, type=slice]; +"1133 fill__3" [id=1133, type=fill_]; +"1134 lift_fresh_copy_4" [id=1134, type=lift_fresh_copy]; +"1135 slice_13" [id=1135, type=slice]; +"1136 slice_14" [id=1136, type=slice]; +"1137 fill__4" [id=1137, type=fill_]; +"1138 lift_fresh_copy_5" [id=1138, type=lift_fresh_copy]; +"1139 slice_15" [id=1139, type=slice]; +"1140 slice_16" [id=1140, type=slice]; +"1141 fill__5" [id=1141, type=fill_]; +"1142 lift_fresh_copy_6" [id=1142, type=lift_fresh_copy]; +"1143 slice_17" [id=1143, type=slice]; +"1144 slice_18" [id=1144, type=slice]; +"1145 fill__6" [id=1145, type=fill_]; +"1146 lift_fresh_copy_7" [id=1146, type=lift_fresh_copy]; +"1147 slice_19" [id=1147, type=slice]; +"1148 slice_20" [id=1148, type=slice]; +"1149 fill__7" [id=1149, type=fill_]; +"1150 lift_fresh_copy_8" [id=1150, type=lift_fresh_copy]; +"1151 slice_21" [id=1151, type=slice]; +"1152 slice_22" [id=1152, type=slice]; +"1153 fill__8" [id=1153, type=fill_]; +"1154 view_7" [id=1154, type=view]; +"1155 permute_8" [id=1155, type=permute]; +"1156 reshape_6" [id=1156, type=reshape]; +"1157 unsqueeze_2" [id=1157, type=unsqueeze]; +"1158 unsqueeze_3" [id=1158, type=unsqueeze]; +"1159 sub" [id=1159, type=sub]; +"1160 ne" [id=1160, type=ne]; +"1161 masked_fill" [id=1161, type=masked_fill]; +"1162 eq" [id=1162, type=eq]; +"1163 masked_fill_1" [id=1163, type=masked_fill]; +"1164 view_8" [id=1164, type=view]; +"1165 unsqueeze_4" [id=1165, type=unsqueeze]; +"1166 unsqueeze_5" [id=1166, type=unsqueeze]; +"1167 add_4" [id=1167, type=add]; +"1168 view_9" [id=1168, type=view]; +"1169 softmax_1" [id=1169, type=softmax]; +"1170 dropout_4" [id=1170, type=dropout]; +"1171 matmul_3" [id=1171, type=matmul]; +"1172 quantize_per_tensor_default_10" [id=1172, type=quantize_per_tensor]; +"1173 dequantize_per_tensor_default_10" [id=1173, type=dequantize_per_tensor]; +"1174 transpose_3" [id=1174, type=transpose]; +"1175 reshape_7" [id=1175, type=reshape]; +"1176 linear_9" [id=1176, type=linear]; +"1177 dropout_5" [id=1177, type=dropout]; +"1178 view_10" [id=1178, type=view]; +"1179 permute_9" [id=1179, type=permute]; +"1180 reshape_8" [id=1180, type=reshape]; +"1181 roll_1" [id=1181, type=roll]; +"1182 slice_23" [id=1182, type=slice]; +"1183 slice_24" [id=1183, type=slice]; +"1184 layer_norm_3" [id=1184, type=layer_norm]; +"1185 add_5" [id=1185, type=add]; +"1186 quantize_per_tensor_default_11" [id=1186, type=quantize_per_tensor]; +"1187 dequantize_per_tensor_default_11" [id=1187, type=dequantize_per_tensor]; +"1188 linear_10" [id=1188, type=linear]; +"1189 gelu_1" [id=1189, type=gelu]; +"1190 quantize_per_tensor_default_12" [id=1190, type=quantize_per_tensor]; +"1191 dequantize_per_tensor_default_12" [id=1191, type=dequantize_per_tensor]; +"1192 dropout_6" [id=1192, type=dropout]; +"1193 linear_11" [id=1193, type=linear]; +"1194 dropout_7" [id=1194, type=dropout]; +"1195 layer_norm_4" [id=1195, type=layer_norm]; +"1196 add_6" [id=1196, type=add]; +"1197 quantize_per_tensor_default_13" [id=1197, type=quantize_per_tensor]; +"1198 dequantize_per_tensor_default_13" [id=1198, type=dequantize_per_tensor]; +"1199 pad_2" [id=1199, type=pad]; +"1200 slice_25" [id=1200, type=slice]; +"1201 slice_26" [id=1201, type=slice]; +"1202 slice_27" [id=1202, type=slice]; +"1203 slice_28" [id=1203, type=slice]; +"1204 slice_29" [id=1204, type=slice]; +"1205 slice_30" [id=1205, type=slice]; +"1206 slice_31" [id=1206, type=slice]; +"1207 slice_32" [id=1207, type=slice]; +"1208 slice_33" [id=1208, type=slice]; +"1209 slice_34" [id=1209, type=slice]; +"1210 slice_35" [id=1210, type=slice]; +"1211 slice_36" [id=1211, type=slice]; +"1212 cat" [id=1212, type=cat]; +"1213 linear_12" [id=1213, type=linear]; +"1214 layer_norm_5" [id=1214, type=layer_norm]; +"1215 quantize_per_tensor_default_14" [id=1215, type=quantize_per_tensor]; +"1216 dequantize_per_tensor_default_14" [id=1216, type=dequantize_per_tensor]; +"1217 linear_13" [id=1217, type=linear]; +"1218 relu__2" [id=1218, type=relu_]; +"1219 linear_14" [id=1219, type=linear]; +"1220 view_11" [id=1220, type=view]; +"1221 index_2" [id=1221, type=index]; +"1222 view_12" [id=1222, type=view]; +"1223 permute_10" [id=1223, type=permute]; +"1224 contiguous_2" [id=1224, type=contiguous]; +"1225 unsqueeze_6" [id=1225, type=unsqueeze]; +"1226 sigmoid_2" [id=1226, type=sigmoid]; +"1227 mul_4" [id=1227, type=mul]; +"1228 pad_3" [id=1228, type=pad]; +"1229 view_13" [id=1229, type=view]; +"1230 permute_11" [id=1230, type=permute]; +"1231 reshape_9" [id=1231, type=reshape]; +"1232 clone_2" [id=1232, type=clone]; +"1233 slice_37" [id=1233, type=slice]; +"1234 zero__2" [id=1234, type=zero_]; +"1235 linear_15" [id=1235, type=linear]; +"1236 reshape_10" [id=1236, type=reshape]; +"1237 permute_12" [id=1237, type=permute]; +"1238 select_6" [id=1238, type=select]; +"1239 select_7" [id=1239, type=select]; +"1240 select_8" [id=1240, type=select]; +"1241 linalg_vector_norm_4" [id=1241, type=linalg_vector_norm]; +"1242 clamp_min_4" [id=1242, type=clamp_min]; +"1243 expand_as_4" [id=1243, type=expand_as]; +"1244 div_4" [id=1244, type=div]; +"1245 quantize_per_tensor_default_15" [id=1245, type=quantize_per_tensor]; +"1246 dequantize_per_tensor_default_15" [id=1246, type=dequantize_per_tensor]; +"1247 linalg_vector_norm_5" [id=1247, type=linalg_vector_norm]; +"1248 clamp_min_5" [id=1248, type=clamp_min]; +"1249 expand_as_5" [id=1249, type=expand_as]; +"1250 div_5" [id=1250, type=div]; +"1251 quantize_per_tensor_default_16" [id=1251, type=quantize_per_tensor]; +"1252 dequantize_per_tensor_default_16" [id=1252, type=dequantize_per_tensor]; +"1253 transpose_4" [id=1253, type=transpose]; +"1254 matmul_4" [id=1254, type=matmul]; +"1255 clamp_2" [id=1255, type=clamp]; +"1256 exp_2" [id=1256, type=exp]; +"1257 mul_5" [id=1257, type=mul]; +"1258 add_7" [id=1258, type=add]; +"1259 softmax_2" [id=1259, type=softmax]; +"1260 dropout_8" [id=1260, type=dropout]; +"1261 matmul_5" [id=1261, type=matmul]; +"1262 quantize_per_tensor_default_17" [id=1262, type=quantize_per_tensor]; +"1263 dequantize_per_tensor_default_17" [id=1263, type=dequantize_per_tensor]; +"1264 transpose_5" [id=1264, type=transpose]; +"1265 reshape_11" [id=1265, type=reshape]; +"1266 linear_16" [id=1266, type=linear]; +"1267 dropout_9" [id=1267, type=dropout]; +"1268 view_14" [id=1268, type=view]; +"1269 permute_13" [id=1269, type=permute]; +"1270 reshape_12" [id=1270, type=reshape]; +"1271 slice_38" [id=1271, type=slice]; +"1272 slice_39" [id=1272, type=slice]; +"1273 slice_40" [id=1273, type=slice]; +"1274 slice_41" [id=1274, type=slice]; +"1275 contiguous_3" [id=1275, type=contiguous]; +"1276 layer_norm_6" [id=1276, type=layer_norm]; +"1277 add_8" [id=1277, type=add]; +"1278 quantize_per_tensor_default_18" [id=1278, type=quantize_per_tensor]; +"1279 dequantize_per_tensor_default_18" [id=1279, type=dequantize_per_tensor]; +"1280 linear_17" [id=1280, type=linear]; +"1281 gelu_2" [id=1281, type=gelu]; +"1282 quantize_per_tensor_default_19" [id=1282, type=quantize_per_tensor]; +"1283 dequantize_per_tensor_default_19" [id=1283, type=dequantize_per_tensor]; +"1284 dropout_10" [id=1284, type=dropout]; +"1285 linear_18" [id=1285, type=linear]; +"1286 dropout_11" [id=1286, type=dropout]; +"1287 layer_norm_7" [id=1287, type=layer_norm]; +"1288 add_9" [id=1288, type=add]; +"1289 linear_19" [id=1289, type=linear]; +"1290 relu__3" [id=1290, type=relu_]; +"1291 linear_20" [id=1291, type=linear]; +"1292 view_15" [id=1292, type=view]; +"1293 index_3" [id=1293, type=index]; +"1294 view_16" [id=1294, type=view]; +"1295 permute_14" [id=1295, type=permute]; +"1296 contiguous_4" [id=1296, type=contiguous]; +"1297 unsqueeze_7" [id=1297, type=unsqueeze]; +"1298 sigmoid_3" [id=1298, type=sigmoid]; +"1299 mul_6" [id=1299, type=mul]; +"1300 pad_4" [id=1300, type=pad]; +"1301 roll_2" [id=1301, type=roll]; +"1302 view_17" [id=1302, type=view]; +"1303 permute_15" [id=1303, type=permute]; +"1304 reshape_13" [id=1304, type=reshape]; +"1305 quantize_per_tensor_default_20" [id=1305, type=quantize_per_tensor]; +"1306 dequantize_per_tensor_default_20" [id=1306, type=dequantize_per_tensor]; +"1307 clone_3" [id=1307, type=clone]; +"1308 slice_42" [id=1308, type=slice]; +"1309 zero__3" [id=1309, type=zero_]; +"1310 linear_21" [id=1310, type=linear]; +"1311 reshape_14" [id=1311, type=reshape]; +"1312 permute_16" [id=1312, type=permute]; +"1313 select_9" [id=1313, type=select]; +"1314 select_10" [id=1314, type=select]; +"1315 select_11" [id=1315, type=select]; +"1316 linalg_vector_norm_6" [id=1316, type=linalg_vector_norm]; +"1317 clamp_min_6" [id=1317, type=clamp_min]; +"1318 expand_as_6" [id=1318, type=expand_as]; +"1319 div_6" [id=1319, type=div]; +"1320 quantize_per_tensor_default_21" [id=1320, type=quantize_per_tensor]; +"1321 dequantize_per_tensor_default_21" [id=1321, type=dequantize_per_tensor]; +"1322 linalg_vector_norm_7" [id=1322, type=linalg_vector_norm]; +"1323 clamp_min_7" [id=1323, type=clamp_min]; +"1324 expand_as_7" [id=1324, type=expand_as]; +"1325 div_7" [id=1325, type=div]; +"1326 quantize_per_tensor_default_22" [id=1326, type=quantize_per_tensor]; +"1327 dequantize_per_tensor_default_22" [id=1327, type=dequantize_per_tensor]; +"1328 transpose_6" [id=1328, type=transpose]; +"1329 matmul_6" [id=1329, type=matmul]; +"1330 clamp_3" [id=1330, type=clamp]; +"1331 exp_3" [id=1331, type=exp]; +"1332 mul_7" [id=1332, type=mul]; +"1333 add_10" [id=1333, type=add]; +"1334 new_zeros_1" [id=1334, type=new_zeros]; +"1335 lift_fresh_copy_9" [id=1335, type=lift_fresh_copy]; +"1336 slice_43" [id=1336, type=slice]; +"1337 slice_44" [id=1337, type=slice]; +"1338 fill__9" [id=1338, type=fill_]; +"1339 lift_fresh_copy_10" [id=1339, type=lift_fresh_copy]; +"1340 slice_45" [id=1340, type=slice]; +"1341 slice_46" [id=1341, type=slice]; +"1342 fill__10" [id=1342, type=fill_]; +"1343 lift_fresh_copy_11" [id=1343, type=lift_fresh_copy]; +"1344 slice_47" [id=1344, type=slice]; +"1345 slice_48" [id=1345, type=slice]; +"1346 fill__11" [id=1346, type=fill_]; +"1347 lift_fresh_copy_12" [id=1347, type=lift_fresh_copy]; +"1348 slice_49" [id=1348, type=slice]; +"1349 slice_50" [id=1349, type=slice]; +"1350 fill__12" [id=1350, type=fill_]; +"1351 lift_fresh_copy_13" [id=1351, type=lift_fresh_copy]; +"1352 slice_51" [id=1352, type=slice]; +"1353 slice_52" [id=1353, type=slice]; +"1354 fill__13" [id=1354, type=fill_]; +"1355 lift_fresh_copy_14" [id=1355, type=lift_fresh_copy]; +"1356 slice_53" [id=1356, type=slice]; +"1357 slice_54" [id=1357, type=slice]; +"1358 fill__14" [id=1358, type=fill_]; +"1359 lift_fresh_copy_15" [id=1359, type=lift_fresh_copy]; +"1360 slice_55" [id=1360, type=slice]; +"1361 slice_56" [id=1361, type=slice]; +"1362 fill__15" [id=1362, type=fill_]; +"1363 lift_fresh_copy_16" [id=1363, type=lift_fresh_copy]; +"1364 slice_57" [id=1364, type=slice]; +"1365 slice_58" [id=1365, type=slice]; +"1366 fill__16" [id=1366, type=fill_]; +"1367 lift_fresh_copy_17" [id=1367, type=lift_fresh_copy]; +"1368 slice_59" [id=1368, type=slice]; +"1369 slice_60" [id=1369, type=slice]; +"1370 fill__17" [id=1370, type=fill_]; +"1371 view_18" [id=1371, type=view]; +"1372 permute_17" [id=1372, type=permute]; +"1373 reshape_15" [id=1373, type=reshape]; +"1374 unsqueeze_8" [id=1374, type=unsqueeze]; +"1375 unsqueeze_9" [id=1375, type=unsqueeze]; +"1376 sub_1" [id=1376, type=sub]; +"1377 ne_1" [id=1377, type=ne]; +"1378 masked_fill_2" [id=1378, type=masked_fill]; +"1379 eq_1" [id=1379, type=eq]; +"1380 masked_fill_3" [id=1380, type=masked_fill]; +"1381 view_19" [id=1381, type=view]; +"1382 unsqueeze_10" [id=1382, type=unsqueeze]; +"1383 unsqueeze_11" [id=1383, type=unsqueeze]; +"1384 add_11" [id=1384, type=add]; +"1385 view_20" [id=1385, type=view]; +"1386 softmax_3" [id=1386, type=softmax]; +"1387 dropout_12" [id=1387, type=dropout]; +"1388 matmul_7" [id=1388, type=matmul]; +"1389 quantize_per_tensor_default_23" [id=1389, type=quantize_per_tensor]; +"1390 dequantize_per_tensor_default_23" [id=1390, type=dequantize_per_tensor]; +"1391 transpose_7" [id=1391, type=transpose]; +"1392 reshape_16" [id=1392, type=reshape]; +"1393 linear_22" [id=1393, type=linear]; +"1394 dropout_13" [id=1394, type=dropout]; +"1395 view_21" [id=1395, type=view]; +"1396 permute_18" [id=1396, type=permute]; +"1397 reshape_17" [id=1397, type=reshape]; +"1398 roll_3" [id=1398, type=roll]; +"1399 slice_61" [id=1399, type=slice]; +"1400 slice_62" [id=1400, type=slice]; +"1401 slice_63" [id=1401, type=slice]; +"1402 slice_64" [id=1402, type=slice]; +"1403 contiguous_5" [id=1403, type=contiguous]; +"1404 layer_norm_8" [id=1404, type=layer_norm]; +"1405 add_12" [id=1405, type=add]; +"1406 quantize_per_tensor_default_24" [id=1406, type=quantize_per_tensor]; +"1407 dequantize_per_tensor_default_24" [id=1407, type=dequantize_per_tensor]; +"1408 linear_23" [id=1408, type=linear]; +"1409 gelu_3" [id=1409, type=gelu]; +"1410 quantize_per_tensor_default_25" [id=1410, type=quantize_per_tensor]; +"1411 dequantize_per_tensor_default_25" [id=1411, type=dequantize_per_tensor]; +"1412 dropout_14" [id=1412, type=dropout]; +"1413 linear_24" [id=1413, type=linear]; +"1414 dropout_15" [id=1414, type=dropout]; +"1415 layer_norm_9" [id=1415, type=layer_norm]; +"1416 add_13" [id=1416, type=add]; +"1417 quantize_per_tensor_default_26" [id=1417, type=quantize_per_tensor]; +"1418 dequantize_per_tensor_default_26" [id=1418, type=dequantize_per_tensor]; +"1419 pad_5" [id=1419, type=pad]; +"1420 slice_65" [id=1420, type=slice]; +"1421 slice_66" [id=1421, type=slice]; +"1422 slice_67" [id=1422, type=slice]; +"1423 slice_68" [id=1423, type=slice]; +"1424 slice_69" [id=1424, type=slice]; +"1425 slice_70" [id=1425, type=slice]; +"1426 slice_71" [id=1426, type=slice]; +"1427 slice_72" [id=1427, type=slice]; +"1428 slice_73" [id=1428, type=slice]; +"1429 slice_74" [id=1429, type=slice]; +"1430 slice_75" [id=1430, type=slice]; +"1431 slice_76" [id=1431, type=slice]; +"1432 cat_1" [id=1432, type=cat]; +"1433 linear_25" [id=1433, type=linear]; +"1434 layer_norm_10" [id=1434, type=layer_norm]; +"1435 quantize_per_tensor_default_27" [id=1435, type=quantize_per_tensor]; +"1436 dequantize_per_tensor_default_27" [id=1436, type=dequantize_per_tensor]; +"1437 linear_26" [id=1437, type=linear]; +"1438 relu__4" [id=1438, type=relu_]; +"1439 linear_27" [id=1439, type=linear]; +"1440 view_22" [id=1440, type=view]; +"1441 index_4" [id=1441, type=index]; +"1442 view_23" [id=1442, type=view]; +"1443 permute_19" [id=1443, type=permute]; +"1444 contiguous_6" [id=1444, type=contiguous]; +"1445 unsqueeze_12" [id=1445, type=unsqueeze]; +"1446 sigmoid_4" [id=1446, type=sigmoid]; +"1447 mul_8" [id=1447, type=mul]; +"1448 pad_6" [id=1448, type=pad]; +"1449 view_24" [id=1449, type=view]; +"1450 permute_20" [id=1450, type=permute]; +"1451 reshape_18" [id=1451, type=reshape]; +"1452 clone_4" [id=1452, type=clone]; +"1453 slice_77" [id=1453, type=slice]; +"1454 zero__4" [id=1454, type=zero_]; +"1455 linear_28" [id=1455, type=linear]; +"1456 reshape_19" [id=1456, type=reshape]; +"1457 permute_21" [id=1457, type=permute]; +"1458 select_12" [id=1458, type=select]; +"1459 select_13" [id=1459, type=select]; +"1460 select_14" [id=1460, type=select]; +"1461 linalg_vector_norm_8" [id=1461, type=linalg_vector_norm]; +"1462 clamp_min_8" [id=1462, type=clamp_min]; +"1463 expand_as_8" [id=1463, type=expand_as]; +"1464 div_8" [id=1464, type=div]; +"1465 quantize_per_tensor_default_28" [id=1465, type=quantize_per_tensor]; +"1466 dequantize_per_tensor_default_28" [id=1466, type=dequantize_per_tensor]; +"1467 linalg_vector_norm_9" [id=1467, type=linalg_vector_norm]; +"1468 clamp_min_9" [id=1468, type=clamp_min]; +"1469 expand_as_9" [id=1469, type=expand_as]; +"1470 div_9" [id=1470, type=div]; +"1471 quantize_per_tensor_default_29" [id=1471, type=quantize_per_tensor]; +"1472 dequantize_per_tensor_default_29" [id=1472, type=dequantize_per_tensor]; +"1473 transpose_8" [id=1473, type=transpose]; +"1474 matmul_8" [id=1474, type=matmul]; +"1475 clamp_4" [id=1475, type=clamp]; +"1476 exp_4" [id=1476, type=exp]; +"1477 mul_9" [id=1477, type=mul]; +"1478 add_14" [id=1478, type=add]; +"1479 softmax_4" [id=1479, type=softmax]; +"1480 dropout_16" [id=1480, type=dropout]; +"1481 matmul_9" [id=1481, type=matmul]; +"1482 quantize_per_tensor_default_30" [id=1482, type=quantize_per_tensor]; +"1483 dequantize_per_tensor_default_30" [id=1483, type=dequantize_per_tensor]; +"1484 transpose_9" [id=1484, type=transpose]; +"1485 reshape_20" [id=1485, type=reshape]; +"1486 linear_29" [id=1486, type=linear]; +"1487 dropout_17" [id=1487, type=dropout]; +"1488 view_25" [id=1488, type=view]; +"1489 permute_22" [id=1489, type=permute]; +"1490 reshape_21" [id=1490, type=reshape]; +"1491 slice_78" [id=1491, type=slice]; +"1492 slice_79" [id=1492, type=slice]; +"1493 slice_80" [id=1493, type=slice]; +"1494 slice_81" [id=1494, type=slice]; +"1495 contiguous_7" [id=1495, type=contiguous]; +"1496 layer_norm_11" [id=1496, type=layer_norm]; +"1497 add_15" [id=1497, type=add]; +"1498 quantize_per_tensor_default_31" [id=1498, type=quantize_per_tensor]; +"1499 dequantize_per_tensor_default_31" [id=1499, type=dequantize_per_tensor]; +"1500 linear_30" [id=1500, type=linear]; +"1501 gelu_4" [id=1501, type=gelu]; +"1502 quantize_per_tensor_default_32" [id=1502, type=quantize_per_tensor]; +"1503 dequantize_per_tensor_default_32" [id=1503, type=dequantize_per_tensor]; +"1504 dropout_18" [id=1504, type=dropout]; +"1505 linear_31" [id=1505, type=linear]; +"1506 dropout_19" [id=1506, type=dropout]; +"1507 layer_norm_12" [id=1507, type=layer_norm]; +"1508 add_16" [id=1508, type=add]; +"1509 linear_32" [id=1509, type=linear]; +"1510 relu__5" [id=1510, type=relu_]; +"1511 linear_33" [id=1511, type=linear]; +"1512 view_26" [id=1512, type=view]; +"1513 index_5" [id=1513, type=index]; +"1514 view_27" [id=1514, type=view]; +"1515 permute_23" [id=1515, type=permute]; +"1516 contiguous_8" [id=1516, type=contiguous]; +"1517 unsqueeze_13" [id=1517, type=unsqueeze]; +"1518 sigmoid_5" [id=1518, type=sigmoid]; +"1519 mul_10" [id=1519, type=mul]; +"1520 pad_7" [id=1520, type=pad]; +"1521 roll_4" [id=1521, type=roll]; +"1522 view_28" [id=1522, type=view]; +"1523 permute_24" [id=1523, type=permute]; +"1524 reshape_22" [id=1524, type=reshape]; +"1525 quantize_per_tensor_default_33" [id=1525, type=quantize_per_tensor]; +"1526 dequantize_per_tensor_default_33" [id=1526, type=dequantize_per_tensor]; +"1527 clone_5" [id=1527, type=clone]; +"1528 slice_82" [id=1528, type=slice]; +"1529 zero__5" [id=1529, type=zero_]; +"1530 linear_34" [id=1530, type=linear]; +"1531 reshape_23" [id=1531, type=reshape]; +"1532 permute_25" [id=1532, type=permute]; +"1533 select_15" [id=1533, type=select]; +"1534 select_16" [id=1534, type=select]; +"1535 select_17" [id=1535, type=select]; +"1536 linalg_vector_norm_10" [id=1536, type=linalg_vector_norm]; +"1537 clamp_min_10" [id=1537, type=clamp_min]; +"1538 expand_as_10" [id=1538, type=expand_as]; +"1539 div_10" [id=1539, type=div]; +"1540 quantize_per_tensor_default_34" [id=1540, type=quantize_per_tensor]; +"1541 dequantize_per_tensor_default_34" [id=1541, type=dequantize_per_tensor]; +"1542 linalg_vector_norm_11" [id=1542, type=linalg_vector_norm]; +"1543 clamp_min_11" [id=1543, type=clamp_min]; +"1544 expand_as_11" [id=1544, type=expand_as]; +"1545 div_11" [id=1545, type=div]; +"1546 quantize_per_tensor_default_35" [id=1546, type=quantize_per_tensor]; +"1547 dequantize_per_tensor_default_35" [id=1547, type=dequantize_per_tensor]; +"1548 transpose_10" [id=1548, type=transpose]; +"1549 matmul_10" [id=1549, type=matmul]; +"1550 clamp_5" [id=1550, type=clamp]; +"1551 exp_5" [id=1551, type=exp]; +"1552 mul_11" [id=1552, type=mul]; +"1553 add_17" [id=1553, type=add]; +"1554 new_zeros_2" [id=1554, type=new_zeros]; +"1555 lift_fresh_copy_18" [id=1555, type=lift_fresh_copy]; +"1556 slice_83" [id=1556, type=slice]; +"1557 slice_84" [id=1557, type=slice]; +"1558 fill__18" [id=1558, type=fill_]; +"1559 lift_fresh_copy_19" [id=1559, type=lift_fresh_copy]; +"1560 slice_85" [id=1560, type=slice]; +"1561 slice_86" [id=1561, type=slice]; +"1562 fill__19" [id=1562, type=fill_]; +"1563 lift_fresh_copy_20" [id=1563, type=lift_fresh_copy]; +"1564 slice_87" [id=1564, type=slice]; +"1565 slice_88" [id=1565, type=slice]; +"1566 fill__20" [id=1566, type=fill_]; +"1567 lift_fresh_copy_21" [id=1567, type=lift_fresh_copy]; +"1568 slice_89" [id=1568, type=slice]; +"1569 slice_90" [id=1569, type=slice]; +"1570 fill__21" [id=1570, type=fill_]; +"1571 lift_fresh_copy_22" [id=1571, type=lift_fresh_copy]; +"1572 slice_91" [id=1572, type=slice]; +"1573 slice_92" [id=1573, type=slice]; +"1574 fill__22" [id=1574, type=fill_]; +"1575 lift_fresh_copy_23" [id=1575, type=lift_fresh_copy]; +"1576 slice_93" [id=1576, type=slice]; +"1577 slice_94" [id=1577, type=slice]; +"1578 fill__23" [id=1578, type=fill_]; +"1579 lift_fresh_copy_24" [id=1579, type=lift_fresh_copy]; +"1580 slice_95" [id=1580, type=slice]; +"1581 slice_96" [id=1581, type=slice]; +"1582 fill__24" [id=1582, type=fill_]; +"1583 lift_fresh_copy_25" [id=1583, type=lift_fresh_copy]; +"1584 slice_97" [id=1584, type=slice]; +"1585 slice_98" [id=1585, type=slice]; +"1586 fill__25" [id=1586, type=fill_]; +"1587 lift_fresh_copy_26" [id=1587, type=lift_fresh_copy]; +"1588 slice_99" [id=1588, type=slice]; +"1589 slice_100" [id=1589, type=slice]; +"1590 fill__26" [id=1590, type=fill_]; +"1591 view_29" [id=1591, type=view]; +"1592 permute_26" [id=1592, type=permute]; +"1593 reshape_24" [id=1593, type=reshape]; +"1594 unsqueeze_14" [id=1594, type=unsqueeze]; +"1595 unsqueeze_15" [id=1595, type=unsqueeze]; +"1596 sub_2" [id=1596, type=sub]; +"1597 ne_2" [id=1597, type=ne]; +"1598 masked_fill_4" [id=1598, type=masked_fill]; +"1599 eq_2" [id=1599, type=eq]; +"1600 masked_fill_5" [id=1600, type=masked_fill]; +"1601 view_30" [id=1601, type=view]; +"1602 unsqueeze_16" [id=1602, type=unsqueeze]; +"1603 unsqueeze_17" [id=1603, type=unsqueeze]; +"1604 add_18" [id=1604, type=add]; +"1605 view_31" [id=1605, type=view]; +"1606 softmax_5" [id=1606, type=softmax]; +"1607 dropout_20" [id=1607, type=dropout]; +"1608 matmul_11" [id=1608, type=matmul]; +"1609 quantize_per_tensor_default_36" [id=1609, type=quantize_per_tensor]; +"1610 dequantize_per_tensor_default_36" [id=1610, type=dequantize_per_tensor]; +"1611 transpose_11" [id=1611, type=transpose]; +"1612 reshape_25" [id=1612, type=reshape]; +"1613 linear_35" [id=1613, type=linear]; +"1614 dropout_21" [id=1614, type=dropout]; +"1615 view_32" [id=1615, type=view]; +"1616 permute_27" [id=1616, type=permute]; +"1617 reshape_26" [id=1617, type=reshape]; +"1618 roll_5" [id=1618, type=roll]; +"1619 slice_101" [id=1619, type=slice]; +"1620 slice_102" [id=1620, type=slice]; +"1621 slice_103" [id=1621, type=slice]; +"1622 slice_104" [id=1622, type=slice]; +"1623 contiguous_9" [id=1623, type=contiguous]; +"1624 layer_norm_13" [id=1624, type=layer_norm]; +"1625 add_19" [id=1625, type=add]; +"1626 quantize_per_tensor_default_37" [id=1626, type=quantize_per_tensor]; +"1627 dequantize_per_tensor_default_37" [id=1627, type=dequantize_per_tensor]; +"1628 linear_36" [id=1628, type=linear]; +"1629 gelu_5" [id=1629, type=gelu]; +"1630 quantize_per_tensor_default_38" [id=1630, type=quantize_per_tensor]; +"1631 dequantize_per_tensor_default_38" [id=1631, type=dequantize_per_tensor]; +"1632 dropout_22" [id=1632, type=dropout]; +"1633 linear_37" [id=1633, type=linear]; +"1634 dropout_23" [id=1634, type=dropout]; +"1635 layer_norm_14" [id=1635, type=layer_norm]; +"1636 add_20" [id=1636, type=add]; +"1637 quantize_per_tensor_default_39" [id=1637, type=quantize_per_tensor]; +"1638 dequantize_per_tensor_default_39" [id=1638, type=dequantize_per_tensor]; +"1639 linear_38" [id=1639, type=linear]; +"1640 relu__6" [id=1640, type=relu_]; +"1641 linear_39" [id=1641, type=linear]; +"1642 view_33" [id=1642, type=view]; +"1643 index_6" [id=1643, type=index]; +"1644 view_34" [id=1644, type=view]; +"1645 permute_28" [id=1645, type=permute]; +"1646 contiguous_10" [id=1646, type=contiguous]; +"1647 unsqueeze_18" [id=1647, type=unsqueeze]; +"1648 sigmoid_6" [id=1648, type=sigmoid]; +"1649 mul_12" [id=1649, type=mul]; +"1650 pad_8" [id=1650, type=pad]; +"1651 view_35" [id=1651, type=view]; +"1652 permute_29" [id=1652, type=permute]; +"1653 reshape_27" [id=1653, type=reshape]; +"1654 clone_6" [id=1654, type=clone]; +"1655 slice_105" [id=1655, type=slice]; +"1656 zero__6" [id=1656, type=zero_]; +"1657 linear_40" [id=1657, type=linear]; +"1658 reshape_28" [id=1658, type=reshape]; +"1659 permute_30" [id=1659, type=permute]; +"1660 select_18" [id=1660, type=select]; +"1661 select_19" [id=1661, type=select]; +"1662 select_20" [id=1662, type=select]; +"1663 linalg_vector_norm_12" [id=1663, type=linalg_vector_norm]; +"1664 clamp_min_12" [id=1664, type=clamp_min]; +"1665 expand_as_12" [id=1665, type=expand_as]; +"1666 div_12" [id=1666, type=div]; +"1667 quantize_per_tensor_default_40" [id=1667, type=quantize_per_tensor]; +"1668 dequantize_per_tensor_default_40" [id=1668, type=dequantize_per_tensor]; +"1669 linalg_vector_norm_13" [id=1669, type=linalg_vector_norm]; +"1670 clamp_min_13" [id=1670, type=clamp_min]; +"1671 expand_as_13" [id=1671, type=expand_as]; +"1672 div_13" [id=1672, type=div]; +"1673 quantize_per_tensor_default_41" [id=1673, type=quantize_per_tensor]; +"1674 dequantize_per_tensor_default_41" [id=1674, type=dequantize_per_tensor]; +"1675 transpose_12" [id=1675, type=transpose]; +"1676 matmul_12" [id=1676, type=matmul]; +"1677 clamp_6" [id=1677, type=clamp]; +"1678 exp_6" [id=1678, type=exp]; +"1679 mul_13" [id=1679, type=mul]; +"1680 add_21" [id=1680, type=add]; +"1681 softmax_6" [id=1681, type=softmax]; +"1682 dropout_24" [id=1682, type=dropout]; +"1683 matmul_13" [id=1683, type=matmul]; +"1684 quantize_per_tensor_default_42" [id=1684, type=quantize_per_tensor]; +"1685 dequantize_per_tensor_default_42" [id=1685, type=dequantize_per_tensor]; +"1686 transpose_13" [id=1686, type=transpose]; +"1687 reshape_29" [id=1687, type=reshape]; +"1688 linear_41" [id=1688, type=linear]; +"1689 dropout_25" [id=1689, type=dropout]; +"1690 view_36" [id=1690, type=view]; +"1691 permute_31" [id=1691, type=permute]; +"1692 reshape_30" [id=1692, type=reshape]; +"1693 slice_106" [id=1693, type=slice]; +"1694 slice_107" [id=1694, type=slice]; +"1695 slice_108" [id=1695, type=slice]; +"1696 slice_109" [id=1696, type=slice]; +"1697 contiguous_11" [id=1697, type=contiguous]; +"1698 layer_norm_15" [id=1698, type=layer_norm]; +"1699 add_22" [id=1699, type=add]; +"1700 quantize_per_tensor_default_43" [id=1700, type=quantize_per_tensor]; +"1701 dequantize_per_tensor_default_43" [id=1701, type=dequantize_per_tensor]; +"1702 linear_42" [id=1702, type=linear]; +"1703 gelu_6" [id=1703, type=gelu]; +"1704 quantize_per_tensor_default_44" [id=1704, type=quantize_per_tensor]; +"1705 dequantize_per_tensor_default_44" [id=1705, type=dequantize_per_tensor]; +"1706 dropout_26" [id=1706, type=dropout]; +"1707 linear_43" [id=1707, type=linear]; +"1708 dropout_27" [id=1708, type=dropout]; +"1709 layer_norm_16" [id=1709, type=layer_norm]; +"1710 add_23" [id=1710, type=add]; +"1711 linear_44" [id=1711, type=linear]; +"1712 relu__7" [id=1712, type=relu_]; +"1713 linear_45" [id=1713, type=linear]; +"1714 view_37" [id=1714, type=view]; +"1715 index_7" [id=1715, type=index]; +"1716 view_38" [id=1716, type=view]; +"1717 permute_32" [id=1717, type=permute]; +"1718 contiguous_12" [id=1718, type=contiguous]; +"1719 unsqueeze_19" [id=1719, type=unsqueeze]; +"1720 sigmoid_7" [id=1720, type=sigmoid]; +"1721 mul_14" [id=1721, type=mul]; +"1722 pad_9" [id=1722, type=pad]; +"1723 roll_6" [id=1723, type=roll]; +"1724 view_39" [id=1724, type=view]; +"1725 permute_33" [id=1725, type=permute]; +"1726 reshape_31" [id=1726, type=reshape]; +"1727 quantize_per_tensor_default_45" [id=1727, type=quantize_per_tensor]; +"1728 dequantize_per_tensor_default_45" [id=1728, type=dequantize_per_tensor]; +"1729 clone_7" [id=1729, type=clone]; +"1730 slice_110" [id=1730, type=slice]; +"1731 zero__7" [id=1731, type=zero_]; +"1732 linear_46" [id=1732, type=linear]; +"1733 reshape_32" [id=1733, type=reshape]; +"1734 permute_34" [id=1734, type=permute]; +"1735 select_21" [id=1735, type=select]; +"1736 select_22" [id=1736, type=select]; +"1737 select_23" [id=1737, type=select]; +"1738 linalg_vector_norm_14" [id=1738, type=linalg_vector_norm]; +"1739 clamp_min_14" [id=1739, type=clamp_min]; +"1740 expand_as_14" [id=1740, type=expand_as]; +"1741 div_14" [id=1741, type=div]; +"1742 quantize_per_tensor_default_46" [id=1742, type=quantize_per_tensor]; +"1743 dequantize_per_tensor_default_46" [id=1743, type=dequantize_per_tensor]; +"1744 linalg_vector_norm_15" [id=1744, type=linalg_vector_norm]; +"1745 clamp_min_15" [id=1745, type=clamp_min]; +"1746 expand_as_15" [id=1746, type=expand_as]; +"1747 div_15" [id=1747, type=div]; +"1748 quantize_per_tensor_default_47" [id=1748, type=quantize_per_tensor]; +"1749 dequantize_per_tensor_default_47" [id=1749, type=dequantize_per_tensor]; +"1750 transpose_14" [id=1750, type=transpose]; +"1751 matmul_14" [id=1751, type=matmul]; +"1752 clamp_7" [id=1752, type=clamp]; +"1753 exp_7" [id=1753, type=exp]; +"1754 mul_15" [id=1754, type=mul]; +"1755 add_24" [id=1755, type=add]; +"1756 new_zeros_3" [id=1756, type=new_zeros]; +"1757 lift_fresh_copy_27" [id=1757, type=lift_fresh_copy]; +"1758 slice_111" [id=1758, type=slice]; +"1759 slice_112" [id=1759, type=slice]; +"1760 fill__27" [id=1760, type=fill_]; +"1761 lift_fresh_copy_28" [id=1761, type=lift_fresh_copy]; +"1762 slice_113" [id=1762, type=slice]; +"1763 slice_114" [id=1763, type=slice]; +"1764 fill__28" [id=1764, type=fill_]; +"1765 lift_fresh_copy_29" [id=1765, type=lift_fresh_copy]; +"1766 slice_115" [id=1766, type=slice]; +"1767 slice_116" [id=1767, type=slice]; +"1768 fill__29" [id=1768, type=fill_]; +"1769 lift_fresh_copy_30" [id=1769, type=lift_fresh_copy]; +"1770 slice_117" [id=1770, type=slice]; +"1771 slice_118" [id=1771, type=slice]; +"1772 fill__30" [id=1772, type=fill_]; +"1773 lift_fresh_copy_31" [id=1773, type=lift_fresh_copy]; +"1774 slice_119" [id=1774, type=slice]; +"1775 slice_120" [id=1775, type=slice]; +"1776 fill__31" [id=1776, type=fill_]; +"1777 lift_fresh_copy_32" [id=1777, type=lift_fresh_copy]; +"1778 slice_121" [id=1778, type=slice]; +"1779 slice_122" [id=1779, type=slice]; +"1780 fill__32" [id=1780, type=fill_]; +"1781 lift_fresh_copy_33" [id=1781, type=lift_fresh_copy]; +"1782 slice_123" [id=1782, type=slice]; +"1783 slice_124" [id=1783, type=slice]; +"1784 fill__33" [id=1784, type=fill_]; +"1785 lift_fresh_copy_34" [id=1785, type=lift_fresh_copy]; +"1786 slice_125" [id=1786, type=slice]; +"1787 slice_126" [id=1787, type=slice]; +"1788 fill__34" [id=1788, type=fill_]; +"1789 lift_fresh_copy_35" [id=1789, type=lift_fresh_copy]; +"1790 slice_127" [id=1790, type=slice]; +"1791 slice_128" [id=1791, type=slice]; +"1792 fill__35" [id=1792, type=fill_]; +"1793 view_40" [id=1793, type=view]; +"1794 permute_35" [id=1794, type=permute]; +"1795 reshape_33" [id=1795, type=reshape]; +"1796 unsqueeze_20" [id=1796, type=unsqueeze]; +"1797 unsqueeze_21" [id=1797, type=unsqueeze]; +"1798 sub_3" [id=1798, type=sub]; +"1799 ne_3" [id=1799, type=ne]; +"1800 masked_fill_6" [id=1800, type=masked_fill]; +"1801 eq_3" [id=1801, type=eq]; +"1802 masked_fill_7" [id=1802, type=masked_fill]; +"1803 view_41" [id=1803, type=view]; +"1804 unsqueeze_22" [id=1804, type=unsqueeze]; +"1805 unsqueeze_23" [id=1805, type=unsqueeze]; +"1806 add_25" [id=1806, type=add]; +"1807 view_42" [id=1807, type=view]; +"1808 softmax_7" [id=1808, type=softmax]; +"1809 dropout_28" [id=1809, type=dropout]; +"1810 matmul_15" [id=1810, type=matmul]; +"1811 quantize_per_tensor_default_48" [id=1811, type=quantize_per_tensor]; +"1812 dequantize_per_tensor_default_48" [id=1812, type=dequantize_per_tensor]; +"1813 transpose_15" [id=1813, type=transpose]; +"1814 reshape_34" [id=1814, type=reshape]; +"1815 linear_47" [id=1815, type=linear]; +"1816 dropout_29" [id=1816, type=dropout]; +"1817 view_43" [id=1817, type=view]; +"1818 permute_36" [id=1818, type=permute]; +"1819 reshape_35" [id=1819, type=reshape]; +"1820 roll_7" [id=1820, type=roll]; +"1821 slice_129" [id=1821, type=slice]; +"1822 slice_130" [id=1822, type=slice]; +"1823 slice_131" [id=1823, type=slice]; +"1824 slice_132" [id=1824, type=slice]; +"1825 contiguous_13" [id=1825, type=contiguous]; +"1826 layer_norm_17" [id=1826, type=layer_norm]; +"1827 add_26" [id=1827, type=add]; +"1828 quantize_per_tensor_default_49" [id=1828, type=quantize_per_tensor]; +"1829 dequantize_per_tensor_default_49" [id=1829, type=dequantize_per_tensor]; +"1830 linear_48" [id=1830, type=linear]; +"1831 gelu_7" [id=1831, type=gelu]; +"1832 quantize_per_tensor_default_50" [id=1832, type=quantize_per_tensor]; +"1833 dequantize_per_tensor_default_50" [id=1833, type=dequantize_per_tensor]; +"1834 dropout_30" [id=1834, type=dropout]; +"1835 linear_49" [id=1835, type=linear]; +"1836 dropout_31" [id=1836, type=dropout]; +"1837 layer_norm_18" [id=1837, type=layer_norm]; +"1838 add_27" [id=1838, type=add]; +"1839 quantize_per_tensor_default_51" [id=1839, type=quantize_per_tensor]; +"1840 dequantize_per_tensor_default_51" [id=1840, type=dequantize_per_tensor]; +"1841 linear_50" [id=1841, type=linear]; +"1842 relu__8" [id=1842, type=relu_]; +"1843 linear_51" [id=1843, type=linear]; +"1844 view_44" [id=1844, type=view]; +"1845 index_8" [id=1845, type=index]; +"1846 view_45" [id=1846, type=view]; +"1847 permute_37" [id=1847, type=permute]; +"1848 contiguous_14" [id=1848, type=contiguous]; +"1849 unsqueeze_24" [id=1849, type=unsqueeze]; +"1850 sigmoid_8" [id=1850, type=sigmoid]; +"1851 mul_16" [id=1851, type=mul]; +"1852 pad_10" [id=1852, type=pad]; +"1853 view_46" [id=1853, type=view]; +"1854 permute_38" [id=1854, type=permute]; +"1855 reshape_36" [id=1855, type=reshape]; +"1856 clone_8" [id=1856, type=clone]; +"1857 slice_133" [id=1857, type=slice]; +"1858 zero__8" [id=1858, type=zero_]; +"1859 linear_52" [id=1859, type=linear]; +"1860 reshape_37" [id=1860, type=reshape]; +"1861 permute_39" [id=1861, type=permute]; +"1862 select_24" [id=1862, type=select]; +"1863 select_25" [id=1863, type=select]; +"1864 select_26" [id=1864, type=select]; +"1865 linalg_vector_norm_16" [id=1865, type=linalg_vector_norm]; +"1866 clamp_min_16" [id=1866, type=clamp_min]; +"1867 expand_as_16" [id=1867, type=expand_as]; +"1868 div_16" [id=1868, type=div]; +"1869 quantize_per_tensor_default_52" [id=1869, type=quantize_per_tensor]; +"1870 dequantize_per_tensor_default_52" [id=1870, type=dequantize_per_tensor]; +"1871 linalg_vector_norm_17" [id=1871, type=linalg_vector_norm]; +"1872 clamp_min_17" [id=1872, type=clamp_min]; +"1873 expand_as_17" [id=1873, type=expand_as]; +"1874 div_17" [id=1874, type=div]; +"1875 quantize_per_tensor_default_53" [id=1875, type=quantize_per_tensor]; +"1876 dequantize_per_tensor_default_53" [id=1876, type=dequantize_per_tensor]; +"1877 transpose_16" [id=1877, type=transpose]; +"1878 matmul_16" [id=1878, type=matmul]; +"1879 clamp_8" [id=1879, type=clamp]; +"1880 exp_8" [id=1880, type=exp]; +"1881 mul_17" [id=1881, type=mul]; +"1882 add_28" [id=1882, type=add]; +"1883 softmax_8" [id=1883, type=softmax]; +"1884 dropout_32" [id=1884, type=dropout]; +"1885 matmul_17" [id=1885, type=matmul]; +"1886 quantize_per_tensor_default_54" [id=1886, type=quantize_per_tensor]; +"1887 dequantize_per_tensor_default_54" [id=1887, type=dequantize_per_tensor]; +"1888 transpose_17" [id=1888, type=transpose]; +"1889 reshape_38" [id=1889, type=reshape]; +"1890 linear_53" [id=1890, type=linear]; +"1891 dropout_33" [id=1891, type=dropout]; +"1892 view_47" [id=1892, type=view]; +"1893 permute_40" [id=1893, type=permute]; +"1894 reshape_39" [id=1894, type=reshape]; +"1895 slice_134" [id=1895, type=slice]; +"1896 slice_135" [id=1896, type=slice]; +"1897 slice_136" [id=1897, type=slice]; +"1898 slice_137" [id=1898, type=slice]; +"1899 contiguous_15" [id=1899, type=contiguous]; +"1900 layer_norm_19" [id=1900, type=layer_norm]; +"1901 add_29" [id=1901, type=add]; +"1902 quantize_per_tensor_default_55" [id=1902, type=quantize_per_tensor]; +"1903 dequantize_per_tensor_default_55" [id=1903, type=dequantize_per_tensor]; +"1904 linear_54" [id=1904, type=linear]; +"1905 gelu_8" [id=1905, type=gelu]; +"1906 quantize_per_tensor_default_56" [id=1906, type=quantize_per_tensor]; +"1907 dequantize_per_tensor_default_56" [id=1907, type=dequantize_per_tensor]; +"1908 dropout_34" [id=1908, type=dropout]; +"1909 linear_55" [id=1909, type=linear]; +"1910 dropout_35" [id=1910, type=dropout]; +"1911 layer_norm_20" [id=1911, type=layer_norm]; +"1912 add_30" [id=1912, type=add]; +"1913 linear_56" [id=1913, type=linear]; +"1914 relu__9" [id=1914, type=relu_]; +"1915 linear_57" [id=1915, type=linear]; +"1916 view_48" [id=1916, type=view]; +"1917 index_9" [id=1917, type=index]; +"1918 view_49" [id=1918, type=view]; +"1919 permute_41" [id=1919, type=permute]; +"1920 contiguous_16" [id=1920, type=contiguous]; +"1921 unsqueeze_25" [id=1921, type=unsqueeze]; +"1922 sigmoid_9" [id=1922, type=sigmoid]; +"1923 mul_18" [id=1923, type=mul]; +"1924 pad_11" [id=1924, type=pad]; +"1925 roll_8" [id=1925, type=roll]; +"1926 view_50" [id=1926, type=view]; +"1927 permute_42" [id=1927, type=permute]; +"1928 reshape_40" [id=1928, type=reshape]; +"1929 quantize_per_tensor_default_57" [id=1929, type=quantize_per_tensor]; +"1930 dequantize_per_tensor_default_57" [id=1930, type=dequantize_per_tensor]; +"1931 clone_9" [id=1931, type=clone]; +"1932 slice_138" [id=1932, type=slice]; +"1933 zero__9" [id=1933, type=zero_]; +"1934 linear_58" [id=1934, type=linear]; +"1935 reshape_41" [id=1935, type=reshape]; +"1936 permute_43" [id=1936, type=permute]; +"1937 select_27" [id=1937, type=select]; +"1938 select_28" [id=1938, type=select]; +"1939 select_29" [id=1939, type=select]; +"1940 linalg_vector_norm_18" [id=1940, type=linalg_vector_norm]; +"1941 clamp_min_18" [id=1941, type=clamp_min]; +"1942 expand_as_18" [id=1942, type=expand_as]; +"1943 div_18" [id=1943, type=div]; +"1944 quantize_per_tensor_default_58" [id=1944, type=quantize_per_tensor]; +"1945 dequantize_per_tensor_default_58" [id=1945, type=dequantize_per_tensor]; +"1946 linalg_vector_norm_19" [id=1946, type=linalg_vector_norm]; +"1947 clamp_min_19" [id=1947, type=clamp_min]; +"1948 expand_as_19" [id=1948, type=expand_as]; +"1949 div_19" [id=1949, type=div]; +"1950 quantize_per_tensor_default_59" [id=1950, type=quantize_per_tensor]; +"1951 dequantize_per_tensor_default_59" [id=1951, type=dequantize_per_tensor]; +"1952 transpose_18" [id=1952, type=transpose]; +"1953 matmul_18" [id=1953, type=matmul]; +"1954 clamp_9" [id=1954, type=clamp]; +"1955 exp_9" [id=1955, type=exp]; +"1956 mul_19" [id=1956, type=mul]; +"1957 add_31" [id=1957, type=add]; +"1958 new_zeros_4" [id=1958, type=new_zeros]; +"1959 lift_fresh_copy_36" [id=1959, type=lift_fresh_copy]; +"1960 slice_139" [id=1960, type=slice]; +"1961 slice_140" [id=1961, type=slice]; +"1962 fill__36" [id=1962, type=fill_]; +"1963 lift_fresh_copy_37" [id=1963, type=lift_fresh_copy]; +"1964 slice_141" [id=1964, type=slice]; +"1965 slice_142" [id=1965, type=slice]; +"1966 fill__37" [id=1966, type=fill_]; +"1967 lift_fresh_copy_38" [id=1967, type=lift_fresh_copy]; +"1968 slice_143" [id=1968, type=slice]; +"1969 slice_144" [id=1969, type=slice]; +"1970 fill__38" [id=1970, type=fill_]; +"1971 lift_fresh_copy_39" [id=1971, type=lift_fresh_copy]; +"1972 slice_145" [id=1972, type=slice]; +"1973 slice_146" [id=1973, type=slice]; +"1974 fill__39" [id=1974, type=fill_]; +"1975 lift_fresh_copy_40" [id=1975, type=lift_fresh_copy]; +"1976 slice_147" [id=1976, type=slice]; +"1977 slice_148" [id=1977, type=slice]; +"1978 fill__40" [id=1978, type=fill_]; +"1979 lift_fresh_copy_41" [id=1979, type=lift_fresh_copy]; +"1980 slice_149" [id=1980, type=slice]; +"1981 slice_150" [id=1981, type=slice]; +"1982 fill__41" [id=1982, type=fill_]; +"1983 lift_fresh_copy_42" [id=1983, type=lift_fresh_copy]; +"1984 slice_151" [id=1984, type=slice]; +"1985 slice_152" [id=1985, type=slice]; +"1986 fill__42" [id=1986, type=fill_]; +"1987 lift_fresh_copy_43" [id=1987, type=lift_fresh_copy]; +"1988 slice_153" [id=1988, type=slice]; +"1989 slice_154" [id=1989, type=slice]; +"1990 fill__43" [id=1990, type=fill_]; +"1991 lift_fresh_copy_44" [id=1991, type=lift_fresh_copy]; +"1992 slice_155" [id=1992, type=slice]; +"1993 slice_156" [id=1993, type=slice]; +"1994 fill__44" [id=1994, type=fill_]; +"1995 view_51" [id=1995, type=view]; +"1996 permute_44" [id=1996, type=permute]; +"1997 reshape_42" [id=1997, type=reshape]; +"1998 unsqueeze_26" [id=1998, type=unsqueeze]; +"1999 unsqueeze_27" [id=1999, type=unsqueeze]; +"2000 sub_4" [id=2000, type=sub]; +"2001 ne_4" [id=2001, type=ne]; +"2002 masked_fill_8" [id=2002, type=masked_fill]; +"2003 eq_4" [id=2003, type=eq]; +"2004 masked_fill_9" [id=2004, type=masked_fill]; +"2005 view_52" [id=2005, type=view]; +"2006 unsqueeze_28" [id=2006, type=unsqueeze]; +"2007 unsqueeze_29" [id=2007, type=unsqueeze]; +"2008 add_32" [id=2008, type=add]; +"2009 view_53" [id=2009, type=view]; +"2010 softmax_9" [id=2010, type=softmax]; +"2011 dropout_36" [id=2011, type=dropout]; +"2012 matmul_19" [id=2012, type=matmul]; +"2013 quantize_per_tensor_default_60" [id=2013, type=quantize_per_tensor]; +"2014 dequantize_per_tensor_default_60" [id=2014, type=dequantize_per_tensor]; +"2015 transpose_19" [id=2015, type=transpose]; +"2016 reshape_43" [id=2016, type=reshape]; +"2017 linear_59" [id=2017, type=linear]; +"2018 dropout_37" [id=2018, type=dropout]; +"2019 view_54" [id=2019, type=view]; +"2020 permute_45" [id=2020, type=permute]; +"2021 reshape_44" [id=2021, type=reshape]; +"2022 roll_9" [id=2022, type=roll]; +"2023 slice_157" [id=2023, type=slice]; +"2024 slice_158" [id=2024, type=slice]; +"2025 slice_159" [id=2025, type=slice]; +"2026 slice_160" [id=2026, type=slice]; +"2027 contiguous_17" [id=2027, type=contiguous]; +"2028 layer_norm_21" [id=2028, type=layer_norm]; +"2029 add_33" [id=2029, type=add]; +"2030 quantize_per_tensor_default_61" [id=2030, type=quantize_per_tensor]; +"2031 dequantize_per_tensor_default_61" [id=2031, type=dequantize_per_tensor]; +"2032 linear_60" [id=2032, type=linear]; +"2033 gelu_9" [id=2033, type=gelu]; +"2034 quantize_per_tensor_default_62" [id=2034, type=quantize_per_tensor]; +"2035 dequantize_per_tensor_default_62" [id=2035, type=dequantize_per_tensor]; +"2036 dropout_38" [id=2036, type=dropout]; +"2037 linear_61" [id=2037, type=linear]; +"2038 dropout_39" [id=2038, type=dropout]; +"2039 layer_norm_22" [id=2039, type=layer_norm]; +"2040 add_34" [id=2040, type=add]; +"2041 quantize_per_tensor_default_63" [id=2041, type=quantize_per_tensor]; +"2042 dequantize_per_tensor_default_63" [id=2042, type=dequantize_per_tensor]; +"2043 linear_62" [id=2043, type=linear]; +"2044 relu__10" [id=2044, type=relu_]; +"2045 linear_63" [id=2045, type=linear]; +"2046 view_55" [id=2046, type=view]; +"2047 index_10" [id=2047, type=index]; +"2048 view_56" [id=2048, type=view]; +"2049 permute_46" [id=2049, type=permute]; +"2050 contiguous_18" [id=2050, type=contiguous]; +"2051 unsqueeze_30" [id=2051, type=unsqueeze]; +"2052 sigmoid_10" [id=2052, type=sigmoid]; +"2053 mul_20" [id=2053, type=mul]; +"2054 pad_12" [id=2054, type=pad]; +"2055 view_57" [id=2055, type=view]; +"2056 permute_47" [id=2056, type=permute]; +"2057 reshape_45" [id=2057, type=reshape]; +"2058 clone_10" [id=2058, type=clone]; +"2059 slice_161" [id=2059, type=slice]; +"2060 zero__10" [id=2060, type=zero_]; +"2061 linear_64" [id=2061, type=linear]; +"2062 reshape_46" [id=2062, type=reshape]; +"2063 permute_48" [id=2063, type=permute]; +"2064 select_30" [id=2064, type=select]; +"2065 select_31" [id=2065, type=select]; +"2066 select_32" [id=2066, type=select]; +"2067 linalg_vector_norm_20" [id=2067, type=linalg_vector_norm]; +"2068 clamp_min_20" [id=2068, type=clamp_min]; +"2069 expand_as_20" [id=2069, type=expand_as]; +"2070 div_20" [id=2070, type=div]; +"2071 quantize_per_tensor_default_64" [id=2071, type=quantize_per_tensor]; +"2072 dequantize_per_tensor_default_64" [id=2072, type=dequantize_per_tensor]; +"2073 linalg_vector_norm_21" [id=2073, type=linalg_vector_norm]; +"2074 clamp_min_21" [id=2074, type=clamp_min]; +"2075 expand_as_21" [id=2075, type=expand_as]; +"2076 div_21" [id=2076, type=div]; +"2077 quantize_per_tensor_default_65" [id=2077, type=quantize_per_tensor]; +"2078 dequantize_per_tensor_default_65" [id=2078, type=dequantize_per_tensor]; +"2079 transpose_20" [id=2079, type=transpose]; +"2080 matmul_20" [id=2080, type=matmul]; +"2081 clamp_10" [id=2081, type=clamp]; +"2082 exp_10" [id=2082, type=exp]; +"2083 mul_21" [id=2083, type=mul]; +"2084 add_35" [id=2084, type=add]; +"2085 softmax_10" [id=2085, type=softmax]; +"2086 dropout_40" [id=2086, type=dropout]; +"2087 matmul_21" [id=2087, type=matmul]; +"2088 quantize_per_tensor_default_66" [id=2088, type=quantize_per_tensor]; +"2089 dequantize_per_tensor_default_66" [id=2089, type=dequantize_per_tensor]; +"2090 transpose_21" [id=2090, type=transpose]; +"2091 reshape_47" [id=2091, type=reshape]; +"2092 linear_65" [id=2092, type=linear]; +"2093 dropout_41" [id=2093, type=dropout]; +"2094 view_58" [id=2094, type=view]; +"2095 permute_49" [id=2095, type=permute]; +"2096 reshape_48" [id=2096, type=reshape]; +"2097 slice_162" [id=2097, type=slice]; +"2098 slice_163" [id=2098, type=slice]; +"2099 slice_164" [id=2099, type=slice]; +"2100 slice_165" [id=2100, type=slice]; +"2101 contiguous_19" [id=2101, type=contiguous]; +"2102 layer_norm_23" [id=2102, type=layer_norm]; +"2103 add_36" [id=2103, type=add]; +"2104 quantize_per_tensor_default_67" [id=2104, type=quantize_per_tensor]; +"2105 dequantize_per_tensor_default_67" [id=2105, type=dequantize_per_tensor]; +"2106 linear_66" [id=2106, type=linear]; +"2107 gelu_10" [id=2107, type=gelu]; +"2108 quantize_per_tensor_default_68" [id=2108, type=quantize_per_tensor]; +"2109 dequantize_per_tensor_default_68" [id=2109, type=dequantize_per_tensor]; +"2110 dropout_42" [id=2110, type=dropout]; +"2111 linear_67" [id=2111, type=linear]; +"2112 dropout_43" [id=2112, type=dropout]; +"2113 layer_norm_24" [id=2113, type=layer_norm]; +"2114 add_37" [id=2114, type=add]; +"2115 linear_68" [id=2115, type=linear]; +"2116 relu__11" [id=2116, type=relu_]; +"2117 linear_69" [id=2117, type=linear]; +"2118 view_59" [id=2118, type=view]; +"2119 index_11" [id=2119, type=index]; +"2120 view_60" [id=2120, type=view]; +"2121 permute_50" [id=2121, type=permute]; +"2122 contiguous_20" [id=2122, type=contiguous]; +"2123 unsqueeze_31" [id=2123, type=unsqueeze]; +"2124 sigmoid_11" [id=2124, type=sigmoid]; +"2125 mul_22" [id=2125, type=mul]; +"2126 pad_13" [id=2126, type=pad]; +"2127 roll_10" [id=2127, type=roll]; +"2128 view_61" [id=2128, type=view]; +"2129 permute_51" [id=2129, type=permute]; +"2130 reshape_49" [id=2130, type=reshape]; +"2131 quantize_per_tensor_default_69" [id=2131, type=quantize_per_tensor]; +"2132 dequantize_per_tensor_default_69" [id=2132, type=dequantize_per_tensor]; +"2133 clone_11" [id=2133, type=clone]; +"2134 slice_166" [id=2134, type=slice]; +"2135 zero__11" [id=2135, type=zero_]; +"2136 linear_70" [id=2136, type=linear]; +"2137 reshape_50" [id=2137, type=reshape]; +"2138 permute_52" [id=2138, type=permute]; +"2139 select_33" [id=2139, type=select]; +"2140 select_34" [id=2140, type=select]; +"2141 select_35" [id=2141, type=select]; +"2142 linalg_vector_norm_22" [id=2142, type=linalg_vector_norm]; +"2143 clamp_min_22" [id=2143, type=clamp_min]; +"2144 expand_as_22" [id=2144, type=expand_as]; +"2145 div_22" [id=2145, type=div]; +"2146 quantize_per_tensor_default_70" [id=2146, type=quantize_per_tensor]; +"2147 dequantize_per_tensor_default_70" [id=2147, type=dequantize_per_tensor]; +"2148 linalg_vector_norm_23" [id=2148, type=linalg_vector_norm]; +"2149 clamp_min_23" [id=2149, type=clamp_min]; +"2150 expand_as_23" [id=2150, type=expand_as]; +"2151 div_23" [id=2151, type=div]; +"2152 quantize_per_tensor_default_71" [id=2152, type=quantize_per_tensor]; +"2153 dequantize_per_tensor_default_71" [id=2153, type=dequantize_per_tensor]; +"2154 transpose_22" [id=2154, type=transpose]; +"2155 matmul_22" [id=2155, type=matmul]; +"2156 clamp_11" [id=2156, type=clamp]; +"2157 exp_11" [id=2157, type=exp]; +"2158 mul_23" [id=2158, type=mul]; +"2159 add_38" [id=2159, type=add]; +"2160 new_zeros_5" [id=2160, type=new_zeros]; +"2161 lift_fresh_copy_45" [id=2161, type=lift_fresh_copy]; +"2162 slice_167" [id=2162, type=slice]; +"2163 slice_168" [id=2163, type=slice]; +"2164 fill__45" [id=2164, type=fill_]; +"2165 lift_fresh_copy_46" [id=2165, type=lift_fresh_copy]; +"2166 slice_169" [id=2166, type=slice]; +"2167 slice_170" [id=2167, type=slice]; +"2168 fill__46" [id=2168, type=fill_]; +"2169 lift_fresh_copy_47" [id=2169, type=lift_fresh_copy]; +"2170 slice_171" [id=2170, type=slice]; +"2171 slice_172" [id=2171, type=slice]; +"2172 fill__47" [id=2172, type=fill_]; +"2173 lift_fresh_copy_48" [id=2173, type=lift_fresh_copy]; +"2174 slice_173" [id=2174, type=slice]; +"2175 slice_174" [id=2175, type=slice]; +"2176 fill__48" [id=2176, type=fill_]; +"2177 lift_fresh_copy_49" [id=2177, type=lift_fresh_copy]; +"2178 slice_175" [id=2178, type=slice]; +"2179 slice_176" [id=2179, type=slice]; +"2180 fill__49" [id=2180, type=fill_]; +"2181 lift_fresh_copy_50" [id=2181, type=lift_fresh_copy]; +"2182 slice_177" [id=2182, type=slice]; +"2183 slice_178" [id=2183, type=slice]; +"2184 fill__50" [id=2184, type=fill_]; +"2185 lift_fresh_copy_51" [id=2185, type=lift_fresh_copy]; +"2186 slice_179" [id=2186, type=slice]; +"2187 slice_180" [id=2187, type=slice]; +"2188 fill__51" [id=2188, type=fill_]; +"2189 lift_fresh_copy_52" [id=2189, type=lift_fresh_copy]; +"2190 slice_181" [id=2190, type=slice]; +"2191 slice_182" [id=2191, type=slice]; +"2192 fill__52" [id=2192, type=fill_]; +"2193 lift_fresh_copy_53" [id=2193, type=lift_fresh_copy]; +"2194 slice_183" [id=2194, type=slice]; +"2195 slice_184" [id=2195, type=slice]; +"2196 fill__53" [id=2196, type=fill_]; +"2197 view_62" [id=2197, type=view]; +"2198 permute_53" [id=2198, type=permute]; +"2199 reshape_51" [id=2199, type=reshape]; +"2200 unsqueeze_32" [id=2200, type=unsqueeze]; +"2201 unsqueeze_33" [id=2201, type=unsqueeze]; +"2202 sub_5" [id=2202, type=sub]; +"2203 ne_5" [id=2203, type=ne]; +"2204 masked_fill_10" [id=2204, type=masked_fill]; +"2205 eq_5" [id=2205, type=eq]; +"2206 masked_fill_11" [id=2206, type=masked_fill]; +"2207 view_63" [id=2207, type=view]; +"2208 unsqueeze_34" [id=2208, type=unsqueeze]; +"2209 unsqueeze_35" [id=2209, type=unsqueeze]; +"2210 add_39" [id=2210, type=add]; +"2211 view_64" [id=2211, type=view]; +"2212 softmax_11" [id=2212, type=softmax]; +"2213 dropout_44" [id=2213, type=dropout]; +"2214 matmul_23" [id=2214, type=matmul]; +"2215 quantize_per_tensor_default_72" [id=2215, type=quantize_per_tensor]; +"2216 dequantize_per_tensor_default_72" [id=2216, type=dequantize_per_tensor]; +"2217 transpose_23" [id=2217, type=transpose]; +"2218 reshape_52" [id=2218, type=reshape]; +"2219 linear_71" [id=2219, type=linear]; +"2220 dropout_45" [id=2220, type=dropout]; +"2221 view_65" [id=2221, type=view]; +"2222 permute_54" [id=2222, type=permute]; +"2223 reshape_53" [id=2223, type=reshape]; +"2224 roll_11" [id=2224, type=roll]; +"2225 slice_185" [id=2225, type=slice]; +"2226 slice_186" [id=2226, type=slice]; +"2227 slice_187" [id=2227, type=slice]; +"2228 slice_188" [id=2228, type=slice]; +"2229 contiguous_21" [id=2229, type=contiguous]; +"2230 layer_norm_25" [id=2230, type=layer_norm]; +"2231 add_40" [id=2231, type=add]; +"2232 quantize_per_tensor_default_73" [id=2232, type=quantize_per_tensor]; +"2233 dequantize_per_tensor_default_73" [id=2233, type=dequantize_per_tensor]; +"2234 linear_72" [id=2234, type=linear]; +"2235 gelu_11" [id=2235, type=gelu]; +"2236 quantize_per_tensor_default_74" [id=2236, type=quantize_per_tensor]; +"2237 dequantize_per_tensor_default_74" [id=2237, type=dequantize_per_tensor]; +"2238 dropout_46" [id=2238, type=dropout]; +"2239 linear_73" [id=2239, type=linear]; +"2240 dropout_47" [id=2240, type=dropout]; +"2241 layer_norm_26" [id=2241, type=layer_norm]; +"2242 add_41" [id=2242, type=add]; +"2243 quantize_per_tensor_default_75" [id=2243, type=quantize_per_tensor]; +"2244 dequantize_per_tensor_default_75" [id=2244, type=dequantize_per_tensor]; +"2245 linear_74" [id=2245, type=linear]; +"2246 relu__12" [id=2246, type=relu_]; +"2247 linear_75" [id=2247, type=linear]; +"2248 view_66" [id=2248, type=view]; +"2249 index_12" [id=2249, type=index]; +"2250 view_67" [id=2250, type=view]; +"2251 permute_55" [id=2251, type=permute]; +"2252 contiguous_22" [id=2252, type=contiguous]; +"2253 unsqueeze_36" [id=2253, type=unsqueeze]; +"2254 sigmoid_12" [id=2254, type=sigmoid]; +"2255 mul_24" [id=2255, type=mul]; +"2256 pad_14" [id=2256, type=pad]; +"2257 view_68" [id=2257, type=view]; +"2258 permute_56" [id=2258, type=permute]; +"2259 reshape_54" [id=2259, type=reshape]; +"2260 clone_12" [id=2260, type=clone]; +"2261 slice_189" [id=2261, type=slice]; +"2262 zero__12" [id=2262, type=zero_]; +"2263 linear_76" [id=2263, type=linear]; +"2264 reshape_55" [id=2264, type=reshape]; +"2265 permute_57" [id=2265, type=permute]; +"2266 select_36" [id=2266, type=select]; +"2267 select_37" [id=2267, type=select]; +"2268 select_38" [id=2268, type=select]; +"2269 linalg_vector_norm_24" [id=2269, type=linalg_vector_norm]; +"2270 clamp_min_24" [id=2270, type=clamp_min]; +"2271 expand_as_24" [id=2271, type=expand_as]; +"2272 div_24" [id=2272, type=div]; +"2273 quantize_per_tensor_default_76" [id=2273, type=quantize_per_tensor]; +"2274 dequantize_per_tensor_default_76" [id=2274, type=dequantize_per_tensor]; +"2275 linalg_vector_norm_25" [id=2275, type=linalg_vector_norm]; +"2276 clamp_min_25" [id=2276, type=clamp_min]; +"2277 expand_as_25" [id=2277, type=expand_as]; +"2278 div_25" [id=2278, type=div]; +"2279 quantize_per_tensor_default_77" [id=2279, type=quantize_per_tensor]; +"2280 dequantize_per_tensor_default_77" [id=2280, type=dequantize_per_tensor]; +"2281 transpose_24" [id=2281, type=transpose]; +"2282 matmul_24" [id=2282, type=matmul]; +"2283 clamp_12" [id=2283, type=clamp]; +"2284 exp_12" [id=2284, type=exp]; +"2285 mul_25" [id=2285, type=mul]; +"2286 add_42" [id=2286, type=add]; +"2287 softmax_12" [id=2287, type=softmax]; +"2288 dropout_48" [id=2288, type=dropout]; +"2289 matmul_25" [id=2289, type=matmul]; +"2290 quantize_per_tensor_default_78" [id=2290, type=quantize_per_tensor]; +"2291 dequantize_per_tensor_default_78" [id=2291, type=dequantize_per_tensor]; +"2292 transpose_25" [id=2292, type=transpose]; +"2293 reshape_56" [id=2293, type=reshape]; +"2294 linear_77" [id=2294, type=linear]; +"2295 dropout_49" [id=2295, type=dropout]; +"2296 view_69" [id=2296, type=view]; +"2297 permute_58" [id=2297, type=permute]; +"2298 reshape_57" [id=2298, type=reshape]; +"2299 slice_190" [id=2299, type=slice]; +"2300 slice_191" [id=2300, type=slice]; +"2301 slice_192" [id=2301, type=slice]; +"2302 slice_193" [id=2302, type=slice]; +"2303 contiguous_23" [id=2303, type=contiguous]; +"2304 layer_norm_27" [id=2304, type=layer_norm]; +"2305 add_43" [id=2305, type=add]; +"2306 quantize_per_tensor_default_79" [id=2306, type=quantize_per_tensor]; +"2307 dequantize_per_tensor_default_79" [id=2307, type=dequantize_per_tensor]; +"2308 linear_78" [id=2308, type=linear]; +"2309 gelu_12" [id=2309, type=gelu]; +"2310 quantize_per_tensor_default_80" [id=2310, type=quantize_per_tensor]; +"2311 dequantize_per_tensor_default_80" [id=2311, type=dequantize_per_tensor]; +"2312 dropout_50" [id=2312, type=dropout]; +"2313 linear_79" [id=2313, type=linear]; +"2314 dropout_51" [id=2314, type=dropout]; +"2315 layer_norm_28" [id=2315, type=layer_norm]; +"2316 add_44" [id=2316, type=add]; +"2317 linear_80" [id=2317, type=linear]; +"2318 relu__13" [id=2318, type=relu_]; +"2319 linear_81" [id=2319, type=linear]; +"2320 view_70" [id=2320, type=view]; +"2321 index_13" [id=2321, type=index]; +"2322 view_71" [id=2322, type=view]; +"2323 permute_59" [id=2323, type=permute]; +"2324 contiguous_24" [id=2324, type=contiguous]; +"2325 unsqueeze_37" [id=2325, type=unsqueeze]; +"2326 sigmoid_13" [id=2326, type=sigmoid]; +"2327 mul_26" [id=2327, type=mul]; +"2328 pad_15" [id=2328, type=pad]; +"2329 roll_12" [id=2329, type=roll]; +"2330 view_72" [id=2330, type=view]; +"2331 permute_60" [id=2331, type=permute]; +"2332 reshape_58" [id=2332, type=reshape]; +"2333 quantize_per_tensor_default_81" [id=2333, type=quantize_per_tensor]; +"2334 dequantize_per_tensor_default_81" [id=2334, type=dequantize_per_tensor]; +"2335 clone_13" [id=2335, type=clone]; +"2336 slice_194" [id=2336, type=slice]; +"2337 zero__13" [id=2337, type=zero_]; +"2338 linear_82" [id=2338, type=linear]; +"2339 reshape_59" [id=2339, type=reshape]; +"2340 permute_61" [id=2340, type=permute]; +"2341 select_39" [id=2341, type=select]; +"2342 select_40" [id=2342, type=select]; +"2343 select_41" [id=2343, type=select]; +"2344 linalg_vector_norm_26" [id=2344, type=linalg_vector_norm]; +"2345 clamp_min_26" [id=2345, type=clamp_min]; +"2346 expand_as_26" [id=2346, type=expand_as]; +"2347 div_26" [id=2347, type=div]; +"2348 quantize_per_tensor_default_82" [id=2348, type=quantize_per_tensor]; +"2349 dequantize_per_tensor_default_82" [id=2349, type=dequantize_per_tensor]; +"2350 linalg_vector_norm_27" [id=2350, type=linalg_vector_norm]; +"2351 clamp_min_27" [id=2351, type=clamp_min]; +"2352 expand_as_27" [id=2352, type=expand_as]; +"2353 div_27" [id=2353, type=div]; +"2354 quantize_per_tensor_default_83" [id=2354, type=quantize_per_tensor]; +"2355 dequantize_per_tensor_default_83" [id=2355, type=dequantize_per_tensor]; +"2356 transpose_26" [id=2356, type=transpose]; +"2357 matmul_26" [id=2357, type=matmul]; +"2358 clamp_13" [id=2358, type=clamp]; +"2359 exp_13" [id=2359, type=exp]; +"2360 mul_27" [id=2360, type=mul]; +"2361 add_45" [id=2361, type=add]; +"2362 new_zeros_6" [id=2362, type=new_zeros]; +"2363 lift_fresh_copy_54" [id=2363, type=lift_fresh_copy]; +"2364 slice_195" [id=2364, type=slice]; +"2365 slice_196" [id=2365, type=slice]; +"2366 fill__54" [id=2366, type=fill_]; +"2367 lift_fresh_copy_55" [id=2367, type=lift_fresh_copy]; +"2368 slice_197" [id=2368, type=slice]; +"2369 slice_198" [id=2369, type=slice]; +"2370 fill__55" [id=2370, type=fill_]; +"2371 lift_fresh_copy_56" [id=2371, type=lift_fresh_copy]; +"2372 slice_199" [id=2372, type=slice]; +"2373 slice_200" [id=2373, type=slice]; +"2374 fill__56" [id=2374, type=fill_]; +"2375 lift_fresh_copy_57" [id=2375, type=lift_fresh_copy]; +"2376 slice_201" [id=2376, type=slice]; +"2377 slice_202" [id=2377, type=slice]; +"2378 fill__57" [id=2378, type=fill_]; +"2379 lift_fresh_copy_58" [id=2379, type=lift_fresh_copy]; +"2380 slice_203" [id=2380, type=slice]; +"2381 slice_204" [id=2381, type=slice]; +"2382 fill__58" [id=2382, type=fill_]; +"2383 lift_fresh_copy_59" [id=2383, type=lift_fresh_copy]; +"2384 slice_205" [id=2384, type=slice]; +"2385 slice_206" [id=2385, type=slice]; +"2386 fill__59" [id=2386, type=fill_]; +"2387 lift_fresh_copy_60" [id=2387, type=lift_fresh_copy]; +"2388 slice_207" [id=2388, type=slice]; +"2389 slice_208" [id=2389, type=slice]; +"2390 fill__60" [id=2390, type=fill_]; +"2391 lift_fresh_copy_61" [id=2391, type=lift_fresh_copy]; +"2392 slice_209" [id=2392, type=slice]; +"2393 slice_210" [id=2393, type=slice]; +"2394 fill__61" [id=2394, type=fill_]; +"2395 lift_fresh_copy_62" [id=2395, type=lift_fresh_copy]; +"2396 slice_211" [id=2396, type=slice]; +"2397 slice_212" [id=2397, type=slice]; +"2398 fill__62" [id=2398, type=fill_]; +"2399 view_73" [id=2399, type=view]; +"2400 permute_62" [id=2400, type=permute]; +"2401 reshape_60" [id=2401, type=reshape]; +"2402 unsqueeze_38" [id=2402, type=unsqueeze]; +"2403 unsqueeze_39" [id=2403, type=unsqueeze]; +"2404 sub_6" [id=2404, type=sub]; +"2405 ne_6" [id=2405, type=ne]; +"2406 masked_fill_12" [id=2406, type=masked_fill]; +"2407 eq_6" [id=2407, type=eq]; +"2408 masked_fill_13" [id=2408, type=masked_fill]; +"2409 view_74" [id=2409, type=view]; +"2410 unsqueeze_40" [id=2410, type=unsqueeze]; +"2411 unsqueeze_41" [id=2411, type=unsqueeze]; +"2412 add_46" [id=2412, type=add]; +"2413 view_75" [id=2413, type=view]; +"2414 softmax_13" [id=2414, type=softmax]; +"2415 dropout_52" [id=2415, type=dropout]; +"2416 matmul_27" [id=2416, type=matmul]; +"2417 quantize_per_tensor_default_84" [id=2417, type=quantize_per_tensor]; +"2418 dequantize_per_tensor_default_84" [id=2418, type=dequantize_per_tensor]; +"2419 transpose_27" [id=2419, type=transpose]; +"2420 reshape_61" [id=2420, type=reshape]; +"2421 linear_83" [id=2421, type=linear]; +"2422 dropout_53" [id=2422, type=dropout]; +"2423 view_76" [id=2423, type=view]; +"2424 permute_63" [id=2424, type=permute]; +"2425 reshape_62" [id=2425, type=reshape]; +"2426 roll_13" [id=2426, type=roll]; +"2427 slice_213" [id=2427, type=slice]; +"2428 slice_214" [id=2428, type=slice]; +"2429 slice_215" [id=2429, type=slice]; +"2430 slice_216" [id=2430, type=slice]; +"2431 contiguous_25" [id=2431, type=contiguous]; +"2432 layer_norm_29" [id=2432, type=layer_norm]; +"2433 add_47" [id=2433, type=add]; +"2434 quantize_per_tensor_default_85" [id=2434, type=quantize_per_tensor]; +"2435 dequantize_per_tensor_default_85" [id=2435, type=dequantize_per_tensor]; +"2436 linear_84" [id=2436, type=linear]; +"2437 gelu_13" [id=2437, type=gelu]; +"2438 quantize_per_tensor_default_86" [id=2438, type=quantize_per_tensor]; +"2439 dequantize_per_tensor_default_86" [id=2439, type=dequantize_per_tensor]; +"2440 dropout_54" [id=2440, type=dropout]; +"2441 linear_85" [id=2441, type=linear]; +"2442 dropout_55" [id=2442, type=dropout]; +"2443 layer_norm_30" [id=2443, type=layer_norm]; +"2444 add_48" [id=2444, type=add]; +"2445 quantize_per_tensor_default_87" [id=2445, type=quantize_per_tensor]; +"2446 dequantize_per_tensor_default_87" [id=2446, type=dequantize_per_tensor]; +"2447 linear_86" [id=2447, type=linear]; +"2448 relu__14" [id=2448, type=relu_]; +"2449 linear_87" [id=2449, type=linear]; +"2450 view_77" [id=2450, type=view]; +"2451 index_14" [id=2451, type=index]; +"2452 view_78" [id=2452, type=view]; +"2453 permute_64" [id=2453, type=permute]; +"2454 contiguous_26" [id=2454, type=contiguous]; +"2455 unsqueeze_42" [id=2455, type=unsqueeze]; +"2456 sigmoid_14" [id=2456, type=sigmoid]; +"2457 mul_28" [id=2457, type=mul]; +"2458 pad_16" [id=2458, type=pad]; +"2459 view_79" [id=2459, type=view]; +"2460 permute_65" [id=2460, type=permute]; +"2461 reshape_63" [id=2461, type=reshape]; +"2462 clone_14" [id=2462, type=clone]; +"2463 slice_217" [id=2463, type=slice]; +"2464 zero__14" [id=2464, type=zero_]; +"2465 linear_88" [id=2465, type=linear]; +"2466 reshape_64" [id=2466, type=reshape]; +"2467 permute_66" [id=2467, type=permute]; +"2468 select_42" [id=2468, type=select]; +"2469 select_43" [id=2469, type=select]; +"2470 select_44" [id=2470, type=select]; +"2471 linalg_vector_norm_28" [id=2471, type=linalg_vector_norm]; +"2472 clamp_min_28" [id=2472, type=clamp_min]; +"2473 expand_as_28" [id=2473, type=expand_as]; +"2474 div_28" [id=2474, type=div]; +"2475 quantize_per_tensor_default_88" [id=2475, type=quantize_per_tensor]; +"2476 dequantize_per_tensor_default_88" [id=2476, type=dequantize_per_tensor]; +"2477 linalg_vector_norm_29" [id=2477, type=linalg_vector_norm]; +"2478 clamp_min_29" [id=2478, type=clamp_min]; +"2479 expand_as_29" [id=2479, type=expand_as]; +"2480 div_29" [id=2480, type=div]; +"2481 quantize_per_tensor_default_89" [id=2481, type=quantize_per_tensor]; +"2482 dequantize_per_tensor_default_89" [id=2482, type=dequantize_per_tensor]; +"2483 transpose_28" [id=2483, type=transpose]; +"2484 matmul_28" [id=2484, type=matmul]; +"2485 clamp_14" [id=2485, type=clamp]; +"2486 exp_14" [id=2486, type=exp]; +"2487 mul_29" [id=2487, type=mul]; +"2488 add_49" [id=2488, type=add]; +"2489 softmax_14" [id=2489, type=softmax]; +"2490 dropout_56" [id=2490, type=dropout]; +"2491 matmul_29" [id=2491, type=matmul]; +"2492 quantize_per_tensor_default_90" [id=2492, type=quantize_per_tensor]; +"2493 dequantize_per_tensor_default_90" [id=2493, type=dequantize_per_tensor]; +"2494 transpose_29" [id=2494, type=transpose]; +"2495 reshape_65" [id=2495, type=reshape]; +"2496 linear_89" [id=2496, type=linear]; +"2497 dropout_57" [id=2497, type=dropout]; +"2498 view_80" [id=2498, type=view]; +"2499 permute_67" [id=2499, type=permute]; +"2500 reshape_66" [id=2500, type=reshape]; +"2501 slice_218" [id=2501, type=slice]; +"2502 slice_219" [id=2502, type=slice]; +"2503 slice_220" [id=2503, type=slice]; +"2504 slice_221" [id=2504, type=slice]; +"2505 contiguous_27" [id=2505, type=contiguous]; +"2506 layer_norm_31" [id=2506, type=layer_norm]; +"2507 add_50" [id=2507, type=add]; +"2508 quantize_per_tensor_default_91" [id=2508, type=quantize_per_tensor]; +"2509 dequantize_per_tensor_default_91" [id=2509, type=dequantize_per_tensor]; +"2510 linear_90" [id=2510, type=linear]; +"2511 gelu_14" [id=2511, type=gelu]; +"2512 quantize_per_tensor_default_92" [id=2512, type=quantize_per_tensor]; +"2513 dequantize_per_tensor_default_92" [id=2513, type=dequantize_per_tensor]; +"2514 dropout_58" [id=2514, type=dropout]; +"2515 linear_91" [id=2515, type=linear]; +"2516 dropout_59" [id=2516, type=dropout]; +"2517 layer_norm_32" [id=2517, type=layer_norm]; +"2518 add_51" [id=2518, type=add]; +"2519 linear_92" [id=2519, type=linear]; +"2520 relu__15" [id=2520, type=relu_]; +"2521 linear_93" [id=2521, type=linear]; +"2522 view_81" [id=2522, type=view]; +"2523 index_15" [id=2523, type=index]; +"2524 view_82" [id=2524, type=view]; +"2525 permute_68" [id=2525, type=permute]; +"2526 contiguous_28" [id=2526, type=contiguous]; +"2527 unsqueeze_43" [id=2527, type=unsqueeze]; +"2528 sigmoid_15" [id=2528, type=sigmoid]; +"2529 mul_30" [id=2529, type=mul]; +"2530 pad_17" [id=2530, type=pad]; +"2531 roll_14" [id=2531, type=roll]; +"2532 view_83" [id=2532, type=view]; +"2533 permute_69" [id=2533, type=permute]; +"2534 reshape_67" [id=2534, type=reshape]; +"2535 quantize_per_tensor_default_93" [id=2535, type=quantize_per_tensor]; +"2536 dequantize_per_tensor_default_93" [id=2536, type=dequantize_per_tensor]; +"2537 clone_15" [id=2537, type=clone]; +"2538 slice_222" [id=2538, type=slice]; +"2539 zero__15" [id=2539, type=zero_]; +"2540 linear_94" [id=2540, type=linear]; +"2541 reshape_68" [id=2541, type=reshape]; +"2542 permute_70" [id=2542, type=permute]; +"2543 select_45" [id=2543, type=select]; +"2544 select_46" [id=2544, type=select]; +"2545 select_47" [id=2545, type=select]; +"2546 linalg_vector_norm_30" [id=2546, type=linalg_vector_norm]; +"2547 clamp_min_30" [id=2547, type=clamp_min]; +"2548 expand_as_30" [id=2548, type=expand_as]; +"2549 div_30" [id=2549, type=div]; +"2550 quantize_per_tensor_default_94" [id=2550, type=quantize_per_tensor]; +"2551 dequantize_per_tensor_default_94" [id=2551, type=dequantize_per_tensor]; +"2552 linalg_vector_norm_31" [id=2552, type=linalg_vector_norm]; +"2553 clamp_min_31" [id=2553, type=clamp_min]; +"2554 expand_as_31" [id=2554, type=expand_as]; +"2555 div_31" [id=2555, type=div]; +"2556 quantize_per_tensor_default_95" [id=2556, type=quantize_per_tensor]; +"2557 dequantize_per_tensor_default_95" [id=2557, type=dequantize_per_tensor]; +"2558 transpose_30" [id=2558, type=transpose]; +"2559 matmul_30" [id=2559, type=matmul]; +"2560 clamp_15" [id=2560, type=clamp]; +"2561 exp_15" [id=2561, type=exp]; +"2562 mul_31" [id=2562, type=mul]; +"2563 add_52" [id=2563, type=add]; +"2564 new_zeros_7" [id=2564, type=new_zeros]; +"2565 lift_fresh_copy_63" [id=2565, type=lift_fresh_copy]; +"2566 slice_223" [id=2566, type=slice]; +"2567 slice_224" [id=2567, type=slice]; +"2568 fill__63" [id=2568, type=fill_]; +"2569 lift_fresh_copy_64" [id=2569, type=lift_fresh_copy]; +"2570 slice_225" [id=2570, type=slice]; +"2571 slice_226" [id=2571, type=slice]; +"2572 fill__64" [id=2572, type=fill_]; +"2573 lift_fresh_copy_65" [id=2573, type=lift_fresh_copy]; +"2574 slice_227" [id=2574, type=slice]; +"2575 slice_228" [id=2575, type=slice]; +"2576 fill__65" [id=2576, type=fill_]; +"2577 lift_fresh_copy_66" [id=2577, type=lift_fresh_copy]; +"2578 slice_229" [id=2578, type=slice]; +"2579 slice_230" [id=2579, type=slice]; +"2580 fill__66" [id=2580, type=fill_]; +"2581 lift_fresh_copy_67" [id=2581, type=lift_fresh_copy]; +"2582 slice_231" [id=2582, type=slice]; +"2583 slice_232" [id=2583, type=slice]; +"2584 fill__67" [id=2584, type=fill_]; +"2585 lift_fresh_copy_68" [id=2585, type=lift_fresh_copy]; +"2586 slice_233" [id=2586, type=slice]; +"2587 slice_234" [id=2587, type=slice]; +"2588 fill__68" [id=2588, type=fill_]; +"2589 lift_fresh_copy_69" [id=2589, type=lift_fresh_copy]; +"2590 slice_235" [id=2590, type=slice]; +"2591 slice_236" [id=2591, type=slice]; +"2592 fill__69" [id=2592, type=fill_]; +"2593 lift_fresh_copy_70" [id=2593, type=lift_fresh_copy]; +"2594 slice_237" [id=2594, type=slice]; +"2595 slice_238" [id=2595, type=slice]; +"2596 fill__70" [id=2596, type=fill_]; +"2597 lift_fresh_copy_71" [id=2597, type=lift_fresh_copy]; +"2598 slice_239" [id=2598, type=slice]; +"2599 slice_240" [id=2599, type=slice]; +"2600 fill__71" [id=2600, type=fill_]; +"2601 view_84" [id=2601, type=view]; +"2602 permute_71" [id=2602, type=permute]; +"2603 reshape_69" [id=2603, type=reshape]; +"2604 unsqueeze_44" [id=2604, type=unsqueeze]; +"2605 unsqueeze_45" [id=2605, type=unsqueeze]; +"2606 sub_7" [id=2606, type=sub]; +"2607 ne_7" [id=2607, type=ne]; +"2608 masked_fill_14" [id=2608, type=masked_fill]; +"2609 eq_7" [id=2609, type=eq]; +"2610 masked_fill_15" [id=2610, type=masked_fill]; +"2611 view_85" [id=2611, type=view]; +"2612 unsqueeze_46" [id=2612, type=unsqueeze]; +"2613 unsqueeze_47" [id=2613, type=unsqueeze]; +"2614 add_53" [id=2614, type=add]; +"2615 view_86" [id=2615, type=view]; +"2616 softmax_15" [id=2616, type=softmax]; +"2617 dropout_60" [id=2617, type=dropout]; +"2618 matmul_31" [id=2618, type=matmul]; +"2619 quantize_per_tensor_default_96" [id=2619, type=quantize_per_tensor]; +"2620 dequantize_per_tensor_default_96" [id=2620, type=dequantize_per_tensor]; +"2621 transpose_31" [id=2621, type=transpose]; +"2622 reshape_70" [id=2622, type=reshape]; +"2623 linear_95" [id=2623, type=linear]; +"2624 dropout_61" [id=2624, type=dropout]; +"2625 view_87" [id=2625, type=view]; +"2626 permute_72" [id=2626, type=permute]; +"2627 reshape_71" [id=2627, type=reshape]; +"2628 roll_15" [id=2628, type=roll]; +"2629 slice_241" [id=2629, type=slice]; +"2630 slice_242" [id=2630, type=slice]; +"2631 slice_243" [id=2631, type=slice]; +"2632 slice_244" [id=2632, type=slice]; +"2633 contiguous_29" [id=2633, type=contiguous]; +"2634 layer_norm_33" [id=2634, type=layer_norm]; +"2635 add_54" [id=2635, type=add]; +"2636 quantize_per_tensor_default_97" [id=2636, type=quantize_per_tensor]; +"2637 dequantize_per_tensor_default_97" [id=2637, type=dequantize_per_tensor]; +"2638 linear_96" [id=2638, type=linear]; +"2639 gelu_15" [id=2639, type=gelu]; +"2640 quantize_per_tensor_default_98" [id=2640, type=quantize_per_tensor]; +"2641 dequantize_per_tensor_default_98" [id=2641, type=dequantize_per_tensor]; +"2642 dropout_62" [id=2642, type=dropout]; +"2643 linear_97" [id=2643, type=linear]; +"2644 dropout_63" [id=2644, type=dropout]; +"2645 layer_norm_34" [id=2645, type=layer_norm]; +"2646 add_55" [id=2646, type=add]; +"2647 quantize_per_tensor_default_99" [id=2647, type=quantize_per_tensor]; +"2648 dequantize_per_tensor_default_99" [id=2648, type=dequantize_per_tensor]; +"2649 linear_98" [id=2649, type=linear]; +"2650 relu__16" [id=2650, type=relu_]; +"2651 linear_99" [id=2651, type=linear]; +"2652 view_88" [id=2652, type=view]; +"2653 index_16" [id=2653, type=index]; +"2654 view_89" [id=2654, type=view]; +"2655 permute_73" [id=2655, type=permute]; +"2656 contiguous_30" [id=2656, type=contiguous]; +"2657 unsqueeze_48" [id=2657, type=unsqueeze]; +"2658 sigmoid_16" [id=2658, type=sigmoid]; +"2659 mul_32" [id=2659, type=mul]; +"2660 pad_18" [id=2660, type=pad]; +"2661 view_90" [id=2661, type=view]; +"2662 permute_74" [id=2662, type=permute]; +"2663 reshape_72" [id=2663, type=reshape]; +"2664 clone_16" [id=2664, type=clone]; +"2665 slice_245" [id=2665, type=slice]; +"2666 zero__16" [id=2666, type=zero_]; +"2667 linear_100" [id=2667, type=linear]; +"2668 reshape_73" [id=2668, type=reshape]; +"2669 permute_75" [id=2669, type=permute]; +"2670 select_48" [id=2670, type=select]; +"2671 select_49" [id=2671, type=select]; +"2672 select_50" [id=2672, type=select]; +"2673 linalg_vector_norm_32" [id=2673, type=linalg_vector_norm]; +"2674 clamp_min_32" [id=2674, type=clamp_min]; +"2675 expand_as_32" [id=2675, type=expand_as]; +"2676 div_32" [id=2676, type=div]; +"2677 quantize_per_tensor_default_100" [id=2677, type=quantize_per_tensor]; +"2678 dequantize_per_tensor_default_100" [id=2678, type=dequantize_per_tensor]; +"2679 linalg_vector_norm_33" [id=2679, type=linalg_vector_norm]; +"2680 clamp_min_33" [id=2680, type=clamp_min]; +"2681 expand_as_33" [id=2681, type=expand_as]; +"2682 div_33" [id=2682, type=div]; +"2683 quantize_per_tensor_default_101" [id=2683, type=quantize_per_tensor]; +"2684 dequantize_per_tensor_default_101" [id=2684, type=dequantize_per_tensor]; +"2685 transpose_32" [id=2685, type=transpose]; +"2686 matmul_32" [id=2686, type=matmul]; +"2687 clamp_16" [id=2687, type=clamp]; +"2688 exp_16" [id=2688, type=exp]; +"2689 mul_33" [id=2689, type=mul]; +"2690 add_56" [id=2690, type=add]; +"2691 softmax_16" [id=2691, type=softmax]; +"2692 dropout_64" [id=2692, type=dropout]; +"2693 matmul_33" [id=2693, type=matmul]; +"2694 quantize_per_tensor_default_102" [id=2694, type=quantize_per_tensor]; +"2695 dequantize_per_tensor_default_102" [id=2695, type=dequantize_per_tensor]; +"2696 transpose_33" [id=2696, type=transpose]; +"2697 reshape_74" [id=2697, type=reshape]; +"2698 linear_101" [id=2698, type=linear]; +"2699 dropout_65" [id=2699, type=dropout]; +"2700 view_91" [id=2700, type=view]; +"2701 permute_76" [id=2701, type=permute]; +"2702 reshape_75" [id=2702, type=reshape]; +"2703 slice_246" [id=2703, type=slice]; +"2704 slice_247" [id=2704, type=slice]; +"2705 slice_248" [id=2705, type=slice]; +"2706 slice_249" [id=2706, type=slice]; +"2707 contiguous_31" [id=2707, type=contiguous]; +"2708 layer_norm_35" [id=2708, type=layer_norm]; +"2709 add_57" [id=2709, type=add]; +"2710 quantize_per_tensor_default_103" [id=2710, type=quantize_per_tensor]; +"2711 dequantize_per_tensor_default_103" [id=2711, type=dequantize_per_tensor]; +"2712 linear_102" [id=2712, type=linear]; +"2713 gelu_16" [id=2713, type=gelu]; +"2714 quantize_per_tensor_default_104" [id=2714, type=quantize_per_tensor]; +"2715 dequantize_per_tensor_default_104" [id=2715, type=dequantize_per_tensor]; +"2716 dropout_66" [id=2716, type=dropout]; +"2717 linear_103" [id=2717, type=linear]; +"2718 dropout_67" [id=2718, type=dropout]; +"2719 layer_norm_36" [id=2719, type=layer_norm]; +"2720 add_58" [id=2720, type=add]; +"2721 linear_104" [id=2721, type=linear]; +"2722 relu__17" [id=2722, type=relu_]; +"2723 linear_105" [id=2723, type=linear]; +"2724 view_92" [id=2724, type=view]; +"2725 index_17" [id=2725, type=index]; +"2726 view_93" [id=2726, type=view]; +"2727 permute_77" [id=2727, type=permute]; +"2728 contiguous_32" [id=2728, type=contiguous]; +"2729 unsqueeze_49" [id=2729, type=unsqueeze]; +"2730 sigmoid_17" [id=2730, type=sigmoid]; +"2731 mul_34" [id=2731, type=mul]; +"2732 pad_19" [id=2732, type=pad]; +"2733 roll_16" [id=2733, type=roll]; +"2734 view_94" [id=2734, type=view]; +"2735 permute_78" [id=2735, type=permute]; +"2736 reshape_76" [id=2736, type=reshape]; +"2737 quantize_per_tensor_default_105" [id=2737, type=quantize_per_tensor]; +"2738 dequantize_per_tensor_default_105" [id=2738, type=dequantize_per_tensor]; +"2739 clone_17" [id=2739, type=clone]; +"2740 slice_250" [id=2740, type=slice]; +"2741 zero__17" [id=2741, type=zero_]; +"2742 linear_106" [id=2742, type=linear]; +"2743 reshape_77" [id=2743, type=reshape]; +"2744 permute_79" [id=2744, type=permute]; +"2745 select_51" [id=2745, type=select]; +"2746 select_52" [id=2746, type=select]; +"2747 select_53" [id=2747, type=select]; +"2748 linalg_vector_norm_34" [id=2748, type=linalg_vector_norm]; +"2749 clamp_min_34" [id=2749, type=clamp_min]; +"2750 expand_as_34" [id=2750, type=expand_as]; +"2751 div_34" [id=2751, type=div]; +"2752 quantize_per_tensor_default_106" [id=2752, type=quantize_per_tensor]; +"2753 dequantize_per_tensor_default_106" [id=2753, type=dequantize_per_tensor]; +"2754 linalg_vector_norm_35" [id=2754, type=linalg_vector_norm]; +"2755 clamp_min_35" [id=2755, type=clamp_min]; +"2756 expand_as_35" [id=2756, type=expand_as]; +"2757 div_35" [id=2757, type=div]; +"2758 quantize_per_tensor_default_107" [id=2758, type=quantize_per_tensor]; +"2759 dequantize_per_tensor_default_107" [id=2759, type=dequantize_per_tensor]; +"2760 transpose_34" [id=2760, type=transpose]; +"2761 matmul_34" [id=2761, type=matmul]; +"2762 clamp_17" [id=2762, type=clamp]; +"2763 exp_17" [id=2763, type=exp]; +"2764 mul_35" [id=2764, type=mul]; +"2765 add_59" [id=2765, type=add]; +"2766 new_zeros_8" [id=2766, type=new_zeros]; +"2767 lift_fresh_copy_72" [id=2767, type=lift_fresh_copy]; +"2768 slice_251" [id=2768, type=slice]; +"2769 slice_252" [id=2769, type=slice]; +"2770 fill__72" [id=2770, type=fill_]; +"2771 lift_fresh_copy_73" [id=2771, type=lift_fresh_copy]; +"2772 slice_253" [id=2772, type=slice]; +"2773 slice_254" [id=2773, type=slice]; +"2774 fill__73" [id=2774, type=fill_]; +"2775 lift_fresh_copy_74" [id=2775, type=lift_fresh_copy]; +"2776 slice_255" [id=2776, type=slice]; +"2777 slice_256" [id=2777, type=slice]; +"2778 fill__74" [id=2778, type=fill_]; +"2779 lift_fresh_copy_75" [id=2779, type=lift_fresh_copy]; +"2780 slice_257" [id=2780, type=slice]; +"2781 slice_258" [id=2781, type=slice]; +"2782 fill__75" [id=2782, type=fill_]; +"2783 lift_fresh_copy_76" [id=2783, type=lift_fresh_copy]; +"2784 slice_259" [id=2784, type=slice]; +"2785 slice_260" [id=2785, type=slice]; +"2786 fill__76" [id=2786, type=fill_]; +"2787 lift_fresh_copy_77" [id=2787, type=lift_fresh_copy]; +"2788 slice_261" [id=2788, type=slice]; +"2789 slice_262" [id=2789, type=slice]; +"2790 fill__77" [id=2790, type=fill_]; +"2791 lift_fresh_copy_78" [id=2791, type=lift_fresh_copy]; +"2792 slice_263" [id=2792, type=slice]; +"2793 slice_264" [id=2793, type=slice]; +"2794 fill__78" [id=2794, type=fill_]; +"2795 lift_fresh_copy_79" [id=2795, type=lift_fresh_copy]; +"2796 slice_265" [id=2796, type=slice]; +"2797 slice_266" [id=2797, type=slice]; +"2798 fill__79" [id=2798, type=fill_]; +"2799 lift_fresh_copy_80" [id=2799, type=lift_fresh_copy]; +"2800 slice_267" [id=2800, type=slice]; +"2801 slice_268" [id=2801, type=slice]; +"2802 fill__80" [id=2802, type=fill_]; +"2803 view_95" [id=2803, type=view]; +"2804 permute_80" [id=2804, type=permute]; +"2805 reshape_78" [id=2805, type=reshape]; +"2806 unsqueeze_50" [id=2806, type=unsqueeze]; +"2807 unsqueeze_51" [id=2807, type=unsqueeze]; +"2808 sub_8" [id=2808, type=sub]; +"2809 ne_8" [id=2809, type=ne]; +"2810 masked_fill_16" [id=2810, type=masked_fill]; +"2811 eq_8" [id=2811, type=eq]; +"2812 masked_fill_17" [id=2812, type=masked_fill]; +"2813 view_96" [id=2813, type=view]; +"2814 unsqueeze_52" [id=2814, type=unsqueeze]; +"2815 unsqueeze_53" [id=2815, type=unsqueeze]; +"2816 add_60" [id=2816, type=add]; +"2817 view_97" [id=2817, type=view]; +"2818 softmax_17" [id=2818, type=softmax]; +"2819 dropout_68" [id=2819, type=dropout]; +"2820 matmul_35" [id=2820, type=matmul]; +"2821 quantize_per_tensor_default_108" [id=2821, type=quantize_per_tensor]; +"2822 dequantize_per_tensor_default_108" [id=2822, type=dequantize_per_tensor]; +"2823 transpose_35" [id=2823, type=transpose]; +"2824 reshape_79" [id=2824, type=reshape]; +"2825 linear_107" [id=2825, type=linear]; +"2826 dropout_69" [id=2826, type=dropout]; +"2827 view_98" [id=2827, type=view]; +"2828 permute_81" [id=2828, type=permute]; +"2829 reshape_80" [id=2829, type=reshape]; +"2830 roll_17" [id=2830, type=roll]; +"2831 slice_269" [id=2831, type=slice]; +"2832 slice_270" [id=2832, type=slice]; +"2833 slice_271" [id=2833, type=slice]; +"2834 slice_272" [id=2834, type=slice]; +"2835 contiguous_33" [id=2835, type=contiguous]; +"2836 layer_norm_37" [id=2836, type=layer_norm]; +"2837 add_61" [id=2837, type=add]; +"2838 quantize_per_tensor_default_109" [id=2838, type=quantize_per_tensor]; +"2839 dequantize_per_tensor_default_109" [id=2839, type=dequantize_per_tensor]; +"2840 linear_108" [id=2840, type=linear]; +"2841 gelu_17" [id=2841, type=gelu]; +"2842 quantize_per_tensor_default_110" [id=2842, type=quantize_per_tensor]; +"2843 dequantize_per_tensor_default_110" [id=2843, type=dequantize_per_tensor]; +"2844 dropout_70" [id=2844, type=dropout]; +"2845 linear_109" [id=2845, type=linear]; +"2846 dropout_71" [id=2846, type=dropout]; +"2847 layer_norm_38" [id=2847, type=layer_norm]; +"2848 add_62" [id=2848, type=add]; +"2849 quantize_per_tensor_default_111" [id=2849, type=quantize_per_tensor]; +"2850 dequantize_per_tensor_default_111" [id=2850, type=dequantize_per_tensor]; +"2851 linear_110" [id=2851, type=linear]; +"2852 relu__18" [id=2852, type=relu_]; +"2853 linear_111" [id=2853, type=linear]; +"2854 view_99" [id=2854, type=view]; +"2855 index_18" [id=2855, type=index]; +"2856 view_100" [id=2856, type=view]; +"2857 permute_82" [id=2857, type=permute]; +"2858 contiguous_34" [id=2858, type=contiguous]; +"2859 unsqueeze_54" [id=2859, type=unsqueeze]; +"2860 sigmoid_18" [id=2860, type=sigmoid]; +"2861 mul_36" [id=2861, type=mul]; +"2862 pad_20" [id=2862, type=pad]; +"2863 view_101" [id=2863, type=view]; +"2864 permute_83" [id=2864, type=permute]; +"2865 reshape_81" [id=2865, type=reshape]; +"2866 clone_18" [id=2866, type=clone]; +"2867 slice_273" [id=2867, type=slice]; +"2868 zero__18" [id=2868, type=zero_]; +"2869 linear_112" [id=2869, type=linear]; +"2870 reshape_82" [id=2870, type=reshape]; +"2871 permute_84" [id=2871, type=permute]; +"2872 select_54" [id=2872, type=select]; +"2873 select_55" [id=2873, type=select]; +"2874 select_56" [id=2874, type=select]; +"2875 linalg_vector_norm_36" [id=2875, type=linalg_vector_norm]; +"2876 clamp_min_36" [id=2876, type=clamp_min]; +"2877 expand_as_36" [id=2877, type=expand_as]; +"2878 div_36" [id=2878, type=div]; +"2879 quantize_per_tensor_default_112" [id=2879, type=quantize_per_tensor]; +"2880 dequantize_per_tensor_default_112" [id=2880, type=dequantize_per_tensor]; +"2881 linalg_vector_norm_37" [id=2881, type=linalg_vector_norm]; +"2882 clamp_min_37" [id=2882, type=clamp_min]; +"2883 expand_as_37" [id=2883, type=expand_as]; +"2884 div_37" [id=2884, type=div]; +"2885 quantize_per_tensor_default_113" [id=2885, type=quantize_per_tensor]; +"2886 dequantize_per_tensor_default_113" [id=2886, type=dequantize_per_tensor]; +"2887 transpose_36" [id=2887, type=transpose]; +"2888 matmul_36" [id=2888, type=matmul]; +"2889 clamp_18" [id=2889, type=clamp]; +"2890 exp_18" [id=2890, type=exp]; +"2891 mul_37" [id=2891, type=mul]; +"2892 add_63" [id=2892, type=add]; +"2893 softmax_18" [id=2893, type=softmax]; +"2894 dropout_72" [id=2894, type=dropout]; +"2895 matmul_37" [id=2895, type=matmul]; +"2896 quantize_per_tensor_default_114" [id=2896, type=quantize_per_tensor]; +"2897 dequantize_per_tensor_default_114" [id=2897, type=dequantize_per_tensor]; +"2898 transpose_37" [id=2898, type=transpose]; +"2899 reshape_83" [id=2899, type=reshape]; +"2900 linear_113" [id=2900, type=linear]; +"2901 dropout_73" [id=2901, type=dropout]; +"2902 view_102" [id=2902, type=view]; +"2903 permute_85" [id=2903, type=permute]; +"2904 reshape_84" [id=2904, type=reshape]; +"2905 slice_274" [id=2905, type=slice]; +"2906 slice_275" [id=2906, type=slice]; +"2907 slice_276" [id=2907, type=slice]; +"2908 slice_277" [id=2908, type=slice]; +"2909 contiguous_35" [id=2909, type=contiguous]; +"2910 layer_norm_39" [id=2910, type=layer_norm]; +"2911 add_64" [id=2911, type=add]; +"2912 quantize_per_tensor_default_115" [id=2912, type=quantize_per_tensor]; +"2913 dequantize_per_tensor_default_115" [id=2913, type=dequantize_per_tensor]; +"2914 linear_114" [id=2914, type=linear]; +"2915 gelu_18" [id=2915, type=gelu]; +"2916 quantize_per_tensor_default_116" [id=2916, type=quantize_per_tensor]; +"2917 dequantize_per_tensor_default_116" [id=2917, type=dequantize_per_tensor]; +"2918 dropout_74" [id=2918, type=dropout]; +"2919 linear_115" [id=2919, type=linear]; +"2920 dropout_75" [id=2920, type=dropout]; +"2921 layer_norm_40" [id=2921, type=layer_norm]; +"2922 add_65" [id=2922, type=add]; +"2923 linear_116" [id=2923, type=linear]; +"2924 relu__19" [id=2924, type=relu_]; +"2925 linear_117" [id=2925, type=linear]; +"2926 view_103" [id=2926, type=view]; +"2927 index_19" [id=2927, type=index]; +"2928 view_104" [id=2928, type=view]; +"2929 permute_86" [id=2929, type=permute]; +"2930 contiguous_36" [id=2930, type=contiguous]; +"2931 unsqueeze_55" [id=2931, type=unsqueeze]; +"2932 sigmoid_19" [id=2932, type=sigmoid]; +"2933 mul_38" [id=2933, type=mul]; +"2934 pad_21" [id=2934, type=pad]; +"2935 roll_18" [id=2935, type=roll]; +"2936 view_105" [id=2936, type=view]; +"2937 permute_87" [id=2937, type=permute]; +"2938 reshape_85" [id=2938, type=reshape]; +"2939 quantize_per_tensor_default_117" [id=2939, type=quantize_per_tensor]; +"2940 dequantize_per_tensor_default_117" [id=2940, type=dequantize_per_tensor]; +"2941 clone_19" [id=2941, type=clone]; +"2942 slice_278" [id=2942, type=slice]; +"2943 zero__19" [id=2943, type=zero_]; +"2944 linear_118" [id=2944, type=linear]; +"2945 reshape_86" [id=2945, type=reshape]; +"2946 permute_88" [id=2946, type=permute]; +"2947 select_57" [id=2947, type=select]; +"2948 select_58" [id=2948, type=select]; +"2949 select_59" [id=2949, type=select]; +"2950 linalg_vector_norm_38" [id=2950, type=linalg_vector_norm]; +"2951 clamp_min_38" [id=2951, type=clamp_min]; +"2952 expand_as_38" [id=2952, type=expand_as]; +"2953 div_38" [id=2953, type=div]; +"2954 quantize_per_tensor_default_118" [id=2954, type=quantize_per_tensor]; +"2955 dequantize_per_tensor_default_118" [id=2955, type=dequantize_per_tensor]; +"2956 linalg_vector_norm_39" [id=2956, type=linalg_vector_norm]; +"2957 clamp_min_39" [id=2957, type=clamp_min]; +"2958 expand_as_39" [id=2958, type=expand_as]; +"2959 div_39" [id=2959, type=div]; +"2960 quantize_per_tensor_default_119" [id=2960, type=quantize_per_tensor]; +"2961 dequantize_per_tensor_default_119" [id=2961, type=dequantize_per_tensor]; +"2962 transpose_38" [id=2962, type=transpose]; +"2963 matmul_38" [id=2963, type=matmul]; +"2964 clamp_19" [id=2964, type=clamp]; +"2965 exp_19" [id=2965, type=exp]; +"2966 mul_39" [id=2966, type=mul]; +"2967 add_66" [id=2967, type=add]; +"2968 new_zeros_9" [id=2968, type=new_zeros]; +"2969 lift_fresh_copy_81" [id=2969, type=lift_fresh_copy]; +"2970 slice_279" [id=2970, type=slice]; +"2971 slice_280" [id=2971, type=slice]; +"2972 fill__81" [id=2972, type=fill_]; +"2973 lift_fresh_copy_82" [id=2973, type=lift_fresh_copy]; +"2974 slice_281" [id=2974, type=slice]; +"2975 slice_282" [id=2975, type=slice]; +"2976 fill__82" [id=2976, type=fill_]; +"2977 lift_fresh_copy_83" [id=2977, type=lift_fresh_copy]; +"2978 slice_283" [id=2978, type=slice]; +"2979 slice_284" [id=2979, type=slice]; +"2980 fill__83" [id=2980, type=fill_]; +"2981 lift_fresh_copy_84" [id=2981, type=lift_fresh_copy]; +"2982 slice_285" [id=2982, type=slice]; +"2983 slice_286" [id=2983, type=slice]; +"2984 fill__84" [id=2984, type=fill_]; +"2985 lift_fresh_copy_85" [id=2985, type=lift_fresh_copy]; +"2986 slice_287" [id=2986, type=slice]; +"2987 slice_288" [id=2987, type=slice]; +"2988 fill__85" [id=2988, type=fill_]; +"2989 lift_fresh_copy_86" [id=2989, type=lift_fresh_copy]; +"2990 slice_289" [id=2990, type=slice]; +"2991 slice_290" [id=2991, type=slice]; +"2992 fill__86" [id=2992, type=fill_]; +"2993 lift_fresh_copy_87" [id=2993, type=lift_fresh_copy]; +"2994 slice_291" [id=2994, type=slice]; +"2995 slice_292" [id=2995, type=slice]; +"2996 fill__87" [id=2996, type=fill_]; +"2997 lift_fresh_copy_88" [id=2997, type=lift_fresh_copy]; +"2998 slice_293" [id=2998, type=slice]; +"2999 slice_294" [id=2999, type=slice]; +"3000 fill__88" [id=3000, type=fill_]; +"3001 lift_fresh_copy_89" [id=3001, type=lift_fresh_copy]; +"3002 slice_295" [id=3002, type=slice]; +"3003 slice_296" [id=3003, type=slice]; +"3004 fill__89" [id=3004, type=fill_]; +"3005 view_106" [id=3005, type=view]; +"3006 permute_89" [id=3006, type=permute]; +"3007 reshape_87" [id=3007, type=reshape]; +"3008 unsqueeze_56" [id=3008, type=unsqueeze]; +"3009 unsqueeze_57" [id=3009, type=unsqueeze]; +"3010 sub_9" [id=3010, type=sub]; +"3011 ne_9" [id=3011, type=ne]; +"3012 masked_fill_18" [id=3012, type=masked_fill]; +"3013 eq_9" [id=3013, type=eq]; +"3014 masked_fill_19" [id=3014, type=masked_fill]; +"3015 view_107" [id=3015, type=view]; +"3016 unsqueeze_58" [id=3016, type=unsqueeze]; +"3017 unsqueeze_59" [id=3017, type=unsqueeze]; +"3018 add_67" [id=3018, type=add]; +"3019 view_108" [id=3019, type=view]; +"3020 softmax_19" [id=3020, type=softmax]; +"3021 dropout_76" [id=3021, type=dropout]; +"3022 matmul_39" [id=3022, type=matmul]; +"3023 quantize_per_tensor_default_120" [id=3023, type=quantize_per_tensor]; +"3024 dequantize_per_tensor_default_120" [id=3024, type=dequantize_per_tensor]; +"3025 transpose_39" [id=3025, type=transpose]; +"3026 reshape_88" [id=3026, type=reshape]; +"3027 linear_119" [id=3027, type=linear]; +"3028 dropout_77" [id=3028, type=dropout]; +"3029 view_109" [id=3029, type=view]; +"3030 permute_90" [id=3030, type=permute]; +"3031 reshape_89" [id=3031, type=reshape]; +"3032 roll_19" [id=3032, type=roll]; +"3033 slice_297" [id=3033, type=slice]; +"3034 slice_298" [id=3034, type=slice]; +"3035 slice_299" [id=3035, type=slice]; +"3036 slice_300" [id=3036, type=slice]; +"3037 contiguous_37" [id=3037, type=contiguous]; +"3038 layer_norm_41" [id=3038, type=layer_norm]; +"3039 add_68" [id=3039, type=add]; +"3040 quantize_per_tensor_default_121" [id=3040, type=quantize_per_tensor]; +"3041 dequantize_per_tensor_default_121" [id=3041, type=dequantize_per_tensor]; +"3042 linear_120" [id=3042, type=linear]; +"3043 gelu_19" [id=3043, type=gelu]; +"3044 quantize_per_tensor_default_122" [id=3044, type=quantize_per_tensor]; +"3045 dequantize_per_tensor_default_122" [id=3045, type=dequantize_per_tensor]; +"3046 dropout_78" [id=3046, type=dropout]; +"3047 linear_121" [id=3047, type=linear]; +"3048 dropout_79" [id=3048, type=dropout]; +"3049 layer_norm_42" [id=3049, type=layer_norm]; +"3050 add_69" [id=3050, type=add]; +"3051 quantize_per_tensor_default_123" [id=3051, type=quantize_per_tensor]; +"3052 dequantize_per_tensor_default_123" [id=3052, type=dequantize_per_tensor]; +"3053 linear_122" [id=3053, type=linear]; +"3054 relu__20" [id=3054, type=relu_]; +"3055 linear_123" [id=3055, type=linear]; +"3056 view_110" [id=3056, type=view]; +"3057 index_20" [id=3057, type=index]; +"3058 view_111" [id=3058, type=view]; +"3059 permute_91" [id=3059, type=permute]; +"3060 contiguous_38" [id=3060, type=contiguous]; +"3061 unsqueeze_60" [id=3061, type=unsqueeze]; +"3062 sigmoid_20" [id=3062, type=sigmoid]; +"3063 mul_40" [id=3063, type=mul]; +"3064 pad_22" [id=3064, type=pad]; +"3065 view_112" [id=3065, type=view]; +"3066 permute_92" [id=3066, type=permute]; +"3067 reshape_90" [id=3067, type=reshape]; +"3068 clone_20" [id=3068, type=clone]; +"3069 slice_301" [id=3069, type=slice]; +"3070 zero__20" [id=3070, type=zero_]; +"3071 linear_124" [id=3071, type=linear]; +"3072 reshape_91" [id=3072, type=reshape]; +"3073 permute_93" [id=3073, type=permute]; +"3074 select_60" [id=3074, type=select]; +"3075 select_61" [id=3075, type=select]; +"3076 select_62" [id=3076, type=select]; +"3077 linalg_vector_norm_40" [id=3077, type=linalg_vector_norm]; +"3078 clamp_min_40" [id=3078, type=clamp_min]; +"3079 expand_as_40" [id=3079, type=expand_as]; +"3080 div_40" [id=3080, type=div]; +"3081 quantize_per_tensor_default_124" [id=3081, type=quantize_per_tensor]; +"3082 dequantize_per_tensor_default_124" [id=3082, type=dequantize_per_tensor]; +"3083 linalg_vector_norm_41" [id=3083, type=linalg_vector_norm]; +"3084 clamp_min_41" [id=3084, type=clamp_min]; +"3085 expand_as_41" [id=3085, type=expand_as]; +"3086 div_41" [id=3086, type=div]; +"3087 quantize_per_tensor_default_125" [id=3087, type=quantize_per_tensor]; +"3088 dequantize_per_tensor_default_125" [id=3088, type=dequantize_per_tensor]; +"3089 transpose_40" [id=3089, type=transpose]; +"3090 matmul_40" [id=3090, type=matmul]; +"3091 clamp_20" [id=3091, type=clamp]; +"3092 exp_20" [id=3092, type=exp]; +"3093 mul_41" [id=3093, type=mul]; +"3094 add_70" [id=3094, type=add]; +"3095 softmax_20" [id=3095, type=softmax]; +"3096 dropout_80" [id=3096, type=dropout]; +"3097 matmul_41" [id=3097, type=matmul]; +"3098 quantize_per_tensor_default_126" [id=3098, type=quantize_per_tensor]; +"3099 dequantize_per_tensor_default_126" [id=3099, type=dequantize_per_tensor]; +"3100 transpose_41" [id=3100, type=transpose]; +"3101 reshape_92" [id=3101, type=reshape]; +"3102 linear_125" [id=3102, type=linear]; +"3103 dropout_81" [id=3103, type=dropout]; +"3104 view_113" [id=3104, type=view]; +"3105 permute_94" [id=3105, type=permute]; +"3106 reshape_93" [id=3106, type=reshape]; +"3107 slice_302" [id=3107, type=slice]; +"3108 slice_303" [id=3108, type=slice]; +"3109 slice_304" [id=3109, type=slice]; +"3110 slice_305" [id=3110, type=slice]; +"3111 contiguous_39" [id=3111, type=contiguous]; +"3112 layer_norm_43" [id=3112, type=layer_norm]; +"3113 add_71" [id=3113, type=add]; +"3114 quantize_per_tensor_default_127" [id=3114, type=quantize_per_tensor]; +"3115 dequantize_per_tensor_default_127" [id=3115, type=dequantize_per_tensor]; +"3116 linear_126" [id=3116, type=linear]; +"3117 gelu_20" [id=3117, type=gelu]; +"3118 quantize_per_tensor_default_128" [id=3118, type=quantize_per_tensor]; +"3119 dequantize_per_tensor_default_128" [id=3119, type=dequantize_per_tensor]; +"3120 dropout_82" [id=3120, type=dropout]; +"3121 linear_127" [id=3121, type=linear]; +"3122 dropout_83" [id=3122, type=dropout]; +"3123 layer_norm_44" [id=3123, type=layer_norm]; +"3124 add_72" [id=3124, type=add]; +"3125 linear_128" [id=3125, type=linear]; +"3126 relu__21" [id=3126, type=relu_]; +"3127 linear_129" [id=3127, type=linear]; +"3128 view_114" [id=3128, type=view]; +"3129 index_21" [id=3129, type=index]; +"3130 view_115" [id=3130, type=view]; +"3131 permute_95" [id=3131, type=permute]; +"3132 contiguous_40" [id=3132, type=contiguous]; +"3133 unsqueeze_61" [id=3133, type=unsqueeze]; +"3134 sigmoid_21" [id=3134, type=sigmoid]; +"3135 mul_42" [id=3135, type=mul]; +"3136 pad_23" [id=3136, type=pad]; +"3137 roll_20" [id=3137, type=roll]; +"3138 view_116" [id=3138, type=view]; +"3139 permute_96" [id=3139, type=permute]; +"3140 reshape_94" [id=3140, type=reshape]; +"3141 quantize_per_tensor_default_129" [id=3141, type=quantize_per_tensor]; +"3142 dequantize_per_tensor_default_129" [id=3142, type=dequantize_per_tensor]; +"3143 clone_21" [id=3143, type=clone]; +"3144 slice_306" [id=3144, type=slice]; +"3145 zero__21" [id=3145, type=zero_]; +"3146 linear_130" [id=3146, type=linear]; +"3147 reshape_95" [id=3147, type=reshape]; +"3148 permute_97" [id=3148, type=permute]; +"3149 select_63" [id=3149, type=select]; +"3150 select_64" [id=3150, type=select]; +"3151 select_65" [id=3151, type=select]; +"3152 linalg_vector_norm_42" [id=3152, type=linalg_vector_norm]; +"3153 clamp_min_42" [id=3153, type=clamp_min]; +"3154 expand_as_42" [id=3154, type=expand_as]; +"3155 div_42" [id=3155, type=div]; +"3156 quantize_per_tensor_default_130" [id=3156, type=quantize_per_tensor]; +"3157 dequantize_per_tensor_default_130" [id=3157, type=dequantize_per_tensor]; +"3158 linalg_vector_norm_43" [id=3158, type=linalg_vector_norm]; +"3159 clamp_min_43" [id=3159, type=clamp_min]; +"3160 expand_as_43" [id=3160, type=expand_as]; +"3161 div_43" [id=3161, type=div]; +"3162 quantize_per_tensor_default_131" [id=3162, type=quantize_per_tensor]; +"3163 dequantize_per_tensor_default_131" [id=3163, type=dequantize_per_tensor]; +"3164 transpose_42" [id=3164, type=transpose]; +"3165 matmul_42" [id=3165, type=matmul]; +"3166 clamp_21" [id=3166, type=clamp]; +"3167 exp_21" [id=3167, type=exp]; +"3168 mul_43" [id=3168, type=mul]; +"3169 add_73" [id=3169, type=add]; +"3170 new_zeros_10" [id=3170, type=new_zeros]; +"3171 lift_fresh_copy_90" [id=3171, type=lift_fresh_copy]; +"3172 slice_307" [id=3172, type=slice]; +"3173 slice_308" [id=3173, type=slice]; +"3174 fill__90" [id=3174, type=fill_]; +"3175 lift_fresh_copy_91" [id=3175, type=lift_fresh_copy]; +"3176 slice_309" [id=3176, type=slice]; +"3177 slice_310" [id=3177, type=slice]; +"3178 fill__91" [id=3178, type=fill_]; +"3179 lift_fresh_copy_92" [id=3179, type=lift_fresh_copy]; +"3180 slice_311" [id=3180, type=slice]; +"3181 slice_312" [id=3181, type=slice]; +"3182 fill__92" [id=3182, type=fill_]; +"3183 lift_fresh_copy_93" [id=3183, type=lift_fresh_copy]; +"3184 slice_313" [id=3184, type=slice]; +"3185 slice_314" [id=3185, type=slice]; +"3186 fill__93" [id=3186, type=fill_]; +"3187 lift_fresh_copy_94" [id=3187, type=lift_fresh_copy]; +"3188 slice_315" [id=3188, type=slice]; +"3189 slice_316" [id=3189, type=slice]; +"3190 fill__94" [id=3190, type=fill_]; +"3191 lift_fresh_copy_95" [id=3191, type=lift_fresh_copy]; +"3192 slice_317" [id=3192, type=slice]; +"3193 slice_318" [id=3193, type=slice]; +"3194 fill__95" [id=3194, type=fill_]; +"3195 lift_fresh_copy_96" [id=3195, type=lift_fresh_copy]; +"3196 slice_319" [id=3196, type=slice]; +"3197 slice_320" [id=3197, type=slice]; +"3198 fill__96" [id=3198, type=fill_]; +"3199 lift_fresh_copy_97" [id=3199, type=lift_fresh_copy]; +"3200 slice_321" [id=3200, type=slice]; +"3201 slice_322" [id=3201, type=slice]; +"3202 fill__97" [id=3202, type=fill_]; +"3203 lift_fresh_copy_98" [id=3203, type=lift_fresh_copy]; +"3204 slice_323" [id=3204, type=slice]; +"3205 slice_324" [id=3205, type=slice]; +"3206 fill__98" [id=3206, type=fill_]; +"3207 view_117" [id=3207, type=view]; +"3208 permute_98" [id=3208, type=permute]; +"3209 reshape_96" [id=3209, type=reshape]; +"3210 unsqueeze_62" [id=3210, type=unsqueeze]; +"3211 unsqueeze_63" [id=3211, type=unsqueeze]; +"3212 sub_10" [id=3212, type=sub]; +"3213 ne_10" [id=3213, type=ne]; +"3214 masked_fill_20" [id=3214, type=masked_fill]; +"3215 eq_10" [id=3215, type=eq]; +"3216 masked_fill_21" [id=3216, type=masked_fill]; +"3217 view_118" [id=3217, type=view]; +"3218 unsqueeze_64" [id=3218, type=unsqueeze]; +"3219 unsqueeze_65" [id=3219, type=unsqueeze]; +"3220 add_74" [id=3220, type=add]; +"3221 view_119" [id=3221, type=view]; +"3222 softmax_21" [id=3222, type=softmax]; +"3223 dropout_84" [id=3223, type=dropout]; +"3224 matmul_43" [id=3224, type=matmul]; +"3225 quantize_per_tensor_default_132" [id=3225, type=quantize_per_tensor]; +"3226 dequantize_per_tensor_default_132" [id=3226, type=dequantize_per_tensor]; +"3227 transpose_43" [id=3227, type=transpose]; +"3228 reshape_97" [id=3228, type=reshape]; +"3229 linear_131" [id=3229, type=linear]; +"3230 dropout_85" [id=3230, type=dropout]; +"3231 view_120" [id=3231, type=view]; +"3232 permute_99" [id=3232, type=permute]; +"3233 reshape_98" [id=3233, type=reshape]; +"3234 roll_21" [id=3234, type=roll]; +"3235 slice_325" [id=3235, type=slice]; +"3236 slice_326" [id=3236, type=slice]; +"3237 slice_327" [id=3237, type=slice]; +"3238 slice_328" [id=3238, type=slice]; +"3239 contiguous_41" [id=3239, type=contiguous]; +"3240 layer_norm_45" [id=3240, type=layer_norm]; +"3241 add_75" [id=3241, type=add]; +"3242 quantize_per_tensor_default_133" [id=3242, type=quantize_per_tensor]; +"3243 dequantize_per_tensor_default_133" [id=3243, type=dequantize_per_tensor]; +"3244 linear_132" [id=3244, type=linear]; +"3245 gelu_21" [id=3245, type=gelu]; +"3246 quantize_per_tensor_default_134" [id=3246, type=quantize_per_tensor]; +"3247 dequantize_per_tensor_default_134" [id=3247, type=dequantize_per_tensor]; +"3248 dropout_86" [id=3248, type=dropout]; +"3249 linear_133" [id=3249, type=linear]; +"3250 dropout_87" [id=3250, type=dropout]; +"3251 layer_norm_46" [id=3251, type=layer_norm]; +"3252 add_76" [id=3252, type=add]; +"3253 quantize_per_tensor_default_135" [id=3253, type=quantize_per_tensor]; +"3254 dequantize_per_tensor_default_135" [id=3254, type=dequantize_per_tensor]; +"3255 pad_24" [id=3255, type=pad]; +"3256 slice_329" [id=3256, type=slice]; +"3257 slice_330" [id=3257, type=slice]; +"3258 slice_331" [id=3258, type=slice]; +"3259 slice_332" [id=3259, type=slice]; +"3260 slice_333" [id=3260, type=slice]; +"3261 slice_334" [id=3261, type=slice]; +"3262 slice_335" [id=3262, type=slice]; +"3263 slice_336" [id=3263, type=slice]; +"3264 slice_337" [id=3264, type=slice]; +"3265 slice_338" [id=3265, type=slice]; +"3266 slice_339" [id=3266, type=slice]; +"3267 slice_340" [id=3267, type=slice]; +"3268 cat_2" [id=3268, type=cat]; +"3269 linear_134" [id=3269, type=linear]; +"3270 layer_norm_47" [id=3270, type=layer_norm]; +"3271 quantize_per_tensor_default_136" [id=3271, type=quantize_per_tensor]; +"3272 dequantize_per_tensor_default_136" [id=3272, type=dequantize_per_tensor]; +"3273 linear_135" [id=3273, type=linear]; +"3274 relu__22" [id=3274, type=relu_]; +"3275 linear_136" [id=3275, type=linear]; +"3276 view_121" [id=3276, type=view]; +"3277 index_22" [id=3277, type=index]; +"3278 view_122" [id=3278, type=view]; +"3279 permute_100" [id=3279, type=permute]; +"3280 contiguous_42" [id=3280, type=contiguous]; +"3281 unsqueeze_66" [id=3281, type=unsqueeze]; +"3282 sigmoid_22" [id=3282, type=sigmoid]; +"3283 mul_44" [id=3283, type=mul]; +"3284 pad_25" [id=3284, type=pad]; +"3285 view_123" [id=3285, type=view]; +"3286 permute_101" [id=3286, type=permute]; +"3287 reshape_99" [id=3287, type=reshape]; +"3288 clone_22" [id=3288, type=clone]; +"3289 slice_341" [id=3289, type=slice]; +"3290 zero__22" [id=3290, type=zero_]; +"3291 linear_137" [id=3291, type=linear]; +"3292 reshape_100" [id=3292, type=reshape]; +"3293 permute_102" [id=3293, type=permute]; +"3294 select_66" [id=3294, type=select]; +"3295 select_67" [id=3295, type=select]; +"3296 select_68" [id=3296, type=select]; +"3297 linalg_vector_norm_44" [id=3297, type=linalg_vector_norm]; +"3298 clamp_min_44" [id=3298, type=clamp_min]; +"3299 expand_as_44" [id=3299, type=expand_as]; +"3300 div_44" [id=3300, type=div]; +"3301 quantize_per_tensor_default_137" [id=3301, type=quantize_per_tensor]; +"3302 dequantize_per_tensor_default_137" [id=3302, type=dequantize_per_tensor]; +"3303 linalg_vector_norm_45" [id=3303, type=linalg_vector_norm]; +"3304 clamp_min_45" [id=3304, type=clamp_min]; +"3305 expand_as_45" [id=3305, type=expand_as]; +"3306 div_45" [id=3306, type=div]; +"3307 quantize_per_tensor_default_138" [id=3307, type=quantize_per_tensor]; +"3308 dequantize_per_tensor_default_138" [id=3308, type=dequantize_per_tensor]; +"3309 transpose_44" [id=3309, type=transpose]; +"3310 matmul_44" [id=3310, type=matmul]; +"3311 clamp_22" [id=3311, type=clamp]; +"3312 exp_22" [id=3312, type=exp]; +"3313 mul_45" [id=3313, type=mul]; +"3314 add_77" [id=3314, type=add]; +"3315 softmax_22" [id=3315, type=softmax]; +"3316 dropout_88" [id=3316, type=dropout]; +"3317 matmul_45" [id=3317, type=matmul]; +"3318 quantize_per_tensor_default_139" [id=3318, type=quantize_per_tensor]; +"3319 dequantize_per_tensor_default_139" [id=3319, type=dequantize_per_tensor]; +"3320 transpose_45" [id=3320, type=transpose]; +"3321 reshape_101" [id=3321, type=reshape]; +"3322 linear_138" [id=3322, type=linear]; +"3323 dropout_89" [id=3323, type=dropout]; +"3324 view_124" [id=3324, type=view]; +"3325 permute_103" [id=3325, type=permute]; +"3326 reshape_102" [id=3326, type=reshape]; +"3327 slice_342" [id=3327, type=slice]; +"3328 slice_343" [id=3328, type=slice]; +"3329 slice_344" [id=3329, type=slice]; +"3330 slice_345" [id=3330, type=slice]; +"3331 contiguous_43" [id=3331, type=contiguous]; +"3332 layer_norm_48" [id=3332, type=layer_norm]; +"3333 add_78" [id=3333, type=add]; +"3334 quantize_per_tensor_default_140" [id=3334, type=quantize_per_tensor]; +"3335 dequantize_per_tensor_default_140" [id=3335, type=dequantize_per_tensor]; +"3336 linear_139" [id=3336, type=linear]; +"3337 gelu_22" [id=3337, type=gelu]; +"3338 quantize_per_tensor_default_141" [id=3338, type=quantize_per_tensor]; +"3339 dequantize_per_tensor_default_141" [id=3339, type=dequantize_per_tensor]; +"3340 dropout_90" [id=3340, type=dropout]; +"3341 linear_140" [id=3341, type=linear]; +"3342 dropout_91" [id=3342, type=dropout]; +"3343 layer_norm_49" [id=3343, type=layer_norm]; +"3344 add_79" [id=3344, type=add]; +"3345 quantize_per_tensor_default_142" [id=3345, type=quantize_per_tensor]; +"3346 dequantize_per_tensor_default_142" [id=3346, type=dequantize_per_tensor]; +"3347 linear_141" [id=3347, type=linear]; +"3348 relu__23" [id=3348, type=relu_]; +"3349 linear_142" [id=3349, type=linear]; +"3350 view_125" [id=3350, type=view]; +"3351 index_23" [id=3351, type=index]; +"3352 view_126" [id=3352, type=view]; +"3353 permute_104" [id=3353, type=permute]; +"3354 contiguous_44" [id=3354, type=contiguous]; +"3355 unsqueeze_67" [id=3355, type=unsqueeze]; +"3356 sigmoid_23" [id=3356, type=sigmoid]; +"3357 mul_46" [id=3357, type=mul]; +"3358 pad_26" [id=3358, type=pad]; +"3359 view_127" [id=3359, type=view]; +"3360 permute_105" [id=3360, type=permute]; +"3361 reshape_103" [id=3361, type=reshape]; +"3362 clone_23" [id=3362, type=clone]; +"3363 slice_346" [id=3363, type=slice]; +"3364 zero__23" [id=3364, type=zero_]; +"3365 linear_143" [id=3365, type=linear]; +"3366 reshape_104" [id=3366, type=reshape]; +"3367 permute_106" [id=3367, type=permute]; +"3368 select_69" [id=3368, type=select]; +"3369 select_70" [id=3369, type=select]; +"3370 select_71" [id=3370, type=select]; +"3371 linalg_vector_norm_46" [id=3371, type=linalg_vector_norm]; +"3372 clamp_min_46" [id=3372, type=clamp_min]; +"3373 expand_as_46" [id=3373, type=expand_as]; +"3374 div_46" [id=3374, type=div]; +"3375 quantize_per_tensor_default_143" [id=3375, type=quantize_per_tensor]; +"3376 dequantize_per_tensor_default_143" [id=3376, type=dequantize_per_tensor]; +"3377 linalg_vector_norm_47" [id=3377, type=linalg_vector_norm]; +"3378 clamp_min_47" [id=3378, type=clamp_min]; +"3379 expand_as_47" [id=3379, type=expand_as]; +"3380 div_47" [id=3380, type=div]; +"3381 quantize_per_tensor_default_144" [id=3381, type=quantize_per_tensor]; +"3382 dequantize_per_tensor_default_144" [id=3382, type=dequantize_per_tensor]; +"3383 transpose_46" [id=3383, type=transpose]; +"3384 matmul_46" [id=3384, type=matmul]; +"3385 clamp_23" [id=3385, type=clamp]; +"3386 exp_23" [id=3386, type=exp]; +"3387 mul_47" [id=3387, type=mul]; +"3388 add_80" [id=3388, type=add]; +"3389 softmax_23" [id=3389, type=softmax]; +"3390 dropout_92" [id=3390, type=dropout]; +"3391 matmul_47" [id=3391, type=matmul]; +"3392 quantize_per_tensor_default_145" [id=3392, type=quantize_per_tensor]; +"3393 dequantize_per_tensor_default_145" [id=3393, type=dequantize_per_tensor]; +"3394 transpose_47" [id=3394, type=transpose]; +"3395 reshape_105" [id=3395, type=reshape]; +"3396 linear_144" [id=3396, type=linear]; +"3397 dropout_93" [id=3397, type=dropout]; +"3398 view_128" [id=3398, type=view]; +"3399 permute_107" [id=3399, type=permute]; +"3400 reshape_106" [id=3400, type=reshape]; +"3401 slice_347" [id=3401, type=slice]; +"3402 slice_348" [id=3402, type=slice]; +"3403 slice_349" [id=3403, type=slice]; +"3404 slice_350" [id=3404, type=slice]; +"3405 contiguous_45" [id=3405, type=contiguous]; +"3406 layer_norm_50" [id=3406, type=layer_norm]; +"3407 add_81" [id=3407, type=add]; +"3408 quantize_per_tensor_default_146" [id=3408, type=quantize_per_tensor]; +"3409 dequantize_per_tensor_default_146" [id=3409, type=dequantize_per_tensor]; +"3410 linear_145" [id=3410, type=linear]; +"3411 gelu_23" [id=3411, type=gelu]; +"3412 quantize_per_tensor_default_147" [id=3412, type=quantize_per_tensor]; +"3413 dequantize_per_tensor_default_147" [id=3413, type=dequantize_per_tensor]; +"3414 dropout_94" [id=3414, type=dropout]; +"3415 linear_146" [id=3415, type=linear]; +"3416 dropout_95" [id=3416, type=dropout]; +"3417 layer_norm_51" [id=3417, type=layer_norm]; +"3418 add_82" [id=3418, type=add]; +"3419 layer_norm_52" [id=3419, type=layer_norm]; +"3420 permute_108" [id=3420, type=permute]; +"3421 adaptive_avg_pool2d" [id=3421, type=adaptive_avg_pool2d]; +"3422 quantize_per_tensor_default_148" [id=3422, type=quantize_per_tensor]; +"3423 dequantize_per_tensor_default_148" [id=3423, type=dequantize_per_tensor]; +"3424 flatten" [id=3424, type=flatten]; +"3425 linear_147" [id=3425, type=linear]; +"3426 output_1" [id=3426, type=output]; +"0 _scale_0" -> "3 dequantize_per_channel_default" [label="(288,)", style=solid]; +"1 _zero_point_0" -> "3 dequantize_per_channel_default" [label="(288,)", style=solid]; +"2 _frozen_param0" -> "3 dequantize_per_channel_default" [label="(288, 96)", style=solid]; +"3 dequantize_per_channel_default" -> "1021 linear_2" [label=None, style=solid]; +"4 _scale_1" -> "7 dequantize_per_channel_default_1" [label="(96,)", style=solid]; +"5 _zero_point_1" -> "7 dequantize_per_channel_default_1" [label="(96,)", style=solid]; +"6 _frozen_param1" -> "7 dequantize_per_channel_default_1" [label="(96, 96)", style=solid]; +"7 dequantize_per_channel_default_1" -> "1052 linear_3" [label=None, style=solid]; +"8 features_1_0_attn_qkv_bias" -> "1018 clone" [label="(288,)", style=solid]; +"9 features_1_0_attn_proj_bias" -> "1052 linear_3" [label="(96,)", style=solid]; +"10 features_1_0_attn_logit_scale" -> "1041 clamp" [label="(3, 1, 1)", style=solid]; +"11 _scale_2" -> "14 dequantize_per_channel_default_2" [label="(288,)", style=solid]; +"12 _zero_point_2" -> "14 dequantize_per_channel_default_2" [label="(288,)", style=solid]; +"13 _frozen_param2" -> "14 dequantize_per_channel_default_2" [label="(288, 96)", style=solid]; +"14 dequantize_per_channel_default_2" -> "1093 linear_8" [label=None, style=solid]; +"15 _scale_3" -> "18 dequantize_per_channel_default_3" [label="(96,)", style=solid]; +"16 _zero_point_3" -> "18 dequantize_per_channel_default_3" [label="(96,)", style=solid]; +"17 _frozen_param3" -> "18 dequantize_per_channel_default_3" [label="(96, 96)", style=solid]; +"18 dequantize_per_channel_default_3" -> "1176 linear_9" [label=None, style=solid]; +"19 features_1_1_attn_qkv_bias" -> "1090 clone_1" [label="(288,)", style=solid]; +"20 features_1_1_attn_proj_bias" -> "1176 linear_9" [label="(96,)", style=solid]; +"21 features_1_1_attn_logit_scale" -> "1113 clamp_1" [label="(3, 1, 1)", style=solid]; +"22 _scale_4" -> "25 dequantize_per_channel_default_4" [label="(576,)", style=solid]; +"23 _zero_point_4" -> "25 dequantize_per_channel_default_4" [label="(576,)", style=solid]; +"24 _frozen_param4" -> "25 dequantize_per_channel_default_4" [label="(576, 192)", style=solid]; +"25 dequantize_per_channel_default_4" -> "1235 linear_15" [label=None, style=solid]; +"26 _scale_5" -> "29 dequantize_per_channel_default_5" [label="(192,)", style=solid]; +"27 _zero_point_5" -> "29 dequantize_per_channel_default_5" [label="(192,)", style=solid]; +"28 _frozen_param5" -> "29 dequantize_per_channel_default_5" [label="(192, 192)", style=solid]; +"29 dequantize_per_channel_default_5" -> "1266 linear_16" [label=None, style=solid]; +"30 features_3_0_attn_qkv_bias" -> "1232 clone_2" [label="(576,)", style=solid]; +"31 features_3_0_attn_proj_bias" -> "1266 linear_16" [label="(192,)", style=solid]; +"32 features_3_0_attn_logit_scale" -> "1255 clamp_2" [label="(6, 1, 1)", style=solid]; +"33 _scale_6" -> "36 dequantize_per_channel_default_6" [label="(576,)", style=solid]; +"34 _zero_point_6" -> "36 dequantize_per_channel_default_6" [label="(576,)", style=solid]; +"35 _frozen_param6" -> "36 dequantize_per_channel_default_6" [label="(576, 192)", style=solid]; +"36 dequantize_per_channel_default_6" -> "1310 linear_21" [label=None, style=solid]; +"37 _scale_7" -> "40 dequantize_per_channel_default_7" [label="(192,)", style=solid]; +"38 _zero_point_7" -> "40 dequantize_per_channel_default_7" [label="(192,)", style=solid]; +"39 _frozen_param7" -> "40 dequantize_per_channel_default_7" [label="(192, 192)", style=solid]; +"40 dequantize_per_channel_default_7" -> "1393 linear_22" [label=None, style=solid]; +"41 features_3_1_attn_qkv_bias" -> "1307 clone_3" [label="(576,)", style=solid]; +"42 features_3_1_attn_proj_bias" -> "1393 linear_22" [label="(192,)", style=solid]; +"43 features_3_1_attn_logit_scale" -> "1330 clamp_3" [label="(6, 1, 1)", style=solid]; +"44 _scale_8" -> "47 dequantize_per_channel_default_8" [label="(1152,)", style=solid]; +"45 _zero_point_8" -> "47 dequantize_per_channel_default_8" [label="(1152,)", style=solid]; +"46 _frozen_param8" -> "47 dequantize_per_channel_default_8" [label="(1152, 384)", style=solid]; +"47 dequantize_per_channel_default_8" -> "1455 linear_28" [label=None, style=solid]; +"48 _scale_9" -> "51 dequantize_per_channel_default_9" [label="(384,)", style=solid]; +"49 _zero_point_9" -> "51 dequantize_per_channel_default_9" [label="(384,)", style=solid]; +"50 _frozen_param9" -> "51 dequantize_per_channel_default_9" [label="(384, 384)", style=solid]; +"51 dequantize_per_channel_default_9" -> "1486 linear_29" [label=None, style=solid]; +"52 features_5_0_attn_qkv_bias" -> "1452 clone_4" [label="(1152,)", style=solid]; +"53 features_5_0_attn_proj_bias" -> "1486 linear_29" [label="(384,)", style=solid]; +"54 features_5_0_attn_logit_scale" -> "1475 clamp_4" [label="(12, 1, 1)", style=solid]; +"55 _scale_10" -> "58 dequantize_per_channel_default_10" [label="(1152,)", style=solid]; +"56 _zero_point_10" -> "58 dequantize_per_channel_default_10" [label="(1152,)", style=solid]; +"57 _frozen_param10" -> "58 dequantize_per_channel_default_10" [label="(1152, 384)", style=solid]; +"58 dequantize_per_channel_default_10" -> "1530 linear_34" [label=None, style=solid]; +"59 _scale_11" -> "62 dequantize_per_channel_default_11" [label="(384,)", style=solid]; +"60 _zero_point_11" -> "62 dequantize_per_channel_default_11" [label="(384,)", style=solid]; +"61 _frozen_param11" -> "62 dequantize_per_channel_default_11" [label="(384, 384)", style=solid]; +"62 dequantize_per_channel_default_11" -> "1613 linear_35" [label=None, style=solid]; +"63 features_5_1_attn_qkv_bias" -> "1527 clone_5" [label="(1152,)", style=solid]; +"64 features_5_1_attn_proj_bias" -> "1613 linear_35" [label="(384,)", style=solid]; +"65 features_5_1_attn_logit_scale" -> "1550 clamp_5" [label="(12, 1, 1)", style=solid]; +"66 _scale_12" -> "69 dequantize_per_channel_default_12" [label="(1152,)", style=solid]; +"67 _zero_point_12" -> "69 dequantize_per_channel_default_12" [label="(1152,)", style=solid]; +"68 _frozen_param12" -> "69 dequantize_per_channel_default_12" [label="(1152, 384)", style=solid]; +"69 dequantize_per_channel_default_12" -> "1657 linear_40" [label=None, style=solid]; +"70 _scale_13" -> "73 dequantize_per_channel_default_13" [label="(384,)", style=solid]; +"71 _zero_point_13" -> "73 dequantize_per_channel_default_13" [label="(384,)", style=solid]; +"72 _frozen_param13" -> "73 dequantize_per_channel_default_13" [label="(384, 384)", style=solid]; +"73 dequantize_per_channel_default_13" -> "1688 linear_41" [label=None, style=solid]; +"74 features_5_2_attn_qkv_bias" -> "1654 clone_6" [label="(1152,)", style=solid]; +"75 features_5_2_attn_proj_bias" -> "1688 linear_41" [label="(384,)", style=solid]; +"76 features_5_2_attn_logit_scale" -> "1677 clamp_6" [label="(12, 1, 1)", style=solid]; +"77 _scale_14" -> "80 dequantize_per_channel_default_14" [label="(1152,)", style=solid]; +"78 _zero_point_14" -> "80 dequantize_per_channel_default_14" [label="(1152,)", style=solid]; +"79 _frozen_param14" -> "80 dequantize_per_channel_default_14" [label="(1152, 384)", style=solid]; +"80 dequantize_per_channel_default_14" -> "1732 linear_46" [label=None, style=solid]; +"81 _scale_15" -> "84 dequantize_per_channel_default_15" [label="(384,)", style=solid]; +"82 _zero_point_15" -> "84 dequantize_per_channel_default_15" [label="(384,)", style=solid]; +"83 _frozen_param15" -> "84 dequantize_per_channel_default_15" [label="(384, 384)", style=solid]; +"84 dequantize_per_channel_default_15" -> "1815 linear_47" [label=None, style=solid]; +"85 features_5_3_attn_qkv_bias" -> "1729 clone_7" [label="(1152,)", style=solid]; +"86 features_5_3_attn_proj_bias" -> "1815 linear_47" [label="(384,)", style=solid]; +"87 features_5_3_attn_logit_scale" -> "1752 clamp_7" [label="(12, 1, 1)", style=solid]; +"88 _scale_16" -> "91 dequantize_per_channel_default_16" [label="(1152,)", style=solid]; +"89 _zero_point_16" -> "91 dequantize_per_channel_default_16" [label="(1152,)", style=solid]; +"90 _frozen_param16" -> "91 dequantize_per_channel_default_16" [label="(1152, 384)", style=solid]; +"91 dequantize_per_channel_default_16" -> "1859 linear_52" [label=None, style=solid]; +"92 _scale_17" -> "95 dequantize_per_channel_default_17" [label="(384,)", style=solid]; +"93 _zero_point_17" -> "95 dequantize_per_channel_default_17" [label="(384,)", style=solid]; +"94 _frozen_param17" -> "95 dequantize_per_channel_default_17" [label="(384, 384)", style=solid]; +"95 dequantize_per_channel_default_17" -> "1890 linear_53" [label=None, style=solid]; +"96 features_5_4_attn_qkv_bias" -> "1856 clone_8" [label="(1152,)", style=solid]; +"97 features_5_4_attn_proj_bias" -> "1890 linear_53" [label="(384,)", style=solid]; +"98 features_5_4_attn_logit_scale" -> "1879 clamp_8" [label="(12, 1, 1)", style=solid]; +"99 _scale_18" -> "102 dequantize_per_channel_default_18" [label="(1152,)", style=solid]; +"100 _zero_point_18" -> "102 dequantize_per_channel_default_18" [label="(1152,)", style=solid]; +"101 _frozen_param18" -> "102 dequantize_per_channel_default_18" [label="(1152, 384)", style=solid]; +"102 dequantize_per_channel_default_18" -> "1934 linear_58" [label=None, style=solid]; +"103 _scale_19" -> "106 dequantize_per_channel_default_19" [label="(384,)", style=solid]; +"104 _zero_point_19" -> "106 dequantize_per_channel_default_19" [label="(384,)", style=solid]; +"105 _frozen_param19" -> "106 dequantize_per_channel_default_19" [label="(384, 384)", style=solid]; +"106 dequantize_per_channel_default_19" -> "2017 linear_59" [label=None, style=solid]; +"107 features_5_5_attn_qkv_bias" -> "1931 clone_9" [label="(1152,)", style=solid]; +"108 features_5_5_attn_proj_bias" -> "2017 linear_59" [label="(384,)", style=solid]; +"109 features_5_5_attn_logit_scale" -> "1954 clamp_9" [label="(12, 1, 1)", style=solid]; +"110 _scale_20" -> "113 dequantize_per_channel_default_20" [label="(1152,)", style=solid]; +"111 _zero_point_20" -> "113 dequantize_per_channel_default_20" [label="(1152,)", style=solid]; +"112 _frozen_param20" -> "113 dequantize_per_channel_default_20" [label="(1152, 384)", style=solid]; +"113 dequantize_per_channel_default_20" -> "2061 linear_64" [label=None, style=solid]; +"114 _scale_21" -> "117 dequantize_per_channel_default_21" [label="(384,)", style=solid]; +"115 _zero_point_21" -> "117 dequantize_per_channel_default_21" [label="(384,)", style=solid]; +"116 _frozen_param21" -> "117 dequantize_per_channel_default_21" [label="(384, 384)", style=solid]; +"117 dequantize_per_channel_default_21" -> "2092 linear_65" [label=None, style=solid]; +"118 features_5_6_attn_qkv_bias" -> "2058 clone_10" [label="(1152,)", style=solid]; +"119 features_5_6_attn_proj_bias" -> "2092 linear_65" [label="(384,)", style=solid]; +"120 features_5_6_attn_logit_scale" -> "2081 clamp_10" [label="(12, 1, 1)", style=solid]; +"121 _scale_22" -> "124 dequantize_per_channel_default_22" [label="(1152,)", style=solid]; +"122 _zero_point_22" -> "124 dequantize_per_channel_default_22" [label="(1152,)", style=solid]; +"123 _frozen_param22" -> "124 dequantize_per_channel_default_22" [label="(1152, 384)", style=solid]; +"124 dequantize_per_channel_default_22" -> "2136 linear_70" [label=None, style=solid]; +"125 _scale_23" -> "128 dequantize_per_channel_default_23" [label="(384,)", style=solid]; +"126 _zero_point_23" -> "128 dequantize_per_channel_default_23" [label="(384,)", style=solid]; +"127 _frozen_param23" -> "128 dequantize_per_channel_default_23" [label="(384, 384)", style=solid]; +"128 dequantize_per_channel_default_23" -> "2219 linear_71" [label=None, style=solid]; +"129 features_5_7_attn_qkv_bias" -> "2133 clone_11" [label="(1152,)", style=solid]; +"130 features_5_7_attn_proj_bias" -> "2219 linear_71" [label="(384,)", style=solid]; +"131 features_5_7_attn_logit_scale" -> "2156 clamp_11" [label="(12, 1, 1)", style=solid]; +"132 _scale_24" -> "135 dequantize_per_channel_default_24" [label="(1152,)", style=solid]; +"133 _zero_point_24" -> "135 dequantize_per_channel_default_24" [label="(1152,)", style=solid]; +"134 _frozen_param24" -> "135 dequantize_per_channel_default_24" [label="(1152, 384)", style=solid]; +"135 dequantize_per_channel_default_24" -> "2263 linear_76" [label=None, style=solid]; +"136 _scale_25" -> "139 dequantize_per_channel_default_25" [label="(384,)", style=solid]; +"137 _zero_point_25" -> "139 dequantize_per_channel_default_25" [label="(384,)", style=solid]; +"138 _frozen_param25" -> "139 dequantize_per_channel_default_25" [label="(384, 384)", style=solid]; +"139 dequantize_per_channel_default_25" -> "2294 linear_77" [label=None, style=solid]; +"140 features_5_8_attn_qkv_bias" -> "2260 clone_12" [label="(1152,)", style=solid]; +"141 features_5_8_attn_proj_bias" -> "2294 linear_77" [label="(384,)", style=solid]; +"142 features_5_8_attn_logit_scale" -> "2283 clamp_12" [label="(12, 1, 1)", style=solid]; +"143 _scale_26" -> "146 dequantize_per_channel_default_26" [label="(1152,)", style=solid]; +"144 _zero_point_26" -> "146 dequantize_per_channel_default_26" [label="(1152,)", style=solid]; +"145 _frozen_param26" -> "146 dequantize_per_channel_default_26" [label="(1152, 384)", style=solid]; +"146 dequantize_per_channel_default_26" -> "2338 linear_82" [label=None, style=solid]; +"147 _scale_27" -> "150 dequantize_per_channel_default_27" [label="(384,)", style=solid]; +"148 _zero_point_27" -> "150 dequantize_per_channel_default_27" [label="(384,)", style=solid]; +"149 _frozen_param27" -> "150 dequantize_per_channel_default_27" [label="(384, 384)", style=solid]; +"150 dequantize_per_channel_default_27" -> "2421 linear_83" [label=None, style=solid]; +"151 features_5_9_attn_qkv_bias" -> "2335 clone_13" [label="(1152,)", style=solid]; +"152 features_5_9_attn_proj_bias" -> "2421 linear_83" [label="(384,)", style=solid]; +"153 features_5_9_attn_logit_scale" -> "2358 clamp_13" [label="(12, 1, 1)", style=solid]; +"154 _scale_28" -> "157 dequantize_per_channel_default_28" [label="(1152,)", style=solid]; +"155 _zero_point_28" -> "157 dequantize_per_channel_default_28" [label="(1152,)", style=solid]; +"156 _frozen_param28" -> "157 dequantize_per_channel_default_28" [label="(1152, 384)", style=solid]; +"157 dequantize_per_channel_default_28" -> "2465 linear_88" [label=None, style=solid]; +"158 _scale_29" -> "161 dequantize_per_channel_default_29" [label="(384,)", style=solid]; +"159 _zero_point_29" -> "161 dequantize_per_channel_default_29" [label="(384,)", style=solid]; +"160 _frozen_param29" -> "161 dequantize_per_channel_default_29" [label="(384, 384)", style=solid]; +"161 dequantize_per_channel_default_29" -> "2496 linear_89" [label=None, style=solid]; +"162 features_5_10_attn_qkv_bias" -> "2462 clone_14" [label="(1152,)", style=solid]; +"163 features_5_10_attn_proj_bias" -> "2496 linear_89" [label="(384,)", style=solid]; +"164 features_5_10_attn_logit_scale" -> "2485 clamp_14" [label="(12, 1, 1)", style=solid]; +"165 _scale_30" -> "168 dequantize_per_channel_default_30" [label="(1152,)", style=solid]; +"166 _zero_point_30" -> "168 dequantize_per_channel_default_30" [label="(1152,)", style=solid]; +"167 _frozen_param30" -> "168 dequantize_per_channel_default_30" [label="(1152, 384)", style=solid]; +"168 dequantize_per_channel_default_30" -> "2540 linear_94" [label=None, style=solid]; +"169 _scale_31" -> "172 dequantize_per_channel_default_31" [label="(384,)", style=solid]; +"170 _zero_point_31" -> "172 dequantize_per_channel_default_31" [label="(384,)", style=solid]; +"171 _frozen_param31" -> "172 dequantize_per_channel_default_31" [label="(384, 384)", style=solid]; +"172 dequantize_per_channel_default_31" -> "2623 linear_95" [label=None, style=solid]; +"173 features_5_11_attn_qkv_bias" -> "2537 clone_15" [label="(1152,)", style=solid]; +"174 features_5_11_attn_proj_bias" -> "2623 linear_95" [label="(384,)", style=solid]; +"175 features_5_11_attn_logit_scale" -> "2560 clamp_15" [label="(12, 1, 1)", style=solid]; +"176 _scale_32" -> "179 dequantize_per_channel_default_32" [label="(1152,)", style=solid]; +"177 _zero_point_32" -> "179 dequantize_per_channel_default_32" [label="(1152,)", style=solid]; +"178 _frozen_param32" -> "179 dequantize_per_channel_default_32" [label="(1152, 384)", style=solid]; +"179 dequantize_per_channel_default_32" -> "2667 linear_100" [label=None, style=solid]; +"180 _scale_33" -> "183 dequantize_per_channel_default_33" [label="(384,)", style=solid]; +"181 _zero_point_33" -> "183 dequantize_per_channel_default_33" [label="(384,)", style=solid]; +"182 _frozen_param33" -> "183 dequantize_per_channel_default_33" [label="(384, 384)", style=solid]; +"183 dequantize_per_channel_default_33" -> "2698 linear_101" [label=None, style=solid]; +"184 features_5_12_attn_qkv_bias" -> "2664 clone_16" [label="(1152,)", style=solid]; +"185 features_5_12_attn_proj_bias" -> "2698 linear_101" [label="(384,)", style=solid]; +"186 features_5_12_attn_logit_scale" -> "2687 clamp_16" [label="(12, 1, 1)", style=solid]; +"187 _scale_34" -> "190 dequantize_per_channel_default_34" [label="(1152,)", style=solid]; +"188 _zero_point_34" -> "190 dequantize_per_channel_default_34" [label="(1152,)", style=solid]; +"189 _frozen_param34" -> "190 dequantize_per_channel_default_34" [label="(1152, 384)", style=solid]; +"190 dequantize_per_channel_default_34" -> "2742 linear_106" [label=None, style=solid]; +"191 _scale_35" -> "194 dequantize_per_channel_default_35" [label="(384,)", style=solid]; +"192 _zero_point_35" -> "194 dequantize_per_channel_default_35" [label="(384,)", style=solid]; +"193 _frozen_param35" -> "194 dequantize_per_channel_default_35" [label="(384, 384)", style=solid]; +"194 dequantize_per_channel_default_35" -> "2825 linear_107" [label=None, style=solid]; +"195 features_5_13_attn_qkv_bias" -> "2739 clone_17" [label="(1152,)", style=solid]; +"196 features_5_13_attn_proj_bias" -> "2825 linear_107" [label="(384,)", style=solid]; +"197 features_5_13_attn_logit_scale" -> "2762 clamp_17" [label="(12, 1, 1)", style=solid]; +"198 _scale_36" -> "201 dequantize_per_channel_default_36" [label="(1152,)", style=solid]; +"199 _zero_point_36" -> "201 dequantize_per_channel_default_36" [label="(1152,)", style=solid]; +"200 _frozen_param36" -> "201 dequantize_per_channel_default_36" [label="(1152, 384)", style=solid]; +"201 dequantize_per_channel_default_36" -> "2869 linear_112" [label=None, style=solid]; +"202 _scale_37" -> "205 dequantize_per_channel_default_37" [label="(384,)", style=solid]; +"203 _zero_point_37" -> "205 dequantize_per_channel_default_37" [label="(384,)", style=solid]; +"204 _frozen_param37" -> "205 dequantize_per_channel_default_37" [label="(384, 384)", style=solid]; +"205 dequantize_per_channel_default_37" -> "2900 linear_113" [label=None, style=solid]; +"206 features_5_14_attn_qkv_bias" -> "2866 clone_18" [label="(1152,)", style=solid]; +"207 features_5_14_attn_proj_bias" -> "2900 linear_113" [label="(384,)", style=solid]; +"208 features_5_14_attn_logit_scale" -> "2889 clamp_18" [label="(12, 1, 1)", style=solid]; +"209 _scale_38" -> "212 dequantize_per_channel_default_38" [label="(1152,)", style=solid]; +"210 _zero_point_38" -> "212 dequantize_per_channel_default_38" [label="(1152,)", style=solid]; +"211 _frozen_param38" -> "212 dequantize_per_channel_default_38" [label="(1152, 384)", style=solid]; +"212 dequantize_per_channel_default_38" -> "2944 linear_118" [label=None, style=solid]; +"213 _scale_39" -> "216 dequantize_per_channel_default_39" [label="(384,)", style=solid]; +"214 _zero_point_39" -> "216 dequantize_per_channel_default_39" [label="(384,)", style=solid]; +"215 _frozen_param39" -> "216 dequantize_per_channel_default_39" [label="(384, 384)", style=solid]; +"216 dequantize_per_channel_default_39" -> "3027 linear_119" [label=None, style=solid]; +"217 features_5_15_attn_qkv_bias" -> "2941 clone_19" [label="(1152,)", style=solid]; +"218 features_5_15_attn_proj_bias" -> "3027 linear_119" [label="(384,)", style=solid]; +"219 features_5_15_attn_logit_scale" -> "2964 clamp_19" [label="(12, 1, 1)", style=solid]; +"220 _scale_40" -> "223 dequantize_per_channel_default_40" [label="(1152,)", style=solid]; +"221 _zero_point_40" -> "223 dequantize_per_channel_default_40" [label="(1152,)", style=solid]; +"222 _frozen_param40" -> "223 dequantize_per_channel_default_40" [label="(1152, 384)", style=solid]; +"223 dequantize_per_channel_default_40" -> "3071 linear_124" [label=None, style=solid]; +"224 _scale_41" -> "227 dequantize_per_channel_default_41" [label="(384,)", style=solid]; +"225 _zero_point_41" -> "227 dequantize_per_channel_default_41" [label="(384,)", style=solid]; +"226 _frozen_param41" -> "227 dequantize_per_channel_default_41" [label="(384, 384)", style=solid]; +"227 dequantize_per_channel_default_41" -> "3102 linear_125" [label=None, style=solid]; +"228 features_5_16_attn_qkv_bias" -> "3068 clone_20" [label="(1152,)", style=solid]; +"229 features_5_16_attn_proj_bias" -> "3102 linear_125" [label="(384,)", style=solid]; +"230 features_5_16_attn_logit_scale" -> "3091 clamp_20" [label="(12, 1, 1)", style=solid]; +"231 _scale_42" -> "234 dequantize_per_channel_default_42" [label="(1152,)", style=solid]; +"232 _zero_point_42" -> "234 dequantize_per_channel_default_42" [label="(1152,)", style=solid]; +"233 _frozen_param42" -> "234 dequantize_per_channel_default_42" [label="(1152, 384)", style=solid]; +"234 dequantize_per_channel_default_42" -> "3146 linear_130" [label=None, style=solid]; +"235 _scale_43" -> "238 dequantize_per_channel_default_43" [label="(384,)", style=solid]; +"236 _zero_point_43" -> "238 dequantize_per_channel_default_43" [label="(384,)", style=solid]; +"237 _frozen_param43" -> "238 dequantize_per_channel_default_43" [label="(384, 384)", style=solid]; +"238 dequantize_per_channel_default_43" -> "3229 linear_131" [label=None, style=solid]; +"239 features_5_17_attn_qkv_bias" -> "3143 clone_21" [label="(1152,)", style=solid]; +"240 features_5_17_attn_proj_bias" -> "3229 linear_131" [label="(384,)", style=solid]; +"241 features_5_17_attn_logit_scale" -> "3166 clamp_21" [label="(12, 1, 1)", style=solid]; +"242 _scale_44" -> "245 dequantize_per_channel_default_44" [label="(2304,)", style=solid]; +"243 _zero_point_44" -> "245 dequantize_per_channel_default_44" [label="(2304,)", style=solid]; +"244 _frozen_param44" -> "245 dequantize_per_channel_default_44" [label="(2304, 768)", style=solid]; +"245 dequantize_per_channel_default_44" -> "3291 linear_137" [label=None, style=solid]; +"246 _scale_45" -> "249 dequantize_per_channel_default_45" [label="(768,)", style=solid]; +"247 _zero_point_45" -> "249 dequantize_per_channel_default_45" [label="(768,)", style=solid]; +"248 _frozen_param45" -> "249 dequantize_per_channel_default_45" [label="(768, 768)", style=solid]; +"249 dequantize_per_channel_default_45" -> "3322 linear_138" [label=None, style=solid]; +"250 features_7_0_attn_qkv_bias" -> "3288 clone_22" [label="(2304,)", style=solid]; +"251 features_7_0_attn_proj_bias" -> "3322 linear_138" [label="(768,)", style=solid]; +"252 features_7_0_attn_logit_scale" -> "3311 clamp_22" [label="(24, 1, 1)", style=solid]; +"253 _scale_46" -> "256 dequantize_per_channel_default_46" [label="(2304,)", style=solid]; +"254 _zero_point_46" -> "256 dequantize_per_channel_default_46" [label="(2304,)", style=solid]; +"255 _frozen_param46" -> "256 dequantize_per_channel_default_46" [label="(2304, 768)", style=solid]; +"256 dequantize_per_channel_default_46" -> "3365 linear_143" [label=None, style=solid]; +"257 _scale_47" -> "260 dequantize_per_channel_default_47" [label="(768,)", style=solid]; +"258 _zero_point_47" -> "260 dequantize_per_channel_default_47" [label="(768,)", style=solid]; +"259 _frozen_param47" -> "260 dequantize_per_channel_default_47" [label="(768, 768)", style=solid]; +"260 dequantize_per_channel_default_47" -> "3396 linear_144" [label=None, style=solid]; +"261 features_7_1_attn_qkv_bias" -> "3362 clone_23" [label="(2304,)", style=solid]; +"262 features_7_1_attn_proj_bias" -> "3396 linear_144" [label="(768,)", style=solid]; +"263 features_7_1_attn_logit_scale" -> "3385 clamp_23" [label="(24, 1, 1)", style=solid]; +"264 _scale_48" -> "267 dequantize_per_channel_default_48" [label="(96,)", style=solid]; +"265 _zero_point_48" -> "267 dequantize_per_channel_default_48" [label="(96,)", style=solid]; +"266 _frozen_param48" -> "267 dequantize_per_channel_default_48" [label="(96, 3, 4, 4)", style=solid]; +"267 dequantize_per_channel_default_48" -> "998 conv2d" [label=None, style=solid]; +"268 features_0_0_bias" -> "998 conv2d" [label="(96,)", style=solid]; +"269 features_0_2_weight" -> "1000 layer_norm" [label="(96,)", style=solid]; +"270 features_0_2_bias" -> "1000 layer_norm" [label="(96,)", style=solid]; +"271 _scale_49" -> "274 dequantize_per_channel_default_49" [label="(512,)", style=solid]; +"272 _zero_point_49" -> "274 dequantize_per_channel_default_49" [label="(512,)", style=solid]; +"273 _frozen_param49" -> "274 dequantize_per_channel_default_49" [label="(512, 2)", style=solid]; +"274 dequantize_per_channel_default_49" -> "1003 linear" [label=None, style=solid]; +"275 features_1_0_attn_cpb_mlp_0_bias" -> "1003 linear" [label="(512,)", style=solid]; +"276 _scale_50" -> "279 dequantize_per_channel_default_50" [label="(3,)", style=solid]; +"277 _zero_point_50" -> "279 dequantize_per_channel_default_50" [label="(3,)", style=solid]; +"278 _frozen_param50" -> "279 dequantize_per_channel_default_50" [label="(3, 512)", style=solid]; +"279 dequantize_per_channel_default_50" -> "1005 linear_1" [label=None, style=solid]; +"280 features_1_0_norm1_weight" -> "1059 layer_norm_1" [label="(96,)", style=solid]; +"281 features_1_0_norm1_bias" -> "1059 layer_norm_1" [label="(96,)", style=solid]; +"282 _scale_51" -> "285 dequantize_per_channel_default_51" [label="(384,)", style=solid]; +"283 _zero_point_51" -> "285 dequantize_per_channel_default_51" [label="(384,)", style=solid]; +"284 _frozen_param51" -> "285 dequantize_per_channel_default_51" [label="(384, 96)", style=solid]; +"285 dequantize_per_channel_default_51" -> "1063 linear_4" [label=None, style=solid]; +"286 features_1_0_mlp_0_bias" -> "1063 linear_4" [label="(384,)", style=solid]; +"287 _scale_52" -> "290 dequantize_per_channel_default_52" [label="(96,)", style=solid]; +"288 _zero_point_52" -> "290 dequantize_per_channel_default_52" [label="(96,)", style=solid]; +"289 _frozen_param52" -> "290 dequantize_per_channel_default_52" [label="(96, 384)", style=solid]; +"290 dequantize_per_channel_default_52" -> "1068 linear_5" [label=None, style=solid]; +"291 features_1_0_mlp_3_bias" -> "1068 linear_5" [label="(96,)", style=solid]; +"292 features_1_0_norm2_weight" -> "1070 layer_norm_2" [label="(96,)", style=solid]; +"293 features_1_0_norm2_bias" -> "1070 layer_norm_2" [label="(96,)", style=solid]; +"294 _scale_53" -> "297 dequantize_per_channel_default_53" [label="(512,)", style=solid]; +"295 _zero_point_53" -> "297 dequantize_per_channel_default_53" [label="(512,)", style=solid]; +"296 _frozen_param53" -> "297 dequantize_per_channel_default_53" [label="(512, 2)", style=solid]; +"297 dequantize_per_channel_default_53" -> "1072 linear_6" [label=None, style=solid]; +"298 features_1_1_attn_cpb_mlp_0_bias" -> "1072 linear_6" [label="(512,)", style=solid]; +"299 _scale_54" -> "302 dequantize_per_channel_default_54" [label="(3,)", style=solid]; +"300 _zero_point_54" -> "302 dequantize_per_channel_default_54" [label="(3,)", style=solid]; +"301 _frozen_param54" -> "302 dequantize_per_channel_default_54" [label="(3, 512)", style=solid]; +"302 dequantize_per_channel_default_54" -> "1074 linear_7" [label=None, style=solid]; +"303 features_1_1_norm1_weight" -> "1184 layer_norm_3" [label="(96,)", style=solid]; +"304 features_1_1_norm1_bias" -> "1184 layer_norm_3" [label="(96,)", style=solid]; +"305 _scale_55" -> "308 dequantize_per_channel_default_55" [label="(384,)", style=solid]; +"306 _zero_point_55" -> "308 dequantize_per_channel_default_55" [label="(384,)", style=solid]; +"307 _frozen_param55" -> "308 dequantize_per_channel_default_55" [label="(384, 96)", style=solid]; +"308 dequantize_per_channel_default_55" -> "1188 linear_10" [label=None, style=solid]; +"309 features_1_1_mlp_0_bias" -> "1188 linear_10" [label="(384,)", style=solid]; +"310 _scale_56" -> "313 dequantize_per_channel_default_56" [label="(96,)", style=solid]; +"311 _zero_point_56" -> "313 dequantize_per_channel_default_56" [label="(96,)", style=solid]; +"312 _frozen_param56" -> "313 dequantize_per_channel_default_56" [label="(96, 384)", style=solid]; +"313 dequantize_per_channel_default_56" -> "1193 linear_11" [label=None, style=solid]; +"314 features_1_1_mlp_3_bias" -> "1193 linear_11" [label="(96,)", style=solid]; +"315 features_1_1_norm2_weight" -> "1195 layer_norm_4" [label="(96,)", style=solid]; +"316 features_1_1_norm2_bias" -> "1195 layer_norm_4" [label="(96,)", style=solid]; +"317 _scale_57" -> "320 dequantize_per_channel_default_57" [label="(192,)", style=solid]; +"318 _zero_point_57" -> "320 dequantize_per_channel_default_57" [label="(192,)", style=solid]; +"319 _frozen_param57" -> "320 dequantize_per_channel_default_57" [label="(192, 384)", style=solid]; +"320 dequantize_per_channel_default_57" -> "1213 linear_12" [label=None, style=solid]; +"321 features_2_norm_weight" -> "1214 layer_norm_5" [label="(192,)", style=solid]; +"322 features_2_norm_bias" -> "1214 layer_norm_5" [label="(192,)", style=solid]; +"323 _scale_58" -> "326 dequantize_per_channel_default_58" [label="(512,)", style=solid]; +"324 _zero_point_58" -> "326 dequantize_per_channel_default_58" [label="(512,)", style=solid]; +"325 _frozen_param58" -> "326 dequantize_per_channel_default_58" [label="(512, 2)", style=solid]; +"326 dequantize_per_channel_default_58" -> "1217 linear_13" [label=None, style=solid]; +"327 features_3_0_attn_cpb_mlp_0_bias" -> "1217 linear_13" [label="(512,)", style=solid]; +"328 _scale_59" -> "331 dequantize_per_channel_default_59" [label="(6,)", style=solid]; +"329 _zero_point_59" -> "331 dequantize_per_channel_default_59" [label="(6,)", style=solid]; +"330 _frozen_param59" -> "331 dequantize_per_channel_default_59" [label="(6, 512)", style=solid]; +"331 dequantize_per_channel_default_59" -> "1219 linear_14" [label=None, style=solid]; +"332 features_3_0_norm1_weight" -> "1276 layer_norm_6" [label="(192,)", style=solid]; +"333 features_3_0_norm1_bias" -> "1276 layer_norm_6" [label="(192,)", style=solid]; +"334 _scale_60" -> "337 dequantize_per_channel_default_60" [label="(768,)", style=solid]; +"335 _zero_point_60" -> "337 dequantize_per_channel_default_60" [label="(768,)", style=solid]; +"336 _frozen_param60" -> "337 dequantize_per_channel_default_60" [label="(768, 192)", style=solid]; +"337 dequantize_per_channel_default_60" -> "1280 linear_17" [label=None, style=solid]; +"338 features_3_0_mlp_0_bias" -> "1280 linear_17" [label="(768,)", style=solid]; +"339 _scale_61" -> "342 dequantize_per_channel_default_61" [label="(192,)", style=solid]; +"340 _zero_point_61" -> "342 dequantize_per_channel_default_61" [label="(192,)", style=solid]; +"341 _frozen_param61" -> "342 dequantize_per_channel_default_61" [label="(192, 768)", style=solid]; +"342 dequantize_per_channel_default_61" -> "1285 linear_18" [label=None, style=solid]; +"343 features_3_0_mlp_3_bias" -> "1285 linear_18" [label="(192,)", style=solid]; +"344 features_3_0_norm2_weight" -> "1287 layer_norm_7" [label="(192,)", style=solid]; +"345 features_3_0_norm2_bias" -> "1287 layer_norm_7" [label="(192,)", style=solid]; +"346 _scale_62" -> "349 dequantize_per_channel_default_62" [label="(512,)", style=solid]; +"347 _zero_point_62" -> "349 dequantize_per_channel_default_62" [label="(512,)", style=solid]; +"348 _frozen_param62" -> "349 dequantize_per_channel_default_62" [label="(512, 2)", style=solid]; +"349 dequantize_per_channel_default_62" -> "1289 linear_19" [label=None, style=solid]; +"350 features_3_1_attn_cpb_mlp_0_bias" -> "1289 linear_19" [label="(512,)", style=solid]; +"351 _scale_63" -> "354 dequantize_per_channel_default_63" [label="(6,)", style=solid]; +"352 _zero_point_63" -> "354 dequantize_per_channel_default_63" [label="(6,)", style=solid]; +"353 _frozen_param63" -> "354 dequantize_per_channel_default_63" [label="(6, 512)", style=solid]; +"354 dequantize_per_channel_default_63" -> "1291 linear_20" [label=None, style=solid]; +"355 features_3_1_norm1_weight" -> "1404 layer_norm_8" [label="(192,)", style=solid]; +"356 features_3_1_norm1_bias" -> "1404 layer_norm_8" [label="(192,)", style=solid]; +"357 _scale_64" -> "360 dequantize_per_channel_default_64" [label="(768,)", style=solid]; +"358 _zero_point_64" -> "360 dequantize_per_channel_default_64" [label="(768,)", style=solid]; +"359 _frozen_param64" -> "360 dequantize_per_channel_default_64" [label="(768, 192)", style=solid]; +"360 dequantize_per_channel_default_64" -> "1408 linear_23" [label=None, style=solid]; +"361 features_3_1_mlp_0_bias" -> "1408 linear_23" [label="(768,)", style=solid]; +"362 _scale_65" -> "365 dequantize_per_channel_default_65" [label="(192,)", style=solid]; +"363 _zero_point_65" -> "365 dequantize_per_channel_default_65" [label="(192,)", style=solid]; +"364 _frozen_param65" -> "365 dequantize_per_channel_default_65" [label="(192, 768)", style=solid]; +"365 dequantize_per_channel_default_65" -> "1413 linear_24" [label=None, style=solid]; +"366 features_3_1_mlp_3_bias" -> "1413 linear_24" [label="(192,)", style=solid]; +"367 features_3_1_norm2_weight" -> "1415 layer_norm_9" [label="(192,)", style=solid]; +"368 features_3_1_norm2_bias" -> "1415 layer_norm_9" [label="(192,)", style=solid]; +"369 _scale_66" -> "372 dequantize_per_channel_default_66" [label="(384,)", style=solid]; +"370 _zero_point_66" -> "372 dequantize_per_channel_default_66" [label="(384,)", style=solid]; +"371 _frozen_param66" -> "372 dequantize_per_channel_default_66" [label="(384, 768)", style=solid]; +"372 dequantize_per_channel_default_66" -> "1433 linear_25" [label=None, style=solid]; +"373 features_4_norm_weight" -> "1434 layer_norm_10" [label="(384,)", style=solid]; +"374 features_4_norm_bias" -> "1434 layer_norm_10" [label="(384,)", style=solid]; +"375 _scale_67" -> "378 dequantize_per_channel_default_67" [label="(512,)", style=solid]; +"376 _zero_point_67" -> "378 dequantize_per_channel_default_67" [label="(512,)", style=solid]; +"377 _frozen_param67" -> "378 dequantize_per_channel_default_67" [label="(512, 2)", style=solid]; +"378 dequantize_per_channel_default_67" -> "1437 linear_26" [label=None, style=solid]; +"379 features_5_0_attn_cpb_mlp_0_bias" -> "1437 linear_26" [label="(512,)", style=solid]; +"380 _scale_68" -> "383 dequantize_per_channel_default_68" [label="(12,)", style=solid]; +"381 _zero_point_68" -> "383 dequantize_per_channel_default_68" [label="(12,)", style=solid]; +"382 _frozen_param68" -> "383 dequantize_per_channel_default_68" [label="(12, 512)", style=solid]; +"383 dequantize_per_channel_default_68" -> "1439 linear_27" [label=None, style=solid]; +"384 features_5_0_norm1_weight" -> "1496 layer_norm_11" [label="(384,)", style=solid]; +"385 features_5_0_norm1_bias" -> "1496 layer_norm_11" [label="(384,)", style=solid]; +"386 _scale_69" -> "389 dequantize_per_channel_default_69" [label="(1536,)", style=solid]; +"387 _zero_point_69" -> "389 dequantize_per_channel_default_69" [label="(1536,)", style=solid]; +"388 _frozen_param69" -> "389 dequantize_per_channel_default_69" [label="(1536, 384)", style=solid]; +"389 dequantize_per_channel_default_69" -> "1500 linear_30" [label=None, style=solid]; +"390 features_5_0_mlp_0_bias" -> "1500 linear_30" [label="(1536,)", style=solid]; +"391 _scale_70" -> "394 dequantize_per_channel_default_70" [label="(384,)", style=solid]; +"392 _zero_point_70" -> "394 dequantize_per_channel_default_70" [label="(384,)", style=solid]; +"393 _frozen_param70" -> "394 dequantize_per_channel_default_70" [label="(384, 1536)", style=solid]; +"394 dequantize_per_channel_default_70" -> "1505 linear_31" [label=None, style=solid]; +"395 features_5_0_mlp_3_bias" -> "1505 linear_31" [label="(384,)", style=solid]; +"396 features_5_0_norm2_weight" -> "1507 layer_norm_12" [label="(384,)", style=solid]; +"397 features_5_0_norm2_bias" -> "1507 layer_norm_12" [label="(384,)", style=solid]; +"398 _scale_71" -> "401 dequantize_per_channel_default_71" [label="(512,)", style=solid]; +"399 _zero_point_71" -> "401 dequantize_per_channel_default_71" [label="(512,)", style=solid]; +"400 _frozen_param71" -> "401 dequantize_per_channel_default_71" [label="(512, 2)", style=solid]; +"401 dequantize_per_channel_default_71" -> "1509 linear_32" [label=None, style=solid]; +"402 features_5_1_attn_cpb_mlp_0_bias" -> "1509 linear_32" [label="(512,)", style=solid]; +"403 _scale_72" -> "406 dequantize_per_channel_default_72" [label="(12,)", style=solid]; +"404 _zero_point_72" -> "406 dequantize_per_channel_default_72" [label="(12,)", style=solid]; +"405 _frozen_param72" -> "406 dequantize_per_channel_default_72" [label="(12, 512)", style=solid]; +"406 dequantize_per_channel_default_72" -> "1511 linear_33" [label=None, style=solid]; +"407 features_5_1_norm1_weight" -> "1624 layer_norm_13" [label="(384,)", style=solid]; +"408 features_5_1_norm1_bias" -> "1624 layer_norm_13" [label="(384,)", style=solid]; +"409 _scale_73" -> "412 dequantize_per_channel_default_73" [label="(1536,)", style=solid]; +"410 _zero_point_73" -> "412 dequantize_per_channel_default_73" [label="(1536,)", style=solid]; +"411 _frozen_param73" -> "412 dequantize_per_channel_default_73" [label="(1536, 384)", style=solid]; +"412 dequantize_per_channel_default_73" -> "1628 linear_36" [label=None, style=solid]; +"413 features_5_1_mlp_0_bias" -> "1628 linear_36" [label="(1536,)", style=solid]; +"414 _scale_74" -> "417 dequantize_per_channel_default_74" [label="(384,)", style=solid]; +"415 _zero_point_74" -> "417 dequantize_per_channel_default_74" [label="(384,)", style=solid]; +"416 _frozen_param74" -> "417 dequantize_per_channel_default_74" [label="(384, 1536)", style=solid]; +"417 dequantize_per_channel_default_74" -> "1633 linear_37" [label=None, style=solid]; +"418 features_5_1_mlp_3_bias" -> "1633 linear_37" [label="(384,)", style=solid]; +"419 features_5_1_norm2_weight" -> "1635 layer_norm_14" [label="(384,)", style=solid]; +"420 features_5_1_norm2_bias" -> "1635 layer_norm_14" [label="(384,)", style=solid]; +"421 _scale_75" -> "424 dequantize_per_channel_default_75" [label="(512,)", style=solid]; +"422 _zero_point_75" -> "424 dequantize_per_channel_default_75" [label="(512,)", style=solid]; +"423 _frozen_param75" -> "424 dequantize_per_channel_default_75" [label="(512, 2)", style=solid]; +"424 dequantize_per_channel_default_75" -> "1639 linear_38" [label=None, style=solid]; +"425 features_5_2_attn_cpb_mlp_0_bias" -> "1639 linear_38" [label="(512,)", style=solid]; +"426 _scale_76" -> "429 dequantize_per_channel_default_76" [label="(12,)", style=solid]; +"427 _zero_point_76" -> "429 dequantize_per_channel_default_76" [label="(12,)", style=solid]; +"428 _frozen_param76" -> "429 dequantize_per_channel_default_76" [label="(12, 512)", style=solid]; +"429 dequantize_per_channel_default_76" -> "1641 linear_39" [label=None, style=solid]; +"430 features_5_2_norm1_weight" -> "1698 layer_norm_15" [label="(384,)", style=solid]; +"431 features_5_2_norm1_bias" -> "1698 layer_norm_15" [label="(384,)", style=solid]; +"432 _scale_77" -> "435 dequantize_per_channel_default_77" [label="(1536,)", style=solid]; +"433 _zero_point_77" -> "435 dequantize_per_channel_default_77" [label="(1536,)", style=solid]; +"434 _frozen_param77" -> "435 dequantize_per_channel_default_77" [label="(1536, 384)", style=solid]; +"435 dequantize_per_channel_default_77" -> "1702 linear_42" [label=None, style=solid]; +"436 features_5_2_mlp_0_bias" -> "1702 linear_42" [label="(1536,)", style=solid]; +"437 _scale_78" -> "440 dequantize_per_channel_default_78" [label="(384,)", style=solid]; +"438 _zero_point_78" -> "440 dequantize_per_channel_default_78" [label="(384,)", style=solid]; +"439 _frozen_param78" -> "440 dequantize_per_channel_default_78" [label="(384, 1536)", style=solid]; +"440 dequantize_per_channel_default_78" -> "1707 linear_43" [label=None, style=solid]; +"441 features_5_2_mlp_3_bias" -> "1707 linear_43" [label="(384,)", style=solid]; +"442 features_5_2_norm2_weight" -> "1709 layer_norm_16" [label="(384,)", style=solid]; +"443 features_5_2_norm2_bias" -> "1709 layer_norm_16" [label="(384,)", style=solid]; +"444 _scale_79" -> "447 dequantize_per_channel_default_79" [label="(512,)", style=solid]; +"445 _zero_point_79" -> "447 dequantize_per_channel_default_79" [label="(512,)", style=solid]; +"446 _frozen_param79" -> "447 dequantize_per_channel_default_79" [label="(512, 2)", style=solid]; +"447 dequantize_per_channel_default_79" -> "1711 linear_44" [label=None, style=solid]; +"448 features_5_3_attn_cpb_mlp_0_bias" -> "1711 linear_44" [label="(512,)", style=solid]; +"449 _scale_80" -> "452 dequantize_per_channel_default_80" [label="(12,)", style=solid]; +"450 _zero_point_80" -> "452 dequantize_per_channel_default_80" [label="(12,)", style=solid]; +"451 _frozen_param80" -> "452 dequantize_per_channel_default_80" [label="(12, 512)", style=solid]; +"452 dequantize_per_channel_default_80" -> "1713 linear_45" [label=None, style=solid]; +"453 features_5_3_norm1_weight" -> "1826 layer_norm_17" [label="(384,)", style=solid]; +"454 features_5_3_norm1_bias" -> "1826 layer_norm_17" [label="(384,)", style=solid]; +"455 _scale_81" -> "458 dequantize_per_channel_default_81" [label="(1536,)", style=solid]; +"456 _zero_point_81" -> "458 dequantize_per_channel_default_81" [label="(1536,)", style=solid]; +"457 _frozen_param81" -> "458 dequantize_per_channel_default_81" [label="(1536, 384)", style=solid]; +"458 dequantize_per_channel_default_81" -> "1830 linear_48" [label=None, style=solid]; +"459 features_5_3_mlp_0_bias" -> "1830 linear_48" [label="(1536,)", style=solid]; +"460 _scale_82" -> "463 dequantize_per_channel_default_82" [label="(384,)", style=solid]; +"461 _zero_point_82" -> "463 dequantize_per_channel_default_82" [label="(384,)", style=solid]; +"462 _frozen_param82" -> "463 dequantize_per_channel_default_82" [label="(384, 1536)", style=solid]; +"463 dequantize_per_channel_default_82" -> "1835 linear_49" [label=None, style=solid]; +"464 features_5_3_mlp_3_bias" -> "1835 linear_49" [label="(384,)", style=solid]; +"465 features_5_3_norm2_weight" -> "1837 layer_norm_18" [label="(384,)", style=solid]; +"466 features_5_3_norm2_bias" -> "1837 layer_norm_18" [label="(384,)", style=solid]; +"467 _scale_83" -> "470 dequantize_per_channel_default_83" [label="(512,)", style=solid]; +"468 _zero_point_83" -> "470 dequantize_per_channel_default_83" [label="(512,)", style=solid]; +"469 _frozen_param83" -> "470 dequantize_per_channel_default_83" [label="(512, 2)", style=solid]; +"470 dequantize_per_channel_default_83" -> "1841 linear_50" [label=None, style=solid]; +"471 features_5_4_attn_cpb_mlp_0_bias" -> "1841 linear_50" [label="(512,)", style=solid]; +"472 _scale_84" -> "475 dequantize_per_channel_default_84" [label="(12,)", style=solid]; +"473 _zero_point_84" -> "475 dequantize_per_channel_default_84" [label="(12,)", style=solid]; +"474 _frozen_param84" -> "475 dequantize_per_channel_default_84" [label="(12, 512)", style=solid]; +"475 dequantize_per_channel_default_84" -> "1843 linear_51" [label=None, style=solid]; +"476 features_5_4_norm1_weight" -> "1900 layer_norm_19" [label="(384,)", style=solid]; +"477 features_5_4_norm1_bias" -> "1900 layer_norm_19" [label="(384,)", style=solid]; +"478 _scale_85" -> "481 dequantize_per_channel_default_85" [label="(1536,)", style=solid]; +"479 _zero_point_85" -> "481 dequantize_per_channel_default_85" [label="(1536,)", style=solid]; +"480 _frozen_param85" -> "481 dequantize_per_channel_default_85" [label="(1536, 384)", style=solid]; +"481 dequantize_per_channel_default_85" -> "1904 linear_54" [label=None, style=solid]; +"482 features_5_4_mlp_0_bias" -> "1904 linear_54" [label="(1536,)", style=solid]; +"483 _scale_86" -> "486 dequantize_per_channel_default_86" [label="(384,)", style=solid]; +"484 _zero_point_86" -> "486 dequantize_per_channel_default_86" [label="(384,)", style=solid]; +"485 _frozen_param86" -> "486 dequantize_per_channel_default_86" [label="(384, 1536)", style=solid]; +"486 dequantize_per_channel_default_86" -> "1909 linear_55" [label=None, style=solid]; +"487 features_5_4_mlp_3_bias" -> "1909 linear_55" [label="(384,)", style=solid]; +"488 features_5_4_norm2_weight" -> "1911 layer_norm_20" [label="(384,)", style=solid]; +"489 features_5_4_norm2_bias" -> "1911 layer_norm_20" [label="(384,)", style=solid]; +"490 _scale_87" -> "493 dequantize_per_channel_default_87" [label="(512,)", style=solid]; +"491 _zero_point_87" -> "493 dequantize_per_channel_default_87" [label="(512,)", style=solid]; +"492 _frozen_param87" -> "493 dequantize_per_channel_default_87" [label="(512, 2)", style=solid]; +"493 dequantize_per_channel_default_87" -> "1913 linear_56" [label=None, style=solid]; +"494 features_5_5_attn_cpb_mlp_0_bias" -> "1913 linear_56" [label="(512,)", style=solid]; +"495 _scale_88" -> "498 dequantize_per_channel_default_88" [label="(12,)", style=solid]; +"496 _zero_point_88" -> "498 dequantize_per_channel_default_88" [label="(12,)", style=solid]; +"497 _frozen_param88" -> "498 dequantize_per_channel_default_88" [label="(12, 512)", style=solid]; +"498 dequantize_per_channel_default_88" -> "1915 linear_57" [label=None, style=solid]; +"499 features_5_5_norm1_weight" -> "2028 layer_norm_21" [label="(384,)", style=solid]; +"500 features_5_5_norm1_bias" -> "2028 layer_norm_21" [label="(384,)", style=solid]; +"501 _scale_89" -> "504 dequantize_per_channel_default_89" [label="(1536,)", style=solid]; +"502 _zero_point_89" -> "504 dequantize_per_channel_default_89" [label="(1536,)", style=solid]; +"503 _frozen_param89" -> "504 dequantize_per_channel_default_89" [label="(1536, 384)", style=solid]; +"504 dequantize_per_channel_default_89" -> "2032 linear_60" [label=None, style=solid]; +"505 features_5_5_mlp_0_bias" -> "2032 linear_60" [label="(1536,)", style=solid]; +"506 _scale_90" -> "509 dequantize_per_channel_default_90" [label="(384,)", style=solid]; +"507 _zero_point_90" -> "509 dequantize_per_channel_default_90" [label="(384,)", style=solid]; +"508 _frozen_param90" -> "509 dequantize_per_channel_default_90" [label="(384, 1536)", style=solid]; +"509 dequantize_per_channel_default_90" -> "2037 linear_61" [label=None, style=solid]; +"510 features_5_5_mlp_3_bias" -> "2037 linear_61" [label="(384,)", style=solid]; +"511 features_5_5_norm2_weight" -> "2039 layer_norm_22" [label="(384,)", style=solid]; +"512 features_5_5_norm2_bias" -> "2039 layer_norm_22" [label="(384,)", style=solid]; +"513 _scale_91" -> "516 dequantize_per_channel_default_91" [label="(512,)", style=solid]; +"514 _zero_point_91" -> "516 dequantize_per_channel_default_91" [label="(512,)", style=solid]; +"515 _frozen_param91" -> "516 dequantize_per_channel_default_91" [label="(512, 2)", style=solid]; +"516 dequantize_per_channel_default_91" -> "2043 linear_62" [label=None, style=solid]; +"517 features_5_6_attn_cpb_mlp_0_bias" -> "2043 linear_62" [label="(512,)", style=solid]; +"518 _scale_92" -> "521 dequantize_per_channel_default_92" [label="(12,)", style=solid]; +"519 _zero_point_92" -> "521 dequantize_per_channel_default_92" [label="(12,)", style=solid]; +"520 _frozen_param92" -> "521 dequantize_per_channel_default_92" [label="(12, 512)", style=solid]; +"521 dequantize_per_channel_default_92" -> "2045 linear_63" [label=None, style=solid]; +"522 features_5_6_norm1_weight" -> "2102 layer_norm_23" [label="(384,)", style=solid]; +"523 features_5_6_norm1_bias" -> "2102 layer_norm_23" [label="(384,)", style=solid]; +"524 _scale_93" -> "527 dequantize_per_channel_default_93" [label="(1536,)", style=solid]; +"525 _zero_point_93" -> "527 dequantize_per_channel_default_93" [label="(1536,)", style=solid]; +"526 _frozen_param93" -> "527 dequantize_per_channel_default_93" [label="(1536, 384)", style=solid]; +"527 dequantize_per_channel_default_93" -> "2106 linear_66" [label=None, style=solid]; +"528 features_5_6_mlp_0_bias" -> "2106 linear_66" [label="(1536,)", style=solid]; +"529 _scale_94" -> "532 dequantize_per_channel_default_94" [label="(384,)", style=solid]; +"530 _zero_point_94" -> "532 dequantize_per_channel_default_94" [label="(384,)", style=solid]; +"531 _frozen_param94" -> "532 dequantize_per_channel_default_94" [label="(384, 1536)", style=solid]; +"532 dequantize_per_channel_default_94" -> "2111 linear_67" [label=None, style=solid]; +"533 features_5_6_mlp_3_bias" -> "2111 linear_67" [label="(384,)", style=solid]; +"534 features_5_6_norm2_weight" -> "2113 layer_norm_24" [label="(384,)", style=solid]; +"535 features_5_6_norm2_bias" -> "2113 layer_norm_24" [label="(384,)", style=solid]; +"536 _scale_95" -> "539 dequantize_per_channel_default_95" [label="(512,)", style=solid]; +"537 _zero_point_95" -> "539 dequantize_per_channel_default_95" [label="(512,)", style=solid]; +"538 _frozen_param95" -> "539 dequantize_per_channel_default_95" [label="(512, 2)", style=solid]; +"539 dequantize_per_channel_default_95" -> "2115 linear_68" [label=None, style=solid]; +"540 features_5_7_attn_cpb_mlp_0_bias" -> "2115 linear_68" [label="(512,)", style=solid]; +"541 _scale_96" -> "544 dequantize_per_channel_default_96" [label="(12,)", style=solid]; +"542 _zero_point_96" -> "544 dequantize_per_channel_default_96" [label="(12,)", style=solid]; +"543 _frozen_param96" -> "544 dequantize_per_channel_default_96" [label="(12, 512)", style=solid]; +"544 dequantize_per_channel_default_96" -> "2117 linear_69" [label=None, style=solid]; +"545 features_5_7_norm1_weight" -> "2230 layer_norm_25" [label="(384,)", style=solid]; +"546 features_5_7_norm1_bias" -> "2230 layer_norm_25" [label="(384,)", style=solid]; +"547 _scale_97" -> "550 dequantize_per_channel_default_97" [label="(1536,)", style=solid]; +"548 _zero_point_97" -> "550 dequantize_per_channel_default_97" [label="(1536,)", style=solid]; +"549 _frozen_param97" -> "550 dequantize_per_channel_default_97" [label="(1536, 384)", style=solid]; +"550 dequantize_per_channel_default_97" -> "2234 linear_72" [label=None, style=solid]; +"551 features_5_7_mlp_0_bias" -> "2234 linear_72" [label="(1536,)", style=solid]; +"552 _scale_98" -> "555 dequantize_per_channel_default_98" [label="(384,)", style=solid]; +"553 _zero_point_98" -> "555 dequantize_per_channel_default_98" [label="(384,)", style=solid]; +"554 _frozen_param98" -> "555 dequantize_per_channel_default_98" [label="(384, 1536)", style=solid]; +"555 dequantize_per_channel_default_98" -> "2239 linear_73" [label=None, style=solid]; +"556 features_5_7_mlp_3_bias" -> "2239 linear_73" [label="(384,)", style=solid]; +"557 features_5_7_norm2_weight" -> "2241 layer_norm_26" [label="(384,)", style=solid]; +"558 features_5_7_norm2_bias" -> "2241 layer_norm_26" [label="(384,)", style=solid]; +"559 _scale_99" -> "562 dequantize_per_channel_default_99" [label="(512,)", style=solid]; +"560 _zero_point_99" -> "562 dequantize_per_channel_default_99" [label="(512,)", style=solid]; +"561 _frozen_param99" -> "562 dequantize_per_channel_default_99" [label="(512, 2)", style=solid]; +"562 dequantize_per_channel_default_99" -> "2245 linear_74" [label=None, style=solid]; +"563 features_5_8_attn_cpb_mlp_0_bias" -> "2245 linear_74" [label="(512,)", style=solid]; +"564 _scale_100" -> "567 dequantize_per_channel_default_100" [label="(12,)", style=solid]; +"565 _zero_point_100" -> "567 dequantize_per_channel_default_100" [label="(12,)", style=solid]; +"566 _frozen_param100" -> "567 dequantize_per_channel_default_100" [label="(12, 512)", style=solid]; +"567 dequantize_per_channel_default_100" -> "2247 linear_75" [label=None, style=solid]; +"568 features_5_8_norm1_weight" -> "2304 layer_norm_27" [label="(384,)", style=solid]; +"569 features_5_8_norm1_bias" -> "2304 layer_norm_27" [label="(384,)", style=solid]; +"570 _scale_101" -> "573 dequantize_per_channel_default_101" [label="(1536,)", style=solid]; +"571 _zero_point_101" -> "573 dequantize_per_channel_default_101" [label="(1536,)", style=solid]; +"572 _frozen_param101" -> "573 dequantize_per_channel_default_101" [label="(1536, 384)", style=solid]; +"573 dequantize_per_channel_default_101" -> "2308 linear_78" [label=None, style=solid]; +"574 features_5_8_mlp_0_bias" -> "2308 linear_78" [label="(1536,)", style=solid]; +"575 _scale_102" -> "578 dequantize_per_channel_default_102" [label="(384,)", style=solid]; +"576 _zero_point_102" -> "578 dequantize_per_channel_default_102" [label="(384,)", style=solid]; +"577 _frozen_param102" -> "578 dequantize_per_channel_default_102" [label="(384, 1536)", style=solid]; +"578 dequantize_per_channel_default_102" -> "2313 linear_79" [label=None, style=solid]; +"579 features_5_8_mlp_3_bias" -> "2313 linear_79" [label="(384,)", style=solid]; +"580 features_5_8_norm2_weight" -> "2315 layer_norm_28" [label="(384,)", style=solid]; +"581 features_5_8_norm2_bias" -> "2315 layer_norm_28" [label="(384,)", style=solid]; +"582 _scale_103" -> "585 dequantize_per_channel_default_103" [label="(512,)", style=solid]; +"583 _zero_point_103" -> "585 dequantize_per_channel_default_103" [label="(512,)", style=solid]; +"584 _frozen_param103" -> "585 dequantize_per_channel_default_103" [label="(512, 2)", style=solid]; +"585 dequantize_per_channel_default_103" -> "2317 linear_80" [label=None, style=solid]; +"586 features_5_9_attn_cpb_mlp_0_bias" -> "2317 linear_80" [label="(512,)", style=solid]; +"587 _scale_104" -> "590 dequantize_per_channel_default_104" [label="(12,)", style=solid]; +"588 _zero_point_104" -> "590 dequantize_per_channel_default_104" [label="(12,)", style=solid]; +"589 _frozen_param104" -> "590 dequantize_per_channel_default_104" [label="(12, 512)", style=solid]; +"590 dequantize_per_channel_default_104" -> "2319 linear_81" [label=None, style=solid]; +"591 features_5_9_norm1_weight" -> "2432 layer_norm_29" [label="(384,)", style=solid]; +"592 features_5_9_norm1_bias" -> "2432 layer_norm_29" [label="(384,)", style=solid]; +"593 _scale_105" -> "596 dequantize_per_channel_default_105" [label="(1536,)", style=solid]; +"594 _zero_point_105" -> "596 dequantize_per_channel_default_105" [label="(1536,)", style=solid]; +"595 _frozen_param105" -> "596 dequantize_per_channel_default_105" [label="(1536, 384)", style=solid]; +"596 dequantize_per_channel_default_105" -> "2436 linear_84" [label=None, style=solid]; +"597 features_5_9_mlp_0_bias" -> "2436 linear_84" [label="(1536,)", style=solid]; +"598 _scale_106" -> "601 dequantize_per_channel_default_106" [label="(384,)", style=solid]; +"599 _zero_point_106" -> "601 dequantize_per_channel_default_106" [label="(384,)", style=solid]; +"600 _frozen_param106" -> "601 dequantize_per_channel_default_106" [label="(384, 1536)", style=solid]; +"601 dequantize_per_channel_default_106" -> "2441 linear_85" [label=None, style=solid]; +"602 features_5_9_mlp_3_bias" -> "2441 linear_85" [label="(384,)", style=solid]; +"603 features_5_9_norm2_weight" -> "2443 layer_norm_30" [label="(384,)", style=solid]; +"604 features_5_9_norm2_bias" -> "2443 layer_norm_30" [label="(384,)", style=solid]; +"605 _scale_107" -> "608 dequantize_per_channel_default_107" [label="(512,)", style=solid]; +"606 _zero_point_107" -> "608 dequantize_per_channel_default_107" [label="(512,)", style=solid]; +"607 _frozen_param107" -> "608 dequantize_per_channel_default_107" [label="(512, 2)", style=solid]; +"608 dequantize_per_channel_default_107" -> "2447 linear_86" [label=None, style=solid]; +"609 features_5_10_attn_cpb_mlp_0_bias" -> "2447 linear_86" [label="(512,)", style=solid]; +"610 _scale_108" -> "613 dequantize_per_channel_default_108" [label="(12,)", style=solid]; +"611 _zero_point_108" -> "613 dequantize_per_channel_default_108" [label="(12,)", style=solid]; +"612 _frozen_param108" -> "613 dequantize_per_channel_default_108" [label="(12, 512)", style=solid]; +"613 dequantize_per_channel_default_108" -> "2449 linear_87" [label=None, style=solid]; +"614 features_5_10_norm1_weight" -> "2506 layer_norm_31" [label="(384,)", style=solid]; +"615 features_5_10_norm1_bias" -> "2506 layer_norm_31" [label="(384,)", style=solid]; +"616 _scale_109" -> "619 dequantize_per_channel_default_109" [label="(1536,)", style=solid]; +"617 _zero_point_109" -> "619 dequantize_per_channel_default_109" [label="(1536,)", style=solid]; +"618 _frozen_param109" -> "619 dequantize_per_channel_default_109" [label="(1536, 384)", style=solid]; +"619 dequantize_per_channel_default_109" -> "2510 linear_90" [label=None, style=solid]; +"620 features_5_10_mlp_0_bias" -> "2510 linear_90" [label="(1536,)", style=solid]; +"621 _scale_110" -> "624 dequantize_per_channel_default_110" [label="(384,)", style=solid]; +"622 _zero_point_110" -> "624 dequantize_per_channel_default_110" [label="(384,)", style=solid]; +"623 _frozen_param110" -> "624 dequantize_per_channel_default_110" [label="(384, 1536)", style=solid]; +"624 dequantize_per_channel_default_110" -> "2515 linear_91" [label=None, style=solid]; +"625 features_5_10_mlp_3_bias" -> "2515 linear_91" [label="(384,)", style=solid]; +"626 features_5_10_norm2_weight" -> "2517 layer_norm_32" [label="(384,)", style=solid]; +"627 features_5_10_norm2_bias" -> "2517 layer_norm_32" [label="(384,)", style=solid]; +"628 _scale_111" -> "631 dequantize_per_channel_default_111" [label="(512,)", style=solid]; +"629 _zero_point_111" -> "631 dequantize_per_channel_default_111" [label="(512,)", style=solid]; +"630 _frozen_param111" -> "631 dequantize_per_channel_default_111" [label="(512, 2)", style=solid]; +"631 dequantize_per_channel_default_111" -> "2519 linear_92" [label=None, style=solid]; +"632 features_5_11_attn_cpb_mlp_0_bias" -> "2519 linear_92" [label="(512,)", style=solid]; +"633 _scale_112" -> "636 dequantize_per_channel_default_112" [label="(12,)", style=solid]; +"634 _zero_point_112" -> "636 dequantize_per_channel_default_112" [label="(12,)", style=solid]; +"635 _frozen_param112" -> "636 dequantize_per_channel_default_112" [label="(12, 512)", style=solid]; +"636 dequantize_per_channel_default_112" -> "2521 linear_93" [label=None, style=solid]; +"637 features_5_11_norm1_weight" -> "2634 layer_norm_33" [label="(384,)", style=solid]; +"638 features_5_11_norm1_bias" -> "2634 layer_norm_33" [label="(384,)", style=solid]; +"639 _scale_113" -> "642 dequantize_per_channel_default_113" [label="(1536,)", style=solid]; +"640 _zero_point_113" -> "642 dequantize_per_channel_default_113" [label="(1536,)", style=solid]; +"641 _frozen_param113" -> "642 dequantize_per_channel_default_113" [label="(1536, 384)", style=solid]; +"642 dequantize_per_channel_default_113" -> "2638 linear_96" [label=None, style=solid]; +"643 features_5_11_mlp_0_bias" -> "2638 linear_96" [label="(1536,)", style=solid]; +"644 _scale_114" -> "647 dequantize_per_channel_default_114" [label="(384,)", style=solid]; +"645 _zero_point_114" -> "647 dequantize_per_channel_default_114" [label="(384,)", style=solid]; +"646 _frozen_param114" -> "647 dequantize_per_channel_default_114" [label="(384, 1536)", style=solid]; +"647 dequantize_per_channel_default_114" -> "2643 linear_97" [label=None, style=solid]; +"648 features_5_11_mlp_3_bias" -> "2643 linear_97" [label="(384,)", style=solid]; +"649 features_5_11_norm2_weight" -> "2645 layer_norm_34" [label="(384,)", style=solid]; +"650 features_5_11_norm2_bias" -> "2645 layer_norm_34" [label="(384,)", style=solid]; +"651 _scale_115" -> "654 dequantize_per_channel_default_115" [label="(512,)", style=solid]; +"652 _zero_point_115" -> "654 dequantize_per_channel_default_115" [label="(512,)", style=solid]; +"653 _frozen_param115" -> "654 dequantize_per_channel_default_115" [label="(512, 2)", style=solid]; +"654 dequantize_per_channel_default_115" -> "2649 linear_98" [label=None, style=solid]; +"655 features_5_12_attn_cpb_mlp_0_bias" -> "2649 linear_98" [label="(512,)", style=solid]; +"656 _scale_116" -> "659 dequantize_per_channel_default_116" [label="(12,)", style=solid]; +"657 _zero_point_116" -> "659 dequantize_per_channel_default_116" [label="(12,)", style=solid]; +"658 _frozen_param116" -> "659 dequantize_per_channel_default_116" [label="(12, 512)", style=solid]; +"659 dequantize_per_channel_default_116" -> "2651 linear_99" [label=None, style=solid]; +"660 features_5_12_norm1_weight" -> "2708 layer_norm_35" [label="(384,)", style=solid]; +"661 features_5_12_norm1_bias" -> "2708 layer_norm_35" [label="(384,)", style=solid]; +"662 _scale_117" -> "665 dequantize_per_channel_default_117" [label="(1536,)", style=solid]; +"663 _zero_point_117" -> "665 dequantize_per_channel_default_117" [label="(1536,)", style=solid]; +"664 _frozen_param117" -> "665 dequantize_per_channel_default_117" [label="(1536, 384)", style=solid]; +"665 dequantize_per_channel_default_117" -> "2712 linear_102" [label=None, style=solid]; +"666 features_5_12_mlp_0_bias" -> "2712 linear_102" [label="(1536,)", style=solid]; +"667 _scale_118" -> "670 dequantize_per_channel_default_118" [label="(384,)", style=solid]; +"668 _zero_point_118" -> "670 dequantize_per_channel_default_118" [label="(384,)", style=solid]; +"669 _frozen_param118" -> "670 dequantize_per_channel_default_118" [label="(384, 1536)", style=solid]; +"670 dequantize_per_channel_default_118" -> "2717 linear_103" [label=None, style=solid]; +"671 features_5_12_mlp_3_bias" -> "2717 linear_103" [label="(384,)", style=solid]; +"672 features_5_12_norm2_weight" -> "2719 layer_norm_36" [label="(384,)", style=solid]; +"673 features_5_12_norm2_bias" -> "2719 layer_norm_36" [label="(384,)", style=solid]; +"674 _scale_119" -> "677 dequantize_per_channel_default_119" [label="(512,)", style=solid]; +"675 _zero_point_119" -> "677 dequantize_per_channel_default_119" [label="(512,)", style=solid]; +"676 _frozen_param119" -> "677 dequantize_per_channel_default_119" [label="(512, 2)", style=solid]; +"677 dequantize_per_channel_default_119" -> "2721 linear_104" [label=None, style=solid]; +"678 features_5_13_attn_cpb_mlp_0_bias" -> "2721 linear_104" [label="(512,)", style=solid]; +"679 _scale_120" -> "682 dequantize_per_channel_default_120" [label="(12,)", style=solid]; +"680 _zero_point_120" -> "682 dequantize_per_channel_default_120" [label="(12,)", style=solid]; +"681 _frozen_param120" -> "682 dequantize_per_channel_default_120" [label="(12, 512)", style=solid]; +"682 dequantize_per_channel_default_120" -> "2723 linear_105" [label=None, style=solid]; +"683 features_5_13_norm1_weight" -> "2836 layer_norm_37" [label="(384,)", style=solid]; +"684 features_5_13_norm1_bias" -> "2836 layer_norm_37" [label="(384,)", style=solid]; +"685 _scale_121" -> "688 dequantize_per_channel_default_121" [label="(1536,)", style=solid]; +"686 _zero_point_121" -> "688 dequantize_per_channel_default_121" [label="(1536,)", style=solid]; +"687 _frozen_param121" -> "688 dequantize_per_channel_default_121" [label="(1536, 384)", style=solid]; +"688 dequantize_per_channel_default_121" -> "2840 linear_108" [label=None, style=solid]; +"689 features_5_13_mlp_0_bias" -> "2840 linear_108" [label="(1536,)", style=solid]; +"690 _scale_122" -> "693 dequantize_per_channel_default_122" [label="(384,)", style=solid]; +"691 _zero_point_122" -> "693 dequantize_per_channel_default_122" [label="(384,)", style=solid]; +"692 _frozen_param122" -> "693 dequantize_per_channel_default_122" [label="(384, 1536)", style=solid]; +"693 dequantize_per_channel_default_122" -> "2845 linear_109" [label=None, style=solid]; +"694 features_5_13_mlp_3_bias" -> "2845 linear_109" [label="(384,)", style=solid]; +"695 features_5_13_norm2_weight" -> "2847 layer_norm_38" [label="(384,)", style=solid]; +"696 features_5_13_norm2_bias" -> "2847 layer_norm_38" [label="(384,)", style=solid]; +"697 _scale_123" -> "700 dequantize_per_channel_default_123" [label="(512,)", style=solid]; +"698 _zero_point_123" -> "700 dequantize_per_channel_default_123" [label="(512,)", style=solid]; +"699 _frozen_param123" -> "700 dequantize_per_channel_default_123" [label="(512, 2)", style=solid]; +"700 dequantize_per_channel_default_123" -> "2851 linear_110" [label=None, style=solid]; +"701 features_5_14_attn_cpb_mlp_0_bias" -> "2851 linear_110" [label="(512,)", style=solid]; +"702 _scale_124" -> "705 dequantize_per_channel_default_124" [label="(12,)", style=solid]; +"703 _zero_point_124" -> "705 dequantize_per_channel_default_124" [label="(12,)", style=solid]; +"704 _frozen_param124" -> "705 dequantize_per_channel_default_124" [label="(12, 512)", style=solid]; +"705 dequantize_per_channel_default_124" -> "2853 linear_111" [label=None, style=solid]; +"706 features_5_14_norm1_weight" -> "2910 layer_norm_39" [label="(384,)", style=solid]; +"707 features_5_14_norm1_bias" -> "2910 layer_norm_39" [label="(384,)", style=solid]; +"708 _scale_125" -> "711 dequantize_per_channel_default_125" [label="(1536,)", style=solid]; +"709 _zero_point_125" -> "711 dequantize_per_channel_default_125" [label="(1536,)", style=solid]; +"710 _frozen_param125" -> "711 dequantize_per_channel_default_125" [label="(1536, 384)", style=solid]; +"711 dequantize_per_channel_default_125" -> "2914 linear_114" [label=None, style=solid]; +"712 features_5_14_mlp_0_bias" -> "2914 linear_114" [label="(1536,)", style=solid]; +"713 _scale_126" -> "716 dequantize_per_channel_default_126" [label="(384,)", style=solid]; +"714 _zero_point_126" -> "716 dequantize_per_channel_default_126" [label="(384,)", style=solid]; +"715 _frozen_param126" -> "716 dequantize_per_channel_default_126" [label="(384, 1536)", style=solid]; +"716 dequantize_per_channel_default_126" -> "2919 linear_115" [label=None, style=solid]; +"717 features_5_14_mlp_3_bias" -> "2919 linear_115" [label="(384,)", style=solid]; +"718 features_5_14_norm2_weight" -> "2921 layer_norm_40" [label="(384,)", style=solid]; +"719 features_5_14_norm2_bias" -> "2921 layer_norm_40" [label="(384,)", style=solid]; +"720 _scale_127" -> "723 dequantize_per_channel_default_127" [label="(512,)", style=solid]; +"721 _zero_point_127" -> "723 dequantize_per_channel_default_127" [label="(512,)", style=solid]; +"722 _frozen_param127" -> "723 dequantize_per_channel_default_127" [label="(512, 2)", style=solid]; +"723 dequantize_per_channel_default_127" -> "2923 linear_116" [label=None, style=solid]; +"724 features_5_15_attn_cpb_mlp_0_bias" -> "2923 linear_116" [label="(512,)", style=solid]; +"725 _scale_128" -> "728 dequantize_per_channel_default_128" [label="(12,)", style=solid]; +"726 _zero_point_128" -> "728 dequantize_per_channel_default_128" [label="(12,)", style=solid]; +"727 _frozen_param128" -> "728 dequantize_per_channel_default_128" [label="(12, 512)", style=solid]; +"728 dequantize_per_channel_default_128" -> "2925 linear_117" [label=None, style=solid]; +"729 features_5_15_norm1_weight" -> "3038 layer_norm_41" [label="(384,)", style=solid]; +"730 features_5_15_norm1_bias" -> "3038 layer_norm_41" [label="(384,)", style=solid]; +"731 _scale_129" -> "734 dequantize_per_channel_default_129" [label="(1536,)", style=solid]; +"732 _zero_point_129" -> "734 dequantize_per_channel_default_129" [label="(1536,)", style=solid]; +"733 _frozen_param129" -> "734 dequantize_per_channel_default_129" [label="(1536, 384)", style=solid]; +"734 dequantize_per_channel_default_129" -> "3042 linear_120" [label=None, style=solid]; +"735 features_5_15_mlp_0_bias" -> "3042 linear_120" [label="(1536,)", style=solid]; +"736 _scale_130" -> "739 dequantize_per_channel_default_130" [label="(384,)", style=solid]; +"737 _zero_point_130" -> "739 dequantize_per_channel_default_130" [label="(384,)", style=solid]; +"738 _frozen_param130" -> "739 dequantize_per_channel_default_130" [label="(384, 1536)", style=solid]; +"739 dequantize_per_channel_default_130" -> "3047 linear_121" [label=None, style=solid]; +"740 features_5_15_mlp_3_bias" -> "3047 linear_121" [label="(384,)", style=solid]; +"741 features_5_15_norm2_weight" -> "3049 layer_norm_42" [label="(384,)", style=solid]; +"742 features_5_15_norm2_bias" -> "3049 layer_norm_42" [label="(384,)", style=solid]; +"743 _scale_131" -> "746 dequantize_per_channel_default_131" [label="(512,)", style=solid]; +"744 _zero_point_131" -> "746 dequantize_per_channel_default_131" [label="(512,)", style=solid]; +"745 _frozen_param131" -> "746 dequantize_per_channel_default_131" [label="(512, 2)", style=solid]; +"746 dequantize_per_channel_default_131" -> "3053 linear_122" [label=None, style=solid]; +"747 features_5_16_attn_cpb_mlp_0_bias" -> "3053 linear_122" [label="(512,)", style=solid]; +"748 _scale_132" -> "751 dequantize_per_channel_default_132" [label="(12,)", style=solid]; +"749 _zero_point_132" -> "751 dequantize_per_channel_default_132" [label="(12,)", style=solid]; +"750 _frozen_param132" -> "751 dequantize_per_channel_default_132" [label="(12, 512)", style=solid]; +"751 dequantize_per_channel_default_132" -> "3055 linear_123" [label=None, style=solid]; +"752 features_5_16_norm1_weight" -> "3112 layer_norm_43" [label="(384,)", style=solid]; +"753 features_5_16_norm1_bias" -> "3112 layer_norm_43" [label="(384,)", style=solid]; +"754 _scale_133" -> "757 dequantize_per_channel_default_133" [label="(1536,)", style=solid]; +"755 _zero_point_133" -> "757 dequantize_per_channel_default_133" [label="(1536,)", style=solid]; +"756 _frozen_param133" -> "757 dequantize_per_channel_default_133" [label="(1536, 384)", style=solid]; +"757 dequantize_per_channel_default_133" -> "3116 linear_126" [label=None, style=solid]; +"758 features_5_16_mlp_0_bias" -> "3116 linear_126" [label="(1536,)", style=solid]; +"759 _scale_134" -> "762 dequantize_per_channel_default_134" [label="(384,)", style=solid]; +"760 _zero_point_134" -> "762 dequantize_per_channel_default_134" [label="(384,)", style=solid]; +"761 _frozen_param134" -> "762 dequantize_per_channel_default_134" [label="(384, 1536)", style=solid]; +"762 dequantize_per_channel_default_134" -> "3121 linear_127" [label=None, style=solid]; +"763 features_5_16_mlp_3_bias" -> "3121 linear_127" [label="(384,)", style=solid]; +"764 features_5_16_norm2_weight" -> "3123 layer_norm_44" [label="(384,)", style=solid]; +"765 features_5_16_norm2_bias" -> "3123 layer_norm_44" [label="(384,)", style=solid]; +"766 _scale_135" -> "769 dequantize_per_channel_default_135" [label="(512,)", style=solid]; +"767 _zero_point_135" -> "769 dequantize_per_channel_default_135" [label="(512,)", style=solid]; +"768 _frozen_param135" -> "769 dequantize_per_channel_default_135" [label="(512, 2)", style=solid]; +"769 dequantize_per_channel_default_135" -> "3125 linear_128" [label=None, style=solid]; +"770 features_5_17_attn_cpb_mlp_0_bias" -> "3125 linear_128" [label="(512,)", style=solid]; +"771 _scale_136" -> "774 dequantize_per_channel_default_136" [label="(12,)", style=solid]; +"772 _zero_point_136" -> "774 dequantize_per_channel_default_136" [label="(12,)", style=solid]; +"773 _frozen_param136" -> "774 dequantize_per_channel_default_136" [label="(12, 512)", style=solid]; +"774 dequantize_per_channel_default_136" -> "3127 linear_129" [label=None, style=solid]; +"775 features_5_17_norm1_weight" -> "3240 layer_norm_45" [label="(384,)", style=solid]; +"776 features_5_17_norm1_bias" -> "3240 layer_norm_45" [label="(384,)", style=solid]; +"777 _scale_137" -> "780 dequantize_per_channel_default_137" [label="(1536,)", style=solid]; +"778 _zero_point_137" -> "780 dequantize_per_channel_default_137" [label="(1536,)", style=solid]; +"779 _frozen_param137" -> "780 dequantize_per_channel_default_137" [label="(1536, 384)", style=solid]; +"780 dequantize_per_channel_default_137" -> "3244 linear_132" [label=None, style=solid]; +"781 features_5_17_mlp_0_bias" -> "3244 linear_132" [label="(1536,)", style=solid]; +"782 _scale_138" -> "785 dequantize_per_channel_default_138" [label="(384,)", style=solid]; +"783 _zero_point_138" -> "785 dequantize_per_channel_default_138" [label="(384,)", style=solid]; +"784 _frozen_param138" -> "785 dequantize_per_channel_default_138" [label="(384, 1536)", style=solid]; +"785 dequantize_per_channel_default_138" -> "3249 linear_133" [label=None, style=solid]; +"786 features_5_17_mlp_3_bias" -> "3249 linear_133" [label="(384,)", style=solid]; +"787 features_5_17_norm2_weight" -> "3251 layer_norm_46" [label="(384,)", style=solid]; +"788 features_5_17_norm2_bias" -> "3251 layer_norm_46" [label="(384,)", style=solid]; +"789 _scale_139" -> "792 dequantize_per_channel_default_139" [label="(768,)", style=solid]; +"790 _zero_point_139" -> "792 dequantize_per_channel_default_139" [label="(768,)", style=solid]; +"791 _frozen_param139" -> "792 dequantize_per_channel_default_139" [label="(768, 1536)", style=solid]; +"792 dequantize_per_channel_default_139" -> "3269 linear_134" [label=None, style=solid]; +"793 features_6_norm_weight" -> "3270 layer_norm_47" [label="(768,)", style=solid]; +"794 features_6_norm_bias" -> "3270 layer_norm_47" [label="(768,)", style=solid]; +"795 _scale_140" -> "798 dequantize_per_channel_default_140" [label="(512,)", style=solid]; +"796 _zero_point_140" -> "798 dequantize_per_channel_default_140" [label="(512,)", style=solid]; +"797 _frozen_param140" -> "798 dequantize_per_channel_default_140" [label="(512, 2)", style=solid]; +"798 dequantize_per_channel_default_140" -> "3273 linear_135" [label=None, style=solid]; +"799 features_7_0_attn_cpb_mlp_0_bias" -> "3273 linear_135" [label="(512,)", style=solid]; +"800 _scale_141" -> "803 dequantize_per_channel_default_141" [label="(24,)", style=solid]; +"801 _zero_point_141" -> "803 dequantize_per_channel_default_141" [label="(24,)", style=solid]; +"802 _frozen_param141" -> "803 dequantize_per_channel_default_141" [label="(24, 512)", style=solid]; +"803 dequantize_per_channel_default_141" -> "3275 linear_136" [label=None, style=solid]; +"804 features_7_0_norm1_weight" -> "3332 layer_norm_48" [label="(768,)", style=solid]; +"805 features_7_0_norm1_bias" -> "3332 layer_norm_48" [label="(768,)", style=solid]; +"806 _scale_142" -> "809 dequantize_per_channel_default_142" [label="(3072,)", style=solid]; +"807 _zero_point_142" -> "809 dequantize_per_channel_default_142" [label="(3072,)", style=solid]; +"808 _frozen_param142" -> "809 dequantize_per_channel_default_142" [label="(3072, 768)", style=solid]; +"809 dequantize_per_channel_default_142" -> "3336 linear_139" [label=None, style=solid]; +"810 features_7_0_mlp_0_bias" -> "3336 linear_139" [label="(3072,)", style=solid]; +"811 _scale_143" -> "814 dequantize_per_channel_default_143" [label="(768,)", style=solid]; +"812 _zero_point_143" -> "814 dequantize_per_channel_default_143" [label="(768,)", style=solid]; +"813 _frozen_param143" -> "814 dequantize_per_channel_default_143" [label="(768, 3072)", style=solid]; +"814 dequantize_per_channel_default_143" -> "3341 linear_140" [label=None, style=solid]; +"815 features_7_0_mlp_3_bias" -> "3341 linear_140" [label="(768,)", style=solid]; +"816 features_7_0_norm2_weight" -> "3343 layer_norm_49" [label="(768,)", style=solid]; +"817 features_7_0_norm2_bias" -> "3343 layer_norm_49" [label="(768,)", style=solid]; +"818 _scale_144" -> "821 dequantize_per_channel_default_144" [label="(512,)", style=solid]; +"819 _zero_point_144" -> "821 dequantize_per_channel_default_144" [label="(512,)", style=solid]; +"820 _frozen_param144" -> "821 dequantize_per_channel_default_144" [label="(512, 2)", style=solid]; +"821 dequantize_per_channel_default_144" -> "3347 linear_141" [label=None, style=solid]; +"822 features_7_1_attn_cpb_mlp_0_bias" -> "3347 linear_141" [label="(512,)", style=solid]; +"823 _scale_145" -> "826 dequantize_per_channel_default_145" [label="(24,)", style=solid]; +"824 _zero_point_145" -> "826 dequantize_per_channel_default_145" [label="(24,)", style=solid]; +"825 _frozen_param145" -> "826 dequantize_per_channel_default_145" [label="(24, 512)", style=solid]; +"826 dequantize_per_channel_default_145" -> "3349 linear_142" [label=None, style=solid]; +"827 features_7_1_norm1_weight" -> "3406 layer_norm_50" [label="(768,)", style=solid]; +"828 features_7_1_norm1_bias" -> "3406 layer_norm_50" [label="(768,)", style=solid]; +"829 _scale_146" -> "832 dequantize_per_channel_default_146" [label="(3072,)", style=solid]; +"830 _zero_point_146" -> "832 dequantize_per_channel_default_146" [label="(3072,)", style=solid]; +"831 _frozen_param146" -> "832 dequantize_per_channel_default_146" [label="(3072, 768)", style=solid]; +"832 dequantize_per_channel_default_146" -> "3410 linear_145" [label=None, style=solid]; +"833 features_7_1_mlp_0_bias" -> "3410 linear_145" [label="(3072,)", style=solid]; +"834 _scale_147" -> "837 dequantize_per_channel_default_147" [label="(768,)", style=solid]; +"835 _zero_point_147" -> "837 dequantize_per_channel_default_147" [label="(768,)", style=solid]; +"836 _frozen_param147" -> "837 dequantize_per_channel_default_147" [label="(768, 3072)", style=solid]; +"837 dequantize_per_channel_default_147" -> "3415 linear_146" [label=None, style=solid]; +"838 features_7_1_mlp_3_bias" -> "3415 linear_146" [label="(768,)", style=solid]; +"839 features_7_1_norm2_weight" -> "3417 layer_norm_51" [label="(768,)", style=solid]; +"840 features_7_1_norm2_bias" -> "3417 layer_norm_51" [label="(768,)", style=solid]; +"841 norm_weight" -> "3419 layer_norm_52" [label="(768,)", style=solid]; +"842 norm_bias" -> "3419 layer_norm_52" [label="(768,)", style=solid]; +"843 _scale_148" -> "846 dequantize_per_channel_default_148" [label="(1000,)", style=solid]; +"844 _zero_point_148" -> "846 dequantize_per_channel_default_148" [label="(1000,)", style=solid]; +"845 _frozen_param148" -> "846 dequantize_per_channel_default_148" [label="(1000, 768)", style=solid]; +"846 dequantize_per_channel_default_148" -> "3425 linear_147" [label=None, style=solid]; +"847 head_bias" -> "3425 linear_147" [label="(1000,)", style=solid]; +"848 features_1_0_attn_relative_coords_table" -> "1003 linear" [label="(1, 15, 15, 2)", style=solid]; +"849 features_1_0_attn_relative_position_index" -> "1007 index" [label="(4096,)", style=solid]; +"850 features_1_1_attn_relative_coords_table" -> "1072 linear_6" [label="(1, 15, 15, 2)", style=solid]; +"851 features_1_1_attn_relative_position_index" -> "1076 index_1" [label="(4096,)", style=solid]; +"852 features_3_0_attn_relative_coords_table" -> "1217 linear_13" [label="(1, 15, 15, 2)", style=solid]; +"853 features_3_0_attn_relative_position_index" -> "1221 index_2" [label="(4096,)", style=solid]; +"854 features_3_1_attn_relative_coords_table" -> "1289 linear_19" [label="(1, 15, 15, 2)", style=solid]; +"855 features_3_1_attn_relative_position_index" -> "1293 index_3" [label="(4096,)", style=solid]; +"856 features_5_0_attn_relative_coords_table" -> "1437 linear_26" [label="(1, 15, 15, 2)", style=solid]; +"857 features_5_0_attn_relative_position_index" -> "1441 index_4" [label="(4096,)", style=solid]; +"858 features_5_1_attn_relative_coords_table" -> "1509 linear_32" [label="(1, 15, 15, 2)", style=solid]; +"859 features_5_1_attn_relative_position_index" -> "1513 index_5" [label="(4096,)", style=solid]; +"860 features_5_2_attn_relative_coords_table" -> "1639 linear_38" [label="(1, 15, 15, 2)", style=solid]; +"861 features_5_2_attn_relative_position_index" -> "1643 index_6" [label="(4096,)", style=solid]; +"862 features_5_3_attn_relative_coords_table" -> "1711 linear_44" [label="(1, 15, 15, 2)", style=solid]; +"863 features_5_3_attn_relative_position_index" -> "1715 index_7" [label="(4096,)", style=solid]; +"864 features_5_4_attn_relative_coords_table" -> "1841 linear_50" [label="(1, 15, 15, 2)", style=solid]; +"865 features_5_4_attn_relative_position_index" -> "1845 index_8" [label="(4096,)", style=solid]; +"866 features_5_5_attn_relative_coords_table" -> "1913 linear_56" [label="(1, 15, 15, 2)", style=solid]; +"867 features_5_5_attn_relative_position_index" -> "1917 index_9" [label="(4096,)", style=solid]; +"868 features_5_6_attn_relative_coords_table" -> "2043 linear_62" [label="(1, 15, 15, 2)", style=solid]; +"869 features_5_6_attn_relative_position_index" -> "2047 index_10" [label="(4096,)", style=solid]; +"870 features_5_7_attn_relative_coords_table" -> "2115 linear_68" [label="(1, 15, 15, 2)", style=solid]; +"871 features_5_7_attn_relative_position_index" -> "2119 index_11" [label="(4096,)", style=solid]; +"872 features_5_8_attn_relative_coords_table" -> "2245 linear_74" [label="(1, 15, 15, 2)", style=solid]; +"873 features_5_8_attn_relative_position_index" -> "2249 index_12" [label="(4096,)", style=solid]; +"874 features_5_9_attn_relative_coords_table" -> "2317 linear_80" [label="(1, 15, 15, 2)", style=solid]; +"875 features_5_9_attn_relative_position_index" -> "2321 index_13" [label="(4096,)", style=solid]; +"876 features_5_10_attn_relative_coords_table" -> "2447 linear_86" [label="(1, 15, 15, 2)", style=solid]; +"877 features_5_10_attn_relative_position_index" -> "2451 index_14" [label="(4096,)", style=solid]; +"878 features_5_11_attn_relative_coords_table" -> "2519 linear_92" [label="(1, 15, 15, 2)", style=solid]; +"879 features_5_11_attn_relative_position_index" -> "2523 index_15" [label="(4096,)", style=solid]; +"880 features_5_12_attn_relative_coords_table" -> "2649 linear_98" [label="(1, 15, 15, 2)", style=solid]; +"881 features_5_12_attn_relative_position_index" -> "2653 index_16" [label="(4096,)", style=solid]; +"882 features_5_13_attn_relative_coords_table" -> "2721 linear_104" [label="(1, 15, 15, 2)", style=solid]; +"883 features_5_13_attn_relative_position_index" -> "2725 index_17" [label="(4096,)", style=solid]; +"884 features_5_14_attn_relative_coords_table" -> "2851 linear_110" [label="(1, 15, 15, 2)", style=solid]; +"885 features_5_14_attn_relative_position_index" -> "2855 index_18" [label="(4096,)", style=solid]; +"886 features_5_15_attn_relative_coords_table" -> "2923 linear_116" [label="(1, 15, 15, 2)", style=solid]; +"887 features_5_15_attn_relative_position_index" -> "2927 index_19" [label="(4096,)", style=solid]; +"888 features_5_16_attn_relative_coords_table" -> "3053 linear_122" [label="(1, 15, 15, 2)", style=solid]; +"889 features_5_16_attn_relative_position_index" -> "3057 index_20" [label="(4096,)", style=solid]; +"890 features_5_17_attn_relative_coords_table" -> "3125 linear_128" [label="(1, 15, 15, 2)", style=solid]; +"891 features_5_17_attn_relative_position_index" -> "3129 index_21" [label="(4096,)", style=solid]; +"892 features_7_0_attn_relative_coords_table" -> "3273 linear_135" [label="(1, 15, 15, 2)", style=solid]; +"893 features_7_0_attn_relative_position_index" -> "3277 index_22" [label="(4096,)", style=solid]; +"894 features_7_1_attn_relative_coords_table" -> "3347 linear_141" [label="(1, 15, 15, 2)", style=solid]; +"895 features_7_1_attn_relative_position_index" -> "3351 index_23" [label="(4096,)", style=solid]; +"896 features_1_1_attn_lifted_tensor_0" -> "1118 lift_fresh_copy" [label="()", style=solid]; +"897 features_1_1_attn_lifted_tensor_1" -> "1122 lift_fresh_copy_1" [label="()", style=solid]; +"898 features_1_1_attn_lifted_tensor_2" -> "1126 lift_fresh_copy_2" [label="()", style=solid]; +"899 features_1_1_attn_lifted_tensor_3" -> "1130 lift_fresh_copy_3" [label="()", style=solid]; +"900 features_1_1_attn_lifted_tensor_4" -> "1134 lift_fresh_copy_4" [label="()", style=solid]; +"901 features_1_1_attn_lifted_tensor_5" -> "1138 lift_fresh_copy_5" [label="()", style=solid]; +"902 features_1_1_attn_lifted_tensor_6" -> "1142 lift_fresh_copy_6" [label="()", style=solid]; +"903 features_1_1_attn_lifted_tensor_7" -> "1146 lift_fresh_copy_7" [label="()", style=solid]; +"904 features_1_1_attn_lifted_tensor_8" -> "1150 lift_fresh_copy_8" [label="()", style=solid]; +"905 features_3_1_attn_lifted_tensor_9" -> "1335 lift_fresh_copy_9" [label="()", style=solid]; +"906 features_3_1_attn_lifted_tensor_10" -> "1339 lift_fresh_copy_10" [label="()", style=solid]; +"907 features_3_1_attn_lifted_tensor_11" -> "1343 lift_fresh_copy_11" [label="()", style=solid]; +"908 features_3_1_attn_lifted_tensor_12" -> "1347 lift_fresh_copy_12" [label="()", style=solid]; +"909 features_3_1_attn_lifted_tensor_13" -> "1351 lift_fresh_copy_13" [label="()", style=solid]; +"910 features_3_1_attn_lifted_tensor_14" -> "1355 lift_fresh_copy_14" [label="()", style=solid]; +"911 features_3_1_attn_lifted_tensor_15" -> "1359 lift_fresh_copy_15" [label="()", style=solid]; +"912 features_3_1_attn_lifted_tensor_16" -> "1363 lift_fresh_copy_16" [label="()", style=solid]; +"913 features_3_1_attn_lifted_tensor_17" -> "1367 lift_fresh_copy_17" [label="()", style=solid]; +"914 features_5_1_attn_lifted_tensor_18" -> "1555 lift_fresh_copy_18" [label="()", style=solid]; +"915 features_5_1_attn_lifted_tensor_19" -> "1559 lift_fresh_copy_19" [label="()", style=solid]; +"916 features_5_1_attn_lifted_tensor_20" -> "1563 lift_fresh_copy_20" [label="()", style=solid]; +"917 features_5_1_attn_lifted_tensor_21" -> "1567 lift_fresh_copy_21" [label="()", style=solid]; +"918 features_5_1_attn_lifted_tensor_22" -> "1571 lift_fresh_copy_22" [label="()", style=solid]; +"919 features_5_1_attn_lifted_tensor_23" -> "1575 lift_fresh_copy_23" [label="()", style=solid]; +"920 features_5_1_attn_lifted_tensor_24" -> "1579 lift_fresh_copy_24" [label="()", style=solid]; +"921 features_5_1_attn_lifted_tensor_25" -> "1583 lift_fresh_copy_25" [label="()", style=solid]; +"922 features_5_1_attn_lifted_tensor_26" -> "1587 lift_fresh_copy_26" [label="()", style=solid]; +"923 features_5_3_attn_lifted_tensor_27" -> "1757 lift_fresh_copy_27" [label="()", style=solid]; +"924 features_5_3_attn_lifted_tensor_28" -> "1761 lift_fresh_copy_28" [label="()", style=solid]; +"925 features_5_3_attn_lifted_tensor_29" -> "1765 lift_fresh_copy_29" [label="()", style=solid]; +"926 features_5_3_attn_lifted_tensor_30" -> "1769 lift_fresh_copy_30" [label="()", style=solid]; +"927 features_5_3_attn_lifted_tensor_31" -> "1773 lift_fresh_copy_31" [label="()", style=solid]; +"928 features_5_3_attn_lifted_tensor_32" -> "1777 lift_fresh_copy_32" [label="()", style=solid]; +"929 features_5_3_attn_lifted_tensor_33" -> "1781 lift_fresh_copy_33" [label="()", style=solid]; +"930 features_5_3_attn_lifted_tensor_34" -> "1785 lift_fresh_copy_34" [label="()", style=solid]; +"931 features_5_3_attn_lifted_tensor_35" -> "1789 lift_fresh_copy_35" [label="()", style=solid]; +"932 features_5_5_attn_lifted_tensor_36" -> "1959 lift_fresh_copy_36" [label="()", style=solid]; +"933 features_5_5_attn_lifted_tensor_37" -> "1963 lift_fresh_copy_37" [label="()", style=solid]; +"934 features_5_5_attn_lifted_tensor_38" -> "1967 lift_fresh_copy_38" [label="()", style=solid]; +"935 features_5_5_attn_lifted_tensor_39" -> "1971 lift_fresh_copy_39" [label="()", style=solid]; +"936 features_5_5_attn_lifted_tensor_40" -> "1975 lift_fresh_copy_40" [label="()", style=solid]; +"937 features_5_5_attn_lifted_tensor_41" -> "1979 lift_fresh_copy_41" [label="()", style=solid]; +"938 features_5_5_attn_lifted_tensor_42" -> "1983 lift_fresh_copy_42" [label="()", style=solid]; +"939 features_5_5_attn_lifted_tensor_43" -> "1987 lift_fresh_copy_43" [label="()", style=solid]; +"940 features_5_5_attn_lifted_tensor_44" -> "1991 lift_fresh_copy_44" [label="()", style=solid]; +"941 features_5_7_attn_lifted_tensor_45" -> "2161 lift_fresh_copy_45" [label="()", style=solid]; +"942 features_5_7_attn_lifted_tensor_46" -> "2165 lift_fresh_copy_46" [label="()", style=solid]; +"943 features_5_7_attn_lifted_tensor_47" -> "2169 lift_fresh_copy_47" [label="()", style=solid]; +"944 features_5_7_attn_lifted_tensor_48" -> "2173 lift_fresh_copy_48" [label="()", style=solid]; +"945 features_5_7_attn_lifted_tensor_49" -> "2177 lift_fresh_copy_49" [label="()", style=solid]; +"946 features_5_7_attn_lifted_tensor_50" -> "2181 lift_fresh_copy_50" [label="()", style=solid]; +"947 features_5_7_attn_lifted_tensor_51" -> "2185 lift_fresh_copy_51" [label="()", style=solid]; +"948 features_5_7_attn_lifted_tensor_52" -> "2189 lift_fresh_copy_52" [label="()", style=solid]; +"949 features_5_7_attn_lifted_tensor_53" -> "2193 lift_fresh_copy_53" [label="()", style=solid]; +"950 features_5_9_attn_lifted_tensor_54" -> "2363 lift_fresh_copy_54" [label="()", style=solid]; +"951 features_5_9_attn_lifted_tensor_55" -> "2367 lift_fresh_copy_55" [label="()", style=solid]; +"952 features_5_9_attn_lifted_tensor_56" -> "2371 lift_fresh_copy_56" [label="()", style=solid]; +"953 features_5_9_attn_lifted_tensor_57" -> "2375 lift_fresh_copy_57" [label="()", style=solid]; +"954 features_5_9_attn_lifted_tensor_58" -> "2379 lift_fresh_copy_58" [label="()", style=solid]; +"955 features_5_9_attn_lifted_tensor_59" -> "2383 lift_fresh_copy_59" [label="()", style=solid]; +"956 features_5_9_attn_lifted_tensor_60" -> "2387 lift_fresh_copy_60" [label="()", style=solid]; +"957 features_5_9_attn_lifted_tensor_61" -> "2391 lift_fresh_copy_61" [label="()", style=solid]; +"958 features_5_9_attn_lifted_tensor_62" -> "2395 lift_fresh_copy_62" [label="()", style=solid]; +"959 features_5_11_attn_lifted_tensor_63" -> "2565 lift_fresh_copy_63" [label="()", style=solid]; +"960 features_5_11_attn_lifted_tensor_64" -> "2569 lift_fresh_copy_64" [label="()", style=solid]; +"961 features_5_11_attn_lifted_tensor_65" -> "2573 lift_fresh_copy_65" [label="()", style=solid]; +"962 features_5_11_attn_lifted_tensor_66" -> "2577 lift_fresh_copy_66" [label="()", style=solid]; +"963 features_5_11_attn_lifted_tensor_67" -> "2581 lift_fresh_copy_67" [label="()", style=solid]; +"964 features_5_11_attn_lifted_tensor_68" -> "2585 lift_fresh_copy_68" [label="()", style=solid]; +"965 features_5_11_attn_lifted_tensor_69" -> "2589 lift_fresh_copy_69" [label="()", style=solid]; +"966 features_5_11_attn_lifted_tensor_70" -> "2593 lift_fresh_copy_70" [label="()", style=solid]; +"967 features_5_11_attn_lifted_tensor_71" -> "2597 lift_fresh_copy_71" [label="()", style=solid]; +"968 features_5_13_attn_lifted_tensor_72" -> "2767 lift_fresh_copy_72" [label="()", style=solid]; +"969 features_5_13_attn_lifted_tensor_73" -> "2771 lift_fresh_copy_73" [label="()", style=solid]; +"970 features_5_13_attn_lifted_tensor_74" -> "2775 lift_fresh_copy_74" [label="()", style=solid]; +"971 features_5_13_attn_lifted_tensor_75" -> "2779 lift_fresh_copy_75" [label="()", style=solid]; +"972 features_5_13_attn_lifted_tensor_76" -> "2783 lift_fresh_copy_76" [label="()", style=solid]; +"973 features_5_13_attn_lifted_tensor_77" -> "2787 lift_fresh_copy_77" [label="()", style=solid]; +"974 features_5_13_attn_lifted_tensor_78" -> "2791 lift_fresh_copy_78" [label="()", style=solid]; +"975 features_5_13_attn_lifted_tensor_79" -> "2795 lift_fresh_copy_79" [label="()", style=solid]; +"976 features_5_13_attn_lifted_tensor_80" -> "2799 lift_fresh_copy_80" [label="()", style=solid]; +"977 features_5_15_attn_lifted_tensor_81" -> "2969 lift_fresh_copy_81" [label="()", style=solid]; +"978 features_5_15_attn_lifted_tensor_82" -> "2973 lift_fresh_copy_82" [label="()", style=solid]; +"979 features_5_15_attn_lifted_tensor_83" -> "2977 lift_fresh_copy_83" [label="()", style=solid]; +"980 features_5_15_attn_lifted_tensor_84" -> "2981 lift_fresh_copy_84" [label="()", style=solid]; +"981 features_5_15_attn_lifted_tensor_85" -> "2985 lift_fresh_copy_85" [label="()", style=solid]; +"982 features_5_15_attn_lifted_tensor_86" -> "2989 lift_fresh_copy_86" [label="()", style=solid]; +"983 features_5_15_attn_lifted_tensor_87" -> "2993 lift_fresh_copy_87" [label="()", style=solid]; +"984 features_5_15_attn_lifted_tensor_88" -> "2997 lift_fresh_copy_88" [label="()", style=solid]; +"985 features_5_15_attn_lifted_tensor_89" -> "3001 lift_fresh_copy_89" [label="()", style=solid]; +"986 features_5_17_attn_lifted_tensor_90" -> "3171 lift_fresh_copy_90" [label="()", style=solid]; +"987 features_5_17_attn_lifted_tensor_91" -> "3175 lift_fresh_copy_91" [label="()", style=solid]; +"988 features_5_17_attn_lifted_tensor_92" -> "3179 lift_fresh_copy_92" [label="()", style=solid]; +"989 features_5_17_attn_lifted_tensor_93" -> "3183 lift_fresh_copy_93" [label="()", style=solid]; +"990 features_5_17_attn_lifted_tensor_94" -> "3187 lift_fresh_copy_94" [label="()", style=solid]; +"991 features_5_17_attn_lifted_tensor_95" -> "3191 lift_fresh_copy_95" [label="()", style=solid]; +"992 features_5_17_attn_lifted_tensor_96" -> "3195 lift_fresh_copy_96" [label="()", style=solid]; +"993 features_5_17_attn_lifted_tensor_97" -> "3199 lift_fresh_copy_97" [label="()", style=solid]; +"994 features_5_17_attn_lifted_tensor_98" -> "3203 lift_fresh_copy_98" [label="()", style=solid]; +"995 x" -> "996 quantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; +"996 quantize_per_tensor_default" -> "997 dequantize_per_tensor_default" [label=None, style=solid]; +"997 dequantize_per_tensor_default" -> "998 conv2d" [label=None, style=solid]; +"998 conv2d" -> "999 permute" [label="(1, 96, 56, 56)", style=solid]; +"999 permute" -> "1000 layer_norm" [label="(1, 56, 56, 96)", style=solid]; +"1000 layer_norm" -> "1001 quantize_per_tensor_default_1" [label="(1, 56, 56, 96)", style=solid]; +"1000 layer_norm" -> "1060 add_1" [label="(1, 56, 56, 96)", style=solid]; +"1001 quantize_per_tensor_default_1" -> "1002 dequantize_per_tensor_default_1" [label=None, style=solid]; +"1002 dequantize_per_tensor_default_1" -> "1014 pad" [label=None, style=solid]; +"1003 linear" -> "1004 relu_" [label="(1, 15, 15, 512)", style=solid]; +"1004 relu_" -> "1005 linear_1" [label="(1, 15, 15, 512)", style=solid]; +"1005 linear_1" -> "1006 view" [label="(1, 15, 15, 3)", style=solid]; +"1006 view" -> "1007 index" [label="(225, 3)", style=solid]; +"1007 index" -> "1008 view_1" [label="(4096, 3)", style=solid]; +"1008 view_1" -> "1009 permute_1" [label="(64, 64, 3)", style=solid]; +"1009 permute_1" -> "1010 contiguous" [label="(3, 64, 64)", style=solid]; +"1010 contiguous" -> "1011 unsqueeze" [label="(3, 64, 64)", style=solid]; +"1011 unsqueeze" -> "1012 sigmoid" [label="(1, 3, 64, 64)", style=solid]; +"1012 sigmoid" -> "1013 mul" [label="(1, 3, 64, 64)", style=solid]; +"1013 mul" -> "1044 add" [label="(1, 3, 64, 64)", style=solid]; +"1014 pad" -> "1015 view_2" [label="(1, 56, 56, 96)", style=solid]; +"1015 view_2" -> "1016 permute_2" [label="(1, 7, 8, 7, 8, 96)", style=solid]; +"1016 permute_2" -> "1017 reshape" [label="(1, 7, 7, 8, 8, 96)", style=solid]; +"1017 reshape" -> "1021 linear_2" [label="(49, 64, 96)", style=solid]; +"1018 clone" -> "1019 slice_1" [label="(288,)", style=solid]; +"1018 clone" -> "1021 linear_2" [label="(288,)", style=solid]; +"1019 slice_1" -> "1020 zero_" [label="(96,)", style=solid]; +"1021 linear_2" -> "1022 reshape_1" [label="(49, 64, 288)", style=solid]; +"1022 reshape_1" -> "1023 permute_3" [label="(49, 64, 3, 3, 32)", style=solid]; +"1023 permute_3" -> "1024 select" [label="(3, 49, 3, 64, 32)", style=solid]; +"1023 permute_3" -> "1025 select_1" [label="(3, 49, 3, 64, 32)", style=solid]; +"1023 permute_3" -> "1026 select_2" [label="(3, 49, 3, 64, 32)", style=solid]; +"1024 select" -> "1027 linalg_vector_norm" [label="(49, 3, 64, 32)", style=solid]; +"1024 select" -> "1029 expand_as" [label="(49, 3, 64, 32)", style=solid]; +"1024 select" -> "1030 div" [label="(49, 3, 64, 32)", style=solid]; +"1025 select_1" -> "1033 linalg_vector_norm_1" [label="(49, 3, 64, 32)", style=solid]; +"1025 select_1" -> "1035 expand_as_1" [label="(49, 3, 64, 32)", style=solid]; +"1025 select_1" -> "1036 div_1" [label="(49, 3, 64, 32)", style=solid]; +"1026 select_2" -> "1047 matmul_1" [label="(49, 3, 64, 32)", style=solid]; +"1027 linalg_vector_norm" -> "1028 clamp_min" [label="(49, 3, 64, 1)", style=solid]; +"1028 clamp_min" -> "1029 expand_as" [label="(49, 3, 64, 1)", style=solid]; +"1029 expand_as" -> "1030 div" [label="(49, 3, 64, 32)", style=solid]; +"1030 div" -> "1031 quantize_per_tensor_default_2" [label="(49, 3, 64, 32)", style=solid]; +"1031 quantize_per_tensor_default_2" -> "1032 dequantize_per_tensor_default_2" [label=None, style=solid]; +"1032 dequantize_per_tensor_default_2" -> "1040 matmul" [label=None, style=solid]; +"1033 linalg_vector_norm_1" -> "1034 clamp_min_1" [label="(49, 3, 64, 1)", style=solid]; +"1034 clamp_min_1" -> "1035 expand_as_1" [label="(49, 3, 64, 1)", style=solid]; +"1035 expand_as_1" -> "1036 div_1" [label="(49, 3, 64, 32)", style=solid]; +"1036 div_1" -> "1037 quantize_per_tensor_default_3" [label="(49, 3, 64, 32)", style=solid]; +"1037 quantize_per_tensor_default_3" -> "1038 dequantize_per_tensor_default_3" [label=None, style=solid]; +"1038 dequantize_per_tensor_default_3" -> "1039 transpose" [label=None, style=solid]; +"1039 transpose" -> "1040 matmul" [label="(49, 3, 32, 64)", style=solid]; +"1040 matmul" -> "1043 mul_1" [label="(49, 3, 64, 64)", style=solid]; +"1041 clamp" -> "1042 exp" [label="(3, 1, 1)", style=solid]; +"1042 exp" -> "1043 mul_1" [label="(3, 1, 1)", style=solid]; +"1043 mul_1" -> "1044 add" [label="(49, 3, 64, 64)", style=solid]; +"1044 add" -> "1045 softmax" [label="(49, 3, 64, 64)", style=solid]; +"1045 softmax" -> "1046 dropout" [label="(49, 3, 64, 64)", style=solid]; +"1046 dropout" -> "1047 matmul_1" [label="(49, 3, 64, 64)", style=solid]; +"1047 matmul_1" -> "1048 quantize_per_tensor_default_4" [label="(49, 3, 64, 32)", style=solid]; +"1048 quantize_per_tensor_default_4" -> "1049 dequantize_per_tensor_default_4" [label=None, style=solid]; +"1049 dequantize_per_tensor_default_4" -> "1050 transpose_1" [label=None, style=solid]; +"1050 transpose_1" -> "1051 reshape_2" [label="(49, 64, 3, 32)", style=solid]; +"1051 reshape_2" -> "1052 linear_3" [label="(49, 64, 96)", style=solid]; +"1052 linear_3" -> "1053 dropout_1" [label="(49, 64, 96)", style=solid]; +"1053 dropout_1" -> "1054 view_3" [label="(49, 64, 96)", style=solid]; +"1054 view_3" -> "1055 permute_4" [label="(1, 7, 7, 8, 8, 96)", style=solid]; +"1055 permute_4" -> "1056 reshape_3" [label="(1, 7, 8, 7, 8, 96)", style=solid]; +"1056 reshape_3" -> "1057 slice_2" [label="(1, 56, 56, 96)", style=solid]; +"1057 slice_2" -> "1058 slice_3" [label="(1, 56, 56, 96)", style=solid]; +"1058 slice_3" -> "1059 layer_norm_1" [label="(1, 56, 56, 96)", style=solid]; +"1059 layer_norm_1" -> "1060 add_1" [label="(1, 56, 56, 96)", style=solid]; +"1060 add_1" -> "1061 quantize_per_tensor_default_5" [label="(1, 56, 56, 96)", style=solid]; +"1060 add_1" -> "1071 add_2" [label="(1, 56, 56, 96)", style=solid]; +"1061 quantize_per_tensor_default_5" -> "1062 dequantize_per_tensor_default_5" [label=None, style=solid]; +"1062 dequantize_per_tensor_default_5" -> "1063 linear_4" [label=None, style=solid]; +"1063 linear_4" -> "1064 gelu" [label="(1, 56, 56, 384)", style=solid]; +"1064 gelu" -> "1065 quantize_per_tensor_default_6" [label="(1, 56, 56, 384)", style=solid]; +"1065 quantize_per_tensor_default_6" -> "1066 dequantize_per_tensor_default_6" [label=None, style=solid]; +"1066 dequantize_per_tensor_default_6" -> "1067 dropout_2" [label=None, style=solid]; +"1067 dropout_2" -> "1068 linear_5" [label="(1, 56, 56, 384)", style=solid]; +"1068 linear_5" -> "1069 dropout_3" [label="(1, 56, 56, 96)", style=solid]; +"1069 dropout_3" -> "1070 layer_norm_2" [label="(1, 56, 56, 96)", style=solid]; +"1070 layer_norm_2" -> "1071 add_2" [label="(1, 56, 56, 96)", style=solid]; +"1071 add_2" -> "1083 pad_1" [label="(1, 56, 56, 96)", style=solid]; +"1071 add_2" -> "1185 add_5" [label="(1, 56, 56, 96)", style=solid]; +"1072 linear_6" -> "1073 relu__1" [label="(1, 15, 15, 512)", style=solid]; +"1073 relu__1" -> "1074 linear_7" [label="(1, 15, 15, 512)", style=solid]; +"1074 linear_7" -> "1075 view_4" [label="(1, 15, 15, 3)", style=solid]; +"1075 view_4" -> "1076 index_1" [label="(225, 3)", style=solid]; +"1076 index_1" -> "1077 view_5" [label="(4096, 3)", style=solid]; +"1077 view_5" -> "1078 permute_5" [label="(64, 64, 3)", style=solid]; +"1078 permute_5" -> "1079 contiguous_1" [label="(3, 64, 64)", style=solid]; +"1079 contiguous_1" -> "1080 unsqueeze_1" [label="(3, 64, 64)", style=solid]; +"1080 unsqueeze_1" -> "1081 sigmoid_1" [label="(1, 3, 64, 64)", style=solid]; +"1081 sigmoid_1" -> "1082 mul_2" [label="(1, 3, 64, 64)", style=solid]; +"1082 mul_2" -> "1116 add_3" [label="(1, 3, 64, 64)", style=solid]; +"1083 pad_1" -> "1084 roll" [label="(1, 56, 56, 96)", style=solid]; +"1084 roll" -> "1085 view_6" [label="(1, 56, 56, 96)", style=solid]; +"1085 view_6" -> "1086 permute_6" [label="(1, 7, 8, 7, 8, 96)", style=solid]; +"1086 permute_6" -> "1087 reshape_4" [label="(1, 7, 7, 8, 8, 96)", style=solid]; +"1087 reshape_4" -> "1088 quantize_per_tensor_default_7" [label="(49, 64, 96)", style=solid]; +"1087 reshape_4" -> "1117 new_zeros" [label="(49, 64, 96)", style=solid]; +"1088 quantize_per_tensor_default_7" -> "1089 dequantize_per_tensor_default_7" [label=None, style=solid]; +"1089 dequantize_per_tensor_default_7" -> "1093 linear_8" [label=None, style=solid]; +"1090 clone_1" -> "1091 slice_4" [label="(288,)", style=solid]; +"1090 clone_1" -> "1093 linear_8" [label="(288,)", style=solid]; +"1091 slice_4" -> "1092 zero__1" [label="(96,)", style=solid]; +"1093 linear_8" -> "1094 reshape_5" [label="(49, 64, 288)", style=solid]; +"1094 reshape_5" -> "1095 permute_7" [label="(49, 64, 3, 3, 32)", style=solid]; +"1095 permute_7" -> "1096 select_3" [label="(3, 49, 3, 64, 32)", style=solid]; +"1095 permute_7" -> "1097 select_4" [label="(3, 49, 3, 64, 32)", style=solid]; +"1095 permute_7" -> "1098 select_5" [label="(3, 49, 3, 64, 32)", style=solid]; +"1096 select_3" -> "1099 linalg_vector_norm_2" [label="(49, 3, 64, 32)", style=solid]; +"1096 select_3" -> "1101 expand_as_2" [label="(49, 3, 64, 32)", style=solid]; +"1096 select_3" -> "1102 div_2" [label="(49, 3, 64, 32)", style=solid]; +"1097 select_4" -> "1105 linalg_vector_norm_3" [label="(49, 3, 64, 32)", style=solid]; +"1097 select_4" -> "1107 expand_as_3" [label="(49, 3, 64, 32)", style=solid]; +"1097 select_4" -> "1108 div_3" [label="(49, 3, 64, 32)", style=solid]; +"1098 select_5" -> "1171 matmul_3" [label="(49, 3, 64, 32)", style=solid]; +"1099 linalg_vector_norm_2" -> "1100 clamp_min_2" [label="(49, 3, 64, 1)", style=solid]; +"1100 clamp_min_2" -> "1101 expand_as_2" [label="(49, 3, 64, 1)", style=solid]; +"1101 expand_as_2" -> "1102 div_2" [label="(49, 3, 64, 32)", style=solid]; +"1102 div_2" -> "1103 quantize_per_tensor_default_8" [label="(49, 3, 64, 32)", style=solid]; +"1103 quantize_per_tensor_default_8" -> "1104 dequantize_per_tensor_default_8" [label=None, style=solid]; +"1104 dequantize_per_tensor_default_8" -> "1112 matmul_2" [label=None, style=solid]; +"1105 linalg_vector_norm_3" -> "1106 clamp_min_3" [label="(49, 3, 64, 1)", style=solid]; +"1106 clamp_min_3" -> "1107 expand_as_3" [label="(49, 3, 64, 1)", style=solid]; +"1107 expand_as_3" -> "1108 div_3" [label="(49, 3, 64, 32)", style=solid]; +"1108 div_3" -> "1109 quantize_per_tensor_default_9" [label="(49, 3, 64, 32)", style=solid]; +"1109 quantize_per_tensor_default_9" -> "1110 dequantize_per_tensor_default_9" [label=None, style=solid]; +"1110 dequantize_per_tensor_default_9" -> "1111 transpose_2" [label=None, style=solid]; +"1111 transpose_2" -> "1112 matmul_2" [label="(49, 3, 32, 64)", style=solid]; +"1112 matmul_2" -> "1115 mul_3" [label="(49, 3, 64, 64)", style=solid]; +"1113 clamp_1" -> "1114 exp_1" [label="(3, 1, 1)", style=solid]; +"1114 exp_1" -> "1115 mul_3" [label="(3, 1, 1)", style=solid]; +"1115 mul_3" -> "1116 add_3" [label="(49, 3, 64, 64)", style=solid]; +"1116 add_3" -> "1164 view_8" [label="(49, 3, 64, 64)", style=solid]; +"1117 new_zeros" -> "1119 slice_5" [label="(56, 56)", style=solid]; +"1117 new_zeros" -> "1123 slice_7" [label="(56, 56)", style=solid]; +"1117 new_zeros" -> "1127 slice_9" [label="(56, 56)", style=solid]; +"1117 new_zeros" -> "1131 slice_11" [label="(56, 56)", style=solid]; +"1117 new_zeros" -> "1135 slice_13" [label="(56, 56)", style=solid]; +"1117 new_zeros" -> "1139 slice_15" [label="(56, 56)", style=solid]; +"1117 new_zeros" -> "1143 slice_17" [label="(56, 56)", style=solid]; +"1117 new_zeros" -> "1147 slice_19" [label="(56, 56)", style=solid]; +"1117 new_zeros" -> "1151 slice_21" [label="(56, 56)", style=solid]; +"1117 new_zeros" -> "1154 view_7" [label="(56, 56)", style=solid]; +"1118 lift_fresh_copy" -> "1121 fill_" [label="()", style=solid]; +"1119 slice_5" -> "1120 slice_6" [label="(48, 56)", style=solid]; +"1120 slice_6" -> "1121 fill_" [label="(48, 48)", style=solid]; +"1122 lift_fresh_copy_1" -> "1125 fill__1" [label="()", style=solid]; +"1123 slice_7" -> "1124 slice_8" [label="(48, 56)", style=solid]; +"1124 slice_8" -> "1125 fill__1" [label="(48, 4)", style=solid]; +"1126 lift_fresh_copy_2" -> "1129 fill__2" [label="()", style=solid]; +"1127 slice_9" -> "1128 slice_10" [label="(48, 56)", style=solid]; +"1128 slice_10" -> "1129 fill__2" [label="(48, 4)", style=solid]; +"1130 lift_fresh_copy_3" -> "1133 fill__3" [label="()", style=solid]; +"1131 slice_11" -> "1132 slice_12" [label="(4, 56)", style=solid]; +"1132 slice_12" -> "1133 fill__3" [label="(4, 48)", style=solid]; +"1134 lift_fresh_copy_4" -> "1137 fill__4" [label="()", style=solid]; +"1135 slice_13" -> "1136 slice_14" [label="(4, 56)", style=solid]; +"1136 slice_14" -> "1137 fill__4" [label="(4, 4)", style=solid]; +"1138 lift_fresh_copy_5" -> "1141 fill__5" [label="()", style=solid]; +"1139 slice_15" -> "1140 slice_16" [label="(4, 56)", style=solid]; +"1140 slice_16" -> "1141 fill__5" [label="(4, 4)", style=solid]; +"1142 lift_fresh_copy_6" -> "1145 fill__6" [label="()", style=solid]; +"1143 slice_17" -> "1144 slice_18" [label="(4, 56)", style=solid]; +"1144 slice_18" -> "1145 fill__6" [label="(4, 48)", style=solid]; +"1146 lift_fresh_copy_7" -> "1149 fill__7" [label="()", style=solid]; +"1147 slice_19" -> "1148 slice_20" [label="(4, 56)", style=solid]; +"1148 slice_20" -> "1149 fill__7" [label="(4, 4)", style=solid]; +"1150 lift_fresh_copy_8" -> "1153 fill__8" [label="()", style=solid]; +"1151 slice_21" -> "1152 slice_22" [label="(4, 56)", style=solid]; +"1152 slice_22" -> "1153 fill__8" [label="(4, 4)", style=solid]; +"1154 view_7" -> "1155 permute_8" [label="(7, 8, 7, 8)", style=solid]; +"1155 permute_8" -> "1156 reshape_6" [label="(7, 7, 8, 8)", style=solid]; +"1156 reshape_6" -> "1157 unsqueeze_2" [label="(49, 64)", style=solid]; +"1156 reshape_6" -> "1158 unsqueeze_3" [label="(49, 64)", style=solid]; +"1157 unsqueeze_2" -> "1159 sub" [label="(49, 1, 64)", style=solid]; +"1158 unsqueeze_3" -> "1159 sub" [label="(49, 64, 1)", style=solid]; +"1159 sub" -> "1160 ne" [label="(49, 64, 64)", style=solid]; +"1159 sub" -> "1161 masked_fill" [label="(49, 64, 64)", style=solid]; +"1159 sub" -> "1162 eq" [label="(49, 64, 64)", style=solid]; +"1160 ne" -> "1161 masked_fill" [label="(49, 64, 64)", style=solid]; +"1161 masked_fill" -> "1163 masked_fill_1" [label="(49, 64, 64)", style=solid]; +"1162 eq" -> "1163 masked_fill_1" [label="(49, 64, 64)", style=solid]; +"1163 masked_fill_1" -> "1165 unsqueeze_4" [label="(49, 64, 64)", style=solid]; +"1164 view_8" -> "1167 add_4" [label="(1, 49, 3, 64, 64)", style=solid]; +"1165 unsqueeze_4" -> "1166 unsqueeze_5" [label="(49, 1, 64, 64)", style=solid]; +"1166 unsqueeze_5" -> "1167 add_4" [label="(1, 49, 1, 64, 64)", style=solid]; +"1167 add_4" -> "1168 view_9" [label="(1, 49, 3, 64, 64)", style=solid]; +"1168 view_9" -> "1169 softmax_1" [label="(49, 3, 64, 64)", style=solid]; +"1169 softmax_1" -> "1170 dropout_4" [label="(49, 3, 64, 64)", style=solid]; +"1170 dropout_4" -> "1171 matmul_3" [label="(49, 3, 64, 64)", style=solid]; +"1171 matmul_3" -> "1172 quantize_per_tensor_default_10" [label="(49, 3, 64, 32)", style=solid]; +"1172 quantize_per_tensor_default_10" -> "1173 dequantize_per_tensor_default_10" [label=None, style=solid]; +"1173 dequantize_per_tensor_default_10" -> "1174 transpose_3" [label=None, style=solid]; +"1174 transpose_3" -> "1175 reshape_7" [label="(49, 64, 3, 32)", style=solid]; +"1175 reshape_7" -> "1176 linear_9" [label="(49, 64, 96)", style=solid]; +"1176 linear_9" -> "1177 dropout_5" [label="(49, 64, 96)", style=solid]; +"1177 dropout_5" -> "1178 view_10" [label="(49, 64, 96)", style=solid]; +"1178 view_10" -> "1179 permute_9" [label="(1, 7, 7, 8, 8, 96)", style=solid]; +"1179 permute_9" -> "1180 reshape_8" [label="(1, 7, 8, 7, 8, 96)", style=solid]; +"1180 reshape_8" -> "1181 roll_1" [label="(1, 56, 56, 96)", style=solid]; +"1181 roll_1" -> "1182 slice_23" [label="(1, 56, 56, 96)", style=solid]; +"1182 slice_23" -> "1183 slice_24" [label="(1, 56, 56, 96)", style=solid]; +"1183 slice_24" -> "1184 layer_norm_3" [label="(1, 56, 56, 96)", style=solid]; +"1184 layer_norm_3" -> "1185 add_5" [label="(1, 56, 56, 96)", style=solid]; +"1185 add_5" -> "1186 quantize_per_tensor_default_11" [label="(1, 56, 56, 96)", style=solid]; +"1185 add_5" -> "1196 add_6" [label="(1, 56, 56, 96)", style=solid]; +"1186 quantize_per_tensor_default_11" -> "1187 dequantize_per_tensor_default_11" [label=None, style=solid]; +"1187 dequantize_per_tensor_default_11" -> "1188 linear_10" [label=None, style=solid]; +"1188 linear_10" -> "1189 gelu_1" [label="(1, 56, 56, 384)", style=solid]; +"1189 gelu_1" -> "1190 quantize_per_tensor_default_12" [label="(1, 56, 56, 384)", style=solid]; +"1190 quantize_per_tensor_default_12" -> "1191 dequantize_per_tensor_default_12" [label=None, style=solid]; +"1191 dequantize_per_tensor_default_12" -> "1192 dropout_6" [label=None, style=solid]; +"1192 dropout_6" -> "1193 linear_11" [label="(1, 56, 56, 384)", style=solid]; +"1193 linear_11" -> "1194 dropout_7" [label="(1, 56, 56, 96)", style=solid]; +"1194 dropout_7" -> "1195 layer_norm_4" [label="(1, 56, 56, 96)", style=solid]; +"1195 layer_norm_4" -> "1196 add_6" [label="(1, 56, 56, 96)", style=solid]; +"1196 add_6" -> "1197 quantize_per_tensor_default_13" [label="(1, 56, 56, 96)", style=solid]; +"1197 quantize_per_tensor_default_13" -> "1198 dequantize_per_tensor_default_13" [label=None, style=solid]; +"1198 dequantize_per_tensor_default_13" -> "1199 pad_2" [label=None, style=solid]; +"1199 pad_2" -> "1200 slice_25" [label="(1, 56, 56, 96)", style=solid]; +"1199 pad_2" -> "1203 slice_28" [label="(1, 56, 56, 96)", style=solid]; +"1199 pad_2" -> "1206 slice_31" [label="(1, 56, 56, 96)", style=solid]; +"1199 pad_2" -> "1209 slice_34" [label="(1, 56, 56, 96)", style=solid]; +"1200 slice_25" -> "1201 slice_26" [label="(1, 28, 56, 96)", style=solid]; +"1201 slice_26" -> "1202 slice_27" [label="(1, 28, 28, 96)", style=solid]; +"1202 slice_27" -> "1212 cat" [label="(1, 28, 28, 96)", style=solid]; +"1203 slice_28" -> "1204 slice_29" [label="(1, 28, 56, 96)", style=solid]; +"1204 slice_29" -> "1205 slice_30" [label="(1, 28, 28, 96)", style=solid]; +"1205 slice_30" -> "1212 cat" [label="(1, 28, 28, 96)", style=solid]; +"1206 slice_31" -> "1207 slice_32" [label="(1, 28, 56, 96)", style=solid]; +"1207 slice_32" -> "1208 slice_33" [label="(1, 28, 28, 96)", style=solid]; +"1208 slice_33" -> "1212 cat" [label="(1, 28, 28, 96)", style=solid]; +"1209 slice_34" -> "1210 slice_35" [label="(1, 28, 56, 96)", style=solid]; +"1210 slice_35" -> "1211 slice_36" [label="(1, 28, 28, 96)", style=solid]; +"1211 slice_36" -> "1212 cat" [label="(1, 28, 28, 96)", style=solid]; +"1212 cat" -> "1213 linear_12" [label="(1, 28, 28, 384)", style=solid]; +"1213 linear_12" -> "1214 layer_norm_5" [label="(1, 28, 28, 192)", style=solid]; +"1214 layer_norm_5" -> "1215 quantize_per_tensor_default_14" [label="(1, 28, 28, 192)", style=solid]; +"1214 layer_norm_5" -> "1277 add_8" [label="(1, 28, 28, 192)", style=solid]; +"1215 quantize_per_tensor_default_14" -> "1216 dequantize_per_tensor_default_14" [label=None, style=solid]; +"1216 dequantize_per_tensor_default_14" -> "1228 pad_3" [label=None, style=solid]; +"1217 linear_13" -> "1218 relu__2" [label="(1, 15, 15, 512)", style=solid]; +"1218 relu__2" -> "1219 linear_14" [label="(1, 15, 15, 512)", style=solid]; +"1219 linear_14" -> "1220 view_11" [label="(1, 15, 15, 6)", style=solid]; +"1220 view_11" -> "1221 index_2" [label="(225, 6)", style=solid]; +"1221 index_2" -> "1222 view_12" [label="(4096, 6)", style=solid]; +"1222 view_12" -> "1223 permute_10" [label="(64, 64, 6)", style=solid]; +"1223 permute_10" -> "1224 contiguous_2" [label="(6, 64, 64)", style=solid]; +"1224 contiguous_2" -> "1225 unsqueeze_6" [label="(6, 64, 64)", style=solid]; +"1225 unsqueeze_6" -> "1226 sigmoid_2" [label="(1, 6, 64, 64)", style=solid]; +"1226 sigmoid_2" -> "1227 mul_4" [label="(1, 6, 64, 64)", style=solid]; +"1227 mul_4" -> "1258 add_7" [label="(1, 6, 64, 64)", style=solid]; +"1228 pad_3" -> "1229 view_13" [label="(1, 32, 32, 192)", style=solid]; +"1229 view_13" -> "1230 permute_11" [label="(1, 4, 8, 4, 8, 192)", style=solid]; +"1230 permute_11" -> "1231 reshape_9" [label="(1, 4, 4, 8, 8, 192)", style=solid]; +"1231 reshape_9" -> "1235 linear_15" [label="(16, 64, 192)", style=solid]; +"1232 clone_2" -> "1233 slice_37" [label="(576,)", style=solid]; +"1232 clone_2" -> "1235 linear_15" [label="(576,)", style=solid]; +"1233 slice_37" -> "1234 zero__2" [label="(192,)", style=solid]; +"1235 linear_15" -> "1236 reshape_10" [label="(16, 64, 576)", style=solid]; +"1236 reshape_10" -> "1237 permute_12" [label="(16, 64, 3, 6, 32)", style=solid]; +"1237 permute_12" -> "1238 select_6" [label="(3, 16, 6, 64, 32)", style=solid]; +"1237 permute_12" -> "1239 select_7" [label="(3, 16, 6, 64, 32)", style=solid]; +"1237 permute_12" -> "1240 select_8" [label="(3, 16, 6, 64, 32)", style=solid]; +"1238 select_6" -> "1241 linalg_vector_norm_4" [label="(16, 6, 64, 32)", style=solid]; +"1238 select_6" -> "1243 expand_as_4" [label="(16, 6, 64, 32)", style=solid]; +"1238 select_6" -> "1244 div_4" [label="(16, 6, 64, 32)", style=solid]; +"1239 select_7" -> "1247 linalg_vector_norm_5" [label="(16, 6, 64, 32)", style=solid]; +"1239 select_7" -> "1249 expand_as_5" [label="(16, 6, 64, 32)", style=solid]; +"1239 select_7" -> "1250 div_5" [label="(16, 6, 64, 32)", style=solid]; +"1240 select_8" -> "1261 matmul_5" [label="(16, 6, 64, 32)", style=solid]; +"1241 linalg_vector_norm_4" -> "1242 clamp_min_4" [label="(16, 6, 64, 1)", style=solid]; +"1242 clamp_min_4" -> "1243 expand_as_4" [label="(16, 6, 64, 1)", style=solid]; +"1243 expand_as_4" -> "1244 div_4" [label="(16, 6, 64, 32)", style=solid]; +"1244 div_4" -> "1245 quantize_per_tensor_default_15" [label="(16, 6, 64, 32)", style=solid]; +"1245 quantize_per_tensor_default_15" -> "1246 dequantize_per_tensor_default_15" [label=None, style=solid]; +"1246 dequantize_per_tensor_default_15" -> "1254 matmul_4" [label=None, style=solid]; +"1247 linalg_vector_norm_5" -> "1248 clamp_min_5" [label="(16, 6, 64, 1)", style=solid]; +"1248 clamp_min_5" -> "1249 expand_as_5" [label="(16, 6, 64, 1)", style=solid]; +"1249 expand_as_5" -> "1250 div_5" [label="(16, 6, 64, 32)", style=solid]; +"1250 div_5" -> "1251 quantize_per_tensor_default_16" [label="(16, 6, 64, 32)", style=solid]; +"1251 quantize_per_tensor_default_16" -> "1252 dequantize_per_tensor_default_16" [label=None, style=solid]; +"1252 dequantize_per_tensor_default_16" -> "1253 transpose_4" [label=None, style=solid]; +"1253 transpose_4" -> "1254 matmul_4" [label="(16, 6, 32, 64)", style=solid]; +"1254 matmul_4" -> "1257 mul_5" [label="(16, 6, 64, 64)", style=solid]; +"1255 clamp_2" -> "1256 exp_2" [label="(6, 1, 1)", style=solid]; +"1256 exp_2" -> "1257 mul_5" [label="(6, 1, 1)", style=solid]; +"1257 mul_5" -> "1258 add_7" [label="(16, 6, 64, 64)", style=solid]; +"1258 add_7" -> "1259 softmax_2" [label="(16, 6, 64, 64)", style=solid]; +"1259 softmax_2" -> "1260 dropout_8" [label="(16, 6, 64, 64)", style=solid]; +"1260 dropout_8" -> "1261 matmul_5" [label="(16, 6, 64, 64)", style=solid]; +"1261 matmul_5" -> "1262 quantize_per_tensor_default_17" [label="(16, 6, 64, 32)", style=solid]; +"1262 quantize_per_tensor_default_17" -> "1263 dequantize_per_tensor_default_17" [label=None, style=solid]; +"1263 dequantize_per_tensor_default_17" -> "1264 transpose_5" [label=None, style=solid]; +"1264 transpose_5" -> "1265 reshape_11" [label="(16, 64, 6, 32)", style=solid]; +"1265 reshape_11" -> "1266 linear_16" [label="(16, 64, 192)", style=solid]; +"1266 linear_16" -> "1267 dropout_9" [label="(16, 64, 192)", style=solid]; +"1267 dropout_9" -> "1268 view_14" [label="(16, 64, 192)", style=solid]; +"1268 view_14" -> "1269 permute_13" [label="(1, 4, 4, 8, 8, 192)", style=solid]; +"1269 permute_13" -> "1270 reshape_12" [label="(1, 4, 8, 4, 8, 192)", style=solid]; +"1270 reshape_12" -> "1271 slice_38" [label="(1, 32, 32, 192)", style=solid]; +"1271 slice_38" -> "1272 slice_39" [label="(1, 32, 32, 192)", style=solid]; +"1272 slice_39" -> "1273 slice_40" [label="(1, 28, 32, 192)", style=solid]; +"1273 slice_40" -> "1274 slice_41" [label="(1, 28, 28, 192)", style=solid]; +"1274 slice_41" -> "1275 contiguous_3" [label="(1, 28, 28, 192)", style=solid]; +"1275 contiguous_3" -> "1276 layer_norm_6" [label="(1, 28, 28, 192)", style=solid]; +"1276 layer_norm_6" -> "1277 add_8" [label="(1, 28, 28, 192)", style=solid]; +"1277 add_8" -> "1278 quantize_per_tensor_default_18" [label="(1, 28, 28, 192)", style=solid]; +"1277 add_8" -> "1288 add_9" [label="(1, 28, 28, 192)", style=solid]; +"1278 quantize_per_tensor_default_18" -> "1279 dequantize_per_tensor_default_18" [label=None, style=solid]; +"1279 dequantize_per_tensor_default_18" -> "1280 linear_17" [label=None, style=solid]; +"1280 linear_17" -> "1281 gelu_2" [label="(1, 28, 28, 768)", style=solid]; +"1281 gelu_2" -> "1282 quantize_per_tensor_default_19" [label="(1, 28, 28, 768)", style=solid]; +"1282 quantize_per_tensor_default_19" -> "1283 dequantize_per_tensor_default_19" [label=None, style=solid]; +"1283 dequantize_per_tensor_default_19" -> "1284 dropout_10" [label=None, style=solid]; +"1284 dropout_10" -> "1285 linear_18" [label="(1, 28, 28, 768)", style=solid]; +"1285 linear_18" -> "1286 dropout_11" [label="(1, 28, 28, 192)", style=solid]; +"1286 dropout_11" -> "1287 layer_norm_7" [label="(1, 28, 28, 192)", style=solid]; +"1287 layer_norm_7" -> "1288 add_9" [label="(1, 28, 28, 192)", style=solid]; +"1288 add_9" -> "1300 pad_4" [label="(1, 28, 28, 192)", style=solid]; +"1288 add_9" -> "1405 add_12" [label="(1, 28, 28, 192)", style=solid]; +"1289 linear_19" -> "1290 relu__3" [label="(1, 15, 15, 512)", style=solid]; +"1290 relu__3" -> "1291 linear_20" [label="(1, 15, 15, 512)", style=solid]; +"1291 linear_20" -> "1292 view_15" [label="(1, 15, 15, 6)", style=solid]; +"1292 view_15" -> "1293 index_3" [label="(225, 6)", style=solid]; +"1293 index_3" -> "1294 view_16" [label="(4096, 6)", style=solid]; +"1294 view_16" -> "1295 permute_14" [label="(64, 64, 6)", style=solid]; +"1295 permute_14" -> "1296 contiguous_4" [label="(6, 64, 64)", style=solid]; +"1296 contiguous_4" -> "1297 unsqueeze_7" [label="(6, 64, 64)", style=solid]; +"1297 unsqueeze_7" -> "1298 sigmoid_3" [label="(1, 6, 64, 64)", style=solid]; +"1298 sigmoid_3" -> "1299 mul_6" [label="(1, 6, 64, 64)", style=solid]; +"1299 mul_6" -> "1333 add_10" [label="(1, 6, 64, 64)", style=solid]; +"1300 pad_4" -> "1301 roll_2" [label="(1, 32, 32, 192)", style=solid]; +"1301 roll_2" -> "1302 view_17" [label="(1, 32, 32, 192)", style=solid]; +"1302 view_17" -> "1303 permute_15" [label="(1, 4, 8, 4, 8, 192)", style=solid]; +"1303 permute_15" -> "1304 reshape_13" [label="(1, 4, 4, 8, 8, 192)", style=solid]; +"1304 reshape_13" -> "1305 quantize_per_tensor_default_20" [label="(16, 64, 192)", style=solid]; +"1304 reshape_13" -> "1334 new_zeros_1" [label="(16, 64, 192)", style=solid]; +"1305 quantize_per_tensor_default_20" -> "1306 dequantize_per_tensor_default_20" [label=None, style=solid]; +"1306 dequantize_per_tensor_default_20" -> "1310 linear_21" [label=None, style=solid]; +"1307 clone_3" -> "1308 slice_42" [label="(576,)", style=solid]; +"1307 clone_3" -> "1310 linear_21" [label="(576,)", style=solid]; +"1308 slice_42" -> "1309 zero__3" [label="(192,)", style=solid]; +"1310 linear_21" -> "1311 reshape_14" [label="(16, 64, 576)", style=solid]; +"1311 reshape_14" -> "1312 permute_16" [label="(16, 64, 3, 6, 32)", style=solid]; +"1312 permute_16" -> "1313 select_9" [label="(3, 16, 6, 64, 32)", style=solid]; +"1312 permute_16" -> "1314 select_10" [label="(3, 16, 6, 64, 32)", style=solid]; +"1312 permute_16" -> "1315 select_11" [label="(3, 16, 6, 64, 32)", style=solid]; +"1313 select_9" -> "1316 linalg_vector_norm_6" [label="(16, 6, 64, 32)", style=solid]; +"1313 select_9" -> "1318 expand_as_6" [label="(16, 6, 64, 32)", style=solid]; +"1313 select_9" -> "1319 div_6" [label="(16, 6, 64, 32)", style=solid]; +"1314 select_10" -> "1322 linalg_vector_norm_7" [label="(16, 6, 64, 32)", style=solid]; +"1314 select_10" -> "1324 expand_as_7" [label="(16, 6, 64, 32)", style=solid]; +"1314 select_10" -> "1325 div_7" [label="(16, 6, 64, 32)", style=solid]; +"1315 select_11" -> "1388 matmul_7" [label="(16, 6, 64, 32)", style=solid]; +"1316 linalg_vector_norm_6" -> "1317 clamp_min_6" [label="(16, 6, 64, 1)", style=solid]; +"1317 clamp_min_6" -> "1318 expand_as_6" [label="(16, 6, 64, 1)", style=solid]; +"1318 expand_as_6" -> "1319 div_6" [label="(16, 6, 64, 32)", style=solid]; +"1319 div_6" -> "1320 quantize_per_tensor_default_21" [label="(16, 6, 64, 32)", style=solid]; +"1320 quantize_per_tensor_default_21" -> "1321 dequantize_per_tensor_default_21" [label=None, style=solid]; +"1321 dequantize_per_tensor_default_21" -> "1329 matmul_6" [label=None, style=solid]; +"1322 linalg_vector_norm_7" -> "1323 clamp_min_7" [label="(16, 6, 64, 1)", style=solid]; +"1323 clamp_min_7" -> "1324 expand_as_7" [label="(16, 6, 64, 1)", style=solid]; +"1324 expand_as_7" -> "1325 div_7" [label="(16, 6, 64, 32)", style=solid]; +"1325 div_7" -> "1326 quantize_per_tensor_default_22" [label="(16, 6, 64, 32)", style=solid]; +"1326 quantize_per_tensor_default_22" -> "1327 dequantize_per_tensor_default_22" [label=None, style=solid]; +"1327 dequantize_per_tensor_default_22" -> "1328 transpose_6" [label=None, style=solid]; +"1328 transpose_6" -> "1329 matmul_6" [label="(16, 6, 32, 64)", style=solid]; +"1329 matmul_6" -> "1332 mul_7" [label="(16, 6, 64, 64)", style=solid]; +"1330 clamp_3" -> "1331 exp_3" [label="(6, 1, 1)", style=solid]; +"1331 exp_3" -> "1332 mul_7" [label="(6, 1, 1)", style=solid]; +"1332 mul_7" -> "1333 add_10" [label="(16, 6, 64, 64)", style=solid]; +"1333 add_10" -> "1381 view_19" [label="(16, 6, 64, 64)", style=solid]; +"1334 new_zeros_1" -> "1336 slice_43" [label="(32, 32)", style=solid]; +"1334 new_zeros_1" -> "1340 slice_45" [label="(32, 32)", style=solid]; +"1334 new_zeros_1" -> "1344 slice_47" [label="(32, 32)", style=solid]; +"1334 new_zeros_1" -> "1348 slice_49" [label="(32, 32)", style=solid]; +"1334 new_zeros_1" -> "1352 slice_51" [label="(32, 32)", style=solid]; +"1334 new_zeros_1" -> "1356 slice_53" [label="(32, 32)", style=solid]; +"1334 new_zeros_1" -> "1360 slice_55" [label="(32, 32)", style=solid]; +"1334 new_zeros_1" -> "1364 slice_57" [label="(32, 32)", style=solid]; +"1334 new_zeros_1" -> "1368 slice_59" [label="(32, 32)", style=solid]; +"1334 new_zeros_1" -> "1371 view_18" [label="(32, 32)", style=solid]; +"1335 lift_fresh_copy_9" -> "1338 fill__9" [label="()", style=solid]; +"1336 slice_43" -> "1337 slice_44" [label="(24, 32)", style=solid]; +"1337 slice_44" -> "1338 fill__9" [label="(24, 24)", style=solid]; +"1339 lift_fresh_copy_10" -> "1342 fill__10" [label="()", style=solid]; +"1340 slice_45" -> "1341 slice_46" [label="(24, 32)", style=solid]; +"1341 slice_46" -> "1342 fill__10" [label="(24, 4)", style=solid]; +"1343 lift_fresh_copy_11" -> "1346 fill__11" [label="()", style=solid]; +"1344 slice_47" -> "1345 slice_48" [label="(24, 32)", style=solid]; +"1345 slice_48" -> "1346 fill__11" [label="(24, 4)", style=solid]; +"1347 lift_fresh_copy_12" -> "1350 fill__12" [label="()", style=solid]; +"1348 slice_49" -> "1349 slice_50" [label="(4, 32)", style=solid]; +"1349 slice_50" -> "1350 fill__12" [label="(4, 24)", style=solid]; +"1351 lift_fresh_copy_13" -> "1354 fill__13" [label="()", style=solid]; +"1352 slice_51" -> "1353 slice_52" [label="(4, 32)", style=solid]; +"1353 slice_52" -> "1354 fill__13" [label="(4, 4)", style=solid]; +"1355 lift_fresh_copy_14" -> "1358 fill__14" [label="()", style=solid]; +"1356 slice_53" -> "1357 slice_54" [label="(4, 32)", style=solid]; +"1357 slice_54" -> "1358 fill__14" [label="(4, 4)", style=solid]; +"1359 lift_fresh_copy_15" -> "1362 fill__15" [label="()", style=solid]; +"1360 slice_55" -> "1361 slice_56" [label="(4, 32)", style=solid]; +"1361 slice_56" -> "1362 fill__15" [label="(4, 24)", style=solid]; +"1363 lift_fresh_copy_16" -> "1366 fill__16" [label="()", style=solid]; +"1364 slice_57" -> "1365 slice_58" [label="(4, 32)", style=solid]; +"1365 slice_58" -> "1366 fill__16" [label="(4, 4)", style=solid]; +"1367 lift_fresh_copy_17" -> "1370 fill__17" [label="()", style=solid]; +"1368 slice_59" -> "1369 slice_60" [label="(4, 32)", style=solid]; +"1369 slice_60" -> "1370 fill__17" [label="(4, 4)", style=solid]; +"1371 view_18" -> "1372 permute_17" [label="(4, 8, 4, 8)", style=solid]; +"1372 permute_17" -> "1373 reshape_15" [label="(4, 4, 8, 8)", style=solid]; +"1373 reshape_15" -> "1374 unsqueeze_8" [label="(16, 64)", style=solid]; +"1373 reshape_15" -> "1375 unsqueeze_9" [label="(16, 64)", style=solid]; +"1374 unsqueeze_8" -> "1376 sub_1" [label="(16, 1, 64)", style=solid]; +"1375 unsqueeze_9" -> "1376 sub_1" [label="(16, 64, 1)", style=solid]; +"1376 sub_1" -> "1377 ne_1" [label="(16, 64, 64)", style=solid]; +"1376 sub_1" -> "1378 masked_fill_2" [label="(16, 64, 64)", style=solid]; +"1376 sub_1" -> "1379 eq_1" [label="(16, 64, 64)", style=solid]; +"1377 ne_1" -> "1378 masked_fill_2" [label="(16, 64, 64)", style=solid]; +"1378 masked_fill_2" -> "1380 masked_fill_3" [label="(16, 64, 64)", style=solid]; +"1379 eq_1" -> "1380 masked_fill_3" [label="(16, 64, 64)", style=solid]; +"1380 masked_fill_3" -> "1382 unsqueeze_10" [label="(16, 64, 64)", style=solid]; +"1381 view_19" -> "1384 add_11" [label="(1, 16, 6, 64, 64)", style=solid]; +"1382 unsqueeze_10" -> "1383 unsqueeze_11" [label="(16, 1, 64, 64)", style=solid]; +"1383 unsqueeze_11" -> "1384 add_11" [label="(1, 16, 1, 64, 64)", style=solid]; +"1384 add_11" -> "1385 view_20" [label="(1, 16, 6, 64, 64)", style=solid]; +"1385 view_20" -> "1386 softmax_3" [label="(16, 6, 64, 64)", style=solid]; +"1386 softmax_3" -> "1387 dropout_12" [label="(16, 6, 64, 64)", style=solid]; +"1387 dropout_12" -> "1388 matmul_7" [label="(16, 6, 64, 64)", style=solid]; +"1388 matmul_7" -> "1389 quantize_per_tensor_default_23" [label="(16, 6, 64, 32)", style=solid]; +"1389 quantize_per_tensor_default_23" -> "1390 dequantize_per_tensor_default_23" [label=None, style=solid]; +"1390 dequantize_per_tensor_default_23" -> "1391 transpose_7" [label=None, style=solid]; +"1391 transpose_7" -> "1392 reshape_16" [label="(16, 64, 6, 32)", style=solid]; +"1392 reshape_16" -> "1393 linear_22" [label="(16, 64, 192)", style=solid]; +"1393 linear_22" -> "1394 dropout_13" [label="(16, 64, 192)", style=solid]; +"1394 dropout_13" -> "1395 view_21" [label="(16, 64, 192)", style=solid]; +"1395 view_21" -> "1396 permute_18" [label="(1, 4, 4, 8, 8, 192)", style=solid]; +"1396 permute_18" -> "1397 reshape_17" [label="(1, 4, 8, 4, 8, 192)", style=solid]; +"1397 reshape_17" -> "1398 roll_3" [label="(1, 32, 32, 192)", style=solid]; +"1398 roll_3" -> "1399 slice_61" [label="(1, 32, 32, 192)", style=solid]; +"1399 slice_61" -> "1400 slice_62" [label="(1, 32, 32, 192)", style=solid]; +"1400 slice_62" -> "1401 slice_63" [label="(1, 28, 32, 192)", style=solid]; +"1401 slice_63" -> "1402 slice_64" [label="(1, 28, 28, 192)", style=solid]; +"1402 slice_64" -> "1403 contiguous_5" [label="(1, 28, 28, 192)", style=solid]; +"1403 contiguous_5" -> "1404 layer_norm_8" [label="(1, 28, 28, 192)", style=solid]; +"1404 layer_norm_8" -> "1405 add_12" [label="(1, 28, 28, 192)", style=solid]; +"1405 add_12" -> "1406 quantize_per_tensor_default_24" [label="(1, 28, 28, 192)", style=solid]; +"1405 add_12" -> "1416 add_13" [label="(1, 28, 28, 192)", style=solid]; +"1406 quantize_per_tensor_default_24" -> "1407 dequantize_per_tensor_default_24" [label=None, style=solid]; +"1407 dequantize_per_tensor_default_24" -> "1408 linear_23" [label=None, style=solid]; +"1408 linear_23" -> "1409 gelu_3" [label="(1, 28, 28, 768)", style=solid]; +"1409 gelu_3" -> "1410 quantize_per_tensor_default_25" [label="(1, 28, 28, 768)", style=solid]; +"1410 quantize_per_tensor_default_25" -> "1411 dequantize_per_tensor_default_25" [label=None, style=solid]; +"1411 dequantize_per_tensor_default_25" -> "1412 dropout_14" [label=None, style=solid]; +"1412 dropout_14" -> "1413 linear_24" [label="(1, 28, 28, 768)", style=solid]; +"1413 linear_24" -> "1414 dropout_15" [label="(1, 28, 28, 192)", style=solid]; +"1414 dropout_15" -> "1415 layer_norm_9" [label="(1, 28, 28, 192)", style=solid]; +"1415 layer_norm_9" -> "1416 add_13" [label="(1, 28, 28, 192)", style=solid]; +"1416 add_13" -> "1417 quantize_per_tensor_default_26" [label="(1, 28, 28, 192)", style=solid]; +"1417 quantize_per_tensor_default_26" -> "1418 dequantize_per_tensor_default_26" [label=None, style=solid]; +"1418 dequantize_per_tensor_default_26" -> "1419 pad_5" [label=None, style=solid]; +"1419 pad_5" -> "1420 slice_65" [label="(1, 28, 28, 192)", style=solid]; +"1419 pad_5" -> "1423 slice_68" [label="(1, 28, 28, 192)", style=solid]; +"1419 pad_5" -> "1426 slice_71" [label="(1, 28, 28, 192)", style=solid]; +"1419 pad_5" -> "1429 slice_74" [label="(1, 28, 28, 192)", style=solid]; +"1420 slice_65" -> "1421 slice_66" [label="(1, 14, 28, 192)", style=solid]; +"1421 slice_66" -> "1422 slice_67" [label="(1, 14, 14, 192)", style=solid]; +"1422 slice_67" -> "1432 cat_1" [label="(1, 14, 14, 192)", style=solid]; +"1423 slice_68" -> "1424 slice_69" [label="(1, 14, 28, 192)", style=solid]; +"1424 slice_69" -> "1425 slice_70" [label="(1, 14, 14, 192)", style=solid]; +"1425 slice_70" -> "1432 cat_1" [label="(1, 14, 14, 192)", style=solid]; +"1426 slice_71" -> "1427 slice_72" [label="(1, 14, 28, 192)", style=solid]; +"1427 slice_72" -> "1428 slice_73" [label="(1, 14, 14, 192)", style=solid]; +"1428 slice_73" -> "1432 cat_1" [label="(1, 14, 14, 192)", style=solid]; +"1429 slice_74" -> "1430 slice_75" [label="(1, 14, 28, 192)", style=solid]; +"1430 slice_75" -> "1431 slice_76" [label="(1, 14, 14, 192)", style=solid]; +"1431 slice_76" -> "1432 cat_1" [label="(1, 14, 14, 192)", style=solid]; +"1432 cat_1" -> "1433 linear_25" [label="(1, 14, 14, 768)", style=solid]; +"1433 linear_25" -> "1434 layer_norm_10" [label="(1, 14, 14, 384)", style=solid]; +"1434 layer_norm_10" -> "1435 quantize_per_tensor_default_27" [label="(1, 14, 14, 384)", style=solid]; +"1434 layer_norm_10" -> "1497 add_15" [label="(1, 14, 14, 384)", style=solid]; +"1435 quantize_per_tensor_default_27" -> "1436 dequantize_per_tensor_default_27" [label=None, style=solid]; +"1436 dequantize_per_tensor_default_27" -> "1448 pad_6" [label=None, style=solid]; +"1437 linear_26" -> "1438 relu__4" [label="(1, 15, 15, 512)", style=solid]; +"1438 relu__4" -> "1439 linear_27" [label="(1, 15, 15, 512)", style=solid]; +"1439 linear_27" -> "1440 view_22" [label="(1, 15, 15, 12)", style=solid]; +"1440 view_22" -> "1441 index_4" [label="(225, 12)", style=solid]; +"1441 index_4" -> "1442 view_23" [label="(4096, 12)", style=solid]; +"1442 view_23" -> "1443 permute_19" [label="(64, 64, 12)", style=solid]; +"1443 permute_19" -> "1444 contiguous_6" [label="(12, 64, 64)", style=solid]; +"1444 contiguous_6" -> "1445 unsqueeze_12" [label="(12, 64, 64)", style=solid]; +"1445 unsqueeze_12" -> "1446 sigmoid_4" [label="(1, 12, 64, 64)", style=solid]; +"1446 sigmoid_4" -> "1447 mul_8" [label="(1, 12, 64, 64)", style=solid]; +"1447 mul_8" -> "1478 add_14" [label="(1, 12, 64, 64)", style=solid]; +"1448 pad_6" -> "1449 view_24" [label="(1, 16, 16, 384)", style=solid]; +"1449 view_24" -> "1450 permute_20" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1450 permute_20" -> "1451 reshape_18" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1451 reshape_18" -> "1455 linear_28" [label="(4, 64, 384)", style=solid]; +"1452 clone_4" -> "1453 slice_77" [label="(1152,)", style=solid]; +"1452 clone_4" -> "1455 linear_28" [label="(1152,)", style=solid]; +"1453 slice_77" -> "1454 zero__4" [label="(384,)", style=solid]; +"1455 linear_28" -> "1456 reshape_19" [label="(4, 64, 1152)", style=solid]; +"1456 reshape_19" -> "1457 permute_21" [label="(4, 64, 3, 12, 32)", style=solid]; +"1457 permute_21" -> "1458 select_12" [label="(3, 4, 12, 64, 32)", style=solid]; +"1457 permute_21" -> "1459 select_13" [label="(3, 4, 12, 64, 32)", style=solid]; +"1457 permute_21" -> "1460 select_14" [label="(3, 4, 12, 64, 32)", style=solid]; +"1458 select_12" -> "1461 linalg_vector_norm_8" [label="(4, 12, 64, 32)", style=solid]; +"1458 select_12" -> "1463 expand_as_8" [label="(4, 12, 64, 32)", style=solid]; +"1458 select_12" -> "1464 div_8" [label="(4, 12, 64, 32)", style=solid]; +"1459 select_13" -> "1467 linalg_vector_norm_9" [label="(4, 12, 64, 32)", style=solid]; +"1459 select_13" -> "1469 expand_as_9" [label="(4, 12, 64, 32)", style=solid]; +"1459 select_13" -> "1470 div_9" [label="(4, 12, 64, 32)", style=solid]; +"1460 select_14" -> "1481 matmul_9" [label="(4, 12, 64, 32)", style=solid]; +"1461 linalg_vector_norm_8" -> "1462 clamp_min_8" [label="(4, 12, 64, 1)", style=solid]; +"1462 clamp_min_8" -> "1463 expand_as_8" [label="(4, 12, 64, 1)", style=solid]; +"1463 expand_as_8" -> "1464 div_8" [label="(4, 12, 64, 32)", style=solid]; +"1464 div_8" -> "1465 quantize_per_tensor_default_28" [label="(4, 12, 64, 32)", style=solid]; +"1465 quantize_per_tensor_default_28" -> "1466 dequantize_per_tensor_default_28" [label=None, style=solid]; +"1466 dequantize_per_tensor_default_28" -> "1474 matmul_8" [label=None, style=solid]; +"1467 linalg_vector_norm_9" -> "1468 clamp_min_9" [label="(4, 12, 64, 1)", style=solid]; +"1468 clamp_min_9" -> "1469 expand_as_9" [label="(4, 12, 64, 1)", style=solid]; +"1469 expand_as_9" -> "1470 div_9" [label="(4, 12, 64, 32)", style=solid]; +"1470 div_9" -> "1471 quantize_per_tensor_default_29" [label="(4, 12, 64, 32)", style=solid]; +"1471 quantize_per_tensor_default_29" -> "1472 dequantize_per_tensor_default_29" [label=None, style=solid]; +"1472 dequantize_per_tensor_default_29" -> "1473 transpose_8" [label=None, style=solid]; +"1473 transpose_8" -> "1474 matmul_8" [label="(4, 12, 32, 64)", style=solid]; +"1474 matmul_8" -> "1477 mul_9" [label="(4, 12, 64, 64)", style=solid]; +"1475 clamp_4" -> "1476 exp_4" [label="(12, 1, 1)", style=solid]; +"1476 exp_4" -> "1477 mul_9" [label="(12, 1, 1)", style=solid]; +"1477 mul_9" -> "1478 add_14" [label="(4, 12, 64, 64)", style=solid]; +"1478 add_14" -> "1479 softmax_4" [label="(4, 12, 64, 64)", style=solid]; +"1479 softmax_4" -> "1480 dropout_16" [label="(4, 12, 64, 64)", style=solid]; +"1480 dropout_16" -> "1481 matmul_9" [label="(4, 12, 64, 64)", style=solid]; +"1481 matmul_9" -> "1482 quantize_per_tensor_default_30" [label="(4, 12, 64, 32)", style=solid]; +"1482 quantize_per_tensor_default_30" -> "1483 dequantize_per_tensor_default_30" [label=None, style=solid]; +"1483 dequantize_per_tensor_default_30" -> "1484 transpose_9" [label=None, style=solid]; +"1484 transpose_9" -> "1485 reshape_20" [label="(4, 64, 12, 32)", style=solid]; +"1485 reshape_20" -> "1486 linear_29" [label="(4, 64, 384)", style=solid]; +"1486 linear_29" -> "1487 dropout_17" [label="(4, 64, 384)", style=solid]; +"1487 dropout_17" -> "1488 view_25" [label="(4, 64, 384)", style=solid]; +"1488 view_25" -> "1489 permute_22" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1489 permute_22" -> "1490 reshape_21" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1490 reshape_21" -> "1491 slice_78" [label="(1, 16, 16, 384)", style=solid]; +"1491 slice_78" -> "1492 slice_79" [label="(1, 16, 16, 384)", style=solid]; +"1492 slice_79" -> "1493 slice_80" [label="(1, 14, 16, 384)", style=solid]; +"1493 slice_80" -> "1494 slice_81" [label="(1, 14, 14, 384)", style=solid]; +"1494 slice_81" -> "1495 contiguous_7" [label="(1, 14, 14, 384)", style=solid]; +"1495 contiguous_7" -> "1496 layer_norm_11" [label="(1, 14, 14, 384)", style=solid]; +"1496 layer_norm_11" -> "1497 add_15" [label="(1, 14, 14, 384)", style=solid]; +"1497 add_15" -> "1498 quantize_per_tensor_default_31" [label="(1, 14, 14, 384)", style=solid]; +"1497 add_15" -> "1508 add_16" [label="(1, 14, 14, 384)", style=solid]; +"1498 quantize_per_tensor_default_31" -> "1499 dequantize_per_tensor_default_31" [label=None, style=solid]; +"1499 dequantize_per_tensor_default_31" -> "1500 linear_30" [label=None, style=solid]; +"1500 linear_30" -> "1501 gelu_4" [label="(1, 14, 14, 1536)", style=solid]; +"1501 gelu_4" -> "1502 quantize_per_tensor_default_32" [label="(1, 14, 14, 1536)", style=solid]; +"1502 quantize_per_tensor_default_32" -> "1503 dequantize_per_tensor_default_32" [label=None, style=solid]; +"1503 dequantize_per_tensor_default_32" -> "1504 dropout_18" [label=None, style=solid]; +"1504 dropout_18" -> "1505 linear_31" [label="(1, 14, 14, 1536)", style=solid]; +"1505 linear_31" -> "1506 dropout_19" [label="(1, 14, 14, 384)", style=solid]; +"1506 dropout_19" -> "1507 layer_norm_12" [label="(1, 14, 14, 384)", style=solid]; +"1507 layer_norm_12" -> "1508 add_16" [label="(1, 14, 14, 384)", style=solid]; +"1508 add_16" -> "1520 pad_7" [label="(1, 14, 14, 384)", style=solid]; +"1508 add_16" -> "1625 add_19" [label="(1, 14, 14, 384)", style=solid]; +"1509 linear_32" -> "1510 relu__5" [label="(1, 15, 15, 512)", style=solid]; +"1510 relu__5" -> "1511 linear_33" [label="(1, 15, 15, 512)", style=solid]; +"1511 linear_33" -> "1512 view_26" [label="(1, 15, 15, 12)", style=solid]; +"1512 view_26" -> "1513 index_5" [label="(225, 12)", style=solid]; +"1513 index_5" -> "1514 view_27" [label="(4096, 12)", style=solid]; +"1514 view_27" -> "1515 permute_23" [label="(64, 64, 12)", style=solid]; +"1515 permute_23" -> "1516 contiguous_8" [label="(12, 64, 64)", style=solid]; +"1516 contiguous_8" -> "1517 unsqueeze_13" [label="(12, 64, 64)", style=solid]; +"1517 unsqueeze_13" -> "1518 sigmoid_5" [label="(1, 12, 64, 64)", style=solid]; +"1518 sigmoid_5" -> "1519 mul_10" [label="(1, 12, 64, 64)", style=solid]; +"1519 mul_10" -> "1553 add_17" [label="(1, 12, 64, 64)", style=solid]; +"1520 pad_7" -> "1521 roll_4" [label="(1, 16, 16, 384)", style=solid]; +"1521 roll_4" -> "1522 view_28" [label="(1, 16, 16, 384)", style=solid]; +"1522 view_28" -> "1523 permute_24" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1523 permute_24" -> "1524 reshape_22" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1524 reshape_22" -> "1525 quantize_per_tensor_default_33" [label="(4, 64, 384)", style=solid]; +"1524 reshape_22" -> "1554 new_zeros_2" [label="(4, 64, 384)", style=solid]; +"1525 quantize_per_tensor_default_33" -> "1526 dequantize_per_tensor_default_33" [label=None, style=solid]; +"1526 dequantize_per_tensor_default_33" -> "1530 linear_34" [label=None, style=solid]; +"1527 clone_5" -> "1528 slice_82" [label="(1152,)", style=solid]; +"1527 clone_5" -> "1530 linear_34" [label="(1152,)", style=solid]; +"1528 slice_82" -> "1529 zero__5" [label="(384,)", style=solid]; +"1530 linear_34" -> "1531 reshape_23" [label="(4, 64, 1152)", style=solid]; +"1531 reshape_23" -> "1532 permute_25" [label="(4, 64, 3, 12, 32)", style=solid]; +"1532 permute_25" -> "1533 select_15" [label="(3, 4, 12, 64, 32)", style=solid]; +"1532 permute_25" -> "1534 select_16" [label="(3, 4, 12, 64, 32)", style=solid]; +"1532 permute_25" -> "1535 select_17" [label="(3, 4, 12, 64, 32)", style=solid]; +"1533 select_15" -> "1536 linalg_vector_norm_10" [label="(4, 12, 64, 32)", style=solid]; +"1533 select_15" -> "1538 expand_as_10" [label="(4, 12, 64, 32)", style=solid]; +"1533 select_15" -> "1539 div_10" [label="(4, 12, 64, 32)", style=solid]; +"1534 select_16" -> "1542 linalg_vector_norm_11" [label="(4, 12, 64, 32)", style=solid]; +"1534 select_16" -> "1544 expand_as_11" [label="(4, 12, 64, 32)", style=solid]; +"1534 select_16" -> "1545 div_11" [label="(4, 12, 64, 32)", style=solid]; +"1535 select_17" -> "1608 matmul_11" [label="(4, 12, 64, 32)", style=solid]; +"1536 linalg_vector_norm_10" -> "1537 clamp_min_10" [label="(4, 12, 64, 1)", style=solid]; +"1537 clamp_min_10" -> "1538 expand_as_10" [label="(4, 12, 64, 1)", style=solid]; +"1538 expand_as_10" -> "1539 div_10" [label="(4, 12, 64, 32)", style=solid]; +"1539 div_10" -> "1540 quantize_per_tensor_default_34" [label="(4, 12, 64, 32)", style=solid]; +"1540 quantize_per_tensor_default_34" -> "1541 dequantize_per_tensor_default_34" [label=None, style=solid]; +"1541 dequantize_per_tensor_default_34" -> "1549 matmul_10" [label=None, style=solid]; +"1542 linalg_vector_norm_11" -> "1543 clamp_min_11" [label="(4, 12, 64, 1)", style=solid]; +"1543 clamp_min_11" -> "1544 expand_as_11" [label="(4, 12, 64, 1)", style=solid]; +"1544 expand_as_11" -> "1545 div_11" [label="(4, 12, 64, 32)", style=solid]; +"1545 div_11" -> "1546 quantize_per_tensor_default_35" [label="(4, 12, 64, 32)", style=solid]; +"1546 quantize_per_tensor_default_35" -> "1547 dequantize_per_tensor_default_35" [label=None, style=solid]; +"1547 dequantize_per_tensor_default_35" -> "1548 transpose_10" [label=None, style=solid]; +"1548 transpose_10" -> "1549 matmul_10" [label="(4, 12, 32, 64)", style=solid]; +"1549 matmul_10" -> "1552 mul_11" [label="(4, 12, 64, 64)", style=solid]; +"1550 clamp_5" -> "1551 exp_5" [label="(12, 1, 1)", style=solid]; +"1551 exp_5" -> "1552 mul_11" [label="(12, 1, 1)", style=solid]; +"1552 mul_11" -> "1553 add_17" [label="(4, 12, 64, 64)", style=solid]; +"1553 add_17" -> "1601 view_30" [label="(4, 12, 64, 64)", style=solid]; +"1554 new_zeros_2" -> "1556 slice_83" [label="(16, 16)", style=solid]; +"1554 new_zeros_2" -> "1560 slice_85" [label="(16, 16)", style=solid]; +"1554 new_zeros_2" -> "1564 slice_87" [label="(16, 16)", style=solid]; +"1554 new_zeros_2" -> "1568 slice_89" [label="(16, 16)", style=solid]; +"1554 new_zeros_2" -> "1572 slice_91" [label="(16, 16)", style=solid]; +"1554 new_zeros_2" -> "1576 slice_93" [label="(16, 16)", style=solid]; +"1554 new_zeros_2" -> "1580 slice_95" [label="(16, 16)", style=solid]; +"1554 new_zeros_2" -> "1584 slice_97" [label="(16, 16)", style=solid]; +"1554 new_zeros_2" -> "1588 slice_99" [label="(16, 16)", style=solid]; +"1554 new_zeros_2" -> "1591 view_29" [label="(16, 16)", style=solid]; +"1555 lift_fresh_copy_18" -> "1558 fill__18" [label="()", style=solid]; +"1556 slice_83" -> "1557 slice_84" [label="(8, 16)", style=solid]; +"1557 slice_84" -> "1558 fill__18" [label="(8, 8)", style=solid]; +"1559 lift_fresh_copy_19" -> "1562 fill__19" [label="()", style=solid]; +"1560 slice_85" -> "1561 slice_86" [label="(8, 16)", style=solid]; +"1561 slice_86" -> "1562 fill__19" [label="(8, 4)", style=solid]; +"1563 lift_fresh_copy_20" -> "1566 fill__20" [label="()", style=solid]; +"1564 slice_87" -> "1565 slice_88" [label="(8, 16)", style=solid]; +"1565 slice_88" -> "1566 fill__20" [label="(8, 4)", style=solid]; +"1567 lift_fresh_copy_21" -> "1570 fill__21" [label="()", style=solid]; +"1568 slice_89" -> "1569 slice_90" [label="(4, 16)", style=solid]; +"1569 slice_90" -> "1570 fill__21" [label="(4, 8)", style=solid]; +"1571 lift_fresh_copy_22" -> "1574 fill__22" [label="()", style=solid]; +"1572 slice_91" -> "1573 slice_92" [label="(4, 16)", style=solid]; +"1573 slice_92" -> "1574 fill__22" [label="(4, 4)", style=solid]; +"1575 lift_fresh_copy_23" -> "1578 fill__23" [label="()", style=solid]; +"1576 slice_93" -> "1577 slice_94" [label="(4, 16)", style=solid]; +"1577 slice_94" -> "1578 fill__23" [label="(4, 4)", style=solid]; +"1579 lift_fresh_copy_24" -> "1582 fill__24" [label="()", style=solid]; +"1580 slice_95" -> "1581 slice_96" [label="(4, 16)", style=solid]; +"1581 slice_96" -> "1582 fill__24" [label="(4, 8)", style=solid]; +"1583 lift_fresh_copy_25" -> "1586 fill__25" [label="()", style=solid]; +"1584 slice_97" -> "1585 slice_98" [label="(4, 16)", style=solid]; +"1585 slice_98" -> "1586 fill__25" [label="(4, 4)", style=solid]; +"1587 lift_fresh_copy_26" -> "1590 fill__26" [label="()", style=solid]; +"1588 slice_99" -> "1589 slice_100" [label="(4, 16)", style=solid]; +"1589 slice_100" -> "1590 fill__26" [label="(4, 4)", style=solid]; +"1591 view_29" -> "1592 permute_26" [label="(2, 8, 2, 8)", style=solid]; +"1592 permute_26" -> "1593 reshape_24" [label="(2, 2, 8, 8)", style=solid]; +"1593 reshape_24" -> "1594 unsqueeze_14" [label="(4, 64)", style=solid]; +"1593 reshape_24" -> "1595 unsqueeze_15" [label="(4, 64)", style=solid]; +"1594 unsqueeze_14" -> "1596 sub_2" [label="(4, 1, 64)", style=solid]; +"1595 unsqueeze_15" -> "1596 sub_2" [label="(4, 64, 1)", style=solid]; +"1596 sub_2" -> "1597 ne_2" [label="(4, 64, 64)", style=solid]; +"1596 sub_2" -> "1598 masked_fill_4" [label="(4, 64, 64)", style=solid]; +"1596 sub_2" -> "1599 eq_2" [label="(4, 64, 64)", style=solid]; +"1597 ne_2" -> "1598 masked_fill_4" [label="(4, 64, 64)", style=solid]; +"1598 masked_fill_4" -> "1600 masked_fill_5" [label="(4, 64, 64)", style=solid]; +"1599 eq_2" -> "1600 masked_fill_5" [label="(4, 64, 64)", style=solid]; +"1600 masked_fill_5" -> "1602 unsqueeze_16" [label="(4, 64, 64)", style=solid]; +"1601 view_30" -> "1604 add_18" [label="(1, 4, 12, 64, 64)", style=solid]; +"1602 unsqueeze_16" -> "1603 unsqueeze_17" [label="(4, 1, 64, 64)", style=solid]; +"1603 unsqueeze_17" -> "1604 add_18" [label="(1, 4, 1, 64, 64)", style=solid]; +"1604 add_18" -> "1605 view_31" [label="(1, 4, 12, 64, 64)", style=solid]; +"1605 view_31" -> "1606 softmax_5" [label="(4, 12, 64, 64)", style=solid]; +"1606 softmax_5" -> "1607 dropout_20" [label="(4, 12, 64, 64)", style=solid]; +"1607 dropout_20" -> "1608 matmul_11" [label="(4, 12, 64, 64)", style=solid]; +"1608 matmul_11" -> "1609 quantize_per_tensor_default_36" [label="(4, 12, 64, 32)", style=solid]; +"1609 quantize_per_tensor_default_36" -> "1610 dequantize_per_tensor_default_36" [label=None, style=solid]; +"1610 dequantize_per_tensor_default_36" -> "1611 transpose_11" [label=None, style=solid]; +"1611 transpose_11" -> "1612 reshape_25" [label="(4, 64, 12, 32)", style=solid]; +"1612 reshape_25" -> "1613 linear_35" [label="(4, 64, 384)", style=solid]; +"1613 linear_35" -> "1614 dropout_21" [label="(4, 64, 384)", style=solid]; +"1614 dropout_21" -> "1615 view_32" [label="(4, 64, 384)", style=solid]; +"1615 view_32" -> "1616 permute_27" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1616 permute_27" -> "1617 reshape_26" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1617 reshape_26" -> "1618 roll_5" [label="(1, 16, 16, 384)", style=solid]; +"1618 roll_5" -> "1619 slice_101" [label="(1, 16, 16, 384)", style=solid]; +"1619 slice_101" -> "1620 slice_102" [label="(1, 16, 16, 384)", style=solid]; +"1620 slice_102" -> "1621 slice_103" [label="(1, 14, 16, 384)", style=solid]; +"1621 slice_103" -> "1622 slice_104" [label="(1, 14, 14, 384)", style=solid]; +"1622 slice_104" -> "1623 contiguous_9" [label="(1, 14, 14, 384)", style=solid]; +"1623 contiguous_9" -> "1624 layer_norm_13" [label="(1, 14, 14, 384)", style=solid]; +"1624 layer_norm_13" -> "1625 add_19" [label="(1, 14, 14, 384)", style=solid]; +"1625 add_19" -> "1626 quantize_per_tensor_default_37" [label="(1, 14, 14, 384)", style=solid]; +"1625 add_19" -> "1636 add_20" [label="(1, 14, 14, 384)", style=solid]; +"1626 quantize_per_tensor_default_37" -> "1627 dequantize_per_tensor_default_37" [label=None, style=solid]; +"1627 dequantize_per_tensor_default_37" -> "1628 linear_36" [label=None, style=solid]; +"1628 linear_36" -> "1629 gelu_5" [label="(1, 14, 14, 1536)", style=solid]; +"1629 gelu_5" -> "1630 quantize_per_tensor_default_38" [label="(1, 14, 14, 1536)", style=solid]; +"1630 quantize_per_tensor_default_38" -> "1631 dequantize_per_tensor_default_38" [label=None, style=solid]; +"1631 dequantize_per_tensor_default_38" -> "1632 dropout_22" [label=None, style=solid]; +"1632 dropout_22" -> "1633 linear_37" [label="(1, 14, 14, 1536)", style=solid]; +"1633 linear_37" -> "1634 dropout_23" [label="(1, 14, 14, 384)", style=solid]; +"1634 dropout_23" -> "1635 layer_norm_14" [label="(1, 14, 14, 384)", style=solid]; +"1635 layer_norm_14" -> "1636 add_20" [label="(1, 14, 14, 384)", style=solid]; +"1636 add_20" -> "1637 quantize_per_tensor_default_39" [label="(1, 14, 14, 384)", style=solid]; +"1636 add_20" -> "1699 add_22" [label="(1, 14, 14, 384)", style=solid]; +"1637 quantize_per_tensor_default_39" -> "1638 dequantize_per_tensor_default_39" [label=None, style=solid]; +"1638 dequantize_per_tensor_default_39" -> "1650 pad_8" [label=None, style=solid]; +"1639 linear_38" -> "1640 relu__6" [label="(1, 15, 15, 512)", style=solid]; +"1640 relu__6" -> "1641 linear_39" [label="(1, 15, 15, 512)", style=solid]; +"1641 linear_39" -> "1642 view_33" [label="(1, 15, 15, 12)", style=solid]; +"1642 view_33" -> "1643 index_6" [label="(225, 12)", style=solid]; +"1643 index_6" -> "1644 view_34" [label="(4096, 12)", style=solid]; +"1644 view_34" -> "1645 permute_28" [label="(64, 64, 12)", style=solid]; +"1645 permute_28" -> "1646 contiguous_10" [label="(12, 64, 64)", style=solid]; +"1646 contiguous_10" -> "1647 unsqueeze_18" [label="(12, 64, 64)", style=solid]; +"1647 unsqueeze_18" -> "1648 sigmoid_6" [label="(1, 12, 64, 64)", style=solid]; +"1648 sigmoid_6" -> "1649 mul_12" [label="(1, 12, 64, 64)", style=solid]; +"1649 mul_12" -> "1680 add_21" [label="(1, 12, 64, 64)", style=solid]; +"1650 pad_8" -> "1651 view_35" [label="(1, 16, 16, 384)", style=solid]; +"1651 view_35" -> "1652 permute_29" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1652 permute_29" -> "1653 reshape_27" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1653 reshape_27" -> "1657 linear_40" [label="(4, 64, 384)", style=solid]; +"1654 clone_6" -> "1655 slice_105" [label="(1152,)", style=solid]; +"1654 clone_6" -> "1657 linear_40" [label="(1152,)", style=solid]; +"1655 slice_105" -> "1656 zero__6" [label="(384,)", style=solid]; +"1657 linear_40" -> "1658 reshape_28" [label="(4, 64, 1152)", style=solid]; +"1658 reshape_28" -> "1659 permute_30" [label="(4, 64, 3, 12, 32)", style=solid]; +"1659 permute_30" -> "1660 select_18" [label="(3, 4, 12, 64, 32)", style=solid]; +"1659 permute_30" -> "1661 select_19" [label="(3, 4, 12, 64, 32)", style=solid]; +"1659 permute_30" -> "1662 select_20" [label="(3, 4, 12, 64, 32)", style=solid]; +"1660 select_18" -> "1663 linalg_vector_norm_12" [label="(4, 12, 64, 32)", style=solid]; +"1660 select_18" -> "1665 expand_as_12" [label="(4, 12, 64, 32)", style=solid]; +"1660 select_18" -> "1666 div_12" [label="(4, 12, 64, 32)", style=solid]; +"1661 select_19" -> "1669 linalg_vector_norm_13" [label="(4, 12, 64, 32)", style=solid]; +"1661 select_19" -> "1671 expand_as_13" [label="(4, 12, 64, 32)", style=solid]; +"1661 select_19" -> "1672 div_13" [label="(4, 12, 64, 32)", style=solid]; +"1662 select_20" -> "1683 matmul_13" [label="(4, 12, 64, 32)", style=solid]; +"1663 linalg_vector_norm_12" -> "1664 clamp_min_12" [label="(4, 12, 64, 1)", style=solid]; +"1664 clamp_min_12" -> "1665 expand_as_12" [label="(4, 12, 64, 1)", style=solid]; +"1665 expand_as_12" -> "1666 div_12" [label="(4, 12, 64, 32)", style=solid]; +"1666 div_12" -> "1667 quantize_per_tensor_default_40" [label="(4, 12, 64, 32)", style=solid]; +"1667 quantize_per_tensor_default_40" -> "1668 dequantize_per_tensor_default_40" [label=None, style=solid]; +"1668 dequantize_per_tensor_default_40" -> "1676 matmul_12" [label=None, style=solid]; +"1669 linalg_vector_norm_13" -> "1670 clamp_min_13" [label="(4, 12, 64, 1)", style=solid]; +"1670 clamp_min_13" -> "1671 expand_as_13" [label="(4, 12, 64, 1)", style=solid]; +"1671 expand_as_13" -> "1672 div_13" [label="(4, 12, 64, 32)", style=solid]; +"1672 div_13" -> "1673 quantize_per_tensor_default_41" [label="(4, 12, 64, 32)", style=solid]; +"1673 quantize_per_tensor_default_41" -> "1674 dequantize_per_tensor_default_41" [label=None, style=solid]; +"1674 dequantize_per_tensor_default_41" -> "1675 transpose_12" [label=None, style=solid]; +"1675 transpose_12" -> "1676 matmul_12" [label="(4, 12, 32, 64)", style=solid]; +"1676 matmul_12" -> "1679 mul_13" [label="(4, 12, 64, 64)", style=solid]; +"1677 clamp_6" -> "1678 exp_6" [label="(12, 1, 1)", style=solid]; +"1678 exp_6" -> "1679 mul_13" [label="(12, 1, 1)", style=solid]; +"1679 mul_13" -> "1680 add_21" [label="(4, 12, 64, 64)", style=solid]; +"1680 add_21" -> "1681 softmax_6" [label="(4, 12, 64, 64)", style=solid]; +"1681 softmax_6" -> "1682 dropout_24" [label="(4, 12, 64, 64)", style=solid]; +"1682 dropout_24" -> "1683 matmul_13" [label="(4, 12, 64, 64)", style=solid]; +"1683 matmul_13" -> "1684 quantize_per_tensor_default_42" [label="(4, 12, 64, 32)", style=solid]; +"1684 quantize_per_tensor_default_42" -> "1685 dequantize_per_tensor_default_42" [label=None, style=solid]; +"1685 dequantize_per_tensor_default_42" -> "1686 transpose_13" [label=None, style=solid]; +"1686 transpose_13" -> "1687 reshape_29" [label="(4, 64, 12, 32)", style=solid]; +"1687 reshape_29" -> "1688 linear_41" [label="(4, 64, 384)", style=solid]; +"1688 linear_41" -> "1689 dropout_25" [label="(4, 64, 384)", style=solid]; +"1689 dropout_25" -> "1690 view_36" [label="(4, 64, 384)", style=solid]; +"1690 view_36" -> "1691 permute_31" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1691 permute_31" -> "1692 reshape_30" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1692 reshape_30" -> "1693 slice_106" [label="(1, 16, 16, 384)", style=solid]; +"1693 slice_106" -> "1694 slice_107" [label="(1, 16, 16, 384)", style=solid]; +"1694 slice_107" -> "1695 slice_108" [label="(1, 14, 16, 384)", style=solid]; +"1695 slice_108" -> "1696 slice_109" [label="(1, 14, 14, 384)", style=solid]; +"1696 slice_109" -> "1697 contiguous_11" [label="(1, 14, 14, 384)", style=solid]; +"1697 contiguous_11" -> "1698 layer_norm_15" [label="(1, 14, 14, 384)", style=solid]; +"1698 layer_norm_15" -> "1699 add_22" [label="(1, 14, 14, 384)", style=solid]; +"1699 add_22" -> "1700 quantize_per_tensor_default_43" [label="(1, 14, 14, 384)", style=solid]; +"1699 add_22" -> "1710 add_23" [label="(1, 14, 14, 384)", style=solid]; +"1700 quantize_per_tensor_default_43" -> "1701 dequantize_per_tensor_default_43" [label=None, style=solid]; +"1701 dequantize_per_tensor_default_43" -> "1702 linear_42" [label=None, style=solid]; +"1702 linear_42" -> "1703 gelu_6" [label="(1, 14, 14, 1536)", style=solid]; +"1703 gelu_6" -> "1704 quantize_per_tensor_default_44" [label="(1, 14, 14, 1536)", style=solid]; +"1704 quantize_per_tensor_default_44" -> "1705 dequantize_per_tensor_default_44" [label=None, style=solid]; +"1705 dequantize_per_tensor_default_44" -> "1706 dropout_26" [label=None, style=solid]; +"1706 dropout_26" -> "1707 linear_43" [label="(1, 14, 14, 1536)", style=solid]; +"1707 linear_43" -> "1708 dropout_27" [label="(1, 14, 14, 384)", style=solid]; +"1708 dropout_27" -> "1709 layer_norm_16" [label="(1, 14, 14, 384)", style=solid]; +"1709 layer_norm_16" -> "1710 add_23" [label="(1, 14, 14, 384)", style=solid]; +"1710 add_23" -> "1722 pad_9" [label="(1, 14, 14, 384)", style=solid]; +"1710 add_23" -> "1827 add_26" [label="(1, 14, 14, 384)", style=solid]; +"1711 linear_44" -> "1712 relu__7" [label="(1, 15, 15, 512)", style=solid]; +"1712 relu__7" -> "1713 linear_45" [label="(1, 15, 15, 512)", style=solid]; +"1713 linear_45" -> "1714 view_37" [label="(1, 15, 15, 12)", style=solid]; +"1714 view_37" -> "1715 index_7" [label="(225, 12)", style=solid]; +"1715 index_7" -> "1716 view_38" [label="(4096, 12)", style=solid]; +"1716 view_38" -> "1717 permute_32" [label="(64, 64, 12)", style=solid]; +"1717 permute_32" -> "1718 contiguous_12" [label="(12, 64, 64)", style=solid]; +"1718 contiguous_12" -> "1719 unsqueeze_19" [label="(12, 64, 64)", style=solid]; +"1719 unsqueeze_19" -> "1720 sigmoid_7" [label="(1, 12, 64, 64)", style=solid]; +"1720 sigmoid_7" -> "1721 mul_14" [label="(1, 12, 64, 64)", style=solid]; +"1721 mul_14" -> "1755 add_24" [label="(1, 12, 64, 64)", style=solid]; +"1722 pad_9" -> "1723 roll_6" [label="(1, 16, 16, 384)", style=solid]; +"1723 roll_6" -> "1724 view_39" [label="(1, 16, 16, 384)", style=solid]; +"1724 view_39" -> "1725 permute_33" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1725 permute_33" -> "1726 reshape_31" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1726 reshape_31" -> "1727 quantize_per_tensor_default_45" [label="(4, 64, 384)", style=solid]; +"1726 reshape_31" -> "1756 new_zeros_3" [label="(4, 64, 384)", style=solid]; +"1727 quantize_per_tensor_default_45" -> "1728 dequantize_per_tensor_default_45" [label=None, style=solid]; +"1728 dequantize_per_tensor_default_45" -> "1732 linear_46" [label=None, style=solid]; +"1729 clone_7" -> "1730 slice_110" [label="(1152,)", style=solid]; +"1729 clone_7" -> "1732 linear_46" [label="(1152,)", style=solid]; +"1730 slice_110" -> "1731 zero__7" [label="(384,)", style=solid]; +"1732 linear_46" -> "1733 reshape_32" [label="(4, 64, 1152)", style=solid]; +"1733 reshape_32" -> "1734 permute_34" [label="(4, 64, 3, 12, 32)", style=solid]; +"1734 permute_34" -> "1735 select_21" [label="(3, 4, 12, 64, 32)", style=solid]; +"1734 permute_34" -> "1736 select_22" [label="(3, 4, 12, 64, 32)", style=solid]; +"1734 permute_34" -> "1737 select_23" [label="(3, 4, 12, 64, 32)", style=solid]; +"1735 select_21" -> "1738 linalg_vector_norm_14" [label="(4, 12, 64, 32)", style=solid]; +"1735 select_21" -> "1740 expand_as_14" [label="(4, 12, 64, 32)", style=solid]; +"1735 select_21" -> "1741 div_14" [label="(4, 12, 64, 32)", style=solid]; +"1736 select_22" -> "1744 linalg_vector_norm_15" [label="(4, 12, 64, 32)", style=solid]; +"1736 select_22" -> "1746 expand_as_15" [label="(4, 12, 64, 32)", style=solid]; +"1736 select_22" -> "1747 div_15" [label="(4, 12, 64, 32)", style=solid]; +"1737 select_23" -> "1810 matmul_15" [label="(4, 12, 64, 32)", style=solid]; +"1738 linalg_vector_norm_14" -> "1739 clamp_min_14" [label="(4, 12, 64, 1)", style=solid]; +"1739 clamp_min_14" -> "1740 expand_as_14" [label="(4, 12, 64, 1)", style=solid]; +"1740 expand_as_14" -> "1741 div_14" [label="(4, 12, 64, 32)", style=solid]; +"1741 div_14" -> "1742 quantize_per_tensor_default_46" [label="(4, 12, 64, 32)", style=solid]; +"1742 quantize_per_tensor_default_46" -> "1743 dequantize_per_tensor_default_46" [label=None, style=solid]; +"1743 dequantize_per_tensor_default_46" -> "1751 matmul_14" [label=None, style=solid]; +"1744 linalg_vector_norm_15" -> "1745 clamp_min_15" [label="(4, 12, 64, 1)", style=solid]; +"1745 clamp_min_15" -> "1746 expand_as_15" [label="(4, 12, 64, 1)", style=solid]; +"1746 expand_as_15" -> "1747 div_15" [label="(4, 12, 64, 32)", style=solid]; +"1747 div_15" -> "1748 quantize_per_tensor_default_47" [label="(4, 12, 64, 32)", style=solid]; +"1748 quantize_per_tensor_default_47" -> "1749 dequantize_per_tensor_default_47" [label=None, style=solid]; +"1749 dequantize_per_tensor_default_47" -> "1750 transpose_14" [label=None, style=solid]; +"1750 transpose_14" -> "1751 matmul_14" [label="(4, 12, 32, 64)", style=solid]; +"1751 matmul_14" -> "1754 mul_15" [label="(4, 12, 64, 64)", style=solid]; +"1752 clamp_7" -> "1753 exp_7" [label="(12, 1, 1)", style=solid]; +"1753 exp_7" -> "1754 mul_15" [label="(12, 1, 1)", style=solid]; +"1754 mul_15" -> "1755 add_24" [label="(4, 12, 64, 64)", style=solid]; +"1755 add_24" -> "1803 view_41" [label="(4, 12, 64, 64)", style=solid]; +"1756 new_zeros_3" -> "1758 slice_111" [label="(16, 16)", style=solid]; +"1756 new_zeros_3" -> "1762 slice_113" [label="(16, 16)", style=solid]; +"1756 new_zeros_3" -> "1766 slice_115" [label="(16, 16)", style=solid]; +"1756 new_zeros_3" -> "1770 slice_117" [label="(16, 16)", style=solid]; +"1756 new_zeros_3" -> "1774 slice_119" [label="(16, 16)", style=solid]; +"1756 new_zeros_3" -> "1778 slice_121" [label="(16, 16)", style=solid]; +"1756 new_zeros_3" -> "1782 slice_123" [label="(16, 16)", style=solid]; +"1756 new_zeros_3" -> "1786 slice_125" [label="(16, 16)", style=solid]; +"1756 new_zeros_3" -> "1790 slice_127" [label="(16, 16)", style=solid]; +"1756 new_zeros_3" -> "1793 view_40" [label="(16, 16)", style=solid]; +"1757 lift_fresh_copy_27" -> "1760 fill__27" [label="()", style=solid]; +"1758 slice_111" -> "1759 slice_112" [label="(8, 16)", style=solid]; +"1759 slice_112" -> "1760 fill__27" [label="(8, 8)", style=solid]; +"1761 lift_fresh_copy_28" -> "1764 fill__28" [label="()", style=solid]; +"1762 slice_113" -> "1763 slice_114" [label="(8, 16)", style=solid]; +"1763 slice_114" -> "1764 fill__28" [label="(8, 4)", style=solid]; +"1765 lift_fresh_copy_29" -> "1768 fill__29" [label="()", style=solid]; +"1766 slice_115" -> "1767 slice_116" [label="(8, 16)", style=solid]; +"1767 slice_116" -> "1768 fill__29" [label="(8, 4)", style=solid]; +"1769 lift_fresh_copy_30" -> "1772 fill__30" [label="()", style=solid]; +"1770 slice_117" -> "1771 slice_118" [label="(4, 16)", style=solid]; +"1771 slice_118" -> "1772 fill__30" [label="(4, 8)", style=solid]; +"1773 lift_fresh_copy_31" -> "1776 fill__31" [label="()", style=solid]; +"1774 slice_119" -> "1775 slice_120" [label="(4, 16)", style=solid]; +"1775 slice_120" -> "1776 fill__31" [label="(4, 4)", style=solid]; +"1777 lift_fresh_copy_32" -> "1780 fill__32" [label="()", style=solid]; +"1778 slice_121" -> "1779 slice_122" [label="(4, 16)", style=solid]; +"1779 slice_122" -> "1780 fill__32" [label="(4, 4)", style=solid]; +"1781 lift_fresh_copy_33" -> "1784 fill__33" [label="()", style=solid]; +"1782 slice_123" -> "1783 slice_124" [label="(4, 16)", style=solid]; +"1783 slice_124" -> "1784 fill__33" [label="(4, 8)", style=solid]; +"1785 lift_fresh_copy_34" -> "1788 fill__34" [label="()", style=solid]; +"1786 slice_125" -> "1787 slice_126" [label="(4, 16)", style=solid]; +"1787 slice_126" -> "1788 fill__34" [label="(4, 4)", style=solid]; +"1789 lift_fresh_copy_35" -> "1792 fill__35" [label="()", style=solid]; +"1790 slice_127" -> "1791 slice_128" [label="(4, 16)", style=solid]; +"1791 slice_128" -> "1792 fill__35" [label="(4, 4)", style=solid]; +"1793 view_40" -> "1794 permute_35" [label="(2, 8, 2, 8)", style=solid]; +"1794 permute_35" -> "1795 reshape_33" [label="(2, 2, 8, 8)", style=solid]; +"1795 reshape_33" -> "1796 unsqueeze_20" [label="(4, 64)", style=solid]; +"1795 reshape_33" -> "1797 unsqueeze_21" [label="(4, 64)", style=solid]; +"1796 unsqueeze_20" -> "1798 sub_3" [label="(4, 1, 64)", style=solid]; +"1797 unsqueeze_21" -> "1798 sub_3" [label="(4, 64, 1)", style=solid]; +"1798 sub_3" -> "1799 ne_3" [label="(4, 64, 64)", style=solid]; +"1798 sub_3" -> "1800 masked_fill_6" [label="(4, 64, 64)", style=solid]; +"1798 sub_3" -> "1801 eq_3" [label="(4, 64, 64)", style=solid]; +"1799 ne_3" -> "1800 masked_fill_6" [label="(4, 64, 64)", style=solid]; +"1800 masked_fill_6" -> "1802 masked_fill_7" [label="(4, 64, 64)", style=solid]; +"1801 eq_3" -> "1802 masked_fill_7" [label="(4, 64, 64)", style=solid]; +"1802 masked_fill_7" -> "1804 unsqueeze_22" [label="(4, 64, 64)", style=solid]; +"1803 view_41" -> "1806 add_25" [label="(1, 4, 12, 64, 64)", style=solid]; +"1804 unsqueeze_22" -> "1805 unsqueeze_23" [label="(4, 1, 64, 64)", style=solid]; +"1805 unsqueeze_23" -> "1806 add_25" [label="(1, 4, 1, 64, 64)", style=solid]; +"1806 add_25" -> "1807 view_42" [label="(1, 4, 12, 64, 64)", style=solid]; +"1807 view_42" -> "1808 softmax_7" [label="(4, 12, 64, 64)", style=solid]; +"1808 softmax_7" -> "1809 dropout_28" [label="(4, 12, 64, 64)", style=solid]; +"1809 dropout_28" -> "1810 matmul_15" [label="(4, 12, 64, 64)", style=solid]; +"1810 matmul_15" -> "1811 quantize_per_tensor_default_48" [label="(4, 12, 64, 32)", style=solid]; +"1811 quantize_per_tensor_default_48" -> "1812 dequantize_per_tensor_default_48" [label=None, style=solid]; +"1812 dequantize_per_tensor_default_48" -> "1813 transpose_15" [label=None, style=solid]; +"1813 transpose_15" -> "1814 reshape_34" [label="(4, 64, 12, 32)", style=solid]; +"1814 reshape_34" -> "1815 linear_47" [label="(4, 64, 384)", style=solid]; +"1815 linear_47" -> "1816 dropout_29" [label="(4, 64, 384)", style=solid]; +"1816 dropout_29" -> "1817 view_43" [label="(4, 64, 384)", style=solid]; +"1817 view_43" -> "1818 permute_36" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1818 permute_36" -> "1819 reshape_35" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1819 reshape_35" -> "1820 roll_7" [label="(1, 16, 16, 384)", style=solid]; +"1820 roll_7" -> "1821 slice_129" [label="(1, 16, 16, 384)", style=solid]; +"1821 slice_129" -> "1822 slice_130" [label="(1, 16, 16, 384)", style=solid]; +"1822 slice_130" -> "1823 slice_131" [label="(1, 14, 16, 384)", style=solid]; +"1823 slice_131" -> "1824 slice_132" [label="(1, 14, 14, 384)", style=solid]; +"1824 slice_132" -> "1825 contiguous_13" [label="(1, 14, 14, 384)", style=solid]; +"1825 contiguous_13" -> "1826 layer_norm_17" [label="(1, 14, 14, 384)", style=solid]; +"1826 layer_norm_17" -> "1827 add_26" [label="(1, 14, 14, 384)", style=solid]; +"1827 add_26" -> "1828 quantize_per_tensor_default_49" [label="(1, 14, 14, 384)", style=solid]; +"1827 add_26" -> "1838 add_27" [label="(1, 14, 14, 384)", style=solid]; +"1828 quantize_per_tensor_default_49" -> "1829 dequantize_per_tensor_default_49" [label=None, style=solid]; +"1829 dequantize_per_tensor_default_49" -> "1830 linear_48" [label=None, style=solid]; +"1830 linear_48" -> "1831 gelu_7" [label="(1, 14, 14, 1536)", style=solid]; +"1831 gelu_7" -> "1832 quantize_per_tensor_default_50" [label="(1, 14, 14, 1536)", style=solid]; +"1832 quantize_per_tensor_default_50" -> "1833 dequantize_per_tensor_default_50" [label=None, style=solid]; +"1833 dequantize_per_tensor_default_50" -> "1834 dropout_30" [label=None, style=solid]; +"1834 dropout_30" -> "1835 linear_49" [label="(1, 14, 14, 1536)", style=solid]; +"1835 linear_49" -> "1836 dropout_31" [label="(1, 14, 14, 384)", style=solid]; +"1836 dropout_31" -> "1837 layer_norm_18" [label="(1, 14, 14, 384)", style=solid]; +"1837 layer_norm_18" -> "1838 add_27" [label="(1, 14, 14, 384)", style=solid]; +"1838 add_27" -> "1839 quantize_per_tensor_default_51" [label="(1, 14, 14, 384)", style=solid]; +"1838 add_27" -> "1901 add_29" [label="(1, 14, 14, 384)", style=solid]; +"1839 quantize_per_tensor_default_51" -> "1840 dequantize_per_tensor_default_51" [label=None, style=solid]; +"1840 dequantize_per_tensor_default_51" -> "1852 pad_10" [label=None, style=solid]; +"1841 linear_50" -> "1842 relu__8" [label="(1, 15, 15, 512)", style=solid]; +"1842 relu__8" -> "1843 linear_51" [label="(1, 15, 15, 512)", style=solid]; +"1843 linear_51" -> "1844 view_44" [label="(1, 15, 15, 12)", style=solid]; +"1844 view_44" -> "1845 index_8" [label="(225, 12)", style=solid]; +"1845 index_8" -> "1846 view_45" [label="(4096, 12)", style=solid]; +"1846 view_45" -> "1847 permute_37" [label="(64, 64, 12)", style=solid]; +"1847 permute_37" -> "1848 contiguous_14" [label="(12, 64, 64)", style=solid]; +"1848 contiguous_14" -> "1849 unsqueeze_24" [label="(12, 64, 64)", style=solid]; +"1849 unsqueeze_24" -> "1850 sigmoid_8" [label="(1, 12, 64, 64)", style=solid]; +"1850 sigmoid_8" -> "1851 mul_16" [label="(1, 12, 64, 64)", style=solid]; +"1851 mul_16" -> "1882 add_28" [label="(1, 12, 64, 64)", style=solid]; +"1852 pad_10" -> "1853 view_46" [label="(1, 16, 16, 384)", style=solid]; +"1853 view_46" -> "1854 permute_38" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1854 permute_38" -> "1855 reshape_36" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1855 reshape_36" -> "1859 linear_52" [label="(4, 64, 384)", style=solid]; +"1856 clone_8" -> "1857 slice_133" [label="(1152,)", style=solid]; +"1856 clone_8" -> "1859 linear_52" [label="(1152,)", style=solid]; +"1857 slice_133" -> "1858 zero__8" [label="(384,)", style=solid]; +"1859 linear_52" -> "1860 reshape_37" [label="(4, 64, 1152)", style=solid]; +"1860 reshape_37" -> "1861 permute_39" [label="(4, 64, 3, 12, 32)", style=solid]; +"1861 permute_39" -> "1862 select_24" [label="(3, 4, 12, 64, 32)", style=solid]; +"1861 permute_39" -> "1863 select_25" [label="(3, 4, 12, 64, 32)", style=solid]; +"1861 permute_39" -> "1864 select_26" [label="(3, 4, 12, 64, 32)", style=solid]; +"1862 select_24" -> "1865 linalg_vector_norm_16" [label="(4, 12, 64, 32)", style=solid]; +"1862 select_24" -> "1867 expand_as_16" [label="(4, 12, 64, 32)", style=solid]; +"1862 select_24" -> "1868 div_16" [label="(4, 12, 64, 32)", style=solid]; +"1863 select_25" -> "1871 linalg_vector_norm_17" [label="(4, 12, 64, 32)", style=solid]; +"1863 select_25" -> "1873 expand_as_17" [label="(4, 12, 64, 32)", style=solid]; +"1863 select_25" -> "1874 div_17" [label="(4, 12, 64, 32)", style=solid]; +"1864 select_26" -> "1885 matmul_17" [label="(4, 12, 64, 32)", style=solid]; +"1865 linalg_vector_norm_16" -> "1866 clamp_min_16" [label="(4, 12, 64, 1)", style=solid]; +"1866 clamp_min_16" -> "1867 expand_as_16" [label="(4, 12, 64, 1)", style=solid]; +"1867 expand_as_16" -> "1868 div_16" [label="(4, 12, 64, 32)", style=solid]; +"1868 div_16" -> "1869 quantize_per_tensor_default_52" [label="(4, 12, 64, 32)", style=solid]; +"1869 quantize_per_tensor_default_52" -> "1870 dequantize_per_tensor_default_52" [label=None, style=solid]; +"1870 dequantize_per_tensor_default_52" -> "1878 matmul_16" [label=None, style=solid]; +"1871 linalg_vector_norm_17" -> "1872 clamp_min_17" [label="(4, 12, 64, 1)", style=solid]; +"1872 clamp_min_17" -> "1873 expand_as_17" [label="(4, 12, 64, 1)", style=solid]; +"1873 expand_as_17" -> "1874 div_17" [label="(4, 12, 64, 32)", style=solid]; +"1874 div_17" -> "1875 quantize_per_tensor_default_53" [label="(4, 12, 64, 32)", style=solid]; +"1875 quantize_per_tensor_default_53" -> "1876 dequantize_per_tensor_default_53" [label=None, style=solid]; +"1876 dequantize_per_tensor_default_53" -> "1877 transpose_16" [label=None, style=solid]; +"1877 transpose_16" -> "1878 matmul_16" [label="(4, 12, 32, 64)", style=solid]; +"1878 matmul_16" -> "1881 mul_17" [label="(4, 12, 64, 64)", style=solid]; +"1879 clamp_8" -> "1880 exp_8" [label="(12, 1, 1)", style=solid]; +"1880 exp_8" -> "1881 mul_17" [label="(12, 1, 1)", style=solid]; +"1881 mul_17" -> "1882 add_28" [label="(4, 12, 64, 64)", style=solid]; +"1882 add_28" -> "1883 softmax_8" [label="(4, 12, 64, 64)", style=solid]; +"1883 softmax_8" -> "1884 dropout_32" [label="(4, 12, 64, 64)", style=solid]; +"1884 dropout_32" -> "1885 matmul_17" [label="(4, 12, 64, 64)", style=solid]; +"1885 matmul_17" -> "1886 quantize_per_tensor_default_54" [label="(4, 12, 64, 32)", style=solid]; +"1886 quantize_per_tensor_default_54" -> "1887 dequantize_per_tensor_default_54" [label=None, style=solid]; +"1887 dequantize_per_tensor_default_54" -> "1888 transpose_17" [label=None, style=solid]; +"1888 transpose_17" -> "1889 reshape_38" [label="(4, 64, 12, 32)", style=solid]; +"1889 reshape_38" -> "1890 linear_53" [label="(4, 64, 384)", style=solid]; +"1890 linear_53" -> "1891 dropout_33" [label="(4, 64, 384)", style=solid]; +"1891 dropout_33" -> "1892 view_47" [label="(4, 64, 384)", style=solid]; +"1892 view_47" -> "1893 permute_40" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1893 permute_40" -> "1894 reshape_39" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1894 reshape_39" -> "1895 slice_134" [label="(1, 16, 16, 384)", style=solid]; +"1895 slice_134" -> "1896 slice_135" [label="(1, 16, 16, 384)", style=solid]; +"1896 slice_135" -> "1897 slice_136" [label="(1, 14, 16, 384)", style=solid]; +"1897 slice_136" -> "1898 slice_137" [label="(1, 14, 14, 384)", style=solid]; +"1898 slice_137" -> "1899 contiguous_15" [label="(1, 14, 14, 384)", style=solid]; +"1899 contiguous_15" -> "1900 layer_norm_19" [label="(1, 14, 14, 384)", style=solid]; +"1900 layer_norm_19" -> "1901 add_29" [label="(1, 14, 14, 384)", style=solid]; +"1901 add_29" -> "1902 quantize_per_tensor_default_55" [label="(1, 14, 14, 384)", style=solid]; +"1901 add_29" -> "1912 add_30" [label="(1, 14, 14, 384)", style=solid]; +"1902 quantize_per_tensor_default_55" -> "1903 dequantize_per_tensor_default_55" [label=None, style=solid]; +"1903 dequantize_per_tensor_default_55" -> "1904 linear_54" [label=None, style=solid]; +"1904 linear_54" -> "1905 gelu_8" [label="(1, 14, 14, 1536)", style=solid]; +"1905 gelu_8" -> "1906 quantize_per_tensor_default_56" [label="(1, 14, 14, 1536)", style=solid]; +"1906 quantize_per_tensor_default_56" -> "1907 dequantize_per_tensor_default_56" [label=None, style=solid]; +"1907 dequantize_per_tensor_default_56" -> "1908 dropout_34" [label=None, style=solid]; +"1908 dropout_34" -> "1909 linear_55" [label="(1, 14, 14, 1536)", style=solid]; +"1909 linear_55" -> "1910 dropout_35" [label="(1, 14, 14, 384)", style=solid]; +"1910 dropout_35" -> "1911 layer_norm_20" [label="(1, 14, 14, 384)", style=solid]; +"1911 layer_norm_20" -> "1912 add_30" [label="(1, 14, 14, 384)", style=solid]; +"1912 add_30" -> "1924 pad_11" [label="(1, 14, 14, 384)", style=solid]; +"1912 add_30" -> "2029 add_33" [label="(1, 14, 14, 384)", style=solid]; +"1913 linear_56" -> "1914 relu__9" [label="(1, 15, 15, 512)", style=solid]; +"1914 relu__9" -> "1915 linear_57" [label="(1, 15, 15, 512)", style=solid]; +"1915 linear_57" -> "1916 view_48" [label="(1, 15, 15, 12)", style=solid]; +"1916 view_48" -> "1917 index_9" [label="(225, 12)", style=solid]; +"1917 index_9" -> "1918 view_49" [label="(4096, 12)", style=solid]; +"1918 view_49" -> "1919 permute_41" [label="(64, 64, 12)", style=solid]; +"1919 permute_41" -> "1920 contiguous_16" [label="(12, 64, 64)", style=solid]; +"1920 contiguous_16" -> "1921 unsqueeze_25" [label="(12, 64, 64)", style=solid]; +"1921 unsqueeze_25" -> "1922 sigmoid_9" [label="(1, 12, 64, 64)", style=solid]; +"1922 sigmoid_9" -> "1923 mul_18" [label="(1, 12, 64, 64)", style=solid]; +"1923 mul_18" -> "1957 add_31" [label="(1, 12, 64, 64)", style=solid]; +"1924 pad_11" -> "1925 roll_8" [label="(1, 16, 16, 384)", style=solid]; +"1925 roll_8" -> "1926 view_50" [label="(1, 16, 16, 384)", style=solid]; +"1926 view_50" -> "1927 permute_42" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1927 permute_42" -> "1928 reshape_40" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1928 reshape_40" -> "1929 quantize_per_tensor_default_57" [label="(4, 64, 384)", style=solid]; +"1928 reshape_40" -> "1958 new_zeros_4" [label="(4, 64, 384)", style=solid]; +"1929 quantize_per_tensor_default_57" -> "1930 dequantize_per_tensor_default_57" [label=None, style=solid]; +"1930 dequantize_per_tensor_default_57" -> "1934 linear_58" [label=None, style=solid]; +"1931 clone_9" -> "1932 slice_138" [label="(1152,)", style=solid]; +"1931 clone_9" -> "1934 linear_58" [label="(1152,)", style=solid]; +"1932 slice_138" -> "1933 zero__9" [label="(384,)", style=solid]; +"1934 linear_58" -> "1935 reshape_41" [label="(4, 64, 1152)", style=solid]; +"1935 reshape_41" -> "1936 permute_43" [label="(4, 64, 3, 12, 32)", style=solid]; +"1936 permute_43" -> "1937 select_27" [label="(3, 4, 12, 64, 32)", style=solid]; +"1936 permute_43" -> "1938 select_28" [label="(3, 4, 12, 64, 32)", style=solid]; +"1936 permute_43" -> "1939 select_29" [label="(3, 4, 12, 64, 32)", style=solid]; +"1937 select_27" -> "1940 linalg_vector_norm_18" [label="(4, 12, 64, 32)", style=solid]; +"1937 select_27" -> "1942 expand_as_18" [label="(4, 12, 64, 32)", style=solid]; +"1937 select_27" -> "1943 div_18" [label="(4, 12, 64, 32)", style=solid]; +"1938 select_28" -> "1946 linalg_vector_norm_19" [label="(4, 12, 64, 32)", style=solid]; +"1938 select_28" -> "1948 expand_as_19" [label="(4, 12, 64, 32)", style=solid]; +"1938 select_28" -> "1949 div_19" [label="(4, 12, 64, 32)", style=solid]; +"1939 select_29" -> "2012 matmul_19" [label="(4, 12, 64, 32)", style=solid]; +"1940 linalg_vector_norm_18" -> "1941 clamp_min_18" [label="(4, 12, 64, 1)", style=solid]; +"1941 clamp_min_18" -> "1942 expand_as_18" [label="(4, 12, 64, 1)", style=solid]; +"1942 expand_as_18" -> "1943 div_18" [label="(4, 12, 64, 32)", style=solid]; +"1943 div_18" -> "1944 quantize_per_tensor_default_58" [label="(4, 12, 64, 32)", style=solid]; +"1944 quantize_per_tensor_default_58" -> "1945 dequantize_per_tensor_default_58" [label=None, style=solid]; +"1945 dequantize_per_tensor_default_58" -> "1953 matmul_18" [label=None, style=solid]; +"1946 linalg_vector_norm_19" -> "1947 clamp_min_19" [label="(4, 12, 64, 1)", style=solid]; +"1947 clamp_min_19" -> "1948 expand_as_19" [label="(4, 12, 64, 1)", style=solid]; +"1948 expand_as_19" -> "1949 div_19" [label="(4, 12, 64, 32)", style=solid]; +"1949 div_19" -> "1950 quantize_per_tensor_default_59" [label="(4, 12, 64, 32)", style=solid]; +"1950 quantize_per_tensor_default_59" -> "1951 dequantize_per_tensor_default_59" [label=None, style=solid]; +"1951 dequantize_per_tensor_default_59" -> "1952 transpose_18" [label=None, style=solid]; +"1952 transpose_18" -> "1953 matmul_18" [label="(4, 12, 32, 64)", style=solid]; +"1953 matmul_18" -> "1956 mul_19" [label="(4, 12, 64, 64)", style=solid]; +"1954 clamp_9" -> "1955 exp_9" [label="(12, 1, 1)", style=solid]; +"1955 exp_9" -> "1956 mul_19" [label="(12, 1, 1)", style=solid]; +"1956 mul_19" -> "1957 add_31" [label="(4, 12, 64, 64)", style=solid]; +"1957 add_31" -> "2005 view_52" [label="(4, 12, 64, 64)", style=solid]; +"1958 new_zeros_4" -> "1960 slice_139" [label="(16, 16)", style=solid]; +"1958 new_zeros_4" -> "1964 slice_141" [label="(16, 16)", style=solid]; +"1958 new_zeros_4" -> "1968 slice_143" [label="(16, 16)", style=solid]; +"1958 new_zeros_4" -> "1972 slice_145" [label="(16, 16)", style=solid]; +"1958 new_zeros_4" -> "1976 slice_147" [label="(16, 16)", style=solid]; +"1958 new_zeros_4" -> "1980 slice_149" [label="(16, 16)", style=solid]; +"1958 new_zeros_4" -> "1984 slice_151" [label="(16, 16)", style=solid]; +"1958 new_zeros_4" -> "1988 slice_153" [label="(16, 16)", style=solid]; +"1958 new_zeros_4" -> "1992 slice_155" [label="(16, 16)", style=solid]; +"1958 new_zeros_4" -> "1995 view_51" [label="(16, 16)", style=solid]; +"1959 lift_fresh_copy_36" -> "1962 fill__36" [label="()", style=solid]; +"1960 slice_139" -> "1961 slice_140" [label="(8, 16)", style=solid]; +"1961 slice_140" -> "1962 fill__36" [label="(8, 8)", style=solid]; +"1963 lift_fresh_copy_37" -> "1966 fill__37" [label="()", style=solid]; +"1964 slice_141" -> "1965 slice_142" [label="(8, 16)", style=solid]; +"1965 slice_142" -> "1966 fill__37" [label="(8, 4)", style=solid]; +"1967 lift_fresh_copy_38" -> "1970 fill__38" [label="()", style=solid]; +"1968 slice_143" -> "1969 slice_144" [label="(8, 16)", style=solid]; +"1969 slice_144" -> "1970 fill__38" [label="(8, 4)", style=solid]; +"1971 lift_fresh_copy_39" -> "1974 fill__39" [label="()", style=solid]; +"1972 slice_145" -> "1973 slice_146" [label="(4, 16)", style=solid]; +"1973 slice_146" -> "1974 fill__39" [label="(4, 8)", style=solid]; +"1975 lift_fresh_copy_40" -> "1978 fill__40" [label="()", style=solid]; +"1976 slice_147" -> "1977 slice_148" [label="(4, 16)", style=solid]; +"1977 slice_148" -> "1978 fill__40" [label="(4, 4)", style=solid]; +"1979 lift_fresh_copy_41" -> "1982 fill__41" [label="()", style=solid]; +"1980 slice_149" -> "1981 slice_150" [label="(4, 16)", style=solid]; +"1981 slice_150" -> "1982 fill__41" [label="(4, 4)", style=solid]; +"1983 lift_fresh_copy_42" -> "1986 fill__42" [label="()", style=solid]; +"1984 slice_151" -> "1985 slice_152" [label="(4, 16)", style=solid]; +"1985 slice_152" -> "1986 fill__42" [label="(4, 8)", style=solid]; +"1987 lift_fresh_copy_43" -> "1990 fill__43" [label="()", style=solid]; +"1988 slice_153" -> "1989 slice_154" [label="(4, 16)", style=solid]; +"1989 slice_154" -> "1990 fill__43" [label="(4, 4)", style=solid]; +"1991 lift_fresh_copy_44" -> "1994 fill__44" [label="()", style=solid]; +"1992 slice_155" -> "1993 slice_156" [label="(4, 16)", style=solid]; +"1993 slice_156" -> "1994 fill__44" [label="(4, 4)", style=solid]; +"1995 view_51" -> "1996 permute_44" [label="(2, 8, 2, 8)", style=solid]; +"1996 permute_44" -> "1997 reshape_42" [label="(2, 2, 8, 8)", style=solid]; +"1997 reshape_42" -> "1998 unsqueeze_26" [label="(4, 64)", style=solid]; +"1997 reshape_42" -> "1999 unsqueeze_27" [label="(4, 64)", style=solid]; +"1998 unsqueeze_26" -> "2000 sub_4" [label="(4, 1, 64)", style=solid]; +"1999 unsqueeze_27" -> "2000 sub_4" [label="(4, 64, 1)", style=solid]; +"2000 sub_4" -> "2001 ne_4" [label="(4, 64, 64)", style=solid]; +"2000 sub_4" -> "2002 masked_fill_8" [label="(4, 64, 64)", style=solid]; +"2000 sub_4" -> "2003 eq_4" [label="(4, 64, 64)", style=solid]; +"2001 ne_4" -> "2002 masked_fill_8" [label="(4, 64, 64)", style=solid]; +"2002 masked_fill_8" -> "2004 masked_fill_9" [label="(4, 64, 64)", style=solid]; +"2003 eq_4" -> "2004 masked_fill_9" [label="(4, 64, 64)", style=solid]; +"2004 masked_fill_9" -> "2006 unsqueeze_28" [label="(4, 64, 64)", style=solid]; +"2005 view_52" -> "2008 add_32" [label="(1, 4, 12, 64, 64)", style=solid]; +"2006 unsqueeze_28" -> "2007 unsqueeze_29" [label="(4, 1, 64, 64)", style=solid]; +"2007 unsqueeze_29" -> "2008 add_32" [label="(1, 4, 1, 64, 64)", style=solid]; +"2008 add_32" -> "2009 view_53" [label="(1, 4, 12, 64, 64)", style=solid]; +"2009 view_53" -> "2010 softmax_9" [label="(4, 12, 64, 64)", style=solid]; +"2010 softmax_9" -> "2011 dropout_36" [label="(4, 12, 64, 64)", style=solid]; +"2011 dropout_36" -> "2012 matmul_19" [label="(4, 12, 64, 64)", style=solid]; +"2012 matmul_19" -> "2013 quantize_per_tensor_default_60" [label="(4, 12, 64, 32)", style=solid]; +"2013 quantize_per_tensor_default_60" -> "2014 dequantize_per_tensor_default_60" [label=None, style=solid]; +"2014 dequantize_per_tensor_default_60" -> "2015 transpose_19" [label=None, style=solid]; +"2015 transpose_19" -> "2016 reshape_43" [label="(4, 64, 12, 32)", style=solid]; +"2016 reshape_43" -> "2017 linear_59" [label="(4, 64, 384)", style=solid]; +"2017 linear_59" -> "2018 dropout_37" [label="(4, 64, 384)", style=solid]; +"2018 dropout_37" -> "2019 view_54" [label="(4, 64, 384)", style=solid]; +"2019 view_54" -> "2020 permute_45" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2020 permute_45" -> "2021 reshape_44" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2021 reshape_44" -> "2022 roll_9" [label="(1, 16, 16, 384)", style=solid]; +"2022 roll_9" -> "2023 slice_157" [label="(1, 16, 16, 384)", style=solid]; +"2023 slice_157" -> "2024 slice_158" [label="(1, 16, 16, 384)", style=solid]; +"2024 slice_158" -> "2025 slice_159" [label="(1, 14, 16, 384)", style=solid]; +"2025 slice_159" -> "2026 slice_160" [label="(1, 14, 14, 384)", style=solid]; +"2026 slice_160" -> "2027 contiguous_17" [label="(1, 14, 14, 384)", style=solid]; +"2027 contiguous_17" -> "2028 layer_norm_21" [label="(1, 14, 14, 384)", style=solid]; +"2028 layer_norm_21" -> "2029 add_33" [label="(1, 14, 14, 384)", style=solid]; +"2029 add_33" -> "2030 quantize_per_tensor_default_61" [label="(1, 14, 14, 384)", style=solid]; +"2029 add_33" -> "2040 add_34" [label="(1, 14, 14, 384)", style=solid]; +"2030 quantize_per_tensor_default_61" -> "2031 dequantize_per_tensor_default_61" [label=None, style=solid]; +"2031 dequantize_per_tensor_default_61" -> "2032 linear_60" [label=None, style=solid]; +"2032 linear_60" -> "2033 gelu_9" [label="(1, 14, 14, 1536)", style=solid]; +"2033 gelu_9" -> "2034 quantize_per_tensor_default_62" [label="(1, 14, 14, 1536)", style=solid]; +"2034 quantize_per_tensor_default_62" -> "2035 dequantize_per_tensor_default_62" [label=None, style=solid]; +"2035 dequantize_per_tensor_default_62" -> "2036 dropout_38" [label=None, style=solid]; +"2036 dropout_38" -> "2037 linear_61" [label="(1, 14, 14, 1536)", style=solid]; +"2037 linear_61" -> "2038 dropout_39" [label="(1, 14, 14, 384)", style=solid]; +"2038 dropout_39" -> "2039 layer_norm_22" [label="(1, 14, 14, 384)", style=solid]; +"2039 layer_norm_22" -> "2040 add_34" [label="(1, 14, 14, 384)", style=solid]; +"2040 add_34" -> "2041 quantize_per_tensor_default_63" [label="(1, 14, 14, 384)", style=solid]; +"2040 add_34" -> "2103 add_36" [label="(1, 14, 14, 384)", style=solid]; +"2041 quantize_per_tensor_default_63" -> "2042 dequantize_per_tensor_default_63" [label=None, style=solid]; +"2042 dequantize_per_tensor_default_63" -> "2054 pad_12" [label=None, style=solid]; +"2043 linear_62" -> "2044 relu__10" [label="(1, 15, 15, 512)", style=solid]; +"2044 relu__10" -> "2045 linear_63" [label="(1, 15, 15, 512)", style=solid]; +"2045 linear_63" -> "2046 view_55" [label="(1, 15, 15, 12)", style=solid]; +"2046 view_55" -> "2047 index_10" [label="(225, 12)", style=solid]; +"2047 index_10" -> "2048 view_56" [label="(4096, 12)", style=solid]; +"2048 view_56" -> "2049 permute_46" [label="(64, 64, 12)", style=solid]; +"2049 permute_46" -> "2050 contiguous_18" [label="(12, 64, 64)", style=solid]; +"2050 contiguous_18" -> "2051 unsqueeze_30" [label="(12, 64, 64)", style=solid]; +"2051 unsqueeze_30" -> "2052 sigmoid_10" [label="(1, 12, 64, 64)", style=solid]; +"2052 sigmoid_10" -> "2053 mul_20" [label="(1, 12, 64, 64)", style=solid]; +"2053 mul_20" -> "2084 add_35" [label="(1, 12, 64, 64)", style=solid]; +"2054 pad_12" -> "2055 view_57" [label="(1, 16, 16, 384)", style=solid]; +"2055 view_57" -> "2056 permute_47" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2056 permute_47" -> "2057 reshape_45" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2057 reshape_45" -> "2061 linear_64" [label="(4, 64, 384)", style=solid]; +"2058 clone_10" -> "2059 slice_161" [label="(1152,)", style=solid]; +"2058 clone_10" -> "2061 linear_64" [label="(1152,)", style=solid]; +"2059 slice_161" -> "2060 zero__10" [label="(384,)", style=solid]; +"2061 linear_64" -> "2062 reshape_46" [label="(4, 64, 1152)", style=solid]; +"2062 reshape_46" -> "2063 permute_48" [label="(4, 64, 3, 12, 32)", style=solid]; +"2063 permute_48" -> "2064 select_30" [label="(3, 4, 12, 64, 32)", style=solid]; +"2063 permute_48" -> "2065 select_31" [label="(3, 4, 12, 64, 32)", style=solid]; +"2063 permute_48" -> "2066 select_32" [label="(3, 4, 12, 64, 32)", style=solid]; +"2064 select_30" -> "2067 linalg_vector_norm_20" [label="(4, 12, 64, 32)", style=solid]; +"2064 select_30" -> "2069 expand_as_20" [label="(4, 12, 64, 32)", style=solid]; +"2064 select_30" -> "2070 div_20" [label="(4, 12, 64, 32)", style=solid]; +"2065 select_31" -> "2073 linalg_vector_norm_21" [label="(4, 12, 64, 32)", style=solid]; +"2065 select_31" -> "2075 expand_as_21" [label="(4, 12, 64, 32)", style=solid]; +"2065 select_31" -> "2076 div_21" [label="(4, 12, 64, 32)", style=solid]; +"2066 select_32" -> "2087 matmul_21" [label="(4, 12, 64, 32)", style=solid]; +"2067 linalg_vector_norm_20" -> "2068 clamp_min_20" [label="(4, 12, 64, 1)", style=solid]; +"2068 clamp_min_20" -> "2069 expand_as_20" [label="(4, 12, 64, 1)", style=solid]; +"2069 expand_as_20" -> "2070 div_20" [label="(4, 12, 64, 32)", style=solid]; +"2070 div_20" -> "2071 quantize_per_tensor_default_64" [label="(4, 12, 64, 32)", style=solid]; +"2071 quantize_per_tensor_default_64" -> "2072 dequantize_per_tensor_default_64" [label=None, style=solid]; +"2072 dequantize_per_tensor_default_64" -> "2080 matmul_20" [label=None, style=solid]; +"2073 linalg_vector_norm_21" -> "2074 clamp_min_21" [label="(4, 12, 64, 1)", style=solid]; +"2074 clamp_min_21" -> "2075 expand_as_21" [label="(4, 12, 64, 1)", style=solid]; +"2075 expand_as_21" -> "2076 div_21" [label="(4, 12, 64, 32)", style=solid]; +"2076 div_21" -> "2077 quantize_per_tensor_default_65" [label="(4, 12, 64, 32)", style=solid]; +"2077 quantize_per_tensor_default_65" -> "2078 dequantize_per_tensor_default_65" [label=None, style=solid]; +"2078 dequantize_per_tensor_default_65" -> "2079 transpose_20" [label=None, style=solid]; +"2079 transpose_20" -> "2080 matmul_20" [label="(4, 12, 32, 64)", style=solid]; +"2080 matmul_20" -> "2083 mul_21" [label="(4, 12, 64, 64)", style=solid]; +"2081 clamp_10" -> "2082 exp_10" [label="(12, 1, 1)", style=solid]; +"2082 exp_10" -> "2083 mul_21" [label="(12, 1, 1)", style=solid]; +"2083 mul_21" -> "2084 add_35" [label="(4, 12, 64, 64)", style=solid]; +"2084 add_35" -> "2085 softmax_10" [label="(4, 12, 64, 64)", style=solid]; +"2085 softmax_10" -> "2086 dropout_40" [label="(4, 12, 64, 64)", style=solid]; +"2086 dropout_40" -> "2087 matmul_21" [label="(4, 12, 64, 64)", style=solid]; +"2087 matmul_21" -> "2088 quantize_per_tensor_default_66" [label="(4, 12, 64, 32)", style=solid]; +"2088 quantize_per_tensor_default_66" -> "2089 dequantize_per_tensor_default_66" [label=None, style=solid]; +"2089 dequantize_per_tensor_default_66" -> "2090 transpose_21" [label=None, style=solid]; +"2090 transpose_21" -> "2091 reshape_47" [label="(4, 64, 12, 32)", style=solid]; +"2091 reshape_47" -> "2092 linear_65" [label="(4, 64, 384)", style=solid]; +"2092 linear_65" -> "2093 dropout_41" [label="(4, 64, 384)", style=solid]; +"2093 dropout_41" -> "2094 view_58" [label="(4, 64, 384)", style=solid]; +"2094 view_58" -> "2095 permute_49" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2095 permute_49" -> "2096 reshape_48" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2096 reshape_48" -> "2097 slice_162" [label="(1, 16, 16, 384)", style=solid]; +"2097 slice_162" -> "2098 slice_163" [label="(1, 16, 16, 384)", style=solid]; +"2098 slice_163" -> "2099 slice_164" [label="(1, 14, 16, 384)", style=solid]; +"2099 slice_164" -> "2100 slice_165" [label="(1, 14, 14, 384)", style=solid]; +"2100 slice_165" -> "2101 contiguous_19" [label="(1, 14, 14, 384)", style=solid]; +"2101 contiguous_19" -> "2102 layer_norm_23" [label="(1, 14, 14, 384)", style=solid]; +"2102 layer_norm_23" -> "2103 add_36" [label="(1, 14, 14, 384)", style=solid]; +"2103 add_36" -> "2104 quantize_per_tensor_default_67" [label="(1, 14, 14, 384)", style=solid]; +"2103 add_36" -> "2114 add_37" [label="(1, 14, 14, 384)", style=solid]; +"2104 quantize_per_tensor_default_67" -> "2105 dequantize_per_tensor_default_67" [label=None, style=solid]; +"2105 dequantize_per_tensor_default_67" -> "2106 linear_66" [label=None, style=solid]; +"2106 linear_66" -> "2107 gelu_10" [label="(1, 14, 14, 1536)", style=solid]; +"2107 gelu_10" -> "2108 quantize_per_tensor_default_68" [label="(1, 14, 14, 1536)", style=solid]; +"2108 quantize_per_tensor_default_68" -> "2109 dequantize_per_tensor_default_68" [label=None, style=solid]; +"2109 dequantize_per_tensor_default_68" -> "2110 dropout_42" [label=None, style=solid]; +"2110 dropout_42" -> "2111 linear_67" [label="(1, 14, 14, 1536)", style=solid]; +"2111 linear_67" -> "2112 dropout_43" [label="(1, 14, 14, 384)", style=solid]; +"2112 dropout_43" -> "2113 layer_norm_24" [label="(1, 14, 14, 384)", style=solid]; +"2113 layer_norm_24" -> "2114 add_37" [label="(1, 14, 14, 384)", style=solid]; +"2114 add_37" -> "2126 pad_13" [label="(1, 14, 14, 384)", style=solid]; +"2114 add_37" -> "2231 add_40" [label="(1, 14, 14, 384)", style=solid]; +"2115 linear_68" -> "2116 relu__11" [label="(1, 15, 15, 512)", style=solid]; +"2116 relu__11" -> "2117 linear_69" [label="(1, 15, 15, 512)", style=solid]; +"2117 linear_69" -> "2118 view_59" [label="(1, 15, 15, 12)", style=solid]; +"2118 view_59" -> "2119 index_11" [label="(225, 12)", style=solid]; +"2119 index_11" -> "2120 view_60" [label="(4096, 12)", style=solid]; +"2120 view_60" -> "2121 permute_50" [label="(64, 64, 12)", style=solid]; +"2121 permute_50" -> "2122 contiguous_20" [label="(12, 64, 64)", style=solid]; +"2122 contiguous_20" -> "2123 unsqueeze_31" [label="(12, 64, 64)", style=solid]; +"2123 unsqueeze_31" -> "2124 sigmoid_11" [label="(1, 12, 64, 64)", style=solid]; +"2124 sigmoid_11" -> "2125 mul_22" [label="(1, 12, 64, 64)", style=solid]; +"2125 mul_22" -> "2159 add_38" [label="(1, 12, 64, 64)", style=solid]; +"2126 pad_13" -> "2127 roll_10" [label="(1, 16, 16, 384)", style=solid]; +"2127 roll_10" -> "2128 view_61" [label="(1, 16, 16, 384)", style=solid]; +"2128 view_61" -> "2129 permute_51" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2129 permute_51" -> "2130 reshape_49" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2130 reshape_49" -> "2131 quantize_per_tensor_default_69" [label="(4, 64, 384)", style=solid]; +"2130 reshape_49" -> "2160 new_zeros_5" [label="(4, 64, 384)", style=solid]; +"2131 quantize_per_tensor_default_69" -> "2132 dequantize_per_tensor_default_69" [label=None, style=solid]; +"2132 dequantize_per_tensor_default_69" -> "2136 linear_70" [label=None, style=solid]; +"2133 clone_11" -> "2134 slice_166" [label="(1152,)", style=solid]; +"2133 clone_11" -> "2136 linear_70" [label="(1152,)", style=solid]; +"2134 slice_166" -> "2135 zero__11" [label="(384,)", style=solid]; +"2136 linear_70" -> "2137 reshape_50" [label="(4, 64, 1152)", style=solid]; +"2137 reshape_50" -> "2138 permute_52" [label="(4, 64, 3, 12, 32)", style=solid]; +"2138 permute_52" -> "2139 select_33" [label="(3, 4, 12, 64, 32)", style=solid]; +"2138 permute_52" -> "2140 select_34" [label="(3, 4, 12, 64, 32)", style=solid]; +"2138 permute_52" -> "2141 select_35" [label="(3, 4, 12, 64, 32)", style=solid]; +"2139 select_33" -> "2142 linalg_vector_norm_22" [label="(4, 12, 64, 32)", style=solid]; +"2139 select_33" -> "2144 expand_as_22" [label="(4, 12, 64, 32)", style=solid]; +"2139 select_33" -> "2145 div_22" [label="(4, 12, 64, 32)", style=solid]; +"2140 select_34" -> "2148 linalg_vector_norm_23" [label="(4, 12, 64, 32)", style=solid]; +"2140 select_34" -> "2150 expand_as_23" [label="(4, 12, 64, 32)", style=solid]; +"2140 select_34" -> "2151 div_23" [label="(4, 12, 64, 32)", style=solid]; +"2141 select_35" -> "2214 matmul_23" [label="(4, 12, 64, 32)", style=solid]; +"2142 linalg_vector_norm_22" -> "2143 clamp_min_22" [label="(4, 12, 64, 1)", style=solid]; +"2143 clamp_min_22" -> "2144 expand_as_22" [label="(4, 12, 64, 1)", style=solid]; +"2144 expand_as_22" -> "2145 div_22" [label="(4, 12, 64, 32)", style=solid]; +"2145 div_22" -> "2146 quantize_per_tensor_default_70" [label="(4, 12, 64, 32)", style=solid]; +"2146 quantize_per_tensor_default_70" -> "2147 dequantize_per_tensor_default_70" [label=None, style=solid]; +"2147 dequantize_per_tensor_default_70" -> "2155 matmul_22" [label=None, style=solid]; +"2148 linalg_vector_norm_23" -> "2149 clamp_min_23" [label="(4, 12, 64, 1)", style=solid]; +"2149 clamp_min_23" -> "2150 expand_as_23" [label="(4, 12, 64, 1)", style=solid]; +"2150 expand_as_23" -> "2151 div_23" [label="(4, 12, 64, 32)", style=solid]; +"2151 div_23" -> "2152 quantize_per_tensor_default_71" [label="(4, 12, 64, 32)", style=solid]; +"2152 quantize_per_tensor_default_71" -> "2153 dequantize_per_tensor_default_71" [label=None, style=solid]; +"2153 dequantize_per_tensor_default_71" -> "2154 transpose_22" [label=None, style=solid]; +"2154 transpose_22" -> "2155 matmul_22" [label="(4, 12, 32, 64)", style=solid]; +"2155 matmul_22" -> "2158 mul_23" [label="(4, 12, 64, 64)", style=solid]; +"2156 clamp_11" -> "2157 exp_11" [label="(12, 1, 1)", style=solid]; +"2157 exp_11" -> "2158 mul_23" [label="(12, 1, 1)", style=solid]; +"2158 mul_23" -> "2159 add_38" [label="(4, 12, 64, 64)", style=solid]; +"2159 add_38" -> "2207 view_63" [label="(4, 12, 64, 64)", style=solid]; +"2160 new_zeros_5" -> "2162 slice_167" [label="(16, 16)", style=solid]; +"2160 new_zeros_5" -> "2166 slice_169" [label="(16, 16)", style=solid]; +"2160 new_zeros_5" -> "2170 slice_171" [label="(16, 16)", style=solid]; +"2160 new_zeros_5" -> "2174 slice_173" [label="(16, 16)", style=solid]; +"2160 new_zeros_5" -> "2178 slice_175" [label="(16, 16)", style=solid]; +"2160 new_zeros_5" -> "2182 slice_177" [label="(16, 16)", style=solid]; +"2160 new_zeros_5" -> "2186 slice_179" [label="(16, 16)", style=solid]; +"2160 new_zeros_5" -> "2190 slice_181" [label="(16, 16)", style=solid]; +"2160 new_zeros_5" -> "2194 slice_183" [label="(16, 16)", style=solid]; +"2160 new_zeros_5" -> "2197 view_62" [label="(16, 16)", style=solid]; +"2161 lift_fresh_copy_45" -> "2164 fill__45" [label="()", style=solid]; +"2162 slice_167" -> "2163 slice_168" [label="(8, 16)", style=solid]; +"2163 slice_168" -> "2164 fill__45" [label="(8, 8)", style=solid]; +"2165 lift_fresh_copy_46" -> "2168 fill__46" [label="()", style=solid]; +"2166 slice_169" -> "2167 slice_170" [label="(8, 16)", style=solid]; +"2167 slice_170" -> "2168 fill__46" [label="(8, 4)", style=solid]; +"2169 lift_fresh_copy_47" -> "2172 fill__47" [label="()", style=solid]; +"2170 slice_171" -> "2171 slice_172" [label="(8, 16)", style=solid]; +"2171 slice_172" -> "2172 fill__47" [label="(8, 4)", style=solid]; +"2173 lift_fresh_copy_48" -> "2176 fill__48" [label="()", style=solid]; +"2174 slice_173" -> "2175 slice_174" [label="(4, 16)", style=solid]; +"2175 slice_174" -> "2176 fill__48" [label="(4, 8)", style=solid]; +"2177 lift_fresh_copy_49" -> "2180 fill__49" [label="()", style=solid]; +"2178 slice_175" -> "2179 slice_176" [label="(4, 16)", style=solid]; +"2179 slice_176" -> "2180 fill__49" [label="(4, 4)", style=solid]; +"2181 lift_fresh_copy_50" -> "2184 fill__50" [label="()", style=solid]; +"2182 slice_177" -> "2183 slice_178" [label="(4, 16)", style=solid]; +"2183 slice_178" -> "2184 fill__50" [label="(4, 4)", style=solid]; +"2185 lift_fresh_copy_51" -> "2188 fill__51" [label="()", style=solid]; +"2186 slice_179" -> "2187 slice_180" [label="(4, 16)", style=solid]; +"2187 slice_180" -> "2188 fill__51" [label="(4, 8)", style=solid]; +"2189 lift_fresh_copy_52" -> "2192 fill__52" [label="()", style=solid]; +"2190 slice_181" -> "2191 slice_182" [label="(4, 16)", style=solid]; +"2191 slice_182" -> "2192 fill__52" [label="(4, 4)", style=solid]; +"2193 lift_fresh_copy_53" -> "2196 fill__53" [label="()", style=solid]; +"2194 slice_183" -> "2195 slice_184" [label="(4, 16)", style=solid]; +"2195 slice_184" -> "2196 fill__53" [label="(4, 4)", style=solid]; +"2197 view_62" -> "2198 permute_53" [label="(2, 8, 2, 8)", style=solid]; +"2198 permute_53" -> "2199 reshape_51" [label="(2, 2, 8, 8)", style=solid]; +"2199 reshape_51" -> "2200 unsqueeze_32" [label="(4, 64)", style=solid]; +"2199 reshape_51" -> "2201 unsqueeze_33" [label="(4, 64)", style=solid]; +"2200 unsqueeze_32" -> "2202 sub_5" [label="(4, 1, 64)", style=solid]; +"2201 unsqueeze_33" -> "2202 sub_5" [label="(4, 64, 1)", style=solid]; +"2202 sub_5" -> "2203 ne_5" [label="(4, 64, 64)", style=solid]; +"2202 sub_5" -> "2204 masked_fill_10" [label="(4, 64, 64)", style=solid]; +"2202 sub_5" -> "2205 eq_5" [label="(4, 64, 64)", style=solid]; +"2203 ne_5" -> "2204 masked_fill_10" [label="(4, 64, 64)", style=solid]; +"2204 masked_fill_10" -> "2206 masked_fill_11" [label="(4, 64, 64)", style=solid]; +"2205 eq_5" -> "2206 masked_fill_11" [label="(4, 64, 64)", style=solid]; +"2206 masked_fill_11" -> "2208 unsqueeze_34" [label="(4, 64, 64)", style=solid]; +"2207 view_63" -> "2210 add_39" [label="(1, 4, 12, 64, 64)", style=solid]; +"2208 unsqueeze_34" -> "2209 unsqueeze_35" [label="(4, 1, 64, 64)", style=solid]; +"2209 unsqueeze_35" -> "2210 add_39" [label="(1, 4, 1, 64, 64)", style=solid]; +"2210 add_39" -> "2211 view_64" [label="(1, 4, 12, 64, 64)", style=solid]; +"2211 view_64" -> "2212 softmax_11" [label="(4, 12, 64, 64)", style=solid]; +"2212 softmax_11" -> "2213 dropout_44" [label="(4, 12, 64, 64)", style=solid]; +"2213 dropout_44" -> "2214 matmul_23" [label="(4, 12, 64, 64)", style=solid]; +"2214 matmul_23" -> "2215 quantize_per_tensor_default_72" [label="(4, 12, 64, 32)", style=solid]; +"2215 quantize_per_tensor_default_72" -> "2216 dequantize_per_tensor_default_72" [label=None, style=solid]; +"2216 dequantize_per_tensor_default_72" -> "2217 transpose_23" [label=None, style=solid]; +"2217 transpose_23" -> "2218 reshape_52" [label="(4, 64, 12, 32)", style=solid]; +"2218 reshape_52" -> "2219 linear_71" [label="(4, 64, 384)", style=solid]; +"2219 linear_71" -> "2220 dropout_45" [label="(4, 64, 384)", style=solid]; +"2220 dropout_45" -> "2221 view_65" [label="(4, 64, 384)", style=solid]; +"2221 view_65" -> "2222 permute_54" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2222 permute_54" -> "2223 reshape_53" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2223 reshape_53" -> "2224 roll_11" [label="(1, 16, 16, 384)", style=solid]; +"2224 roll_11" -> "2225 slice_185" [label="(1, 16, 16, 384)", style=solid]; +"2225 slice_185" -> "2226 slice_186" [label="(1, 16, 16, 384)", style=solid]; +"2226 slice_186" -> "2227 slice_187" [label="(1, 14, 16, 384)", style=solid]; +"2227 slice_187" -> "2228 slice_188" [label="(1, 14, 14, 384)", style=solid]; +"2228 slice_188" -> "2229 contiguous_21" [label="(1, 14, 14, 384)", style=solid]; +"2229 contiguous_21" -> "2230 layer_norm_25" [label="(1, 14, 14, 384)", style=solid]; +"2230 layer_norm_25" -> "2231 add_40" [label="(1, 14, 14, 384)", style=solid]; +"2231 add_40" -> "2232 quantize_per_tensor_default_73" [label="(1, 14, 14, 384)", style=solid]; +"2231 add_40" -> "2242 add_41" [label="(1, 14, 14, 384)", style=solid]; +"2232 quantize_per_tensor_default_73" -> "2233 dequantize_per_tensor_default_73" [label=None, style=solid]; +"2233 dequantize_per_tensor_default_73" -> "2234 linear_72" [label=None, style=solid]; +"2234 linear_72" -> "2235 gelu_11" [label="(1, 14, 14, 1536)", style=solid]; +"2235 gelu_11" -> "2236 quantize_per_tensor_default_74" [label="(1, 14, 14, 1536)", style=solid]; +"2236 quantize_per_tensor_default_74" -> "2237 dequantize_per_tensor_default_74" [label=None, style=solid]; +"2237 dequantize_per_tensor_default_74" -> "2238 dropout_46" [label=None, style=solid]; +"2238 dropout_46" -> "2239 linear_73" [label="(1, 14, 14, 1536)", style=solid]; +"2239 linear_73" -> "2240 dropout_47" [label="(1, 14, 14, 384)", style=solid]; +"2240 dropout_47" -> "2241 layer_norm_26" [label="(1, 14, 14, 384)", style=solid]; +"2241 layer_norm_26" -> "2242 add_41" [label="(1, 14, 14, 384)", style=solid]; +"2242 add_41" -> "2243 quantize_per_tensor_default_75" [label="(1, 14, 14, 384)", style=solid]; +"2242 add_41" -> "2305 add_43" [label="(1, 14, 14, 384)", style=solid]; +"2243 quantize_per_tensor_default_75" -> "2244 dequantize_per_tensor_default_75" [label=None, style=solid]; +"2244 dequantize_per_tensor_default_75" -> "2256 pad_14" [label=None, style=solid]; +"2245 linear_74" -> "2246 relu__12" [label="(1, 15, 15, 512)", style=solid]; +"2246 relu__12" -> "2247 linear_75" [label="(1, 15, 15, 512)", style=solid]; +"2247 linear_75" -> "2248 view_66" [label="(1, 15, 15, 12)", style=solid]; +"2248 view_66" -> "2249 index_12" [label="(225, 12)", style=solid]; +"2249 index_12" -> "2250 view_67" [label="(4096, 12)", style=solid]; +"2250 view_67" -> "2251 permute_55" [label="(64, 64, 12)", style=solid]; +"2251 permute_55" -> "2252 contiguous_22" [label="(12, 64, 64)", style=solid]; +"2252 contiguous_22" -> "2253 unsqueeze_36" [label="(12, 64, 64)", style=solid]; +"2253 unsqueeze_36" -> "2254 sigmoid_12" [label="(1, 12, 64, 64)", style=solid]; +"2254 sigmoid_12" -> "2255 mul_24" [label="(1, 12, 64, 64)", style=solid]; +"2255 mul_24" -> "2286 add_42" [label="(1, 12, 64, 64)", style=solid]; +"2256 pad_14" -> "2257 view_68" [label="(1, 16, 16, 384)", style=solid]; +"2257 view_68" -> "2258 permute_56" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2258 permute_56" -> "2259 reshape_54" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2259 reshape_54" -> "2263 linear_76" [label="(4, 64, 384)", style=solid]; +"2260 clone_12" -> "2261 slice_189" [label="(1152,)", style=solid]; +"2260 clone_12" -> "2263 linear_76" [label="(1152,)", style=solid]; +"2261 slice_189" -> "2262 zero__12" [label="(384,)", style=solid]; +"2263 linear_76" -> "2264 reshape_55" [label="(4, 64, 1152)", style=solid]; +"2264 reshape_55" -> "2265 permute_57" [label="(4, 64, 3, 12, 32)", style=solid]; +"2265 permute_57" -> "2266 select_36" [label="(3, 4, 12, 64, 32)", style=solid]; +"2265 permute_57" -> "2267 select_37" [label="(3, 4, 12, 64, 32)", style=solid]; +"2265 permute_57" -> "2268 select_38" [label="(3, 4, 12, 64, 32)", style=solid]; +"2266 select_36" -> "2269 linalg_vector_norm_24" [label="(4, 12, 64, 32)", style=solid]; +"2266 select_36" -> "2271 expand_as_24" [label="(4, 12, 64, 32)", style=solid]; +"2266 select_36" -> "2272 div_24" [label="(4, 12, 64, 32)", style=solid]; +"2267 select_37" -> "2275 linalg_vector_norm_25" [label="(4, 12, 64, 32)", style=solid]; +"2267 select_37" -> "2277 expand_as_25" [label="(4, 12, 64, 32)", style=solid]; +"2267 select_37" -> "2278 div_25" [label="(4, 12, 64, 32)", style=solid]; +"2268 select_38" -> "2289 matmul_25" [label="(4, 12, 64, 32)", style=solid]; +"2269 linalg_vector_norm_24" -> "2270 clamp_min_24" [label="(4, 12, 64, 1)", style=solid]; +"2270 clamp_min_24" -> "2271 expand_as_24" [label="(4, 12, 64, 1)", style=solid]; +"2271 expand_as_24" -> "2272 div_24" [label="(4, 12, 64, 32)", style=solid]; +"2272 div_24" -> "2273 quantize_per_tensor_default_76" [label="(4, 12, 64, 32)", style=solid]; +"2273 quantize_per_tensor_default_76" -> "2274 dequantize_per_tensor_default_76" [label=None, style=solid]; +"2274 dequantize_per_tensor_default_76" -> "2282 matmul_24" [label=None, style=solid]; +"2275 linalg_vector_norm_25" -> "2276 clamp_min_25" [label="(4, 12, 64, 1)", style=solid]; +"2276 clamp_min_25" -> "2277 expand_as_25" [label="(4, 12, 64, 1)", style=solid]; +"2277 expand_as_25" -> "2278 div_25" [label="(4, 12, 64, 32)", style=solid]; +"2278 div_25" -> "2279 quantize_per_tensor_default_77" [label="(4, 12, 64, 32)", style=solid]; +"2279 quantize_per_tensor_default_77" -> "2280 dequantize_per_tensor_default_77" [label=None, style=solid]; +"2280 dequantize_per_tensor_default_77" -> "2281 transpose_24" [label=None, style=solid]; +"2281 transpose_24" -> "2282 matmul_24" [label="(4, 12, 32, 64)", style=solid]; +"2282 matmul_24" -> "2285 mul_25" [label="(4, 12, 64, 64)", style=solid]; +"2283 clamp_12" -> "2284 exp_12" [label="(12, 1, 1)", style=solid]; +"2284 exp_12" -> "2285 mul_25" [label="(12, 1, 1)", style=solid]; +"2285 mul_25" -> "2286 add_42" [label="(4, 12, 64, 64)", style=solid]; +"2286 add_42" -> "2287 softmax_12" [label="(4, 12, 64, 64)", style=solid]; +"2287 softmax_12" -> "2288 dropout_48" [label="(4, 12, 64, 64)", style=solid]; +"2288 dropout_48" -> "2289 matmul_25" [label="(4, 12, 64, 64)", style=solid]; +"2289 matmul_25" -> "2290 quantize_per_tensor_default_78" [label="(4, 12, 64, 32)", style=solid]; +"2290 quantize_per_tensor_default_78" -> "2291 dequantize_per_tensor_default_78" [label=None, style=solid]; +"2291 dequantize_per_tensor_default_78" -> "2292 transpose_25" [label=None, style=solid]; +"2292 transpose_25" -> "2293 reshape_56" [label="(4, 64, 12, 32)", style=solid]; +"2293 reshape_56" -> "2294 linear_77" [label="(4, 64, 384)", style=solid]; +"2294 linear_77" -> "2295 dropout_49" [label="(4, 64, 384)", style=solid]; +"2295 dropout_49" -> "2296 view_69" [label="(4, 64, 384)", style=solid]; +"2296 view_69" -> "2297 permute_58" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2297 permute_58" -> "2298 reshape_57" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2298 reshape_57" -> "2299 slice_190" [label="(1, 16, 16, 384)", style=solid]; +"2299 slice_190" -> "2300 slice_191" [label="(1, 16, 16, 384)", style=solid]; +"2300 slice_191" -> "2301 slice_192" [label="(1, 14, 16, 384)", style=solid]; +"2301 slice_192" -> "2302 slice_193" [label="(1, 14, 14, 384)", style=solid]; +"2302 slice_193" -> "2303 contiguous_23" [label="(1, 14, 14, 384)", style=solid]; +"2303 contiguous_23" -> "2304 layer_norm_27" [label="(1, 14, 14, 384)", style=solid]; +"2304 layer_norm_27" -> "2305 add_43" [label="(1, 14, 14, 384)", style=solid]; +"2305 add_43" -> "2306 quantize_per_tensor_default_79" [label="(1, 14, 14, 384)", style=solid]; +"2305 add_43" -> "2316 add_44" [label="(1, 14, 14, 384)", style=solid]; +"2306 quantize_per_tensor_default_79" -> "2307 dequantize_per_tensor_default_79" [label=None, style=solid]; +"2307 dequantize_per_tensor_default_79" -> "2308 linear_78" [label=None, style=solid]; +"2308 linear_78" -> "2309 gelu_12" [label="(1, 14, 14, 1536)", style=solid]; +"2309 gelu_12" -> "2310 quantize_per_tensor_default_80" [label="(1, 14, 14, 1536)", style=solid]; +"2310 quantize_per_tensor_default_80" -> "2311 dequantize_per_tensor_default_80" [label=None, style=solid]; +"2311 dequantize_per_tensor_default_80" -> "2312 dropout_50" [label=None, style=solid]; +"2312 dropout_50" -> "2313 linear_79" [label="(1, 14, 14, 1536)", style=solid]; +"2313 linear_79" -> "2314 dropout_51" [label="(1, 14, 14, 384)", style=solid]; +"2314 dropout_51" -> "2315 layer_norm_28" [label="(1, 14, 14, 384)", style=solid]; +"2315 layer_norm_28" -> "2316 add_44" [label="(1, 14, 14, 384)", style=solid]; +"2316 add_44" -> "2328 pad_15" [label="(1, 14, 14, 384)", style=solid]; +"2316 add_44" -> "2433 add_47" [label="(1, 14, 14, 384)", style=solid]; +"2317 linear_80" -> "2318 relu__13" [label="(1, 15, 15, 512)", style=solid]; +"2318 relu__13" -> "2319 linear_81" [label="(1, 15, 15, 512)", style=solid]; +"2319 linear_81" -> "2320 view_70" [label="(1, 15, 15, 12)", style=solid]; +"2320 view_70" -> "2321 index_13" [label="(225, 12)", style=solid]; +"2321 index_13" -> "2322 view_71" [label="(4096, 12)", style=solid]; +"2322 view_71" -> "2323 permute_59" [label="(64, 64, 12)", style=solid]; +"2323 permute_59" -> "2324 contiguous_24" [label="(12, 64, 64)", style=solid]; +"2324 contiguous_24" -> "2325 unsqueeze_37" [label="(12, 64, 64)", style=solid]; +"2325 unsqueeze_37" -> "2326 sigmoid_13" [label="(1, 12, 64, 64)", style=solid]; +"2326 sigmoid_13" -> "2327 mul_26" [label="(1, 12, 64, 64)", style=solid]; +"2327 mul_26" -> "2361 add_45" [label="(1, 12, 64, 64)", style=solid]; +"2328 pad_15" -> "2329 roll_12" [label="(1, 16, 16, 384)", style=solid]; +"2329 roll_12" -> "2330 view_72" [label="(1, 16, 16, 384)", style=solid]; +"2330 view_72" -> "2331 permute_60" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2331 permute_60" -> "2332 reshape_58" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2332 reshape_58" -> "2333 quantize_per_tensor_default_81" [label="(4, 64, 384)", style=solid]; +"2332 reshape_58" -> "2362 new_zeros_6" [label="(4, 64, 384)", style=solid]; +"2333 quantize_per_tensor_default_81" -> "2334 dequantize_per_tensor_default_81" [label=None, style=solid]; +"2334 dequantize_per_tensor_default_81" -> "2338 linear_82" [label=None, style=solid]; +"2335 clone_13" -> "2336 slice_194" [label="(1152,)", style=solid]; +"2335 clone_13" -> "2338 linear_82" [label="(1152,)", style=solid]; +"2336 slice_194" -> "2337 zero__13" [label="(384,)", style=solid]; +"2338 linear_82" -> "2339 reshape_59" [label="(4, 64, 1152)", style=solid]; +"2339 reshape_59" -> "2340 permute_61" [label="(4, 64, 3, 12, 32)", style=solid]; +"2340 permute_61" -> "2341 select_39" [label="(3, 4, 12, 64, 32)", style=solid]; +"2340 permute_61" -> "2342 select_40" [label="(3, 4, 12, 64, 32)", style=solid]; +"2340 permute_61" -> "2343 select_41" [label="(3, 4, 12, 64, 32)", style=solid]; +"2341 select_39" -> "2344 linalg_vector_norm_26" [label="(4, 12, 64, 32)", style=solid]; +"2341 select_39" -> "2346 expand_as_26" [label="(4, 12, 64, 32)", style=solid]; +"2341 select_39" -> "2347 div_26" [label="(4, 12, 64, 32)", style=solid]; +"2342 select_40" -> "2350 linalg_vector_norm_27" [label="(4, 12, 64, 32)", style=solid]; +"2342 select_40" -> "2352 expand_as_27" [label="(4, 12, 64, 32)", style=solid]; +"2342 select_40" -> "2353 div_27" [label="(4, 12, 64, 32)", style=solid]; +"2343 select_41" -> "2416 matmul_27" [label="(4, 12, 64, 32)", style=solid]; +"2344 linalg_vector_norm_26" -> "2345 clamp_min_26" [label="(4, 12, 64, 1)", style=solid]; +"2345 clamp_min_26" -> "2346 expand_as_26" [label="(4, 12, 64, 1)", style=solid]; +"2346 expand_as_26" -> "2347 div_26" [label="(4, 12, 64, 32)", style=solid]; +"2347 div_26" -> "2348 quantize_per_tensor_default_82" [label="(4, 12, 64, 32)", style=solid]; +"2348 quantize_per_tensor_default_82" -> "2349 dequantize_per_tensor_default_82" [label=None, style=solid]; +"2349 dequantize_per_tensor_default_82" -> "2357 matmul_26" [label=None, style=solid]; +"2350 linalg_vector_norm_27" -> "2351 clamp_min_27" [label="(4, 12, 64, 1)", style=solid]; +"2351 clamp_min_27" -> "2352 expand_as_27" [label="(4, 12, 64, 1)", style=solid]; +"2352 expand_as_27" -> "2353 div_27" [label="(4, 12, 64, 32)", style=solid]; +"2353 div_27" -> "2354 quantize_per_tensor_default_83" [label="(4, 12, 64, 32)", style=solid]; +"2354 quantize_per_tensor_default_83" -> "2355 dequantize_per_tensor_default_83" [label=None, style=solid]; +"2355 dequantize_per_tensor_default_83" -> "2356 transpose_26" [label=None, style=solid]; +"2356 transpose_26" -> "2357 matmul_26" [label="(4, 12, 32, 64)", style=solid]; +"2357 matmul_26" -> "2360 mul_27" [label="(4, 12, 64, 64)", style=solid]; +"2358 clamp_13" -> "2359 exp_13" [label="(12, 1, 1)", style=solid]; +"2359 exp_13" -> "2360 mul_27" [label="(12, 1, 1)", style=solid]; +"2360 mul_27" -> "2361 add_45" [label="(4, 12, 64, 64)", style=solid]; +"2361 add_45" -> "2409 view_74" [label="(4, 12, 64, 64)", style=solid]; +"2362 new_zeros_6" -> "2364 slice_195" [label="(16, 16)", style=solid]; +"2362 new_zeros_6" -> "2368 slice_197" [label="(16, 16)", style=solid]; +"2362 new_zeros_6" -> "2372 slice_199" [label="(16, 16)", style=solid]; +"2362 new_zeros_6" -> "2376 slice_201" [label="(16, 16)", style=solid]; +"2362 new_zeros_6" -> "2380 slice_203" [label="(16, 16)", style=solid]; +"2362 new_zeros_6" -> "2384 slice_205" [label="(16, 16)", style=solid]; +"2362 new_zeros_6" -> "2388 slice_207" [label="(16, 16)", style=solid]; +"2362 new_zeros_6" -> "2392 slice_209" [label="(16, 16)", style=solid]; +"2362 new_zeros_6" -> "2396 slice_211" [label="(16, 16)", style=solid]; +"2362 new_zeros_6" -> "2399 view_73" [label="(16, 16)", style=solid]; +"2363 lift_fresh_copy_54" -> "2366 fill__54" [label="()", style=solid]; +"2364 slice_195" -> "2365 slice_196" [label="(8, 16)", style=solid]; +"2365 slice_196" -> "2366 fill__54" [label="(8, 8)", style=solid]; +"2367 lift_fresh_copy_55" -> "2370 fill__55" [label="()", style=solid]; +"2368 slice_197" -> "2369 slice_198" [label="(8, 16)", style=solid]; +"2369 slice_198" -> "2370 fill__55" [label="(8, 4)", style=solid]; +"2371 lift_fresh_copy_56" -> "2374 fill__56" [label="()", style=solid]; +"2372 slice_199" -> "2373 slice_200" [label="(8, 16)", style=solid]; +"2373 slice_200" -> "2374 fill__56" [label="(8, 4)", style=solid]; +"2375 lift_fresh_copy_57" -> "2378 fill__57" [label="()", style=solid]; +"2376 slice_201" -> "2377 slice_202" [label="(4, 16)", style=solid]; +"2377 slice_202" -> "2378 fill__57" [label="(4, 8)", style=solid]; +"2379 lift_fresh_copy_58" -> "2382 fill__58" [label="()", style=solid]; +"2380 slice_203" -> "2381 slice_204" [label="(4, 16)", style=solid]; +"2381 slice_204" -> "2382 fill__58" [label="(4, 4)", style=solid]; +"2383 lift_fresh_copy_59" -> "2386 fill__59" [label="()", style=solid]; +"2384 slice_205" -> "2385 slice_206" [label="(4, 16)", style=solid]; +"2385 slice_206" -> "2386 fill__59" [label="(4, 4)", style=solid]; +"2387 lift_fresh_copy_60" -> "2390 fill__60" [label="()", style=solid]; +"2388 slice_207" -> "2389 slice_208" [label="(4, 16)", style=solid]; +"2389 slice_208" -> "2390 fill__60" [label="(4, 8)", style=solid]; +"2391 lift_fresh_copy_61" -> "2394 fill__61" [label="()", style=solid]; +"2392 slice_209" -> "2393 slice_210" [label="(4, 16)", style=solid]; +"2393 slice_210" -> "2394 fill__61" [label="(4, 4)", style=solid]; +"2395 lift_fresh_copy_62" -> "2398 fill__62" [label="()", style=solid]; +"2396 slice_211" -> "2397 slice_212" [label="(4, 16)", style=solid]; +"2397 slice_212" -> "2398 fill__62" [label="(4, 4)", style=solid]; +"2399 view_73" -> "2400 permute_62" [label="(2, 8, 2, 8)", style=solid]; +"2400 permute_62" -> "2401 reshape_60" [label="(2, 2, 8, 8)", style=solid]; +"2401 reshape_60" -> "2402 unsqueeze_38" [label="(4, 64)", style=solid]; +"2401 reshape_60" -> "2403 unsqueeze_39" [label="(4, 64)", style=solid]; +"2402 unsqueeze_38" -> "2404 sub_6" [label="(4, 1, 64)", style=solid]; +"2403 unsqueeze_39" -> "2404 sub_6" [label="(4, 64, 1)", style=solid]; +"2404 sub_6" -> "2405 ne_6" [label="(4, 64, 64)", style=solid]; +"2404 sub_6" -> "2406 masked_fill_12" [label="(4, 64, 64)", style=solid]; +"2404 sub_6" -> "2407 eq_6" [label="(4, 64, 64)", style=solid]; +"2405 ne_6" -> "2406 masked_fill_12" [label="(4, 64, 64)", style=solid]; +"2406 masked_fill_12" -> "2408 masked_fill_13" [label="(4, 64, 64)", style=solid]; +"2407 eq_6" -> "2408 masked_fill_13" [label="(4, 64, 64)", style=solid]; +"2408 masked_fill_13" -> "2410 unsqueeze_40" [label="(4, 64, 64)", style=solid]; +"2409 view_74" -> "2412 add_46" [label="(1, 4, 12, 64, 64)", style=solid]; +"2410 unsqueeze_40" -> "2411 unsqueeze_41" [label="(4, 1, 64, 64)", style=solid]; +"2411 unsqueeze_41" -> "2412 add_46" [label="(1, 4, 1, 64, 64)", style=solid]; +"2412 add_46" -> "2413 view_75" [label="(1, 4, 12, 64, 64)", style=solid]; +"2413 view_75" -> "2414 softmax_13" [label="(4, 12, 64, 64)", style=solid]; +"2414 softmax_13" -> "2415 dropout_52" [label="(4, 12, 64, 64)", style=solid]; +"2415 dropout_52" -> "2416 matmul_27" [label="(4, 12, 64, 64)", style=solid]; +"2416 matmul_27" -> "2417 quantize_per_tensor_default_84" [label="(4, 12, 64, 32)", style=solid]; +"2417 quantize_per_tensor_default_84" -> "2418 dequantize_per_tensor_default_84" [label=None, style=solid]; +"2418 dequantize_per_tensor_default_84" -> "2419 transpose_27" [label=None, style=solid]; +"2419 transpose_27" -> "2420 reshape_61" [label="(4, 64, 12, 32)", style=solid]; +"2420 reshape_61" -> "2421 linear_83" [label="(4, 64, 384)", style=solid]; +"2421 linear_83" -> "2422 dropout_53" [label="(4, 64, 384)", style=solid]; +"2422 dropout_53" -> "2423 view_76" [label="(4, 64, 384)", style=solid]; +"2423 view_76" -> "2424 permute_63" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2424 permute_63" -> "2425 reshape_62" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2425 reshape_62" -> "2426 roll_13" [label="(1, 16, 16, 384)", style=solid]; +"2426 roll_13" -> "2427 slice_213" [label="(1, 16, 16, 384)", style=solid]; +"2427 slice_213" -> "2428 slice_214" [label="(1, 16, 16, 384)", style=solid]; +"2428 slice_214" -> "2429 slice_215" [label="(1, 14, 16, 384)", style=solid]; +"2429 slice_215" -> "2430 slice_216" [label="(1, 14, 14, 384)", style=solid]; +"2430 slice_216" -> "2431 contiguous_25" [label="(1, 14, 14, 384)", style=solid]; +"2431 contiguous_25" -> "2432 layer_norm_29" [label="(1, 14, 14, 384)", style=solid]; +"2432 layer_norm_29" -> "2433 add_47" [label="(1, 14, 14, 384)", style=solid]; +"2433 add_47" -> "2434 quantize_per_tensor_default_85" [label="(1, 14, 14, 384)", style=solid]; +"2433 add_47" -> "2444 add_48" [label="(1, 14, 14, 384)", style=solid]; +"2434 quantize_per_tensor_default_85" -> "2435 dequantize_per_tensor_default_85" [label=None, style=solid]; +"2435 dequantize_per_tensor_default_85" -> "2436 linear_84" [label=None, style=solid]; +"2436 linear_84" -> "2437 gelu_13" [label="(1, 14, 14, 1536)", style=solid]; +"2437 gelu_13" -> "2438 quantize_per_tensor_default_86" [label="(1, 14, 14, 1536)", style=solid]; +"2438 quantize_per_tensor_default_86" -> "2439 dequantize_per_tensor_default_86" [label=None, style=solid]; +"2439 dequantize_per_tensor_default_86" -> "2440 dropout_54" [label=None, style=solid]; +"2440 dropout_54" -> "2441 linear_85" [label="(1, 14, 14, 1536)", style=solid]; +"2441 linear_85" -> "2442 dropout_55" [label="(1, 14, 14, 384)", style=solid]; +"2442 dropout_55" -> "2443 layer_norm_30" [label="(1, 14, 14, 384)", style=solid]; +"2443 layer_norm_30" -> "2444 add_48" [label="(1, 14, 14, 384)", style=solid]; +"2444 add_48" -> "2445 quantize_per_tensor_default_87" [label="(1, 14, 14, 384)", style=solid]; +"2444 add_48" -> "2507 add_50" [label="(1, 14, 14, 384)", style=solid]; +"2445 quantize_per_tensor_default_87" -> "2446 dequantize_per_tensor_default_87" [label=None, style=solid]; +"2446 dequantize_per_tensor_default_87" -> "2458 pad_16" [label=None, style=solid]; +"2447 linear_86" -> "2448 relu__14" [label="(1, 15, 15, 512)", style=solid]; +"2448 relu__14" -> "2449 linear_87" [label="(1, 15, 15, 512)", style=solid]; +"2449 linear_87" -> "2450 view_77" [label="(1, 15, 15, 12)", style=solid]; +"2450 view_77" -> "2451 index_14" [label="(225, 12)", style=solid]; +"2451 index_14" -> "2452 view_78" [label="(4096, 12)", style=solid]; +"2452 view_78" -> "2453 permute_64" [label="(64, 64, 12)", style=solid]; +"2453 permute_64" -> "2454 contiguous_26" [label="(12, 64, 64)", style=solid]; +"2454 contiguous_26" -> "2455 unsqueeze_42" [label="(12, 64, 64)", style=solid]; +"2455 unsqueeze_42" -> "2456 sigmoid_14" [label="(1, 12, 64, 64)", style=solid]; +"2456 sigmoid_14" -> "2457 mul_28" [label="(1, 12, 64, 64)", style=solid]; +"2457 mul_28" -> "2488 add_49" [label="(1, 12, 64, 64)", style=solid]; +"2458 pad_16" -> "2459 view_79" [label="(1, 16, 16, 384)", style=solid]; +"2459 view_79" -> "2460 permute_65" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2460 permute_65" -> "2461 reshape_63" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2461 reshape_63" -> "2465 linear_88" [label="(4, 64, 384)", style=solid]; +"2462 clone_14" -> "2463 slice_217" [label="(1152,)", style=solid]; +"2462 clone_14" -> "2465 linear_88" [label="(1152,)", style=solid]; +"2463 slice_217" -> "2464 zero__14" [label="(384,)", style=solid]; +"2465 linear_88" -> "2466 reshape_64" [label="(4, 64, 1152)", style=solid]; +"2466 reshape_64" -> "2467 permute_66" [label="(4, 64, 3, 12, 32)", style=solid]; +"2467 permute_66" -> "2468 select_42" [label="(3, 4, 12, 64, 32)", style=solid]; +"2467 permute_66" -> "2469 select_43" [label="(3, 4, 12, 64, 32)", style=solid]; +"2467 permute_66" -> "2470 select_44" [label="(3, 4, 12, 64, 32)", style=solid]; +"2468 select_42" -> "2471 linalg_vector_norm_28" [label="(4, 12, 64, 32)", style=solid]; +"2468 select_42" -> "2473 expand_as_28" [label="(4, 12, 64, 32)", style=solid]; +"2468 select_42" -> "2474 div_28" [label="(4, 12, 64, 32)", style=solid]; +"2469 select_43" -> "2477 linalg_vector_norm_29" [label="(4, 12, 64, 32)", style=solid]; +"2469 select_43" -> "2479 expand_as_29" [label="(4, 12, 64, 32)", style=solid]; +"2469 select_43" -> "2480 div_29" [label="(4, 12, 64, 32)", style=solid]; +"2470 select_44" -> "2491 matmul_29" [label="(4, 12, 64, 32)", style=solid]; +"2471 linalg_vector_norm_28" -> "2472 clamp_min_28" [label="(4, 12, 64, 1)", style=solid]; +"2472 clamp_min_28" -> "2473 expand_as_28" [label="(4, 12, 64, 1)", style=solid]; +"2473 expand_as_28" -> "2474 div_28" [label="(4, 12, 64, 32)", style=solid]; +"2474 div_28" -> "2475 quantize_per_tensor_default_88" [label="(4, 12, 64, 32)", style=solid]; +"2475 quantize_per_tensor_default_88" -> "2476 dequantize_per_tensor_default_88" [label=None, style=solid]; +"2476 dequantize_per_tensor_default_88" -> "2484 matmul_28" [label=None, style=solid]; +"2477 linalg_vector_norm_29" -> "2478 clamp_min_29" [label="(4, 12, 64, 1)", style=solid]; +"2478 clamp_min_29" -> "2479 expand_as_29" [label="(4, 12, 64, 1)", style=solid]; +"2479 expand_as_29" -> "2480 div_29" [label="(4, 12, 64, 32)", style=solid]; +"2480 div_29" -> "2481 quantize_per_tensor_default_89" [label="(4, 12, 64, 32)", style=solid]; +"2481 quantize_per_tensor_default_89" -> "2482 dequantize_per_tensor_default_89" [label=None, style=solid]; +"2482 dequantize_per_tensor_default_89" -> "2483 transpose_28" [label=None, style=solid]; +"2483 transpose_28" -> "2484 matmul_28" [label="(4, 12, 32, 64)", style=solid]; +"2484 matmul_28" -> "2487 mul_29" [label="(4, 12, 64, 64)", style=solid]; +"2485 clamp_14" -> "2486 exp_14" [label="(12, 1, 1)", style=solid]; +"2486 exp_14" -> "2487 mul_29" [label="(12, 1, 1)", style=solid]; +"2487 mul_29" -> "2488 add_49" [label="(4, 12, 64, 64)", style=solid]; +"2488 add_49" -> "2489 softmax_14" [label="(4, 12, 64, 64)", style=solid]; +"2489 softmax_14" -> "2490 dropout_56" [label="(4, 12, 64, 64)", style=solid]; +"2490 dropout_56" -> "2491 matmul_29" [label="(4, 12, 64, 64)", style=solid]; +"2491 matmul_29" -> "2492 quantize_per_tensor_default_90" [label="(4, 12, 64, 32)", style=solid]; +"2492 quantize_per_tensor_default_90" -> "2493 dequantize_per_tensor_default_90" [label=None, style=solid]; +"2493 dequantize_per_tensor_default_90" -> "2494 transpose_29" [label=None, style=solid]; +"2494 transpose_29" -> "2495 reshape_65" [label="(4, 64, 12, 32)", style=solid]; +"2495 reshape_65" -> "2496 linear_89" [label="(4, 64, 384)", style=solid]; +"2496 linear_89" -> "2497 dropout_57" [label="(4, 64, 384)", style=solid]; +"2497 dropout_57" -> "2498 view_80" [label="(4, 64, 384)", style=solid]; +"2498 view_80" -> "2499 permute_67" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2499 permute_67" -> "2500 reshape_66" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2500 reshape_66" -> "2501 slice_218" [label="(1, 16, 16, 384)", style=solid]; +"2501 slice_218" -> "2502 slice_219" [label="(1, 16, 16, 384)", style=solid]; +"2502 slice_219" -> "2503 slice_220" [label="(1, 14, 16, 384)", style=solid]; +"2503 slice_220" -> "2504 slice_221" [label="(1, 14, 14, 384)", style=solid]; +"2504 slice_221" -> "2505 contiguous_27" [label="(1, 14, 14, 384)", style=solid]; +"2505 contiguous_27" -> "2506 layer_norm_31" [label="(1, 14, 14, 384)", style=solid]; +"2506 layer_norm_31" -> "2507 add_50" [label="(1, 14, 14, 384)", style=solid]; +"2507 add_50" -> "2508 quantize_per_tensor_default_91" [label="(1, 14, 14, 384)", style=solid]; +"2507 add_50" -> "2518 add_51" [label="(1, 14, 14, 384)", style=solid]; +"2508 quantize_per_tensor_default_91" -> "2509 dequantize_per_tensor_default_91" [label=None, style=solid]; +"2509 dequantize_per_tensor_default_91" -> "2510 linear_90" [label=None, style=solid]; +"2510 linear_90" -> "2511 gelu_14" [label="(1, 14, 14, 1536)", style=solid]; +"2511 gelu_14" -> "2512 quantize_per_tensor_default_92" [label="(1, 14, 14, 1536)", style=solid]; +"2512 quantize_per_tensor_default_92" -> "2513 dequantize_per_tensor_default_92" [label=None, style=solid]; +"2513 dequantize_per_tensor_default_92" -> "2514 dropout_58" [label=None, style=solid]; +"2514 dropout_58" -> "2515 linear_91" [label="(1, 14, 14, 1536)", style=solid]; +"2515 linear_91" -> "2516 dropout_59" [label="(1, 14, 14, 384)", style=solid]; +"2516 dropout_59" -> "2517 layer_norm_32" [label="(1, 14, 14, 384)", style=solid]; +"2517 layer_norm_32" -> "2518 add_51" [label="(1, 14, 14, 384)", style=solid]; +"2518 add_51" -> "2530 pad_17" [label="(1, 14, 14, 384)", style=solid]; +"2518 add_51" -> "2635 add_54" [label="(1, 14, 14, 384)", style=solid]; +"2519 linear_92" -> "2520 relu__15" [label="(1, 15, 15, 512)", style=solid]; +"2520 relu__15" -> "2521 linear_93" [label="(1, 15, 15, 512)", style=solid]; +"2521 linear_93" -> "2522 view_81" [label="(1, 15, 15, 12)", style=solid]; +"2522 view_81" -> "2523 index_15" [label="(225, 12)", style=solid]; +"2523 index_15" -> "2524 view_82" [label="(4096, 12)", style=solid]; +"2524 view_82" -> "2525 permute_68" [label="(64, 64, 12)", style=solid]; +"2525 permute_68" -> "2526 contiguous_28" [label="(12, 64, 64)", style=solid]; +"2526 contiguous_28" -> "2527 unsqueeze_43" [label="(12, 64, 64)", style=solid]; +"2527 unsqueeze_43" -> "2528 sigmoid_15" [label="(1, 12, 64, 64)", style=solid]; +"2528 sigmoid_15" -> "2529 mul_30" [label="(1, 12, 64, 64)", style=solid]; +"2529 mul_30" -> "2563 add_52" [label="(1, 12, 64, 64)", style=solid]; +"2530 pad_17" -> "2531 roll_14" [label="(1, 16, 16, 384)", style=solid]; +"2531 roll_14" -> "2532 view_83" [label="(1, 16, 16, 384)", style=solid]; +"2532 view_83" -> "2533 permute_69" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2533 permute_69" -> "2534 reshape_67" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2534 reshape_67" -> "2535 quantize_per_tensor_default_93" [label="(4, 64, 384)", style=solid]; +"2534 reshape_67" -> "2564 new_zeros_7" [label="(4, 64, 384)", style=solid]; +"2535 quantize_per_tensor_default_93" -> "2536 dequantize_per_tensor_default_93" [label=None, style=solid]; +"2536 dequantize_per_tensor_default_93" -> "2540 linear_94" [label=None, style=solid]; +"2537 clone_15" -> "2538 slice_222" [label="(1152,)", style=solid]; +"2537 clone_15" -> "2540 linear_94" [label="(1152,)", style=solid]; +"2538 slice_222" -> "2539 zero__15" [label="(384,)", style=solid]; +"2540 linear_94" -> "2541 reshape_68" [label="(4, 64, 1152)", style=solid]; +"2541 reshape_68" -> "2542 permute_70" [label="(4, 64, 3, 12, 32)", style=solid]; +"2542 permute_70" -> "2543 select_45" [label="(3, 4, 12, 64, 32)", style=solid]; +"2542 permute_70" -> "2544 select_46" [label="(3, 4, 12, 64, 32)", style=solid]; +"2542 permute_70" -> "2545 select_47" [label="(3, 4, 12, 64, 32)", style=solid]; +"2543 select_45" -> "2546 linalg_vector_norm_30" [label="(4, 12, 64, 32)", style=solid]; +"2543 select_45" -> "2548 expand_as_30" [label="(4, 12, 64, 32)", style=solid]; +"2543 select_45" -> "2549 div_30" [label="(4, 12, 64, 32)", style=solid]; +"2544 select_46" -> "2552 linalg_vector_norm_31" [label="(4, 12, 64, 32)", style=solid]; +"2544 select_46" -> "2554 expand_as_31" [label="(4, 12, 64, 32)", style=solid]; +"2544 select_46" -> "2555 div_31" [label="(4, 12, 64, 32)", style=solid]; +"2545 select_47" -> "2618 matmul_31" [label="(4, 12, 64, 32)", style=solid]; +"2546 linalg_vector_norm_30" -> "2547 clamp_min_30" [label="(4, 12, 64, 1)", style=solid]; +"2547 clamp_min_30" -> "2548 expand_as_30" [label="(4, 12, 64, 1)", style=solid]; +"2548 expand_as_30" -> "2549 div_30" [label="(4, 12, 64, 32)", style=solid]; +"2549 div_30" -> "2550 quantize_per_tensor_default_94" [label="(4, 12, 64, 32)", style=solid]; +"2550 quantize_per_tensor_default_94" -> "2551 dequantize_per_tensor_default_94" [label=None, style=solid]; +"2551 dequantize_per_tensor_default_94" -> "2559 matmul_30" [label=None, style=solid]; +"2552 linalg_vector_norm_31" -> "2553 clamp_min_31" [label="(4, 12, 64, 1)", style=solid]; +"2553 clamp_min_31" -> "2554 expand_as_31" [label="(4, 12, 64, 1)", style=solid]; +"2554 expand_as_31" -> "2555 div_31" [label="(4, 12, 64, 32)", style=solid]; +"2555 div_31" -> "2556 quantize_per_tensor_default_95" [label="(4, 12, 64, 32)", style=solid]; +"2556 quantize_per_tensor_default_95" -> "2557 dequantize_per_tensor_default_95" [label=None, style=solid]; +"2557 dequantize_per_tensor_default_95" -> "2558 transpose_30" [label=None, style=solid]; +"2558 transpose_30" -> "2559 matmul_30" [label="(4, 12, 32, 64)", style=solid]; +"2559 matmul_30" -> "2562 mul_31" [label="(4, 12, 64, 64)", style=solid]; +"2560 clamp_15" -> "2561 exp_15" [label="(12, 1, 1)", style=solid]; +"2561 exp_15" -> "2562 mul_31" [label="(12, 1, 1)", style=solid]; +"2562 mul_31" -> "2563 add_52" [label="(4, 12, 64, 64)", style=solid]; +"2563 add_52" -> "2611 view_85" [label="(4, 12, 64, 64)", style=solid]; +"2564 new_zeros_7" -> "2566 slice_223" [label="(16, 16)", style=solid]; +"2564 new_zeros_7" -> "2570 slice_225" [label="(16, 16)", style=solid]; +"2564 new_zeros_7" -> "2574 slice_227" [label="(16, 16)", style=solid]; +"2564 new_zeros_7" -> "2578 slice_229" [label="(16, 16)", style=solid]; +"2564 new_zeros_7" -> "2582 slice_231" [label="(16, 16)", style=solid]; +"2564 new_zeros_7" -> "2586 slice_233" [label="(16, 16)", style=solid]; +"2564 new_zeros_7" -> "2590 slice_235" [label="(16, 16)", style=solid]; +"2564 new_zeros_7" -> "2594 slice_237" [label="(16, 16)", style=solid]; +"2564 new_zeros_7" -> "2598 slice_239" [label="(16, 16)", style=solid]; +"2564 new_zeros_7" -> "2601 view_84" [label="(16, 16)", style=solid]; +"2565 lift_fresh_copy_63" -> "2568 fill__63" [label="()", style=solid]; +"2566 slice_223" -> "2567 slice_224" [label="(8, 16)", style=solid]; +"2567 slice_224" -> "2568 fill__63" [label="(8, 8)", style=solid]; +"2569 lift_fresh_copy_64" -> "2572 fill__64" [label="()", style=solid]; +"2570 slice_225" -> "2571 slice_226" [label="(8, 16)", style=solid]; +"2571 slice_226" -> "2572 fill__64" [label="(8, 4)", style=solid]; +"2573 lift_fresh_copy_65" -> "2576 fill__65" [label="()", style=solid]; +"2574 slice_227" -> "2575 slice_228" [label="(8, 16)", style=solid]; +"2575 slice_228" -> "2576 fill__65" [label="(8, 4)", style=solid]; +"2577 lift_fresh_copy_66" -> "2580 fill__66" [label="()", style=solid]; +"2578 slice_229" -> "2579 slice_230" [label="(4, 16)", style=solid]; +"2579 slice_230" -> "2580 fill__66" [label="(4, 8)", style=solid]; +"2581 lift_fresh_copy_67" -> "2584 fill__67" [label="()", style=solid]; +"2582 slice_231" -> "2583 slice_232" [label="(4, 16)", style=solid]; +"2583 slice_232" -> "2584 fill__67" [label="(4, 4)", style=solid]; +"2585 lift_fresh_copy_68" -> "2588 fill__68" [label="()", style=solid]; +"2586 slice_233" -> "2587 slice_234" [label="(4, 16)", style=solid]; +"2587 slice_234" -> "2588 fill__68" [label="(4, 4)", style=solid]; +"2589 lift_fresh_copy_69" -> "2592 fill__69" [label="()", style=solid]; +"2590 slice_235" -> "2591 slice_236" [label="(4, 16)", style=solid]; +"2591 slice_236" -> "2592 fill__69" [label="(4, 8)", style=solid]; +"2593 lift_fresh_copy_70" -> "2596 fill__70" [label="()", style=solid]; +"2594 slice_237" -> "2595 slice_238" [label="(4, 16)", style=solid]; +"2595 slice_238" -> "2596 fill__70" [label="(4, 4)", style=solid]; +"2597 lift_fresh_copy_71" -> "2600 fill__71" [label="()", style=solid]; +"2598 slice_239" -> "2599 slice_240" [label="(4, 16)", style=solid]; +"2599 slice_240" -> "2600 fill__71" [label="(4, 4)", style=solid]; +"2601 view_84" -> "2602 permute_71" [label="(2, 8, 2, 8)", style=solid]; +"2602 permute_71" -> "2603 reshape_69" [label="(2, 2, 8, 8)", style=solid]; +"2603 reshape_69" -> "2604 unsqueeze_44" [label="(4, 64)", style=solid]; +"2603 reshape_69" -> "2605 unsqueeze_45" [label="(4, 64)", style=solid]; +"2604 unsqueeze_44" -> "2606 sub_7" [label="(4, 1, 64)", style=solid]; +"2605 unsqueeze_45" -> "2606 sub_7" [label="(4, 64, 1)", style=solid]; +"2606 sub_7" -> "2607 ne_7" [label="(4, 64, 64)", style=solid]; +"2606 sub_7" -> "2608 masked_fill_14" [label="(4, 64, 64)", style=solid]; +"2606 sub_7" -> "2609 eq_7" [label="(4, 64, 64)", style=solid]; +"2607 ne_7" -> "2608 masked_fill_14" [label="(4, 64, 64)", style=solid]; +"2608 masked_fill_14" -> "2610 masked_fill_15" [label="(4, 64, 64)", style=solid]; +"2609 eq_7" -> "2610 masked_fill_15" [label="(4, 64, 64)", style=solid]; +"2610 masked_fill_15" -> "2612 unsqueeze_46" [label="(4, 64, 64)", style=solid]; +"2611 view_85" -> "2614 add_53" [label="(1, 4, 12, 64, 64)", style=solid]; +"2612 unsqueeze_46" -> "2613 unsqueeze_47" [label="(4, 1, 64, 64)", style=solid]; +"2613 unsqueeze_47" -> "2614 add_53" [label="(1, 4, 1, 64, 64)", style=solid]; +"2614 add_53" -> "2615 view_86" [label="(1, 4, 12, 64, 64)", style=solid]; +"2615 view_86" -> "2616 softmax_15" [label="(4, 12, 64, 64)", style=solid]; +"2616 softmax_15" -> "2617 dropout_60" [label="(4, 12, 64, 64)", style=solid]; +"2617 dropout_60" -> "2618 matmul_31" [label="(4, 12, 64, 64)", style=solid]; +"2618 matmul_31" -> "2619 quantize_per_tensor_default_96" [label="(4, 12, 64, 32)", style=solid]; +"2619 quantize_per_tensor_default_96" -> "2620 dequantize_per_tensor_default_96" [label=None, style=solid]; +"2620 dequantize_per_tensor_default_96" -> "2621 transpose_31" [label=None, style=solid]; +"2621 transpose_31" -> "2622 reshape_70" [label="(4, 64, 12, 32)", style=solid]; +"2622 reshape_70" -> "2623 linear_95" [label="(4, 64, 384)", style=solid]; +"2623 linear_95" -> "2624 dropout_61" [label="(4, 64, 384)", style=solid]; +"2624 dropout_61" -> "2625 view_87" [label="(4, 64, 384)", style=solid]; +"2625 view_87" -> "2626 permute_72" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2626 permute_72" -> "2627 reshape_71" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2627 reshape_71" -> "2628 roll_15" [label="(1, 16, 16, 384)", style=solid]; +"2628 roll_15" -> "2629 slice_241" [label="(1, 16, 16, 384)", style=solid]; +"2629 slice_241" -> "2630 slice_242" [label="(1, 16, 16, 384)", style=solid]; +"2630 slice_242" -> "2631 slice_243" [label="(1, 14, 16, 384)", style=solid]; +"2631 slice_243" -> "2632 slice_244" [label="(1, 14, 14, 384)", style=solid]; +"2632 slice_244" -> "2633 contiguous_29" [label="(1, 14, 14, 384)", style=solid]; +"2633 contiguous_29" -> "2634 layer_norm_33" [label="(1, 14, 14, 384)", style=solid]; +"2634 layer_norm_33" -> "2635 add_54" [label="(1, 14, 14, 384)", style=solid]; +"2635 add_54" -> "2636 quantize_per_tensor_default_97" [label="(1, 14, 14, 384)", style=solid]; +"2635 add_54" -> "2646 add_55" [label="(1, 14, 14, 384)", style=solid]; +"2636 quantize_per_tensor_default_97" -> "2637 dequantize_per_tensor_default_97" [label=None, style=solid]; +"2637 dequantize_per_tensor_default_97" -> "2638 linear_96" [label=None, style=solid]; +"2638 linear_96" -> "2639 gelu_15" [label="(1, 14, 14, 1536)", style=solid]; +"2639 gelu_15" -> "2640 quantize_per_tensor_default_98" [label="(1, 14, 14, 1536)", style=solid]; +"2640 quantize_per_tensor_default_98" -> "2641 dequantize_per_tensor_default_98" [label=None, style=solid]; +"2641 dequantize_per_tensor_default_98" -> "2642 dropout_62" [label=None, style=solid]; +"2642 dropout_62" -> "2643 linear_97" [label="(1, 14, 14, 1536)", style=solid]; +"2643 linear_97" -> "2644 dropout_63" [label="(1, 14, 14, 384)", style=solid]; +"2644 dropout_63" -> "2645 layer_norm_34" [label="(1, 14, 14, 384)", style=solid]; +"2645 layer_norm_34" -> "2646 add_55" [label="(1, 14, 14, 384)", style=solid]; +"2646 add_55" -> "2647 quantize_per_tensor_default_99" [label="(1, 14, 14, 384)", style=solid]; +"2646 add_55" -> "2709 add_57" [label="(1, 14, 14, 384)", style=solid]; +"2647 quantize_per_tensor_default_99" -> "2648 dequantize_per_tensor_default_99" [label=None, style=solid]; +"2648 dequantize_per_tensor_default_99" -> "2660 pad_18" [label=None, style=solid]; +"2649 linear_98" -> "2650 relu__16" [label="(1, 15, 15, 512)", style=solid]; +"2650 relu__16" -> "2651 linear_99" [label="(1, 15, 15, 512)", style=solid]; +"2651 linear_99" -> "2652 view_88" [label="(1, 15, 15, 12)", style=solid]; +"2652 view_88" -> "2653 index_16" [label="(225, 12)", style=solid]; +"2653 index_16" -> "2654 view_89" [label="(4096, 12)", style=solid]; +"2654 view_89" -> "2655 permute_73" [label="(64, 64, 12)", style=solid]; +"2655 permute_73" -> "2656 contiguous_30" [label="(12, 64, 64)", style=solid]; +"2656 contiguous_30" -> "2657 unsqueeze_48" [label="(12, 64, 64)", style=solid]; +"2657 unsqueeze_48" -> "2658 sigmoid_16" [label="(1, 12, 64, 64)", style=solid]; +"2658 sigmoid_16" -> "2659 mul_32" [label="(1, 12, 64, 64)", style=solid]; +"2659 mul_32" -> "2690 add_56" [label="(1, 12, 64, 64)", style=solid]; +"2660 pad_18" -> "2661 view_90" [label="(1, 16, 16, 384)", style=solid]; +"2661 view_90" -> "2662 permute_74" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2662 permute_74" -> "2663 reshape_72" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2663 reshape_72" -> "2667 linear_100" [label="(4, 64, 384)", style=solid]; +"2664 clone_16" -> "2665 slice_245" [label="(1152,)", style=solid]; +"2664 clone_16" -> "2667 linear_100" [label="(1152,)", style=solid]; +"2665 slice_245" -> "2666 zero__16" [label="(384,)", style=solid]; +"2667 linear_100" -> "2668 reshape_73" [label="(4, 64, 1152)", style=solid]; +"2668 reshape_73" -> "2669 permute_75" [label="(4, 64, 3, 12, 32)", style=solid]; +"2669 permute_75" -> "2670 select_48" [label="(3, 4, 12, 64, 32)", style=solid]; +"2669 permute_75" -> "2671 select_49" [label="(3, 4, 12, 64, 32)", style=solid]; +"2669 permute_75" -> "2672 select_50" [label="(3, 4, 12, 64, 32)", style=solid]; +"2670 select_48" -> "2673 linalg_vector_norm_32" [label="(4, 12, 64, 32)", style=solid]; +"2670 select_48" -> "2675 expand_as_32" [label="(4, 12, 64, 32)", style=solid]; +"2670 select_48" -> "2676 div_32" [label="(4, 12, 64, 32)", style=solid]; +"2671 select_49" -> "2679 linalg_vector_norm_33" [label="(4, 12, 64, 32)", style=solid]; +"2671 select_49" -> "2681 expand_as_33" [label="(4, 12, 64, 32)", style=solid]; +"2671 select_49" -> "2682 div_33" [label="(4, 12, 64, 32)", style=solid]; +"2672 select_50" -> "2693 matmul_33" [label="(4, 12, 64, 32)", style=solid]; +"2673 linalg_vector_norm_32" -> "2674 clamp_min_32" [label="(4, 12, 64, 1)", style=solid]; +"2674 clamp_min_32" -> "2675 expand_as_32" [label="(4, 12, 64, 1)", style=solid]; +"2675 expand_as_32" -> "2676 div_32" [label="(4, 12, 64, 32)", style=solid]; +"2676 div_32" -> "2677 quantize_per_tensor_default_100" [label="(4, 12, 64, 32)", style=solid]; +"2677 quantize_per_tensor_default_100" -> "2678 dequantize_per_tensor_default_100" [label=None, style=solid]; +"2678 dequantize_per_tensor_default_100" -> "2686 matmul_32" [label=None, style=solid]; +"2679 linalg_vector_norm_33" -> "2680 clamp_min_33" [label="(4, 12, 64, 1)", style=solid]; +"2680 clamp_min_33" -> "2681 expand_as_33" [label="(4, 12, 64, 1)", style=solid]; +"2681 expand_as_33" -> "2682 div_33" [label="(4, 12, 64, 32)", style=solid]; +"2682 div_33" -> "2683 quantize_per_tensor_default_101" [label="(4, 12, 64, 32)", style=solid]; +"2683 quantize_per_tensor_default_101" -> "2684 dequantize_per_tensor_default_101" [label=None, style=solid]; +"2684 dequantize_per_tensor_default_101" -> "2685 transpose_32" [label=None, style=solid]; +"2685 transpose_32" -> "2686 matmul_32" [label="(4, 12, 32, 64)", style=solid]; +"2686 matmul_32" -> "2689 mul_33" [label="(4, 12, 64, 64)", style=solid]; +"2687 clamp_16" -> "2688 exp_16" [label="(12, 1, 1)", style=solid]; +"2688 exp_16" -> "2689 mul_33" [label="(12, 1, 1)", style=solid]; +"2689 mul_33" -> "2690 add_56" [label="(4, 12, 64, 64)", style=solid]; +"2690 add_56" -> "2691 softmax_16" [label="(4, 12, 64, 64)", style=solid]; +"2691 softmax_16" -> "2692 dropout_64" [label="(4, 12, 64, 64)", style=solid]; +"2692 dropout_64" -> "2693 matmul_33" [label="(4, 12, 64, 64)", style=solid]; +"2693 matmul_33" -> "2694 quantize_per_tensor_default_102" [label="(4, 12, 64, 32)", style=solid]; +"2694 quantize_per_tensor_default_102" -> "2695 dequantize_per_tensor_default_102" [label=None, style=solid]; +"2695 dequantize_per_tensor_default_102" -> "2696 transpose_33" [label=None, style=solid]; +"2696 transpose_33" -> "2697 reshape_74" [label="(4, 64, 12, 32)", style=solid]; +"2697 reshape_74" -> "2698 linear_101" [label="(4, 64, 384)", style=solid]; +"2698 linear_101" -> "2699 dropout_65" [label="(4, 64, 384)", style=solid]; +"2699 dropout_65" -> "2700 view_91" [label="(4, 64, 384)", style=solid]; +"2700 view_91" -> "2701 permute_76" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2701 permute_76" -> "2702 reshape_75" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2702 reshape_75" -> "2703 slice_246" [label="(1, 16, 16, 384)", style=solid]; +"2703 slice_246" -> "2704 slice_247" [label="(1, 16, 16, 384)", style=solid]; +"2704 slice_247" -> "2705 slice_248" [label="(1, 14, 16, 384)", style=solid]; +"2705 slice_248" -> "2706 slice_249" [label="(1, 14, 14, 384)", style=solid]; +"2706 slice_249" -> "2707 contiguous_31" [label="(1, 14, 14, 384)", style=solid]; +"2707 contiguous_31" -> "2708 layer_norm_35" [label="(1, 14, 14, 384)", style=solid]; +"2708 layer_norm_35" -> "2709 add_57" [label="(1, 14, 14, 384)", style=solid]; +"2709 add_57" -> "2710 quantize_per_tensor_default_103" [label="(1, 14, 14, 384)", style=solid]; +"2709 add_57" -> "2720 add_58" [label="(1, 14, 14, 384)", style=solid]; +"2710 quantize_per_tensor_default_103" -> "2711 dequantize_per_tensor_default_103" [label=None, style=solid]; +"2711 dequantize_per_tensor_default_103" -> "2712 linear_102" [label=None, style=solid]; +"2712 linear_102" -> "2713 gelu_16" [label="(1, 14, 14, 1536)", style=solid]; +"2713 gelu_16" -> "2714 quantize_per_tensor_default_104" [label="(1, 14, 14, 1536)", style=solid]; +"2714 quantize_per_tensor_default_104" -> "2715 dequantize_per_tensor_default_104" [label=None, style=solid]; +"2715 dequantize_per_tensor_default_104" -> "2716 dropout_66" [label=None, style=solid]; +"2716 dropout_66" -> "2717 linear_103" [label="(1, 14, 14, 1536)", style=solid]; +"2717 linear_103" -> "2718 dropout_67" [label="(1, 14, 14, 384)", style=solid]; +"2718 dropout_67" -> "2719 layer_norm_36" [label="(1, 14, 14, 384)", style=solid]; +"2719 layer_norm_36" -> "2720 add_58" [label="(1, 14, 14, 384)", style=solid]; +"2720 add_58" -> "2732 pad_19" [label="(1, 14, 14, 384)", style=solid]; +"2720 add_58" -> "2837 add_61" [label="(1, 14, 14, 384)", style=solid]; +"2721 linear_104" -> "2722 relu__17" [label="(1, 15, 15, 512)", style=solid]; +"2722 relu__17" -> "2723 linear_105" [label="(1, 15, 15, 512)", style=solid]; +"2723 linear_105" -> "2724 view_92" [label="(1, 15, 15, 12)", style=solid]; +"2724 view_92" -> "2725 index_17" [label="(225, 12)", style=solid]; +"2725 index_17" -> "2726 view_93" [label="(4096, 12)", style=solid]; +"2726 view_93" -> "2727 permute_77" [label="(64, 64, 12)", style=solid]; +"2727 permute_77" -> "2728 contiguous_32" [label="(12, 64, 64)", style=solid]; +"2728 contiguous_32" -> "2729 unsqueeze_49" [label="(12, 64, 64)", style=solid]; +"2729 unsqueeze_49" -> "2730 sigmoid_17" [label="(1, 12, 64, 64)", style=solid]; +"2730 sigmoid_17" -> "2731 mul_34" [label="(1, 12, 64, 64)", style=solid]; +"2731 mul_34" -> "2765 add_59" [label="(1, 12, 64, 64)", style=solid]; +"2732 pad_19" -> "2733 roll_16" [label="(1, 16, 16, 384)", style=solid]; +"2733 roll_16" -> "2734 view_94" [label="(1, 16, 16, 384)", style=solid]; +"2734 view_94" -> "2735 permute_78" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2735 permute_78" -> "2736 reshape_76" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2736 reshape_76" -> "2737 quantize_per_tensor_default_105" [label="(4, 64, 384)", style=solid]; +"2736 reshape_76" -> "2766 new_zeros_8" [label="(4, 64, 384)", style=solid]; +"2737 quantize_per_tensor_default_105" -> "2738 dequantize_per_tensor_default_105" [label=None, style=solid]; +"2738 dequantize_per_tensor_default_105" -> "2742 linear_106" [label=None, style=solid]; +"2739 clone_17" -> "2740 slice_250" [label="(1152,)", style=solid]; +"2739 clone_17" -> "2742 linear_106" [label="(1152,)", style=solid]; +"2740 slice_250" -> "2741 zero__17" [label="(384,)", style=solid]; +"2742 linear_106" -> "2743 reshape_77" [label="(4, 64, 1152)", style=solid]; +"2743 reshape_77" -> "2744 permute_79" [label="(4, 64, 3, 12, 32)", style=solid]; +"2744 permute_79" -> "2745 select_51" [label="(3, 4, 12, 64, 32)", style=solid]; +"2744 permute_79" -> "2746 select_52" [label="(3, 4, 12, 64, 32)", style=solid]; +"2744 permute_79" -> "2747 select_53" [label="(3, 4, 12, 64, 32)", style=solid]; +"2745 select_51" -> "2748 linalg_vector_norm_34" [label="(4, 12, 64, 32)", style=solid]; +"2745 select_51" -> "2750 expand_as_34" [label="(4, 12, 64, 32)", style=solid]; +"2745 select_51" -> "2751 div_34" [label="(4, 12, 64, 32)", style=solid]; +"2746 select_52" -> "2754 linalg_vector_norm_35" [label="(4, 12, 64, 32)", style=solid]; +"2746 select_52" -> "2756 expand_as_35" [label="(4, 12, 64, 32)", style=solid]; +"2746 select_52" -> "2757 div_35" [label="(4, 12, 64, 32)", style=solid]; +"2747 select_53" -> "2820 matmul_35" [label="(4, 12, 64, 32)", style=solid]; +"2748 linalg_vector_norm_34" -> "2749 clamp_min_34" [label="(4, 12, 64, 1)", style=solid]; +"2749 clamp_min_34" -> "2750 expand_as_34" [label="(4, 12, 64, 1)", style=solid]; +"2750 expand_as_34" -> "2751 div_34" [label="(4, 12, 64, 32)", style=solid]; +"2751 div_34" -> "2752 quantize_per_tensor_default_106" [label="(4, 12, 64, 32)", style=solid]; +"2752 quantize_per_tensor_default_106" -> "2753 dequantize_per_tensor_default_106" [label=None, style=solid]; +"2753 dequantize_per_tensor_default_106" -> "2761 matmul_34" [label=None, style=solid]; +"2754 linalg_vector_norm_35" -> "2755 clamp_min_35" [label="(4, 12, 64, 1)", style=solid]; +"2755 clamp_min_35" -> "2756 expand_as_35" [label="(4, 12, 64, 1)", style=solid]; +"2756 expand_as_35" -> "2757 div_35" [label="(4, 12, 64, 32)", style=solid]; +"2757 div_35" -> "2758 quantize_per_tensor_default_107" [label="(4, 12, 64, 32)", style=solid]; +"2758 quantize_per_tensor_default_107" -> "2759 dequantize_per_tensor_default_107" [label=None, style=solid]; +"2759 dequantize_per_tensor_default_107" -> "2760 transpose_34" [label=None, style=solid]; +"2760 transpose_34" -> "2761 matmul_34" [label="(4, 12, 32, 64)", style=solid]; +"2761 matmul_34" -> "2764 mul_35" [label="(4, 12, 64, 64)", style=solid]; +"2762 clamp_17" -> "2763 exp_17" [label="(12, 1, 1)", style=solid]; +"2763 exp_17" -> "2764 mul_35" [label="(12, 1, 1)", style=solid]; +"2764 mul_35" -> "2765 add_59" [label="(4, 12, 64, 64)", style=solid]; +"2765 add_59" -> "2813 view_96" [label="(4, 12, 64, 64)", style=solid]; +"2766 new_zeros_8" -> "2768 slice_251" [label="(16, 16)", style=solid]; +"2766 new_zeros_8" -> "2772 slice_253" [label="(16, 16)", style=solid]; +"2766 new_zeros_8" -> "2776 slice_255" [label="(16, 16)", style=solid]; +"2766 new_zeros_8" -> "2780 slice_257" [label="(16, 16)", style=solid]; +"2766 new_zeros_8" -> "2784 slice_259" [label="(16, 16)", style=solid]; +"2766 new_zeros_8" -> "2788 slice_261" [label="(16, 16)", style=solid]; +"2766 new_zeros_8" -> "2792 slice_263" [label="(16, 16)", style=solid]; +"2766 new_zeros_8" -> "2796 slice_265" [label="(16, 16)", style=solid]; +"2766 new_zeros_8" -> "2800 slice_267" [label="(16, 16)", style=solid]; +"2766 new_zeros_8" -> "2803 view_95" [label="(16, 16)", style=solid]; +"2767 lift_fresh_copy_72" -> "2770 fill__72" [label="()", style=solid]; +"2768 slice_251" -> "2769 slice_252" [label="(8, 16)", style=solid]; +"2769 slice_252" -> "2770 fill__72" [label="(8, 8)", style=solid]; +"2771 lift_fresh_copy_73" -> "2774 fill__73" [label="()", style=solid]; +"2772 slice_253" -> "2773 slice_254" [label="(8, 16)", style=solid]; +"2773 slice_254" -> "2774 fill__73" [label="(8, 4)", style=solid]; +"2775 lift_fresh_copy_74" -> "2778 fill__74" [label="()", style=solid]; +"2776 slice_255" -> "2777 slice_256" [label="(8, 16)", style=solid]; +"2777 slice_256" -> "2778 fill__74" [label="(8, 4)", style=solid]; +"2779 lift_fresh_copy_75" -> "2782 fill__75" [label="()", style=solid]; +"2780 slice_257" -> "2781 slice_258" [label="(4, 16)", style=solid]; +"2781 slice_258" -> "2782 fill__75" [label="(4, 8)", style=solid]; +"2783 lift_fresh_copy_76" -> "2786 fill__76" [label="()", style=solid]; +"2784 slice_259" -> "2785 slice_260" [label="(4, 16)", style=solid]; +"2785 slice_260" -> "2786 fill__76" [label="(4, 4)", style=solid]; +"2787 lift_fresh_copy_77" -> "2790 fill__77" [label="()", style=solid]; +"2788 slice_261" -> "2789 slice_262" [label="(4, 16)", style=solid]; +"2789 slice_262" -> "2790 fill__77" [label="(4, 4)", style=solid]; +"2791 lift_fresh_copy_78" -> "2794 fill__78" [label="()", style=solid]; +"2792 slice_263" -> "2793 slice_264" [label="(4, 16)", style=solid]; +"2793 slice_264" -> "2794 fill__78" [label="(4, 8)", style=solid]; +"2795 lift_fresh_copy_79" -> "2798 fill__79" [label="()", style=solid]; +"2796 slice_265" -> "2797 slice_266" [label="(4, 16)", style=solid]; +"2797 slice_266" -> "2798 fill__79" [label="(4, 4)", style=solid]; +"2799 lift_fresh_copy_80" -> "2802 fill__80" [label="()", style=solid]; +"2800 slice_267" -> "2801 slice_268" [label="(4, 16)", style=solid]; +"2801 slice_268" -> "2802 fill__80" [label="(4, 4)", style=solid]; +"2803 view_95" -> "2804 permute_80" [label="(2, 8, 2, 8)", style=solid]; +"2804 permute_80" -> "2805 reshape_78" [label="(2, 2, 8, 8)", style=solid]; +"2805 reshape_78" -> "2806 unsqueeze_50" [label="(4, 64)", style=solid]; +"2805 reshape_78" -> "2807 unsqueeze_51" [label="(4, 64)", style=solid]; +"2806 unsqueeze_50" -> "2808 sub_8" [label="(4, 1, 64)", style=solid]; +"2807 unsqueeze_51" -> "2808 sub_8" [label="(4, 64, 1)", style=solid]; +"2808 sub_8" -> "2809 ne_8" [label="(4, 64, 64)", style=solid]; +"2808 sub_8" -> "2810 masked_fill_16" [label="(4, 64, 64)", style=solid]; +"2808 sub_8" -> "2811 eq_8" [label="(4, 64, 64)", style=solid]; +"2809 ne_8" -> "2810 masked_fill_16" [label="(4, 64, 64)", style=solid]; +"2810 masked_fill_16" -> "2812 masked_fill_17" [label="(4, 64, 64)", style=solid]; +"2811 eq_8" -> "2812 masked_fill_17" [label="(4, 64, 64)", style=solid]; +"2812 masked_fill_17" -> "2814 unsqueeze_52" [label="(4, 64, 64)", style=solid]; +"2813 view_96" -> "2816 add_60" [label="(1, 4, 12, 64, 64)", style=solid]; +"2814 unsqueeze_52" -> "2815 unsqueeze_53" [label="(4, 1, 64, 64)", style=solid]; +"2815 unsqueeze_53" -> "2816 add_60" [label="(1, 4, 1, 64, 64)", style=solid]; +"2816 add_60" -> "2817 view_97" [label="(1, 4, 12, 64, 64)", style=solid]; +"2817 view_97" -> "2818 softmax_17" [label="(4, 12, 64, 64)", style=solid]; +"2818 softmax_17" -> "2819 dropout_68" [label="(4, 12, 64, 64)", style=solid]; +"2819 dropout_68" -> "2820 matmul_35" [label="(4, 12, 64, 64)", style=solid]; +"2820 matmul_35" -> "2821 quantize_per_tensor_default_108" [label="(4, 12, 64, 32)", style=solid]; +"2821 quantize_per_tensor_default_108" -> "2822 dequantize_per_tensor_default_108" [label=None, style=solid]; +"2822 dequantize_per_tensor_default_108" -> "2823 transpose_35" [label=None, style=solid]; +"2823 transpose_35" -> "2824 reshape_79" [label="(4, 64, 12, 32)", style=solid]; +"2824 reshape_79" -> "2825 linear_107" [label="(4, 64, 384)", style=solid]; +"2825 linear_107" -> "2826 dropout_69" [label="(4, 64, 384)", style=solid]; +"2826 dropout_69" -> "2827 view_98" [label="(4, 64, 384)", style=solid]; +"2827 view_98" -> "2828 permute_81" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2828 permute_81" -> "2829 reshape_80" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2829 reshape_80" -> "2830 roll_17" [label="(1, 16, 16, 384)", style=solid]; +"2830 roll_17" -> "2831 slice_269" [label="(1, 16, 16, 384)", style=solid]; +"2831 slice_269" -> "2832 slice_270" [label="(1, 16, 16, 384)", style=solid]; +"2832 slice_270" -> "2833 slice_271" [label="(1, 14, 16, 384)", style=solid]; +"2833 slice_271" -> "2834 slice_272" [label="(1, 14, 14, 384)", style=solid]; +"2834 slice_272" -> "2835 contiguous_33" [label="(1, 14, 14, 384)", style=solid]; +"2835 contiguous_33" -> "2836 layer_norm_37" [label="(1, 14, 14, 384)", style=solid]; +"2836 layer_norm_37" -> "2837 add_61" [label="(1, 14, 14, 384)", style=solid]; +"2837 add_61" -> "2838 quantize_per_tensor_default_109" [label="(1, 14, 14, 384)", style=solid]; +"2837 add_61" -> "2848 add_62" [label="(1, 14, 14, 384)", style=solid]; +"2838 quantize_per_tensor_default_109" -> "2839 dequantize_per_tensor_default_109" [label=None, style=solid]; +"2839 dequantize_per_tensor_default_109" -> "2840 linear_108" [label=None, style=solid]; +"2840 linear_108" -> "2841 gelu_17" [label="(1, 14, 14, 1536)", style=solid]; +"2841 gelu_17" -> "2842 quantize_per_tensor_default_110" [label="(1, 14, 14, 1536)", style=solid]; +"2842 quantize_per_tensor_default_110" -> "2843 dequantize_per_tensor_default_110" [label=None, style=solid]; +"2843 dequantize_per_tensor_default_110" -> "2844 dropout_70" [label=None, style=solid]; +"2844 dropout_70" -> "2845 linear_109" [label="(1, 14, 14, 1536)", style=solid]; +"2845 linear_109" -> "2846 dropout_71" [label="(1, 14, 14, 384)", style=solid]; +"2846 dropout_71" -> "2847 layer_norm_38" [label="(1, 14, 14, 384)", style=solid]; +"2847 layer_norm_38" -> "2848 add_62" [label="(1, 14, 14, 384)", style=solid]; +"2848 add_62" -> "2849 quantize_per_tensor_default_111" [label="(1, 14, 14, 384)", style=solid]; +"2848 add_62" -> "2911 add_64" [label="(1, 14, 14, 384)", style=solid]; +"2849 quantize_per_tensor_default_111" -> "2850 dequantize_per_tensor_default_111" [label=None, style=solid]; +"2850 dequantize_per_tensor_default_111" -> "2862 pad_20" [label=None, style=solid]; +"2851 linear_110" -> "2852 relu__18" [label="(1, 15, 15, 512)", style=solid]; +"2852 relu__18" -> "2853 linear_111" [label="(1, 15, 15, 512)", style=solid]; +"2853 linear_111" -> "2854 view_99" [label="(1, 15, 15, 12)", style=solid]; +"2854 view_99" -> "2855 index_18" [label="(225, 12)", style=solid]; +"2855 index_18" -> "2856 view_100" [label="(4096, 12)", style=solid]; +"2856 view_100" -> "2857 permute_82" [label="(64, 64, 12)", style=solid]; +"2857 permute_82" -> "2858 contiguous_34" [label="(12, 64, 64)", style=solid]; +"2858 contiguous_34" -> "2859 unsqueeze_54" [label="(12, 64, 64)", style=solid]; +"2859 unsqueeze_54" -> "2860 sigmoid_18" [label="(1, 12, 64, 64)", style=solid]; +"2860 sigmoid_18" -> "2861 mul_36" [label="(1, 12, 64, 64)", style=solid]; +"2861 mul_36" -> "2892 add_63" [label="(1, 12, 64, 64)", style=solid]; +"2862 pad_20" -> "2863 view_101" [label="(1, 16, 16, 384)", style=solid]; +"2863 view_101" -> "2864 permute_83" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2864 permute_83" -> "2865 reshape_81" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2865 reshape_81" -> "2869 linear_112" [label="(4, 64, 384)", style=solid]; +"2866 clone_18" -> "2867 slice_273" [label="(1152,)", style=solid]; +"2866 clone_18" -> "2869 linear_112" [label="(1152,)", style=solid]; +"2867 slice_273" -> "2868 zero__18" [label="(384,)", style=solid]; +"2869 linear_112" -> "2870 reshape_82" [label="(4, 64, 1152)", style=solid]; +"2870 reshape_82" -> "2871 permute_84" [label="(4, 64, 3, 12, 32)", style=solid]; +"2871 permute_84" -> "2872 select_54" [label="(3, 4, 12, 64, 32)", style=solid]; +"2871 permute_84" -> "2873 select_55" [label="(3, 4, 12, 64, 32)", style=solid]; +"2871 permute_84" -> "2874 select_56" [label="(3, 4, 12, 64, 32)", style=solid]; +"2872 select_54" -> "2875 linalg_vector_norm_36" [label="(4, 12, 64, 32)", style=solid]; +"2872 select_54" -> "2877 expand_as_36" [label="(4, 12, 64, 32)", style=solid]; +"2872 select_54" -> "2878 div_36" [label="(4, 12, 64, 32)", style=solid]; +"2873 select_55" -> "2881 linalg_vector_norm_37" [label="(4, 12, 64, 32)", style=solid]; +"2873 select_55" -> "2883 expand_as_37" [label="(4, 12, 64, 32)", style=solid]; +"2873 select_55" -> "2884 div_37" [label="(4, 12, 64, 32)", style=solid]; +"2874 select_56" -> "2895 matmul_37" [label="(4, 12, 64, 32)", style=solid]; +"2875 linalg_vector_norm_36" -> "2876 clamp_min_36" [label="(4, 12, 64, 1)", style=solid]; +"2876 clamp_min_36" -> "2877 expand_as_36" [label="(4, 12, 64, 1)", style=solid]; +"2877 expand_as_36" -> "2878 div_36" [label="(4, 12, 64, 32)", style=solid]; +"2878 div_36" -> "2879 quantize_per_tensor_default_112" [label="(4, 12, 64, 32)", style=solid]; +"2879 quantize_per_tensor_default_112" -> "2880 dequantize_per_tensor_default_112" [label=None, style=solid]; +"2880 dequantize_per_tensor_default_112" -> "2888 matmul_36" [label=None, style=solid]; +"2881 linalg_vector_norm_37" -> "2882 clamp_min_37" [label="(4, 12, 64, 1)", style=solid]; +"2882 clamp_min_37" -> "2883 expand_as_37" [label="(4, 12, 64, 1)", style=solid]; +"2883 expand_as_37" -> "2884 div_37" [label="(4, 12, 64, 32)", style=solid]; +"2884 div_37" -> "2885 quantize_per_tensor_default_113" [label="(4, 12, 64, 32)", style=solid]; +"2885 quantize_per_tensor_default_113" -> "2886 dequantize_per_tensor_default_113" [label=None, style=solid]; +"2886 dequantize_per_tensor_default_113" -> "2887 transpose_36" [label=None, style=solid]; +"2887 transpose_36" -> "2888 matmul_36" [label="(4, 12, 32, 64)", style=solid]; +"2888 matmul_36" -> "2891 mul_37" [label="(4, 12, 64, 64)", style=solid]; +"2889 clamp_18" -> "2890 exp_18" [label="(12, 1, 1)", style=solid]; +"2890 exp_18" -> "2891 mul_37" [label="(12, 1, 1)", style=solid]; +"2891 mul_37" -> "2892 add_63" [label="(4, 12, 64, 64)", style=solid]; +"2892 add_63" -> "2893 softmax_18" [label="(4, 12, 64, 64)", style=solid]; +"2893 softmax_18" -> "2894 dropout_72" [label="(4, 12, 64, 64)", style=solid]; +"2894 dropout_72" -> "2895 matmul_37" [label="(4, 12, 64, 64)", style=solid]; +"2895 matmul_37" -> "2896 quantize_per_tensor_default_114" [label="(4, 12, 64, 32)", style=solid]; +"2896 quantize_per_tensor_default_114" -> "2897 dequantize_per_tensor_default_114" [label=None, style=solid]; +"2897 dequantize_per_tensor_default_114" -> "2898 transpose_37" [label=None, style=solid]; +"2898 transpose_37" -> "2899 reshape_83" [label="(4, 64, 12, 32)", style=solid]; +"2899 reshape_83" -> "2900 linear_113" [label="(4, 64, 384)", style=solid]; +"2900 linear_113" -> "2901 dropout_73" [label="(4, 64, 384)", style=solid]; +"2901 dropout_73" -> "2902 view_102" [label="(4, 64, 384)", style=solid]; +"2902 view_102" -> "2903 permute_85" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2903 permute_85" -> "2904 reshape_84" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2904 reshape_84" -> "2905 slice_274" [label="(1, 16, 16, 384)", style=solid]; +"2905 slice_274" -> "2906 slice_275" [label="(1, 16, 16, 384)", style=solid]; +"2906 slice_275" -> "2907 slice_276" [label="(1, 14, 16, 384)", style=solid]; +"2907 slice_276" -> "2908 slice_277" [label="(1, 14, 14, 384)", style=solid]; +"2908 slice_277" -> "2909 contiguous_35" [label="(1, 14, 14, 384)", style=solid]; +"2909 contiguous_35" -> "2910 layer_norm_39" [label="(1, 14, 14, 384)", style=solid]; +"2910 layer_norm_39" -> "2911 add_64" [label="(1, 14, 14, 384)", style=solid]; +"2911 add_64" -> "2912 quantize_per_tensor_default_115" [label="(1, 14, 14, 384)", style=solid]; +"2911 add_64" -> "2922 add_65" [label="(1, 14, 14, 384)", style=solid]; +"2912 quantize_per_tensor_default_115" -> "2913 dequantize_per_tensor_default_115" [label=None, style=solid]; +"2913 dequantize_per_tensor_default_115" -> "2914 linear_114" [label=None, style=solid]; +"2914 linear_114" -> "2915 gelu_18" [label="(1, 14, 14, 1536)", style=solid]; +"2915 gelu_18" -> "2916 quantize_per_tensor_default_116" [label="(1, 14, 14, 1536)", style=solid]; +"2916 quantize_per_tensor_default_116" -> "2917 dequantize_per_tensor_default_116" [label=None, style=solid]; +"2917 dequantize_per_tensor_default_116" -> "2918 dropout_74" [label=None, style=solid]; +"2918 dropout_74" -> "2919 linear_115" [label="(1, 14, 14, 1536)", style=solid]; +"2919 linear_115" -> "2920 dropout_75" [label="(1, 14, 14, 384)", style=solid]; +"2920 dropout_75" -> "2921 layer_norm_40" [label="(1, 14, 14, 384)", style=solid]; +"2921 layer_norm_40" -> "2922 add_65" [label="(1, 14, 14, 384)", style=solid]; +"2922 add_65" -> "2934 pad_21" [label="(1, 14, 14, 384)", style=solid]; +"2922 add_65" -> "3039 add_68" [label="(1, 14, 14, 384)", style=solid]; +"2923 linear_116" -> "2924 relu__19" [label="(1, 15, 15, 512)", style=solid]; +"2924 relu__19" -> "2925 linear_117" [label="(1, 15, 15, 512)", style=solid]; +"2925 linear_117" -> "2926 view_103" [label="(1, 15, 15, 12)", style=solid]; +"2926 view_103" -> "2927 index_19" [label="(225, 12)", style=solid]; +"2927 index_19" -> "2928 view_104" [label="(4096, 12)", style=solid]; +"2928 view_104" -> "2929 permute_86" [label="(64, 64, 12)", style=solid]; +"2929 permute_86" -> "2930 contiguous_36" [label="(12, 64, 64)", style=solid]; +"2930 contiguous_36" -> "2931 unsqueeze_55" [label="(12, 64, 64)", style=solid]; +"2931 unsqueeze_55" -> "2932 sigmoid_19" [label="(1, 12, 64, 64)", style=solid]; +"2932 sigmoid_19" -> "2933 mul_38" [label="(1, 12, 64, 64)", style=solid]; +"2933 mul_38" -> "2967 add_66" [label="(1, 12, 64, 64)", style=solid]; +"2934 pad_21" -> "2935 roll_18" [label="(1, 16, 16, 384)", style=solid]; +"2935 roll_18" -> "2936 view_105" [label="(1, 16, 16, 384)", style=solid]; +"2936 view_105" -> "2937 permute_87" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2937 permute_87" -> "2938 reshape_85" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2938 reshape_85" -> "2939 quantize_per_tensor_default_117" [label="(4, 64, 384)", style=solid]; +"2938 reshape_85" -> "2968 new_zeros_9" [label="(4, 64, 384)", style=solid]; +"2939 quantize_per_tensor_default_117" -> "2940 dequantize_per_tensor_default_117" [label=None, style=solid]; +"2940 dequantize_per_tensor_default_117" -> "2944 linear_118" [label=None, style=solid]; +"2941 clone_19" -> "2942 slice_278" [label="(1152,)", style=solid]; +"2941 clone_19" -> "2944 linear_118" [label="(1152,)", style=solid]; +"2942 slice_278" -> "2943 zero__19" [label="(384,)", style=solid]; +"2944 linear_118" -> "2945 reshape_86" [label="(4, 64, 1152)", style=solid]; +"2945 reshape_86" -> "2946 permute_88" [label="(4, 64, 3, 12, 32)", style=solid]; +"2946 permute_88" -> "2947 select_57" [label="(3, 4, 12, 64, 32)", style=solid]; +"2946 permute_88" -> "2948 select_58" [label="(3, 4, 12, 64, 32)", style=solid]; +"2946 permute_88" -> "2949 select_59" [label="(3, 4, 12, 64, 32)", style=solid]; +"2947 select_57" -> "2950 linalg_vector_norm_38" [label="(4, 12, 64, 32)", style=solid]; +"2947 select_57" -> "2952 expand_as_38" [label="(4, 12, 64, 32)", style=solid]; +"2947 select_57" -> "2953 div_38" [label="(4, 12, 64, 32)", style=solid]; +"2948 select_58" -> "2956 linalg_vector_norm_39" [label="(4, 12, 64, 32)", style=solid]; +"2948 select_58" -> "2958 expand_as_39" [label="(4, 12, 64, 32)", style=solid]; +"2948 select_58" -> "2959 div_39" [label="(4, 12, 64, 32)", style=solid]; +"2949 select_59" -> "3022 matmul_39" [label="(4, 12, 64, 32)", style=solid]; +"2950 linalg_vector_norm_38" -> "2951 clamp_min_38" [label="(4, 12, 64, 1)", style=solid]; +"2951 clamp_min_38" -> "2952 expand_as_38" [label="(4, 12, 64, 1)", style=solid]; +"2952 expand_as_38" -> "2953 div_38" [label="(4, 12, 64, 32)", style=solid]; +"2953 div_38" -> "2954 quantize_per_tensor_default_118" [label="(4, 12, 64, 32)", style=solid]; +"2954 quantize_per_tensor_default_118" -> "2955 dequantize_per_tensor_default_118" [label=None, style=solid]; +"2955 dequantize_per_tensor_default_118" -> "2963 matmul_38" [label=None, style=solid]; +"2956 linalg_vector_norm_39" -> "2957 clamp_min_39" [label="(4, 12, 64, 1)", style=solid]; +"2957 clamp_min_39" -> "2958 expand_as_39" [label="(4, 12, 64, 1)", style=solid]; +"2958 expand_as_39" -> "2959 div_39" [label="(4, 12, 64, 32)", style=solid]; +"2959 div_39" -> "2960 quantize_per_tensor_default_119" [label="(4, 12, 64, 32)", style=solid]; +"2960 quantize_per_tensor_default_119" -> "2961 dequantize_per_tensor_default_119" [label=None, style=solid]; +"2961 dequantize_per_tensor_default_119" -> "2962 transpose_38" [label=None, style=solid]; +"2962 transpose_38" -> "2963 matmul_38" [label="(4, 12, 32, 64)", style=solid]; +"2963 matmul_38" -> "2966 mul_39" [label="(4, 12, 64, 64)", style=solid]; +"2964 clamp_19" -> "2965 exp_19" [label="(12, 1, 1)", style=solid]; +"2965 exp_19" -> "2966 mul_39" [label="(12, 1, 1)", style=solid]; +"2966 mul_39" -> "2967 add_66" [label="(4, 12, 64, 64)", style=solid]; +"2967 add_66" -> "3015 view_107" [label="(4, 12, 64, 64)", style=solid]; +"2968 new_zeros_9" -> "2970 slice_279" [label="(16, 16)", style=solid]; +"2968 new_zeros_9" -> "2974 slice_281" [label="(16, 16)", style=solid]; +"2968 new_zeros_9" -> "2978 slice_283" [label="(16, 16)", style=solid]; +"2968 new_zeros_9" -> "2982 slice_285" [label="(16, 16)", style=solid]; +"2968 new_zeros_9" -> "2986 slice_287" [label="(16, 16)", style=solid]; +"2968 new_zeros_9" -> "2990 slice_289" [label="(16, 16)", style=solid]; +"2968 new_zeros_9" -> "2994 slice_291" [label="(16, 16)", style=solid]; +"2968 new_zeros_9" -> "2998 slice_293" [label="(16, 16)", style=solid]; +"2968 new_zeros_9" -> "3002 slice_295" [label="(16, 16)", style=solid]; +"2968 new_zeros_9" -> "3005 view_106" [label="(16, 16)", style=solid]; +"2969 lift_fresh_copy_81" -> "2972 fill__81" [label="()", style=solid]; +"2970 slice_279" -> "2971 slice_280" [label="(8, 16)", style=solid]; +"2971 slice_280" -> "2972 fill__81" [label="(8, 8)", style=solid]; +"2973 lift_fresh_copy_82" -> "2976 fill__82" [label="()", style=solid]; +"2974 slice_281" -> "2975 slice_282" [label="(8, 16)", style=solid]; +"2975 slice_282" -> "2976 fill__82" [label="(8, 4)", style=solid]; +"2977 lift_fresh_copy_83" -> "2980 fill__83" [label="()", style=solid]; +"2978 slice_283" -> "2979 slice_284" [label="(8, 16)", style=solid]; +"2979 slice_284" -> "2980 fill__83" [label="(8, 4)", style=solid]; +"2981 lift_fresh_copy_84" -> "2984 fill__84" [label="()", style=solid]; +"2982 slice_285" -> "2983 slice_286" [label="(4, 16)", style=solid]; +"2983 slice_286" -> "2984 fill__84" [label="(4, 8)", style=solid]; +"2985 lift_fresh_copy_85" -> "2988 fill__85" [label="()", style=solid]; +"2986 slice_287" -> "2987 slice_288" [label="(4, 16)", style=solid]; +"2987 slice_288" -> "2988 fill__85" [label="(4, 4)", style=solid]; +"2989 lift_fresh_copy_86" -> "2992 fill__86" [label="()", style=solid]; +"2990 slice_289" -> "2991 slice_290" [label="(4, 16)", style=solid]; +"2991 slice_290" -> "2992 fill__86" [label="(4, 4)", style=solid]; +"2993 lift_fresh_copy_87" -> "2996 fill__87" [label="()", style=solid]; +"2994 slice_291" -> "2995 slice_292" [label="(4, 16)", style=solid]; +"2995 slice_292" -> "2996 fill__87" [label="(4, 8)", style=solid]; +"2997 lift_fresh_copy_88" -> "3000 fill__88" [label="()", style=solid]; +"2998 slice_293" -> "2999 slice_294" [label="(4, 16)", style=solid]; +"2999 slice_294" -> "3000 fill__88" [label="(4, 4)", style=solid]; +"3001 lift_fresh_copy_89" -> "3004 fill__89" [label="()", style=solid]; +"3002 slice_295" -> "3003 slice_296" [label="(4, 16)", style=solid]; +"3003 slice_296" -> "3004 fill__89" [label="(4, 4)", style=solid]; +"3005 view_106" -> "3006 permute_89" [label="(2, 8, 2, 8)", style=solid]; +"3006 permute_89" -> "3007 reshape_87" [label="(2, 2, 8, 8)", style=solid]; +"3007 reshape_87" -> "3008 unsqueeze_56" [label="(4, 64)", style=solid]; +"3007 reshape_87" -> "3009 unsqueeze_57" [label="(4, 64)", style=solid]; +"3008 unsqueeze_56" -> "3010 sub_9" [label="(4, 1, 64)", style=solid]; +"3009 unsqueeze_57" -> "3010 sub_9" [label="(4, 64, 1)", style=solid]; +"3010 sub_9" -> "3011 ne_9" [label="(4, 64, 64)", style=solid]; +"3010 sub_9" -> "3012 masked_fill_18" [label="(4, 64, 64)", style=solid]; +"3010 sub_9" -> "3013 eq_9" [label="(4, 64, 64)", style=solid]; +"3011 ne_9" -> "3012 masked_fill_18" [label="(4, 64, 64)", style=solid]; +"3012 masked_fill_18" -> "3014 masked_fill_19" [label="(4, 64, 64)", style=solid]; +"3013 eq_9" -> "3014 masked_fill_19" [label="(4, 64, 64)", style=solid]; +"3014 masked_fill_19" -> "3016 unsqueeze_58" [label="(4, 64, 64)", style=solid]; +"3015 view_107" -> "3018 add_67" [label="(1, 4, 12, 64, 64)", style=solid]; +"3016 unsqueeze_58" -> "3017 unsqueeze_59" [label="(4, 1, 64, 64)", style=solid]; +"3017 unsqueeze_59" -> "3018 add_67" [label="(1, 4, 1, 64, 64)", style=solid]; +"3018 add_67" -> "3019 view_108" [label="(1, 4, 12, 64, 64)", style=solid]; +"3019 view_108" -> "3020 softmax_19" [label="(4, 12, 64, 64)", style=solid]; +"3020 softmax_19" -> "3021 dropout_76" [label="(4, 12, 64, 64)", style=solid]; +"3021 dropout_76" -> "3022 matmul_39" [label="(4, 12, 64, 64)", style=solid]; +"3022 matmul_39" -> "3023 quantize_per_tensor_default_120" [label="(4, 12, 64, 32)", style=solid]; +"3023 quantize_per_tensor_default_120" -> "3024 dequantize_per_tensor_default_120" [label=None, style=solid]; +"3024 dequantize_per_tensor_default_120" -> "3025 transpose_39" [label=None, style=solid]; +"3025 transpose_39" -> "3026 reshape_88" [label="(4, 64, 12, 32)", style=solid]; +"3026 reshape_88" -> "3027 linear_119" [label="(4, 64, 384)", style=solid]; +"3027 linear_119" -> "3028 dropout_77" [label="(4, 64, 384)", style=solid]; +"3028 dropout_77" -> "3029 view_109" [label="(4, 64, 384)", style=solid]; +"3029 view_109" -> "3030 permute_90" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"3030 permute_90" -> "3031 reshape_89" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"3031 reshape_89" -> "3032 roll_19" [label="(1, 16, 16, 384)", style=solid]; +"3032 roll_19" -> "3033 slice_297" [label="(1, 16, 16, 384)", style=solid]; +"3033 slice_297" -> "3034 slice_298" [label="(1, 16, 16, 384)", style=solid]; +"3034 slice_298" -> "3035 slice_299" [label="(1, 14, 16, 384)", style=solid]; +"3035 slice_299" -> "3036 slice_300" [label="(1, 14, 14, 384)", style=solid]; +"3036 slice_300" -> "3037 contiguous_37" [label="(1, 14, 14, 384)", style=solid]; +"3037 contiguous_37" -> "3038 layer_norm_41" [label="(1, 14, 14, 384)", style=solid]; +"3038 layer_norm_41" -> "3039 add_68" [label="(1, 14, 14, 384)", style=solid]; +"3039 add_68" -> "3040 quantize_per_tensor_default_121" [label="(1, 14, 14, 384)", style=solid]; +"3039 add_68" -> "3050 add_69" [label="(1, 14, 14, 384)", style=solid]; +"3040 quantize_per_tensor_default_121" -> "3041 dequantize_per_tensor_default_121" [label=None, style=solid]; +"3041 dequantize_per_tensor_default_121" -> "3042 linear_120" [label=None, style=solid]; +"3042 linear_120" -> "3043 gelu_19" [label="(1, 14, 14, 1536)", style=solid]; +"3043 gelu_19" -> "3044 quantize_per_tensor_default_122" [label="(1, 14, 14, 1536)", style=solid]; +"3044 quantize_per_tensor_default_122" -> "3045 dequantize_per_tensor_default_122" [label=None, style=solid]; +"3045 dequantize_per_tensor_default_122" -> "3046 dropout_78" [label=None, style=solid]; +"3046 dropout_78" -> "3047 linear_121" [label="(1, 14, 14, 1536)", style=solid]; +"3047 linear_121" -> "3048 dropout_79" [label="(1, 14, 14, 384)", style=solid]; +"3048 dropout_79" -> "3049 layer_norm_42" [label="(1, 14, 14, 384)", style=solid]; +"3049 layer_norm_42" -> "3050 add_69" [label="(1, 14, 14, 384)", style=solid]; +"3050 add_69" -> "3051 quantize_per_tensor_default_123" [label="(1, 14, 14, 384)", style=solid]; +"3050 add_69" -> "3113 add_71" [label="(1, 14, 14, 384)", style=solid]; +"3051 quantize_per_tensor_default_123" -> "3052 dequantize_per_tensor_default_123" [label=None, style=solid]; +"3052 dequantize_per_tensor_default_123" -> "3064 pad_22" [label=None, style=solid]; +"3053 linear_122" -> "3054 relu__20" [label="(1, 15, 15, 512)", style=solid]; +"3054 relu__20" -> "3055 linear_123" [label="(1, 15, 15, 512)", style=solid]; +"3055 linear_123" -> "3056 view_110" [label="(1, 15, 15, 12)", style=solid]; +"3056 view_110" -> "3057 index_20" [label="(225, 12)", style=solid]; +"3057 index_20" -> "3058 view_111" [label="(4096, 12)", style=solid]; +"3058 view_111" -> "3059 permute_91" [label="(64, 64, 12)", style=solid]; +"3059 permute_91" -> "3060 contiguous_38" [label="(12, 64, 64)", style=solid]; +"3060 contiguous_38" -> "3061 unsqueeze_60" [label="(12, 64, 64)", style=solid]; +"3061 unsqueeze_60" -> "3062 sigmoid_20" [label="(1, 12, 64, 64)", style=solid]; +"3062 sigmoid_20" -> "3063 mul_40" [label="(1, 12, 64, 64)", style=solid]; +"3063 mul_40" -> "3094 add_70" [label="(1, 12, 64, 64)", style=solid]; +"3064 pad_22" -> "3065 view_112" [label="(1, 16, 16, 384)", style=solid]; +"3065 view_112" -> "3066 permute_92" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"3066 permute_92" -> "3067 reshape_90" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"3067 reshape_90" -> "3071 linear_124" [label="(4, 64, 384)", style=solid]; +"3068 clone_20" -> "3069 slice_301" [label="(1152,)", style=solid]; +"3068 clone_20" -> "3071 linear_124" [label="(1152,)", style=solid]; +"3069 slice_301" -> "3070 zero__20" [label="(384,)", style=solid]; +"3071 linear_124" -> "3072 reshape_91" [label="(4, 64, 1152)", style=solid]; +"3072 reshape_91" -> "3073 permute_93" [label="(4, 64, 3, 12, 32)", style=solid]; +"3073 permute_93" -> "3074 select_60" [label="(3, 4, 12, 64, 32)", style=solid]; +"3073 permute_93" -> "3075 select_61" [label="(3, 4, 12, 64, 32)", style=solid]; +"3073 permute_93" -> "3076 select_62" [label="(3, 4, 12, 64, 32)", style=solid]; +"3074 select_60" -> "3077 linalg_vector_norm_40" [label="(4, 12, 64, 32)", style=solid]; +"3074 select_60" -> "3079 expand_as_40" [label="(4, 12, 64, 32)", style=solid]; +"3074 select_60" -> "3080 div_40" [label="(4, 12, 64, 32)", style=solid]; +"3075 select_61" -> "3083 linalg_vector_norm_41" [label="(4, 12, 64, 32)", style=solid]; +"3075 select_61" -> "3085 expand_as_41" [label="(4, 12, 64, 32)", style=solid]; +"3075 select_61" -> "3086 div_41" [label="(4, 12, 64, 32)", style=solid]; +"3076 select_62" -> "3097 matmul_41" [label="(4, 12, 64, 32)", style=solid]; +"3077 linalg_vector_norm_40" -> "3078 clamp_min_40" [label="(4, 12, 64, 1)", style=solid]; +"3078 clamp_min_40" -> "3079 expand_as_40" [label="(4, 12, 64, 1)", style=solid]; +"3079 expand_as_40" -> "3080 div_40" [label="(4, 12, 64, 32)", style=solid]; +"3080 div_40" -> "3081 quantize_per_tensor_default_124" [label="(4, 12, 64, 32)", style=solid]; +"3081 quantize_per_tensor_default_124" -> "3082 dequantize_per_tensor_default_124" [label=None, style=solid]; +"3082 dequantize_per_tensor_default_124" -> "3090 matmul_40" [label=None, style=solid]; +"3083 linalg_vector_norm_41" -> "3084 clamp_min_41" [label="(4, 12, 64, 1)", style=solid]; +"3084 clamp_min_41" -> "3085 expand_as_41" [label="(4, 12, 64, 1)", style=solid]; +"3085 expand_as_41" -> "3086 div_41" [label="(4, 12, 64, 32)", style=solid]; +"3086 div_41" -> "3087 quantize_per_tensor_default_125" [label="(4, 12, 64, 32)", style=solid]; +"3087 quantize_per_tensor_default_125" -> "3088 dequantize_per_tensor_default_125" [label=None, style=solid]; +"3088 dequantize_per_tensor_default_125" -> "3089 transpose_40" [label=None, style=solid]; +"3089 transpose_40" -> "3090 matmul_40" [label="(4, 12, 32, 64)", style=solid]; +"3090 matmul_40" -> "3093 mul_41" [label="(4, 12, 64, 64)", style=solid]; +"3091 clamp_20" -> "3092 exp_20" [label="(12, 1, 1)", style=solid]; +"3092 exp_20" -> "3093 mul_41" [label="(12, 1, 1)", style=solid]; +"3093 mul_41" -> "3094 add_70" [label="(4, 12, 64, 64)", style=solid]; +"3094 add_70" -> "3095 softmax_20" [label="(4, 12, 64, 64)", style=solid]; +"3095 softmax_20" -> "3096 dropout_80" [label="(4, 12, 64, 64)", style=solid]; +"3096 dropout_80" -> "3097 matmul_41" [label="(4, 12, 64, 64)", style=solid]; +"3097 matmul_41" -> "3098 quantize_per_tensor_default_126" [label="(4, 12, 64, 32)", style=solid]; +"3098 quantize_per_tensor_default_126" -> "3099 dequantize_per_tensor_default_126" [label=None, style=solid]; +"3099 dequantize_per_tensor_default_126" -> "3100 transpose_41" [label=None, style=solid]; +"3100 transpose_41" -> "3101 reshape_92" [label="(4, 64, 12, 32)", style=solid]; +"3101 reshape_92" -> "3102 linear_125" [label="(4, 64, 384)", style=solid]; +"3102 linear_125" -> "3103 dropout_81" [label="(4, 64, 384)", style=solid]; +"3103 dropout_81" -> "3104 view_113" [label="(4, 64, 384)", style=solid]; +"3104 view_113" -> "3105 permute_94" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"3105 permute_94" -> "3106 reshape_93" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"3106 reshape_93" -> "3107 slice_302" [label="(1, 16, 16, 384)", style=solid]; +"3107 slice_302" -> "3108 slice_303" [label="(1, 16, 16, 384)", style=solid]; +"3108 slice_303" -> "3109 slice_304" [label="(1, 14, 16, 384)", style=solid]; +"3109 slice_304" -> "3110 slice_305" [label="(1, 14, 14, 384)", style=solid]; +"3110 slice_305" -> "3111 contiguous_39" [label="(1, 14, 14, 384)", style=solid]; +"3111 contiguous_39" -> "3112 layer_norm_43" [label="(1, 14, 14, 384)", style=solid]; +"3112 layer_norm_43" -> "3113 add_71" [label="(1, 14, 14, 384)", style=solid]; +"3113 add_71" -> "3114 quantize_per_tensor_default_127" [label="(1, 14, 14, 384)", style=solid]; +"3113 add_71" -> "3124 add_72" [label="(1, 14, 14, 384)", style=solid]; +"3114 quantize_per_tensor_default_127" -> "3115 dequantize_per_tensor_default_127" [label=None, style=solid]; +"3115 dequantize_per_tensor_default_127" -> "3116 linear_126" [label=None, style=solid]; +"3116 linear_126" -> "3117 gelu_20" [label="(1, 14, 14, 1536)", style=solid]; +"3117 gelu_20" -> "3118 quantize_per_tensor_default_128" [label="(1, 14, 14, 1536)", style=solid]; +"3118 quantize_per_tensor_default_128" -> "3119 dequantize_per_tensor_default_128" [label=None, style=solid]; +"3119 dequantize_per_tensor_default_128" -> "3120 dropout_82" [label=None, style=solid]; +"3120 dropout_82" -> "3121 linear_127" [label="(1, 14, 14, 1536)", style=solid]; +"3121 linear_127" -> "3122 dropout_83" [label="(1, 14, 14, 384)", style=solid]; +"3122 dropout_83" -> "3123 layer_norm_44" [label="(1, 14, 14, 384)", style=solid]; +"3123 layer_norm_44" -> "3124 add_72" [label="(1, 14, 14, 384)", style=solid]; +"3124 add_72" -> "3136 pad_23" [label="(1, 14, 14, 384)", style=solid]; +"3124 add_72" -> "3241 add_75" [label="(1, 14, 14, 384)", style=solid]; +"3125 linear_128" -> "3126 relu__21" [label="(1, 15, 15, 512)", style=solid]; +"3126 relu__21" -> "3127 linear_129" [label="(1, 15, 15, 512)", style=solid]; +"3127 linear_129" -> "3128 view_114" [label="(1, 15, 15, 12)", style=solid]; +"3128 view_114" -> "3129 index_21" [label="(225, 12)", style=solid]; +"3129 index_21" -> "3130 view_115" [label="(4096, 12)", style=solid]; +"3130 view_115" -> "3131 permute_95" [label="(64, 64, 12)", style=solid]; +"3131 permute_95" -> "3132 contiguous_40" [label="(12, 64, 64)", style=solid]; +"3132 contiguous_40" -> "3133 unsqueeze_61" [label="(12, 64, 64)", style=solid]; +"3133 unsqueeze_61" -> "3134 sigmoid_21" [label="(1, 12, 64, 64)", style=solid]; +"3134 sigmoid_21" -> "3135 mul_42" [label="(1, 12, 64, 64)", style=solid]; +"3135 mul_42" -> "3169 add_73" [label="(1, 12, 64, 64)", style=solid]; +"3136 pad_23" -> "3137 roll_20" [label="(1, 16, 16, 384)", style=solid]; +"3137 roll_20" -> "3138 view_116" [label="(1, 16, 16, 384)", style=solid]; +"3138 view_116" -> "3139 permute_96" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"3139 permute_96" -> "3140 reshape_94" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"3140 reshape_94" -> "3141 quantize_per_tensor_default_129" [label="(4, 64, 384)", style=solid]; +"3140 reshape_94" -> "3170 new_zeros_10" [label="(4, 64, 384)", style=solid]; +"3141 quantize_per_tensor_default_129" -> "3142 dequantize_per_tensor_default_129" [label=None, style=solid]; +"3142 dequantize_per_tensor_default_129" -> "3146 linear_130" [label=None, style=solid]; +"3143 clone_21" -> "3144 slice_306" [label="(1152,)", style=solid]; +"3143 clone_21" -> "3146 linear_130" [label="(1152,)", style=solid]; +"3144 slice_306" -> "3145 zero__21" [label="(384,)", style=solid]; +"3146 linear_130" -> "3147 reshape_95" [label="(4, 64, 1152)", style=solid]; +"3147 reshape_95" -> "3148 permute_97" [label="(4, 64, 3, 12, 32)", style=solid]; +"3148 permute_97" -> "3149 select_63" [label="(3, 4, 12, 64, 32)", style=solid]; +"3148 permute_97" -> "3150 select_64" [label="(3, 4, 12, 64, 32)", style=solid]; +"3148 permute_97" -> "3151 select_65" [label="(3, 4, 12, 64, 32)", style=solid]; +"3149 select_63" -> "3152 linalg_vector_norm_42" [label="(4, 12, 64, 32)", style=solid]; +"3149 select_63" -> "3154 expand_as_42" [label="(4, 12, 64, 32)", style=solid]; +"3149 select_63" -> "3155 div_42" [label="(4, 12, 64, 32)", style=solid]; +"3150 select_64" -> "3158 linalg_vector_norm_43" [label="(4, 12, 64, 32)", style=solid]; +"3150 select_64" -> "3160 expand_as_43" [label="(4, 12, 64, 32)", style=solid]; +"3150 select_64" -> "3161 div_43" [label="(4, 12, 64, 32)", style=solid]; +"3151 select_65" -> "3224 matmul_43" [label="(4, 12, 64, 32)", style=solid]; +"3152 linalg_vector_norm_42" -> "3153 clamp_min_42" [label="(4, 12, 64, 1)", style=solid]; +"3153 clamp_min_42" -> "3154 expand_as_42" [label="(4, 12, 64, 1)", style=solid]; +"3154 expand_as_42" -> "3155 div_42" [label="(4, 12, 64, 32)", style=solid]; +"3155 div_42" -> "3156 quantize_per_tensor_default_130" [label="(4, 12, 64, 32)", style=solid]; +"3156 quantize_per_tensor_default_130" -> "3157 dequantize_per_tensor_default_130" [label=None, style=solid]; +"3157 dequantize_per_tensor_default_130" -> "3165 matmul_42" [label=None, style=solid]; +"3158 linalg_vector_norm_43" -> "3159 clamp_min_43" [label="(4, 12, 64, 1)", style=solid]; +"3159 clamp_min_43" -> "3160 expand_as_43" [label="(4, 12, 64, 1)", style=solid]; +"3160 expand_as_43" -> "3161 div_43" [label="(4, 12, 64, 32)", style=solid]; +"3161 div_43" -> "3162 quantize_per_tensor_default_131" [label="(4, 12, 64, 32)", style=solid]; +"3162 quantize_per_tensor_default_131" -> "3163 dequantize_per_tensor_default_131" [label=None, style=solid]; +"3163 dequantize_per_tensor_default_131" -> "3164 transpose_42" [label=None, style=solid]; +"3164 transpose_42" -> "3165 matmul_42" [label="(4, 12, 32, 64)", style=solid]; +"3165 matmul_42" -> "3168 mul_43" [label="(4, 12, 64, 64)", style=solid]; +"3166 clamp_21" -> "3167 exp_21" [label="(12, 1, 1)", style=solid]; +"3167 exp_21" -> "3168 mul_43" [label="(12, 1, 1)", style=solid]; +"3168 mul_43" -> "3169 add_73" [label="(4, 12, 64, 64)", style=solid]; +"3169 add_73" -> "3217 view_118" [label="(4, 12, 64, 64)", style=solid]; +"3170 new_zeros_10" -> "3172 slice_307" [label="(16, 16)", style=solid]; +"3170 new_zeros_10" -> "3176 slice_309" [label="(16, 16)", style=solid]; +"3170 new_zeros_10" -> "3180 slice_311" [label="(16, 16)", style=solid]; +"3170 new_zeros_10" -> "3184 slice_313" [label="(16, 16)", style=solid]; +"3170 new_zeros_10" -> "3188 slice_315" [label="(16, 16)", style=solid]; +"3170 new_zeros_10" -> "3192 slice_317" [label="(16, 16)", style=solid]; +"3170 new_zeros_10" -> "3196 slice_319" [label="(16, 16)", style=solid]; +"3170 new_zeros_10" -> "3200 slice_321" [label="(16, 16)", style=solid]; +"3170 new_zeros_10" -> "3204 slice_323" [label="(16, 16)", style=solid]; +"3170 new_zeros_10" -> "3207 view_117" [label="(16, 16)", style=solid]; +"3171 lift_fresh_copy_90" -> "3174 fill__90" [label="()", style=solid]; +"3172 slice_307" -> "3173 slice_308" [label="(8, 16)", style=solid]; +"3173 slice_308" -> "3174 fill__90" [label="(8, 8)", style=solid]; +"3175 lift_fresh_copy_91" -> "3178 fill__91" [label="()", style=solid]; +"3176 slice_309" -> "3177 slice_310" [label="(8, 16)", style=solid]; +"3177 slice_310" -> "3178 fill__91" [label="(8, 4)", style=solid]; +"3179 lift_fresh_copy_92" -> "3182 fill__92" [label="()", style=solid]; +"3180 slice_311" -> "3181 slice_312" [label="(8, 16)", style=solid]; +"3181 slice_312" -> "3182 fill__92" [label="(8, 4)", style=solid]; +"3183 lift_fresh_copy_93" -> "3186 fill__93" [label="()", style=solid]; +"3184 slice_313" -> "3185 slice_314" [label="(4, 16)", style=solid]; +"3185 slice_314" -> "3186 fill__93" [label="(4, 8)", style=solid]; +"3187 lift_fresh_copy_94" -> "3190 fill__94" [label="()", style=solid]; +"3188 slice_315" -> "3189 slice_316" [label="(4, 16)", style=solid]; +"3189 slice_316" -> "3190 fill__94" [label="(4, 4)", style=solid]; +"3191 lift_fresh_copy_95" -> "3194 fill__95" [label="()", style=solid]; +"3192 slice_317" -> "3193 slice_318" [label="(4, 16)", style=solid]; +"3193 slice_318" -> "3194 fill__95" [label="(4, 4)", style=solid]; +"3195 lift_fresh_copy_96" -> "3198 fill__96" [label="()", style=solid]; +"3196 slice_319" -> "3197 slice_320" [label="(4, 16)", style=solid]; +"3197 slice_320" -> "3198 fill__96" [label="(4, 8)", style=solid]; +"3199 lift_fresh_copy_97" -> "3202 fill__97" [label="()", style=solid]; +"3200 slice_321" -> "3201 slice_322" [label="(4, 16)", style=solid]; +"3201 slice_322" -> "3202 fill__97" [label="(4, 4)", style=solid]; +"3203 lift_fresh_copy_98" -> "3206 fill__98" [label="()", style=solid]; +"3204 slice_323" -> "3205 slice_324" [label="(4, 16)", style=solid]; +"3205 slice_324" -> "3206 fill__98" [label="(4, 4)", style=solid]; +"3207 view_117" -> "3208 permute_98" [label="(2, 8, 2, 8)", style=solid]; +"3208 permute_98" -> "3209 reshape_96" [label="(2, 2, 8, 8)", style=solid]; +"3209 reshape_96" -> "3210 unsqueeze_62" [label="(4, 64)", style=solid]; +"3209 reshape_96" -> "3211 unsqueeze_63" [label="(4, 64)", style=solid]; +"3210 unsqueeze_62" -> "3212 sub_10" [label="(4, 1, 64)", style=solid]; +"3211 unsqueeze_63" -> "3212 sub_10" [label="(4, 64, 1)", style=solid]; +"3212 sub_10" -> "3213 ne_10" [label="(4, 64, 64)", style=solid]; +"3212 sub_10" -> "3214 masked_fill_20" [label="(4, 64, 64)", style=solid]; +"3212 sub_10" -> "3215 eq_10" [label="(4, 64, 64)", style=solid]; +"3213 ne_10" -> "3214 masked_fill_20" [label="(4, 64, 64)", style=solid]; +"3214 masked_fill_20" -> "3216 masked_fill_21" [label="(4, 64, 64)", style=solid]; +"3215 eq_10" -> "3216 masked_fill_21" [label="(4, 64, 64)", style=solid]; +"3216 masked_fill_21" -> "3218 unsqueeze_64" [label="(4, 64, 64)", style=solid]; +"3217 view_118" -> "3220 add_74" [label="(1, 4, 12, 64, 64)", style=solid]; +"3218 unsqueeze_64" -> "3219 unsqueeze_65" [label="(4, 1, 64, 64)", style=solid]; +"3219 unsqueeze_65" -> "3220 add_74" [label="(1, 4, 1, 64, 64)", style=solid]; +"3220 add_74" -> "3221 view_119" [label="(1, 4, 12, 64, 64)", style=solid]; +"3221 view_119" -> "3222 softmax_21" [label="(4, 12, 64, 64)", style=solid]; +"3222 softmax_21" -> "3223 dropout_84" [label="(4, 12, 64, 64)", style=solid]; +"3223 dropout_84" -> "3224 matmul_43" [label="(4, 12, 64, 64)", style=solid]; +"3224 matmul_43" -> "3225 quantize_per_tensor_default_132" [label="(4, 12, 64, 32)", style=solid]; +"3225 quantize_per_tensor_default_132" -> "3226 dequantize_per_tensor_default_132" [label=None, style=solid]; +"3226 dequantize_per_tensor_default_132" -> "3227 transpose_43" [label=None, style=solid]; +"3227 transpose_43" -> "3228 reshape_97" [label="(4, 64, 12, 32)", style=solid]; +"3228 reshape_97" -> "3229 linear_131" [label="(4, 64, 384)", style=solid]; +"3229 linear_131" -> "3230 dropout_85" [label="(4, 64, 384)", style=solid]; +"3230 dropout_85" -> "3231 view_120" [label="(4, 64, 384)", style=solid]; +"3231 view_120" -> "3232 permute_99" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"3232 permute_99" -> "3233 reshape_98" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"3233 reshape_98" -> "3234 roll_21" [label="(1, 16, 16, 384)", style=solid]; +"3234 roll_21" -> "3235 slice_325" [label="(1, 16, 16, 384)", style=solid]; +"3235 slice_325" -> "3236 slice_326" [label="(1, 16, 16, 384)", style=solid]; +"3236 slice_326" -> "3237 slice_327" [label="(1, 14, 16, 384)", style=solid]; +"3237 slice_327" -> "3238 slice_328" [label="(1, 14, 14, 384)", style=solid]; +"3238 slice_328" -> "3239 contiguous_41" [label="(1, 14, 14, 384)", style=solid]; +"3239 contiguous_41" -> "3240 layer_norm_45" [label="(1, 14, 14, 384)", style=solid]; +"3240 layer_norm_45" -> "3241 add_75" [label="(1, 14, 14, 384)", style=solid]; +"3241 add_75" -> "3242 quantize_per_tensor_default_133" [label="(1, 14, 14, 384)", style=solid]; +"3241 add_75" -> "3252 add_76" [label="(1, 14, 14, 384)", style=solid]; +"3242 quantize_per_tensor_default_133" -> "3243 dequantize_per_tensor_default_133" [label=None, style=solid]; +"3243 dequantize_per_tensor_default_133" -> "3244 linear_132" [label=None, style=solid]; +"3244 linear_132" -> "3245 gelu_21" [label="(1, 14, 14, 1536)", style=solid]; +"3245 gelu_21" -> "3246 quantize_per_tensor_default_134" [label="(1, 14, 14, 1536)", style=solid]; +"3246 quantize_per_tensor_default_134" -> "3247 dequantize_per_tensor_default_134" [label=None, style=solid]; +"3247 dequantize_per_tensor_default_134" -> "3248 dropout_86" [label=None, style=solid]; +"3248 dropout_86" -> "3249 linear_133" [label="(1, 14, 14, 1536)", style=solid]; +"3249 linear_133" -> "3250 dropout_87" [label="(1, 14, 14, 384)", style=solid]; +"3250 dropout_87" -> "3251 layer_norm_46" [label="(1, 14, 14, 384)", style=solid]; +"3251 layer_norm_46" -> "3252 add_76" [label="(1, 14, 14, 384)", style=solid]; +"3252 add_76" -> "3253 quantize_per_tensor_default_135" [label="(1, 14, 14, 384)", style=solid]; +"3253 quantize_per_tensor_default_135" -> "3254 dequantize_per_tensor_default_135" [label=None, style=solid]; +"3254 dequantize_per_tensor_default_135" -> "3255 pad_24" [label=None, style=solid]; +"3255 pad_24" -> "3256 slice_329" [label="(1, 14, 14, 384)", style=solid]; +"3255 pad_24" -> "3259 slice_332" [label="(1, 14, 14, 384)", style=solid]; +"3255 pad_24" -> "3262 slice_335" [label="(1, 14, 14, 384)", style=solid]; +"3255 pad_24" -> "3265 slice_338" [label="(1, 14, 14, 384)", style=solid]; +"3256 slice_329" -> "3257 slice_330" [label="(1, 7, 14, 384)", style=solid]; +"3257 slice_330" -> "3258 slice_331" [label="(1, 7, 7, 384)", style=solid]; +"3258 slice_331" -> "3268 cat_2" [label="(1, 7, 7, 384)", style=solid]; +"3259 slice_332" -> "3260 slice_333" [label="(1, 7, 14, 384)", style=solid]; +"3260 slice_333" -> "3261 slice_334" [label="(1, 7, 7, 384)", style=solid]; +"3261 slice_334" -> "3268 cat_2" [label="(1, 7, 7, 384)", style=solid]; +"3262 slice_335" -> "3263 slice_336" [label="(1, 7, 14, 384)", style=solid]; +"3263 slice_336" -> "3264 slice_337" [label="(1, 7, 7, 384)", style=solid]; +"3264 slice_337" -> "3268 cat_2" [label="(1, 7, 7, 384)", style=solid]; +"3265 slice_338" -> "3266 slice_339" [label="(1, 7, 14, 384)", style=solid]; +"3266 slice_339" -> "3267 slice_340" [label="(1, 7, 7, 384)", style=solid]; +"3267 slice_340" -> "3268 cat_2" [label="(1, 7, 7, 384)", style=solid]; +"3268 cat_2" -> "3269 linear_134" [label="(1, 7, 7, 1536)", style=solid]; +"3269 linear_134" -> "3270 layer_norm_47" [label="(1, 7, 7, 768)", style=solid]; +"3270 layer_norm_47" -> "3271 quantize_per_tensor_default_136" [label="(1, 7, 7, 768)", style=solid]; +"3270 layer_norm_47" -> "3333 add_78" [label="(1, 7, 7, 768)", style=solid]; +"3271 quantize_per_tensor_default_136" -> "3272 dequantize_per_tensor_default_136" [label=None, style=solid]; +"3272 dequantize_per_tensor_default_136" -> "3284 pad_25" [label=None, style=solid]; +"3273 linear_135" -> "3274 relu__22" [label="(1, 15, 15, 512)", style=solid]; +"3274 relu__22" -> "3275 linear_136" [label="(1, 15, 15, 512)", style=solid]; +"3275 linear_136" -> "3276 view_121" [label="(1, 15, 15, 24)", style=solid]; +"3276 view_121" -> "3277 index_22" [label="(225, 24)", style=solid]; +"3277 index_22" -> "3278 view_122" [label="(4096, 24)", style=solid]; +"3278 view_122" -> "3279 permute_100" [label="(64, 64, 24)", style=solid]; +"3279 permute_100" -> "3280 contiguous_42" [label="(24, 64, 64)", style=solid]; +"3280 contiguous_42" -> "3281 unsqueeze_66" [label="(24, 64, 64)", style=solid]; +"3281 unsqueeze_66" -> "3282 sigmoid_22" [label="(1, 24, 64, 64)", style=solid]; +"3282 sigmoid_22" -> "3283 mul_44" [label="(1, 24, 64, 64)", style=solid]; +"3283 mul_44" -> "3314 add_77" [label="(1, 24, 64, 64)", style=solid]; +"3284 pad_25" -> "3285 view_123" [label="(1, 8, 8, 768)", style=solid]; +"3285 view_123" -> "3286 permute_101" [label="(1, 1, 8, 1, 8, 768)", style=solid]; +"3286 permute_101" -> "3287 reshape_99" [label="(1, 1, 1, 8, 8, 768)", style=solid]; +"3287 reshape_99" -> "3291 linear_137" [label="(1, 64, 768)", style=solid]; +"3288 clone_22" -> "3289 slice_341" [label="(2304,)", style=solid]; +"3288 clone_22" -> "3291 linear_137" [label="(2304,)", style=solid]; +"3289 slice_341" -> "3290 zero__22" [label="(768,)", style=solid]; +"3291 linear_137" -> "3292 reshape_100" [label="(1, 64, 2304)", style=solid]; +"3292 reshape_100" -> "3293 permute_102" [label="(1, 64, 3, 24, 32)", style=solid]; +"3293 permute_102" -> "3294 select_66" [label="(3, 1, 24, 64, 32)", style=solid]; +"3293 permute_102" -> "3295 select_67" [label="(3, 1, 24, 64, 32)", style=solid]; +"3293 permute_102" -> "3296 select_68" [label="(3, 1, 24, 64, 32)", style=solid]; +"3294 select_66" -> "3297 linalg_vector_norm_44" [label="(1, 24, 64, 32)", style=solid]; +"3294 select_66" -> "3299 expand_as_44" [label="(1, 24, 64, 32)", style=solid]; +"3294 select_66" -> "3300 div_44" [label="(1, 24, 64, 32)", style=solid]; +"3295 select_67" -> "3303 linalg_vector_norm_45" [label="(1, 24, 64, 32)", style=solid]; +"3295 select_67" -> "3305 expand_as_45" [label="(1, 24, 64, 32)", style=solid]; +"3295 select_67" -> "3306 div_45" [label="(1, 24, 64, 32)", style=solid]; +"3296 select_68" -> "3317 matmul_45" [label="(1, 24, 64, 32)", style=solid]; +"3297 linalg_vector_norm_44" -> "3298 clamp_min_44" [label="(1, 24, 64, 1)", style=solid]; +"3298 clamp_min_44" -> "3299 expand_as_44" [label="(1, 24, 64, 1)", style=solid]; +"3299 expand_as_44" -> "3300 div_44" [label="(1, 24, 64, 32)", style=solid]; +"3300 div_44" -> "3301 quantize_per_tensor_default_137" [label="(1, 24, 64, 32)", style=solid]; +"3301 quantize_per_tensor_default_137" -> "3302 dequantize_per_tensor_default_137" [label=None, style=solid]; +"3302 dequantize_per_tensor_default_137" -> "3310 matmul_44" [label=None, style=solid]; +"3303 linalg_vector_norm_45" -> "3304 clamp_min_45" [label="(1, 24, 64, 1)", style=solid]; +"3304 clamp_min_45" -> "3305 expand_as_45" [label="(1, 24, 64, 1)", style=solid]; +"3305 expand_as_45" -> "3306 div_45" [label="(1, 24, 64, 32)", style=solid]; +"3306 div_45" -> "3307 quantize_per_tensor_default_138" [label="(1, 24, 64, 32)", style=solid]; +"3307 quantize_per_tensor_default_138" -> "3308 dequantize_per_tensor_default_138" [label=None, style=solid]; +"3308 dequantize_per_tensor_default_138" -> "3309 transpose_44" [label=None, style=solid]; +"3309 transpose_44" -> "3310 matmul_44" [label="(1, 24, 32, 64)", style=solid]; +"3310 matmul_44" -> "3313 mul_45" [label="(1, 24, 64, 64)", style=solid]; +"3311 clamp_22" -> "3312 exp_22" [label="(24, 1, 1)", style=solid]; +"3312 exp_22" -> "3313 mul_45" [label="(24, 1, 1)", style=solid]; +"3313 mul_45" -> "3314 add_77" [label="(1, 24, 64, 64)", style=solid]; +"3314 add_77" -> "3315 softmax_22" [label="(1, 24, 64, 64)", style=solid]; +"3315 softmax_22" -> "3316 dropout_88" [label="(1, 24, 64, 64)", style=solid]; +"3316 dropout_88" -> "3317 matmul_45" [label="(1, 24, 64, 64)", style=solid]; +"3317 matmul_45" -> "3318 quantize_per_tensor_default_139" [label="(1, 24, 64, 32)", style=solid]; +"3318 quantize_per_tensor_default_139" -> "3319 dequantize_per_tensor_default_139" [label=None, style=solid]; +"3319 dequantize_per_tensor_default_139" -> "3320 transpose_45" [label=None, style=solid]; +"3320 transpose_45" -> "3321 reshape_101" [label="(1, 64, 24, 32)", style=solid]; +"3321 reshape_101" -> "3322 linear_138" [label="(1, 64, 768)", style=solid]; +"3322 linear_138" -> "3323 dropout_89" [label="(1, 64, 768)", style=solid]; +"3323 dropout_89" -> "3324 view_124" [label="(1, 64, 768)", style=solid]; +"3324 view_124" -> "3325 permute_103" [label="(1, 1, 1, 8, 8, 768)", style=solid]; +"3325 permute_103" -> "3326 reshape_102" [label="(1, 1, 8, 1, 8, 768)", style=solid]; +"3326 reshape_102" -> "3327 slice_342" [label="(1, 8, 8, 768)", style=solid]; +"3327 slice_342" -> "3328 slice_343" [label="(1, 8, 8, 768)", style=solid]; +"3328 slice_343" -> "3329 slice_344" [label="(1, 7, 8, 768)", style=solid]; +"3329 slice_344" -> "3330 slice_345" [label="(1, 7, 7, 768)", style=solid]; +"3330 slice_345" -> "3331 contiguous_43" [label="(1, 7, 7, 768)", style=solid]; +"3331 contiguous_43" -> "3332 layer_norm_48" [label="(1, 7, 7, 768)", style=solid]; +"3332 layer_norm_48" -> "3333 add_78" [label="(1, 7, 7, 768)", style=solid]; +"3333 add_78" -> "3334 quantize_per_tensor_default_140" [label="(1, 7, 7, 768)", style=solid]; +"3333 add_78" -> "3344 add_79" [label="(1, 7, 7, 768)", style=solid]; +"3334 quantize_per_tensor_default_140" -> "3335 dequantize_per_tensor_default_140" [label=None, style=solid]; +"3335 dequantize_per_tensor_default_140" -> "3336 linear_139" [label=None, style=solid]; +"3336 linear_139" -> "3337 gelu_22" [label="(1, 7, 7, 3072)", style=solid]; +"3337 gelu_22" -> "3338 quantize_per_tensor_default_141" [label="(1, 7, 7, 3072)", style=solid]; +"3338 quantize_per_tensor_default_141" -> "3339 dequantize_per_tensor_default_141" [label=None, style=solid]; +"3339 dequantize_per_tensor_default_141" -> "3340 dropout_90" [label=None, style=solid]; +"3340 dropout_90" -> "3341 linear_140" [label="(1, 7, 7, 3072)", style=solid]; +"3341 linear_140" -> "3342 dropout_91" [label="(1, 7, 7, 768)", style=solid]; +"3342 dropout_91" -> "3343 layer_norm_49" [label="(1, 7, 7, 768)", style=solid]; +"3343 layer_norm_49" -> "3344 add_79" [label="(1, 7, 7, 768)", style=solid]; +"3344 add_79" -> "3345 quantize_per_tensor_default_142" [label="(1, 7, 7, 768)", style=solid]; +"3344 add_79" -> "3407 add_81" [label="(1, 7, 7, 768)", style=solid]; +"3345 quantize_per_tensor_default_142" -> "3346 dequantize_per_tensor_default_142" [label=None, style=solid]; +"3346 dequantize_per_tensor_default_142" -> "3358 pad_26" [label=None, style=solid]; +"3347 linear_141" -> "3348 relu__23" [label="(1, 15, 15, 512)", style=solid]; +"3348 relu__23" -> "3349 linear_142" [label="(1, 15, 15, 512)", style=solid]; +"3349 linear_142" -> "3350 view_125" [label="(1, 15, 15, 24)", style=solid]; +"3350 view_125" -> "3351 index_23" [label="(225, 24)", style=solid]; +"3351 index_23" -> "3352 view_126" [label="(4096, 24)", style=solid]; +"3352 view_126" -> "3353 permute_104" [label="(64, 64, 24)", style=solid]; +"3353 permute_104" -> "3354 contiguous_44" [label="(24, 64, 64)", style=solid]; +"3354 contiguous_44" -> "3355 unsqueeze_67" [label="(24, 64, 64)", style=solid]; +"3355 unsqueeze_67" -> "3356 sigmoid_23" [label="(1, 24, 64, 64)", style=solid]; +"3356 sigmoid_23" -> "3357 mul_46" [label="(1, 24, 64, 64)", style=solid]; +"3357 mul_46" -> "3388 add_80" [label="(1, 24, 64, 64)", style=solid]; +"3358 pad_26" -> "3359 view_127" [label="(1, 8, 8, 768)", style=solid]; +"3359 view_127" -> "3360 permute_105" [label="(1, 1, 8, 1, 8, 768)", style=solid]; +"3360 permute_105" -> "3361 reshape_103" [label="(1, 1, 1, 8, 8, 768)", style=solid]; +"3361 reshape_103" -> "3365 linear_143" [label="(1, 64, 768)", style=solid]; +"3362 clone_23" -> "3363 slice_346" [label="(2304,)", style=solid]; +"3362 clone_23" -> "3365 linear_143" [label="(2304,)", style=solid]; +"3363 slice_346" -> "3364 zero__23" [label="(768,)", style=solid]; +"3365 linear_143" -> "3366 reshape_104" [label="(1, 64, 2304)", style=solid]; +"3366 reshape_104" -> "3367 permute_106" [label="(1, 64, 3, 24, 32)", style=solid]; +"3367 permute_106" -> "3368 select_69" [label="(3, 1, 24, 64, 32)", style=solid]; +"3367 permute_106" -> "3369 select_70" [label="(3, 1, 24, 64, 32)", style=solid]; +"3367 permute_106" -> "3370 select_71" [label="(3, 1, 24, 64, 32)", style=solid]; +"3368 select_69" -> "3371 linalg_vector_norm_46" [label="(1, 24, 64, 32)", style=solid]; +"3368 select_69" -> "3373 expand_as_46" [label="(1, 24, 64, 32)", style=solid]; +"3368 select_69" -> "3374 div_46" [label="(1, 24, 64, 32)", style=solid]; +"3369 select_70" -> "3377 linalg_vector_norm_47" [label="(1, 24, 64, 32)", style=solid]; +"3369 select_70" -> "3379 expand_as_47" [label="(1, 24, 64, 32)", style=solid]; +"3369 select_70" -> "3380 div_47" [label="(1, 24, 64, 32)", style=solid]; +"3370 select_71" -> "3391 matmul_47" [label="(1, 24, 64, 32)", style=solid]; +"3371 linalg_vector_norm_46" -> "3372 clamp_min_46" [label="(1, 24, 64, 1)", style=solid]; +"3372 clamp_min_46" -> "3373 expand_as_46" [label="(1, 24, 64, 1)", style=solid]; +"3373 expand_as_46" -> "3374 div_46" [label="(1, 24, 64, 32)", style=solid]; +"3374 div_46" -> "3375 quantize_per_tensor_default_143" [label="(1, 24, 64, 32)", style=solid]; +"3375 quantize_per_tensor_default_143" -> "3376 dequantize_per_tensor_default_143" [label=None, style=solid]; +"3376 dequantize_per_tensor_default_143" -> "3384 matmul_46" [label=None, style=solid]; +"3377 linalg_vector_norm_47" -> "3378 clamp_min_47" [label="(1, 24, 64, 1)", style=solid]; +"3378 clamp_min_47" -> "3379 expand_as_47" [label="(1, 24, 64, 1)", style=solid]; +"3379 expand_as_47" -> "3380 div_47" [label="(1, 24, 64, 32)", style=solid]; +"3380 div_47" -> "3381 quantize_per_tensor_default_144" [label="(1, 24, 64, 32)", style=solid]; +"3381 quantize_per_tensor_default_144" -> "3382 dequantize_per_tensor_default_144" [label=None, style=solid]; +"3382 dequantize_per_tensor_default_144" -> "3383 transpose_46" [label=None, style=solid]; +"3383 transpose_46" -> "3384 matmul_46" [label="(1, 24, 32, 64)", style=solid]; +"3384 matmul_46" -> "3387 mul_47" [label="(1, 24, 64, 64)", style=solid]; +"3385 clamp_23" -> "3386 exp_23" [label="(24, 1, 1)", style=solid]; +"3386 exp_23" -> "3387 mul_47" [label="(24, 1, 1)", style=solid]; +"3387 mul_47" -> "3388 add_80" [label="(1, 24, 64, 64)", style=solid]; +"3388 add_80" -> "3389 softmax_23" [label="(1, 24, 64, 64)", style=solid]; +"3389 softmax_23" -> "3390 dropout_92" [label="(1, 24, 64, 64)", style=solid]; +"3390 dropout_92" -> "3391 matmul_47" [label="(1, 24, 64, 64)", style=solid]; +"3391 matmul_47" -> "3392 quantize_per_tensor_default_145" [label="(1, 24, 64, 32)", style=solid]; +"3392 quantize_per_tensor_default_145" -> "3393 dequantize_per_tensor_default_145" [label=None, style=solid]; +"3393 dequantize_per_tensor_default_145" -> "3394 transpose_47" [label=None, style=solid]; +"3394 transpose_47" -> "3395 reshape_105" [label="(1, 64, 24, 32)", style=solid]; +"3395 reshape_105" -> "3396 linear_144" [label="(1, 64, 768)", style=solid]; +"3396 linear_144" -> "3397 dropout_93" [label="(1, 64, 768)", style=solid]; +"3397 dropout_93" -> "3398 view_128" [label="(1, 64, 768)", style=solid]; +"3398 view_128" -> "3399 permute_107" [label="(1, 1, 1, 8, 8, 768)", style=solid]; +"3399 permute_107" -> "3400 reshape_106" [label="(1, 1, 8, 1, 8, 768)", style=solid]; +"3400 reshape_106" -> "3401 slice_347" [label="(1, 8, 8, 768)", style=solid]; +"3401 slice_347" -> "3402 slice_348" [label="(1, 8, 8, 768)", style=solid]; +"3402 slice_348" -> "3403 slice_349" [label="(1, 7, 8, 768)", style=solid]; +"3403 slice_349" -> "3404 slice_350" [label="(1, 7, 7, 768)", style=solid]; +"3404 slice_350" -> "3405 contiguous_45" [label="(1, 7, 7, 768)", style=solid]; +"3405 contiguous_45" -> "3406 layer_norm_50" [label="(1, 7, 7, 768)", style=solid]; +"3406 layer_norm_50" -> "3407 add_81" [label="(1, 7, 7, 768)", style=solid]; +"3407 add_81" -> "3408 quantize_per_tensor_default_146" [label="(1, 7, 7, 768)", style=solid]; +"3407 add_81" -> "3418 add_82" [label="(1, 7, 7, 768)", style=solid]; +"3408 quantize_per_tensor_default_146" -> "3409 dequantize_per_tensor_default_146" [label=None, style=solid]; +"3409 dequantize_per_tensor_default_146" -> "3410 linear_145" [label=None, style=solid]; +"3410 linear_145" -> "3411 gelu_23" [label="(1, 7, 7, 3072)", style=solid]; +"3411 gelu_23" -> "3412 quantize_per_tensor_default_147" [label="(1, 7, 7, 3072)", style=solid]; +"3412 quantize_per_tensor_default_147" -> "3413 dequantize_per_tensor_default_147" [label=None, style=solid]; +"3413 dequantize_per_tensor_default_147" -> "3414 dropout_94" [label=None, style=solid]; +"3414 dropout_94" -> "3415 linear_146" [label="(1, 7, 7, 3072)", style=solid]; +"3415 linear_146" -> "3416 dropout_95" [label="(1, 7, 7, 768)", style=solid]; +"3416 dropout_95" -> "3417 layer_norm_51" [label="(1, 7, 7, 768)", style=solid]; +"3417 layer_norm_51" -> "3418 add_82" [label="(1, 7, 7, 768)", style=solid]; +"3418 add_82" -> "3419 layer_norm_52" [label="(1, 7, 7, 768)", style=solid]; +"3419 layer_norm_52" -> "3420 permute_108" [label="(1, 7, 7, 768)", style=solid]; +"3420 permute_108" -> "3421 adaptive_avg_pool2d" [label="(1, 768, 7, 7)", style=solid]; +"3421 adaptive_avg_pool2d" -> "3422 quantize_per_tensor_default_148" [label="(1, 768, 1, 1)", style=solid]; +"3422 quantize_per_tensor_default_148" -> "3423 dequantize_per_tensor_default_148" [label=None, style=solid]; +"3423 dequantize_per_tensor_default_148" -> "3424 flatten" [label=None, style=solid]; +"3424 flatten" -> "3425 linear_147" [label="(1, 768)", style=solid]; +"3425 linear_147" -> "3426 output_1" [label="(1, 1000)", style=solid]; +} diff --git a/tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/synthetic_transformer.dot b/tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/synthetic_transformer.dot new file mode 100644 index 00000000000..ef95b18fdfa --- /dev/null +++ b/tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/synthetic_transformer.dot @@ -0,0 +1,41 @@ +strict digraph { +"0 wte_weight" [id=0, type=get_attr]; +"1 _scale_0" [id=1, type=get_attr]; +"2 _zero_point_0" [id=2, type=get_attr]; +"3 _frozen_param0" [id=3, type=get_attr]; +"4 dequantize_per_channel_default" [id=4, type=dequantize_per_channel]; +"5 linear_bias" [id=5, type=get_attr]; +"6 _scale_1" [id=6, type=get_attr]; +"7 _zero_point_1" [id=7, type=get_attr]; +"8 _frozen_param1" [id=8, type=get_attr]; +"9 dequantize_per_channel_default_1" [id=9, type=dequantize_per_channel]; +"10 lm_head_bias" [id=10, type=get_attr]; +"11 input_ids" [id=11, type=input]; +"12 embedding" [id=12, type=embedding]; +"13 quantize_per_tensor_default" [id=13, type=quantize_per_tensor]; +"14 dequantize_per_tensor_default" [id=14, type=dequantize_per_tensor]; +"15 linear" [id=15, type=linear]; +"16 quantize_per_tensor_default_1" [id=16, type=quantize_per_tensor]; +"17 dequantize_per_tensor_default_1" [id=17, type=dequantize_per_tensor]; +"18 linear_1" [id=18, type=linear]; +"19 output_1" [id=19, type=output]; +"0 wte_weight" -> "12 embedding" [label="(10, 5)", style=solid]; +"1 _scale_0" -> "4 dequantize_per_channel_default" [label="(5,)", style=solid]; +"2 _zero_point_0" -> "4 dequantize_per_channel_default" [label="(5,)", style=solid]; +"3 _frozen_param0" -> "4 dequantize_per_channel_default" [label="(5, 5)", style=solid]; +"4 dequantize_per_channel_default" -> "15 linear" [label=None, style=solid]; +"5 linear_bias" -> "15 linear" [label="(5,)", style=solid]; +"6 _scale_1" -> "9 dequantize_per_channel_default_1" [label="(10,)", style=solid]; +"7 _zero_point_1" -> "9 dequantize_per_channel_default_1" [label="(10,)", style=solid]; +"8 _frozen_param1" -> "9 dequantize_per_channel_default_1" [label="(10, 5)", style=solid]; +"9 dequantize_per_channel_default_1" -> "18 linear_1" [label=None, style=solid]; +"10 lm_head_bias" -> "18 linear_1" [label="(10,)", style=solid]; +"11 input_ids" -> "12 embedding" [label="(5,)", style=solid]; +"12 embedding" -> "13 quantize_per_tensor_default" [label="(5, 5)", style=solid]; +"13 quantize_per_tensor_default" -> "14 dequantize_per_tensor_default" [label=None, style=solid]; +"14 dequantize_per_tensor_default" -> "15 linear" [label=None, style=solid]; +"15 linear" -> "16 quantize_per_tensor_default_1" [label="(5, 5)", style=solid]; +"16 quantize_per_tensor_default_1" -> "17 dequantize_per_tensor_default_1" [label=None, style=solid]; +"17 dequantize_per_tensor_default_1" -> "18 linear_1" [label=None, style=solid]; +"18 linear_1" -> "19 output_1" [label="(5, 10)", style=solid]; +} diff --git a/tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/unet.dot b/tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/unet.dot new file mode 100644 index 00000000000..6c7e8e88eff --- /dev/null +++ b/tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/unet.dot @@ -0,0 +1,461 @@ +strict digraph { +"0 _scale_0" [id=0, type=get_attr]; +"1 _zero_point_0" [id=1, type=get_attr]; +"2 _frozen_param0" [id=2, type=get_attr]; +"3 dequantize_per_channel_default" [id=3, type=dequantize_per_channel]; +"4 down_path_0_block_0_bias" [id=4, type=get_attr]; +"5 _scale_1" [id=5, type=get_attr]; +"6 _zero_point_1" [id=6, type=get_attr]; +"7 _frozen_param1" [id=7, type=get_attr]; +"8 dequantize_per_channel_default_1" [id=8, type=dequantize_per_channel]; +"9 down_path_0_block_3_bias" [id=9, type=get_attr]; +"10 _scale_2" [id=10, type=get_attr]; +"11 _zero_point_2" [id=11, type=get_attr]; +"12 _frozen_param2" [id=12, type=get_attr]; +"13 dequantize_per_channel_default_2" [id=13, type=dequantize_per_channel]; +"14 down_path_1_block_0_bias" [id=14, type=get_attr]; +"15 _scale_3" [id=15, type=get_attr]; +"16 _zero_point_3" [id=16, type=get_attr]; +"17 _frozen_param3" [id=17, type=get_attr]; +"18 dequantize_per_channel_default_3" [id=18, type=dequantize_per_channel]; +"19 down_path_1_block_3_bias" [id=19, type=get_attr]; +"20 _scale_4" [id=20, type=get_attr]; +"21 _zero_point_4" [id=21, type=get_attr]; +"22 _frozen_param4" [id=22, type=get_attr]; +"23 dequantize_per_channel_default_4" [id=23, type=dequantize_per_channel]; +"24 down_path_2_block_0_bias" [id=24, type=get_attr]; +"25 _scale_5" [id=25, type=get_attr]; +"26 _zero_point_5" [id=26, type=get_attr]; +"27 _frozen_param5" [id=27, type=get_attr]; +"28 dequantize_per_channel_default_5" [id=28, type=dequantize_per_channel]; +"29 down_path_2_block_3_bias" [id=29, type=get_attr]; +"30 _scale_6" [id=30, type=get_attr]; +"31 _zero_point_6" [id=31, type=get_attr]; +"32 _frozen_param6" [id=32, type=get_attr]; +"33 dequantize_per_channel_default_6" [id=33, type=dequantize_per_channel]; +"34 down_path_3_block_0_bias" [id=34, type=get_attr]; +"35 _scale_7" [id=35, type=get_attr]; +"36 _zero_point_7" [id=36, type=get_attr]; +"37 _frozen_param7" [id=37, type=get_attr]; +"38 dequantize_per_channel_default_7" [id=38, type=dequantize_per_channel]; +"39 down_path_3_block_3_bias" [id=39, type=get_attr]; +"40 _scale_8" [id=40, type=get_attr]; +"41 _zero_point_8" [id=41, type=get_attr]; +"42 _frozen_param8" [id=42, type=get_attr]; +"43 dequantize_per_channel_default_8" [id=43, type=dequantize_per_channel]; +"44 down_path_4_block_0_bias" [id=44, type=get_attr]; +"45 _scale_9" [id=45, type=get_attr]; +"46 _zero_point_9" [id=46, type=get_attr]; +"47 _frozen_param9" [id=47, type=get_attr]; +"48 dequantize_per_channel_default_9" [id=48, type=dequantize_per_channel]; +"49 down_path_4_block_3_bias" [id=49, type=get_attr]; +"50 _scale_10" [id=50, type=get_attr]; +"51 _zero_point_10" [id=51, type=get_attr]; +"52 _frozen_param10" [id=52, type=get_attr]; +"53 dequantize_per_channel_default_10" [id=53, type=dequantize_per_channel]; +"54 up_path_0_up_bias" [id=54, type=get_attr]; +"55 _scale_11" [id=55, type=get_attr]; +"56 _zero_point_11" [id=56, type=get_attr]; +"57 _frozen_param11" [id=57, type=get_attr]; +"58 dequantize_per_channel_default_11" [id=58, type=dequantize_per_channel]; +"59 up_path_0_conv_block_block_0_bias" [id=59, type=get_attr]; +"60 _scale_12" [id=60, type=get_attr]; +"61 _zero_point_12" [id=61, type=get_attr]; +"62 _frozen_param12" [id=62, type=get_attr]; +"63 dequantize_per_channel_default_12" [id=63, type=dequantize_per_channel]; +"64 up_path_0_conv_block_block_3_bias" [id=64, type=get_attr]; +"65 _scale_13" [id=65, type=get_attr]; +"66 _zero_point_13" [id=66, type=get_attr]; +"67 _frozen_param13" [id=67, type=get_attr]; +"68 dequantize_per_channel_default_13" [id=68, type=dequantize_per_channel]; +"69 up_path_1_up_bias" [id=69, type=get_attr]; +"70 _scale_14" [id=70, type=get_attr]; +"71 _zero_point_14" [id=71, type=get_attr]; +"72 _frozen_param14" [id=72, type=get_attr]; +"73 dequantize_per_channel_default_14" [id=73, type=dequantize_per_channel]; +"74 up_path_1_conv_block_block_0_bias" [id=74, type=get_attr]; +"75 _scale_15" [id=75, type=get_attr]; +"76 _zero_point_15" [id=76, type=get_attr]; +"77 _frozen_param15" [id=77, type=get_attr]; +"78 dequantize_per_channel_default_15" [id=78, type=dequantize_per_channel]; +"79 up_path_1_conv_block_block_3_bias" [id=79, type=get_attr]; +"80 _scale_16" [id=80, type=get_attr]; +"81 _zero_point_16" [id=81, type=get_attr]; +"82 _frozen_param16" [id=82, type=get_attr]; +"83 dequantize_per_channel_default_16" [id=83, type=dequantize_per_channel]; +"84 up_path_2_up_bias" [id=84, type=get_attr]; +"85 _scale_17" [id=85, type=get_attr]; +"86 _zero_point_17" [id=86, type=get_attr]; +"87 _frozen_param17" [id=87, type=get_attr]; +"88 dequantize_per_channel_default_17" [id=88, type=dequantize_per_channel]; +"89 up_path_2_conv_block_block_0_bias" [id=89, type=get_attr]; +"90 _scale_18" [id=90, type=get_attr]; +"91 _zero_point_18" [id=91, type=get_attr]; +"92 _frozen_param18" [id=92, type=get_attr]; +"93 dequantize_per_channel_default_18" [id=93, type=dequantize_per_channel]; +"94 up_path_2_conv_block_block_3_bias" [id=94, type=get_attr]; +"95 _scale_19" [id=95, type=get_attr]; +"96 _zero_point_19" [id=96, type=get_attr]; +"97 _frozen_param19" [id=97, type=get_attr]; +"98 dequantize_per_channel_default_19" [id=98, type=dequantize_per_channel]; +"99 up_path_3_up_bias" [id=99, type=get_attr]; +"100 _scale_20" [id=100, type=get_attr]; +"101 _zero_point_20" [id=101, type=get_attr]; +"102 _frozen_param20" [id=102, type=get_attr]; +"103 dequantize_per_channel_default_20" [id=103, type=dequantize_per_channel]; +"104 up_path_3_conv_block_block_0_bias" [id=104, type=get_attr]; +"105 _scale_21" [id=105, type=get_attr]; +"106 _zero_point_21" [id=106, type=get_attr]; +"107 _frozen_param21" [id=107, type=get_attr]; +"108 dequantize_per_channel_default_21" [id=108, type=dequantize_per_channel]; +"109 up_path_3_conv_block_block_3_bias" [id=109, type=get_attr]; +"110 _scale_22" [id=110, type=get_attr]; +"111 _zero_point_22" [id=111, type=get_attr]; +"112 _frozen_param22" [id=112, type=get_attr]; +"113 dequantize_per_channel_default_22" [id=113, type=dequantize_per_channel]; +"114 last_bias" [id=114, type=get_attr]; +"115 x" [id=115, type=input]; +"116 quantize_per_tensor_default" [id=116, type=quantize_per_tensor]; +"117 dequantize_per_tensor_default" [id=117, type=dequantize_per_tensor]; +"118 conv2d" [id=118, type=conv2d]; +"119 relu" [id=119, type=relu]; +"120 quantize_per_tensor_default_1" [id=120, type=quantize_per_tensor]; +"121 dequantize_per_tensor_default_1" [id=121, type=dequantize_per_tensor]; +"122 conv2d_1" [id=122, type=conv2d]; +"123 relu_1" [id=123, type=relu]; +"124 quantize_per_tensor_default_2" [id=124, type=quantize_per_tensor]; +"125 dequantize_per_tensor_default_2" [id=125, type=dequantize_per_tensor]; +"126 max_pool2d" [id=126, type=max_pool2d]; +"127 conv2d_2" [id=127, type=conv2d]; +"128 relu_2" [id=128, type=relu]; +"129 quantize_per_tensor_default_3" [id=129, type=quantize_per_tensor]; +"130 dequantize_per_tensor_default_3" [id=130, type=dequantize_per_tensor]; +"131 conv2d_3" [id=131, type=conv2d]; +"132 relu_3" [id=132, type=relu]; +"133 quantize_per_tensor_default_4" [id=133, type=quantize_per_tensor]; +"134 dequantize_per_tensor_default_4" [id=134, type=dequantize_per_tensor]; +"135 max_pool2d_1" [id=135, type=max_pool2d]; +"136 conv2d_4" [id=136, type=conv2d]; +"137 relu_4" [id=137, type=relu]; +"138 quantize_per_tensor_default_5" [id=138, type=quantize_per_tensor]; +"139 dequantize_per_tensor_default_5" [id=139, type=dequantize_per_tensor]; +"140 conv2d_5" [id=140, type=conv2d]; +"141 relu_5" [id=141, type=relu]; +"142 quantize_per_tensor_default_6" [id=142, type=quantize_per_tensor]; +"143 dequantize_per_tensor_default_6" [id=143, type=dequantize_per_tensor]; +"144 max_pool2d_2" [id=144, type=max_pool2d]; +"145 conv2d_6" [id=145, type=conv2d]; +"146 relu_6" [id=146, type=relu]; +"147 quantize_per_tensor_default_7" [id=147, type=quantize_per_tensor]; +"148 dequantize_per_tensor_default_7" [id=148, type=dequantize_per_tensor]; +"149 conv2d_7" [id=149, type=conv2d]; +"150 relu_7" [id=150, type=relu]; +"151 quantize_per_tensor_default_8" [id=151, type=quantize_per_tensor]; +"152 dequantize_per_tensor_default_8" [id=152, type=dequantize_per_tensor]; +"153 max_pool2d_3" [id=153, type=max_pool2d]; +"154 conv2d_8" [id=154, type=conv2d]; +"155 relu_8" [id=155, type=relu]; +"156 quantize_per_tensor_default_9" [id=156, type=quantize_per_tensor]; +"157 dequantize_per_tensor_default_9" [id=157, type=dequantize_per_tensor]; +"158 conv2d_9" [id=158, type=conv2d]; +"159 relu_9" [id=159, type=relu]; +"160 quantize_per_tensor_default_10" [id=160, type=quantize_per_tensor]; +"161 dequantize_per_tensor_default_10" [id=161, type=dequantize_per_tensor]; +"162 conv_transpose2d" [id=162, type=conv_transpose2d]; +"163 quantize_per_tensor_default_11" [id=163, type=quantize_per_tensor]; +"164 dequantize_per_tensor_default_11" [id=164, type=dequantize_per_tensor]; +"165 slice_1" [id=165, type=slice]; +"166 slice_2" [id=166, type=slice]; +"167 slice_3" [id=167, type=slice]; +"168 slice_4" [id=168, type=slice]; +"169 cat" [id=169, type=cat]; +"170 conv2d_10" [id=170, type=conv2d]; +"171 relu_10" [id=171, type=relu]; +"172 quantize_per_tensor_default_12" [id=172, type=quantize_per_tensor]; +"173 dequantize_per_tensor_default_12" [id=173, type=dequantize_per_tensor]; +"174 conv2d_11" [id=174, type=conv2d]; +"175 relu_11" [id=175, type=relu]; +"176 quantize_per_tensor_default_13" [id=176, type=quantize_per_tensor]; +"177 dequantize_per_tensor_default_13" [id=177, type=dequantize_per_tensor]; +"178 conv_transpose2d_1" [id=178, type=conv_transpose2d]; +"179 quantize_per_tensor_default_14" [id=179, type=quantize_per_tensor]; +"180 dequantize_per_tensor_default_14" [id=180, type=dequantize_per_tensor]; +"181 slice_5" [id=181, type=slice]; +"182 slice_6" [id=182, type=slice]; +"183 slice_7" [id=183, type=slice]; +"184 slice_8" [id=184, type=slice]; +"185 cat_1" [id=185, type=cat]; +"186 conv2d_12" [id=186, type=conv2d]; +"187 relu_12" [id=187, type=relu]; +"188 quantize_per_tensor_default_15" [id=188, type=quantize_per_tensor]; +"189 dequantize_per_tensor_default_15" [id=189, type=dequantize_per_tensor]; +"190 conv2d_13" [id=190, type=conv2d]; +"191 relu_13" [id=191, type=relu]; +"192 quantize_per_tensor_default_16" [id=192, type=quantize_per_tensor]; +"193 dequantize_per_tensor_default_16" [id=193, type=dequantize_per_tensor]; +"194 conv_transpose2d_2" [id=194, type=conv_transpose2d]; +"195 quantize_per_tensor_default_17" [id=195, type=quantize_per_tensor]; +"196 dequantize_per_tensor_default_17" [id=196, type=dequantize_per_tensor]; +"197 slice_9" [id=197, type=slice]; +"198 slice_10" [id=198, type=slice]; +"199 slice_11" [id=199, type=slice]; +"200 slice_12" [id=200, type=slice]; +"201 cat_2" [id=201, type=cat]; +"202 conv2d_14" [id=202, type=conv2d]; +"203 relu_14" [id=203, type=relu]; +"204 quantize_per_tensor_default_18" [id=204, type=quantize_per_tensor]; +"205 dequantize_per_tensor_default_18" [id=205, type=dequantize_per_tensor]; +"206 conv2d_15" [id=206, type=conv2d]; +"207 relu_15" [id=207, type=relu]; +"208 quantize_per_tensor_default_19" [id=208, type=quantize_per_tensor]; +"209 dequantize_per_tensor_default_19" [id=209, type=dequantize_per_tensor]; +"210 conv_transpose2d_3" [id=210, type=conv_transpose2d]; +"211 quantize_per_tensor_default_20" [id=211, type=quantize_per_tensor]; +"212 dequantize_per_tensor_default_20" [id=212, type=dequantize_per_tensor]; +"213 slice_13" [id=213, type=slice]; +"214 slice_14" [id=214, type=slice]; +"215 slice_15" [id=215, type=slice]; +"216 slice_16" [id=216, type=slice]; +"217 cat_3" [id=217, type=cat]; +"218 conv2d_16" [id=218, type=conv2d]; +"219 relu_16" [id=219, type=relu]; +"220 quantize_per_tensor_default_21" [id=220, type=quantize_per_tensor]; +"221 dequantize_per_tensor_default_21" [id=221, type=dequantize_per_tensor]; +"222 conv2d_17" [id=222, type=conv2d]; +"223 relu_17" [id=223, type=relu]; +"224 quantize_per_tensor_default_22" [id=224, type=quantize_per_tensor]; +"225 dequantize_per_tensor_default_22" [id=225, type=dequantize_per_tensor]; +"226 conv2d_18" [id=226, type=conv2d]; +"227 output_1" [id=227, type=output]; +"0 _scale_0" -> "3 dequantize_per_channel_default" [label="(64,)", style=solid]; +"1 _zero_point_0" -> "3 dequantize_per_channel_default" [label="(64,)", style=solid]; +"2 _frozen_param0" -> "3 dequantize_per_channel_default" [label="(64, 3, 3, 3)", style=solid]; +"3 dequantize_per_channel_default" -> "118 conv2d" [label=None, style=solid]; +"4 down_path_0_block_0_bias" -> "118 conv2d" [label="(64,)", style=solid]; +"5 _scale_1" -> "8 dequantize_per_channel_default_1" [label="(64,)", style=solid]; +"6 _zero_point_1" -> "8 dequantize_per_channel_default_1" [label="(64,)", style=solid]; +"7 _frozen_param1" -> "8 dequantize_per_channel_default_1" [label="(64, 64, 3, 3)", style=solid]; +"8 dequantize_per_channel_default_1" -> "122 conv2d_1" [label=None, style=solid]; +"9 down_path_0_block_3_bias" -> "122 conv2d_1" [label="(64,)", style=solid]; +"10 _scale_2" -> "13 dequantize_per_channel_default_2" [label="(128,)", style=solid]; +"11 _zero_point_2" -> "13 dequantize_per_channel_default_2" [label="(128,)", style=solid]; +"12 _frozen_param2" -> "13 dequantize_per_channel_default_2" [label="(128, 64, 3, 3)", style=solid]; +"13 dequantize_per_channel_default_2" -> "127 conv2d_2" [label=None, style=solid]; +"14 down_path_1_block_0_bias" -> "127 conv2d_2" [label="(128,)", style=solid]; +"15 _scale_3" -> "18 dequantize_per_channel_default_3" [label="(128,)", style=solid]; +"16 _zero_point_3" -> "18 dequantize_per_channel_default_3" [label="(128,)", style=solid]; +"17 _frozen_param3" -> "18 dequantize_per_channel_default_3" [label="(128, 128, 3, 3)", style=solid]; +"18 dequantize_per_channel_default_3" -> "131 conv2d_3" [label=None, style=solid]; +"19 down_path_1_block_3_bias" -> "131 conv2d_3" [label="(128,)", style=solid]; +"20 _scale_4" -> "23 dequantize_per_channel_default_4" [label="(256,)", style=solid]; +"21 _zero_point_4" -> "23 dequantize_per_channel_default_4" [label="(256,)", style=solid]; +"22 _frozen_param4" -> "23 dequantize_per_channel_default_4" [label="(256, 128, 3, 3)", style=solid]; +"23 dequantize_per_channel_default_4" -> "136 conv2d_4" [label=None, style=solid]; +"24 down_path_2_block_0_bias" -> "136 conv2d_4" [label="(256,)", style=solid]; +"25 _scale_5" -> "28 dequantize_per_channel_default_5" [label="(256,)", style=solid]; +"26 _zero_point_5" -> "28 dequantize_per_channel_default_5" [label="(256,)", style=solid]; +"27 _frozen_param5" -> "28 dequantize_per_channel_default_5" [label="(256, 256, 3, 3)", style=solid]; +"28 dequantize_per_channel_default_5" -> "140 conv2d_5" [label=None, style=solid]; +"29 down_path_2_block_3_bias" -> "140 conv2d_5" [label="(256,)", style=solid]; +"30 _scale_6" -> "33 dequantize_per_channel_default_6" [label="(512,)", style=solid]; +"31 _zero_point_6" -> "33 dequantize_per_channel_default_6" [label="(512,)", style=solid]; +"32 _frozen_param6" -> "33 dequantize_per_channel_default_6" [label="(512, 256, 3, 3)", style=solid]; +"33 dequantize_per_channel_default_6" -> "145 conv2d_6" [label=None, style=solid]; +"34 down_path_3_block_0_bias" -> "145 conv2d_6" [label="(512,)", style=solid]; +"35 _scale_7" -> "38 dequantize_per_channel_default_7" [label="(512,)", style=solid]; +"36 _zero_point_7" -> "38 dequantize_per_channel_default_7" [label="(512,)", style=solid]; +"37 _frozen_param7" -> "38 dequantize_per_channel_default_7" [label="(512, 512, 3, 3)", style=solid]; +"38 dequantize_per_channel_default_7" -> "149 conv2d_7" [label=None, style=solid]; +"39 down_path_3_block_3_bias" -> "149 conv2d_7" [label="(512,)", style=solid]; +"40 _scale_8" -> "43 dequantize_per_channel_default_8" [label="(1024,)", style=solid]; +"41 _zero_point_8" -> "43 dequantize_per_channel_default_8" [label="(1024,)", style=solid]; +"42 _frozen_param8" -> "43 dequantize_per_channel_default_8" [label="(1024, 512, 3, 3)", style=solid]; +"43 dequantize_per_channel_default_8" -> "154 conv2d_8" [label=None, style=solid]; +"44 down_path_4_block_0_bias" -> "154 conv2d_8" [label="(1024,)", style=solid]; +"45 _scale_9" -> "48 dequantize_per_channel_default_9" [label="(1024,)", style=solid]; +"46 _zero_point_9" -> "48 dequantize_per_channel_default_9" [label="(1024,)", style=solid]; +"47 _frozen_param9" -> "48 dequantize_per_channel_default_9" [label="(1024, 1024, 3, 3)", style=solid]; +"48 dequantize_per_channel_default_9" -> "158 conv2d_9" [label=None, style=solid]; +"49 down_path_4_block_3_bias" -> "158 conv2d_9" [label="(1024,)", style=solid]; +"50 _scale_10" -> "53 dequantize_per_channel_default_10" [label="(1024,)", style=solid]; +"51 _zero_point_10" -> "53 dequantize_per_channel_default_10" [label="(1024,)", style=solid]; +"52 _frozen_param10" -> "53 dequantize_per_channel_default_10" [label="(1024, 512, 2, 2)", style=solid]; +"53 dequantize_per_channel_default_10" -> "162 conv_transpose2d" [label=None, style=solid]; +"54 up_path_0_up_bias" -> "162 conv_transpose2d" [label="(512,)", style=solid]; +"55 _scale_11" -> "58 dequantize_per_channel_default_11" [label="(512,)", style=solid]; +"56 _zero_point_11" -> "58 dequantize_per_channel_default_11" [label="(512,)", style=solid]; +"57 _frozen_param11" -> "58 dequantize_per_channel_default_11" [label="(512, 1024, 3, 3)", style=solid]; +"58 dequantize_per_channel_default_11" -> "170 conv2d_10" [label=None, style=solid]; +"59 up_path_0_conv_block_block_0_bias" -> "170 conv2d_10" [label="(512,)", style=solid]; +"60 _scale_12" -> "63 dequantize_per_channel_default_12" [label="(512,)", style=solid]; +"61 _zero_point_12" -> "63 dequantize_per_channel_default_12" [label="(512,)", style=solid]; +"62 _frozen_param12" -> "63 dequantize_per_channel_default_12" [label="(512, 512, 3, 3)", style=solid]; +"63 dequantize_per_channel_default_12" -> "174 conv2d_11" [label=None, style=solid]; +"64 up_path_0_conv_block_block_3_bias" -> "174 conv2d_11" [label="(512,)", style=solid]; +"65 _scale_13" -> "68 dequantize_per_channel_default_13" [label="(512,)", style=solid]; +"66 _zero_point_13" -> "68 dequantize_per_channel_default_13" [label="(512,)", style=solid]; +"67 _frozen_param13" -> "68 dequantize_per_channel_default_13" [label="(512, 256, 2, 2)", style=solid]; +"68 dequantize_per_channel_default_13" -> "178 conv_transpose2d_1" [label=None, style=solid]; +"69 up_path_1_up_bias" -> "178 conv_transpose2d_1" [label="(256,)", style=solid]; +"70 _scale_14" -> "73 dequantize_per_channel_default_14" [label="(256,)", style=solid]; +"71 _zero_point_14" -> "73 dequantize_per_channel_default_14" [label="(256,)", style=solid]; +"72 _frozen_param14" -> "73 dequantize_per_channel_default_14" [label="(256, 512, 3, 3)", style=solid]; +"73 dequantize_per_channel_default_14" -> "186 conv2d_12" [label=None, style=solid]; +"74 up_path_1_conv_block_block_0_bias" -> "186 conv2d_12" [label="(256,)", style=solid]; +"75 _scale_15" -> "78 dequantize_per_channel_default_15" [label="(256,)", style=solid]; +"76 _zero_point_15" -> "78 dequantize_per_channel_default_15" [label="(256,)", style=solid]; +"77 _frozen_param15" -> "78 dequantize_per_channel_default_15" [label="(256, 256, 3, 3)", style=solid]; +"78 dequantize_per_channel_default_15" -> "190 conv2d_13" [label=None, style=solid]; +"79 up_path_1_conv_block_block_3_bias" -> "190 conv2d_13" [label="(256,)", style=solid]; +"80 _scale_16" -> "83 dequantize_per_channel_default_16" [label="(256,)", style=solid]; +"81 _zero_point_16" -> "83 dequantize_per_channel_default_16" [label="(256,)", style=solid]; +"82 _frozen_param16" -> "83 dequantize_per_channel_default_16" [label="(256, 128, 2, 2)", style=solid]; +"83 dequantize_per_channel_default_16" -> "194 conv_transpose2d_2" [label=None, style=solid]; +"84 up_path_2_up_bias" -> "194 conv_transpose2d_2" [label="(128,)", style=solid]; +"85 _scale_17" -> "88 dequantize_per_channel_default_17" [label="(128,)", style=solid]; +"86 _zero_point_17" -> "88 dequantize_per_channel_default_17" [label="(128,)", style=solid]; +"87 _frozen_param17" -> "88 dequantize_per_channel_default_17" [label="(128, 256, 3, 3)", style=solid]; +"88 dequantize_per_channel_default_17" -> "202 conv2d_14" [label=None, style=solid]; +"89 up_path_2_conv_block_block_0_bias" -> "202 conv2d_14" [label="(128,)", style=solid]; +"90 _scale_18" -> "93 dequantize_per_channel_default_18" [label="(128,)", style=solid]; +"91 _zero_point_18" -> "93 dequantize_per_channel_default_18" [label="(128,)", style=solid]; +"92 _frozen_param18" -> "93 dequantize_per_channel_default_18" [label="(128, 128, 3, 3)", style=solid]; +"93 dequantize_per_channel_default_18" -> "206 conv2d_15" [label=None, style=solid]; +"94 up_path_2_conv_block_block_3_bias" -> "206 conv2d_15" [label="(128,)", style=solid]; +"95 _scale_19" -> "98 dequantize_per_channel_default_19" [label="(128,)", style=solid]; +"96 _zero_point_19" -> "98 dequantize_per_channel_default_19" [label="(128,)", style=solid]; +"97 _frozen_param19" -> "98 dequantize_per_channel_default_19" [label="(128, 64, 2, 2)", style=solid]; +"98 dequantize_per_channel_default_19" -> "210 conv_transpose2d_3" [label=None, style=solid]; +"99 up_path_3_up_bias" -> "210 conv_transpose2d_3" [label="(64,)", style=solid]; +"100 _scale_20" -> "103 dequantize_per_channel_default_20" [label="(64,)", style=solid]; +"101 _zero_point_20" -> "103 dequantize_per_channel_default_20" [label="(64,)", style=solid]; +"102 _frozen_param20" -> "103 dequantize_per_channel_default_20" [label="(64, 128, 3, 3)", style=solid]; +"103 dequantize_per_channel_default_20" -> "218 conv2d_16" [label=None, style=solid]; +"104 up_path_3_conv_block_block_0_bias" -> "218 conv2d_16" [label="(64,)", style=solid]; +"105 _scale_21" -> "108 dequantize_per_channel_default_21" [label="(64,)", style=solid]; +"106 _zero_point_21" -> "108 dequantize_per_channel_default_21" [label="(64,)", style=solid]; +"107 _frozen_param21" -> "108 dequantize_per_channel_default_21" [label="(64, 64, 3, 3)", style=solid]; +"108 dequantize_per_channel_default_21" -> "222 conv2d_17" [label=None, style=solid]; +"109 up_path_3_conv_block_block_3_bias" -> "222 conv2d_17" [label="(64,)", style=solid]; +"110 _scale_22" -> "113 dequantize_per_channel_default_22" [label="(12,)", style=solid]; +"111 _zero_point_22" -> "113 dequantize_per_channel_default_22" [label="(12,)", style=solid]; +"112 _frozen_param22" -> "113 dequantize_per_channel_default_22" [label="(12, 64, 1, 1)", style=solid]; +"113 dequantize_per_channel_default_22" -> "226 conv2d_18" [label=None, style=solid]; +"114 last_bias" -> "226 conv2d_18" [label="(12,)", style=solid]; +"115 x" -> "116 quantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; +"116 quantize_per_tensor_default" -> "117 dequantize_per_tensor_default" [label=None, style=solid]; +"117 dequantize_per_tensor_default" -> "118 conv2d" [label=None, style=solid]; +"118 conv2d" -> "119 relu" [label="(1, 64, 222, 222)", style=solid]; +"119 relu" -> "120 quantize_per_tensor_default_1" [label="(1, 64, 222, 222)", style=solid]; +"120 quantize_per_tensor_default_1" -> "121 dequantize_per_tensor_default_1" [label=None, style=solid]; +"121 dequantize_per_tensor_default_1" -> "122 conv2d_1" [label=None, style=solid]; +"122 conv2d_1" -> "123 relu_1" [label="(1, 64, 220, 220)", style=solid]; +"123 relu_1" -> "124 quantize_per_tensor_default_2" [label="(1, 64, 220, 220)", style=solid]; +"124 quantize_per_tensor_default_2" -> "125 dequantize_per_tensor_default_2" [label=None, style=solid]; +"125 dequantize_per_tensor_default_2" -> "126 max_pool2d" [label=None, style=solid]; +"125 dequantize_per_tensor_default_2" -> "213 slice_13" [label=None, style=solid]; +"126 max_pool2d" -> "127 conv2d_2" [label="(1, 64, 110, 110)", style=solid]; +"127 conv2d_2" -> "128 relu_2" [label="(1, 128, 108, 108)", style=solid]; +"128 relu_2" -> "129 quantize_per_tensor_default_3" [label="(1, 128, 108, 108)", style=solid]; +"129 quantize_per_tensor_default_3" -> "130 dequantize_per_tensor_default_3" [label=None, style=solid]; +"130 dequantize_per_tensor_default_3" -> "131 conv2d_3" [label=None, style=solid]; +"131 conv2d_3" -> "132 relu_3" [label="(1, 128, 106, 106)", style=solid]; +"132 relu_3" -> "133 quantize_per_tensor_default_4" [label="(1, 128, 106, 106)", style=solid]; +"133 quantize_per_tensor_default_4" -> "134 dequantize_per_tensor_default_4" [label=None, style=solid]; +"134 dequantize_per_tensor_default_4" -> "135 max_pool2d_1" [label=None, style=solid]; +"134 dequantize_per_tensor_default_4" -> "197 slice_9" [label=None, style=solid]; +"135 max_pool2d_1" -> "136 conv2d_4" [label="(1, 128, 53, 53)", style=solid]; +"136 conv2d_4" -> "137 relu_4" [label="(1, 256, 51, 51)", style=solid]; +"137 relu_4" -> "138 quantize_per_tensor_default_5" [label="(1, 256, 51, 51)", style=solid]; +"138 quantize_per_tensor_default_5" -> "139 dequantize_per_tensor_default_5" [label=None, style=solid]; +"139 dequantize_per_tensor_default_5" -> "140 conv2d_5" [label=None, style=solid]; +"140 conv2d_5" -> "141 relu_5" [label="(1, 256, 49, 49)", style=solid]; +"141 relu_5" -> "142 quantize_per_tensor_default_6" [label="(1, 256, 49, 49)", style=solid]; +"142 quantize_per_tensor_default_6" -> "143 dequantize_per_tensor_default_6" [label=None, style=solid]; +"143 dequantize_per_tensor_default_6" -> "144 max_pool2d_2" [label=None, style=solid]; +"143 dequantize_per_tensor_default_6" -> "181 slice_5" [label=None, style=solid]; +"144 max_pool2d_2" -> "145 conv2d_6" [label="(1, 256, 24, 24)", style=solid]; +"145 conv2d_6" -> "146 relu_6" [label="(1, 512, 22, 22)", style=solid]; +"146 relu_6" -> "147 quantize_per_tensor_default_7" [label="(1, 512, 22, 22)", style=solid]; +"147 quantize_per_tensor_default_7" -> "148 dequantize_per_tensor_default_7" [label=None, style=solid]; +"148 dequantize_per_tensor_default_7" -> "149 conv2d_7" [label=None, style=solid]; +"149 conv2d_7" -> "150 relu_7" [label="(1, 512, 20, 20)", style=solid]; +"150 relu_7" -> "151 quantize_per_tensor_default_8" [label="(1, 512, 20, 20)", style=solid]; +"151 quantize_per_tensor_default_8" -> "152 dequantize_per_tensor_default_8" [label=None, style=solid]; +"152 dequantize_per_tensor_default_8" -> "153 max_pool2d_3" [label=None, style=solid]; +"152 dequantize_per_tensor_default_8" -> "165 slice_1" [label=None, style=solid]; +"153 max_pool2d_3" -> "154 conv2d_8" [label="(1, 512, 10, 10)", style=solid]; +"154 conv2d_8" -> "155 relu_8" [label="(1, 1024, 8, 8)", style=solid]; +"155 relu_8" -> "156 quantize_per_tensor_default_9" [label="(1, 1024, 8, 8)", style=solid]; +"156 quantize_per_tensor_default_9" -> "157 dequantize_per_tensor_default_9" [label=None, style=solid]; +"157 dequantize_per_tensor_default_9" -> "158 conv2d_9" [label=None, style=solid]; +"158 conv2d_9" -> "159 relu_9" [label="(1, 1024, 6, 6)", style=solid]; +"159 relu_9" -> "160 quantize_per_tensor_default_10" [label="(1, 1024, 6, 6)", style=solid]; +"160 quantize_per_tensor_default_10" -> "161 dequantize_per_tensor_default_10" [label=None, style=solid]; +"161 dequantize_per_tensor_default_10" -> "162 conv_transpose2d" [label=None, style=solid]; +"162 conv_transpose2d" -> "163 quantize_per_tensor_default_11" [label="(1, 512, 12, 12)", style=solid]; +"163 quantize_per_tensor_default_11" -> "164 dequantize_per_tensor_default_11" [label=None, style=solid]; +"164 dequantize_per_tensor_default_11" -> "169 cat" [label=None, style=solid]; +"165 slice_1" -> "166 slice_2" [label="(1, 512, 20, 20)", style=solid]; +"166 slice_2" -> "167 slice_3" [label="(1, 512, 20, 20)", style=solid]; +"167 slice_3" -> "168 slice_4" [label="(1, 512, 12, 20)", style=solid]; +"168 slice_4" -> "169 cat" [label="(1, 512, 12, 12)", style=solid]; +"169 cat" -> "170 conv2d_10" [label="(1, 1024, 12, 12)", style=solid]; +"170 conv2d_10" -> "171 relu_10" [label="(1, 512, 10, 10)", style=solid]; +"171 relu_10" -> "172 quantize_per_tensor_default_12" [label="(1, 512, 10, 10)", style=solid]; +"172 quantize_per_tensor_default_12" -> "173 dequantize_per_tensor_default_12" [label=None, style=solid]; +"173 dequantize_per_tensor_default_12" -> "174 conv2d_11" [label=None, style=solid]; +"174 conv2d_11" -> "175 relu_11" [label="(1, 512, 8, 8)", style=solid]; +"175 relu_11" -> "176 quantize_per_tensor_default_13" [label="(1, 512, 8, 8)", style=solid]; +"176 quantize_per_tensor_default_13" -> "177 dequantize_per_tensor_default_13" [label=None, style=solid]; +"177 dequantize_per_tensor_default_13" -> "178 conv_transpose2d_1" [label=None, style=solid]; +"178 conv_transpose2d_1" -> "179 quantize_per_tensor_default_14" [label="(1, 256, 16, 16)", style=solid]; +"179 quantize_per_tensor_default_14" -> "180 dequantize_per_tensor_default_14" [label=None, style=solid]; +"180 dequantize_per_tensor_default_14" -> "185 cat_1" [label=None, style=solid]; +"181 slice_5" -> "182 slice_6" [label="(1, 256, 49, 49)", style=solid]; +"182 slice_6" -> "183 slice_7" [label="(1, 256, 49, 49)", style=solid]; +"183 slice_7" -> "184 slice_8" [label="(1, 256, 16, 49)", style=solid]; +"184 slice_8" -> "185 cat_1" [label="(1, 256, 16, 16)", style=solid]; +"185 cat_1" -> "186 conv2d_12" [label="(1, 512, 16, 16)", style=solid]; +"186 conv2d_12" -> "187 relu_12" [label="(1, 256, 14, 14)", style=solid]; +"187 relu_12" -> "188 quantize_per_tensor_default_15" [label="(1, 256, 14, 14)", style=solid]; +"188 quantize_per_tensor_default_15" -> "189 dequantize_per_tensor_default_15" [label=None, style=solid]; +"189 dequantize_per_tensor_default_15" -> "190 conv2d_13" [label=None, style=solid]; +"190 conv2d_13" -> "191 relu_13" [label="(1, 256, 12, 12)", style=solid]; +"191 relu_13" -> "192 quantize_per_tensor_default_16" [label="(1, 256, 12, 12)", style=solid]; +"192 quantize_per_tensor_default_16" -> "193 dequantize_per_tensor_default_16" [label=None, style=solid]; +"193 dequantize_per_tensor_default_16" -> "194 conv_transpose2d_2" [label=None, style=solid]; +"194 conv_transpose2d_2" -> "195 quantize_per_tensor_default_17" [label="(1, 128, 24, 24)", style=solid]; +"195 quantize_per_tensor_default_17" -> "196 dequantize_per_tensor_default_17" [label=None, style=solid]; +"196 dequantize_per_tensor_default_17" -> "201 cat_2" [label=None, style=solid]; +"197 slice_9" -> "198 slice_10" [label="(1, 128, 106, 106)", style=solid]; +"198 slice_10" -> "199 slice_11" [label="(1, 128, 106, 106)", style=solid]; +"199 slice_11" -> "200 slice_12" [label="(1, 128, 24, 106)", style=solid]; +"200 slice_12" -> "201 cat_2" [label="(1, 128, 24, 24)", style=solid]; +"201 cat_2" -> "202 conv2d_14" [label="(1, 256, 24, 24)", style=solid]; +"202 conv2d_14" -> "203 relu_14" [label="(1, 128, 22, 22)", style=solid]; +"203 relu_14" -> "204 quantize_per_tensor_default_18" [label="(1, 128, 22, 22)", style=solid]; +"204 quantize_per_tensor_default_18" -> "205 dequantize_per_tensor_default_18" [label=None, style=solid]; +"205 dequantize_per_tensor_default_18" -> "206 conv2d_15" [label=None, style=solid]; +"206 conv2d_15" -> "207 relu_15" [label="(1, 128, 20, 20)", style=solid]; +"207 relu_15" -> "208 quantize_per_tensor_default_19" [label="(1, 128, 20, 20)", style=solid]; +"208 quantize_per_tensor_default_19" -> "209 dequantize_per_tensor_default_19" [label=None, style=solid]; +"209 dequantize_per_tensor_default_19" -> "210 conv_transpose2d_3" [label=None, style=solid]; +"210 conv_transpose2d_3" -> "211 quantize_per_tensor_default_20" [label="(1, 64, 40, 40)", style=solid]; +"211 quantize_per_tensor_default_20" -> "212 dequantize_per_tensor_default_20" [label=None, style=solid]; +"212 dequantize_per_tensor_default_20" -> "217 cat_3" [label=None, style=solid]; +"213 slice_13" -> "214 slice_14" [label="(1, 64, 220, 220)", style=solid]; +"214 slice_14" -> "215 slice_15" [label="(1, 64, 220, 220)", style=solid]; +"215 slice_15" -> "216 slice_16" [label="(1, 64, 40, 220)", style=solid]; +"216 slice_16" -> "217 cat_3" [label="(1, 64, 40, 40)", style=solid]; +"217 cat_3" -> "218 conv2d_16" [label="(1, 128, 40, 40)", style=solid]; +"218 conv2d_16" -> "219 relu_16" [label="(1, 64, 38, 38)", style=solid]; +"219 relu_16" -> "220 quantize_per_tensor_default_21" [label="(1, 64, 38, 38)", style=solid]; +"220 quantize_per_tensor_default_21" -> "221 dequantize_per_tensor_default_21" [label=None, style=solid]; +"221 dequantize_per_tensor_default_21" -> "222 conv2d_17" [label=None, style=solid]; +"222 conv2d_17" -> "223 relu_17" [label="(1, 64, 36, 36)", style=solid]; +"223 relu_17" -> "224 quantize_per_tensor_default_22" [label="(1, 64, 36, 36)", style=solid]; +"224 quantize_per_tensor_default_22" -> "225 dequantize_per_tensor_default_22" [label=None, style=solid]; +"225 dequantize_per_tensor_default_22" -> "226 conv2d_18" [label=None, style=solid]; +"226 conv2d_18" -> "227 output_1" [label="(1, 12, 36, 36)", style=solid]; +} diff --git a/tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/vit_b_16.dot b/tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/vit_b_16.dot new file mode 100644 index 00000000000..d3c40ce0d67 --- /dev/null +++ b/tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/vit_b_16.dot @@ -0,0 +1,1791 @@ +strict digraph { +"0 class_token" [id=0, type=get_attr]; +"1 encoder_pos_embedding" [id=1, type=get_attr]; +"2 _scale_0" [id=2, type=get_attr]; +"3 _zero_point_0" [id=3, type=get_attr]; +"4 _frozen_param0" [id=4, type=get_attr]; +"5 dequantize_per_channel_default" [id=5, type=dequantize_per_channel]; +"6 conv_proj_bias" [id=6, type=get_attr]; +"7 encoder_layers_encoder_layer_0_ln_1_weight" [id=7, type=get_attr]; +"8 encoder_layers_encoder_layer_0_ln_1_bias" [id=8, type=get_attr]; +"9 _scale_1" [id=9, type=get_attr]; +"10 _zero_point_1" [id=10, type=get_attr]; +"11 _frozen_param1" [id=11, type=get_attr]; +"12 dequantize_per_channel_default_1" [id=12, type=dequantize_per_channel]; +"13 encoder_layers_encoder_layer_0_self_attention_in_proj_bias" [id=13, type=get_attr]; +"14 _scale_2" [id=14, type=get_attr]; +"15 _zero_point_2" [id=15, type=get_attr]; +"16 _frozen_param2" [id=16, type=get_attr]; +"17 dequantize_per_channel_default_2" [id=17, type=dequantize_per_channel]; +"18 encoder_layers_encoder_layer_0_self_attention_out_proj_bias" [id=18, type=get_attr]; +"19 encoder_layers_encoder_layer_0_ln_2_weight" [id=19, type=get_attr]; +"20 encoder_layers_encoder_layer_0_ln_2_bias" [id=20, type=get_attr]; +"21 _scale_3" [id=21, type=get_attr]; +"22 _zero_point_3" [id=22, type=get_attr]; +"23 _frozen_param3" [id=23, type=get_attr]; +"24 dequantize_per_channel_default_3" [id=24, type=dequantize_per_channel]; +"25 encoder_layers_encoder_layer_0_mlp_0_bias" [id=25, type=get_attr]; +"26 _scale_4" [id=26, type=get_attr]; +"27 _zero_point_4" [id=27, type=get_attr]; +"28 _frozen_param4" [id=28, type=get_attr]; +"29 dequantize_per_channel_default_4" [id=29, type=dequantize_per_channel]; +"30 encoder_layers_encoder_layer_0_mlp_3_bias" [id=30, type=get_attr]; +"31 encoder_layers_encoder_layer_1_ln_1_weight" [id=31, type=get_attr]; +"32 encoder_layers_encoder_layer_1_ln_1_bias" [id=32, type=get_attr]; +"33 _scale_5" [id=33, type=get_attr]; +"34 _zero_point_5" [id=34, type=get_attr]; +"35 _frozen_param5" [id=35, type=get_attr]; +"36 dequantize_per_channel_default_5" [id=36, type=dequantize_per_channel]; +"37 encoder_layers_encoder_layer_1_self_attention_in_proj_bias" [id=37, type=get_attr]; +"38 _scale_6" [id=38, type=get_attr]; +"39 _zero_point_6" [id=39, type=get_attr]; +"40 _frozen_param6" [id=40, type=get_attr]; +"41 dequantize_per_channel_default_6" [id=41, type=dequantize_per_channel]; +"42 encoder_layers_encoder_layer_1_self_attention_out_proj_bias" [id=42, type=get_attr]; +"43 encoder_layers_encoder_layer_1_ln_2_weight" [id=43, type=get_attr]; +"44 encoder_layers_encoder_layer_1_ln_2_bias" [id=44, type=get_attr]; +"45 _scale_7" [id=45, type=get_attr]; +"46 _zero_point_7" [id=46, type=get_attr]; +"47 _frozen_param7" [id=47, type=get_attr]; +"48 dequantize_per_channel_default_7" [id=48, type=dequantize_per_channel]; +"49 encoder_layers_encoder_layer_1_mlp_0_bias" [id=49, type=get_attr]; +"50 _scale_8" [id=50, type=get_attr]; +"51 _zero_point_8" [id=51, type=get_attr]; +"52 _frozen_param8" [id=52, type=get_attr]; +"53 dequantize_per_channel_default_8" [id=53, type=dequantize_per_channel]; +"54 encoder_layers_encoder_layer_1_mlp_3_bias" [id=54, type=get_attr]; +"55 encoder_layers_encoder_layer_2_ln_1_weight" [id=55, type=get_attr]; +"56 encoder_layers_encoder_layer_2_ln_1_bias" [id=56, type=get_attr]; +"57 _scale_9" [id=57, type=get_attr]; +"58 _zero_point_9" [id=58, type=get_attr]; +"59 _frozen_param9" [id=59, type=get_attr]; +"60 dequantize_per_channel_default_9" [id=60, type=dequantize_per_channel]; +"61 encoder_layers_encoder_layer_2_self_attention_in_proj_bias" [id=61, type=get_attr]; +"62 _scale_10" [id=62, type=get_attr]; +"63 _zero_point_10" [id=63, type=get_attr]; +"64 _frozen_param10" [id=64, type=get_attr]; +"65 dequantize_per_channel_default_10" [id=65, type=dequantize_per_channel]; +"66 encoder_layers_encoder_layer_2_self_attention_out_proj_bias" [id=66, type=get_attr]; +"67 encoder_layers_encoder_layer_2_ln_2_weight" [id=67, type=get_attr]; +"68 encoder_layers_encoder_layer_2_ln_2_bias" [id=68, type=get_attr]; +"69 _scale_11" [id=69, type=get_attr]; +"70 _zero_point_11" [id=70, type=get_attr]; +"71 _frozen_param11" [id=71, type=get_attr]; +"72 dequantize_per_channel_default_11" [id=72, type=dequantize_per_channel]; +"73 encoder_layers_encoder_layer_2_mlp_0_bias" [id=73, type=get_attr]; +"74 _scale_12" [id=74, type=get_attr]; +"75 _zero_point_12" [id=75, type=get_attr]; +"76 _frozen_param12" [id=76, type=get_attr]; +"77 dequantize_per_channel_default_12" [id=77, type=dequantize_per_channel]; +"78 encoder_layers_encoder_layer_2_mlp_3_bias" [id=78, type=get_attr]; +"79 encoder_layers_encoder_layer_3_ln_1_weight" [id=79, type=get_attr]; +"80 encoder_layers_encoder_layer_3_ln_1_bias" [id=80, type=get_attr]; +"81 _scale_13" [id=81, type=get_attr]; +"82 _zero_point_13" [id=82, type=get_attr]; +"83 _frozen_param13" [id=83, type=get_attr]; +"84 dequantize_per_channel_default_13" [id=84, type=dequantize_per_channel]; +"85 encoder_layers_encoder_layer_3_self_attention_in_proj_bias" [id=85, type=get_attr]; +"86 _scale_14" [id=86, type=get_attr]; +"87 _zero_point_14" [id=87, type=get_attr]; +"88 _frozen_param14" [id=88, type=get_attr]; +"89 dequantize_per_channel_default_14" [id=89, type=dequantize_per_channel]; +"90 encoder_layers_encoder_layer_3_self_attention_out_proj_bias" [id=90, type=get_attr]; +"91 encoder_layers_encoder_layer_3_ln_2_weight" [id=91, type=get_attr]; +"92 encoder_layers_encoder_layer_3_ln_2_bias" [id=92, type=get_attr]; +"93 _scale_15" [id=93, type=get_attr]; +"94 _zero_point_15" [id=94, type=get_attr]; +"95 _frozen_param15" [id=95, type=get_attr]; +"96 dequantize_per_channel_default_15" [id=96, type=dequantize_per_channel]; +"97 encoder_layers_encoder_layer_3_mlp_0_bias" [id=97, type=get_attr]; +"98 _scale_16" [id=98, type=get_attr]; +"99 _zero_point_16" [id=99, type=get_attr]; +"100 _frozen_param16" [id=100, type=get_attr]; +"101 dequantize_per_channel_default_16" [id=101, type=dequantize_per_channel]; +"102 encoder_layers_encoder_layer_3_mlp_3_bias" [id=102, type=get_attr]; +"103 encoder_layers_encoder_layer_4_ln_1_weight" [id=103, type=get_attr]; +"104 encoder_layers_encoder_layer_4_ln_1_bias" [id=104, type=get_attr]; +"105 _scale_17" [id=105, type=get_attr]; +"106 _zero_point_17" [id=106, type=get_attr]; +"107 _frozen_param17" [id=107, type=get_attr]; +"108 dequantize_per_channel_default_17" [id=108, type=dequantize_per_channel]; +"109 encoder_layers_encoder_layer_4_self_attention_in_proj_bias" [id=109, type=get_attr]; +"110 _scale_18" [id=110, type=get_attr]; +"111 _zero_point_18" [id=111, type=get_attr]; +"112 _frozen_param18" [id=112, type=get_attr]; +"113 dequantize_per_channel_default_18" [id=113, type=dequantize_per_channel]; +"114 encoder_layers_encoder_layer_4_self_attention_out_proj_bias" [id=114, type=get_attr]; +"115 encoder_layers_encoder_layer_4_ln_2_weight" [id=115, type=get_attr]; +"116 encoder_layers_encoder_layer_4_ln_2_bias" [id=116, type=get_attr]; +"117 _scale_19" [id=117, type=get_attr]; +"118 _zero_point_19" [id=118, type=get_attr]; +"119 _frozen_param19" [id=119, type=get_attr]; +"120 dequantize_per_channel_default_19" [id=120, type=dequantize_per_channel]; +"121 encoder_layers_encoder_layer_4_mlp_0_bias" [id=121, type=get_attr]; +"122 _scale_20" [id=122, type=get_attr]; +"123 _zero_point_20" [id=123, type=get_attr]; +"124 _frozen_param20" [id=124, type=get_attr]; +"125 dequantize_per_channel_default_20" [id=125, type=dequantize_per_channel]; +"126 encoder_layers_encoder_layer_4_mlp_3_bias" [id=126, type=get_attr]; +"127 encoder_layers_encoder_layer_5_ln_1_weight" [id=127, type=get_attr]; +"128 encoder_layers_encoder_layer_5_ln_1_bias" [id=128, type=get_attr]; +"129 _scale_21" [id=129, type=get_attr]; +"130 _zero_point_21" [id=130, type=get_attr]; +"131 _frozen_param21" [id=131, type=get_attr]; +"132 dequantize_per_channel_default_21" [id=132, type=dequantize_per_channel]; +"133 encoder_layers_encoder_layer_5_self_attention_in_proj_bias" [id=133, type=get_attr]; +"134 _scale_22" [id=134, type=get_attr]; +"135 _zero_point_22" [id=135, type=get_attr]; +"136 _frozen_param22" [id=136, type=get_attr]; +"137 dequantize_per_channel_default_22" [id=137, type=dequantize_per_channel]; +"138 encoder_layers_encoder_layer_5_self_attention_out_proj_bias" [id=138, type=get_attr]; +"139 encoder_layers_encoder_layer_5_ln_2_weight" [id=139, type=get_attr]; +"140 encoder_layers_encoder_layer_5_ln_2_bias" [id=140, type=get_attr]; +"141 _scale_23" [id=141, type=get_attr]; +"142 _zero_point_23" [id=142, type=get_attr]; +"143 _frozen_param23" [id=143, type=get_attr]; +"144 dequantize_per_channel_default_23" [id=144, type=dequantize_per_channel]; +"145 encoder_layers_encoder_layer_5_mlp_0_bias" [id=145, type=get_attr]; +"146 _scale_24" [id=146, type=get_attr]; +"147 _zero_point_24" [id=147, type=get_attr]; +"148 _frozen_param24" [id=148, type=get_attr]; +"149 dequantize_per_channel_default_24" [id=149, type=dequantize_per_channel]; +"150 encoder_layers_encoder_layer_5_mlp_3_bias" [id=150, type=get_attr]; +"151 encoder_layers_encoder_layer_6_ln_1_weight" [id=151, type=get_attr]; +"152 encoder_layers_encoder_layer_6_ln_1_bias" [id=152, type=get_attr]; +"153 _scale_25" [id=153, type=get_attr]; +"154 _zero_point_25" [id=154, type=get_attr]; +"155 _frozen_param25" [id=155, type=get_attr]; +"156 dequantize_per_channel_default_25" [id=156, type=dequantize_per_channel]; +"157 encoder_layers_encoder_layer_6_self_attention_in_proj_bias" [id=157, type=get_attr]; +"158 _scale_26" [id=158, type=get_attr]; +"159 _zero_point_26" [id=159, type=get_attr]; +"160 _frozen_param26" [id=160, type=get_attr]; +"161 dequantize_per_channel_default_26" [id=161, type=dequantize_per_channel]; +"162 encoder_layers_encoder_layer_6_self_attention_out_proj_bias" [id=162, type=get_attr]; +"163 encoder_layers_encoder_layer_6_ln_2_weight" [id=163, type=get_attr]; +"164 encoder_layers_encoder_layer_6_ln_2_bias" [id=164, type=get_attr]; +"165 _scale_27" [id=165, type=get_attr]; +"166 _zero_point_27" [id=166, type=get_attr]; +"167 _frozen_param27" [id=167, type=get_attr]; +"168 dequantize_per_channel_default_27" [id=168, type=dequantize_per_channel]; +"169 encoder_layers_encoder_layer_6_mlp_0_bias" [id=169, type=get_attr]; +"170 _scale_28" [id=170, type=get_attr]; +"171 _zero_point_28" [id=171, type=get_attr]; +"172 _frozen_param28" [id=172, type=get_attr]; +"173 dequantize_per_channel_default_28" [id=173, type=dequantize_per_channel]; +"174 encoder_layers_encoder_layer_6_mlp_3_bias" [id=174, type=get_attr]; +"175 encoder_layers_encoder_layer_7_ln_1_weight" [id=175, type=get_attr]; +"176 encoder_layers_encoder_layer_7_ln_1_bias" [id=176, type=get_attr]; +"177 _scale_29" [id=177, type=get_attr]; +"178 _zero_point_29" [id=178, type=get_attr]; +"179 _frozen_param29" [id=179, type=get_attr]; +"180 dequantize_per_channel_default_29" [id=180, type=dequantize_per_channel]; +"181 encoder_layers_encoder_layer_7_self_attention_in_proj_bias" [id=181, type=get_attr]; +"182 _scale_30" [id=182, type=get_attr]; +"183 _zero_point_30" [id=183, type=get_attr]; +"184 _frozen_param30" [id=184, type=get_attr]; +"185 dequantize_per_channel_default_30" [id=185, type=dequantize_per_channel]; +"186 encoder_layers_encoder_layer_7_self_attention_out_proj_bias" [id=186, type=get_attr]; +"187 encoder_layers_encoder_layer_7_ln_2_weight" [id=187, type=get_attr]; +"188 encoder_layers_encoder_layer_7_ln_2_bias" [id=188, type=get_attr]; +"189 _scale_31" [id=189, type=get_attr]; +"190 _zero_point_31" [id=190, type=get_attr]; +"191 _frozen_param31" [id=191, type=get_attr]; +"192 dequantize_per_channel_default_31" [id=192, type=dequantize_per_channel]; +"193 encoder_layers_encoder_layer_7_mlp_0_bias" [id=193, type=get_attr]; +"194 _scale_32" [id=194, type=get_attr]; +"195 _zero_point_32" [id=195, type=get_attr]; +"196 _frozen_param32" [id=196, type=get_attr]; +"197 dequantize_per_channel_default_32" [id=197, type=dequantize_per_channel]; +"198 encoder_layers_encoder_layer_7_mlp_3_bias" [id=198, type=get_attr]; +"199 encoder_layers_encoder_layer_8_ln_1_weight" [id=199, type=get_attr]; +"200 encoder_layers_encoder_layer_8_ln_1_bias" [id=200, type=get_attr]; +"201 _scale_33" [id=201, type=get_attr]; +"202 _zero_point_33" [id=202, type=get_attr]; +"203 _frozen_param33" [id=203, type=get_attr]; +"204 dequantize_per_channel_default_33" [id=204, type=dequantize_per_channel]; +"205 encoder_layers_encoder_layer_8_self_attention_in_proj_bias" [id=205, type=get_attr]; +"206 _scale_34" [id=206, type=get_attr]; +"207 _zero_point_34" [id=207, type=get_attr]; +"208 _frozen_param34" [id=208, type=get_attr]; +"209 dequantize_per_channel_default_34" [id=209, type=dequantize_per_channel]; +"210 encoder_layers_encoder_layer_8_self_attention_out_proj_bias" [id=210, type=get_attr]; +"211 encoder_layers_encoder_layer_8_ln_2_weight" [id=211, type=get_attr]; +"212 encoder_layers_encoder_layer_8_ln_2_bias" [id=212, type=get_attr]; +"213 _scale_35" [id=213, type=get_attr]; +"214 _zero_point_35" [id=214, type=get_attr]; +"215 _frozen_param35" [id=215, type=get_attr]; +"216 dequantize_per_channel_default_35" [id=216, type=dequantize_per_channel]; +"217 encoder_layers_encoder_layer_8_mlp_0_bias" [id=217, type=get_attr]; +"218 _scale_36" [id=218, type=get_attr]; +"219 _zero_point_36" [id=219, type=get_attr]; +"220 _frozen_param36" [id=220, type=get_attr]; +"221 dequantize_per_channel_default_36" [id=221, type=dequantize_per_channel]; +"222 encoder_layers_encoder_layer_8_mlp_3_bias" [id=222, type=get_attr]; +"223 encoder_layers_encoder_layer_9_ln_1_weight" [id=223, type=get_attr]; +"224 encoder_layers_encoder_layer_9_ln_1_bias" [id=224, type=get_attr]; +"225 _scale_37" [id=225, type=get_attr]; +"226 _zero_point_37" [id=226, type=get_attr]; +"227 _frozen_param37" [id=227, type=get_attr]; +"228 dequantize_per_channel_default_37" [id=228, type=dequantize_per_channel]; +"229 encoder_layers_encoder_layer_9_self_attention_in_proj_bias" [id=229, type=get_attr]; +"230 _scale_38" [id=230, type=get_attr]; +"231 _zero_point_38" [id=231, type=get_attr]; +"232 _frozen_param38" [id=232, type=get_attr]; +"233 dequantize_per_channel_default_38" [id=233, type=dequantize_per_channel]; +"234 encoder_layers_encoder_layer_9_self_attention_out_proj_bias" [id=234, type=get_attr]; +"235 encoder_layers_encoder_layer_9_ln_2_weight" [id=235, type=get_attr]; +"236 encoder_layers_encoder_layer_9_ln_2_bias" [id=236, type=get_attr]; +"237 _scale_39" [id=237, type=get_attr]; +"238 _zero_point_39" [id=238, type=get_attr]; +"239 _frozen_param39" [id=239, type=get_attr]; +"240 dequantize_per_channel_default_39" [id=240, type=dequantize_per_channel]; +"241 encoder_layers_encoder_layer_9_mlp_0_bias" [id=241, type=get_attr]; +"242 _scale_40" [id=242, type=get_attr]; +"243 _zero_point_40" [id=243, type=get_attr]; +"244 _frozen_param40" [id=244, type=get_attr]; +"245 dequantize_per_channel_default_40" [id=245, type=dequantize_per_channel]; +"246 encoder_layers_encoder_layer_9_mlp_3_bias" [id=246, type=get_attr]; +"247 encoder_layers_encoder_layer_10_ln_1_weight" [id=247, type=get_attr]; +"248 encoder_layers_encoder_layer_10_ln_1_bias" [id=248, type=get_attr]; +"249 _scale_41" [id=249, type=get_attr]; +"250 _zero_point_41" [id=250, type=get_attr]; +"251 _frozen_param41" [id=251, type=get_attr]; +"252 dequantize_per_channel_default_41" [id=252, type=dequantize_per_channel]; +"253 encoder_layers_encoder_layer_10_self_attention_in_proj_bias" [id=253, type=get_attr]; +"254 _scale_42" [id=254, type=get_attr]; +"255 _zero_point_42" [id=255, type=get_attr]; +"256 _frozen_param42" [id=256, type=get_attr]; +"257 dequantize_per_channel_default_42" [id=257, type=dequantize_per_channel]; +"258 encoder_layers_encoder_layer_10_self_attention_out_proj_bias" [id=258, type=get_attr]; +"259 encoder_layers_encoder_layer_10_ln_2_weight" [id=259, type=get_attr]; +"260 encoder_layers_encoder_layer_10_ln_2_bias" [id=260, type=get_attr]; +"261 _scale_43" [id=261, type=get_attr]; +"262 _zero_point_43" [id=262, type=get_attr]; +"263 _frozen_param43" [id=263, type=get_attr]; +"264 dequantize_per_channel_default_43" [id=264, type=dequantize_per_channel]; +"265 encoder_layers_encoder_layer_10_mlp_0_bias" [id=265, type=get_attr]; +"266 _scale_44" [id=266, type=get_attr]; +"267 _zero_point_44" [id=267, type=get_attr]; +"268 _frozen_param44" [id=268, type=get_attr]; +"269 dequantize_per_channel_default_44" [id=269, type=dequantize_per_channel]; +"270 encoder_layers_encoder_layer_10_mlp_3_bias" [id=270, type=get_attr]; +"271 encoder_layers_encoder_layer_11_ln_1_weight" [id=271, type=get_attr]; +"272 encoder_layers_encoder_layer_11_ln_1_bias" [id=272, type=get_attr]; +"273 _scale_45" [id=273, type=get_attr]; +"274 _zero_point_45" [id=274, type=get_attr]; +"275 _frozen_param45" [id=275, type=get_attr]; +"276 dequantize_per_channel_default_45" [id=276, type=dequantize_per_channel]; +"277 encoder_layers_encoder_layer_11_self_attention_in_proj_bias" [id=277, type=get_attr]; +"278 _scale_46" [id=278, type=get_attr]; +"279 _zero_point_46" [id=279, type=get_attr]; +"280 _frozen_param46" [id=280, type=get_attr]; +"281 dequantize_per_channel_default_46" [id=281, type=dequantize_per_channel]; +"282 encoder_layers_encoder_layer_11_self_attention_out_proj_bias" [id=282, type=get_attr]; +"283 encoder_layers_encoder_layer_11_ln_2_weight" [id=283, type=get_attr]; +"284 encoder_layers_encoder_layer_11_ln_2_bias" [id=284, type=get_attr]; +"285 _scale_47" [id=285, type=get_attr]; +"286 _zero_point_47" [id=286, type=get_attr]; +"287 _frozen_param47" [id=287, type=get_attr]; +"288 dequantize_per_channel_default_47" [id=288, type=dequantize_per_channel]; +"289 encoder_layers_encoder_layer_11_mlp_0_bias" [id=289, type=get_attr]; +"290 _scale_48" [id=290, type=get_attr]; +"291 _zero_point_48" [id=291, type=get_attr]; +"292 _frozen_param48" [id=292, type=get_attr]; +"293 dequantize_per_channel_default_48" [id=293, type=dequantize_per_channel]; +"294 encoder_layers_encoder_layer_11_mlp_3_bias" [id=294, type=get_attr]; +"295 encoder_ln_weight" [id=295, type=get_attr]; +"296 encoder_ln_bias" [id=296, type=get_attr]; +"297 _scale_49" [id=297, type=get_attr]; +"298 _zero_point_49" [id=298, type=get_attr]; +"299 _frozen_param49" [id=299, type=get_attr]; +"300 dequantize_per_channel_default_49" [id=300, type=dequantize_per_channel]; +"301 heads_head_bias" [id=301, type=get_attr]; +"302 x" [id=302, type=input]; +"303 quantize_per_tensor_default" [id=303, type=quantize_per_tensor]; +"304 dequantize_per_tensor_default" [id=304, type=dequantize_per_tensor]; +"305 conv2d" [id=305, type=conv2d]; +"306 reshape" [id=306, type=reshape]; +"307 permute" [id=307, type=permute]; +"308 expand" [id=308, type=expand]; +"309 cat" [id=309, type=cat]; +"310 add" [id=310, type=add]; +"311 dropout" [id=311, type=dropout]; +"312 layer_norm" [id=312, type=layer_norm]; +"313 quantize_per_tensor_default_1" [id=313, type=quantize_per_tensor]; +"314 dequantize_per_tensor_default_1" [id=314, type=dequantize_per_tensor]; +"315 transpose" [id=315, type=transpose]; +"316 linear" [id=316, type=linear]; +"317 unflatten" [id=317, type=unflatten]; +"318 unsqueeze" [id=318, type=unsqueeze]; +"319 transpose_1" [id=319, type=transpose]; +"320 squeeze" [id=320, type=squeeze]; +"321 contiguous" [id=321, type=contiguous]; +"322 quantize_per_tensor_default_2" [id=322, type=quantize_per_tensor]; +"323 dequantize_per_tensor_default_63" [id=323, type=dequantize_per_tensor]; +"324 dequantize_per_tensor_default_62" [id=324, type=dequantize_per_tensor]; +"325 select" [id=325, type=select]; +"326 select_1" [id=326, type=select]; +"327 select_2" [id=327, type=select]; +"328 view" [id=328, type=view]; +"329 transpose_2" [id=329, type=transpose]; +"330 view_1" [id=330, type=view]; +"331 transpose_3" [id=331, type=transpose]; +"332 view_2" [id=332, type=view]; +"333 transpose_4" [id=333, type=transpose]; +"334 view_3" [id=334, type=view]; +"335 view_4" [id=335, type=view]; +"336 view_5" [id=336, type=view]; +"337 scaled_dot_product_attention" [id=337, type=scaled_dot_product_attention]; +"338 quantize_per_tensor_default_3" [id=338, type=quantize_per_tensor]; +"339 dequantize_per_tensor_default_3" [id=339, type=dequantize_per_tensor]; +"340 permute_1" [id=340, type=permute]; +"341 view_6" [id=341, type=view]; +"342 linear_1" [id=342, type=linear]; +"343 view_7" [id=343, type=view]; +"344 transpose_5" [id=344, type=transpose]; +"345 dropout_1" [id=345, type=dropout]; +"346 add_1" [id=346, type=add]; +"347 layer_norm_1" [id=347, type=layer_norm]; +"348 quantize_per_tensor_default_4" [id=348, type=quantize_per_tensor]; +"349 dequantize_per_tensor_default_4" [id=349, type=dequantize_per_tensor]; +"350 linear_2" [id=350, type=linear]; +"351 gelu" [id=351, type=gelu]; +"352 quantize_per_tensor_default_5" [id=352, type=quantize_per_tensor]; +"353 dequantize_per_tensor_default_5" [id=353, type=dequantize_per_tensor]; +"354 dropout_2" [id=354, type=dropout]; +"355 linear_3" [id=355, type=linear]; +"356 dropout_3" [id=356, type=dropout]; +"357 add_2" [id=357, type=add]; +"358 layer_norm_2" [id=358, type=layer_norm]; +"359 quantize_per_tensor_default_6" [id=359, type=quantize_per_tensor]; +"360 dequantize_per_tensor_default_6" [id=360, type=dequantize_per_tensor]; +"361 transpose_6" [id=361, type=transpose]; +"362 linear_4" [id=362, type=linear]; +"363 unflatten_1" [id=363, type=unflatten]; +"364 unsqueeze_1" [id=364, type=unsqueeze]; +"365 transpose_7" [id=365, type=transpose]; +"366 squeeze_1" [id=366, type=squeeze]; +"367 contiguous_1" [id=367, type=contiguous]; +"368 quantize_per_tensor_default_7" [id=368, type=quantize_per_tensor]; +"369 dequantize_per_tensor_default_65" [id=369, type=dequantize_per_tensor]; +"370 dequantize_per_tensor_default_64" [id=370, type=dequantize_per_tensor]; +"371 select_3" [id=371, type=select]; +"372 select_4" [id=372, type=select]; +"373 select_5" [id=373, type=select]; +"374 view_8" [id=374, type=view]; +"375 transpose_8" [id=375, type=transpose]; +"376 view_9" [id=376, type=view]; +"377 transpose_9" [id=377, type=transpose]; +"378 view_10" [id=378, type=view]; +"379 transpose_10" [id=379, type=transpose]; +"380 view_11" [id=380, type=view]; +"381 view_12" [id=381, type=view]; +"382 view_13" [id=382, type=view]; +"383 scaled_dot_product_attention_1" [id=383, type=scaled_dot_product_attention]; +"384 quantize_per_tensor_default_8" [id=384, type=quantize_per_tensor]; +"385 dequantize_per_tensor_default_8" [id=385, type=dequantize_per_tensor]; +"386 permute_2" [id=386, type=permute]; +"387 view_14" [id=387, type=view]; +"388 linear_5" [id=388, type=linear]; +"389 view_15" [id=389, type=view]; +"390 transpose_11" [id=390, type=transpose]; +"391 dropout_4" [id=391, type=dropout]; +"392 add_3" [id=392, type=add]; +"393 layer_norm_3" [id=393, type=layer_norm]; +"394 quantize_per_tensor_default_9" [id=394, type=quantize_per_tensor]; +"395 dequantize_per_tensor_default_9" [id=395, type=dequantize_per_tensor]; +"396 linear_6" [id=396, type=linear]; +"397 gelu_1" [id=397, type=gelu]; +"398 quantize_per_tensor_default_10" [id=398, type=quantize_per_tensor]; +"399 dequantize_per_tensor_default_10" [id=399, type=dequantize_per_tensor]; +"400 dropout_5" [id=400, type=dropout]; +"401 linear_7" [id=401, type=linear]; +"402 dropout_6" [id=402, type=dropout]; +"403 add_4" [id=403, type=add]; +"404 layer_norm_4" [id=404, type=layer_norm]; +"405 quantize_per_tensor_default_11" [id=405, type=quantize_per_tensor]; +"406 dequantize_per_tensor_default_11" [id=406, type=dequantize_per_tensor]; +"407 transpose_12" [id=407, type=transpose]; +"408 linear_8" [id=408, type=linear]; +"409 unflatten_2" [id=409, type=unflatten]; +"410 unsqueeze_2" [id=410, type=unsqueeze]; +"411 transpose_13" [id=411, type=transpose]; +"412 squeeze_2" [id=412, type=squeeze]; +"413 contiguous_2" [id=413, type=contiguous]; +"414 quantize_per_tensor_default_12" [id=414, type=quantize_per_tensor]; +"415 dequantize_per_tensor_default_67" [id=415, type=dequantize_per_tensor]; +"416 dequantize_per_tensor_default_66" [id=416, type=dequantize_per_tensor]; +"417 select_6" [id=417, type=select]; +"418 select_7" [id=418, type=select]; +"419 select_8" [id=419, type=select]; +"420 view_16" [id=420, type=view]; +"421 transpose_14" [id=421, type=transpose]; +"422 view_17" [id=422, type=view]; +"423 transpose_15" [id=423, type=transpose]; +"424 view_18" [id=424, type=view]; +"425 transpose_16" [id=425, type=transpose]; +"426 view_19" [id=426, type=view]; +"427 view_20" [id=427, type=view]; +"428 view_21" [id=428, type=view]; +"429 scaled_dot_product_attention_2" [id=429, type=scaled_dot_product_attention]; +"430 quantize_per_tensor_default_13" [id=430, type=quantize_per_tensor]; +"431 dequantize_per_tensor_default_13" [id=431, type=dequantize_per_tensor]; +"432 permute_3" [id=432, type=permute]; +"433 view_22" [id=433, type=view]; +"434 linear_9" [id=434, type=linear]; +"435 view_23" [id=435, type=view]; +"436 transpose_17" [id=436, type=transpose]; +"437 dropout_7" [id=437, type=dropout]; +"438 add_5" [id=438, type=add]; +"439 layer_norm_5" [id=439, type=layer_norm]; +"440 quantize_per_tensor_default_14" [id=440, type=quantize_per_tensor]; +"441 dequantize_per_tensor_default_14" [id=441, type=dequantize_per_tensor]; +"442 linear_10" [id=442, type=linear]; +"443 gelu_2" [id=443, type=gelu]; +"444 quantize_per_tensor_default_15" [id=444, type=quantize_per_tensor]; +"445 dequantize_per_tensor_default_15" [id=445, type=dequantize_per_tensor]; +"446 dropout_8" [id=446, type=dropout]; +"447 linear_11" [id=447, type=linear]; +"448 dropout_9" [id=448, type=dropout]; +"449 add_6" [id=449, type=add]; +"450 layer_norm_6" [id=450, type=layer_norm]; +"451 quantize_per_tensor_default_16" [id=451, type=quantize_per_tensor]; +"452 dequantize_per_tensor_default_16" [id=452, type=dequantize_per_tensor]; +"453 transpose_18" [id=453, type=transpose]; +"454 linear_12" [id=454, type=linear]; +"455 unflatten_3" [id=455, type=unflatten]; +"456 unsqueeze_3" [id=456, type=unsqueeze]; +"457 transpose_19" [id=457, type=transpose]; +"458 squeeze_3" [id=458, type=squeeze]; +"459 contiguous_3" [id=459, type=contiguous]; +"460 quantize_per_tensor_default_17" [id=460, type=quantize_per_tensor]; +"461 dequantize_per_tensor_default_69" [id=461, type=dequantize_per_tensor]; +"462 dequantize_per_tensor_default_68" [id=462, type=dequantize_per_tensor]; +"463 select_9" [id=463, type=select]; +"464 select_10" [id=464, type=select]; +"465 select_11" [id=465, type=select]; +"466 view_24" [id=466, type=view]; +"467 transpose_20" [id=467, type=transpose]; +"468 view_25" [id=468, type=view]; +"469 transpose_21" [id=469, type=transpose]; +"470 view_26" [id=470, type=view]; +"471 transpose_22" [id=471, type=transpose]; +"472 view_27" [id=472, type=view]; +"473 view_28" [id=473, type=view]; +"474 view_29" [id=474, type=view]; +"475 scaled_dot_product_attention_3" [id=475, type=scaled_dot_product_attention]; +"476 quantize_per_tensor_default_18" [id=476, type=quantize_per_tensor]; +"477 dequantize_per_tensor_default_18" [id=477, type=dequantize_per_tensor]; +"478 permute_4" [id=478, type=permute]; +"479 view_30" [id=479, type=view]; +"480 linear_13" [id=480, type=linear]; +"481 view_31" [id=481, type=view]; +"482 transpose_23" [id=482, type=transpose]; +"483 dropout_10" [id=483, type=dropout]; +"484 add_7" [id=484, type=add]; +"485 layer_norm_7" [id=485, type=layer_norm]; +"486 quantize_per_tensor_default_19" [id=486, type=quantize_per_tensor]; +"487 dequantize_per_tensor_default_19" [id=487, type=dequantize_per_tensor]; +"488 linear_14" [id=488, type=linear]; +"489 gelu_3" [id=489, type=gelu]; +"490 quantize_per_tensor_default_20" [id=490, type=quantize_per_tensor]; +"491 dequantize_per_tensor_default_20" [id=491, type=dequantize_per_tensor]; +"492 dropout_11" [id=492, type=dropout]; +"493 linear_15" [id=493, type=linear]; +"494 dropout_12" [id=494, type=dropout]; +"495 add_8" [id=495, type=add]; +"496 layer_norm_8" [id=496, type=layer_norm]; +"497 quantize_per_tensor_default_21" [id=497, type=quantize_per_tensor]; +"498 dequantize_per_tensor_default_21" [id=498, type=dequantize_per_tensor]; +"499 transpose_24" [id=499, type=transpose]; +"500 linear_16" [id=500, type=linear]; +"501 unflatten_4" [id=501, type=unflatten]; +"502 unsqueeze_4" [id=502, type=unsqueeze]; +"503 transpose_25" [id=503, type=transpose]; +"504 squeeze_4" [id=504, type=squeeze]; +"505 contiguous_4" [id=505, type=contiguous]; +"506 quantize_per_tensor_default_22" [id=506, type=quantize_per_tensor]; +"507 dequantize_per_tensor_default_71" [id=507, type=dequantize_per_tensor]; +"508 dequantize_per_tensor_default_70" [id=508, type=dequantize_per_tensor]; +"509 select_12" [id=509, type=select]; +"510 select_13" [id=510, type=select]; +"511 select_14" [id=511, type=select]; +"512 view_32" [id=512, type=view]; +"513 transpose_26" [id=513, type=transpose]; +"514 view_33" [id=514, type=view]; +"515 transpose_27" [id=515, type=transpose]; +"516 view_34" [id=516, type=view]; +"517 transpose_28" [id=517, type=transpose]; +"518 view_35" [id=518, type=view]; +"519 view_36" [id=519, type=view]; +"520 view_37" [id=520, type=view]; +"521 scaled_dot_product_attention_4" [id=521, type=scaled_dot_product_attention]; +"522 quantize_per_tensor_default_23" [id=522, type=quantize_per_tensor]; +"523 dequantize_per_tensor_default_23" [id=523, type=dequantize_per_tensor]; +"524 permute_5" [id=524, type=permute]; +"525 view_38" [id=525, type=view]; +"526 linear_17" [id=526, type=linear]; +"527 view_39" [id=527, type=view]; +"528 transpose_29" [id=528, type=transpose]; +"529 dropout_13" [id=529, type=dropout]; +"530 add_9" [id=530, type=add]; +"531 layer_norm_9" [id=531, type=layer_norm]; +"532 quantize_per_tensor_default_24" [id=532, type=quantize_per_tensor]; +"533 dequantize_per_tensor_default_24" [id=533, type=dequantize_per_tensor]; +"534 linear_18" [id=534, type=linear]; +"535 gelu_4" [id=535, type=gelu]; +"536 quantize_per_tensor_default_25" [id=536, type=quantize_per_tensor]; +"537 dequantize_per_tensor_default_25" [id=537, type=dequantize_per_tensor]; +"538 dropout_14" [id=538, type=dropout]; +"539 linear_19" [id=539, type=linear]; +"540 dropout_15" [id=540, type=dropout]; +"541 add_10" [id=541, type=add]; +"542 layer_norm_10" [id=542, type=layer_norm]; +"543 quantize_per_tensor_default_26" [id=543, type=quantize_per_tensor]; +"544 dequantize_per_tensor_default_26" [id=544, type=dequantize_per_tensor]; +"545 transpose_30" [id=545, type=transpose]; +"546 linear_20" [id=546, type=linear]; +"547 unflatten_5" [id=547, type=unflatten]; +"548 unsqueeze_5" [id=548, type=unsqueeze]; +"549 transpose_31" [id=549, type=transpose]; +"550 squeeze_5" [id=550, type=squeeze]; +"551 contiguous_5" [id=551, type=contiguous]; +"552 quantize_per_tensor_default_27" [id=552, type=quantize_per_tensor]; +"553 dequantize_per_tensor_default_73" [id=553, type=dequantize_per_tensor]; +"554 dequantize_per_tensor_default_72" [id=554, type=dequantize_per_tensor]; +"555 select_15" [id=555, type=select]; +"556 select_16" [id=556, type=select]; +"557 select_17" [id=557, type=select]; +"558 view_40" [id=558, type=view]; +"559 transpose_32" [id=559, type=transpose]; +"560 view_41" [id=560, type=view]; +"561 transpose_33" [id=561, type=transpose]; +"562 view_42" [id=562, type=view]; +"563 transpose_34" [id=563, type=transpose]; +"564 view_43" [id=564, type=view]; +"565 view_44" [id=565, type=view]; +"566 view_45" [id=566, type=view]; +"567 scaled_dot_product_attention_5" [id=567, type=scaled_dot_product_attention]; +"568 quantize_per_tensor_default_28" [id=568, type=quantize_per_tensor]; +"569 dequantize_per_tensor_default_28" [id=569, type=dequantize_per_tensor]; +"570 permute_6" [id=570, type=permute]; +"571 view_46" [id=571, type=view]; +"572 linear_21" [id=572, type=linear]; +"573 view_47" [id=573, type=view]; +"574 transpose_35" [id=574, type=transpose]; +"575 dropout_16" [id=575, type=dropout]; +"576 add_11" [id=576, type=add]; +"577 layer_norm_11" [id=577, type=layer_norm]; +"578 quantize_per_tensor_default_29" [id=578, type=quantize_per_tensor]; +"579 dequantize_per_tensor_default_29" [id=579, type=dequantize_per_tensor]; +"580 linear_22" [id=580, type=linear]; +"581 gelu_5" [id=581, type=gelu]; +"582 quantize_per_tensor_default_30" [id=582, type=quantize_per_tensor]; +"583 dequantize_per_tensor_default_30" [id=583, type=dequantize_per_tensor]; +"584 dropout_17" [id=584, type=dropout]; +"585 linear_23" [id=585, type=linear]; +"586 dropout_18" [id=586, type=dropout]; +"587 add_12" [id=587, type=add]; +"588 layer_norm_12" [id=588, type=layer_norm]; +"589 quantize_per_tensor_default_31" [id=589, type=quantize_per_tensor]; +"590 dequantize_per_tensor_default_31" [id=590, type=dequantize_per_tensor]; +"591 transpose_36" [id=591, type=transpose]; +"592 linear_24" [id=592, type=linear]; +"593 unflatten_6" [id=593, type=unflatten]; +"594 unsqueeze_6" [id=594, type=unsqueeze]; +"595 transpose_37" [id=595, type=transpose]; +"596 squeeze_6" [id=596, type=squeeze]; +"597 contiguous_6" [id=597, type=contiguous]; +"598 quantize_per_tensor_default_32" [id=598, type=quantize_per_tensor]; +"599 dequantize_per_tensor_default_75" [id=599, type=dequantize_per_tensor]; +"600 dequantize_per_tensor_default_74" [id=600, type=dequantize_per_tensor]; +"601 select_18" [id=601, type=select]; +"602 select_19" [id=602, type=select]; +"603 select_20" [id=603, type=select]; +"604 view_48" [id=604, type=view]; +"605 transpose_38" [id=605, type=transpose]; +"606 view_49" [id=606, type=view]; +"607 transpose_39" [id=607, type=transpose]; +"608 view_50" [id=608, type=view]; +"609 transpose_40" [id=609, type=transpose]; +"610 view_51" [id=610, type=view]; +"611 view_52" [id=611, type=view]; +"612 view_53" [id=612, type=view]; +"613 scaled_dot_product_attention_6" [id=613, type=scaled_dot_product_attention]; +"614 quantize_per_tensor_default_33" [id=614, type=quantize_per_tensor]; +"615 dequantize_per_tensor_default_33" [id=615, type=dequantize_per_tensor]; +"616 permute_7" [id=616, type=permute]; +"617 view_54" [id=617, type=view]; +"618 linear_25" [id=618, type=linear]; +"619 view_55" [id=619, type=view]; +"620 transpose_41" [id=620, type=transpose]; +"621 dropout_19" [id=621, type=dropout]; +"622 add_13" [id=622, type=add]; +"623 layer_norm_13" [id=623, type=layer_norm]; +"624 quantize_per_tensor_default_34" [id=624, type=quantize_per_tensor]; +"625 dequantize_per_tensor_default_34" [id=625, type=dequantize_per_tensor]; +"626 linear_26" [id=626, type=linear]; +"627 gelu_6" [id=627, type=gelu]; +"628 quantize_per_tensor_default_35" [id=628, type=quantize_per_tensor]; +"629 dequantize_per_tensor_default_35" [id=629, type=dequantize_per_tensor]; +"630 dropout_20" [id=630, type=dropout]; +"631 linear_27" [id=631, type=linear]; +"632 dropout_21" [id=632, type=dropout]; +"633 add_14" [id=633, type=add]; +"634 layer_norm_14" [id=634, type=layer_norm]; +"635 quantize_per_tensor_default_36" [id=635, type=quantize_per_tensor]; +"636 dequantize_per_tensor_default_36" [id=636, type=dequantize_per_tensor]; +"637 transpose_42" [id=637, type=transpose]; +"638 linear_28" [id=638, type=linear]; +"639 unflatten_7" [id=639, type=unflatten]; +"640 unsqueeze_7" [id=640, type=unsqueeze]; +"641 transpose_43" [id=641, type=transpose]; +"642 squeeze_7" [id=642, type=squeeze]; +"643 contiguous_7" [id=643, type=contiguous]; +"644 quantize_per_tensor_default_37" [id=644, type=quantize_per_tensor]; +"645 dequantize_per_tensor_default_77" [id=645, type=dequantize_per_tensor]; +"646 dequantize_per_tensor_default_76" [id=646, type=dequantize_per_tensor]; +"647 select_21" [id=647, type=select]; +"648 select_22" [id=648, type=select]; +"649 select_23" [id=649, type=select]; +"650 view_56" [id=650, type=view]; +"651 transpose_44" [id=651, type=transpose]; +"652 view_57" [id=652, type=view]; +"653 transpose_45" [id=653, type=transpose]; +"654 view_58" [id=654, type=view]; +"655 transpose_46" [id=655, type=transpose]; +"656 view_59" [id=656, type=view]; +"657 view_60" [id=657, type=view]; +"658 view_61" [id=658, type=view]; +"659 scaled_dot_product_attention_7" [id=659, type=scaled_dot_product_attention]; +"660 quantize_per_tensor_default_38" [id=660, type=quantize_per_tensor]; +"661 dequantize_per_tensor_default_38" [id=661, type=dequantize_per_tensor]; +"662 permute_8" [id=662, type=permute]; +"663 view_62" [id=663, type=view]; +"664 linear_29" [id=664, type=linear]; +"665 view_63" [id=665, type=view]; +"666 transpose_47" [id=666, type=transpose]; +"667 dropout_22" [id=667, type=dropout]; +"668 add_15" [id=668, type=add]; +"669 layer_norm_15" [id=669, type=layer_norm]; +"670 quantize_per_tensor_default_39" [id=670, type=quantize_per_tensor]; +"671 dequantize_per_tensor_default_39" [id=671, type=dequantize_per_tensor]; +"672 linear_30" [id=672, type=linear]; +"673 gelu_7" [id=673, type=gelu]; +"674 quantize_per_tensor_default_40" [id=674, type=quantize_per_tensor]; +"675 dequantize_per_tensor_default_40" [id=675, type=dequantize_per_tensor]; +"676 dropout_23" [id=676, type=dropout]; +"677 linear_31" [id=677, type=linear]; +"678 dropout_24" [id=678, type=dropout]; +"679 add_16" [id=679, type=add]; +"680 layer_norm_16" [id=680, type=layer_norm]; +"681 quantize_per_tensor_default_41" [id=681, type=quantize_per_tensor]; +"682 dequantize_per_tensor_default_41" [id=682, type=dequantize_per_tensor]; +"683 transpose_48" [id=683, type=transpose]; +"684 linear_32" [id=684, type=linear]; +"685 unflatten_8" [id=685, type=unflatten]; +"686 unsqueeze_8" [id=686, type=unsqueeze]; +"687 transpose_49" [id=687, type=transpose]; +"688 squeeze_8" [id=688, type=squeeze]; +"689 contiguous_8" [id=689, type=contiguous]; +"690 quantize_per_tensor_default_42" [id=690, type=quantize_per_tensor]; +"691 dequantize_per_tensor_default_79" [id=691, type=dequantize_per_tensor]; +"692 dequantize_per_tensor_default_78" [id=692, type=dequantize_per_tensor]; +"693 select_24" [id=693, type=select]; +"694 select_25" [id=694, type=select]; +"695 select_26" [id=695, type=select]; +"696 view_64" [id=696, type=view]; +"697 transpose_50" [id=697, type=transpose]; +"698 view_65" [id=698, type=view]; +"699 transpose_51" [id=699, type=transpose]; +"700 view_66" [id=700, type=view]; +"701 transpose_52" [id=701, type=transpose]; +"702 view_67" [id=702, type=view]; +"703 view_68" [id=703, type=view]; +"704 view_69" [id=704, type=view]; +"705 scaled_dot_product_attention_8" [id=705, type=scaled_dot_product_attention]; +"706 quantize_per_tensor_default_43" [id=706, type=quantize_per_tensor]; +"707 dequantize_per_tensor_default_43" [id=707, type=dequantize_per_tensor]; +"708 permute_9" [id=708, type=permute]; +"709 view_70" [id=709, type=view]; +"710 linear_33" [id=710, type=linear]; +"711 view_71" [id=711, type=view]; +"712 transpose_53" [id=712, type=transpose]; +"713 dropout_25" [id=713, type=dropout]; +"714 add_17" [id=714, type=add]; +"715 layer_norm_17" [id=715, type=layer_norm]; +"716 quantize_per_tensor_default_44" [id=716, type=quantize_per_tensor]; +"717 dequantize_per_tensor_default_44" [id=717, type=dequantize_per_tensor]; +"718 linear_34" [id=718, type=linear]; +"719 gelu_8" [id=719, type=gelu]; +"720 quantize_per_tensor_default_45" [id=720, type=quantize_per_tensor]; +"721 dequantize_per_tensor_default_45" [id=721, type=dequantize_per_tensor]; +"722 dropout_26" [id=722, type=dropout]; +"723 linear_35" [id=723, type=linear]; +"724 dropout_27" [id=724, type=dropout]; +"725 add_18" [id=725, type=add]; +"726 layer_norm_18" [id=726, type=layer_norm]; +"727 quantize_per_tensor_default_46" [id=727, type=quantize_per_tensor]; +"728 dequantize_per_tensor_default_46" [id=728, type=dequantize_per_tensor]; +"729 transpose_54" [id=729, type=transpose]; +"730 linear_36" [id=730, type=linear]; +"731 unflatten_9" [id=731, type=unflatten]; +"732 unsqueeze_9" [id=732, type=unsqueeze]; +"733 transpose_55" [id=733, type=transpose]; +"734 squeeze_9" [id=734, type=squeeze]; +"735 contiguous_9" [id=735, type=contiguous]; +"736 quantize_per_tensor_default_47" [id=736, type=quantize_per_tensor]; +"737 dequantize_per_tensor_default_81" [id=737, type=dequantize_per_tensor]; +"738 dequantize_per_tensor_default_80" [id=738, type=dequantize_per_tensor]; +"739 select_27" [id=739, type=select]; +"740 select_28" [id=740, type=select]; +"741 select_29" [id=741, type=select]; +"742 view_72" [id=742, type=view]; +"743 transpose_56" [id=743, type=transpose]; +"744 view_73" [id=744, type=view]; +"745 transpose_57" [id=745, type=transpose]; +"746 view_74" [id=746, type=view]; +"747 transpose_58" [id=747, type=transpose]; +"748 view_75" [id=748, type=view]; +"749 view_76" [id=749, type=view]; +"750 view_77" [id=750, type=view]; +"751 scaled_dot_product_attention_9" [id=751, type=scaled_dot_product_attention]; +"752 quantize_per_tensor_default_48" [id=752, type=quantize_per_tensor]; +"753 dequantize_per_tensor_default_48" [id=753, type=dequantize_per_tensor]; +"754 permute_10" [id=754, type=permute]; +"755 view_78" [id=755, type=view]; +"756 linear_37" [id=756, type=linear]; +"757 view_79" [id=757, type=view]; +"758 transpose_59" [id=758, type=transpose]; +"759 dropout_28" [id=759, type=dropout]; +"760 add_19" [id=760, type=add]; +"761 layer_norm_19" [id=761, type=layer_norm]; +"762 quantize_per_tensor_default_49" [id=762, type=quantize_per_tensor]; +"763 dequantize_per_tensor_default_49" [id=763, type=dequantize_per_tensor]; +"764 linear_38" [id=764, type=linear]; +"765 gelu_9" [id=765, type=gelu]; +"766 quantize_per_tensor_default_50" [id=766, type=quantize_per_tensor]; +"767 dequantize_per_tensor_default_50" [id=767, type=dequantize_per_tensor]; +"768 dropout_29" [id=768, type=dropout]; +"769 linear_39" [id=769, type=linear]; +"770 dropout_30" [id=770, type=dropout]; +"771 add_20" [id=771, type=add]; +"772 layer_norm_20" [id=772, type=layer_norm]; +"773 quantize_per_tensor_default_51" [id=773, type=quantize_per_tensor]; +"774 dequantize_per_tensor_default_51" [id=774, type=dequantize_per_tensor]; +"775 transpose_60" [id=775, type=transpose]; +"776 linear_40" [id=776, type=linear]; +"777 unflatten_10" [id=777, type=unflatten]; +"778 unsqueeze_10" [id=778, type=unsqueeze]; +"779 transpose_61" [id=779, type=transpose]; +"780 squeeze_10" [id=780, type=squeeze]; +"781 contiguous_10" [id=781, type=contiguous]; +"782 quantize_per_tensor_default_52" [id=782, type=quantize_per_tensor]; +"783 dequantize_per_tensor_default_83" [id=783, type=dequantize_per_tensor]; +"784 dequantize_per_tensor_default_82" [id=784, type=dequantize_per_tensor]; +"785 select_30" [id=785, type=select]; +"786 select_31" [id=786, type=select]; +"787 select_32" [id=787, type=select]; +"788 view_80" [id=788, type=view]; +"789 transpose_62" [id=789, type=transpose]; +"790 view_81" [id=790, type=view]; +"791 transpose_63" [id=791, type=transpose]; +"792 view_82" [id=792, type=view]; +"793 transpose_64" [id=793, type=transpose]; +"794 view_83" [id=794, type=view]; +"795 view_84" [id=795, type=view]; +"796 view_85" [id=796, type=view]; +"797 scaled_dot_product_attention_10" [id=797, type=scaled_dot_product_attention]; +"798 quantize_per_tensor_default_53" [id=798, type=quantize_per_tensor]; +"799 dequantize_per_tensor_default_53" [id=799, type=dequantize_per_tensor]; +"800 permute_11" [id=800, type=permute]; +"801 view_86" [id=801, type=view]; +"802 linear_41" [id=802, type=linear]; +"803 view_87" [id=803, type=view]; +"804 transpose_65" [id=804, type=transpose]; +"805 dropout_31" [id=805, type=dropout]; +"806 add_21" [id=806, type=add]; +"807 layer_norm_21" [id=807, type=layer_norm]; +"808 quantize_per_tensor_default_54" [id=808, type=quantize_per_tensor]; +"809 dequantize_per_tensor_default_54" [id=809, type=dequantize_per_tensor]; +"810 linear_42" [id=810, type=linear]; +"811 gelu_10" [id=811, type=gelu]; +"812 quantize_per_tensor_default_55" [id=812, type=quantize_per_tensor]; +"813 dequantize_per_tensor_default_55" [id=813, type=dequantize_per_tensor]; +"814 dropout_32" [id=814, type=dropout]; +"815 linear_43" [id=815, type=linear]; +"816 dropout_33" [id=816, type=dropout]; +"817 add_22" [id=817, type=add]; +"818 layer_norm_22" [id=818, type=layer_norm]; +"819 quantize_per_tensor_default_56" [id=819, type=quantize_per_tensor]; +"820 dequantize_per_tensor_default_56" [id=820, type=dequantize_per_tensor]; +"821 transpose_66" [id=821, type=transpose]; +"822 linear_44" [id=822, type=linear]; +"823 unflatten_11" [id=823, type=unflatten]; +"824 unsqueeze_11" [id=824, type=unsqueeze]; +"825 transpose_67" [id=825, type=transpose]; +"826 squeeze_11" [id=826, type=squeeze]; +"827 contiguous_11" [id=827, type=contiguous]; +"828 quantize_per_tensor_default_57" [id=828, type=quantize_per_tensor]; +"829 dequantize_per_tensor_default_85" [id=829, type=dequantize_per_tensor]; +"830 dequantize_per_tensor_default_84" [id=830, type=dequantize_per_tensor]; +"831 select_33" [id=831, type=select]; +"832 select_34" [id=832, type=select]; +"833 select_35" [id=833, type=select]; +"834 view_88" [id=834, type=view]; +"835 transpose_68" [id=835, type=transpose]; +"836 view_89" [id=836, type=view]; +"837 transpose_69" [id=837, type=transpose]; +"838 view_90" [id=838, type=view]; +"839 transpose_70" [id=839, type=transpose]; +"840 view_91" [id=840, type=view]; +"841 view_92" [id=841, type=view]; +"842 view_93" [id=842, type=view]; +"843 scaled_dot_product_attention_11" [id=843, type=scaled_dot_product_attention]; +"844 quantize_per_tensor_default_58" [id=844, type=quantize_per_tensor]; +"845 dequantize_per_tensor_default_58" [id=845, type=dequantize_per_tensor]; +"846 permute_12" [id=846, type=permute]; +"847 view_94" [id=847, type=view]; +"848 linear_45" [id=848, type=linear]; +"849 view_95" [id=849, type=view]; +"850 transpose_71" [id=850, type=transpose]; +"851 dropout_34" [id=851, type=dropout]; +"852 add_23" [id=852, type=add]; +"853 layer_norm_23" [id=853, type=layer_norm]; +"854 quantize_per_tensor_default_59" [id=854, type=quantize_per_tensor]; +"855 dequantize_per_tensor_default_59" [id=855, type=dequantize_per_tensor]; +"856 linear_46" [id=856, type=linear]; +"857 gelu_11" [id=857, type=gelu]; +"858 quantize_per_tensor_default_60" [id=858, type=quantize_per_tensor]; +"859 dequantize_per_tensor_default_60" [id=859, type=dequantize_per_tensor]; +"860 dropout_35" [id=860, type=dropout]; +"861 linear_47" [id=861, type=linear]; +"862 dropout_36" [id=862, type=dropout]; +"863 add_24" [id=863, type=add]; +"864 layer_norm_24" [id=864, type=layer_norm]; +"865 quantize_per_tensor_default_61" [id=865, type=quantize_per_tensor]; +"866 dequantize_per_tensor_default_61" [id=866, type=dequantize_per_tensor]; +"867 slice_1" [id=867, type=slice]; +"868 select_36" [id=868, type=select]; +"869 linear_48" [id=869, type=linear]; +"870 output_1" [id=870, type=output]; +"0 class_token" -> "308 expand" [label="(1, 1, 768)", style=solid]; +"1 encoder_pos_embedding" -> "310 add" [label="(1, 197, 768)", style=solid]; +"2 _scale_0" -> "5 dequantize_per_channel_default" [label="(768,)", style=solid]; +"3 _zero_point_0" -> "5 dequantize_per_channel_default" [label="(768,)", style=solid]; +"4 _frozen_param0" -> "5 dequantize_per_channel_default" [label="(768, 3, 16, 16)", style=solid]; +"5 dequantize_per_channel_default" -> "305 conv2d" [label=None, style=solid]; +"6 conv_proj_bias" -> "305 conv2d" [label="(768,)", style=solid]; +"7 encoder_layers_encoder_layer_0_ln_1_weight" -> "312 layer_norm" [label="(768,)", style=solid]; +"8 encoder_layers_encoder_layer_0_ln_1_bias" -> "312 layer_norm" [label="(768,)", style=solid]; +"9 _scale_1" -> "12 dequantize_per_channel_default_1" [label="(2304,)", style=solid]; +"10 _zero_point_1" -> "12 dequantize_per_channel_default_1" [label="(2304,)", style=solid]; +"11 _frozen_param1" -> "12 dequantize_per_channel_default_1" [label="(2304, 768)", style=solid]; +"12 dequantize_per_channel_default_1" -> "316 linear" [label=None, style=solid]; +"13 encoder_layers_encoder_layer_0_self_attention_in_proj_bias" -> "316 linear" [label="(2304,)", style=solid]; +"14 _scale_2" -> "17 dequantize_per_channel_default_2" [label="(768,)", style=solid]; +"15 _zero_point_2" -> "17 dequantize_per_channel_default_2" [label="(768,)", style=solid]; +"16 _frozen_param2" -> "17 dequantize_per_channel_default_2" [label="(768, 768)", style=solid]; +"17 dequantize_per_channel_default_2" -> "342 linear_1" [label=None, style=solid]; +"18 encoder_layers_encoder_layer_0_self_attention_out_proj_bias" -> "342 linear_1" [label="(768,)", style=solid]; +"19 encoder_layers_encoder_layer_0_ln_2_weight" -> "347 layer_norm_1" [label="(768,)", style=solid]; +"20 encoder_layers_encoder_layer_0_ln_2_bias" -> "347 layer_norm_1" [label="(768,)", style=solid]; +"21 _scale_3" -> "24 dequantize_per_channel_default_3" [label="(3072,)", style=solid]; +"22 _zero_point_3" -> "24 dequantize_per_channel_default_3" [label="(3072,)", style=solid]; +"23 _frozen_param3" -> "24 dequantize_per_channel_default_3" [label="(3072, 768)", style=solid]; +"24 dequantize_per_channel_default_3" -> "350 linear_2" [label=None, style=solid]; +"25 encoder_layers_encoder_layer_0_mlp_0_bias" -> "350 linear_2" [label="(3072,)", style=solid]; +"26 _scale_4" -> "29 dequantize_per_channel_default_4" [label="(768,)", style=solid]; +"27 _zero_point_4" -> "29 dequantize_per_channel_default_4" [label="(768,)", style=solid]; +"28 _frozen_param4" -> "29 dequantize_per_channel_default_4" [label="(768, 3072)", style=solid]; +"29 dequantize_per_channel_default_4" -> "355 linear_3" [label=None, style=solid]; +"30 encoder_layers_encoder_layer_0_mlp_3_bias" -> "355 linear_3" [label="(768,)", style=solid]; +"31 encoder_layers_encoder_layer_1_ln_1_weight" -> "358 layer_norm_2" [label="(768,)", style=solid]; +"32 encoder_layers_encoder_layer_1_ln_1_bias" -> "358 layer_norm_2" [label="(768,)", style=solid]; +"33 _scale_5" -> "36 dequantize_per_channel_default_5" [label="(2304,)", style=solid]; +"34 _zero_point_5" -> "36 dequantize_per_channel_default_5" [label="(2304,)", style=solid]; +"35 _frozen_param5" -> "36 dequantize_per_channel_default_5" [label="(2304, 768)", style=solid]; +"36 dequantize_per_channel_default_5" -> "362 linear_4" [label=None, style=solid]; +"37 encoder_layers_encoder_layer_1_self_attention_in_proj_bias" -> "362 linear_4" [label="(2304,)", style=solid]; +"38 _scale_6" -> "41 dequantize_per_channel_default_6" [label="(768,)", style=solid]; +"39 _zero_point_6" -> "41 dequantize_per_channel_default_6" [label="(768,)", style=solid]; +"40 _frozen_param6" -> "41 dequantize_per_channel_default_6" [label="(768, 768)", style=solid]; +"41 dequantize_per_channel_default_6" -> "388 linear_5" [label=None, style=solid]; +"42 encoder_layers_encoder_layer_1_self_attention_out_proj_bias" -> "388 linear_5" [label="(768,)", style=solid]; +"43 encoder_layers_encoder_layer_1_ln_2_weight" -> "393 layer_norm_3" [label="(768,)", style=solid]; +"44 encoder_layers_encoder_layer_1_ln_2_bias" -> "393 layer_norm_3" [label="(768,)", style=solid]; +"45 _scale_7" -> "48 dequantize_per_channel_default_7" [label="(3072,)", style=solid]; +"46 _zero_point_7" -> "48 dequantize_per_channel_default_7" [label="(3072,)", style=solid]; +"47 _frozen_param7" -> "48 dequantize_per_channel_default_7" [label="(3072, 768)", style=solid]; +"48 dequantize_per_channel_default_7" -> "396 linear_6" [label=None, style=solid]; +"49 encoder_layers_encoder_layer_1_mlp_0_bias" -> "396 linear_6" [label="(3072,)", style=solid]; +"50 _scale_8" -> "53 dequantize_per_channel_default_8" [label="(768,)", style=solid]; +"51 _zero_point_8" -> "53 dequantize_per_channel_default_8" [label="(768,)", style=solid]; +"52 _frozen_param8" -> "53 dequantize_per_channel_default_8" [label="(768, 3072)", style=solid]; +"53 dequantize_per_channel_default_8" -> "401 linear_7" [label=None, style=solid]; +"54 encoder_layers_encoder_layer_1_mlp_3_bias" -> "401 linear_7" [label="(768,)", style=solid]; +"55 encoder_layers_encoder_layer_2_ln_1_weight" -> "404 layer_norm_4" [label="(768,)", style=solid]; +"56 encoder_layers_encoder_layer_2_ln_1_bias" -> "404 layer_norm_4" [label="(768,)", style=solid]; +"57 _scale_9" -> "60 dequantize_per_channel_default_9" [label="(2304,)", style=solid]; +"58 _zero_point_9" -> "60 dequantize_per_channel_default_9" [label="(2304,)", style=solid]; +"59 _frozen_param9" -> "60 dequantize_per_channel_default_9" [label="(2304, 768)", style=solid]; +"60 dequantize_per_channel_default_9" -> "408 linear_8" [label=None, style=solid]; +"61 encoder_layers_encoder_layer_2_self_attention_in_proj_bias" -> "408 linear_8" [label="(2304,)", style=solid]; +"62 _scale_10" -> "65 dequantize_per_channel_default_10" [label="(768,)", style=solid]; +"63 _zero_point_10" -> "65 dequantize_per_channel_default_10" [label="(768,)", style=solid]; +"64 _frozen_param10" -> "65 dequantize_per_channel_default_10" [label="(768, 768)", style=solid]; +"65 dequantize_per_channel_default_10" -> "434 linear_9" [label=None, style=solid]; +"66 encoder_layers_encoder_layer_2_self_attention_out_proj_bias" -> "434 linear_9" [label="(768,)", style=solid]; +"67 encoder_layers_encoder_layer_2_ln_2_weight" -> "439 layer_norm_5" [label="(768,)", style=solid]; +"68 encoder_layers_encoder_layer_2_ln_2_bias" -> "439 layer_norm_5" [label="(768,)", style=solid]; +"69 _scale_11" -> "72 dequantize_per_channel_default_11" [label="(3072,)", style=solid]; +"70 _zero_point_11" -> "72 dequantize_per_channel_default_11" [label="(3072,)", style=solid]; +"71 _frozen_param11" -> "72 dequantize_per_channel_default_11" [label="(3072, 768)", style=solid]; +"72 dequantize_per_channel_default_11" -> "442 linear_10" [label=None, style=solid]; +"73 encoder_layers_encoder_layer_2_mlp_0_bias" -> "442 linear_10" [label="(3072,)", style=solid]; +"74 _scale_12" -> "77 dequantize_per_channel_default_12" [label="(768,)", style=solid]; +"75 _zero_point_12" -> "77 dequantize_per_channel_default_12" [label="(768,)", style=solid]; +"76 _frozen_param12" -> "77 dequantize_per_channel_default_12" [label="(768, 3072)", style=solid]; +"77 dequantize_per_channel_default_12" -> "447 linear_11" [label=None, style=solid]; +"78 encoder_layers_encoder_layer_2_mlp_3_bias" -> "447 linear_11" [label="(768,)", style=solid]; +"79 encoder_layers_encoder_layer_3_ln_1_weight" -> "450 layer_norm_6" [label="(768,)", style=solid]; +"80 encoder_layers_encoder_layer_3_ln_1_bias" -> "450 layer_norm_6" [label="(768,)", style=solid]; +"81 _scale_13" -> "84 dequantize_per_channel_default_13" [label="(2304,)", style=solid]; +"82 _zero_point_13" -> "84 dequantize_per_channel_default_13" [label="(2304,)", style=solid]; +"83 _frozen_param13" -> "84 dequantize_per_channel_default_13" [label="(2304, 768)", style=solid]; +"84 dequantize_per_channel_default_13" -> "454 linear_12" [label=None, style=solid]; +"85 encoder_layers_encoder_layer_3_self_attention_in_proj_bias" -> "454 linear_12" [label="(2304,)", style=solid]; +"86 _scale_14" -> "89 dequantize_per_channel_default_14" [label="(768,)", style=solid]; +"87 _zero_point_14" -> "89 dequantize_per_channel_default_14" [label="(768,)", style=solid]; +"88 _frozen_param14" -> "89 dequantize_per_channel_default_14" [label="(768, 768)", style=solid]; +"89 dequantize_per_channel_default_14" -> "480 linear_13" [label=None, style=solid]; +"90 encoder_layers_encoder_layer_3_self_attention_out_proj_bias" -> "480 linear_13" [label="(768,)", style=solid]; +"91 encoder_layers_encoder_layer_3_ln_2_weight" -> "485 layer_norm_7" [label="(768,)", style=solid]; +"92 encoder_layers_encoder_layer_3_ln_2_bias" -> "485 layer_norm_7" [label="(768,)", style=solid]; +"93 _scale_15" -> "96 dequantize_per_channel_default_15" [label="(3072,)", style=solid]; +"94 _zero_point_15" -> "96 dequantize_per_channel_default_15" [label="(3072,)", style=solid]; +"95 _frozen_param15" -> "96 dequantize_per_channel_default_15" [label="(3072, 768)", style=solid]; +"96 dequantize_per_channel_default_15" -> "488 linear_14" [label=None, style=solid]; +"97 encoder_layers_encoder_layer_3_mlp_0_bias" -> "488 linear_14" [label="(3072,)", style=solid]; +"98 _scale_16" -> "101 dequantize_per_channel_default_16" [label="(768,)", style=solid]; +"99 _zero_point_16" -> "101 dequantize_per_channel_default_16" [label="(768,)", style=solid]; +"100 _frozen_param16" -> "101 dequantize_per_channel_default_16" [label="(768, 3072)", style=solid]; +"101 dequantize_per_channel_default_16" -> "493 linear_15" [label=None, style=solid]; +"102 encoder_layers_encoder_layer_3_mlp_3_bias" -> "493 linear_15" [label="(768,)", style=solid]; +"103 encoder_layers_encoder_layer_4_ln_1_weight" -> "496 layer_norm_8" [label="(768,)", style=solid]; +"104 encoder_layers_encoder_layer_4_ln_1_bias" -> "496 layer_norm_8" [label="(768,)", style=solid]; +"105 _scale_17" -> "108 dequantize_per_channel_default_17" [label="(2304,)", style=solid]; +"106 _zero_point_17" -> "108 dequantize_per_channel_default_17" [label="(2304,)", style=solid]; +"107 _frozen_param17" -> "108 dequantize_per_channel_default_17" [label="(2304, 768)", style=solid]; +"108 dequantize_per_channel_default_17" -> "500 linear_16" [label=None, style=solid]; +"109 encoder_layers_encoder_layer_4_self_attention_in_proj_bias" -> "500 linear_16" [label="(2304,)", style=solid]; +"110 _scale_18" -> "113 dequantize_per_channel_default_18" [label="(768,)", style=solid]; +"111 _zero_point_18" -> "113 dequantize_per_channel_default_18" [label="(768,)", style=solid]; +"112 _frozen_param18" -> "113 dequantize_per_channel_default_18" [label="(768, 768)", style=solid]; +"113 dequantize_per_channel_default_18" -> "526 linear_17" [label=None, style=solid]; +"114 encoder_layers_encoder_layer_4_self_attention_out_proj_bias" -> "526 linear_17" [label="(768,)", style=solid]; +"115 encoder_layers_encoder_layer_4_ln_2_weight" -> "531 layer_norm_9" [label="(768,)", style=solid]; +"116 encoder_layers_encoder_layer_4_ln_2_bias" -> "531 layer_norm_9" [label="(768,)", style=solid]; +"117 _scale_19" -> "120 dequantize_per_channel_default_19" [label="(3072,)", style=solid]; +"118 _zero_point_19" -> "120 dequantize_per_channel_default_19" [label="(3072,)", style=solid]; +"119 _frozen_param19" -> "120 dequantize_per_channel_default_19" [label="(3072, 768)", style=solid]; +"120 dequantize_per_channel_default_19" -> "534 linear_18" [label=None, style=solid]; +"121 encoder_layers_encoder_layer_4_mlp_0_bias" -> "534 linear_18" [label="(3072,)", style=solid]; +"122 _scale_20" -> "125 dequantize_per_channel_default_20" [label="(768,)", style=solid]; +"123 _zero_point_20" -> "125 dequantize_per_channel_default_20" [label="(768,)", style=solid]; +"124 _frozen_param20" -> "125 dequantize_per_channel_default_20" [label="(768, 3072)", style=solid]; +"125 dequantize_per_channel_default_20" -> "539 linear_19" [label=None, style=solid]; +"126 encoder_layers_encoder_layer_4_mlp_3_bias" -> "539 linear_19" [label="(768,)", style=solid]; +"127 encoder_layers_encoder_layer_5_ln_1_weight" -> "542 layer_norm_10" [label="(768,)", style=solid]; +"128 encoder_layers_encoder_layer_5_ln_1_bias" -> "542 layer_norm_10" [label="(768,)", style=solid]; +"129 _scale_21" -> "132 dequantize_per_channel_default_21" [label="(2304,)", style=solid]; +"130 _zero_point_21" -> "132 dequantize_per_channel_default_21" [label="(2304,)", style=solid]; +"131 _frozen_param21" -> "132 dequantize_per_channel_default_21" [label="(2304, 768)", style=solid]; +"132 dequantize_per_channel_default_21" -> "546 linear_20" [label=None, style=solid]; +"133 encoder_layers_encoder_layer_5_self_attention_in_proj_bias" -> "546 linear_20" [label="(2304,)", style=solid]; +"134 _scale_22" -> "137 dequantize_per_channel_default_22" [label="(768,)", style=solid]; +"135 _zero_point_22" -> "137 dequantize_per_channel_default_22" [label="(768,)", style=solid]; +"136 _frozen_param22" -> "137 dequantize_per_channel_default_22" [label="(768, 768)", style=solid]; +"137 dequantize_per_channel_default_22" -> "572 linear_21" [label=None, style=solid]; +"138 encoder_layers_encoder_layer_5_self_attention_out_proj_bias" -> "572 linear_21" [label="(768,)", style=solid]; +"139 encoder_layers_encoder_layer_5_ln_2_weight" -> "577 layer_norm_11" [label="(768,)", style=solid]; +"140 encoder_layers_encoder_layer_5_ln_2_bias" -> "577 layer_norm_11" [label="(768,)", style=solid]; +"141 _scale_23" -> "144 dequantize_per_channel_default_23" [label="(3072,)", style=solid]; +"142 _zero_point_23" -> "144 dequantize_per_channel_default_23" [label="(3072,)", style=solid]; +"143 _frozen_param23" -> "144 dequantize_per_channel_default_23" [label="(3072, 768)", style=solid]; +"144 dequantize_per_channel_default_23" -> "580 linear_22" [label=None, style=solid]; +"145 encoder_layers_encoder_layer_5_mlp_0_bias" -> "580 linear_22" [label="(3072,)", style=solid]; +"146 _scale_24" -> "149 dequantize_per_channel_default_24" [label="(768,)", style=solid]; +"147 _zero_point_24" -> "149 dequantize_per_channel_default_24" [label="(768,)", style=solid]; +"148 _frozen_param24" -> "149 dequantize_per_channel_default_24" [label="(768, 3072)", style=solid]; +"149 dequantize_per_channel_default_24" -> "585 linear_23" [label=None, style=solid]; +"150 encoder_layers_encoder_layer_5_mlp_3_bias" -> "585 linear_23" [label="(768,)", style=solid]; +"151 encoder_layers_encoder_layer_6_ln_1_weight" -> "588 layer_norm_12" [label="(768,)", style=solid]; +"152 encoder_layers_encoder_layer_6_ln_1_bias" -> "588 layer_norm_12" [label="(768,)", style=solid]; +"153 _scale_25" -> "156 dequantize_per_channel_default_25" [label="(2304,)", style=solid]; +"154 _zero_point_25" -> "156 dequantize_per_channel_default_25" [label="(2304,)", style=solid]; +"155 _frozen_param25" -> "156 dequantize_per_channel_default_25" [label="(2304, 768)", style=solid]; +"156 dequantize_per_channel_default_25" -> "592 linear_24" [label=None, style=solid]; +"157 encoder_layers_encoder_layer_6_self_attention_in_proj_bias" -> "592 linear_24" [label="(2304,)", style=solid]; +"158 _scale_26" -> "161 dequantize_per_channel_default_26" [label="(768,)", style=solid]; +"159 _zero_point_26" -> "161 dequantize_per_channel_default_26" [label="(768,)", style=solid]; +"160 _frozen_param26" -> "161 dequantize_per_channel_default_26" [label="(768, 768)", style=solid]; +"161 dequantize_per_channel_default_26" -> "618 linear_25" [label=None, style=solid]; +"162 encoder_layers_encoder_layer_6_self_attention_out_proj_bias" -> "618 linear_25" [label="(768,)", style=solid]; +"163 encoder_layers_encoder_layer_6_ln_2_weight" -> "623 layer_norm_13" [label="(768,)", style=solid]; +"164 encoder_layers_encoder_layer_6_ln_2_bias" -> "623 layer_norm_13" [label="(768,)", style=solid]; +"165 _scale_27" -> "168 dequantize_per_channel_default_27" [label="(3072,)", style=solid]; +"166 _zero_point_27" -> "168 dequantize_per_channel_default_27" [label="(3072,)", style=solid]; +"167 _frozen_param27" -> "168 dequantize_per_channel_default_27" [label="(3072, 768)", style=solid]; +"168 dequantize_per_channel_default_27" -> "626 linear_26" [label=None, style=solid]; +"169 encoder_layers_encoder_layer_6_mlp_0_bias" -> "626 linear_26" [label="(3072,)", style=solid]; +"170 _scale_28" -> "173 dequantize_per_channel_default_28" [label="(768,)", style=solid]; +"171 _zero_point_28" -> "173 dequantize_per_channel_default_28" [label="(768,)", style=solid]; +"172 _frozen_param28" -> "173 dequantize_per_channel_default_28" [label="(768, 3072)", style=solid]; +"173 dequantize_per_channel_default_28" -> "631 linear_27" [label=None, style=solid]; +"174 encoder_layers_encoder_layer_6_mlp_3_bias" -> "631 linear_27" [label="(768,)", style=solid]; +"175 encoder_layers_encoder_layer_7_ln_1_weight" -> "634 layer_norm_14" [label="(768,)", style=solid]; +"176 encoder_layers_encoder_layer_7_ln_1_bias" -> "634 layer_norm_14" [label="(768,)", style=solid]; +"177 _scale_29" -> "180 dequantize_per_channel_default_29" [label="(2304,)", style=solid]; +"178 _zero_point_29" -> "180 dequantize_per_channel_default_29" [label="(2304,)", style=solid]; +"179 _frozen_param29" -> "180 dequantize_per_channel_default_29" [label="(2304, 768)", style=solid]; +"180 dequantize_per_channel_default_29" -> "638 linear_28" [label=None, style=solid]; +"181 encoder_layers_encoder_layer_7_self_attention_in_proj_bias" -> "638 linear_28" [label="(2304,)", style=solid]; +"182 _scale_30" -> "185 dequantize_per_channel_default_30" [label="(768,)", style=solid]; +"183 _zero_point_30" -> "185 dequantize_per_channel_default_30" [label="(768,)", style=solid]; +"184 _frozen_param30" -> "185 dequantize_per_channel_default_30" [label="(768, 768)", style=solid]; +"185 dequantize_per_channel_default_30" -> "664 linear_29" [label=None, style=solid]; +"186 encoder_layers_encoder_layer_7_self_attention_out_proj_bias" -> "664 linear_29" [label="(768,)", style=solid]; +"187 encoder_layers_encoder_layer_7_ln_2_weight" -> "669 layer_norm_15" [label="(768,)", style=solid]; +"188 encoder_layers_encoder_layer_7_ln_2_bias" -> "669 layer_norm_15" [label="(768,)", style=solid]; +"189 _scale_31" -> "192 dequantize_per_channel_default_31" [label="(3072,)", style=solid]; +"190 _zero_point_31" -> "192 dequantize_per_channel_default_31" [label="(3072,)", style=solid]; +"191 _frozen_param31" -> "192 dequantize_per_channel_default_31" [label="(3072, 768)", style=solid]; +"192 dequantize_per_channel_default_31" -> "672 linear_30" [label=None, style=solid]; +"193 encoder_layers_encoder_layer_7_mlp_0_bias" -> "672 linear_30" [label="(3072,)", style=solid]; +"194 _scale_32" -> "197 dequantize_per_channel_default_32" [label="(768,)", style=solid]; +"195 _zero_point_32" -> "197 dequantize_per_channel_default_32" [label="(768,)", style=solid]; +"196 _frozen_param32" -> "197 dequantize_per_channel_default_32" [label="(768, 3072)", style=solid]; +"197 dequantize_per_channel_default_32" -> "677 linear_31" [label=None, style=solid]; +"198 encoder_layers_encoder_layer_7_mlp_3_bias" -> "677 linear_31" [label="(768,)", style=solid]; +"199 encoder_layers_encoder_layer_8_ln_1_weight" -> "680 layer_norm_16" [label="(768,)", style=solid]; +"200 encoder_layers_encoder_layer_8_ln_1_bias" -> "680 layer_norm_16" [label="(768,)", style=solid]; +"201 _scale_33" -> "204 dequantize_per_channel_default_33" [label="(2304,)", style=solid]; +"202 _zero_point_33" -> "204 dequantize_per_channel_default_33" [label="(2304,)", style=solid]; +"203 _frozen_param33" -> "204 dequantize_per_channel_default_33" [label="(2304, 768)", style=solid]; +"204 dequantize_per_channel_default_33" -> "684 linear_32" [label=None, style=solid]; +"205 encoder_layers_encoder_layer_8_self_attention_in_proj_bias" -> "684 linear_32" [label="(2304,)", style=solid]; +"206 _scale_34" -> "209 dequantize_per_channel_default_34" [label="(768,)", style=solid]; +"207 _zero_point_34" -> "209 dequantize_per_channel_default_34" [label="(768,)", style=solid]; +"208 _frozen_param34" -> "209 dequantize_per_channel_default_34" [label="(768, 768)", style=solid]; +"209 dequantize_per_channel_default_34" -> "710 linear_33" [label=None, style=solid]; +"210 encoder_layers_encoder_layer_8_self_attention_out_proj_bias" -> "710 linear_33" [label="(768,)", style=solid]; +"211 encoder_layers_encoder_layer_8_ln_2_weight" -> "715 layer_norm_17" [label="(768,)", style=solid]; +"212 encoder_layers_encoder_layer_8_ln_2_bias" -> "715 layer_norm_17" [label="(768,)", style=solid]; +"213 _scale_35" -> "216 dequantize_per_channel_default_35" [label="(3072,)", style=solid]; +"214 _zero_point_35" -> "216 dequantize_per_channel_default_35" [label="(3072,)", style=solid]; +"215 _frozen_param35" -> "216 dequantize_per_channel_default_35" [label="(3072, 768)", style=solid]; +"216 dequantize_per_channel_default_35" -> "718 linear_34" [label=None, style=solid]; +"217 encoder_layers_encoder_layer_8_mlp_0_bias" -> "718 linear_34" [label="(3072,)", style=solid]; +"218 _scale_36" -> "221 dequantize_per_channel_default_36" [label="(768,)", style=solid]; +"219 _zero_point_36" -> "221 dequantize_per_channel_default_36" [label="(768,)", style=solid]; +"220 _frozen_param36" -> "221 dequantize_per_channel_default_36" [label="(768, 3072)", style=solid]; +"221 dequantize_per_channel_default_36" -> "723 linear_35" [label=None, style=solid]; +"222 encoder_layers_encoder_layer_8_mlp_3_bias" -> "723 linear_35" [label="(768,)", style=solid]; +"223 encoder_layers_encoder_layer_9_ln_1_weight" -> "726 layer_norm_18" [label="(768,)", style=solid]; +"224 encoder_layers_encoder_layer_9_ln_1_bias" -> "726 layer_norm_18" [label="(768,)", style=solid]; +"225 _scale_37" -> "228 dequantize_per_channel_default_37" [label="(2304,)", style=solid]; +"226 _zero_point_37" -> "228 dequantize_per_channel_default_37" [label="(2304,)", style=solid]; +"227 _frozen_param37" -> "228 dequantize_per_channel_default_37" [label="(2304, 768)", style=solid]; +"228 dequantize_per_channel_default_37" -> "730 linear_36" [label=None, style=solid]; +"229 encoder_layers_encoder_layer_9_self_attention_in_proj_bias" -> "730 linear_36" [label="(2304,)", style=solid]; +"230 _scale_38" -> "233 dequantize_per_channel_default_38" [label="(768,)", style=solid]; +"231 _zero_point_38" -> "233 dequantize_per_channel_default_38" [label="(768,)", style=solid]; +"232 _frozen_param38" -> "233 dequantize_per_channel_default_38" [label="(768, 768)", style=solid]; +"233 dequantize_per_channel_default_38" -> "756 linear_37" [label=None, style=solid]; +"234 encoder_layers_encoder_layer_9_self_attention_out_proj_bias" -> "756 linear_37" [label="(768,)", style=solid]; +"235 encoder_layers_encoder_layer_9_ln_2_weight" -> "761 layer_norm_19" [label="(768,)", style=solid]; +"236 encoder_layers_encoder_layer_9_ln_2_bias" -> "761 layer_norm_19" [label="(768,)", style=solid]; +"237 _scale_39" -> "240 dequantize_per_channel_default_39" [label="(3072,)", style=solid]; +"238 _zero_point_39" -> "240 dequantize_per_channel_default_39" [label="(3072,)", style=solid]; +"239 _frozen_param39" -> "240 dequantize_per_channel_default_39" [label="(3072, 768)", style=solid]; +"240 dequantize_per_channel_default_39" -> "764 linear_38" [label=None, style=solid]; +"241 encoder_layers_encoder_layer_9_mlp_0_bias" -> "764 linear_38" [label="(3072,)", style=solid]; +"242 _scale_40" -> "245 dequantize_per_channel_default_40" [label="(768,)", style=solid]; +"243 _zero_point_40" -> "245 dequantize_per_channel_default_40" [label="(768,)", style=solid]; +"244 _frozen_param40" -> "245 dequantize_per_channel_default_40" [label="(768, 3072)", style=solid]; +"245 dequantize_per_channel_default_40" -> "769 linear_39" [label=None, style=solid]; +"246 encoder_layers_encoder_layer_9_mlp_3_bias" -> "769 linear_39" [label="(768,)", style=solid]; +"247 encoder_layers_encoder_layer_10_ln_1_weight" -> "772 layer_norm_20" [label="(768,)", style=solid]; +"248 encoder_layers_encoder_layer_10_ln_1_bias" -> "772 layer_norm_20" [label="(768,)", style=solid]; +"249 _scale_41" -> "252 dequantize_per_channel_default_41" [label="(2304,)", style=solid]; +"250 _zero_point_41" -> "252 dequantize_per_channel_default_41" [label="(2304,)", style=solid]; +"251 _frozen_param41" -> "252 dequantize_per_channel_default_41" [label="(2304, 768)", style=solid]; +"252 dequantize_per_channel_default_41" -> "776 linear_40" [label=None, style=solid]; +"253 encoder_layers_encoder_layer_10_self_attention_in_proj_bias" -> "776 linear_40" [label="(2304,)", style=solid]; +"254 _scale_42" -> "257 dequantize_per_channel_default_42" [label="(768,)", style=solid]; +"255 _zero_point_42" -> "257 dequantize_per_channel_default_42" [label="(768,)", style=solid]; +"256 _frozen_param42" -> "257 dequantize_per_channel_default_42" [label="(768, 768)", style=solid]; +"257 dequantize_per_channel_default_42" -> "802 linear_41" [label=None, style=solid]; +"258 encoder_layers_encoder_layer_10_self_attention_out_proj_bias" -> "802 linear_41" [label="(768,)", style=solid]; +"259 encoder_layers_encoder_layer_10_ln_2_weight" -> "807 layer_norm_21" [label="(768,)", style=solid]; +"260 encoder_layers_encoder_layer_10_ln_2_bias" -> "807 layer_norm_21" [label="(768,)", style=solid]; +"261 _scale_43" -> "264 dequantize_per_channel_default_43" [label="(3072,)", style=solid]; +"262 _zero_point_43" -> "264 dequantize_per_channel_default_43" [label="(3072,)", style=solid]; +"263 _frozen_param43" -> "264 dequantize_per_channel_default_43" [label="(3072, 768)", style=solid]; +"264 dequantize_per_channel_default_43" -> "810 linear_42" [label=None, style=solid]; +"265 encoder_layers_encoder_layer_10_mlp_0_bias" -> "810 linear_42" [label="(3072,)", style=solid]; +"266 _scale_44" -> "269 dequantize_per_channel_default_44" [label="(768,)", style=solid]; +"267 _zero_point_44" -> "269 dequantize_per_channel_default_44" [label="(768,)", style=solid]; +"268 _frozen_param44" -> "269 dequantize_per_channel_default_44" [label="(768, 3072)", style=solid]; +"269 dequantize_per_channel_default_44" -> "815 linear_43" [label=None, style=solid]; +"270 encoder_layers_encoder_layer_10_mlp_3_bias" -> "815 linear_43" [label="(768,)", style=solid]; +"271 encoder_layers_encoder_layer_11_ln_1_weight" -> "818 layer_norm_22" [label="(768,)", style=solid]; +"272 encoder_layers_encoder_layer_11_ln_1_bias" -> "818 layer_norm_22" [label="(768,)", style=solid]; +"273 _scale_45" -> "276 dequantize_per_channel_default_45" [label="(2304,)", style=solid]; +"274 _zero_point_45" -> "276 dequantize_per_channel_default_45" [label="(2304,)", style=solid]; +"275 _frozen_param45" -> "276 dequantize_per_channel_default_45" [label="(2304, 768)", style=solid]; +"276 dequantize_per_channel_default_45" -> "822 linear_44" [label=None, style=solid]; +"277 encoder_layers_encoder_layer_11_self_attention_in_proj_bias" -> "822 linear_44" [label="(2304,)", style=solid]; +"278 _scale_46" -> "281 dequantize_per_channel_default_46" [label="(768,)", style=solid]; +"279 _zero_point_46" -> "281 dequantize_per_channel_default_46" [label="(768,)", style=solid]; +"280 _frozen_param46" -> "281 dequantize_per_channel_default_46" [label="(768, 768)", style=solid]; +"281 dequantize_per_channel_default_46" -> "848 linear_45" [label=None, style=solid]; +"282 encoder_layers_encoder_layer_11_self_attention_out_proj_bias" -> "848 linear_45" [label="(768,)", style=solid]; +"283 encoder_layers_encoder_layer_11_ln_2_weight" -> "853 layer_norm_23" [label="(768,)", style=solid]; +"284 encoder_layers_encoder_layer_11_ln_2_bias" -> "853 layer_norm_23" [label="(768,)", style=solid]; +"285 _scale_47" -> "288 dequantize_per_channel_default_47" [label="(3072,)", style=solid]; +"286 _zero_point_47" -> "288 dequantize_per_channel_default_47" [label="(3072,)", style=solid]; +"287 _frozen_param47" -> "288 dequantize_per_channel_default_47" [label="(3072, 768)", style=solid]; +"288 dequantize_per_channel_default_47" -> "856 linear_46" [label=None, style=solid]; +"289 encoder_layers_encoder_layer_11_mlp_0_bias" -> "856 linear_46" [label="(3072,)", style=solid]; +"290 _scale_48" -> "293 dequantize_per_channel_default_48" [label="(768,)", style=solid]; +"291 _zero_point_48" -> "293 dequantize_per_channel_default_48" [label="(768,)", style=solid]; +"292 _frozen_param48" -> "293 dequantize_per_channel_default_48" [label="(768, 3072)", style=solid]; +"293 dequantize_per_channel_default_48" -> "861 linear_47" [label=None, style=solid]; +"294 encoder_layers_encoder_layer_11_mlp_3_bias" -> "861 linear_47" [label="(768,)", style=solid]; +"295 encoder_ln_weight" -> "864 layer_norm_24" [label="(768,)", style=solid]; +"296 encoder_ln_bias" -> "864 layer_norm_24" [label="(768,)", style=solid]; +"297 _scale_49" -> "300 dequantize_per_channel_default_49" [label="(1000,)", style=solid]; +"298 _zero_point_49" -> "300 dequantize_per_channel_default_49" [label="(1000,)", style=solid]; +"299 _frozen_param49" -> "300 dequantize_per_channel_default_49" [label="(1000, 768)", style=solid]; +"300 dequantize_per_channel_default_49" -> "869 linear_48" [label=None, style=solid]; +"301 heads_head_bias" -> "869 linear_48" [label="(1000,)", style=solid]; +"302 x" -> "303 quantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; +"303 quantize_per_tensor_default" -> "304 dequantize_per_tensor_default" [label=None, style=solid]; +"304 dequantize_per_tensor_default" -> "305 conv2d" [label=None, style=solid]; +"305 conv2d" -> "306 reshape" [label="(1, 768, 14, 14)", style=solid]; +"306 reshape" -> "307 permute" [label="(1, 768, 196)", style=solid]; +"307 permute" -> "309 cat" [label="(1, 196, 768)", style=solid]; +"308 expand" -> "309 cat" [label="(1, 1, 768)", style=solid]; +"309 cat" -> "310 add" [label="(1, 197, 768)", style=solid]; +"310 add" -> "311 dropout" [label="(1, 197, 768)", style=solid]; +"311 dropout" -> "312 layer_norm" [label="(1, 197, 768)", style=solid]; +"311 dropout" -> "346 add_1" [label="(1, 197, 768)", style=solid]; +"312 layer_norm" -> "313 quantize_per_tensor_default_1" [label="(1, 197, 768)", style=solid]; +"313 quantize_per_tensor_default_1" -> "314 dequantize_per_tensor_default_1" [label=None, style=solid]; +"314 dequantize_per_tensor_default_1" -> "315 transpose" [label=None, style=solid]; +"315 transpose" -> "316 linear" [label="(197, 1, 768)", style=solid]; +"316 linear" -> "317 unflatten" [label="(197, 1, 2304)", style=solid]; +"317 unflatten" -> "318 unsqueeze" [label="(197, 1, 3, 768)", style=solid]; +"318 unsqueeze" -> "319 transpose_1" [label="(1, 197, 1, 3, 768)", style=solid]; +"319 transpose_1" -> "320 squeeze" [label="(3, 197, 1, 1, 768)", style=solid]; +"320 squeeze" -> "321 contiguous" [label="(3, 197, 1, 768)", style=solid]; +"321 contiguous" -> "322 quantize_per_tensor_default_2" [label="(3, 197, 1, 768)", style=solid]; +"321 contiguous" -> "327 select_2" [label="(3, 197, 1, 768)", style=solid]; +"322 quantize_per_tensor_default_2" -> "323 dequantize_per_tensor_default_63" [label=None, style=solid]; +"322 quantize_per_tensor_default_2" -> "324 dequantize_per_tensor_default_62" [label=None, style=solid]; +"323 dequantize_per_tensor_default_63" -> "326 select_1" [label=None, style=solid]; +"324 dequantize_per_tensor_default_62" -> "325 select" [label=None, style=solid]; +"325 select" -> "328 view" [label="(197, 1, 768)", style=solid]; +"326 select_1" -> "330 view_1" [label="(197, 1, 768)", style=solid]; +"327 select_2" -> "332 view_2" [label="(197, 1, 768)", style=solid]; +"328 view" -> "329 transpose_2" [label="(197, 12, 64)", style=solid]; +"329 transpose_2" -> "334 view_3" [label="(12, 197, 64)", style=solid]; +"330 view_1" -> "331 transpose_3" [label="(197, 12, 64)", style=solid]; +"331 transpose_3" -> "335 view_4" [label="(12, 197, 64)", style=solid]; +"332 view_2" -> "333 transpose_4" [label="(197, 12, 64)", style=solid]; +"333 transpose_4" -> "336 view_5" [label="(12, 197, 64)", style=solid]; +"334 view_3" -> "337 scaled_dot_product_attention" [label="(1, 12, 197, 64)", style=solid]; +"335 view_4" -> "337 scaled_dot_product_attention" [label="(1, 12, 197, 64)", style=solid]; +"336 view_5" -> "337 scaled_dot_product_attention" [label="(1, 12, 197, 64)", style=solid]; +"337 scaled_dot_product_attention" -> "338 quantize_per_tensor_default_3" [label="(1, 12, 197, 64)", style=solid]; +"338 quantize_per_tensor_default_3" -> "339 dequantize_per_tensor_default_3" [label=None, style=solid]; +"339 dequantize_per_tensor_default_3" -> "340 permute_1" [label=None, style=solid]; +"340 permute_1" -> "341 view_6" [label="(197, 1, 12, 64)", style=solid]; +"341 view_6" -> "342 linear_1" [label="(197, 768)", style=solid]; +"342 linear_1" -> "343 view_7" [label="(197, 768)", style=solid]; +"343 view_7" -> "344 transpose_5" [label="(197, 1, 768)", style=solid]; +"344 transpose_5" -> "345 dropout_1" [label="(1, 197, 768)", style=solid]; +"345 dropout_1" -> "346 add_1" [label="(1, 197, 768)", style=solid]; +"346 add_1" -> "347 layer_norm_1" [label="(1, 197, 768)", style=solid]; +"346 add_1" -> "357 add_2" [label="(1, 197, 768)", style=solid]; +"347 layer_norm_1" -> "348 quantize_per_tensor_default_4" [label="(1, 197, 768)", style=solid]; +"348 quantize_per_tensor_default_4" -> "349 dequantize_per_tensor_default_4" [label=None, style=solid]; +"349 dequantize_per_tensor_default_4" -> "350 linear_2" [label=None, style=solid]; +"350 linear_2" -> "351 gelu" [label="(1, 197, 3072)", style=solid]; +"351 gelu" -> "352 quantize_per_tensor_default_5" [label="(1, 197, 3072)", style=solid]; +"352 quantize_per_tensor_default_5" -> "353 dequantize_per_tensor_default_5" [label=None, style=solid]; +"353 dequantize_per_tensor_default_5" -> "354 dropout_2" [label=None, style=solid]; +"354 dropout_2" -> "355 linear_3" [label="(1, 197, 3072)", style=solid]; +"355 linear_3" -> "356 dropout_3" [label="(1, 197, 768)", style=solid]; +"356 dropout_3" -> "357 add_2" [label="(1, 197, 768)", style=solid]; +"357 add_2" -> "358 layer_norm_2" [label="(1, 197, 768)", style=solid]; +"357 add_2" -> "392 add_3" [label="(1, 197, 768)", style=solid]; +"358 layer_norm_2" -> "359 quantize_per_tensor_default_6" [label="(1, 197, 768)", style=solid]; +"359 quantize_per_tensor_default_6" -> "360 dequantize_per_tensor_default_6" [label=None, style=solid]; +"360 dequantize_per_tensor_default_6" -> "361 transpose_6" [label=None, style=solid]; +"361 transpose_6" -> "362 linear_4" [label="(197, 1, 768)", style=solid]; +"362 linear_4" -> "363 unflatten_1" [label="(197, 1, 2304)", style=solid]; +"363 unflatten_1" -> "364 unsqueeze_1" [label="(197, 1, 3, 768)", style=solid]; +"364 unsqueeze_1" -> "365 transpose_7" [label="(1, 197, 1, 3, 768)", style=solid]; +"365 transpose_7" -> "366 squeeze_1" [label="(3, 197, 1, 1, 768)", style=solid]; +"366 squeeze_1" -> "367 contiguous_1" [label="(3, 197, 1, 768)", style=solid]; +"367 contiguous_1" -> "368 quantize_per_tensor_default_7" [label="(3, 197, 1, 768)", style=solid]; +"367 contiguous_1" -> "373 select_5" [label="(3, 197, 1, 768)", style=solid]; +"368 quantize_per_tensor_default_7" -> "369 dequantize_per_tensor_default_65" [label=None, style=solid]; +"368 quantize_per_tensor_default_7" -> "370 dequantize_per_tensor_default_64" [label=None, style=solid]; +"369 dequantize_per_tensor_default_65" -> "372 select_4" [label=None, style=solid]; +"370 dequantize_per_tensor_default_64" -> "371 select_3" [label=None, style=solid]; +"371 select_3" -> "374 view_8" [label="(197, 1, 768)", style=solid]; +"372 select_4" -> "376 view_9" [label="(197, 1, 768)", style=solid]; +"373 select_5" -> "378 view_10" [label="(197, 1, 768)", style=solid]; +"374 view_8" -> "375 transpose_8" [label="(197, 12, 64)", style=solid]; +"375 transpose_8" -> "380 view_11" [label="(12, 197, 64)", style=solid]; +"376 view_9" -> "377 transpose_9" [label="(197, 12, 64)", style=solid]; +"377 transpose_9" -> "381 view_12" [label="(12, 197, 64)", style=solid]; +"378 view_10" -> "379 transpose_10" [label="(197, 12, 64)", style=solid]; +"379 transpose_10" -> "382 view_13" [label="(12, 197, 64)", style=solid]; +"380 view_11" -> "383 scaled_dot_product_attention_1" [label="(1, 12, 197, 64)", style=solid]; +"381 view_12" -> "383 scaled_dot_product_attention_1" [label="(1, 12, 197, 64)", style=solid]; +"382 view_13" -> "383 scaled_dot_product_attention_1" [label="(1, 12, 197, 64)", style=solid]; +"383 scaled_dot_product_attention_1" -> "384 quantize_per_tensor_default_8" [label="(1, 12, 197, 64)", style=solid]; +"384 quantize_per_tensor_default_8" -> "385 dequantize_per_tensor_default_8" [label=None, style=solid]; +"385 dequantize_per_tensor_default_8" -> "386 permute_2" [label=None, style=solid]; +"386 permute_2" -> "387 view_14" [label="(197, 1, 12, 64)", style=solid]; +"387 view_14" -> "388 linear_5" [label="(197, 768)", style=solid]; +"388 linear_5" -> "389 view_15" [label="(197, 768)", style=solid]; +"389 view_15" -> "390 transpose_11" [label="(197, 1, 768)", style=solid]; +"390 transpose_11" -> "391 dropout_4" [label="(1, 197, 768)", style=solid]; +"391 dropout_4" -> "392 add_3" [label="(1, 197, 768)", style=solid]; +"392 add_3" -> "393 layer_norm_3" [label="(1, 197, 768)", style=solid]; +"392 add_3" -> "403 add_4" [label="(1, 197, 768)", style=solid]; +"393 layer_norm_3" -> "394 quantize_per_tensor_default_9" [label="(1, 197, 768)", style=solid]; +"394 quantize_per_tensor_default_9" -> "395 dequantize_per_tensor_default_9" [label=None, style=solid]; +"395 dequantize_per_tensor_default_9" -> "396 linear_6" [label=None, style=solid]; +"396 linear_6" -> "397 gelu_1" [label="(1, 197, 3072)", style=solid]; +"397 gelu_1" -> "398 quantize_per_tensor_default_10" [label="(1, 197, 3072)", style=solid]; +"398 quantize_per_tensor_default_10" -> "399 dequantize_per_tensor_default_10" [label=None, style=solid]; +"399 dequantize_per_tensor_default_10" -> "400 dropout_5" [label=None, style=solid]; +"400 dropout_5" -> "401 linear_7" [label="(1, 197, 3072)", style=solid]; +"401 linear_7" -> "402 dropout_6" [label="(1, 197, 768)", style=solid]; +"402 dropout_6" -> "403 add_4" [label="(1, 197, 768)", style=solid]; +"403 add_4" -> "404 layer_norm_4" [label="(1, 197, 768)", style=solid]; +"403 add_4" -> "438 add_5" [label="(1, 197, 768)", style=solid]; +"404 layer_norm_4" -> "405 quantize_per_tensor_default_11" [label="(1, 197, 768)", style=solid]; +"405 quantize_per_tensor_default_11" -> "406 dequantize_per_tensor_default_11" [label=None, style=solid]; +"406 dequantize_per_tensor_default_11" -> "407 transpose_12" [label=None, style=solid]; +"407 transpose_12" -> "408 linear_8" [label="(197, 1, 768)", style=solid]; +"408 linear_8" -> "409 unflatten_2" [label="(197, 1, 2304)", style=solid]; +"409 unflatten_2" -> "410 unsqueeze_2" [label="(197, 1, 3, 768)", style=solid]; +"410 unsqueeze_2" -> "411 transpose_13" [label="(1, 197, 1, 3, 768)", style=solid]; +"411 transpose_13" -> "412 squeeze_2" [label="(3, 197, 1, 1, 768)", style=solid]; +"412 squeeze_2" -> "413 contiguous_2" [label="(3, 197, 1, 768)", style=solid]; +"413 contiguous_2" -> "414 quantize_per_tensor_default_12" [label="(3, 197, 1, 768)", style=solid]; +"413 contiguous_2" -> "419 select_8" [label="(3, 197, 1, 768)", style=solid]; +"414 quantize_per_tensor_default_12" -> "415 dequantize_per_tensor_default_67" [label=None, style=solid]; +"414 quantize_per_tensor_default_12" -> "416 dequantize_per_tensor_default_66" [label=None, style=solid]; +"415 dequantize_per_tensor_default_67" -> "418 select_7" [label=None, style=solid]; +"416 dequantize_per_tensor_default_66" -> "417 select_6" [label=None, style=solid]; +"417 select_6" -> "420 view_16" [label="(197, 1, 768)", style=solid]; +"418 select_7" -> "422 view_17" [label="(197, 1, 768)", style=solid]; +"419 select_8" -> "424 view_18" [label="(197, 1, 768)", style=solid]; +"420 view_16" -> "421 transpose_14" [label="(197, 12, 64)", style=solid]; +"421 transpose_14" -> "426 view_19" [label="(12, 197, 64)", style=solid]; +"422 view_17" -> "423 transpose_15" [label="(197, 12, 64)", style=solid]; +"423 transpose_15" -> "427 view_20" [label="(12, 197, 64)", style=solid]; +"424 view_18" -> "425 transpose_16" [label="(197, 12, 64)", style=solid]; +"425 transpose_16" -> "428 view_21" [label="(12, 197, 64)", style=solid]; +"426 view_19" -> "429 scaled_dot_product_attention_2" [label="(1, 12, 197, 64)", style=solid]; +"427 view_20" -> "429 scaled_dot_product_attention_2" [label="(1, 12, 197, 64)", style=solid]; +"428 view_21" -> "429 scaled_dot_product_attention_2" [label="(1, 12, 197, 64)", style=solid]; +"429 scaled_dot_product_attention_2" -> "430 quantize_per_tensor_default_13" [label="(1, 12, 197, 64)", style=solid]; +"430 quantize_per_tensor_default_13" -> "431 dequantize_per_tensor_default_13" [label=None, style=solid]; +"431 dequantize_per_tensor_default_13" -> "432 permute_3" [label=None, style=solid]; +"432 permute_3" -> "433 view_22" [label="(197, 1, 12, 64)", style=solid]; +"433 view_22" -> "434 linear_9" [label="(197, 768)", style=solid]; +"434 linear_9" -> "435 view_23" [label="(197, 768)", style=solid]; +"435 view_23" -> "436 transpose_17" [label="(197, 1, 768)", style=solid]; +"436 transpose_17" -> "437 dropout_7" [label="(1, 197, 768)", style=solid]; +"437 dropout_7" -> "438 add_5" [label="(1, 197, 768)", style=solid]; +"438 add_5" -> "439 layer_norm_5" [label="(1, 197, 768)", style=solid]; +"438 add_5" -> "449 add_6" [label="(1, 197, 768)", style=solid]; +"439 layer_norm_5" -> "440 quantize_per_tensor_default_14" [label="(1, 197, 768)", style=solid]; +"440 quantize_per_tensor_default_14" -> "441 dequantize_per_tensor_default_14" [label=None, style=solid]; +"441 dequantize_per_tensor_default_14" -> "442 linear_10" [label=None, style=solid]; +"442 linear_10" -> "443 gelu_2" [label="(1, 197, 3072)", style=solid]; +"443 gelu_2" -> "444 quantize_per_tensor_default_15" [label="(1, 197, 3072)", style=solid]; +"444 quantize_per_tensor_default_15" -> "445 dequantize_per_tensor_default_15" [label=None, style=solid]; +"445 dequantize_per_tensor_default_15" -> "446 dropout_8" [label=None, style=solid]; +"446 dropout_8" -> "447 linear_11" [label="(1, 197, 3072)", style=solid]; +"447 linear_11" -> "448 dropout_9" [label="(1, 197, 768)", style=solid]; +"448 dropout_9" -> "449 add_6" [label="(1, 197, 768)", style=solid]; +"449 add_6" -> "450 layer_norm_6" [label="(1, 197, 768)", style=solid]; +"449 add_6" -> "484 add_7" [label="(1, 197, 768)", style=solid]; +"450 layer_norm_6" -> "451 quantize_per_tensor_default_16" [label="(1, 197, 768)", style=solid]; +"451 quantize_per_tensor_default_16" -> "452 dequantize_per_tensor_default_16" [label=None, style=solid]; +"452 dequantize_per_tensor_default_16" -> "453 transpose_18" [label=None, style=solid]; +"453 transpose_18" -> "454 linear_12" [label="(197, 1, 768)", style=solid]; +"454 linear_12" -> "455 unflatten_3" [label="(197, 1, 2304)", style=solid]; +"455 unflatten_3" -> "456 unsqueeze_3" [label="(197, 1, 3, 768)", style=solid]; +"456 unsqueeze_3" -> "457 transpose_19" [label="(1, 197, 1, 3, 768)", style=solid]; +"457 transpose_19" -> "458 squeeze_3" [label="(3, 197, 1, 1, 768)", style=solid]; +"458 squeeze_3" -> "459 contiguous_3" [label="(3, 197, 1, 768)", style=solid]; +"459 contiguous_3" -> "460 quantize_per_tensor_default_17" [label="(3, 197, 1, 768)", style=solid]; +"459 contiguous_3" -> "465 select_11" [label="(3, 197, 1, 768)", style=solid]; +"460 quantize_per_tensor_default_17" -> "461 dequantize_per_tensor_default_69" [label=None, style=solid]; +"460 quantize_per_tensor_default_17" -> "462 dequantize_per_tensor_default_68" [label=None, style=solid]; +"461 dequantize_per_tensor_default_69" -> "464 select_10" [label=None, style=solid]; +"462 dequantize_per_tensor_default_68" -> "463 select_9" [label=None, style=solid]; +"463 select_9" -> "466 view_24" [label="(197, 1, 768)", style=solid]; +"464 select_10" -> "468 view_25" [label="(197, 1, 768)", style=solid]; +"465 select_11" -> "470 view_26" [label="(197, 1, 768)", style=solid]; +"466 view_24" -> "467 transpose_20" [label="(197, 12, 64)", style=solid]; +"467 transpose_20" -> "472 view_27" [label="(12, 197, 64)", style=solid]; +"468 view_25" -> "469 transpose_21" [label="(197, 12, 64)", style=solid]; +"469 transpose_21" -> "473 view_28" [label="(12, 197, 64)", style=solid]; +"470 view_26" -> "471 transpose_22" [label="(197, 12, 64)", style=solid]; +"471 transpose_22" -> "474 view_29" [label="(12, 197, 64)", style=solid]; +"472 view_27" -> "475 scaled_dot_product_attention_3" [label="(1, 12, 197, 64)", style=solid]; +"473 view_28" -> "475 scaled_dot_product_attention_3" [label="(1, 12, 197, 64)", style=solid]; +"474 view_29" -> "475 scaled_dot_product_attention_3" [label="(1, 12, 197, 64)", style=solid]; +"475 scaled_dot_product_attention_3" -> "476 quantize_per_tensor_default_18" [label="(1, 12, 197, 64)", style=solid]; +"476 quantize_per_tensor_default_18" -> "477 dequantize_per_tensor_default_18" [label=None, style=solid]; +"477 dequantize_per_tensor_default_18" -> "478 permute_4" [label=None, style=solid]; +"478 permute_4" -> "479 view_30" [label="(197, 1, 12, 64)", style=solid]; +"479 view_30" -> "480 linear_13" [label="(197, 768)", style=solid]; +"480 linear_13" -> "481 view_31" [label="(197, 768)", style=solid]; +"481 view_31" -> "482 transpose_23" [label="(197, 1, 768)", style=solid]; +"482 transpose_23" -> "483 dropout_10" [label="(1, 197, 768)", style=solid]; +"483 dropout_10" -> "484 add_7" [label="(1, 197, 768)", style=solid]; +"484 add_7" -> "485 layer_norm_7" [label="(1, 197, 768)", style=solid]; +"484 add_7" -> "495 add_8" [label="(1, 197, 768)", style=solid]; +"485 layer_norm_7" -> "486 quantize_per_tensor_default_19" [label="(1, 197, 768)", style=solid]; +"486 quantize_per_tensor_default_19" -> "487 dequantize_per_tensor_default_19" [label=None, style=solid]; +"487 dequantize_per_tensor_default_19" -> "488 linear_14" [label=None, style=solid]; +"488 linear_14" -> "489 gelu_3" [label="(1, 197, 3072)", style=solid]; +"489 gelu_3" -> "490 quantize_per_tensor_default_20" [label="(1, 197, 3072)", style=solid]; +"490 quantize_per_tensor_default_20" -> "491 dequantize_per_tensor_default_20" [label=None, style=solid]; +"491 dequantize_per_tensor_default_20" -> "492 dropout_11" [label=None, style=solid]; +"492 dropout_11" -> "493 linear_15" [label="(1, 197, 3072)", style=solid]; +"493 linear_15" -> "494 dropout_12" [label="(1, 197, 768)", style=solid]; +"494 dropout_12" -> "495 add_8" [label="(1, 197, 768)", style=solid]; +"495 add_8" -> "496 layer_norm_8" [label="(1, 197, 768)", style=solid]; +"495 add_8" -> "530 add_9" [label="(1, 197, 768)", style=solid]; +"496 layer_norm_8" -> "497 quantize_per_tensor_default_21" [label="(1, 197, 768)", style=solid]; +"497 quantize_per_tensor_default_21" -> "498 dequantize_per_tensor_default_21" [label=None, style=solid]; +"498 dequantize_per_tensor_default_21" -> "499 transpose_24" [label=None, style=solid]; +"499 transpose_24" -> "500 linear_16" [label="(197, 1, 768)", style=solid]; +"500 linear_16" -> "501 unflatten_4" [label="(197, 1, 2304)", style=solid]; +"501 unflatten_4" -> "502 unsqueeze_4" [label="(197, 1, 3, 768)", style=solid]; +"502 unsqueeze_4" -> "503 transpose_25" [label="(1, 197, 1, 3, 768)", style=solid]; +"503 transpose_25" -> "504 squeeze_4" [label="(3, 197, 1, 1, 768)", style=solid]; +"504 squeeze_4" -> "505 contiguous_4" [label="(3, 197, 1, 768)", style=solid]; +"505 contiguous_4" -> "506 quantize_per_tensor_default_22" [label="(3, 197, 1, 768)", style=solid]; +"505 contiguous_4" -> "511 select_14" [label="(3, 197, 1, 768)", style=solid]; +"506 quantize_per_tensor_default_22" -> "507 dequantize_per_tensor_default_71" [label=None, style=solid]; +"506 quantize_per_tensor_default_22" -> "508 dequantize_per_tensor_default_70" [label=None, style=solid]; +"507 dequantize_per_tensor_default_71" -> "510 select_13" [label=None, style=solid]; +"508 dequantize_per_tensor_default_70" -> "509 select_12" [label=None, style=solid]; +"509 select_12" -> "512 view_32" [label="(197, 1, 768)", style=solid]; +"510 select_13" -> "514 view_33" [label="(197, 1, 768)", style=solid]; +"511 select_14" -> "516 view_34" [label="(197, 1, 768)", style=solid]; +"512 view_32" -> "513 transpose_26" [label="(197, 12, 64)", style=solid]; +"513 transpose_26" -> "518 view_35" [label="(12, 197, 64)", style=solid]; +"514 view_33" -> "515 transpose_27" [label="(197, 12, 64)", style=solid]; +"515 transpose_27" -> "519 view_36" [label="(12, 197, 64)", style=solid]; +"516 view_34" -> "517 transpose_28" [label="(197, 12, 64)", style=solid]; +"517 transpose_28" -> "520 view_37" [label="(12, 197, 64)", style=solid]; +"518 view_35" -> "521 scaled_dot_product_attention_4" [label="(1, 12, 197, 64)", style=solid]; +"519 view_36" -> "521 scaled_dot_product_attention_4" [label="(1, 12, 197, 64)", style=solid]; +"520 view_37" -> "521 scaled_dot_product_attention_4" [label="(1, 12, 197, 64)", style=solid]; +"521 scaled_dot_product_attention_4" -> "522 quantize_per_tensor_default_23" [label="(1, 12, 197, 64)", style=solid]; +"522 quantize_per_tensor_default_23" -> "523 dequantize_per_tensor_default_23" [label=None, style=solid]; +"523 dequantize_per_tensor_default_23" -> "524 permute_5" [label=None, style=solid]; +"524 permute_5" -> "525 view_38" [label="(197, 1, 12, 64)", style=solid]; +"525 view_38" -> "526 linear_17" [label="(197, 768)", style=solid]; +"526 linear_17" -> "527 view_39" [label="(197, 768)", style=solid]; +"527 view_39" -> "528 transpose_29" [label="(197, 1, 768)", style=solid]; +"528 transpose_29" -> "529 dropout_13" [label="(1, 197, 768)", style=solid]; +"529 dropout_13" -> "530 add_9" [label="(1, 197, 768)", style=solid]; +"530 add_9" -> "531 layer_norm_9" [label="(1, 197, 768)", style=solid]; +"530 add_9" -> "541 add_10" [label="(1, 197, 768)", style=solid]; +"531 layer_norm_9" -> "532 quantize_per_tensor_default_24" [label="(1, 197, 768)", style=solid]; +"532 quantize_per_tensor_default_24" -> "533 dequantize_per_tensor_default_24" [label=None, style=solid]; +"533 dequantize_per_tensor_default_24" -> "534 linear_18" [label=None, style=solid]; +"534 linear_18" -> "535 gelu_4" [label="(1, 197, 3072)", style=solid]; +"535 gelu_4" -> "536 quantize_per_tensor_default_25" [label="(1, 197, 3072)", style=solid]; +"536 quantize_per_tensor_default_25" -> "537 dequantize_per_tensor_default_25" [label=None, style=solid]; +"537 dequantize_per_tensor_default_25" -> "538 dropout_14" [label=None, style=solid]; +"538 dropout_14" -> "539 linear_19" [label="(1, 197, 3072)", style=solid]; +"539 linear_19" -> "540 dropout_15" [label="(1, 197, 768)", style=solid]; +"540 dropout_15" -> "541 add_10" [label="(1, 197, 768)", style=solid]; +"541 add_10" -> "542 layer_norm_10" [label="(1, 197, 768)", style=solid]; +"541 add_10" -> "576 add_11" [label="(1, 197, 768)", style=solid]; +"542 layer_norm_10" -> "543 quantize_per_tensor_default_26" [label="(1, 197, 768)", style=solid]; +"543 quantize_per_tensor_default_26" -> "544 dequantize_per_tensor_default_26" [label=None, style=solid]; +"544 dequantize_per_tensor_default_26" -> "545 transpose_30" [label=None, style=solid]; +"545 transpose_30" -> "546 linear_20" [label="(197, 1, 768)", style=solid]; +"546 linear_20" -> "547 unflatten_5" [label="(197, 1, 2304)", style=solid]; +"547 unflatten_5" -> "548 unsqueeze_5" [label="(197, 1, 3, 768)", style=solid]; +"548 unsqueeze_5" -> "549 transpose_31" [label="(1, 197, 1, 3, 768)", style=solid]; +"549 transpose_31" -> "550 squeeze_5" [label="(3, 197, 1, 1, 768)", style=solid]; +"550 squeeze_5" -> "551 contiguous_5" [label="(3, 197, 1, 768)", style=solid]; +"551 contiguous_5" -> "552 quantize_per_tensor_default_27" [label="(3, 197, 1, 768)", style=solid]; +"551 contiguous_5" -> "557 select_17" [label="(3, 197, 1, 768)", style=solid]; +"552 quantize_per_tensor_default_27" -> "553 dequantize_per_tensor_default_73" [label=None, style=solid]; +"552 quantize_per_tensor_default_27" -> "554 dequantize_per_tensor_default_72" [label=None, style=solid]; +"553 dequantize_per_tensor_default_73" -> "556 select_16" [label=None, style=solid]; +"554 dequantize_per_tensor_default_72" -> "555 select_15" [label=None, style=solid]; +"555 select_15" -> "558 view_40" [label="(197, 1, 768)", style=solid]; +"556 select_16" -> "560 view_41" [label="(197, 1, 768)", style=solid]; +"557 select_17" -> "562 view_42" [label="(197, 1, 768)", style=solid]; +"558 view_40" -> "559 transpose_32" [label="(197, 12, 64)", style=solid]; +"559 transpose_32" -> "564 view_43" [label="(12, 197, 64)", style=solid]; +"560 view_41" -> "561 transpose_33" [label="(197, 12, 64)", style=solid]; +"561 transpose_33" -> "565 view_44" [label="(12, 197, 64)", style=solid]; +"562 view_42" -> "563 transpose_34" [label="(197, 12, 64)", style=solid]; +"563 transpose_34" -> "566 view_45" [label="(12, 197, 64)", style=solid]; +"564 view_43" -> "567 scaled_dot_product_attention_5" [label="(1, 12, 197, 64)", style=solid]; +"565 view_44" -> "567 scaled_dot_product_attention_5" [label="(1, 12, 197, 64)", style=solid]; +"566 view_45" -> "567 scaled_dot_product_attention_5" [label="(1, 12, 197, 64)", style=solid]; +"567 scaled_dot_product_attention_5" -> "568 quantize_per_tensor_default_28" [label="(1, 12, 197, 64)", style=solid]; +"568 quantize_per_tensor_default_28" -> "569 dequantize_per_tensor_default_28" [label=None, style=solid]; +"569 dequantize_per_tensor_default_28" -> "570 permute_6" [label=None, style=solid]; +"570 permute_6" -> "571 view_46" [label="(197, 1, 12, 64)", style=solid]; +"571 view_46" -> "572 linear_21" [label="(197, 768)", style=solid]; +"572 linear_21" -> "573 view_47" [label="(197, 768)", style=solid]; +"573 view_47" -> "574 transpose_35" [label="(197, 1, 768)", style=solid]; +"574 transpose_35" -> "575 dropout_16" [label="(1, 197, 768)", style=solid]; +"575 dropout_16" -> "576 add_11" [label="(1, 197, 768)", style=solid]; +"576 add_11" -> "577 layer_norm_11" [label="(1, 197, 768)", style=solid]; +"576 add_11" -> "587 add_12" [label="(1, 197, 768)", style=solid]; +"577 layer_norm_11" -> "578 quantize_per_tensor_default_29" [label="(1, 197, 768)", style=solid]; +"578 quantize_per_tensor_default_29" -> "579 dequantize_per_tensor_default_29" [label=None, style=solid]; +"579 dequantize_per_tensor_default_29" -> "580 linear_22" [label=None, style=solid]; +"580 linear_22" -> "581 gelu_5" [label="(1, 197, 3072)", style=solid]; +"581 gelu_5" -> "582 quantize_per_tensor_default_30" [label="(1, 197, 3072)", style=solid]; +"582 quantize_per_tensor_default_30" -> "583 dequantize_per_tensor_default_30" [label=None, style=solid]; +"583 dequantize_per_tensor_default_30" -> "584 dropout_17" [label=None, style=solid]; +"584 dropout_17" -> "585 linear_23" [label="(1, 197, 3072)", style=solid]; +"585 linear_23" -> "586 dropout_18" [label="(1, 197, 768)", style=solid]; +"586 dropout_18" -> "587 add_12" [label="(1, 197, 768)", style=solid]; +"587 add_12" -> "588 layer_norm_12" [label="(1, 197, 768)", style=solid]; +"587 add_12" -> "622 add_13" [label="(1, 197, 768)", style=solid]; +"588 layer_norm_12" -> "589 quantize_per_tensor_default_31" [label="(1, 197, 768)", style=solid]; +"589 quantize_per_tensor_default_31" -> "590 dequantize_per_tensor_default_31" [label=None, style=solid]; +"590 dequantize_per_tensor_default_31" -> "591 transpose_36" [label=None, style=solid]; +"591 transpose_36" -> "592 linear_24" [label="(197, 1, 768)", style=solid]; +"592 linear_24" -> "593 unflatten_6" [label="(197, 1, 2304)", style=solid]; +"593 unflatten_6" -> "594 unsqueeze_6" [label="(197, 1, 3, 768)", style=solid]; +"594 unsqueeze_6" -> "595 transpose_37" [label="(1, 197, 1, 3, 768)", style=solid]; +"595 transpose_37" -> "596 squeeze_6" [label="(3, 197, 1, 1, 768)", style=solid]; +"596 squeeze_6" -> "597 contiguous_6" [label="(3, 197, 1, 768)", style=solid]; +"597 contiguous_6" -> "598 quantize_per_tensor_default_32" [label="(3, 197, 1, 768)", style=solid]; +"597 contiguous_6" -> "603 select_20" [label="(3, 197, 1, 768)", style=solid]; +"598 quantize_per_tensor_default_32" -> "599 dequantize_per_tensor_default_75" [label=None, style=solid]; +"598 quantize_per_tensor_default_32" -> "600 dequantize_per_tensor_default_74" [label=None, style=solid]; +"599 dequantize_per_tensor_default_75" -> "602 select_19" [label=None, style=solid]; +"600 dequantize_per_tensor_default_74" -> "601 select_18" [label=None, style=solid]; +"601 select_18" -> "604 view_48" [label="(197, 1, 768)", style=solid]; +"602 select_19" -> "606 view_49" [label="(197, 1, 768)", style=solid]; +"603 select_20" -> "608 view_50" [label="(197, 1, 768)", style=solid]; +"604 view_48" -> "605 transpose_38" [label="(197, 12, 64)", style=solid]; +"605 transpose_38" -> "610 view_51" [label="(12, 197, 64)", style=solid]; +"606 view_49" -> "607 transpose_39" [label="(197, 12, 64)", style=solid]; +"607 transpose_39" -> "611 view_52" [label="(12, 197, 64)", style=solid]; +"608 view_50" -> "609 transpose_40" [label="(197, 12, 64)", style=solid]; +"609 transpose_40" -> "612 view_53" [label="(12, 197, 64)", style=solid]; +"610 view_51" -> "613 scaled_dot_product_attention_6" [label="(1, 12, 197, 64)", style=solid]; +"611 view_52" -> "613 scaled_dot_product_attention_6" [label="(1, 12, 197, 64)", style=solid]; +"612 view_53" -> "613 scaled_dot_product_attention_6" [label="(1, 12, 197, 64)", style=solid]; +"613 scaled_dot_product_attention_6" -> "614 quantize_per_tensor_default_33" [label="(1, 12, 197, 64)", style=solid]; +"614 quantize_per_tensor_default_33" -> "615 dequantize_per_tensor_default_33" [label=None, style=solid]; +"615 dequantize_per_tensor_default_33" -> "616 permute_7" [label=None, style=solid]; +"616 permute_7" -> "617 view_54" [label="(197, 1, 12, 64)", style=solid]; +"617 view_54" -> "618 linear_25" [label="(197, 768)", style=solid]; +"618 linear_25" -> "619 view_55" [label="(197, 768)", style=solid]; +"619 view_55" -> "620 transpose_41" [label="(197, 1, 768)", style=solid]; +"620 transpose_41" -> "621 dropout_19" [label="(1, 197, 768)", style=solid]; +"621 dropout_19" -> "622 add_13" [label="(1, 197, 768)", style=solid]; +"622 add_13" -> "623 layer_norm_13" [label="(1, 197, 768)", style=solid]; +"622 add_13" -> "633 add_14" [label="(1, 197, 768)", style=solid]; +"623 layer_norm_13" -> "624 quantize_per_tensor_default_34" [label="(1, 197, 768)", style=solid]; +"624 quantize_per_tensor_default_34" -> "625 dequantize_per_tensor_default_34" [label=None, style=solid]; +"625 dequantize_per_tensor_default_34" -> "626 linear_26" [label=None, style=solid]; +"626 linear_26" -> "627 gelu_6" [label="(1, 197, 3072)", style=solid]; +"627 gelu_6" -> "628 quantize_per_tensor_default_35" [label="(1, 197, 3072)", style=solid]; +"628 quantize_per_tensor_default_35" -> "629 dequantize_per_tensor_default_35" [label=None, style=solid]; +"629 dequantize_per_tensor_default_35" -> "630 dropout_20" [label=None, style=solid]; +"630 dropout_20" -> "631 linear_27" [label="(1, 197, 3072)", style=solid]; +"631 linear_27" -> "632 dropout_21" [label="(1, 197, 768)", style=solid]; +"632 dropout_21" -> "633 add_14" [label="(1, 197, 768)", style=solid]; +"633 add_14" -> "634 layer_norm_14" [label="(1, 197, 768)", style=solid]; +"633 add_14" -> "668 add_15" [label="(1, 197, 768)", style=solid]; +"634 layer_norm_14" -> "635 quantize_per_tensor_default_36" [label="(1, 197, 768)", style=solid]; +"635 quantize_per_tensor_default_36" -> "636 dequantize_per_tensor_default_36" [label=None, style=solid]; +"636 dequantize_per_tensor_default_36" -> "637 transpose_42" [label=None, style=solid]; +"637 transpose_42" -> "638 linear_28" [label="(197, 1, 768)", style=solid]; +"638 linear_28" -> "639 unflatten_7" [label="(197, 1, 2304)", style=solid]; +"639 unflatten_7" -> "640 unsqueeze_7" [label="(197, 1, 3, 768)", style=solid]; +"640 unsqueeze_7" -> "641 transpose_43" [label="(1, 197, 1, 3, 768)", style=solid]; +"641 transpose_43" -> "642 squeeze_7" [label="(3, 197, 1, 1, 768)", style=solid]; +"642 squeeze_7" -> "643 contiguous_7" [label="(3, 197, 1, 768)", style=solid]; +"643 contiguous_7" -> "644 quantize_per_tensor_default_37" [label="(3, 197, 1, 768)", style=solid]; +"643 contiguous_7" -> "649 select_23" [label="(3, 197, 1, 768)", style=solid]; +"644 quantize_per_tensor_default_37" -> "645 dequantize_per_tensor_default_77" [label=None, style=solid]; +"644 quantize_per_tensor_default_37" -> "646 dequantize_per_tensor_default_76" [label=None, style=solid]; +"645 dequantize_per_tensor_default_77" -> "648 select_22" [label=None, style=solid]; +"646 dequantize_per_tensor_default_76" -> "647 select_21" [label=None, style=solid]; +"647 select_21" -> "650 view_56" [label="(197, 1, 768)", style=solid]; +"648 select_22" -> "652 view_57" [label="(197, 1, 768)", style=solid]; +"649 select_23" -> "654 view_58" [label="(197, 1, 768)", style=solid]; +"650 view_56" -> "651 transpose_44" [label="(197, 12, 64)", style=solid]; +"651 transpose_44" -> "656 view_59" [label="(12, 197, 64)", style=solid]; +"652 view_57" -> "653 transpose_45" [label="(197, 12, 64)", style=solid]; +"653 transpose_45" -> "657 view_60" [label="(12, 197, 64)", style=solid]; +"654 view_58" -> "655 transpose_46" [label="(197, 12, 64)", style=solid]; +"655 transpose_46" -> "658 view_61" [label="(12, 197, 64)", style=solid]; +"656 view_59" -> "659 scaled_dot_product_attention_7" [label="(1, 12, 197, 64)", style=solid]; +"657 view_60" -> "659 scaled_dot_product_attention_7" [label="(1, 12, 197, 64)", style=solid]; +"658 view_61" -> "659 scaled_dot_product_attention_7" [label="(1, 12, 197, 64)", style=solid]; +"659 scaled_dot_product_attention_7" -> "660 quantize_per_tensor_default_38" [label="(1, 12, 197, 64)", style=solid]; +"660 quantize_per_tensor_default_38" -> "661 dequantize_per_tensor_default_38" [label=None, style=solid]; +"661 dequantize_per_tensor_default_38" -> "662 permute_8" [label=None, style=solid]; +"662 permute_8" -> "663 view_62" [label="(197, 1, 12, 64)", style=solid]; +"663 view_62" -> "664 linear_29" [label="(197, 768)", style=solid]; +"664 linear_29" -> "665 view_63" [label="(197, 768)", style=solid]; +"665 view_63" -> "666 transpose_47" [label="(197, 1, 768)", style=solid]; +"666 transpose_47" -> "667 dropout_22" [label="(1, 197, 768)", style=solid]; +"667 dropout_22" -> "668 add_15" [label="(1, 197, 768)", style=solid]; +"668 add_15" -> "669 layer_norm_15" [label="(1, 197, 768)", style=solid]; +"668 add_15" -> "679 add_16" [label="(1, 197, 768)", style=solid]; +"669 layer_norm_15" -> "670 quantize_per_tensor_default_39" [label="(1, 197, 768)", style=solid]; +"670 quantize_per_tensor_default_39" -> "671 dequantize_per_tensor_default_39" [label=None, style=solid]; +"671 dequantize_per_tensor_default_39" -> "672 linear_30" [label=None, style=solid]; +"672 linear_30" -> "673 gelu_7" [label="(1, 197, 3072)", style=solid]; +"673 gelu_7" -> "674 quantize_per_tensor_default_40" [label="(1, 197, 3072)", style=solid]; +"674 quantize_per_tensor_default_40" -> "675 dequantize_per_tensor_default_40" [label=None, style=solid]; +"675 dequantize_per_tensor_default_40" -> "676 dropout_23" [label=None, style=solid]; +"676 dropout_23" -> "677 linear_31" [label="(1, 197, 3072)", style=solid]; +"677 linear_31" -> "678 dropout_24" [label="(1, 197, 768)", style=solid]; +"678 dropout_24" -> "679 add_16" [label="(1, 197, 768)", style=solid]; +"679 add_16" -> "680 layer_norm_16" [label="(1, 197, 768)", style=solid]; +"679 add_16" -> "714 add_17" [label="(1, 197, 768)", style=solid]; +"680 layer_norm_16" -> "681 quantize_per_tensor_default_41" [label="(1, 197, 768)", style=solid]; +"681 quantize_per_tensor_default_41" -> "682 dequantize_per_tensor_default_41" [label=None, style=solid]; +"682 dequantize_per_tensor_default_41" -> "683 transpose_48" [label=None, style=solid]; +"683 transpose_48" -> "684 linear_32" [label="(197, 1, 768)", style=solid]; +"684 linear_32" -> "685 unflatten_8" [label="(197, 1, 2304)", style=solid]; +"685 unflatten_8" -> "686 unsqueeze_8" [label="(197, 1, 3, 768)", style=solid]; +"686 unsqueeze_8" -> "687 transpose_49" [label="(1, 197, 1, 3, 768)", style=solid]; +"687 transpose_49" -> "688 squeeze_8" [label="(3, 197, 1, 1, 768)", style=solid]; +"688 squeeze_8" -> "689 contiguous_8" [label="(3, 197, 1, 768)", style=solid]; +"689 contiguous_8" -> "690 quantize_per_tensor_default_42" [label="(3, 197, 1, 768)", style=solid]; +"689 contiguous_8" -> "695 select_26" [label="(3, 197, 1, 768)", style=solid]; +"690 quantize_per_tensor_default_42" -> "691 dequantize_per_tensor_default_79" [label=None, style=solid]; +"690 quantize_per_tensor_default_42" -> "692 dequantize_per_tensor_default_78" [label=None, style=solid]; +"691 dequantize_per_tensor_default_79" -> "694 select_25" [label=None, style=solid]; +"692 dequantize_per_tensor_default_78" -> "693 select_24" [label=None, style=solid]; +"693 select_24" -> "696 view_64" [label="(197, 1, 768)", style=solid]; +"694 select_25" -> "698 view_65" [label="(197, 1, 768)", style=solid]; +"695 select_26" -> "700 view_66" [label="(197, 1, 768)", style=solid]; +"696 view_64" -> "697 transpose_50" [label="(197, 12, 64)", style=solid]; +"697 transpose_50" -> "702 view_67" [label="(12, 197, 64)", style=solid]; +"698 view_65" -> "699 transpose_51" [label="(197, 12, 64)", style=solid]; +"699 transpose_51" -> "703 view_68" [label="(12, 197, 64)", style=solid]; +"700 view_66" -> "701 transpose_52" [label="(197, 12, 64)", style=solid]; +"701 transpose_52" -> "704 view_69" [label="(12, 197, 64)", style=solid]; +"702 view_67" -> "705 scaled_dot_product_attention_8" [label="(1, 12, 197, 64)", style=solid]; +"703 view_68" -> "705 scaled_dot_product_attention_8" [label="(1, 12, 197, 64)", style=solid]; +"704 view_69" -> "705 scaled_dot_product_attention_8" [label="(1, 12, 197, 64)", style=solid]; +"705 scaled_dot_product_attention_8" -> "706 quantize_per_tensor_default_43" [label="(1, 12, 197, 64)", style=solid]; +"706 quantize_per_tensor_default_43" -> "707 dequantize_per_tensor_default_43" [label=None, style=solid]; +"707 dequantize_per_tensor_default_43" -> "708 permute_9" [label=None, style=solid]; +"708 permute_9" -> "709 view_70" [label="(197, 1, 12, 64)", style=solid]; +"709 view_70" -> "710 linear_33" [label="(197, 768)", style=solid]; +"710 linear_33" -> "711 view_71" [label="(197, 768)", style=solid]; +"711 view_71" -> "712 transpose_53" [label="(197, 1, 768)", style=solid]; +"712 transpose_53" -> "713 dropout_25" [label="(1, 197, 768)", style=solid]; +"713 dropout_25" -> "714 add_17" [label="(1, 197, 768)", style=solid]; +"714 add_17" -> "715 layer_norm_17" [label="(1, 197, 768)", style=solid]; +"714 add_17" -> "725 add_18" [label="(1, 197, 768)", style=solid]; +"715 layer_norm_17" -> "716 quantize_per_tensor_default_44" [label="(1, 197, 768)", style=solid]; +"716 quantize_per_tensor_default_44" -> "717 dequantize_per_tensor_default_44" [label=None, style=solid]; +"717 dequantize_per_tensor_default_44" -> "718 linear_34" [label=None, style=solid]; +"718 linear_34" -> "719 gelu_8" [label="(1, 197, 3072)", style=solid]; +"719 gelu_8" -> "720 quantize_per_tensor_default_45" [label="(1, 197, 3072)", style=solid]; +"720 quantize_per_tensor_default_45" -> "721 dequantize_per_tensor_default_45" [label=None, style=solid]; +"721 dequantize_per_tensor_default_45" -> "722 dropout_26" [label=None, style=solid]; +"722 dropout_26" -> "723 linear_35" [label="(1, 197, 3072)", style=solid]; +"723 linear_35" -> "724 dropout_27" [label="(1, 197, 768)", style=solid]; +"724 dropout_27" -> "725 add_18" [label="(1, 197, 768)", style=solid]; +"725 add_18" -> "726 layer_norm_18" [label="(1, 197, 768)", style=solid]; +"725 add_18" -> "760 add_19" [label="(1, 197, 768)", style=solid]; +"726 layer_norm_18" -> "727 quantize_per_tensor_default_46" [label="(1, 197, 768)", style=solid]; +"727 quantize_per_tensor_default_46" -> "728 dequantize_per_tensor_default_46" [label=None, style=solid]; +"728 dequantize_per_tensor_default_46" -> "729 transpose_54" [label=None, style=solid]; +"729 transpose_54" -> "730 linear_36" [label="(197, 1, 768)", style=solid]; +"730 linear_36" -> "731 unflatten_9" [label="(197, 1, 2304)", style=solid]; +"731 unflatten_9" -> "732 unsqueeze_9" [label="(197, 1, 3, 768)", style=solid]; +"732 unsqueeze_9" -> "733 transpose_55" [label="(1, 197, 1, 3, 768)", style=solid]; +"733 transpose_55" -> "734 squeeze_9" [label="(3, 197, 1, 1, 768)", style=solid]; +"734 squeeze_9" -> "735 contiguous_9" [label="(3, 197, 1, 768)", style=solid]; +"735 contiguous_9" -> "736 quantize_per_tensor_default_47" [label="(3, 197, 1, 768)", style=solid]; +"735 contiguous_9" -> "741 select_29" [label="(3, 197, 1, 768)", style=solid]; +"736 quantize_per_tensor_default_47" -> "737 dequantize_per_tensor_default_81" [label=None, style=solid]; +"736 quantize_per_tensor_default_47" -> "738 dequantize_per_tensor_default_80" [label=None, style=solid]; +"737 dequantize_per_tensor_default_81" -> "740 select_28" [label=None, style=solid]; +"738 dequantize_per_tensor_default_80" -> "739 select_27" [label=None, style=solid]; +"739 select_27" -> "742 view_72" [label="(197, 1, 768)", style=solid]; +"740 select_28" -> "744 view_73" [label="(197, 1, 768)", style=solid]; +"741 select_29" -> "746 view_74" [label="(197, 1, 768)", style=solid]; +"742 view_72" -> "743 transpose_56" [label="(197, 12, 64)", style=solid]; +"743 transpose_56" -> "748 view_75" [label="(12, 197, 64)", style=solid]; +"744 view_73" -> "745 transpose_57" [label="(197, 12, 64)", style=solid]; +"745 transpose_57" -> "749 view_76" [label="(12, 197, 64)", style=solid]; +"746 view_74" -> "747 transpose_58" [label="(197, 12, 64)", style=solid]; +"747 transpose_58" -> "750 view_77" [label="(12, 197, 64)", style=solid]; +"748 view_75" -> "751 scaled_dot_product_attention_9" [label="(1, 12, 197, 64)", style=solid]; +"749 view_76" -> "751 scaled_dot_product_attention_9" [label="(1, 12, 197, 64)", style=solid]; +"750 view_77" -> "751 scaled_dot_product_attention_9" [label="(1, 12, 197, 64)", style=solid]; +"751 scaled_dot_product_attention_9" -> "752 quantize_per_tensor_default_48" [label="(1, 12, 197, 64)", style=solid]; +"752 quantize_per_tensor_default_48" -> "753 dequantize_per_tensor_default_48" [label=None, style=solid]; +"753 dequantize_per_tensor_default_48" -> "754 permute_10" [label=None, style=solid]; +"754 permute_10" -> "755 view_78" [label="(197, 1, 12, 64)", style=solid]; +"755 view_78" -> "756 linear_37" [label="(197, 768)", style=solid]; +"756 linear_37" -> "757 view_79" [label="(197, 768)", style=solid]; +"757 view_79" -> "758 transpose_59" [label="(197, 1, 768)", style=solid]; +"758 transpose_59" -> "759 dropout_28" [label="(1, 197, 768)", style=solid]; +"759 dropout_28" -> "760 add_19" [label="(1, 197, 768)", style=solid]; +"760 add_19" -> "761 layer_norm_19" [label="(1, 197, 768)", style=solid]; +"760 add_19" -> "771 add_20" [label="(1, 197, 768)", style=solid]; +"761 layer_norm_19" -> "762 quantize_per_tensor_default_49" [label="(1, 197, 768)", style=solid]; +"762 quantize_per_tensor_default_49" -> "763 dequantize_per_tensor_default_49" [label=None, style=solid]; +"763 dequantize_per_tensor_default_49" -> "764 linear_38" [label=None, style=solid]; +"764 linear_38" -> "765 gelu_9" [label="(1, 197, 3072)", style=solid]; +"765 gelu_9" -> "766 quantize_per_tensor_default_50" [label="(1, 197, 3072)", style=solid]; +"766 quantize_per_tensor_default_50" -> "767 dequantize_per_tensor_default_50" [label=None, style=solid]; +"767 dequantize_per_tensor_default_50" -> "768 dropout_29" [label=None, style=solid]; +"768 dropout_29" -> "769 linear_39" [label="(1, 197, 3072)", style=solid]; +"769 linear_39" -> "770 dropout_30" [label="(1, 197, 768)", style=solid]; +"770 dropout_30" -> "771 add_20" [label="(1, 197, 768)", style=solid]; +"771 add_20" -> "772 layer_norm_20" [label="(1, 197, 768)", style=solid]; +"771 add_20" -> "806 add_21" [label="(1, 197, 768)", style=solid]; +"772 layer_norm_20" -> "773 quantize_per_tensor_default_51" [label="(1, 197, 768)", style=solid]; +"773 quantize_per_tensor_default_51" -> "774 dequantize_per_tensor_default_51" [label=None, style=solid]; +"774 dequantize_per_tensor_default_51" -> "775 transpose_60" [label=None, style=solid]; +"775 transpose_60" -> "776 linear_40" [label="(197, 1, 768)", style=solid]; +"776 linear_40" -> "777 unflatten_10" [label="(197, 1, 2304)", style=solid]; +"777 unflatten_10" -> "778 unsqueeze_10" [label="(197, 1, 3, 768)", style=solid]; +"778 unsqueeze_10" -> "779 transpose_61" [label="(1, 197, 1, 3, 768)", style=solid]; +"779 transpose_61" -> "780 squeeze_10" [label="(3, 197, 1, 1, 768)", style=solid]; +"780 squeeze_10" -> "781 contiguous_10" [label="(3, 197, 1, 768)", style=solid]; +"781 contiguous_10" -> "782 quantize_per_tensor_default_52" [label="(3, 197, 1, 768)", style=solid]; +"781 contiguous_10" -> "787 select_32" [label="(3, 197, 1, 768)", style=solid]; +"782 quantize_per_tensor_default_52" -> "783 dequantize_per_tensor_default_83" [label=None, style=solid]; +"782 quantize_per_tensor_default_52" -> "784 dequantize_per_tensor_default_82" [label=None, style=solid]; +"783 dequantize_per_tensor_default_83" -> "786 select_31" [label=None, style=solid]; +"784 dequantize_per_tensor_default_82" -> "785 select_30" [label=None, style=solid]; +"785 select_30" -> "788 view_80" [label="(197, 1, 768)", style=solid]; +"786 select_31" -> "790 view_81" [label="(197, 1, 768)", style=solid]; +"787 select_32" -> "792 view_82" [label="(197, 1, 768)", style=solid]; +"788 view_80" -> "789 transpose_62" [label="(197, 12, 64)", style=solid]; +"789 transpose_62" -> "794 view_83" [label="(12, 197, 64)", style=solid]; +"790 view_81" -> "791 transpose_63" [label="(197, 12, 64)", style=solid]; +"791 transpose_63" -> "795 view_84" [label="(12, 197, 64)", style=solid]; +"792 view_82" -> "793 transpose_64" [label="(197, 12, 64)", style=solid]; +"793 transpose_64" -> "796 view_85" [label="(12, 197, 64)", style=solid]; +"794 view_83" -> "797 scaled_dot_product_attention_10" [label="(1, 12, 197, 64)", style=solid]; +"795 view_84" -> "797 scaled_dot_product_attention_10" [label="(1, 12, 197, 64)", style=solid]; +"796 view_85" -> "797 scaled_dot_product_attention_10" [label="(1, 12, 197, 64)", style=solid]; +"797 scaled_dot_product_attention_10" -> "798 quantize_per_tensor_default_53" [label="(1, 12, 197, 64)", style=solid]; +"798 quantize_per_tensor_default_53" -> "799 dequantize_per_tensor_default_53" [label=None, style=solid]; +"799 dequantize_per_tensor_default_53" -> "800 permute_11" [label=None, style=solid]; +"800 permute_11" -> "801 view_86" [label="(197, 1, 12, 64)", style=solid]; +"801 view_86" -> "802 linear_41" [label="(197, 768)", style=solid]; +"802 linear_41" -> "803 view_87" [label="(197, 768)", style=solid]; +"803 view_87" -> "804 transpose_65" [label="(197, 1, 768)", style=solid]; +"804 transpose_65" -> "805 dropout_31" [label="(1, 197, 768)", style=solid]; +"805 dropout_31" -> "806 add_21" [label="(1, 197, 768)", style=solid]; +"806 add_21" -> "807 layer_norm_21" [label="(1, 197, 768)", style=solid]; +"806 add_21" -> "817 add_22" [label="(1, 197, 768)", style=solid]; +"807 layer_norm_21" -> "808 quantize_per_tensor_default_54" [label="(1, 197, 768)", style=solid]; +"808 quantize_per_tensor_default_54" -> "809 dequantize_per_tensor_default_54" [label=None, style=solid]; +"809 dequantize_per_tensor_default_54" -> "810 linear_42" [label=None, style=solid]; +"810 linear_42" -> "811 gelu_10" [label="(1, 197, 3072)", style=solid]; +"811 gelu_10" -> "812 quantize_per_tensor_default_55" [label="(1, 197, 3072)", style=solid]; +"812 quantize_per_tensor_default_55" -> "813 dequantize_per_tensor_default_55" [label=None, style=solid]; +"813 dequantize_per_tensor_default_55" -> "814 dropout_32" [label=None, style=solid]; +"814 dropout_32" -> "815 linear_43" [label="(1, 197, 3072)", style=solid]; +"815 linear_43" -> "816 dropout_33" [label="(1, 197, 768)", style=solid]; +"816 dropout_33" -> "817 add_22" [label="(1, 197, 768)", style=solid]; +"817 add_22" -> "818 layer_norm_22" [label="(1, 197, 768)", style=solid]; +"817 add_22" -> "852 add_23" [label="(1, 197, 768)", style=solid]; +"818 layer_norm_22" -> "819 quantize_per_tensor_default_56" [label="(1, 197, 768)", style=solid]; +"819 quantize_per_tensor_default_56" -> "820 dequantize_per_tensor_default_56" [label=None, style=solid]; +"820 dequantize_per_tensor_default_56" -> "821 transpose_66" [label=None, style=solid]; +"821 transpose_66" -> "822 linear_44" [label="(197, 1, 768)", style=solid]; +"822 linear_44" -> "823 unflatten_11" [label="(197, 1, 2304)", style=solid]; +"823 unflatten_11" -> "824 unsqueeze_11" [label="(197, 1, 3, 768)", style=solid]; +"824 unsqueeze_11" -> "825 transpose_67" [label="(1, 197, 1, 3, 768)", style=solid]; +"825 transpose_67" -> "826 squeeze_11" [label="(3, 197, 1, 1, 768)", style=solid]; +"826 squeeze_11" -> "827 contiguous_11" [label="(3, 197, 1, 768)", style=solid]; +"827 contiguous_11" -> "828 quantize_per_tensor_default_57" [label="(3, 197, 1, 768)", style=solid]; +"827 contiguous_11" -> "833 select_35" [label="(3, 197, 1, 768)", style=solid]; +"828 quantize_per_tensor_default_57" -> "829 dequantize_per_tensor_default_85" [label=None, style=solid]; +"828 quantize_per_tensor_default_57" -> "830 dequantize_per_tensor_default_84" [label=None, style=solid]; +"829 dequantize_per_tensor_default_85" -> "832 select_34" [label=None, style=solid]; +"830 dequantize_per_tensor_default_84" -> "831 select_33" [label=None, style=solid]; +"831 select_33" -> "834 view_88" [label="(197, 1, 768)", style=solid]; +"832 select_34" -> "836 view_89" [label="(197, 1, 768)", style=solid]; +"833 select_35" -> "838 view_90" [label="(197, 1, 768)", style=solid]; +"834 view_88" -> "835 transpose_68" [label="(197, 12, 64)", style=solid]; +"835 transpose_68" -> "840 view_91" [label="(12, 197, 64)", style=solid]; +"836 view_89" -> "837 transpose_69" [label="(197, 12, 64)", style=solid]; +"837 transpose_69" -> "841 view_92" [label="(12, 197, 64)", style=solid]; +"838 view_90" -> "839 transpose_70" [label="(197, 12, 64)", style=solid]; +"839 transpose_70" -> "842 view_93" [label="(12, 197, 64)", style=solid]; +"840 view_91" -> "843 scaled_dot_product_attention_11" [label="(1, 12, 197, 64)", style=solid]; +"841 view_92" -> "843 scaled_dot_product_attention_11" [label="(1, 12, 197, 64)", style=solid]; +"842 view_93" -> "843 scaled_dot_product_attention_11" [label="(1, 12, 197, 64)", style=solid]; +"843 scaled_dot_product_attention_11" -> "844 quantize_per_tensor_default_58" [label="(1, 12, 197, 64)", style=solid]; +"844 quantize_per_tensor_default_58" -> "845 dequantize_per_tensor_default_58" [label=None, style=solid]; +"845 dequantize_per_tensor_default_58" -> "846 permute_12" [label=None, style=solid]; +"846 permute_12" -> "847 view_94" [label="(197, 1, 12, 64)", style=solid]; +"847 view_94" -> "848 linear_45" [label="(197, 768)", style=solid]; +"848 linear_45" -> "849 view_95" [label="(197, 768)", style=solid]; +"849 view_95" -> "850 transpose_71" [label="(197, 1, 768)", style=solid]; +"850 transpose_71" -> "851 dropout_34" [label="(1, 197, 768)", style=solid]; +"851 dropout_34" -> "852 add_23" [label="(1, 197, 768)", style=solid]; +"852 add_23" -> "853 layer_norm_23" [label="(1, 197, 768)", style=solid]; +"852 add_23" -> "863 add_24" [label="(1, 197, 768)", style=solid]; +"853 layer_norm_23" -> "854 quantize_per_tensor_default_59" [label="(1, 197, 768)", style=solid]; +"854 quantize_per_tensor_default_59" -> "855 dequantize_per_tensor_default_59" [label=None, style=solid]; +"855 dequantize_per_tensor_default_59" -> "856 linear_46" [label=None, style=solid]; +"856 linear_46" -> "857 gelu_11" [label="(1, 197, 3072)", style=solid]; +"857 gelu_11" -> "858 quantize_per_tensor_default_60" [label="(1, 197, 3072)", style=solid]; +"858 quantize_per_tensor_default_60" -> "859 dequantize_per_tensor_default_60" [label=None, style=solid]; +"859 dequantize_per_tensor_default_60" -> "860 dropout_35" [label=None, style=solid]; +"860 dropout_35" -> "861 linear_47" [label="(1, 197, 3072)", style=solid]; +"861 linear_47" -> "862 dropout_36" [label="(1, 197, 768)", style=solid]; +"862 dropout_36" -> "863 add_24" [label="(1, 197, 768)", style=solid]; +"863 add_24" -> "864 layer_norm_24" [label="(1, 197, 768)", style=solid]; +"864 layer_norm_24" -> "865 quantize_per_tensor_default_61" [label="(1, 197, 768)", style=solid]; +"865 quantize_per_tensor_default_61" -> "866 dequantize_per_tensor_default_61" [label=None, style=solid]; +"866 dequantize_per_tensor_default_61" -> "867 slice_1" [label=None, style=solid]; +"867 slice_1" -> "868 select_36" [label="(1, 197, 768)", style=solid]; +"868 select_36" -> "869 linear_48" [label="(1, 768)", style=solid]; +"869 linear_48" -> "870 output_1" [label="(1, 1000)", style=solid]; +} diff --git a/tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/yolo11n_sdpa_block.dot b/tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/yolo11n_sdpa_block.dot new file mode 100644 index 00000000000..95cea03006d --- /dev/null +++ b/tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/yolo11n_sdpa_block.dot @@ -0,0 +1,57 @@ +strict digraph { +"0 _scale_0" [id=0, type=get_attr]; +"1 _zero_point_0" [id=1, type=get_attr]; +"2 _frozen_param0" [id=2, type=get_attr]; +"3 dequantize_per_channel_default" [id=3, type=dequantize_per_channel]; +"4 x" [id=4, type=input]; +"5 quantize_per_tensor_default" [id=5, type=quantize_per_tensor]; +"6 dequantize_per_tensor_default" [id=6, type=dequantize_per_tensor]; +"7 linear" [id=7, type=linear]; +"8 quantize_per_tensor_default_1" [id=8, type=quantize_per_tensor]; +"9 dequantize_per_tensor_default_3" [id=9, type=dequantize_per_tensor]; +"10 dequantize_per_tensor_default_2" [id=10, type=dequantize_per_tensor]; +"11 slice_1" [id=11, type=slice]; +"12 slice_2" [id=12, type=slice]; +"13 slice_3" [id=13, type=slice]; +"14 slice_4" [id=14, type=slice]; +"15 slice_5" [id=15, type=slice]; +"16 slice_6" [id=16, type=slice]; +"17 slice_7" [id=17, type=slice]; +"18 slice_8" [id=18, type=slice]; +"19 slice_9" [id=19, type=slice]; +"20 transpose" [id=20, type=transpose]; +"21 matmul" [id=21, type=matmul]; +"22 div_" [id=22, type=div_]; +"23 softmax" [id=23, type=softmax]; +"24 transpose_1" [id=24, type=transpose]; +"25 matmul_1" [id=25, type=matmul]; +"26 output_1" [id=26, type=output]; +"0 _scale_0" -> "3 dequantize_per_channel_default" [label="(12,)", style=solid]; +"1 _zero_point_0" -> "3 dequantize_per_channel_default" [label="(12,)", style=solid]; +"2 _frozen_param0" -> "3 dequantize_per_channel_default" [label="(12, 4)", style=solid]; +"3 dequantize_per_channel_default" -> "7 linear" [label=None, style=solid]; +"4 x" -> "5 quantize_per_tensor_default" [label="(1, 2, 4)", style=solid]; +"5 quantize_per_tensor_default" -> "6 dequantize_per_tensor_default" [label=None, style=solid]; +"6 dequantize_per_tensor_default" -> "7 linear" [label=None, style=solid]; +"7 linear" -> "8 quantize_per_tensor_default_1" [label="(1, 2, 12)", style=solid]; +"7 linear" -> "17 slice_7" [label="(1, 2, 12)", style=solid]; +"8 quantize_per_tensor_default_1" -> "9 dequantize_per_tensor_default_3" [label=None, style=solid]; +"8 quantize_per_tensor_default_1" -> "10 dequantize_per_tensor_default_2" [label=None, style=solid]; +"9 dequantize_per_tensor_default_3" -> "14 slice_4" [label=None, style=solid]; +"10 dequantize_per_tensor_default_2" -> "11 slice_1" [label=None, style=solid]; +"11 slice_1" -> "12 slice_2" [label="(1, 2, 12)", style=solid]; +"12 slice_2" -> "13 slice_3" [label="(1, 2, 12)", style=solid]; +"13 slice_3" -> "21 matmul" [label="(1, 2, 4)", style=solid]; +"14 slice_4" -> "15 slice_5" [label="(1, 2, 12)", style=solid]; +"15 slice_5" -> "16 slice_6" [label="(1, 2, 12)", style=solid]; +"16 slice_6" -> "20 transpose" [label="(1, 2, 4)", style=solid]; +"17 slice_7" -> "18 slice_8" [label="(1, 2, 12)", style=solid]; +"18 slice_8" -> "19 slice_9" [label="(1, 2, 12)", style=solid]; +"19 slice_9" -> "25 matmul_1" [label="(1, 2, 4)", style=solid]; +"20 transpose" -> "21 matmul" [label="(1, 4, 2)", style=solid]; +"21 matmul" -> "22 div_" [label="(1, 2, 2)", style=solid]; +"22 div_" -> "23 softmax" [label="(1, 2, 2)", style=solid]; +"23 softmax" -> "24 transpose_1" [label="(1, 2, 2)", style=solid]; +"24 transpose_1" -> "25 matmul_1" [label="(1, 2, 2)", style=solid]; +"25 matmul_1" -> "26 output_1" [label="(1, 2, 4)", style=solid]; +} diff --git a/tests/torch/fx/test_quantizer.py b/tests/torch/fx/test_quantizer.py index 9c6fbabd2a9..12b8f5bec04 100644 --- a/tests/torch/fx/test_quantizer.py +++ b/tests/torch/fx/test_quantizer.py @@ -22,6 +22,8 @@ import torch.utils.data import torch.utils.data.distributed import torchvision.models as models +from torch.ao.quantization.quantize_pt2e import convert_pt2e +from torch.ao.quantization.quantize_pt2e import prepare_pt2e from torch.ao.quantization.quantizer.quantizer import Quantizer from torch.ao.quantization.quantizer.x86_inductor_quantizer import X86InductorQuantizer from torch.ao.quantization.quantizer.x86_inductor_quantizer import get_default_x86_inductor_quantization_config @@ -165,3 +167,25 @@ def test_quantized_model( # ao_quantized_model = convert_pt2e(prepared_model) # ao_nncf_graph = GraphConverter.create_nncf_graph(ao_quantized_model) # ao_nncf_graph.visualize_graph("ao_" + get_dot_filename(model_case.model_id)) + + +@pytest.mark.parametrize( + "model_case,quantizer_params", + [(m[0], m[1]) for m in TEST_MODELS_QUANIZED], + ids=[m[0].model_id for m in TEST_MODELS_QUANIZED], +) +def test_openvino_quantizer_with_torch_ao_convert_pt2e(model_case: ModelCase, quantizer_params): + quantizer = get_openvino_quantizer(**quantizer_params) + fx_model, example_input = _build_torch_fx_model(model_case) + + prepared_model = prepare_pt2e(fx_model, quantizer) + prepared_model(example_input) + ao_quantized_model = convert_pt2e(prepared_model) + + nncf_graph = GraphConverter.create_nncf_graph(ao_quantized_model) + check_graph( + nncf_graph, + get_dot_filename(model_case.model_id), + FX_QUANTIZED_DIR_NAME / "ao_export_quantization_OpenVINOQuantizer", + extended=True, + ) From 3440d823c55e095cb336bca9ce63a768bd8efa21 Mon Sep 17 00:00:00 2001 From: dlyakhov Date: Mon, 2 Dec 2024 17:00:57 +0100 Subject: [PATCH 10/15] Constant folding is enabled for OpenVINOQuantizer --- .../quantizer/openvino_quantizer.py | 5 + .../OpenVINOQuantizer/swin_v2_s.dot | 11368 +++++++-------- .../OpenVINOQuantizer/vit_b_16.dot | 3820 +++-- .../swin_v2_s.dot | 11938 +++++++--------- .../vit_b_16.dot | 3576 +++-- 5 files changed, 13611 insertions(+), 17096 deletions(-) diff --git a/nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer.py b/nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer.py index 45a1c69be24..7683d1efb2d 100644 --- a/nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer.py +++ b/nncf/experimental/common/quantization/algorithms/quantizer/openvino_quantizer.py @@ -29,6 +29,7 @@ from nncf.experimental.common.quantization.algorithms.quantizer.base_quantizer import NNCFQuantizer from nncf.experimental.torch.fx.nncf_graph_builder import GraphConverter from nncf.experimental.torch.fx.node_utils import get_graph_node_by_name +from nncf.experimental.torch.fx.transformations import fold_constant_except_qdq from nncf.parameters import ModelType from nncf.parameters import QuantizationMode from nncf.parameters import TargetDevice @@ -164,3 +165,7 @@ def _convert_nncf_qspec_to_inductor_qspec(self, qspec: NNCFQuantizerConfig, is_w def validate(self, model: torch.fx.GraphModule) -> None: pass + + def transform_for_annotation(self, model: torch.fx.GraphModule) -> torch.fx.GraphModule: + fold_constant_except_qdq(model) + return model diff --git a/tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/swin_v2_s.dot b/tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/swin_v2_s.dot index 0e655b8fce3..718f98ddc6c 100644 --- a/tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/swin_v2_s.dot +++ b/tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/swin_v2_s.dot @@ -1,6310 +1,5062 @@ strict digraph { -"0 features_1_0_attn_qkv_bias" [id=0, type=get_attr]; -"1 features_1_0_attn_proj_bias" [id=1, type=get_attr]; -"2 features_1_0_attn_logit_scale" [id=2, type=get_attr]; -"3 features_1_1_attn_qkv_bias" [id=3, type=get_attr]; -"4 features_1_1_attn_proj_bias" [id=4, type=get_attr]; -"5 features_1_1_attn_logit_scale" [id=5, type=get_attr]; -"6 features_3_0_attn_qkv_bias" [id=6, type=get_attr]; -"7 features_3_0_attn_proj_bias" [id=7, type=get_attr]; -"8 features_3_0_attn_logit_scale" [id=8, type=get_attr]; -"9 features_3_1_attn_qkv_bias" [id=9, type=get_attr]; -"10 features_3_1_attn_proj_bias" [id=10, type=get_attr]; -"11 features_3_1_attn_logit_scale" [id=11, type=get_attr]; -"12 features_5_0_attn_qkv_bias" [id=12, type=get_attr]; -"13 features_5_0_attn_proj_bias" [id=13, type=get_attr]; -"14 features_5_0_attn_logit_scale" [id=14, type=get_attr]; -"15 features_5_1_attn_qkv_bias" [id=15, type=get_attr]; -"16 features_5_1_attn_proj_bias" [id=16, type=get_attr]; -"17 features_5_1_attn_logit_scale" [id=17, type=get_attr]; -"18 features_5_2_attn_qkv_bias" [id=18, type=get_attr]; -"19 features_5_2_attn_proj_bias" [id=19, type=get_attr]; -"20 features_5_2_attn_logit_scale" [id=20, type=get_attr]; -"21 features_5_3_attn_qkv_bias" [id=21, type=get_attr]; -"22 features_5_3_attn_proj_bias" [id=22, type=get_attr]; -"23 features_5_3_attn_logit_scale" [id=23, type=get_attr]; -"24 features_5_4_attn_qkv_bias" [id=24, type=get_attr]; -"25 features_5_4_attn_proj_bias" [id=25, type=get_attr]; -"26 features_5_4_attn_logit_scale" [id=26, type=get_attr]; -"27 features_5_5_attn_qkv_bias" [id=27, type=get_attr]; -"28 features_5_5_attn_proj_bias" [id=28, type=get_attr]; -"29 features_5_5_attn_logit_scale" [id=29, type=get_attr]; -"30 features_5_6_attn_qkv_bias" [id=30, type=get_attr]; -"31 features_5_6_attn_proj_bias" [id=31, type=get_attr]; -"32 features_5_6_attn_logit_scale" [id=32, type=get_attr]; -"33 features_5_7_attn_qkv_bias" [id=33, type=get_attr]; -"34 features_5_7_attn_proj_bias" [id=34, type=get_attr]; -"35 features_5_7_attn_logit_scale" [id=35, type=get_attr]; -"36 features_5_8_attn_qkv_bias" [id=36, type=get_attr]; -"37 features_5_8_attn_proj_bias" [id=37, type=get_attr]; -"38 features_5_8_attn_logit_scale" [id=38, type=get_attr]; -"39 features_5_9_attn_qkv_bias" [id=39, type=get_attr]; -"40 features_5_9_attn_proj_bias" [id=40, type=get_attr]; -"41 features_5_9_attn_logit_scale" [id=41, type=get_attr]; -"42 features_5_10_attn_qkv_bias" [id=42, type=get_attr]; -"43 features_5_10_attn_proj_bias" [id=43, type=get_attr]; -"44 features_5_10_attn_logit_scale" [id=44, type=get_attr]; -"45 features_5_11_attn_qkv_bias" [id=45, type=get_attr]; -"46 features_5_11_attn_proj_bias" [id=46, type=get_attr]; -"47 features_5_11_attn_logit_scale" [id=47, type=get_attr]; -"48 features_5_12_attn_qkv_bias" [id=48, type=get_attr]; -"49 features_5_12_attn_proj_bias" [id=49, type=get_attr]; -"50 features_5_12_attn_logit_scale" [id=50, type=get_attr]; -"51 features_5_13_attn_qkv_bias" [id=51, type=get_attr]; -"52 features_5_13_attn_proj_bias" [id=52, type=get_attr]; -"53 features_5_13_attn_logit_scale" [id=53, type=get_attr]; -"54 features_5_14_attn_qkv_bias" [id=54, type=get_attr]; -"55 features_5_14_attn_proj_bias" [id=55, type=get_attr]; -"56 features_5_14_attn_logit_scale" [id=56, type=get_attr]; -"57 features_5_15_attn_qkv_bias" [id=57, type=get_attr]; -"58 features_5_15_attn_proj_bias" [id=58, type=get_attr]; -"59 features_5_15_attn_logit_scale" [id=59, type=get_attr]; -"60 features_5_16_attn_qkv_bias" [id=60, type=get_attr]; -"61 features_5_16_attn_proj_bias" [id=61, type=get_attr]; -"62 features_5_16_attn_logit_scale" [id=62, type=get_attr]; -"63 features_5_17_attn_qkv_bias" [id=63, type=get_attr]; -"64 features_5_17_attn_proj_bias" [id=64, type=get_attr]; -"65 features_5_17_attn_logit_scale" [id=65, type=get_attr]; -"66 features_7_0_attn_qkv_bias" [id=66, type=get_attr]; -"67 features_7_0_attn_proj_bias" [id=67, type=get_attr]; -"68 features_7_0_attn_logit_scale" [id=68, type=get_attr]; -"69 features_7_1_attn_qkv_bias" [id=69, type=get_attr]; -"70 features_7_1_attn_proj_bias" [id=70, type=get_attr]; -"71 features_7_1_attn_logit_scale" [id=71, type=get_attr]; -"72 features_0_0_bias" [id=72, type=get_attr]; -"73 features_0_2_weight" [id=73, type=get_attr]; -"74 features_0_2_bias" [id=74, type=get_attr]; -"75 features_1_0_attn_cpb_mlp_0_bias" [id=75, type=get_attr]; -"76 features_1_0_norm1_weight" [id=76, type=get_attr]; -"77 features_1_0_norm1_bias" [id=77, type=get_attr]; -"78 features_1_0_mlp_0_bias" [id=78, type=get_attr]; -"79 features_1_0_mlp_3_bias" [id=79, type=get_attr]; -"80 features_1_0_norm2_weight" [id=80, type=get_attr]; -"81 features_1_0_norm2_bias" [id=81, type=get_attr]; -"82 features_1_1_attn_cpb_mlp_0_bias" [id=82, type=get_attr]; -"83 features_1_1_norm1_weight" [id=83, type=get_attr]; -"84 features_1_1_norm1_bias" [id=84, type=get_attr]; -"85 features_1_1_mlp_0_bias" [id=85, type=get_attr]; -"86 features_1_1_mlp_3_bias" [id=86, type=get_attr]; -"87 features_1_1_norm2_weight" [id=87, type=get_attr]; -"88 features_1_1_norm2_bias" [id=88, type=get_attr]; -"89 features_2_norm_weight" [id=89, type=get_attr]; -"90 features_2_norm_bias" [id=90, type=get_attr]; -"91 features_3_0_attn_cpb_mlp_0_bias" [id=91, type=get_attr]; -"92 features_3_0_norm1_weight" [id=92, type=get_attr]; -"93 features_3_0_norm1_bias" [id=93, type=get_attr]; -"94 features_3_0_mlp_0_bias" [id=94, type=get_attr]; -"95 features_3_0_mlp_3_bias" [id=95, type=get_attr]; -"96 features_3_0_norm2_weight" [id=96, type=get_attr]; -"97 features_3_0_norm2_bias" [id=97, type=get_attr]; -"98 features_3_1_attn_cpb_mlp_0_bias" [id=98, type=get_attr]; -"99 features_3_1_norm1_weight" [id=99, type=get_attr]; -"100 features_3_1_norm1_bias" [id=100, type=get_attr]; -"101 features_3_1_mlp_0_bias" [id=101, type=get_attr]; -"102 features_3_1_mlp_3_bias" [id=102, type=get_attr]; -"103 features_3_1_norm2_weight" [id=103, type=get_attr]; -"104 features_3_1_norm2_bias" [id=104, type=get_attr]; -"105 features_4_norm_weight" [id=105, type=get_attr]; -"106 features_4_norm_bias" [id=106, type=get_attr]; -"107 features_5_0_attn_cpb_mlp_0_bias" [id=107, type=get_attr]; -"108 features_5_0_norm1_weight" [id=108, type=get_attr]; -"109 features_5_0_norm1_bias" [id=109, type=get_attr]; -"110 features_5_0_mlp_0_bias" [id=110, type=get_attr]; -"111 features_5_0_mlp_3_bias" [id=111, type=get_attr]; -"112 features_5_0_norm2_weight" [id=112, type=get_attr]; -"113 features_5_0_norm2_bias" [id=113, type=get_attr]; -"114 features_5_1_attn_cpb_mlp_0_bias" [id=114, type=get_attr]; -"115 features_5_1_norm1_weight" [id=115, type=get_attr]; -"116 features_5_1_norm1_bias" [id=116, type=get_attr]; -"117 features_5_1_mlp_0_bias" [id=117, type=get_attr]; -"118 features_5_1_mlp_3_bias" [id=118, type=get_attr]; -"119 features_5_1_norm2_weight" [id=119, type=get_attr]; -"120 features_5_1_norm2_bias" [id=120, type=get_attr]; -"121 features_5_2_attn_cpb_mlp_0_bias" [id=121, type=get_attr]; -"122 features_5_2_norm1_weight" [id=122, type=get_attr]; -"123 features_5_2_norm1_bias" [id=123, type=get_attr]; -"124 features_5_2_mlp_0_bias" [id=124, type=get_attr]; -"125 features_5_2_mlp_3_bias" [id=125, type=get_attr]; -"126 features_5_2_norm2_weight" [id=126, type=get_attr]; -"127 features_5_2_norm2_bias" [id=127, type=get_attr]; -"128 features_5_3_attn_cpb_mlp_0_bias" [id=128, type=get_attr]; -"129 features_5_3_norm1_weight" [id=129, type=get_attr]; -"130 features_5_3_norm1_bias" [id=130, type=get_attr]; -"131 features_5_3_mlp_0_bias" [id=131, type=get_attr]; -"132 features_5_3_mlp_3_bias" [id=132, type=get_attr]; -"133 features_5_3_norm2_weight" [id=133, type=get_attr]; -"134 features_5_3_norm2_bias" [id=134, type=get_attr]; -"135 features_5_4_attn_cpb_mlp_0_bias" [id=135, type=get_attr]; -"136 features_5_4_norm1_weight" [id=136, type=get_attr]; -"137 features_5_4_norm1_bias" [id=137, type=get_attr]; -"138 features_5_4_mlp_0_bias" [id=138, type=get_attr]; -"139 features_5_4_mlp_3_bias" [id=139, type=get_attr]; -"140 features_5_4_norm2_weight" [id=140, type=get_attr]; -"141 features_5_4_norm2_bias" [id=141, type=get_attr]; -"142 features_5_5_attn_cpb_mlp_0_bias" [id=142, type=get_attr]; -"143 features_5_5_norm1_weight" [id=143, type=get_attr]; -"144 features_5_5_norm1_bias" [id=144, type=get_attr]; -"145 features_5_5_mlp_0_bias" [id=145, type=get_attr]; -"146 features_5_5_mlp_3_bias" [id=146, type=get_attr]; -"147 features_5_5_norm2_weight" [id=147, type=get_attr]; -"148 features_5_5_norm2_bias" [id=148, type=get_attr]; -"149 features_5_6_attn_cpb_mlp_0_bias" [id=149, type=get_attr]; -"150 features_5_6_norm1_weight" [id=150, type=get_attr]; -"151 features_5_6_norm1_bias" [id=151, type=get_attr]; -"152 features_5_6_mlp_0_bias" [id=152, type=get_attr]; -"153 features_5_6_mlp_3_bias" [id=153, type=get_attr]; -"154 features_5_6_norm2_weight" [id=154, type=get_attr]; -"155 features_5_6_norm2_bias" [id=155, type=get_attr]; -"156 features_5_7_attn_cpb_mlp_0_bias" [id=156, type=get_attr]; -"157 features_5_7_norm1_weight" [id=157, type=get_attr]; -"158 features_5_7_norm1_bias" [id=158, type=get_attr]; -"159 features_5_7_mlp_0_bias" [id=159, type=get_attr]; -"160 features_5_7_mlp_3_bias" [id=160, type=get_attr]; -"161 features_5_7_norm2_weight" [id=161, type=get_attr]; -"162 features_5_7_norm2_bias" [id=162, type=get_attr]; -"163 features_5_8_attn_cpb_mlp_0_bias" [id=163, type=get_attr]; -"164 features_5_8_norm1_weight" [id=164, type=get_attr]; -"165 features_5_8_norm1_bias" [id=165, type=get_attr]; -"166 features_5_8_mlp_0_bias" [id=166, type=get_attr]; -"167 features_5_8_mlp_3_bias" [id=167, type=get_attr]; -"168 features_5_8_norm2_weight" [id=168, type=get_attr]; -"169 features_5_8_norm2_bias" [id=169, type=get_attr]; -"170 features_5_9_attn_cpb_mlp_0_bias" [id=170, type=get_attr]; -"171 features_5_9_norm1_weight" [id=171, type=get_attr]; -"172 features_5_9_norm1_bias" [id=172, type=get_attr]; -"173 features_5_9_mlp_0_bias" [id=173, type=get_attr]; -"174 features_5_9_mlp_3_bias" [id=174, type=get_attr]; -"175 features_5_9_norm2_weight" [id=175, type=get_attr]; -"176 features_5_9_norm2_bias" [id=176, type=get_attr]; -"177 features_5_10_attn_cpb_mlp_0_bias" [id=177, type=get_attr]; -"178 features_5_10_norm1_weight" [id=178, type=get_attr]; -"179 features_5_10_norm1_bias" [id=179, type=get_attr]; -"180 features_5_10_mlp_0_bias" [id=180, type=get_attr]; -"181 features_5_10_mlp_3_bias" [id=181, type=get_attr]; -"182 features_5_10_norm2_weight" [id=182, type=get_attr]; -"183 features_5_10_norm2_bias" [id=183, type=get_attr]; -"184 features_5_11_attn_cpb_mlp_0_bias" [id=184, type=get_attr]; -"185 features_5_11_norm1_weight" [id=185, type=get_attr]; -"186 features_5_11_norm1_bias" [id=186, type=get_attr]; -"187 features_5_11_mlp_0_bias" [id=187, type=get_attr]; -"188 features_5_11_mlp_3_bias" [id=188, type=get_attr]; -"189 features_5_11_norm2_weight" [id=189, type=get_attr]; -"190 features_5_11_norm2_bias" [id=190, type=get_attr]; -"191 features_5_12_attn_cpb_mlp_0_bias" [id=191, type=get_attr]; -"192 features_5_12_norm1_weight" [id=192, type=get_attr]; -"193 features_5_12_norm1_bias" [id=193, type=get_attr]; -"194 features_5_12_mlp_0_bias" [id=194, type=get_attr]; -"195 features_5_12_mlp_3_bias" [id=195, type=get_attr]; -"196 features_5_12_norm2_weight" [id=196, type=get_attr]; -"197 features_5_12_norm2_bias" [id=197, type=get_attr]; -"198 features_5_13_attn_cpb_mlp_0_bias" [id=198, type=get_attr]; -"199 features_5_13_norm1_weight" [id=199, type=get_attr]; -"200 features_5_13_norm1_bias" [id=200, type=get_attr]; -"201 features_5_13_mlp_0_bias" [id=201, type=get_attr]; -"202 features_5_13_mlp_3_bias" [id=202, type=get_attr]; -"203 features_5_13_norm2_weight" [id=203, type=get_attr]; -"204 features_5_13_norm2_bias" [id=204, type=get_attr]; -"205 features_5_14_attn_cpb_mlp_0_bias" [id=205, type=get_attr]; -"206 features_5_14_norm1_weight" [id=206, type=get_attr]; -"207 features_5_14_norm1_bias" [id=207, type=get_attr]; -"208 features_5_14_mlp_0_bias" [id=208, type=get_attr]; -"209 features_5_14_mlp_3_bias" [id=209, type=get_attr]; -"210 features_5_14_norm2_weight" [id=210, type=get_attr]; -"211 features_5_14_norm2_bias" [id=211, type=get_attr]; -"212 features_5_15_attn_cpb_mlp_0_bias" [id=212, type=get_attr]; -"213 features_5_15_norm1_weight" [id=213, type=get_attr]; -"214 features_5_15_norm1_bias" [id=214, type=get_attr]; -"215 features_5_15_mlp_0_bias" [id=215, type=get_attr]; -"216 features_5_15_mlp_3_bias" [id=216, type=get_attr]; -"217 features_5_15_norm2_weight" [id=217, type=get_attr]; -"218 features_5_15_norm2_bias" [id=218, type=get_attr]; -"219 features_5_16_attn_cpb_mlp_0_bias" [id=219, type=get_attr]; -"220 features_5_16_norm1_weight" [id=220, type=get_attr]; -"221 features_5_16_norm1_bias" [id=221, type=get_attr]; -"222 features_5_16_mlp_0_bias" [id=222, type=get_attr]; -"223 features_5_16_mlp_3_bias" [id=223, type=get_attr]; -"224 features_5_16_norm2_weight" [id=224, type=get_attr]; -"225 features_5_16_norm2_bias" [id=225, type=get_attr]; -"226 features_5_17_attn_cpb_mlp_0_bias" [id=226, type=get_attr]; -"227 features_5_17_norm1_weight" [id=227, type=get_attr]; -"228 features_5_17_norm1_bias" [id=228, type=get_attr]; -"229 features_5_17_mlp_0_bias" [id=229, type=get_attr]; -"230 features_5_17_mlp_3_bias" [id=230, type=get_attr]; -"231 features_5_17_norm2_weight" [id=231, type=get_attr]; -"232 features_5_17_norm2_bias" [id=232, type=get_attr]; -"233 features_6_norm_weight" [id=233, type=get_attr]; -"234 features_6_norm_bias" [id=234, type=get_attr]; -"235 features_7_0_attn_cpb_mlp_0_bias" [id=235, type=get_attr]; -"236 features_7_0_norm1_weight" [id=236, type=get_attr]; -"237 features_7_0_norm1_bias" [id=237, type=get_attr]; -"238 features_7_0_mlp_0_bias" [id=238, type=get_attr]; -"239 features_7_0_mlp_3_bias" [id=239, type=get_attr]; -"240 features_7_0_norm2_weight" [id=240, type=get_attr]; -"241 features_7_0_norm2_bias" [id=241, type=get_attr]; -"242 features_7_1_attn_cpb_mlp_0_bias" [id=242, type=get_attr]; -"243 features_7_1_norm1_weight" [id=243, type=get_attr]; -"244 features_7_1_norm1_bias" [id=244, type=get_attr]; -"245 features_7_1_mlp_0_bias" [id=245, type=get_attr]; -"246 features_7_1_mlp_3_bias" [id=246, type=get_attr]; -"247 features_7_1_norm2_weight" [id=247, type=get_attr]; -"248 features_7_1_norm2_bias" [id=248, type=get_attr]; -"249 norm_weight" [id=249, type=get_attr]; -"250 norm_bias" [id=250, type=get_attr]; -"251 head_bias" [id=251, type=get_attr]; -"252 features_1_0_attn_relative_coords_table" [id=252, type=get_attr]; -"253 features_1_0_attn_relative_position_index" [id=253, type=get_attr]; -"254 features_1_1_attn_relative_coords_table" [id=254, type=get_attr]; -"255 features_1_1_attn_relative_position_index" [id=255, type=get_attr]; -"256 features_3_0_attn_relative_coords_table" [id=256, type=get_attr]; -"257 features_3_0_attn_relative_position_index" [id=257, type=get_attr]; -"258 features_3_1_attn_relative_coords_table" [id=258, type=get_attr]; -"259 features_3_1_attn_relative_position_index" [id=259, type=get_attr]; -"260 features_5_0_attn_relative_coords_table" [id=260, type=get_attr]; -"261 features_5_0_attn_relative_position_index" [id=261, type=get_attr]; -"262 features_5_1_attn_relative_coords_table" [id=262, type=get_attr]; -"263 features_5_1_attn_relative_position_index" [id=263, type=get_attr]; -"264 features_5_2_attn_relative_coords_table" [id=264, type=get_attr]; -"265 features_5_2_attn_relative_position_index" [id=265, type=get_attr]; -"266 features_5_3_attn_relative_coords_table" [id=266, type=get_attr]; -"267 features_5_3_attn_relative_position_index" [id=267, type=get_attr]; -"268 features_5_4_attn_relative_coords_table" [id=268, type=get_attr]; -"269 features_5_4_attn_relative_position_index" [id=269, type=get_attr]; -"270 features_5_5_attn_relative_coords_table" [id=270, type=get_attr]; -"271 features_5_5_attn_relative_position_index" [id=271, type=get_attr]; -"272 features_5_6_attn_relative_coords_table" [id=272, type=get_attr]; -"273 features_5_6_attn_relative_position_index" [id=273, type=get_attr]; -"274 features_5_7_attn_relative_coords_table" [id=274, type=get_attr]; -"275 features_5_7_attn_relative_position_index" [id=275, type=get_attr]; -"276 features_5_8_attn_relative_coords_table" [id=276, type=get_attr]; -"277 features_5_8_attn_relative_position_index" [id=277, type=get_attr]; -"278 features_5_9_attn_relative_coords_table" [id=278, type=get_attr]; -"279 features_5_9_attn_relative_position_index" [id=279, type=get_attr]; -"280 features_5_10_attn_relative_coords_table" [id=280, type=get_attr]; -"281 features_5_10_attn_relative_position_index" [id=281, type=get_attr]; -"282 features_5_11_attn_relative_coords_table" [id=282, type=get_attr]; -"283 features_5_11_attn_relative_position_index" [id=283, type=get_attr]; -"284 features_5_12_attn_relative_coords_table" [id=284, type=get_attr]; -"285 features_5_12_attn_relative_position_index" [id=285, type=get_attr]; -"286 features_5_13_attn_relative_coords_table" [id=286, type=get_attr]; -"287 features_5_13_attn_relative_position_index" [id=287, type=get_attr]; -"288 features_5_14_attn_relative_coords_table" [id=288, type=get_attr]; -"289 features_5_14_attn_relative_position_index" [id=289, type=get_attr]; -"290 features_5_15_attn_relative_coords_table" [id=290, type=get_attr]; -"291 features_5_15_attn_relative_position_index" [id=291, type=get_attr]; -"292 features_5_16_attn_relative_coords_table" [id=292, type=get_attr]; -"293 features_5_16_attn_relative_position_index" [id=293, type=get_attr]; -"294 features_5_17_attn_relative_coords_table" [id=294, type=get_attr]; -"295 features_5_17_attn_relative_position_index" [id=295, type=get_attr]; -"296 features_7_0_attn_relative_coords_table" [id=296, type=get_attr]; -"297 features_7_0_attn_relative_position_index" [id=297, type=get_attr]; -"298 features_7_1_attn_relative_coords_table" [id=298, type=get_attr]; -"299 features_7_1_attn_relative_position_index" [id=299, type=get_attr]; -"300 x" [id=300, type=input]; -"301 quantize_per_tensor_default" [id=301, type=quantize_per_tensor]; -"302 dequantize_per_tensor_default" [id=302, type=dequantize_per_tensor]; -"303 conv2d_scale_0" [id=303, type=get_attr]; -"304 conv2d_zero_point_0" [id=304, type=get_attr]; -"305 _frozen_param0" [id=305, type=get_attr]; -"306 dequantize_per_channel_default" [id=306, type=dequantize_per_channel]; -"307 conv2d" [id=307, type=conv2d]; -"308 permute" [id=308, type=permute]; -"309 layer_norm" [id=309, type=layer_norm]; -"310 features_1_0_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=310, type=call_module]; -"311 linear_scale_0" [id=311, type=get_attr]; -"312 linear_zero_point_0" [id=312, type=get_attr]; -"313 _frozen_param1" [id=313, type=get_attr]; -"314 dequantize_per_channel_default_1" [id=314, type=dequantize_per_channel]; -"315 linear" [id=315, type=linear]; -"316 relu_" [id=316, type=relu_]; -"317 relu__0_0_nncf_smooth_quant_0" [id=317, type=call_module]; -"318 linear_1_scale_0" [id=318, type=get_attr]; -"319 linear_1_zero_point_0" [id=319, type=get_attr]; -"320 _frozen_param2" [id=320, type=get_attr]; -"321 dequantize_per_channel_default_2" [id=321, type=dequantize_per_channel]; -"322 linear_1" [id=322, type=linear]; -"323 view" [id=323, type=view]; -"324 index" [id=324, type=index]; -"325 view_1" [id=325, type=view]; -"326 permute_1" [id=326, type=permute]; -"327 contiguous" [id=327, type=contiguous]; -"328 unsqueeze" [id=328, type=unsqueeze]; -"329 sigmoid" [id=329, type=sigmoid]; -"330 mul" [id=330, type=mul]; -"331 pad" [id=331, type=pad]; -"332 view_2" [id=332, type=view]; -"333 permute_2" [id=333, type=permute]; -"334 reshape" [id=334, type=reshape]; -"335 clone" [id=335, type=clone]; -"336 reshape_0_0_nncf_smooth_quant_0" [id=336, type=call_module]; -"337 quantize_per_tensor_default_1" [id=337, type=quantize_per_tensor]; -"338 dequantize_per_tensor_default_1" [id=338, type=dequantize_per_tensor]; -"339 linear_2_scale_0" [id=339, type=get_attr]; -"340 linear_2_zero_point_0" [id=340, type=get_attr]; -"341 _frozen_param3" [id=341, type=get_attr]; -"342 dequantize_per_channel_default_3" [id=342, type=dequantize_per_channel]; -"343 linear_2" [id=343, type=linear]; -"344 reshape_1" [id=344, type=reshape]; -"345 permute_3" [id=345, type=permute]; -"346 select" [id=346, type=select]; -"347 select_1" [id=347, type=select]; -"348 select_2" [id=348, type=select]; -"349 linalg_vector_norm" [id=349, type=linalg_vector_norm]; -"350 clamp_min" [id=350, type=clamp_min]; -"351 expand_as" [id=351, type=expand_as]; -"352 div" [id=352, type=div]; -"353 quantize_per_tensor_default_2" [id=353, type=quantize_per_tensor]; -"354 dequantize_per_tensor_default_2" [id=354, type=dequantize_per_tensor]; -"355 linalg_vector_norm_1" [id=355, type=linalg_vector_norm]; -"356 clamp_min_1" [id=356, type=clamp_min]; -"357 expand_as_1" [id=357, type=expand_as]; -"358 div_1" [id=358, type=div]; -"359 quantize_per_tensor_default_3" [id=359, type=quantize_per_tensor]; -"360 dequantize_per_tensor_default_3" [id=360, type=dequantize_per_tensor]; -"361 transpose" [id=361, type=transpose]; -"362 matmul" [id=362, type=matmul]; -"363 clamp" [id=363, type=clamp]; -"364 exp" [id=364, type=exp]; -"365 mul_1" [id=365, type=mul]; -"366 add" [id=366, type=add]; -"367 softmax" [id=367, type=softmax]; -"368 dropout" [id=368, type=dropout]; -"369 matmul_1" [id=369, type=matmul]; -"370 transpose_1" [id=370, type=transpose]; -"371 reshape_2" [id=371, type=reshape]; -"372 reshape_2_0_0_nncf_smooth_quant_0" [id=372, type=call_module]; -"373 quantize_per_tensor_default_4" [id=373, type=quantize_per_tensor]; -"374 dequantize_per_tensor_default_4" [id=374, type=dequantize_per_tensor]; -"375 linear_3_scale_0" [id=375, type=get_attr]; -"376 linear_3_zero_point_0" [id=376, type=get_attr]; -"377 _frozen_param4" [id=377, type=get_attr]; -"378 dequantize_per_channel_default_4" [id=378, type=dequantize_per_channel]; -"379 linear_3" [id=379, type=linear]; -"380 dropout_1" [id=380, type=dropout]; -"381 view_3" [id=381, type=view]; -"382 permute_4" [id=382, type=permute]; -"383 reshape_3" [id=383, type=reshape]; -"384 slice_2" [id=384, type=slice]; -"385 slice_3" [id=385, type=slice]; -"386 layer_norm_1" [id=386, type=layer_norm]; -"387 add_1" [id=387, type=add]; -"388 add_1_0_0_nncf_smooth_quant_0" [id=388, type=call_module]; -"389 quantize_per_tensor_default_5" [id=389, type=quantize_per_tensor]; -"390 dequantize_per_tensor_default_5" [id=390, type=dequantize_per_tensor]; -"391 linear_4_scale_0" [id=391, type=get_attr]; -"392 linear_4_zero_point_0" [id=392, type=get_attr]; -"393 _frozen_param5" [id=393, type=get_attr]; -"394 dequantize_per_channel_default_5" [id=394, type=dequantize_per_channel]; -"395 linear_4" [id=395, type=linear]; -"396 gelu" [id=396, type=gelu]; -"397 dropout_2" [id=397, type=dropout]; -"398 dropout_2_0_0_nncf_smooth_quant_0" [id=398, type=call_module]; -"399 quantize_per_tensor_default_6" [id=399, type=quantize_per_tensor]; -"400 dequantize_per_tensor_default_6" [id=400, type=dequantize_per_tensor]; -"401 linear_5_scale_0" [id=401, type=get_attr]; -"402 linear_5_zero_point_0" [id=402, type=get_attr]; -"403 _frozen_param6" [id=403, type=get_attr]; -"404 dequantize_per_channel_default_6" [id=404, type=dequantize_per_channel]; -"405 linear_5" [id=405, type=linear]; -"406 dropout_3" [id=406, type=dropout]; -"407 layer_norm_2" [id=407, type=layer_norm]; -"408 add_2" [id=408, type=add]; -"409 features_1_1_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=409, type=call_module]; -"410 linear_6_scale_0" [id=410, type=get_attr]; -"411 linear_6_zero_point_0" [id=411, type=get_attr]; -"412 _frozen_param7" [id=412, type=get_attr]; -"413 dequantize_per_channel_default_7" [id=413, type=dequantize_per_channel]; -"414 linear_6" [id=414, type=linear]; -"415 relu__1" [id=415, type=relu_]; -"416 relu__1_0_0_nncf_smooth_quant_0" [id=416, type=call_module]; -"417 linear_7_scale_0" [id=417, type=get_attr]; -"418 linear_7_zero_point_0" [id=418, type=get_attr]; -"419 _frozen_param8" [id=419, type=get_attr]; -"420 dequantize_per_channel_default_8" [id=420, type=dequantize_per_channel]; -"421 linear_7" [id=421, type=linear]; -"422 view_4" [id=422, type=view]; -"423 index_1" [id=423, type=index]; -"424 view_5" [id=424, type=view]; -"425 permute_5" [id=425, type=permute]; -"426 contiguous_1" [id=426, type=contiguous]; -"427 unsqueeze_1" [id=427, type=unsqueeze]; -"428 sigmoid_1" [id=428, type=sigmoid]; -"429 mul_2" [id=429, type=mul]; -"430 pad_1" [id=430, type=pad]; -"431 roll" [id=431, type=roll]; -"432 view_6" [id=432, type=view]; -"433 permute_6" [id=433, type=permute]; -"434 reshape_4" [id=434, type=reshape]; -"435 clone_1" [id=435, type=clone]; -"436 reshape_4_0_0_nncf_smooth_quant_0" [id=436, type=call_module]; -"437 quantize_per_tensor_default_7" [id=437, type=quantize_per_tensor]; -"438 dequantize_per_tensor_default_7" [id=438, type=dequantize_per_tensor]; -"439 linear_8_scale_0" [id=439, type=get_attr]; -"440 linear_8_zero_point_0" [id=440, type=get_attr]; -"441 _frozen_param9" [id=441, type=get_attr]; -"442 dequantize_per_channel_default_9" [id=442, type=dequantize_per_channel]; -"443 linear_8" [id=443, type=linear]; -"444 reshape_5" [id=444, type=reshape]; -"445 permute_7" [id=445, type=permute]; -"446 select_3" [id=446, type=select]; -"447 select_4" [id=447, type=select]; -"448 select_5" [id=448, type=select]; -"449 linalg_vector_norm_2" [id=449, type=linalg_vector_norm]; -"450 clamp_min_2" [id=450, type=clamp_min]; -"451 expand_as_2" [id=451, type=expand_as]; -"452 div_2" [id=452, type=div]; -"453 quantize_per_tensor_default_8" [id=453, type=quantize_per_tensor]; -"454 dequantize_per_tensor_default_8" [id=454, type=dequantize_per_tensor]; -"455 linalg_vector_norm_3" [id=455, type=linalg_vector_norm]; -"456 clamp_min_3" [id=456, type=clamp_min]; -"457 expand_as_3" [id=457, type=expand_as]; -"458 div_3" [id=458, type=div]; -"459 quantize_per_tensor_default_9" [id=459, type=quantize_per_tensor]; -"460 dequantize_per_tensor_default_9" [id=460, type=dequantize_per_tensor]; -"461 transpose_2" [id=461, type=transpose]; -"462 matmul_2" [id=462, type=matmul]; -"463 clamp_1" [id=463, type=clamp]; -"464 exp_1" [id=464, type=exp]; -"465 mul_3" [id=465, type=mul]; -"466 add_3" [id=466, type=add]; -"467 new_zeros" [id=467, type=new_zeros]; -"468 view_7" [id=468, type=view]; -"469 permute_8" [id=469, type=permute]; -"470 reshape_6" [id=470, type=reshape]; -"471 unsqueeze_2" [id=471, type=unsqueeze]; -"472 unsqueeze_3" [id=472, type=unsqueeze]; -"473 sub" [id=473, type=sub]; -"474 ne" [id=474, type=ne]; -"475 masked_fill" [id=475, type=masked_fill]; -"476 eq" [id=476, type=eq]; -"477 masked_fill_1" [id=477, type=masked_fill]; -"478 view_8" [id=478, type=view]; -"479 unsqueeze_4" [id=479, type=unsqueeze]; -"480 unsqueeze_5" [id=480, type=unsqueeze]; -"481 add_4" [id=481, type=add]; -"482 view_9" [id=482, type=view]; -"483 softmax_1" [id=483, type=softmax]; -"484 dropout_4" [id=484, type=dropout]; -"485 matmul_3" [id=485, type=matmul]; -"486 transpose_3" [id=486, type=transpose]; -"487 reshape_7" [id=487, type=reshape]; -"488 reshape_7_0_0_nncf_smooth_quant_0" [id=488, type=call_module]; -"489 quantize_per_tensor_default_10" [id=489, type=quantize_per_tensor]; -"490 dequantize_per_tensor_default_10" [id=490, type=dequantize_per_tensor]; -"491 linear_9_scale_0" [id=491, type=get_attr]; -"492 linear_9_zero_point_0" [id=492, type=get_attr]; -"493 _frozen_param10" [id=493, type=get_attr]; -"494 dequantize_per_channel_default_10" [id=494, type=dequantize_per_channel]; -"495 linear_9" [id=495, type=linear]; -"496 dropout_5" [id=496, type=dropout]; -"497 view_10" [id=497, type=view]; -"498 permute_9" [id=498, type=permute]; -"499 reshape_8" [id=499, type=reshape]; -"500 roll_1" [id=500, type=roll]; -"501 slice_23" [id=501, type=slice]; -"502 slice_24" [id=502, type=slice]; -"503 layer_norm_3" [id=503, type=layer_norm]; -"504 add_5" [id=504, type=add]; -"505 add_5_0_0_nncf_smooth_quant_0" [id=505, type=call_module]; -"506 quantize_per_tensor_default_11" [id=506, type=quantize_per_tensor]; -"507 dequantize_per_tensor_default_11" [id=507, type=dequantize_per_tensor]; -"508 linear_10_scale_0" [id=508, type=get_attr]; -"509 linear_10_zero_point_0" [id=509, type=get_attr]; -"510 _frozen_param11" [id=510, type=get_attr]; -"511 dequantize_per_channel_default_11" [id=511, type=dequantize_per_channel]; -"512 linear_10" [id=512, type=linear]; -"513 gelu_1" [id=513, type=gelu]; -"514 dropout_6" [id=514, type=dropout]; -"515 dropout_6_0_0_nncf_smooth_quant_0" [id=515, type=call_module]; -"516 quantize_per_tensor_default_12" [id=516, type=quantize_per_tensor]; -"517 dequantize_per_tensor_default_12" [id=517, type=dequantize_per_tensor]; -"518 linear_11_scale_0" [id=518, type=get_attr]; -"519 linear_11_zero_point_0" [id=519, type=get_attr]; -"520 _frozen_param12" [id=520, type=get_attr]; -"521 dequantize_per_channel_default_12" [id=521, type=dequantize_per_channel]; -"522 linear_11" [id=522, type=linear]; -"523 dropout_7" [id=523, type=dropout]; -"524 layer_norm_4" [id=524, type=layer_norm]; -"525 add_6" [id=525, type=add]; -"526 pad_2" [id=526, type=pad]; -"527 slice_25" [id=527, type=slice]; -"528 slice_26" [id=528, type=slice]; -"529 slice_27" [id=529, type=slice]; -"530 slice_28" [id=530, type=slice]; -"531 slice_29" [id=531, type=slice]; -"532 slice_30" [id=532, type=slice]; -"533 slice_31" [id=533, type=slice]; -"534 slice_32" [id=534, type=slice]; -"535 slice_33" [id=535, type=slice]; -"536 slice_34" [id=536, type=slice]; -"537 slice_35" [id=537, type=slice]; -"538 slice_36" [id=538, type=slice]; -"539 cat" [id=539, type=cat]; -"540 cat_0_0_nncf_smooth_quant_0" [id=540, type=call_module]; -"541 quantize_per_tensor_default_13" [id=541, type=quantize_per_tensor]; -"542 dequantize_per_tensor_default_13" [id=542, type=dequantize_per_tensor]; -"543 linear_12_scale_0" [id=543, type=get_attr]; -"544 linear_12_zero_point_0" [id=544, type=get_attr]; -"545 _frozen_param13" [id=545, type=get_attr]; -"546 dequantize_per_channel_default_13" [id=546, type=dequantize_per_channel]; -"547 linear_12" [id=547, type=linear]; -"548 layer_norm_5" [id=548, type=layer_norm]; -"549 features_3_0_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=549, type=call_module]; -"550 linear_13_scale_0" [id=550, type=get_attr]; -"551 linear_13_zero_point_0" [id=551, type=get_attr]; -"552 _frozen_param14" [id=552, type=get_attr]; -"553 dequantize_per_channel_default_14" [id=553, type=dequantize_per_channel]; -"554 linear_13" [id=554, type=linear]; -"555 relu__2" [id=555, type=relu_]; -"556 relu__2_0_0_nncf_smooth_quant_0" [id=556, type=call_module]; -"557 linear_14_scale_0" [id=557, type=get_attr]; -"558 linear_14_zero_point_0" [id=558, type=get_attr]; -"559 _frozen_param15" [id=559, type=get_attr]; -"560 dequantize_per_channel_default_15" [id=560, type=dequantize_per_channel]; -"561 linear_14" [id=561, type=linear]; -"562 view_11" [id=562, type=view]; -"563 index_2" [id=563, type=index]; -"564 view_12" [id=564, type=view]; -"565 permute_10" [id=565, type=permute]; -"566 contiguous_2" [id=566, type=contiguous]; -"567 unsqueeze_6" [id=567, type=unsqueeze]; -"568 sigmoid_2" [id=568, type=sigmoid]; -"569 mul_4" [id=569, type=mul]; -"570 pad_3" [id=570, type=pad]; -"571 view_13" [id=571, type=view]; -"572 permute_11" [id=572, type=permute]; -"573 reshape_9" [id=573, type=reshape]; -"574 clone_2" [id=574, type=clone]; -"575 reshape_9_0_0_nncf_smooth_quant_0" [id=575, type=call_module]; -"576 quantize_per_tensor_default_14" [id=576, type=quantize_per_tensor]; -"577 dequantize_per_tensor_default_14" [id=577, type=dequantize_per_tensor]; -"578 linear_15_scale_0" [id=578, type=get_attr]; -"579 linear_15_zero_point_0" [id=579, type=get_attr]; -"580 _frozen_param16" [id=580, type=get_attr]; -"581 dequantize_per_channel_default_16" [id=581, type=dequantize_per_channel]; -"582 linear_15" [id=582, type=linear]; -"583 reshape_10" [id=583, type=reshape]; -"584 permute_12" [id=584, type=permute]; -"585 select_6" [id=585, type=select]; -"586 select_7" [id=586, type=select]; -"587 select_8" [id=587, type=select]; -"588 linalg_vector_norm_4" [id=588, type=linalg_vector_norm]; -"589 clamp_min_4" [id=589, type=clamp_min]; -"590 expand_as_4" [id=590, type=expand_as]; -"591 div_4" [id=591, type=div]; -"592 quantize_per_tensor_default_15" [id=592, type=quantize_per_tensor]; -"593 dequantize_per_tensor_default_15" [id=593, type=dequantize_per_tensor]; -"594 linalg_vector_norm_5" [id=594, type=linalg_vector_norm]; -"595 clamp_min_5" [id=595, type=clamp_min]; -"596 expand_as_5" [id=596, type=expand_as]; -"597 div_5" [id=597, type=div]; -"598 quantize_per_tensor_default_16" [id=598, type=quantize_per_tensor]; -"599 dequantize_per_tensor_default_16" [id=599, type=dequantize_per_tensor]; -"600 transpose_4" [id=600, type=transpose]; -"601 matmul_4" [id=601, type=matmul]; -"602 clamp_2" [id=602, type=clamp]; -"603 exp_2" [id=603, type=exp]; -"604 mul_5" [id=604, type=mul]; -"605 add_7" [id=605, type=add]; -"606 softmax_2" [id=606, type=softmax]; -"607 dropout_8" [id=607, type=dropout]; -"608 matmul_5" [id=608, type=matmul]; -"609 transpose_5" [id=609, type=transpose]; -"610 reshape_11" [id=610, type=reshape]; -"611 reshape_11_0_0_nncf_smooth_quant_0" [id=611, type=call_module]; -"612 quantize_per_tensor_default_17" [id=612, type=quantize_per_tensor]; -"613 dequantize_per_tensor_default_17" [id=613, type=dequantize_per_tensor]; -"614 linear_16_scale_0" [id=614, type=get_attr]; -"615 linear_16_zero_point_0" [id=615, type=get_attr]; -"616 _frozen_param17" [id=616, type=get_attr]; -"617 dequantize_per_channel_default_17" [id=617, type=dequantize_per_channel]; -"618 linear_16" [id=618, type=linear]; -"619 dropout_9" [id=619, type=dropout]; -"620 view_14" [id=620, type=view]; -"621 permute_13" [id=621, type=permute]; -"622 reshape_12" [id=622, type=reshape]; -"623 slice_38" [id=623, type=slice]; -"624 slice_39" [id=624, type=slice]; -"625 slice_40" [id=625, type=slice]; -"626 slice_41" [id=626, type=slice]; -"627 contiguous_3" [id=627, type=contiguous]; -"628 layer_norm_6" [id=628, type=layer_norm]; -"629 add_8" [id=629, type=add]; -"630 add_8_0_0_nncf_smooth_quant_0" [id=630, type=call_module]; -"631 quantize_per_tensor_default_18" [id=631, type=quantize_per_tensor]; -"632 dequantize_per_tensor_default_18" [id=632, type=dequantize_per_tensor]; -"633 linear_17_scale_0" [id=633, type=get_attr]; -"634 linear_17_zero_point_0" [id=634, type=get_attr]; -"635 _frozen_param18" [id=635, type=get_attr]; -"636 dequantize_per_channel_default_18" [id=636, type=dequantize_per_channel]; -"637 linear_17" [id=637, type=linear]; -"638 gelu_2" [id=638, type=gelu]; -"639 dropout_10" [id=639, type=dropout]; -"640 dropout_10_0_0_nncf_smooth_quant_0" [id=640, type=call_module]; -"641 quantize_per_tensor_default_19" [id=641, type=quantize_per_tensor]; -"642 dequantize_per_tensor_default_19" [id=642, type=dequantize_per_tensor]; -"643 linear_18_scale_0" [id=643, type=get_attr]; -"644 linear_18_zero_point_0" [id=644, type=get_attr]; -"645 _frozen_param19" [id=645, type=get_attr]; -"646 dequantize_per_channel_default_19" [id=646, type=dequantize_per_channel]; -"647 linear_18" [id=647, type=linear]; -"648 dropout_11" [id=648, type=dropout]; -"649 layer_norm_7" [id=649, type=layer_norm]; -"650 add_9" [id=650, type=add]; -"651 features_3_1_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=651, type=call_module]; -"652 linear_19_scale_0" [id=652, type=get_attr]; -"653 linear_19_zero_point_0" [id=653, type=get_attr]; -"654 _frozen_param20" [id=654, type=get_attr]; -"655 dequantize_per_channel_default_20" [id=655, type=dequantize_per_channel]; -"656 linear_19" [id=656, type=linear]; -"657 relu__3" [id=657, type=relu_]; -"658 relu__3_0_0_nncf_smooth_quant_0" [id=658, type=call_module]; -"659 linear_20_scale_0" [id=659, type=get_attr]; -"660 linear_20_zero_point_0" [id=660, type=get_attr]; -"661 _frozen_param21" [id=661, type=get_attr]; -"662 dequantize_per_channel_default_21" [id=662, type=dequantize_per_channel]; -"663 linear_20" [id=663, type=linear]; -"664 view_15" [id=664, type=view]; -"665 index_3" [id=665, type=index]; -"666 view_16" [id=666, type=view]; -"667 permute_14" [id=667, type=permute]; -"668 contiguous_4" [id=668, type=contiguous]; -"669 unsqueeze_7" [id=669, type=unsqueeze]; -"670 sigmoid_3" [id=670, type=sigmoid]; -"671 mul_6" [id=671, type=mul]; -"672 pad_4" [id=672, type=pad]; -"673 roll_2" [id=673, type=roll]; -"674 view_17" [id=674, type=view]; -"675 permute_15" [id=675, type=permute]; -"676 reshape_13" [id=676, type=reshape]; -"677 clone_3" [id=677, type=clone]; -"678 reshape_13_0_0_nncf_smooth_quant_0" [id=678, type=call_module]; -"679 quantize_per_tensor_default_20" [id=679, type=quantize_per_tensor]; -"680 dequantize_per_tensor_default_20" [id=680, type=dequantize_per_tensor]; -"681 linear_21_scale_0" [id=681, type=get_attr]; -"682 linear_21_zero_point_0" [id=682, type=get_attr]; -"683 _frozen_param22" [id=683, type=get_attr]; -"684 dequantize_per_channel_default_22" [id=684, type=dequantize_per_channel]; -"685 linear_21" [id=685, type=linear]; -"686 reshape_14" [id=686, type=reshape]; -"687 permute_16" [id=687, type=permute]; -"688 select_9" [id=688, type=select]; -"689 select_10" [id=689, type=select]; -"690 select_11" [id=690, type=select]; -"691 linalg_vector_norm_6" [id=691, type=linalg_vector_norm]; -"692 clamp_min_6" [id=692, type=clamp_min]; -"693 expand_as_6" [id=693, type=expand_as]; -"694 div_6" [id=694, type=div]; -"695 quantize_per_tensor_default_21" [id=695, type=quantize_per_tensor]; -"696 dequantize_per_tensor_default_21" [id=696, type=dequantize_per_tensor]; -"697 linalg_vector_norm_7" [id=697, type=linalg_vector_norm]; -"698 clamp_min_7" [id=698, type=clamp_min]; -"699 expand_as_7" [id=699, type=expand_as]; -"700 div_7" [id=700, type=div]; -"701 quantize_per_tensor_default_22" [id=701, type=quantize_per_tensor]; -"702 dequantize_per_tensor_default_22" [id=702, type=dequantize_per_tensor]; -"703 transpose_6" [id=703, type=transpose]; -"704 matmul_6" [id=704, type=matmul]; -"705 clamp_3" [id=705, type=clamp]; -"706 exp_3" [id=706, type=exp]; -"707 mul_7" [id=707, type=mul]; -"708 add_10" [id=708, type=add]; -"709 new_zeros_1" [id=709, type=new_zeros]; -"710 view_18" [id=710, type=view]; -"711 permute_17" [id=711, type=permute]; -"712 reshape_15" [id=712, type=reshape]; -"713 unsqueeze_8" [id=713, type=unsqueeze]; -"714 unsqueeze_9" [id=714, type=unsqueeze]; -"715 sub_1" [id=715, type=sub]; -"716 ne_1" [id=716, type=ne]; -"717 masked_fill_2" [id=717, type=masked_fill]; -"718 eq_1" [id=718, type=eq]; -"719 masked_fill_3" [id=719, type=masked_fill]; -"720 view_19" [id=720, type=view]; -"721 unsqueeze_10" [id=721, type=unsqueeze]; -"722 unsqueeze_11" [id=722, type=unsqueeze]; -"723 add_11" [id=723, type=add]; -"724 view_20" [id=724, type=view]; -"725 softmax_3" [id=725, type=softmax]; -"726 dropout_12" [id=726, type=dropout]; -"727 matmul_7" [id=727, type=matmul]; -"728 transpose_7" [id=728, type=transpose]; -"729 reshape_16" [id=729, type=reshape]; -"730 reshape_16_0_0_nncf_smooth_quant_0" [id=730, type=call_module]; -"731 quantize_per_tensor_default_23" [id=731, type=quantize_per_tensor]; -"732 dequantize_per_tensor_default_23" [id=732, type=dequantize_per_tensor]; -"733 linear_22_scale_0" [id=733, type=get_attr]; -"734 linear_22_zero_point_0" [id=734, type=get_attr]; -"735 _frozen_param23" [id=735, type=get_attr]; -"736 dequantize_per_channel_default_23" [id=736, type=dequantize_per_channel]; -"737 linear_22" [id=737, type=linear]; -"738 dropout_13" [id=738, type=dropout]; -"739 view_21" [id=739, type=view]; -"740 permute_18" [id=740, type=permute]; -"741 reshape_17" [id=741, type=reshape]; -"742 roll_3" [id=742, type=roll]; -"743 slice_61" [id=743, type=slice]; -"744 slice_62" [id=744, type=slice]; -"745 slice_63" [id=745, type=slice]; -"746 slice_64" [id=746, type=slice]; -"747 contiguous_5" [id=747, type=contiguous]; -"748 layer_norm_8" [id=748, type=layer_norm]; -"749 add_12" [id=749, type=add]; -"750 add_12_0_0_nncf_smooth_quant_0" [id=750, type=call_module]; -"751 quantize_per_tensor_default_24" [id=751, type=quantize_per_tensor]; -"752 dequantize_per_tensor_default_24" [id=752, type=dequantize_per_tensor]; -"753 linear_23_scale_0" [id=753, type=get_attr]; -"754 linear_23_zero_point_0" [id=754, type=get_attr]; -"755 _frozen_param24" [id=755, type=get_attr]; -"756 dequantize_per_channel_default_24" [id=756, type=dequantize_per_channel]; -"757 linear_23" [id=757, type=linear]; -"758 gelu_3" [id=758, type=gelu]; -"759 dropout_14" [id=759, type=dropout]; -"760 dropout_14_0_0_nncf_smooth_quant_0" [id=760, type=call_module]; -"761 quantize_per_tensor_default_25" [id=761, type=quantize_per_tensor]; -"762 dequantize_per_tensor_default_25" [id=762, type=dequantize_per_tensor]; -"763 linear_24_scale_0" [id=763, type=get_attr]; -"764 linear_24_zero_point_0" [id=764, type=get_attr]; -"765 _frozen_param25" [id=765, type=get_attr]; -"766 dequantize_per_channel_default_25" [id=766, type=dequantize_per_channel]; -"767 linear_24" [id=767, type=linear]; -"768 dropout_15" [id=768, type=dropout]; -"769 layer_norm_9" [id=769, type=layer_norm]; -"770 add_13" [id=770, type=add]; -"771 pad_5" [id=771, type=pad]; -"772 slice_65" [id=772, type=slice]; -"773 slice_66" [id=773, type=slice]; -"774 slice_67" [id=774, type=slice]; -"775 slice_68" [id=775, type=slice]; -"776 slice_69" [id=776, type=slice]; -"777 slice_70" [id=777, type=slice]; -"778 slice_71" [id=778, type=slice]; -"779 slice_72" [id=779, type=slice]; -"780 slice_73" [id=780, type=slice]; -"781 slice_74" [id=781, type=slice]; -"782 slice_75" [id=782, type=slice]; -"783 slice_76" [id=783, type=slice]; -"784 cat_1" [id=784, type=cat]; -"785 cat_1_0_0_nncf_smooth_quant_0" [id=785, type=call_module]; -"786 quantize_per_tensor_default_26" [id=786, type=quantize_per_tensor]; -"787 dequantize_per_tensor_default_26" [id=787, type=dequantize_per_tensor]; -"788 linear_25_scale_0" [id=788, type=get_attr]; -"789 linear_25_zero_point_0" [id=789, type=get_attr]; -"790 _frozen_param26" [id=790, type=get_attr]; -"791 dequantize_per_channel_default_26" [id=791, type=dequantize_per_channel]; -"792 linear_25" [id=792, type=linear]; -"793 layer_norm_10" [id=793, type=layer_norm]; -"794 features_5_0_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=794, type=call_module]; -"795 linear_26_scale_0" [id=795, type=get_attr]; -"796 linear_26_zero_point_0" [id=796, type=get_attr]; -"797 _frozen_param27" [id=797, type=get_attr]; -"798 dequantize_per_channel_default_27" [id=798, type=dequantize_per_channel]; -"799 linear_26" [id=799, type=linear]; -"800 relu__4" [id=800, type=relu_]; -"801 relu__4_0_0_nncf_smooth_quant_0" [id=801, type=call_module]; -"802 linear_27_scale_0" [id=802, type=get_attr]; -"803 linear_27_zero_point_0" [id=803, type=get_attr]; -"804 _frozen_param28" [id=804, type=get_attr]; -"805 dequantize_per_channel_default_28" [id=805, type=dequantize_per_channel]; -"806 linear_27" [id=806, type=linear]; -"807 view_22" [id=807, type=view]; -"808 index_4" [id=808, type=index]; -"809 view_23" [id=809, type=view]; -"810 permute_19" [id=810, type=permute]; -"811 contiguous_6" [id=811, type=contiguous]; -"812 unsqueeze_12" [id=812, type=unsqueeze]; -"813 sigmoid_4" [id=813, type=sigmoid]; -"814 mul_8" [id=814, type=mul]; -"815 pad_6" [id=815, type=pad]; -"816 view_24" [id=816, type=view]; -"817 permute_20" [id=817, type=permute]; -"818 reshape_18" [id=818, type=reshape]; -"819 clone_4" [id=819, type=clone]; -"820 reshape_18_0_0_nncf_smooth_quant_0" [id=820, type=call_module]; -"821 quantize_per_tensor_default_27" [id=821, type=quantize_per_tensor]; -"822 dequantize_per_tensor_default_27" [id=822, type=dequantize_per_tensor]; -"823 linear_28_scale_0" [id=823, type=get_attr]; -"824 linear_28_zero_point_0" [id=824, type=get_attr]; -"825 _frozen_param29" [id=825, type=get_attr]; -"826 dequantize_per_channel_default_29" [id=826, type=dequantize_per_channel]; -"827 linear_28" [id=827, type=linear]; -"828 reshape_19" [id=828, type=reshape]; -"829 permute_21" [id=829, type=permute]; -"830 select_12" [id=830, type=select]; -"831 select_13" [id=831, type=select]; -"832 select_14" [id=832, type=select]; -"833 linalg_vector_norm_8" [id=833, type=linalg_vector_norm]; -"834 clamp_min_8" [id=834, type=clamp_min]; -"835 expand_as_8" [id=835, type=expand_as]; -"836 div_8" [id=836, type=div]; -"837 quantize_per_tensor_default_28" [id=837, type=quantize_per_tensor]; -"838 dequantize_per_tensor_default_28" [id=838, type=dequantize_per_tensor]; -"839 linalg_vector_norm_9" [id=839, type=linalg_vector_norm]; -"840 clamp_min_9" [id=840, type=clamp_min]; -"841 expand_as_9" [id=841, type=expand_as]; -"842 div_9" [id=842, type=div]; -"843 quantize_per_tensor_default_29" [id=843, type=quantize_per_tensor]; -"844 dequantize_per_tensor_default_29" [id=844, type=dequantize_per_tensor]; -"845 transpose_8" [id=845, type=transpose]; -"846 matmul_8" [id=846, type=matmul]; -"847 clamp_4" [id=847, type=clamp]; -"848 exp_4" [id=848, type=exp]; -"849 mul_9" [id=849, type=mul]; -"850 add_14" [id=850, type=add]; -"851 softmax_4" [id=851, type=softmax]; -"852 dropout_16" [id=852, type=dropout]; -"853 matmul_9" [id=853, type=matmul]; -"854 transpose_9" [id=854, type=transpose]; -"855 reshape_20" [id=855, type=reshape]; -"856 reshape_20_0_0_nncf_smooth_quant_0" [id=856, type=call_module]; -"857 quantize_per_tensor_default_30" [id=857, type=quantize_per_tensor]; -"858 dequantize_per_tensor_default_30" [id=858, type=dequantize_per_tensor]; -"859 linear_29_scale_0" [id=859, type=get_attr]; -"860 linear_29_zero_point_0" [id=860, type=get_attr]; -"861 _frozen_param30" [id=861, type=get_attr]; -"862 dequantize_per_channel_default_30" [id=862, type=dequantize_per_channel]; -"863 linear_29" [id=863, type=linear]; -"864 dropout_17" [id=864, type=dropout]; -"865 view_25" [id=865, type=view]; -"866 permute_22" [id=866, type=permute]; -"867 reshape_21" [id=867, type=reshape]; -"868 slice_78" [id=868, type=slice]; -"869 slice_79" [id=869, type=slice]; -"870 slice_80" [id=870, type=slice]; -"871 slice_81" [id=871, type=slice]; -"872 contiguous_7" [id=872, type=contiguous]; -"873 layer_norm_11" [id=873, type=layer_norm]; -"874 add_15" [id=874, type=add]; -"875 add_15_0_0_nncf_smooth_quant_0" [id=875, type=call_module]; -"876 quantize_per_tensor_default_31" [id=876, type=quantize_per_tensor]; -"877 dequantize_per_tensor_default_31" [id=877, type=dequantize_per_tensor]; -"878 linear_30_scale_0" [id=878, type=get_attr]; -"879 linear_30_zero_point_0" [id=879, type=get_attr]; -"880 _frozen_param31" [id=880, type=get_attr]; -"881 dequantize_per_channel_default_31" [id=881, type=dequantize_per_channel]; -"882 linear_30" [id=882, type=linear]; -"883 gelu_4" [id=883, type=gelu]; -"884 dropout_18" [id=884, type=dropout]; -"885 dropout_18_0_0_nncf_smooth_quant_0" [id=885, type=call_module]; -"886 quantize_per_tensor_default_32" [id=886, type=quantize_per_tensor]; -"887 dequantize_per_tensor_default_32" [id=887, type=dequantize_per_tensor]; -"888 linear_31_scale_0" [id=888, type=get_attr]; -"889 linear_31_zero_point_0" [id=889, type=get_attr]; -"890 _frozen_param32" [id=890, type=get_attr]; -"891 dequantize_per_channel_default_32" [id=891, type=dequantize_per_channel]; -"892 linear_31" [id=892, type=linear]; -"893 dropout_19" [id=893, type=dropout]; -"894 layer_norm_12" [id=894, type=layer_norm]; -"895 add_16" [id=895, type=add]; -"896 features_5_1_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=896, type=call_module]; -"897 linear_32_scale_0" [id=897, type=get_attr]; -"898 linear_32_zero_point_0" [id=898, type=get_attr]; -"899 _frozen_param33" [id=899, type=get_attr]; -"900 dequantize_per_channel_default_33" [id=900, type=dequantize_per_channel]; -"901 linear_32" [id=901, type=linear]; -"902 relu__5" [id=902, type=relu_]; -"903 relu__5_0_0_nncf_smooth_quant_0" [id=903, type=call_module]; -"904 linear_33_scale_0" [id=904, type=get_attr]; -"905 linear_33_zero_point_0" [id=905, type=get_attr]; -"906 _frozen_param34" [id=906, type=get_attr]; -"907 dequantize_per_channel_default_34" [id=907, type=dequantize_per_channel]; -"908 linear_33" [id=908, type=linear]; -"909 view_26" [id=909, type=view]; -"910 index_5" [id=910, type=index]; -"911 view_27" [id=911, type=view]; -"912 permute_23" [id=912, type=permute]; -"913 contiguous_8" [id=913, type=contiguous]; -"914 unsqueeze_13" [id=914, type=unsqueeze]; -"915 sigmoid_5" [id=915, type=sigmoid]; -"916 mul_10" [id=916, type=mul]; -"917 pad_7" [id=917, type=pad]; -"918 roll_4" [id=918, type=roll]; -"919 view_28" [id=919, type=view]; -"920 permute_24" [id=920, type=permute]; -"921 reshape_22" [id=921, type=reshape]; -"922 clone_5" [id=922, type=clone]; -"923 reshape_22_0_0_nncf_smooth_quant_0" [id=923, type=call_module]; -"924 quantize_per_tensor_default_33" [id=924, type=quantize_per_tensor]; -"925 dequantize_per_tensor_default_33" [id=925, type=dequantize_per_tensor]; -"926 linear_34_scale_0" [id=926, type=get_attr]; -"927 linear_34_zero_point_0" [id=927, type=get_attr]; -"928 _frozen_param35" [id=928, type=get_attr]; -"929 dequantize_per_channel_default_35" [id=929, type=dequantize_per_channel]; -"930 linear_34" [id=930, type=linear]; -"931 reshape_23" [id=931, type=reshape]; -"932 permute_25" [id=932, type=permute]; -"933 select_15" [id=933, type=select]; -"934 select_16" [id=934, type=select]; -"935 select_17" [id=935, type=select]; -"936 linalg_vector_norm_10" [id=936, type=linalg_vector_norm]; -"937 clamp_min_10" [id=937, type=clamp_min]; -"938 expand_as_10" [id=938, type=expand_as]; -"939 div_10" [id=939, type=div]; -"940 quantize_per_tensor_default_34" [id=940, type=quantize_per_tensor]; -"941 dequantize_per_tensor_default_34" [id=941, type=dequantize_per_tensor]; -"942 linalg_vector_norm_11" [id=942, type=linalg_vector_norm]; -"943 clamp_min_11" [id=943, type=clamp_min]; -"944 expand_as_11" [id=944, type=expand_as]; -"945 div_11" [id=945, type=div]; -"946 quantize_per_tensor_default_35" [id=946, type=quantize_per_tensor]; -"947 dequantize_per_tensor_default_35" [id=947, type=dequantize_per_tensor]; -"948 transpose_10" [id=948, type=transpose]; -"949 matmul_10" [id=949, type=matmul]; -"950 clamp_5" [id=950, type=clamp]; -"951 exp_5" [id=951, type=exp]; -"952 mul_11" [id=952, type=mul]; -"953 add_17" [id=953, type=add]; -"954 new_zeros_2" [id=954, type=new_zeros]; -"955 view_29" [id=955, type=view]; -"956 permute_26" [id=956, type=permute]; -"957 reshape_24" [id=957, type=reshape]; -"958 unsqueeze_14" [id=958, type=unsqueeze]; -"959 unsqueeze_15" [id=959, type=unsqueeze]; -"960 sub_2" [id=960, type=sub]; -"961 ne_2" [id=961, type=ne]; -"962 masked_fill_4" [id=962, type=masked_fill]; -"963 eq_2" [id=963, type=eq]; -"964 masked_fill_5" [id=964, type=masked_fill]; -"965 view_30" [id=965, type=view]; -"966 unsqueeze_16" [id=966, type=unsqueeze]; -"967 unsqueeze_17" [id=967, type=unsqueeze]; -"968 add_18" [id=968, type=add]; -"969 view_31" [id=969, type=view]; -"970 softmax_5" [id=970, type=softmax]; -"971 dropout_20" [id=971, type=dropout]; -"972 matmul_11" [id=972, type=matmul]; -"973 transpose_11" [id=973, type=transpose]; -"974 reshape_25" [id=974, type=reshape]; -"975 reshape_25_0_0_nncf_smooth_quant_0" [id=975, type=call_module]; -"976 quantize_per_tensor_default_36" [id=976, type=quantize_per_tensor]; -"977 dequantize_per_tensor_default_36" [id=977, type=dequantize_per_tensor]; -"978 linear_35_scale_0" [id=978, type=get_attr]; -"979 linear_35_zero_point_0" [id=979, type=get_attr]; -"980 _frozen_param36" [id=980, type=get_attr]; -"981 dequantize_per_channel_default_36" [id=981, type=dequantize_per_channel]; -"982 linear_35" [id=982, type=linear]; -"983 dropout_21" [id=983, type=dropout]; -"984 view_32" [id=984, type=view]; -"985 permute_27" [id=985, type=permute]; -"986 reshape_26" [id=986, type=reshape]; -"987 roll_5" [id=987, type=roll]; -"988 slice_101" [id=988, type=slice]; -"989 slice_102" [id=989, type=slice]; -"990 slice_103" [id=990, type=slice]; -"991 slice_104" [id=991, type=slice]; -"992 contiguous_9" [id=992, type=contiguous]; -"993 layer_norm_13" [id=993, type=layer_norm]; -"994 add_19" [id=994, type=add]; -"995 add_19_0_0_nncf_smooth_quant_0" [id=995, type=call_module]; -"996 quantize_per_tensor_default_37" [id=996, type=quantize_per_tensor]; -"997 dequantize_per_tensor_default_37" [id=997, type=dequantize_per_tensor]; -"998 linear_36_scale_0" [id=998, type=get_attr]; -"999 linear_36_zero_point_0" [id=999, type=get_attr]; -"1000 _frozen_param37" [id=1000, type=get_attr]; -"1001 dequantize_per_channel_default_37" [id=1001, type=dequantize_per_channel]; -"1002 linear_36" [id=1002, type=linear]; -"1003 gelu_5" [id=1003, type=gelu]; -"1004 dropout_22" [id=1004, type=dropout]; -"1005 dropout_22_0_0_nncf_smooth_quant_0" [id=1005, type=call_module]; -"1006 quantize_per_tensor_default_38" [id=1006, type=quantize_per_tensor]; -"1007 dequantize_per_tensor_default_38" [id=1007, type=dequantize_per_tensor]; -"1008 linear_37_scale_0" [id=1008, type=get_attr]; -"1009 linear_37_zero_point_0" [id=1009, type=get_attr]; -"1010 _frozen_param38" [id=1010, type=get_attr]; -"1011 dequantize_per_channel_default_38" [id=1011, type=dequantize_per_channel]; -"1012 linear_37" [id=1012, type=linear]; -"1013 dropout_23" [id=1013, type=dropout]; -"1014 layer_norm_14" [id=1014, type=layer_norm]; -"1015 add_20" [id=1015, type=add]; -"1016 features_5_2_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=1016, type=call_module]; -"1017 linear_38_scale_0" [id=1017, type=get_attr]; -"1018 linear_38_zero_point_0" [id=1018, type=get_attr]; -"1019 _frozen_param39" [id=1019, type=get_attr]; -"1020 dequantize_per_channel_default_39" [id=1020, type=dequantize_per_channel]; -"1021 linear_38" [id=1021, type=linear]; -"1022 relu__6" [id=1022, type=relu_]; -"1023 relu__6_0_0_nncf_smooth_quant_0" [id=1023, type=call_module]; -"1024 linear_39_scale_0" [id=1024, type=get_attr]; -"1025 linear_39_zero_point_0" [id=1025, type=get_attr]; -"1026 _frozen_param40" [id=1026, type=get_attr]; -"1027 dequantize_per_channel_default_40" [id=1027, type=dequantize_per_channel]; -"1028 linear_39" [id=1028, type=linear]; -"1029 view_33" [id=1029, type=view]; -"1030 index_6" [id=1030, type=index]; -"1031 view_34" [id=1031, type=view]; -"1032 permute_28" [id=1032, type=permute]; -"1033 contiguous_10" [id=1033, type=contiguous]; -"1034 unsqueeze_18" [id=1034, type=unsqueeze]; -"1035 sigmoid_6" [id=1035, type=sigmoid]; -"1036 mul_12" [id=1036, type=mul]; -"1037 pad_8" [id=1037, type=pad]; -"1038 view_35" [id=1038, type=view]; -"1039 permute_29" [id=1039, type=permute]; -"1040 reshape_27" [id=1040, type=reshape]; -"1041 clone_6" [id=1041, type=clone]; -"1042 reshape_27_0_0_nncf_smooth_quant_0" [id=1042, type=call_module]; -"1043 quantize_per_tensor_default_39" [id=1043, type=quantize_per_tensor]; -"1044 dequantize_per_tensor_default_39" [id=1044, type=dequantize_per_tensor]; -"1045 linear_40_scale_0" [id=1045, type=get_attr]; -"1046 linear_40_zero_point_0" [id=1046, type=get_attr]; -"1047 _frozen_param41" [id=1047, type=get_attr]; -"1048 dequantize_per_channel_default_41" [id=1048, type=dequantize_per_channel]; -"1049 linear_40" [id=1049, type=linear]; -"1050 reshape_28" [id=1050, type=reshape]; -"1051 permute_30" [id=1051, type=permute]; -"1052 select_18" [id=1052, type=select]; -"1053 select_19" [id=1053, type=select]; -"1054 select_20" [id=1054, type=select]; -"1055 linalg_vector_norm_12" [id=1055, type=linalg_vector_norm]; -"1056 clamp_min_12" [id=1056, type=clamp_min]; -"1057 expand_as_12" [id=1057, type=expand_as]; -"1058 div_12" [id=1058, type=div]; -"1059 quantize_per_tensor_default_40" [id=1059, type=quantize_per_tensor]; -"1060 dequantize_per_tensor_default_40" [id=1060, type=dequantize_per_tensor]; -"1061 linalg_vector_norm_13" [id=1061, type=linalg_vector_norm]; -"1062 clamp_min_13" [id=1062, type=clamp_min]; -"1063 expand_as_13" [id=1063, type=expand_as]; -"1064 div_13" [id=1064, type=div]; -"1065 quantize_per_tensor_default_41" [id=1065, type=quantize_per_tensor]; -"1066 dequantize_per_tensor_default_41" [id=1066, type=dequantize_per_tensor]; -"1067 transpose_12" [id=1067, type=transpose]; -"1068 matmul_12" [id=1068, type=matmul]; -"1069 clamp_6" [id=1069, type=clamp]; -"1070 exp_6" [id=1070, type=exp]; -"1071 mul_13" [id=1071, type=mul]; -"1072 add_21" [id=1072, type=add]; -"1073 softmax_6" [id=1073, type=softmax]; -"1074 dropout_24" [id=1074, type=dropout]; -"1075 matmul_13" [id=1075, type=matmul]; -"1076 transpose_13" [id=1076, type=transpose]; -"1077 reshape_29" [id=1077, type=reshape]; -"1078 reshape_29_0_0_nncf_smooth_quant_0" [id=1078, type=call_module]; -"1079 quantize_per_tensor_default_42" [id=1079, type=quantize_per_tensor]; -"1080 dequantize_per_tensor_default_42" [id=1080, type=dequantize_per_tensor]; -"1081 linear_41_scale_0" [id=1081, type=get_attr]; -"1082 linear_41_zero_point_0" [id=1082, type=get_attr]; -"1083 _frozen_param42" [id=1083, type=get_attr]; -"1084 dequantize_per_channel_default_42" [id=1084, type=dequantize_per_channel]; -"1085 linear_41" [id=1085, type=linear]; -"1086 dropout_25" [id=1086, type=dropout]; -"1087 view_36" [id=1087, type=view]; -"1088 permute_31" [id=1088, type=permute]; -"1089 reshape_30" [id=1089, type=reshape]; -"1090 slice_106" [id=1090, type=slice]; -"1091 slice_107" [id=1091, type=slice]; -"1092 slice_108" [id=1092, type=slice]; -"1093 slice_109" [id=1093, type=slice]; -"1094 contiguous_11" [id=1094, type=contiguous]; -"1095 layer_norm_15" [id=1095, type=layer_norm]; -"1096 add_22" [id=1096, type=add]; -"1097 add_22_0_0_nncf_smooth_quant_0" [id=1097, type=call_module]; -"1098 quantize_per_tensor_default_43" [id=1098, type=quantize_per_tensor]; -"1099 dequantize_per_tensor_default_43" [id=1099, type=dequantize_per_tensor]; -"1100 linear_42_scale_0" [id=1100, type=get_attr]; -"1101 linear_42_zero_point_0" [id=1101, type=get_attr]; -"1102 _frozen_param43" [id=1102, type=get_attr]; -"1103 dequantize_per_channel_default_43" [id=1103, type=dequantize_per_channel]; -"1104 linear_42" [id=1104, type=linear]; -"1105 gelu_6" [id=1105, type=gelu]; -"1106 dropout_26" [id=1106, type=dropout]; -"1107 dropout_26_0_0_nncf_smooth_quant_0" [id=1107, type=call_module]; -"1108 quantize_per_tensor_default_44" [id=1108, type=quantize_per_tensor]; -"1109 dequantize_per_tensor_default_44" [id=1109, type=dequantize_per_tensor]; -"1110 linear_43_scale_0" [id=1110, type=get_attr]; -"1111 linear_43_zero_point_0" [id=1111, type=get_attr]; -"1112 _frozen_param44" [id=1112, type=get_attr]; -"1113 dequantize_per_channel_default_44" [id=1113, type=dequantize_per_channel]; -"1114 linear_43" [id=1114, type=linear]; -"1115 dropout_27" [id=1115, type=dropout]; -"1116 layer_norm_16" [id=1116, type=layer_norm]; -"1117 add_23" [id=1117, type=add]; -"1118 features_5_3_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=1118, type=call_module]; -"1119 linear_44_scale_0" [id=1119, type=get_attr]; -"1120 linear_44_zero_point_0" [id=1120, type=get_attr]; -"1121 _frozen_param45" [id=1121, type=get_attr]; -"1122 dequantize_per_channel_default_45" [id=1122, type=dequantize_per_channel]; -"1123 linear_44" [id=1123, type=linear]; -"1124 relu__7" [id=1124, type=relu_]; -"1125 relu__7_0_0_nncf_smooth_quant_0" [id=1125, type=call_module]; -"1126 linear_45_scale_0" [id=1126, type=get_attr]; -"1127 linear_45_zero_point_0" [id=1127, type=get_attr]; -"1128 _frozen_param46" [id=1128, type=get_attr]; -"1129 dequantize_per_channel_default_46" [id=1129, type=dequantize_per_channel]; -"1130 linear_45" [id=1130, type=linear]; -"1131 view_37" [id=1131, type=view]; -"1132 index_7" [id=1132, type=index]; -"1133 view_38" [id=1133, type=view]; -"1134 permute_32" [id=1134, type=permute]; -"1135 contiguous_12" [id=1135, type=contiguous]; -"1136 unsqueeze_19" [id=1136, type=unsqueeze]; -"1137 sigmoid_7" [id=1137, type=sigmoid]; -"1138 mul_14" [id=1138, type=mul]; -"1139 pad_9" [id=1139, type=pad]; -"1140 roll_6" [id=1140, type=roll]; -"1141 view_39" [id=1141, type=view]; -"1142 permute_33" [id=1142, type=permute]; -"1143 reshape_31" [id=1143, type=reshape]; -"1144 clone_7" [id=1144, type=clone]; -"1145 reshape_31_0_0_nncf_smooth_quant_0" [id=1145, type=call_module]; -"1146 quantize_per_tensor_default_45" [id=1146, type=quantize_per_tensor]; -"1147 dequantize_per_tensor_default_45" [id=1147, type=dequantize_per_tensor]; -"1148 linear_46_scale_0" [id=1148, type=get_attr]; -"1149 linear_46_zero_point_0" [id=1149, type=get_attr]; -"1150 _frozen_param47" [id=1150, type=get_attr]; -"1151 dequantize_per_channel_default_47" [id=1151, type=dequantize_per_channel]; -"1152 linear_46" [id=1152, type=linear]; -"1153 reshape_32" [id=1153, type=reshape]; -"1154 permute_34" [id=1154, type=permute]; -"1155 select_21" [id=1155, type=select]; -"1156 select_22" [id=1156, type=select]; -"1157 select_23" [id=1157, type=select]; -"1158 linalg_vector_norm_14" [id=1158, type=linalg_vector_norm]; -"1159 clamp_min_14" [id=1159, type=clamp_min]; -"1160 expand_as_14" [id=1160, type=expand_as]; -"1161 div_14" [id=1161, type=div]; -"1162 quantize_per_tensor_default_46" [id=1162, type=quantize_per_tensor]; -"1163 dequantize_per_tensor_default_46" [id=1163, type=dequantize_per_tensor]; -"1164 linalg_vector_norm_15" [id=1164, type=linalg_vector_norm]; -"1165 clamp_min_15" [id=1165, type=clamp_min]; -"1166 expand_as_15" [id=1166, type=expand_as]; -"1167 div_15" [id=1167, type=div]; -"1168 quantize_per_tensor_default_47" [id=1168, type=quantize_per_tensor]; -"1169 dequantize_per_tensor_default_47" [id=1169, type=dequantize_per_tensor]; -"1170 transpose_14" [id=1170, type=transpose]; -"1171 matmul_14" [id=1171, type=matmul]; -"1172 clamp_7" [id=1172, type=clamp]; -"1173 exp_7" [id=1173, type=exp]; -"1174 mul_15" [id=1174, type=mul]; -"1175 add_24" [id=1175, type=add]; -"1176 new_zeros_3" [id=1176, type=new_zeros]; -"1177 view_40" [id=1177, type=view]; -"1178 permute_35" [id=1178, type=permute]; -"1179 reshape_33" [id=1179, type=reshape]; -"1180 unsqueeze_20" [id=1180, type=unsqueeze]; -"1181 unsqueeze_21" [id=1181, type=unsqueeze]; -"1182 sub_3" [id=1182, type=sub]; -"1183 ne_3" [id=1183, type=ne]; -"1184 masked_fill_6" [id=1184, type=masked_fill]; -"1185 eq_3" [id=1185, type=eq]; -"1186 masked_fill_7" [id=1186, type=masked_fill]; -"1187 view_41" [id=1187, type=view]; -"1188 unsqueeze_22" [id=1188, type=unsqueeze]; -"1189 unsqueeze_23" [id=1189, type=unsqueeze]; -"1190 add_25" [id=1190, type=add]; -"1191 view_42" [id=1191, type=view]; -"1192 softmax_7" [id=1192, type=softmax]; -"1193 dropout_28" [id=1193, type=dropout]; -"1194 matmul_15" [id=1194, type=matmul]; -"1195 transpose_15" [id=1195, type=transpose]; -"1196 reshape_34" [id=1196, type=reshape]; -"1197 reshape_34_0_0_nncf_smooth_quant_0" [id=1197, type=call_module]; -"1198 quantize_per_tensor_default_48" [id=1198, type=quantize_per_tensor]; -"1199 dequantize_per_tensor_default_48" [id=1199, type=dequantize_per_tensor]; -"1200 linear_47_scale_0" [id=1200, type=get_attr]; -"1201 linear_47_zero_point_0" [id=1201, type=get_attr]; -"1202 _frozen_param48" [id=1202, type=get_attr]; -"1203 dequantize_per_channel_default_48" [id=1203, type=dequantize_per_channel]; -"1204 linear_47" [id=1204, type=linear]; -"1205 dropout_29" [id=1205, type=dropout]; -"1206 view_43" [id=1206, type=view]; -"1207 permute_36" [id=1207, type=permute]; -"1208 reshape_35" [id=1208, type=reshape]; -"1209 roll_7" [id=1209, type=roll]; -"1210 slice_129" [id=1210, type=slice]; -"1211 slice_130" [id=1211, type=slice]; -"1212 slice_131" [id=1212, type=slice]; -"1213 slice_132" [id=1213, type=slice]; -"1214 contiguous_13" [id=1214, type=contiguous]; -"1215 layer_norm_17" [id=1215, type=layer_norm]; -"1216 add_26" [id=1216, type=add]; -"1217 add_26_0_0_nncf_smooth_quant_0" [id=1217, type=call_module]; -"1218 quantize_per_tensor_default_49" [id=1218, type=quantize_per_tensor]; -"1219 dequantize_per_tensor_default_49" [id=1219, type=dequantize_per_tensor]; -"1220 linear_48_scale_0" [id=1220, type=get_attr]; -"1221 linear_48_zero_point_0" [id=1221, type=get_attr]; -"1222 _frozen_param49" [id=1222, type=get_attr]; -"1223 dequantize_per_channel_default_49" [id=1223, type=dequantize_per_channel]; -"1224 linear_48" [id=1224, type=linear]; -"1225 gelu_7" [id=1225, type=gelu]; -"1226 dropout_30" [id=1226, type=dropout]; -"1227 dropout_30_0_0_nncf_smooth_quant_0" [id=1227, type=call_module]; -"1228 quantize_per_tensor_default_50" [id=1228, type=quantize_per_tensor]; -"1229 dequantize_per_tensor_default_50" [id=1229, type=dequantize_per_tensor]; -"1230 linear_49_scale_0" [id=1230, type=get_attr]; -"1231 linear_49_zero_point_0" [id=1231, type=get_attr]; -"1232 _frozen_param50" [id=1232, type=get_attr]; -"1233 dequantize_per_channel_default_50" [id=1233, type=dequantize_per_channel]; -"1234 linear_49" [id=1234, type=linear]; -"1235 dropout_31" [id=1235, type=dropout]; -"1236 layer_norm_18" [id=1236, type=layer_norm]; -"1237 add_27" [id=1237, type=add]; -"1238 features_5_4_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=1238, type=call_module]; -"1239 linear_50_scale_0" [id=1239, type=get_attr]; -"1240 linear_50_zero_point_0" [id=1240, type=get_attr]; -"1241 _frozen_param51" [id=1241, type=get_attr]; -"1242 dequantize_per_channel_default_51" [id=1242, type=dequantize_per_channel]; -"1243 linear_50" [id=1243, type=linear]; -"1244 relu__8" [id=1244, type=relu_]; -"1245 relu__8_0_0_nncf_smooth_quant_0" [id=1245, type=call_module]; -"1246 linear_51_scale_0" [id=1246, type=get_attr]; -"1247 linear_51_zero_point_0" [id=1247, type=get_attr]; -"1248 _frozen_param52" [id=1248, type=get_attr]; -"1249 dequantize_per_channel_default_52" [id=1249, type=dequantize_per_channel]; -"1250 linear_51" [id=1250, type=linear]; -"1251 view_44" [id=1251, type=view]; -"1252 index_8" [id=1252, type=index]; -"1253 view_45" [id=1253, type=view]; -"1254 permute_37" [id=1254, type=permute]; -"1255 contiguous_14" [id=1255, type=contiguous]; -"1256 unsqueeze_24" [id=1256, type=unsqueeze]; -"1257 sigmoid_8" [id=1257, type=sigmoid]; -"1258 mul_16" [id=1258, type=mul]; -"1259 pad_10" [id=1259, type=pad]; -"1260 view_46" [id=1260, type=view]; -"1261 permute_38" [id=1261, type=permute]; -"1262 reshape_36" [id=1262, type=reshape]; -"1263 clone_8" [id=1263, type=clone]; -"1264 reshape_36_0_0_nncf_smooth_quant_0" [id=1264, type=call_module]; -"1265 quantize_per_tensor_default_51" [id=1265, type=quantize_per_tensor]; -"1266 dequantize_per_tensor_default_51" [id=1266, type=dequantize_per_tensor]; -"1267 linear_52_scale_0" [id=1267, type=get_attr]; -"1268 linear_52_zero_point_0" [id=1268, type=get_attr]; -"1269 _frozen_param53" [id=1269, type=get_attr]; -"1270 dequantize_per_channel_default_53" [id=1270, type=dequantize_per_channel]; -"1271 linear_52" [id=1271, type=linear]; -"1272 reshape_37" [id=1272, type=reshape]; -"1273 permute_39" [id=1273, type=permute]; -"1274 select_24" [id=1274, type=select]; -"1275 select_25" [id=1275, type=select]; -"1276 select_26" [id=1276, type=select]; -"1277 linalg_vector_norm_16" [id=1277, type=linalg_vector_norm]; -"1278 clamp_min_16" [id=1278, type=clamp_min]; -"1279 expand_as_16" [id=1279, type=expand_as]; -"1280 div_16" [id=1280, type=div]; -"1281 quantize_per_tensor_default_52" [id=1281, type=quantize_per_tensor]; -"1282 dequantize_per_tensor_default_52" [id=1282, type=dequantize_per_tensor]; -"1283 linalg_vector_norm_17" [id=1283, type=linalg_vector_norm]; -"1284 clamp_min_17" [id=1284, type=clamp_min]; -"1285 expand_as_17" [id=1285, type=expand_as]; -"1286 div_17" [id=1286, type=div]; -"1287 quantize_per_tensor_default_53" [id=1287, type=quantize_per_tensor]; -"1288 dequantize_per_tensor_default_53" [id=1288, type=dequantize_per_tensor]; -"1289 transpose_16" [id=1289, type=transpose]; -"1290 matmul_16" [id=1290, type=matmul]; -"1291 clamp_8" [id=1291, type=clamp]; -"1292 exp_8" [id=1292, type=exp]; -"1293 mul_17" [id=1293, type=mul]; -"1294 add_28" [id=1294, type=add]; -"1295 softmax_8" [id=1295, type=softmax]; -"1296 dropout_32" [id=1296, type=dropout]; -"1297 matmul_17" [id=1297, type=matmul]; -"1298 transpose_17" [id=1298, type=transpose]; -"1299 reshape_38" [id=1299, type=reshape]; -"1300 reshape_38_0_0_nncf_smooth_quant_0" [id=1300, type=call_module]; -"1301 quantize_per_tensor_default_54" [id=1301, type=quantize_per_tensor]; -"1302 dequantize_per_tensor_default_54" [id=1302, type=dequantize_per_tensor]; -"1303 linear_53_scale_0" [id=1303, type=get_attr]; -"1304 linear_53_zero_point_0" [id=1304, type=get_attr]; -"1305 _frozen_param54" [id=1305, type=get_attr]; -"1306 dequantize_per_channel_default_54" [id=1306, type=dequantize_per_channel]; -"1307 linear_53" [id=1307, type=linear]; -"1308 dropout_33" [id=1308, type=dropout]; -"1309 view_47" [id=1309, type=view]; -"1310 permute_40" [id=1310, type=permute]; -"1311 reshape_39" [id=1311, type=reshape]; -"1312 slice_134" [id=1312, type=slice]; -"1313 slice_135" [id=1313, type=slice]; -"1314 slice_136" [id=1314, type=slice]; -"1315 slice_137" [id=1315, type=slice]; -"1316 contiguous_15" [id=1316, type=contiguous]; -"1317 layer_norm_19" [id=1317, type=layer_norm]; -"1318 add_29" [id=1318, type=add]; -"1319 add_29_0_0_nncf_smooth_quant_0" [id=1319, type=call_module]; -"1320 quantize_per_tensor_default_55" [id=1320, type=quantize_per_tensor]; -"1321 dequantize_per_tensor_default_55" [id=1321, type=dequantize_per_tensor]; -"1322 linear_54_scale_0" [id=1322, type=get_attr]; -"1323 linear_54_zero_point_0" [id=1323, type=get_attr]; -"1324 _frozen_param55" [id=1324, type=get_attr]; -"1325 dequantize_per_channel_default_55" [id=1325, type=dequantize_per_channel]; -"1326 linear_54" [id=1326, type=linear]; -"1327 gelu_8" [id=1327, type=gelu]; -"1328 dropout_34" [id=1328, type=dropout]; -"1329 dropout_34_0_0_nncf_smooth_quant_0" [id=1329, type=call_module]; -"1330 quantize_per_tensor_default_56" [id=1330, type=quantize_per_tensor]; -"1331 dequantize_per_tensor_default_56" [id=1331, type=dequantize_per_tensor]; -"1332 linear_55_scale_0" [id=1332, type=get_attr]; -"1333 linear_55_zero_point_0" [id=1333, type=get_attr]; -"1334 _frozen_param56" [id=1334, type=get_attr]; -"1335 dequantize_per_channel_default_56" [id=1335, type=dequantize_per_channel]; -"1336 linear_55" [id=1336, type=linear]; -"1337 dropout_35" [id=1337, type=dropout]; -"1338 layer_norm_20" [id=1338, type=layer_norm]; -"1339 add_30" [id=1339, type=add]; -"1340 features_5_5_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=1340, type=call_module]; -"1341 linear_56_scale_0" [id=1341, type=get_attr]; -"1342 linear_56_zero_point_0" [id=1342, type=get_attr]; -"1343 _frozen_param57" [id=1343, type=get_attr]; -"1344 dequantize_per_channel_default_57" [id=1344, type=dequantize_per_channel]; -"1345 linear_56" [id=1345, type=linear]; -"1346 relu__9" [id=1346, type=relu_]; -"1347 relu__9_0_0_nncf_smooth_quant_0" [id=1347, type=call_module]; -"1348 linear_57_scale_0" [id=1348, type=get_attr]; -"1349 linear_57_zero_point_0" [id=1349, type=get_attr]; -"1350 _frozen_param58" [id=1350, type=get_attr]; -"1351 dequantize_per_channel_default_58" [id=1351, type=dequantize_per_channel]; -"1352 linear_57" [id=1352, type=linear]; -"1353 view_48" [id=1353, type=view]; -"1354 index_9" [id=1354, type=index]; -"1355 view_49" [id=1355, type=view]; -"1356 permute_41" [id=1356, type=permute]; -"1357 contiguous_16" [id=1357, type=contiguous]; -"1358 unsqueeze_25" [id=1358, type=unsqueeze]; -"1359 sigmoid_9" [id=1359, type=sigmoid]; -"1360 mul_18" [id=1360, type=mul]; -"1361 pad_11" [id=1361, type=pad]; -"1362 roll_8" [id=1362, type=roll]; -"1363 view_50" [id=1363, type=view]; -"1364 permute_42" [id=1364, type=permute]; -"1365 reshape_40" [id=1365, type=reshape]; -"1366 clone_9" [id=1366, type=clone]; -"1367 reshape_40_0_0_nncf_smooth_quant_0" [id=1367, type=call_module]; -"1368 quantize_per_tensor_default_57" [id=1368, type=quantize_per_tensor]; -"1369 dequantize_per_tensor_default_57" [id=1369, type=dequantize_per_tensor]; -"1370 linear_58_scale_0" [id=1370, type=get_attr]; -"1371 linear_58_zero_point_0" [id=1371, type=get_attr]; -"1372 _frozen_param59" [id=1372, type=get_attr]; -"1373 dequantize_per_channel_default_59" [id=1373, type=dequantize_per_channel]; -"1374 linear_58" [id=1374, type=linear]; -"1375 reshape_41" [id=1375, type=reshape]; -"1376 permute_43" [id=1376, type=permute]; -"1377 select_27" [id=1377, type=select]; -"1378 select_28" [id=1378, type=select]; -"1379 select_29" [id=1379, type=select]; -"1380 linalg_vector_norm_18" [id=1380, type=linalg_vector_norm]; -"1381 clamp_min_18" [id=1381, type=clamp_min]; -"1382 expand_as_18" [id=1382, type=expand_as]; -"1383 div_18" [id=1383, type=div]; -"1384 quantize_per_tensor_default_58" [id=1384, type=quantize_per_tensor]; -"1385 dequantize_per_tensor_default_58" [id=1385, type=dequantize_per_tensor]; -"1386 linalg_vector_norm_19" [id=1386, type=linalg_vector_norm]; -"1387 clamp_min_19" [id=1387, type=clamp_min]; -"1388 expand_as_19" [id=1388, type=expand_as]; -"1389 div_19" [id=1389, type=div]; -"1390 quantize_per_tensor_default_59" [id=1390, type=quantize_per_tensor]; -"1391 dequantize_per_tensor_default_59" [id=1391, type=dequantize_per_tensor]; -"1392 transpose_18" [id=1392, type=transpose]; -"1393 matmul_18" [id=1393, type=matmul]; -"1394 clamp_9" [id=1394, type=clamp]; -"1395 exp_9" [id=1395, type=exp]; -"1396 mul_19" [id=1396, type=mul]; -"1397 add_31" [id=1397, type=add]; -"1398 new_zeros_4" [id=1398, type=new_zeros]; -"1399 view_51" [id=1399, type=view]; -"1400 permute_44" [id=1400, type=permute]; -"1401 reshape_42" [id=1401, type=reshape]; -"1402 unsqueeze_26" [id=1402, type=unsqueeze]; -"1403 unsqueeze_27" [id=1403, type=unsqueeze]; -"1404 sub_4" [id=1404, type=sub]; -"1405 ne_4" [id=1405, type=ne]; -"1406 masked_fill_8" [id=1406, type=masked_fill]; -"1407 eq_4" [id=1407, type=eq]; -"1408 masked_fill_9" [id=1408, type=masked_fill]; -"1409 view_52" [id=1409, type=view]; -"1410 unsqueeze_28" [id=1410, type=unsqueeze]; -"1411 unsqueeze_29" [id=1411, type=unsqueeze]; -"1412 add_32" [id=1412, type=add]; -"1413 view_53" [id=1413, type=view]; -"1414 softmax_9" [id=1414, type=softmax]; -"1415 dropout_36" [id=1415, type=dropout]; -"1416 matmul_19" [id=1416, type=matmul]; -"1417 transpose_19" [id=1417, type=transpose]; -"1418 reshape_43" [id=1418, type=reshape]; -"1419 reshape_43_0_0_nncf_smooth_quant_0" [id=1419, type=call_module]; -"1420 quantize_per_tensor_default_60" [id=1420, type=quantize_per_tensor]; -"1421 dequantize_per_tensor_default_60" [id=1421, type=dequantize_per_tensor]; -"1422 linear_59_scale_0" [id=1422, type=get_attr]; -"1423 linear_59_zero_point_0" [id=1423, type=get_attr]; -"1424 _frozen_param60" [id=1424, type=get_attr]; -"1425 dequantize_per_channel_default_60" [id=1425, type=dequantize_per_channel]; -"1426 linear_59" [id=1426, type=linear]; -"1427 dropout_37" [id=1427, type=dropout]; -"1428 view_54" [id=1428, type=view]; -"1429 permute_45" [id=1429, type=permute]; -"1430 reshape_44" [id=1430, type=reshape]; -"1431 roll_9" [id=1431, type=roll]; -"1432 slice_157" [id=1432, type=slice]; -"1433 slice_158" [id=1433, type=slice]; -"1434 slice_159" [id=1434, type=slice]; -"1435 slice_160" [id=1435, type=slice]; -"1436 contiguous_17" [id=1436, type=contiguous]; -"1437 layer_norm_21" [id=1437, type=layer_norm]; -"1438 add_33" [id=1438, type=add]; -"1439 add_33_0_0_nncf_smooth_quant_0" [id=1439, type=call_module]; -"1440 quantize_per_tensor_default_61" [id=1440, type=quantize_per_tensor]; -"1441 dequantize_per_tensor_default_61" [id=1441, type=dequantize_per_tensor]; -"1442 linear_60_scale_0" [id=1442, type=get_attr]; -"1443 linear_60_zero_point_0" [id=1443, type=get_attr]; -"1444 _frozen_param61" [id=1444, type=get_attr]; -"1445 dequantize_per_channel_default_61" [id=1445, type=dequantize_per_channel]; -"1446 linear_60" [id=1446, type=linear]; -"1447 gelu_9" [id=1447, type=gelu]; -"1448 dropout_38" [id=1448, type=dropout]; -"1449 dropout_38_0_0_nncf_smooth_quant_0" [id=1449, type=call_module]; -"1450 quantize_per_tensor_default_62" [id=1450, type=quantize_per_tensor]; -"1451 dequantize_per_tensor_default_62" [id=1451, type=dequantize_per_tensor]; -"1452 linear_61_scale_0" [id=1452, type=get_attr]; -"1453 linear_61_zero_point_0" [id=1453, type=get_attr]; -"1454 _frozen_param62" [id=1454, type=get_attr]; -"1455 dequantize_per_channel_default_62" [id=1455, type=dequantize_per_channel]; -"1456 linear_61" [id=1456, type=linear]; -"1457 dropout_39" [id=1457, type=dropout]; -"1458 layer_norm_22" [id=1458, type=layer_norm]; -"1459 add_34" [id=1459, type=add]; -"1460 features_5_6_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=1460, type=call_module]; -"1461 linear_62_scale_0" [id=1461, type=get_attr]; -"1462 linear_62_zero_point_0" [id=1462, type=get_attr]; -"1463 _frozen_param63" [id=1463, type=get_attr]; -"1464 dequantize_per_channel_default_63" [id=1464, type=dequantize_per_channel]; -"1465 linear_62" [id=1465, type=linear]; -"1466 relu__10" [id=1466, type=relu_]; -"1467 relu__10_0_0_nncf_smooth_quant_0" [id=1467, type=call_module]; -"1468 linear_63_scale_0" [id=1468, type=get_attr]; -"1469 linear_63_zero_point_0" [id=1469, type=get_attr]; -"1470 _frozen_param64" [id=1470, type=get_attr]; -"1471 dequantize_per_channel_default_64" [id=1471, type=dequantize_per_channel]; -"1472 linear_63" [id=1472, type=linear]; -"1473 view_55" [id=1473, type=view]; -"1474 index_10" [id=1474, type=index]; -"1475 view_56" [id=1475, type=view]; -"1476 permute_46" [id=1476, type=permute]; -"1477 contiguous_18" [id=1477, type=contiguous]; -"1478 unsqueeze_30" [id=1478, type=unsqueeze]; -"1479 sigmoid_10" [id=1479, type=sigmoid]; -"1480 mul_20" [id=1480, type=mul]; -"1481 pad_12" [id=1481, type=pad]; -"1482 view_57" [id=1482, type=view]; -"1483 permute_47" [id=1483, type=permute]; -"1484 reshape_45" [id=1484, type=reshape]; -"1485 clone_10" [id=1485, type=clone]; -"1486 reshape_45_0_0_nncf_smooth_quant_0" [id=1486, type=call_module]; -"1487 quantize_per_tensor_default_63" [id=1487, type=quantize_per_tensor]; -"1488 dequantize_per_tensor_default_63" [id=1488, type=dequantize_per_tensor]; -"1489 linear_64_scale_0" [id=1489, type=get_attr]; -"1490 linear_64_zero_point_0" [id=1490, type=get_attr]; -"1491 _frozen_param65" [id=1491, type=get_attr]; -"1492 dequantize_per_channel_default_65" [id=1492, type=dequantize_per_channel]; -"1493 linear_64" [id=1493, type=linear]; -"1494 reshape_46" [id=1494, type=reshape]; -"1495 permute_48" [id=1495, type=permute]; -"1496 select_30" [id=1496, type=select]; -"1497 select_31" [id=1497, type=select]; -"1498 select_32" [id=1498, type=select]; -"1499 linalg_vector_norm_20" [id=1499, type=linalg_vector_norm]; -"1500 clamp_min_20" [id=1500, type=clamp_min]; -"1501 expand_as_20" [id=1501, type=expand_as]; -"1502 div_20" [id=1502, type=div]; -"1503 quantize_per_tensor_default_64" [id=1503, type=quantize_per_tensor]; -"1504 dequantize_per_tensor_default_64" [id=1504, type=dequantize_per_tensor]; -"1505 linalg_vector_norm_21" [id=1505, type=linalg_vector_norm]; -"1506 clamp_min_21" [id=1506, type=clamp_min]; -"1507 expand_as_21" [id=1507, type=expand_as]; -"1508 div_21" [id=1508, type=div]; -"1509 quantize_per_tensor_default_65" [id=1509, type=quantize_per_tensor]; -"1510 dequantize_per_tensor_default_65" [id=1510, type=dequantize_per_tensor]; -"1511 transpose_20" [id=1511, type=transpose]; -"1512 matmul_20" [id=1512, type=matmul]; -"1513 clamp_10" [id=1513, type=clamp]; -"1514 exp_10" [id=1514, type=exp]; -"1515 mul_21" [id=1515, type=mul]; -"1516 add_35" [id=1516, type=add]; -"1517 softmax_10" [id=1517, type=softmax]; -"1518 dropout_40" [id=1518, type=dropout]; -"1519 matmul_21" [id=1519, type=matmul]; -"1520 transpose_21" [id=1520, type=transpose]; -"1521 reshape_47" [id=1521, type=reshape]; -"1522 reshape_47_0_0_nncf_smooth_quant_0" [id=1522, type=call_module]; -"1523 quantize_per_tensor_default_66" [id=1523, type=quantize_per_tensor]; -"1524 dequantize_per_tensor_default_66" [id=1524, type=dequantize_per_tensor]; -"1525 linear_65_scale_0" [id=1525, type=get_attr]; -"1526 linear_65_zero_point_0" [id=1526, type=get_attr]; -"1527 _frozen_param66" [id=1527, type=get_attr]; -"1528 dequantize_per_channel_default_66" [id=1528, type=dequantize_per_channel]; -"1529 linear_65" [id=1529, type=linear]; -"1530 dropout_41" [id=1530, type=dropout]; -"1531 view_58" [id=1531, type=view]; -"1532 permute_49" [id=1532, type=permute]; -"1533 reshape_48" [id=1533, type=reshape]; -"1534 slice_162" [id=1534, type=slice]; -"1535 slice_163" [id=1535, type=slice]; -"1536 slice_164" [id=1536, type=slice]; -"1537 slice_165" [id=1537, type=slice]; -"1538 contiguous_19" [id=1538, type=contiguous]; -"1539 layer_norm_23" [id=1539, type=layer_norm]; -"1540 add_36" [id=1540, type=add]; -"1541 add_36_0_0_nncf_smooth_quant_0" [id=1541, type=call_module]; -"1542 quantize_per_tensor_default_67" [id=1542, type=quantize_per_tensor]; -"1543 dequantize_per_tensor_default_67" [id=1543, type=dequantize_per_tensor]; -"1544 linear_66_scale_0" [id=1544, type=get_attr]; -"1545 linear_66_zero_point_0" [id=1545, type=get_attr]; -"1546 _frozen_param67" [id=1546, type=get_attr]; -"1547 dequantize_per_channel_default_67" [id=1547, type=dequantize_per_channel]; -"1548 linear_66" [id=1548, type=linear]; -"1549 gelu_10" [id=1549, type=gelu]; -"1550 dropout_42" [id=1550, type=dropout]; -"1551 dropout_42_0_0_nncf_smooth_quant_0" [id=1551, type=call_module]; -"1552 quantize_per_tensor_default_68" [id=1552, type=quantize_per_tensor]; -"1553 dequantize_per_tensor_default_68" [id=1553, type=dequantize_per_tensor]; -"1554 linear_67_scale_0" [id=1554, type=get_attr]; -"1555 linear_67_zero_point_0" [id=1555, type=get_attr]; -"1556 _frozen_param68" [id=1556, type=get_attr]; -"1557 dequantize_per_channel_default_68" [id=1557, type=dequantize_per_channel]; -"1558 linear_67" [id=1558, type=linear]; -"1559 dropout_43" [id=1559, type=dropout]; -"1560 layer_norm_24" [id=1560, type=layer_norm]; -"1561 add_37" [id=1561, type=add]; -"1562 features_5_7_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=1562, type=call_module]; -"1563 linear_68_scale_0" [id=1563, type=get_attr]; -"1564 linear_68_zero_point_0" [id=1564, type=get_attr]; -"1565 _frozen_param69" [id=1565, type=get_attr]; -"1566 dequantize_per_channel_default_69" [id=1566, type=dequantize_per_channel]; -"1567 linear_68" [id=1567, type=linear]; -"1568 relu__11" [id=1568, type=relu_]; -"1569 relu__11_0_0_nncf_smooth_quant_0" [id=1569, type=call_module]; -"1570 linear_69_scale_0" [id=1570, type=get_attr]; -"1571 linear_69_zero_point_0" [id=1571, type=get_attr]; -"1572 _frozen_param70" [id=1572, type=get_attr]; -"1573 dequantize_per_channel_default_70" [id=1573, type=dequantize_per_channel]; -"1574 linear_69" [id=1574, type=linear]; -"1575 view_59" [id=1575, type=view]; -"1576 index_11" [id=1576, type=index]; -"1577 view_60" [id=1577, type=view]; -"1578 permute_50" [id=1578, type=permute]; -"1579 contiguous_20" [id=1579, type=contiguous]; -"1580 unsqueeze_31" [id=1580, type=unsqueeze]; -"1581 sigmoid_11" [id=1581, type=sigmoid]; -"1582 mul_22" [id=1582, type=mul]; -"1583 pad_13" [id=1583, type=pad]; -"1584 roll_10" [id=1584, type=roll]; -"1585 view_61" [id=1585, type=view]; -"1586 permute_51" [id=1586, type=permute]; -"1587 reshape_49" [id=1587, type=reshape]; -"1588 clone_11" [id=1588, type=clone]; -"1589 reshape_49_0_0_nncf_smooth_quant_0" [id=1589, type=call_module]; -"1590 quantize_per_tensor_default_69" [id=1590, type=quantize_per_tensor]; -"1591 dequantize_per_tensor_default_69" [id=1591, type=dequantize_per_tensor]; -"1592 linear_70_scale_0" [id=1592, type=get_attr]; -"1593 linear_70_zero_point_0" [id=1593, type=get_attr]; -"1594 _frozen_param71" [id=1594, type=get_attr]; -"1595 dequantize_per_channel_default_71" [id=1595, type=dequantize_per_channel]; -"1596 linear_70" [id=1596, type=linear]; -"1597 reshape_50" [id=1597, type=reshape]; -"1598 permute_52" [id=1598, type=permute]; -"1599 select_33" [id=1599, type=select]; -"1600 select_34" [id=1600, type=select]; -"1601 select_35" [id=1601, type=select]; -"1602 linalg_vector_norm_22" [id=1602, type=linalg_vector_norm]; -"1603 clamp_min_22" [id=1603, type=clamp_min]; -"1604 expand_as_22" [id=1604, type=expand_as]; -"1605 div_22" [id=1605, type=div]; -"1606 quantize_per_tensor_default_70" [id=1606, type=quantize_per_tensor]; -"1607 dequantize_per_tensor_default_70" [id=1607, type=dequantize_per_tensor]; -"1608 linalg_vector_norm_23" [id=1608, type=linalg_vector_norm]; -"1609 clamp_min_23" [id=1609, type=clamp_min]; -"1610 expand_as_23" [id=1610, type=expand_as]; -"1611 div_23" [id=1611, type=div]; -"1612 quantize_per_tensor_default_71" [id=1612, type=quantize_per_tensor]; -"1613 dequantize_per_tensor_default_71" [id=1613, type=dequantize_per_tensor]; -"1614 transpose_22" [id=1614, type=transpose]; -"1615 matmul_22" [id=1615, type=matmul]; -"1616 clamp_11" [id=1616, type=clamp]; -"1617 exp_11" [id=1617, type=exp]; -"1618 mul_23" [id=1618, type=mul]; -"1619 add_38" [id=1619, type=add]; -"1620 new_zeros_5" [id=1620, type=new_zeros]; -"1621 view_62" [id=1621, type=view]; -"1622 permute_53" [id=1622, type=permute]; -"1623 reshape_51" [id=1623, type=reshape]; -"1624 unsqueeze_32" [id=1624, type=unsqueeze]; -"1625 unsqueeze_33" [id=1625, type=unsqueeze]; -"1626 sub_5" [id=1626, type=sub]; -"1627 ne_5" [id=1627, type=ne]; -"1628 masked_fill_10" [id=1628, type=masked_fill]; -"1629 eq_5" [id=1629, type=eq]; -"1630 masked_fill_11" [id=1630, type=masked_fill]; -"1631 view_63" [id=1631, type=view]; -"1632 unsqueeze_34" [id=1632, type=unsqueeze]; -"1633 unsqueeze_35" [id=1633, type=unsqueeze]; -"1634 add_39" [id=1634, type=add]; -"1635 view_64" [id=1635, type=view]; -"1636 softmax_11" [id=1636, type=softmax]; -"1637 dropout_44" [id=1637, type=dropout]; -"1638 matmul_23" [id=1638, type=matmul]; -"1639 transpose_23" [id=1639, type=transpose]; -"1640 reshape_52" [id=1640, type=reshape]; -"1641 reshape_52_0_0_nncf_smooth_quant_0" [id=1641, type=call_module]; -"1642 quantize_per_tensor_default_72" [id=1642, type=quantize_per_tensor]; -"1643 dequantize_per_tensor_default_72" [id=1643, type=dequantize_per_tensor]; -"1644 linear_71_scale_0" [id=1644, type=get_attr]; -"1645 linear_71_zero_point_0" [id=1645, type=get_attr]; -"1646 _frozen_param72" [id=1646, type=get_attr]; -"1647 dequantize_per_channel_default_72" [id=1647, type=dequantize_per_channel]; -"1648 linear_71" [id=1648, type=linear]; -"1649 dropout_45" [id=1649, type=dropout]; -"1650 view_65" [id=1650, type=view]; -"1651 permute_54" [id=1651, type=permute]; -"1652 reshape_53" [id=1652, type=reshape]; -"1653 roll_11" [id=1653, type=roll]; -"1654 slice_185" [id=1654, type=slice]; -"1655 slice_186" [id=1655, type=slice]; -"1656 slice_187" [id=1656, type=slice]; -"1657 slice_188" [id=1657, type=slice]; -"1658 contiguous_21" [id=1658, type=contiguous]; -"1659 layer_norm_25" [id=1659, type=layer_norm]; -"1660 add_40" [id=1660, type=add]; -"1661 add_40_0_0_nncf_smooth_quant_0" [id=1661, type=call_module]; -"1662 quantize_per_tensor_default_73" [id=1662, type=quantize_per_tensor]; -"1663 dequantize_per_tensor_default_73" [id=1663, type=dequantize_per_tensor]; -"1664 linear_72_scale_0" [id=1664, type=get_attr]; -"1665 linear_72_zero_point_0" [id=1665, type=get_attr]; -"1666 _frozen_param73" [id=1666, type=get_attr]; -"1667 dequantize_per_channel_default_73" [id=1667, type=dequantize_per_channel]; -"1668 linear_72" [id=1668, type=linear]; -"1669 gelu_11" [id=1669, type=gelu]; -"1670 dropout_46" [id=1670, type=dropout]; -"1671 dropout_46_0_0_nncf_smooth_quant_0" [id=1671, type=call_module]; -"1672 quantize_per_tensor_default_74" [id=1672, type=quantize_per_tensor]; -"1673 dequantize_per_tensor_default_74" [id=1673, type=dequantize_per_tensor]; -"1674 linear_73_scale_0" [id=1674, type=get_attr]; -"1675 linear_73_zero_point_0" [id=1675, type=get_attr]; -"1676 _frozen_param74" [id=1676, type=get_attr]; -"1677 dequantize_per_channel_default_74" [id=1677, type=dequantize_per_channel]; -"1678 linear_73" [id=1678, type=linear]; -"1679 dropout_47" [id=1679, type=dropout]; -"1680 layer_norm_26" [id=1680, type=layer_norm]; -"1681 add_41" [id=1681, type=add]; -"1682 features_5_8_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=1682, type=call_module]; -"1683 linear_74_scale_0" [id=1683, type=get_attr]; -"1684 linear_74_zero_point_0" [id=1684, type=get_attr]; -"1685 _frozen_param75" [id=1685, type=get_attr]; -"1686 dequantize_per_channel_default_75" [id=1686, type=dequantize_per_channel]; -"1687 linear_74" [id=1687, type=linear]; -"1688 relu__12" [id=1688, type=relu_]; -"1689 relu__12_0_0_nncf_smooth_quant_0" [id=1689, type=call_module]; -"1690 linear_75_scale_0" [id=1690, type=get_attr]; -"1691 linear_75_zero_point_0" [id=1691, type=get_attr]; -"1692 _frozen_param76" [id=1692, type=get_attr]; -"1693 dequantize_per_channel_default_76" [id=1693, type=dequantize_per_channel]; -"1694 linear_75" [id=1694, type=linear]; -"1695 view_66" [id=1695, type=view]; -"1696 index_12" [id=1696, type=index]; -"1697 view_67" [id=1697, type=view]; -"1698 permute_55" [id=1698, type=permute]; -"1699 contiguous_22" [id=1699, type=contiguous]; -"1700 unsqueeze_36" [id=1700, type=unsqueeze]; -"1701 sigmoid_12" [id=1701, type=sigmoid]; -"1702 mul_24" [id=1702, type=mul]; -"1703 pad_14" [id=1703, type=pad]; -"1704 view_68" [id=1704, type=view]; -"1705 permute_56" [id=1705, type=permute]; -"1706 reshape_54" [id=1706, type=reshape]; -"1707 clone_12" [id=1707, type=clone]; -"1708 reshape_54_0_0_nncf_smooth_quant_0" [id=1708, type=call_module]; -"1709 quantize_per_tensor_default_75" [id=1709, type=quantize_per_tensor]; -"1710 dequantize_per_tensor_default_75" [id=1710, type=dequantize_per_tensor]; -"1711 linear_76_scale_0" [id=1711, type=get_attr]; -"1712 linear_76_zero_point_0" [id=1712, type=get_attr]; -"1713 _frozen_param77" [id=1713, type=get_attr]; -"1714 dequantize_per_channel_default_77" [id=1714, type=dequantize_per_channel]; -"1715 linear_76" [id=1715, type=linear]; -"1716 reshape_55" [id=1716, type=reshape]; -"1717 permute_57" [id=1717, type=permute]; -"1718 select_36" [id=1718, type=select]; -"1719 select_37" [id=1719, type=select]; -"1720 select_38" [id=1720, type=select]; -"1721 linalg_vector_norm_24" [id=1721, type=linalg_vector_norm]; -"1722 clamp_min_24" [id=1722, type=clamp_min]; -"1723 expand_as_24" [id=1723, type=expand_as]; -"1724 div_24" [id=1724, type=div]; -"1725 quantize_per_tensor_default_76" [id=1725, type=quantize_per_tensor]; -"1726 dequantize_per_tensor_default_76" [id=1726, type=dequantize_per_tensor]; -"1727 linalg_vector_norm_25" [id=1727, type=linalg_vector_norm]; -"1728 clamp_min_25" [id=1728, type=clamp_min]; -"1729 expand_as_25" [id=1729, type=expand_as]; -"1730 div_25" [id=1730, type=div]; -"1731 quantize_per_tensor_default_77" [id=1731, type=quantize_per_tensor]; -"1732 dequantize_per_tensor_default_77" [id=1732, type=dequantize_per_tensor]; -"1733 transpose_24" [id=1733, type=transpose]; -"1734 matmul_24" [id=1734, type=matmul]; -"1735 clamp_12" [id=1735, type=clamp]; -"1736 exp_12" [id=1736, type=exp]; -"1737 mul_25" [id=1737, type=mul]; -"1738 add_42" [id=1738, type=add]; -"1739 softmax_12" [id=1739, type=softmax]; -"1740 dropout_48" [id=1740, type=dropout]; -"1741 matmul_25" [id=1741, type=matmul]; -"1742 transpose_25" [id=1742, type=transpose]; -"1743 reshape_56" [id=1743, type=reshape]; -"1744 reshape_56_0_0_nncf_smooth_quant_0" [id=1744, type=call_module]; -"1745 quantize_per_tensor_default_78" [id=1745, type=quantize_per_tensor]; -"1746 dequantize_per_tensor_default_78" [id=1746, type=dequantize_per_tensor]; -"1747 linear_77_scale_0" [id=1747, type=get_attr]; -"1748 linear_77_zero_point_0" [id=1748, type=get_attr]; -"1749 _frozen_param78" [id=1749, type=get_attr]; -"1750 dequantize_per_channel_default_78" [id=1750, type=dequantize_per_channel]; -"1751 linear_77" [id=1751, type=linear]; -"1752 dropout_49" [id=1752, type=dropout]; -"1753 view_69" [id=1753, type=view]; -"1754 permute_58" [id=1754, type=permute]; -"1755 reshape_57" [id=1755, type=reshape]; -"1756 slice_190" [id=1756, type=slice]; -"1757 slice_191" [id=1757, type=slice]; -"1758 slice_192" [id=1758, type=slice]; -"1759 slice_193" [id=1759, type=slice]; -"1760 contiguous_23" [id=1760, type=contiguous]; -"1761 layer_norm_27" [id=1761, type=layer_norm]; -"1762 add_43" [id=1762, type=add]; -"1763 add_43_0_0_nncf_smooth_quant_0" [id=1763, type=call_module]; -"1764 quantize_per_tensor_default_79" [id=1764, type=quantize_per_tensor]; -"1765 dequantize_per_tensor_default_79" [id=1765, type=dequantize_per_tensor]; -"1766 linear_78_scale_0" [id=1766, type=get_attr]; -"1767 linear_78_zero_point_0" [id=1767, type=get_attr]; -"1768 _frozen_param79" [id=1768, type=get_attr]; -"1769 dequantize_per_channel_default_79" [id=1769, type=dequantize_per_channel]; -"1770 linear_78" [id=1770, type=linear]; -"1771 gelu_12" [id=1771, type=gelu]; -"1772 dropout_50" [id=1772, type=dropout]; -"1773 dropout_50_0_0_nncf_smooth_quant_0" [id=1773, type=call_module]; -"1774 quantize_per_tensor_default_80" [id=1774, type=quantize_per_tensor]; -"1775 dequantize_per_tensor_default_80" [id=1775, type=dequantize_per_tensor]; -"1776 linear_79_scale_0" [id=1776, type=get_attr]; -"1777 linear_79_zero_point_0" [id=1777, type=get_attr]; -"1778 _frozen_param80" [id=1778, type=get_attr]; -"1779 dequantize_per_channel_default_80" [id=1779, type=dequantize_per_channel]; -"1780 linear_79" [id=1780, type=linear]; -"1781 dropout_51" [id=1781, type=dropout]; -"1782 layer_norm_28" [id=1782, type=layer_norm]; -"1783 add_44" [id=1783, type=add]; -"1784 features_5_9_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=1784, type=call_module]; -"1785 linear_80_scale_0" [id=1785, type=get_attr]; -"1786 linear_80_zero_point_0" [id=1786, type=get_attr]; -"1787 _frozen_param81" [id=1787, type=get_attr]; -"1788 dequantize_per_channel_default_81" [id=1788, type=dequantize_per_channel]; -"1789 linear_80" [id=1789, type=linear]; -"1790 relu__13" [id=1790, type=relu_]; -"1791 relu__13_0_0_nncf_smooth_quant_0" [id=1791, type=call_module]; -"1792 linear_81_scale_0" [id=1792, type=get_attr]; -"1793 linear_81_zero_point_0" [id=1793, type=get_attr]; -"1794 _frozen_param82" [id=1794, type=get_attr]; -"1795 dequantize_per_channel_default_82" [id=1795, type=dequantize_per_channel]; -"1796 linear_81" [id=1796, type=linear]; -"1797 view_70" [id=1797, type=view]; -"1798 index_13" [id=1798, type=index]; -"1799 view_71" [id=1799, type=view]; -"1800 permute_59" [id=1800, type=permute]; -"1801 contiguous_24" [id=1801, type=contiguous]; -"1802 unsqueeze_37" [id=1802, type=unsqueeze]; -"1803 sigmoid_13" [id=1803, type=sigmoid]; -"1804 mul_26" [id=1804, type=mul]; -"1805 pad_15" [id=1805, type=pad]; -"1806 roll_12" [id=1806, type=roll]; -"1807 view_72" [id=1807, type=view]; -"1808 permute_60" [id=1808, type=permute]; -"1809 reshape_58" [id=1809, type=reshape]; -"1810 clone_13" [id=1810, type=clone]; -"1811 reshape_58_0_0_nncf_smooth_quant_0" [id=1811, type=call_module]; -"1812 quantize_per_tensor_default_81" [id=1812, type=quantize_per_tensor]; -"1813 dequantize_per_tensor_default_81" [id=1813, type=dequantize_per_tensor]; -"1814 linear_82_scale_0" [id=1814, type=get_attr]; -"1815 linear_82_zero_point_0" [id=1815, type=get_attr]; -"1816 _frozen_param83" [id=1816, type=get_attr]; -"1817 dequantize_per_channel_default_83" [id=1817, type=dequantize_per_channel]; -"1818 linear_82" [id=1818, type=linear]; -"1819 reshape_59" [id=1819, type=reshape]; -"1820 permute_61" [id=1820, type=permute]; -"1821 select_39" [id=1821, type=select]; -"1822 select_40" [id=1822, type=select]; -"1823 select_41" [id=1823, type=select]; -"1824 linalg_vector_norm_26" [id=1824, type=linalg_vector_norm]; -"1825 clamp_min_26" [id=1825, type=clamp_min]; -"1826 expand_as_26" [id=1826, type=expand_as]; -"1827 div_26" [id=1827, type=div]; -"1828 quantize_per_tensor_default_82" [id=1828, type=quantize_per_tensor]; -"1829 dequantize_per_tensor_default_82" [id=1829, type=dequantize_per_tensor]; -"1830 linalg_vector_norm_27" [id=1830, type=linalg_vector_norm]; -"1831 clamp_min_27" [id=1831, type=clamp_min]; -"1832 expand_as_27" [id=1832, type=expand_as]; -"1833 div_27" [id=1833, type=div]; -"1834 quantize_per_tensor_default_83" [id=1834, type=quantize_per_tensor]; -"1835 dequantize_per_tensor_default_83" [id=1835, type=dequantize_per_tensor]; -"1836 transpose_26" [id=1836, type=transpose]; -"1837 matmul_26" [id=1837, type=matmul]; -"1838 clamp_13" [id=1838, type=clamp]; -"1839 exp_13" [id=1839, type=exp]; -"1840 mul_27" [id=1840, type=mul]; -"1841 add_45" [id=1841, type=add]; -"1842 new_zeros_6" [id=1842, type=new_zeros]; -"1843 view_73" [id=1843, type=view]; -"1844 permute_62" [id=1844, type=permute]; -"1845 reshape_60" [id=1845, type=reshape]; -"1846 unsqueeze_38" [id=1846, type=unsqueeze]; -"1847 unsqueeze_39" [id=1847, type=unsqueeze]; -"1848 sub_6" [id=1848, type=sub]; -"1849 ne_6" [id=1849, type=ne]; -"1850 masked_fill_12" [id=1850, type=masked_fill]; -"1851 eq_6" [id=1851, type=eq]; -"1852 masked_fill_13" [id=1852, type=masked_fill]; -"1853 view_74" [id=1853, type=view]; -"1854 unsqueeze_40" [id=1854, type=unsqueeze]; -"1855 unsqueeze_41" [id=1855, type=unsqueeze]; -"1856 add_46" [id=1856, type=add]; -"1857 view_75" [id=1857, type=view]; -"1858 softmax_13" [id=1858, type=softmax]; -"1859 dropout_52" [id=1859, type=dropout]; -"1860 matmul_27" [id=1860, type=matmul]; -"1861 transpose_27" [id=1861, type=transpose]; -"1862 reshape_61" [id=1862, type=reshape]; -"1863 reshape_61_0_0_nncf_smooth_quant_0" [id=1863, type=call_module]; -"1864 quantize_per_tensor_default_84" [id=1864, type=quantize_per_tensor]; -"1865 dequantize_per_tensor_default_84" [id=1865, type=dequantize_per_tensor]; -"1866 linear_83_scale_0" [id=1866, type=get_attr]; -"1867 linear_83_zero_point_0" [id=1867, type=get_attr]; -"1868 _frozen_param84" [id=1868, type=get_attr]; -"1869 dequantize_per_channel_default_84" [id=1869, type=dequantize_per_channel]; -"1870 linear_83" [id=1870, type=linear]; -"1871 dropout_53" [id=1871, type=dropout]; -"1872 view_76" [id=1872, type=view]; -"1873 permute_63" [id=1873, type=permute]; -"1874 reshape_62" [id=1874, type=reshape]; -"1875 roll_13" [id=1875, type=roll]; -"1876 slice_213" [id=1876, type=slice]; -"1877 slice_214" [id=1877, type=slice]; -"1878 slice_215" [id=1878, type=slice]; -"1879 slice_216" [id=1879, type=slice]; -"1880 contiguous_25" [id=1880, type=contiguous]; -"1881 layer_norm_29" [id=1881, type=layer_norm]; -"1882 add_47" [id=1882, type=add]; -"1883 add_47_0_0_nncf_smooth_quant_0" [id=1883, type=call_module]; -"1884 quantize_per_tensor_default_85" [id=1884, type=quantize_per_tensor]; -"1885 dequantize_per_tensor_default_85" [id=1885, type=dequantize_per_tensor]; -"1886 linear_84_scale_0" [id=1886, type=get_attr]; -"1887 linear_84_zero_point_0" [id=1887, type=get_attr]; -"1888 _frozen_param85" [id=1888, type=get_attr]; -"1889 dequantize_per_channel_default_85" [id=1889, type=dequantize_per_channel]; -"1890 linear_84" [id=1890, type=linear]; -"1891 gelu_13" [id=1891, type=gelu]; -"1892 dropout_54" [id=1892, type=dropout]; -"1893 dropout_54_0_0_nncf_smooth_quant_0" [id=1893, type=call_module]; -"1894 quantize_per_tensor_default_86" [id=1894, type=quantize_per_tensor]; -"1895 dequantize_per_tensor_default_86" [id=1895, type=dequantize_per_tensor]; -"1896 linear_85_scale_0" [id=1896, type=get_attr]; -"1897 linear_85_zero_point_0" [id=1897, type=get_attr]; -"1898 _frozen_param86" [id=1898, type=get_attr]; -"1899 dequantize_per_channel_default_86" [id=1899, type=dequantize_per_channel]; -"1900 linear_85" [id=1900, type=linear]; -"1901 dropout_55" [id=1901, type=dropout]; -"1902 layer_norm_30" [id=1902, type=layer_norm]; -"1903 add_48" [id=1903, type=add]; -"1904 features_5_10_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=1904, type=call_module]; -"1905 linear_86_scale_0" [id=1905, type=get_attr]; -"1906 linear_86_zero_point_0" [id=1906, type=get_attr]; -"1907 _frozen_param87" [id=1907, type=get_attr]; -"1908 dequantize_per_channel_default_87" [id=1908, type=dequantize_per_channel]; -"1909 linear_86" [id=1909, type=linear]; -"1910 relu__14" [id=1910, type=relu_]; -"1911 relu__14_0_0_nncf_smooth_quant_0" [id=1911, type=call_module]; -"1912 linear_87_scale_0" [id=1912, type=get_attr]; -"1913 linear_87_zero_point_0" [id=1913, type=get_attr]; -"1914 _frozen_param88" [id=1914, type=get_attr]; -"1915 dequantize_per_channel_default_88" [id=1915, type=dequantize_per_channel]; -"1916 linear_87" [id=1916, type=linear]; -"1917 view_77" [id=1917, type=view]; -"1918 index_14" [id=1918, type=index]; -"1919 view_78" [id=1919, type=view]; -"1920 permute_64" [id=1920, type=permute]; -"1921 contiguous_26" [id=1921, type=contiguous]; -"1922 unsqueeze_42" [id=1922, type=unsqueeze]; -"1923 sigmoid_14" [id=1923, type=sigmoid]; -"1924 mul_28" [id=1924, type=mul]; -"1925 pad_16" [id=1925, type=pad]; -"1926 view_79" [id=1926, type=view]; -"1927 permute_65" [id=1927, type=permute]; -"1928 reshape_63" [id=1928, type=reshape]; -"1929 clone_14" [id=1929, type=clone]; -"1930 reshape_63_0_0_nncf_smooth_quant_0" [id=1930, type=call_module]; -"1931 quantize_per_tensor_default_87" [id=1931, type=quantize_per_tensor]; -"1932 dequantize_per_tensor_default_87" [id=1932, type=dequantize_per_tensor]; -"1933 linear_88_scale_0" [id=1933, type=get_attr]; -"1934 linear_88_zero_point_0" [id=1934, type=get_attr]; -"1935 _frozen_param89" [id=1935, type=get_attr]; -"1936 dequantize_per_channel_default_89" [id=1936, type=dequantize_per_channel]; -"1937 linear_88" [id=1937, type=linear]; -"1938 reshape_64" [id=1938, type=reshape]; -"1939 permute_66" [id=1939, type=permute]; -"1940 select_42" [id=1940, type=select]; -"1941 select_43" [id=1941, type=select]; -"1942 select_44" [id=1942, type=select]; -"1943 linalg_vector_norm_28" [id=1943, type=linalg_vector_norm]; -"1944 clamp_min_28" [id=1944, type=clamp_min]; -"1945 expand_as_28" [id=1945, type=expand_as]; -"1946 div_28" [id=1946, type=div]; -"1947 quantize_per_tensor_default_88" [id=1947, type=quantize_per_tensor]; -"1948 dequantize_per_tensor_default_88" [id=1948, type=dequantize_per_tensor]; -"1949 linalg_vector_norm_29" [id=1949, type=linalg_vector_norm]; -"1950 clamp_min_29" [id=1950, type=clamp_min]; -"1951 expand_as_29" [id=1951, type=expand_as]; -"1952 div_29" [id=1952, type=div]; -"1953 quantize_per_tensor_default_89" [id=1953, type=quantize_per_tensor]; -"1954 dequantize_per_tensor_default_89" [id=1954, type=dequantize_per_tensor]; -"1955 transpose_28" [id=1955, type=transpose]; -"1956 matmul_28" [id=1956, type=matmul]; -"1957 clamp_14" [id=1957, type=clamp]; -"1958 exp_14" [id=1958, type=exp]; -"1959 mul_29" [id=1959, type=mul]; -"1960 add_49" [id=1960, type=add]; -"1961 softmax_14" [id=1961, type=softmax]; -"1962 dropout_56" [id=1962, type=dropout]; -"1963 matmul_29" [id=1963, type=matmul]; -"1964 transpose_29" [id=1964, type=transpose]; -"1965 reshape_65" [id=1965, type=reshape]; -"1966 reshape_65_0_0_nncf_smooth_quant_0" [id=1966, type=call_module]; -"1967 quantize_per_tensor_default_90" [id=1967, type=quantize_per_tensor]; -"1968 dequantize_per_tensor_default_90" [id=1968, type=dequantize_per_tensor]; -"1969 linear_89_scale_0" [id=1969, type=get_attr]; -"1970 linear_89_zero_point_0" [id=1970, type=get_attr]; -"1971 _frozen_param90" [id=1971, type=get_attr]; -"1972 dequantize_per_channel_default_90" [id=1972, type=dequantize_per_channel]; -"1973 linear_89" [id=1973, type=linear]; -"1974 dropout_57" [id=1974, type=dropout]; -"1975 view_80" [id=1975, type=view]; -"1976 permute_67" [id=1976, type=permute]; -"1977 reshape_66" [id=1977, type=reshape]; -"1978 slice_218" [id=1978, type=slice]; -"1979 slice_219" [id=1979, type=slice]; -"1980 slice_220" [id=1980, type=slice]; -"1981 slice_221" [id=1981, type=slice]; -"1982 contiguous_27" [id=1982, type=contiguous]; -"1983 layer_norm_31" [id=1983, type=layer_norm]; -"1984 add_50" [id=1984, type=add]; -"1985 add_50_0_0_nncf_smooth_quant_0" [id=1985, type=call_module]; -"1986 quantize_per_tensor_default_91" [id=1986, type=quantize_per_tensor]; -"1987 dequantize_per_tensor_default_91" [id=1987, type=dequantize_per_tensor]; -"1988 linear_90_scale_0" [id=1988, type=get_attr]; -"1989 linear_90_zero_point_0" [id=1989, type=get_attr]; -"1990 _frozen_param91" [id=1990, type=get_attr]; -"1991 dequantize_per_channel_default_91" [id=1991, type=dequantize_per_channel]; -"1992 linear_90" [id=1992, type=linear]; -"1993 gelu_14" [id=1993, type=gelu]; -"1994 dropout_58" [id=1994, type=dropout]; -"1995 dropout_58_0_0_nncf_smooth_quant_0" [id=1995, type=call_module]; -"1996 quantize_per_tensor_default_92" [id=1996, type=quantize_per_tensor]; -"1997 dequantize_per_tensor_default_92" [id=1997, type=dequantize_per_tensor]; -"1998 linear_91_scale_0" [id=1998, type=get_attr]; -"1999 linear_91_zero_point_0" [id=1999, type=get_attr]; -"2000 _frozen_param92" [id=2000, type=get_attr]; -"2001 dequantize_per_channel_default_92" [id=2001, type=dequantize_per_channel]; -"2002 linear_91" [id=2002, type=linear]; -"2003 dropout_59" [id=2003, type=dropout]; -"2004 layer_norm_32" [id=2004, type=layer_norm]; -"2005 add_51" [id=2005, type=add]; -"2006 features_5_11_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=2006, type=call_module]; -"2007 linear_92_scale_0" [id=2007, type=get_attr]; -"2008 linear_92_zero_point_0" [id=2008, type=get_attr]; -"2009 _frozen_param93" [id=2009, type=get_attr]; -"2010 dequantize_per_channel_default_93" [id=2010, type=dequantize_per_channel]; -"2011 linear_92" [id=2011, type=linear]; -"2012 relu__15" [id=2012, type=relu_]; -"2013 relu__15_0_0_nncf_smooth_quant_0" [id=2013, type=call_module]; -"2014 linear_93_scale_0" [id=2014, type=get_attr]; -"2015 linear_93_zero_point_0" [id=2015, type=get_attr]; -"2016 _frozen_param94" [id=2016, type=get_attr]; -"2017 dequantize_per_channel_default_94" [id=2017, type=dequantize_per_channel]; -"2018 linear_93" [id=2018, type=linear]; -"2019 view_81" [id=2019, type=view]; -"2020 index_15" [id=2020, type=index]; -"2021 view_82" [id=2021, type=view]; -"2022 permute_68" [id=2022, type=permute]; -"2023 contiguous_28" [id=2023, type=contiguous]; -"2024 unsqueeze_43" [id=2024, type=unsqueeze]; -"2025 sigmoid_15" [id=2025, type=sigmoid]; -"2026 mul_30" [id=2026, type=mul]; -"2027 pad_17" [id=2027, type=pad]; -"2028 roll_14" [id=2028, type=roll]; -"2029 view_83" [id=2029, type=view]; -"2030 permute_69" [id=2030, type=permute]; -"2031 reshape_67" [id=2031, type=reshape]; -"2032 clone_15" [id=2032, type=clone]; -"2033 reshape_67_0_0_nncf_smooth_quant_0" [id=2033, type=call_module]; -"2034 quantize_per_tensor_default_93" [id=2034, type=quantize_per_tensor]; -"2035 dequantize_per_tensor_default_93" [id=2035, type=dequantize_per_tensor]; -"2036 linear_94_scale_0" [id=2036, type=get_attr]; -"2037 linear_94_zero_point_0" [id=2037, type=get_attr]; -"2038 _frozen_param95" [id=2038, type=get_attr]; -"2039 dequantize_per_channel_default_95" [id=2039, type=dequantize_per_channel]; -"2040 linear_94" [id=2040, type=linear]; -"2041 reshape_68" [id=2041, type=reshape]; -"2042 permute_70" [id=2042, type=permute]; -"2043 select_45" [id=2043, type=select]; -"2044 select_46" [id=2044, type=select]; -"2045 select_47" [id=2045, type=select]; -"2046 linalg_vector_norm_30" [id=2046, type=linalg_vector_norm]; -"2047 clamp_min_30" [id=2047, type=clamp_min]; -"2048 expand_as_30" [id=2048, type=expand_as]; -"2049 div_30" [id=2049, type=div]; -"2050 quantize_per_tensor_default_94" [id=2050, type=quantize_per_tensor]; -"2051 dequantize_per_tensor_default_94" [id=2051, type=dequantize_per_tensor]; -"2052 linalg_vector_norm_31" [id=2052, type=linalg_vector_norm]; -"2053 clamp_min_31" [id=2053, type=clamp_min]; -"2054 expand_as_31" [id=2054, type=expand_as]; -"2055 div_31" [id=2055, type=div]; -"2056 quantize_per_tensor_default_95" [id=2056, type=quantize_per_tensor]; -"2057 dequantize_per_tensor_default_95" [id=2057, type=dequantize_per_tensor]; -"2058 transpose_30" [id=2058, type=transpose]; -"2059 matmul_30" [id=2059, type=matmul]; -"2060 clamp_15" [id=2060, type=clamp]; -"2061 exp_15" [id=2061, type=exp]; -"2062 mul_31" [id=2062, type=mul]; -"2063 add_52" [id=2063, type=add]; -"2064 new_zeros_7" [id=2064, type=new_zeros]; -"2065 view_84" [id=2065, type=view]; -"2066 permute_71" [id=2066, type=permute]; -"2067 reshape_69" [id=2067, type=reshape]; -"2068 unsqueeze_44" [id=2068, type=unsqueeze]; -"2069 unsqueeze_45" [id=2069, type=unsqueeze]; -"2070 sub_7" [id=2070, type=sub]; -"2071 ne_7" [id=2071, type=ne]; -"2072 masked_fill_14" [id=2072, type=masked_fill]; -"2073 eq_7" [id=2073, type=eq]; -"2074 masked_fill_15" [id=2074, type=masked_fill]; -"2075 view_85" [id=2075, type=view]; -"2076 unsqueeze_46" [id=2076, type=unsqueeze]; -"2077 unsqueeze_47" [id=2077, type=unsqueeze]; -"2078 add_53" [id=2078, type=add]; -"2079 view_86" [id=2079, type=view]; -"2080 softmax_15" [id=2080, type=softmax]; -"2081 dropout_60" [id=2081, type=dropout]; -"2082 matmul_31" [id=2082, type=matmul]; -"2083 transpose_31" [id=2083, type=transpose]; -"2084 reshape_70" [id=2084, type=reshape]; -"2085 reshape_70_0_0_nncf_smooth_quant_0" [id=2085, type=call_module]; -"2086 quantize_per_tensor_default_96" [id=2086, type=quantize_per_tensor]; -"2087 dequantize_per_tensor_default_96" [id=2087, type=dequantize_per_tensor]; -"2088 linear_95_scale_0" [id=2088, type=get_attr]; -"2089 linear_95_zero_point_0" [id=2089, type=get_attr]; -"2090 _frozen_param96" [id=2090, type=get_attr]; -"2091 dequantize_per_channel_default_96" [id=2091, type=dequantize_per_channel]; -"2092 linear_95" [id=2092, type=linear]; -"2093 dropout_61" [id=2093, type=dropout]; -"2094 view_87" [id=2094, type=view]; -"2095 permute_72" [id=2095, type=permute]; -"2096 reshape_71" [id=2096, type=reshape]; -"2097 roll_15" [id=2097, type=roll]; -"2098 slice_241" [id=2098, type=slice]; -"2099 slice_242" [id=2099, type=slice]; -"2100 slice_243" [id=2100, type=slice]; -"2101 slice_244" [id=2101, type=slice]; -"2102 contiguous_29" [id=2102, type=contiguous]; -"2103 layer_norm_33" [id=2103, type=layer_norm]; -"2104 add_54" [id=2104, type=add]; -"2105 add_54_0_0_nncf_smooth_quant_0" [id=2105, type=call_module]; -"2106 quantize_per_tensor_default_97" [id=2106, type=quantize_per_tensor]; -"2107 dequantize_per_tensor_default_97" [id=2107, type=dequantize_per_tensor]; -"2108 linear_96_scale_0" [id=2108, type=get_attr]; -"2109 linear_96_zero_point_0" [id=2109, type=get_attr]; -"2110 _frozen_param97" [id=2110, type=get_attr]; -"2111 dequantize_per_channel_default_97" [id=2111, type=dequantize_per_channel]; -"2112 linear_96" [id=2112, type=linear]; -"2113 gelu_15" [id=2113, type=gelu]; -"2114 dropout_62" [id=2114, type=dropout]; -"2115 dropout_62_0_0_nncf_smooth_quant_0" [id=2115, type=call_module]; -"2116 quantize_per_tensor_default_98" [id=2116, type=quantize_per_tensor]; -"2117 dequantize_per_tensor_default_98" [id=2117, type=dequantize_per_tensor]; -"2118 linear_97_scale_0" [id=2118, type=get_attr]; -"2119 linear_97_zero_point_0" [id=2119, type=get_attr]; -"2120 _frozen_param98" [id=2120, type=get_attr]; -"2121 dequantize_per_channel_default_98" [id=2121, type=dequantize_per_channel]; -"2122 linear_97" [id=2122, type=linear]; -"2123 dropout_63" [id=2123, type=dropout]; -"2124 layer_norm_34" [id=2124, type=layer_norm]; -"2125 add_55" [id=2125, type=add]; -"2126 features_5_12_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=2126, type=call_module]; -"2127 linear_98_scale_0" [id=2127, type=get_attr]; -"2128 linear_98_zero_point_0" [id=2128, type=get_attr]; -"2129 _frozen_param99" [id=2129, type=get_attr]; -"2130 dequantize_per_channel_default_99" [id=2130, type=dequantize_per_channel]; -"2131 linear_98" [id=2131, type=linear]; -"2132 relu__16" [id=2132, type=relu_]; -"2133 relu__16_0_0_nncf_smooth_quant_0" [id=2133, type=call_module]; -"2134 linear_99_scale_0" [id=2134, type=get_attr]; -"2135 linear_99_zero_point_0" [id=2135, type=get_attr]; -"2136 _frozen_param100" [id=2136, type=get_attr]; -"2137 dequantize_per_channel_default_100" [id=2137, type=dequantize_per_channel]; -"2138 linear_99" [id=2138, type=linear]; -"2139 view_88" [id=2139, type=view]; -"2140 index_16" [id=2140, type=index]; -"2141 view_89" [id=2141, type=view]; -"2142 permute_73" [id=2142, type=permute]; -"2143 contiguous_30" [id=2143, type=contiguous]; -"2144 unsqueeze_48" [id=2144, type=unsqueeze]; -"2145 sigmoid_16" [id=2145, type=sigmoid]; -"2146 mul_32" [id=2146, type=mul]; -"2147 pad_18" [id=2147, type=pad]; -"2148 view_90" [id=2148, type=view]; -"2149 permute_74" [id=2149, type=permute]; -"2150 reshape_72" [id=2150, type=reshape]; -"2151 clone_16" [id=2151, type=clone]; -"2152 reshape_72_0_0_nncf_smooth_quant_0" [id=2152, type=call_module]; -"2153 quantize_per_tensor_default_99" [id=2153, type=quantize_per_tensor]; -"2154 dequantize_per_tensor_default_99" [id=2154, type=dequantize_per_tensor]; -"2155 linear_100_scale_0" [id=2155, type=get_attr]; -"2156 linear_100_zero_point_0" [id=2156, type=get_attr]; -"2157 _frozen_param101" [id=2157, type=get_attr]; -"2158 dequantize_per_channel_default_101" [id=2158, type=dequantize_per_channel]; -"2159 linear_100" [id=2159, type=linear]; -"2160 reshape_73" [id=2160, type=reshape]; -"2161 permute_75" [id=2161, type=permute]; -"2162 select_48" [id=2162, type=select]; -"2163 select_49" [id=2163, type=select]; -"2164 select_50" [id=2164, type=select]; -"2165 linalg_vector_norm_32" [id=2165, type=linalg_vector_norm]; -"2166 clamp_min_32" [id=2166, type=clamp_min]; -"2167 expand_as_32" [id=2167, type=expand_as]; -"2168 div_32" [id=2168, type=div]; -"2169 quantize_per_tensor_default_100" [id=2169, type=quantize_per_tensor]; -"2170 dequantize_per_tensor_default_100" [id=2170, type=dequantize_per_tensor]; -"2171 linalg_vector_norm_33" [id=2171, type=linalg_vector_norm]; -"2172 clamp_min_33" [id=2172, type=clamp_min]; -"2173 expand_as_33" [id=2173, type=expand_as]; -"2174 div_33" [id=2174, type=div]; -"2175 quantize_per_tensor_default_101" [id=2175, type=quantize_per_tensor]; -"2176 dequantize_per_tensor_default_101" [id=2176, type=dequantize_per_tensor]; -"2177 transpose_32" [id=2177, type=transpose]; -"2178 matmul_32" [id=2178, type=matmul]; -"2179 clamp_16" [id=2179, type=clamp]; -"2180 exp_16" [id=2180, type=exp]; -"2181 mul_33" [id=2181, type=mul]; -"2182 add_56" [id=2182, type=add]; -"2183 softmax_16" [id=2183, type=softmax]; -"2184 dropout_64" [id=2184, type=dropout]; -"2185 matmul_33" [id=2185, type=matmul]; -"2186 transpose_33" [id=2186, type=transpose]; -"2187 reshape_74" [id=2187, type=reshape]; -"2188 reshape_74_0_0_nncf_smooth_quant_0" [id=2188, type=call_module]; -"2189 quantize_per_tensor_default_102" [id=2189, type=quantize_per_tensor]; -"2190 dequantize_per_tensor_default_102" [id=2190, type=dequantize_per_tensor]; -"2191 linear_101_scale_0" [id=2191, type=get_attr]; -"2192 linear_101_zero_point_0" [id=2192, type=get_attr]; -"2193 _frozen_param102" [id=2193, type=get_attr]; -"2194 dequantize_per_channel_default_102" [id=2194, type=dequantize_per_channel]; -"2195 linear_101" [id=2195, type=linear]; -"2196 dropout_65" [id=2196, type=dropout]; -"2197 view_91" [id=2197, type=view]; -"2198 permute_76" [id=2198, type=permute]; -"2199 reshape_75" [id=2199, type=reshape]; -"2200 slice_246" [id=2200, type=slice]; -"2201 slice_247" [id=2201, type=slice]; -"2202 slice_248" [id=2202, type=slice]; -"2203 slice_249" [id=2203, type=slice]; -"2204 contiguous_31" [id=2204, type=contiguous]; -"2205 layer_norm_35" [id=2205, type=layer_norm]; -"2206 add_57" [id=2206, type=add]; -"2207 add_57_0_0_nncf_smooth_quant_0" [id=2207, type=call_module]; -"2208 quantize_per_tensor_default_103" [id=2208, type=quantize_per_tensor]; -"2209 dequantize_per_tensor_default_103" [id=2209, type=dequantize_per_tensor]; -"2210 linear_102_scale_0" [id=2210, type=get_attr]; -"2211 linear_102_zero_point_0" [id=2211, type=get_attr]; -"2212 _frozen_param103" [id=2212, type=get_attr]; -"2213 dequantize_per_channel_default_103" [id=2213, type=dequantize_per_channel]; -"2214 linear_102" [id=2214, type=linear]; -"2215 gelu_16" [id=2215, type=gelu]; -"2216 dropout_66" [id=2216, type=dropout]; -"2217 dropout_66_0_0_nncf_smooth_quant_0" [id=2217, type=call_module]; -"2218 quantize_per_tensor_default_104" [id=2218, type=quantize_per_tensor]; -"2219 dequantize_per_tensor_default_104" [id=2219, type=dequantize_per_tensor]; -"2220 linear_103_scale_0" [id=2220, type=get_attr]; -"2221 linear_103_zero_point_0" [id=2221, type=get_attr]; -"2222 _frozen_param104" [id=2222, type=get_attr]; -"2223 dequantize_per_channel_default_104" [id=2223, type=dequantize_per_channel]; -"2224 linear_103" [id=2224, type=linear]; -"2225 dropout_67" [id=2225, type=dropout]; -"2226 layer_norm_36" [id=2226, type=layer_norm]; -"2227 add_58" [id=2227, type=add]; -"2228 features_5_13_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=2228, type=call_module]; -"2229 linear_104_scale_0" [id=2229, type=get_attr]; -"2230 linear_104_zero_point_0" [id=2230, type=get_attr]; -"2231 _frozen_param105" [id=2231, type=get_attr]; -"2232 dequantize_per_channel_default_105" [id=2232, type=dequantize_per_channel]; -"2233 linear_104" [id=2233, type=linear]; -"2234 relu__17" [id=2234, type=relu_]; -"2235 relu__17_0_0_nncf_smooth_quant_0" [id=2235, type=call_module]; -"2236 linear_105_scale_0" [id=2236, type=get_attr]; -"2237 linear_105_zero_point_0" [id=2237, type=get_attr]; -"2238 _frozen_param106" [id=2238, type=get_attr]; -"2239 dequantize_per_channel_default_106" [id=2239, type=dequantize_per_channel]; -"2240 linear_105" [id=2240, type=linear]; -"2241 view_92" [id=2241, type=view]; -"2242 index_17" [id=2242, type=index]; -"2243 view_93" [id=2243, type=view]; -"2244 permute_77" [id=2244, type=permute]; -"2245 contiguous_32" [id=2245, type=contiguous]; -"2246 unsqueeze_49" [id=2246, type=unsqueeze]; -"2247 sigmoid_17" [id=2247, type=sigmoid]; -"2248 mul_34" [id=2248, type=mul]; -"2249 pad_19" [id=2249, type=pad]; -"2250 roll_16" [id=2250, type=roll]; -"2251 view_94" [id=2251, type=view]; -"2252 permute_78" [id=2252, type=permute]; -"2253 reshape_76" [id=2253, type=reshape]; -"2254 clone_17" [id=2254, type=clone]; -"2255 reshape_76_0_0_nncf_smooth_quant_0" [id=2255, type=call_module]; -"2256 quantize_per_tensor_default_105" [id=2256, type=quantize_per_tensor]; -"2257 dequantize_per_tensor_default_105" [id=2257, type=dequantize_per_tensor]; -"2258 linear_106_scale_0" [id=2258, type=get_attr]; -"2259 linear_106_zero_point_0" [id=2259, type=get_attr]; -"2260 _frozen_param107" [id=2260, type=get_attr]; -"2261 dequantize_per_channel_default_107" [id=2261, type=dequantize_per_channel]; -"2262 linear_106" [id=2262, type=linear]; -"2263 reshape_77" [id=2263, type=reshape]; -"2264 permute_79" [id=2264, type=permute]; -"2265 select_51" [id=2265, type=select]; -"2266 select_52" [id=2266, type=select]; -"2267 select_53" [id=2267, type=select]; -"2268 linalg_vector_norm_34" [id=2268, type=linalg_vector_norm]; -"2269 clamp_min_34" [id=2269, type=clamp_min]; -"2270 expand_as_34" [id=2270, type=expand_as]; -"2271 div_34" [id=2271, type=div]; -"2272 quantize_per_tensor_default_106" [id=2272, type=quantize_per_tensor]; -"2273 dequantize_per_tensor_default_106" [id=2273, type=dequantize_per_tensor]; -"2274 linalg_vector_norm_35" [id=2274, type=linalg_vector_norm]; -"2275 clamp_min_35" [id=2275, type=clamp_min]; -"2276 expand_as_35" [id=2276, type=expand_as]; -"2277 div_35" [id=2277, type=div]; -"2278 quantize_per_tensor_default_107" [id=2278, type=quantize_per_tensor]; -"2279 dequantize_per_tensor_default_107" [id=2279, type=dequantize_per_tensor]; -"2280 transpose_34" [id=2280, type=transpose]; -"2281 matmul_34" [id=2281, type=matmul]; -"2282 clamp_17" [id=2282, type=clamp]; -"2283 exp_17" [id=2283, type=exp]; -"2284 mul_35" [id=2284, type=mul]; -"2285 add_59" [id=2285, type=add]; -"2286 new_zeros_8" [id=2286, type=new_zeros]; -"2287 view_95" [id=2287, type=view]; -"2288 permute_80" [id=2288, type=permute]; -"2289 reshape_78" [id=2289, type=reshape]; -"2290 unsqueeze_50" [id=2290, type=unsqueeze]; -"2291 unsqueeze_51" [id=2291, type=unsqueeze]; -"2292 sub_8" [id=2292, type=sub]; -"2293 ne_8" [id=2293, type=ne]; -"2294 masked_fill_16" [id=2294, type=masked_fill]; -"2295 eq_8" [id=2295, type=eq]; -"2296 masked_fill_17" [id=2296, type=masked_fill]; -"2297 view_96" [id=2297, type=view]; -"2298 unsqueeze_52" [id=2298, type=unsqueeze]; -"2299 unsqueeze_53" [id=2299, type=unsqueeze]; -"2300 add_60" [id=2300, type=add]; -"2301 view_97" [id=2301, type=view]; -"2302 softmax_17" [id=2302, type=softmax]; -"2303 dropout_68" [id=2303, type=dropout]; -"2304 matmul_35" [id=2304, type=matmul]; -"2305 transpose_35" [id=2305, type=transpose]; -"2306 reshape_79" [id=2306, type=reshape]; -"2307 reshape_79_0_0_nncf_smooth_quant_0" [id=2307, type=call_module]; -"2308 quantize_per_tensor_default_108" [id=2308, type=quantize_per_tensor]; -"2309 dequantize_per_tensor_default_108" [id=2309, type=dequantize_per_tensor]; -"2310 linear_107_scale_0" [id=2310, type=get_attr]; -"2311 linear_107_zero_point_0" [id=2311, type=get_attr]; -"2312 _frozen_param108" [id=2312, type=get_attr]; -"2313 dequantize_per_channel_default_108" [id=2313, type=dequantize_per_channel]; -"2314 linear_107" [id=2314, type=linear]; -"2315 dropout_69" [id=2315, type=dropout]; -"2316 view_98" [id=2316, type=view]; -"2317 permute_81" [id=2317, type=permute]; -"2318 reshape_80" [id=2318, type=reshape]; -"2319 roll_17" [id=2319, type=roll]; -"2320 slice_269" [id=2320, type=slice]; -"2321 slice_270" [id=2321, type=slice]; -"2322 slice_271" [id=2322, type=slice]; -"2323 slice_272" [id=2323, type=slice]; -"2324 contiguous_33" [id=2324, type=contiguous]; -"2325 layer_norm_37" [id=2325, type=layer_norm]; -"2326 add_61" [id=2326, type=add]; -"2327 add_61_0_0_nncf_smooth_quant_0" [id=2327, type=call_module]; -"2328 quantize_per_tensor_default_109" [id=2328, type=quantize_per_tensor]; -"2329 dequantize_per_tensor_default_109" [id=2329, type=dequantize_per_tensor]; -"2330 linear_108_scale_0" [id=2330, type=get_attr]; -"2331 linear_108_zero_point_0" [id=2331, type=get_attr]; -"2332 _frozen_param109" [id=2332, type=get_attr]; -"2333 dequantize_per_channel_default_109" [id=2333, type=dequantize_per_channel]; -"2334 linear_108" [id=2334, type=linear]; -"2335 gelu_17" [id=2335, type=gelu]; -"2336 dropout_70" [id=2336, type=dropout]; -"2337 dropout_70_0_0_nncf_smooth_quant_0" [id=2337, type=call_module]; -"2338 quantize_per_tensor_default_110" [id=2338, type=quantize_per_tensor]; -"2339 dequantize_per_tensor_default_110" [id=2339, type=dequantize_per_tensor]; -"2340 linear_109_scale_0" [id=2340, type=get_attr]; -"2341 linear_109_zero_point_0" [id=2341, type=get_attr]; -"2342 _frozen_param110" [id=2342, type=get_attr]; -"2343 dequantize_per_channel_default_110" [id=2343, type=dequantize_per_channel]; -"2344 linear_109" [id=2344, type=linear]; -"2345 dropout_71" [id=2345, type=dropout]; -"2346 layer_norm_38" [id=2346, type=layer_norm]; -"2347 add_62" [id=2347, type=add]; -"2348 features_5_14_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=2348, type=call_module]; -"2349 linear_110_scale_0" [id=2349, type=get_attr]; -"2350 linear_110_zero_point_0" [id=2350, type=get_attr]; -"2351 _frozen_param111" [id=2351, type=get_attr]; -"2352 dequantize_per_channel_default_111" [id=2352, type=dequantize_per_channel]; -"2353 linear_110" [id=2353, type=linear]; -"2354 relu__18" [id=2354, type=relu_]; -"2355 relu__18_0_0_nncf_smooth_quant_0" [id=2355, type=call_module]; -"2356 linear_111_scale_0" [id=2356, type=get_attr]; -"2357 linear_111_zero_point_0" [id=2357, type=get_attr]; -"2358 _frozen_param112" [id=2358, type=get_attr]; -"2359 dequantize_per_channel_default_112" [id=2359, type=dequantize_per_channel]; -"2360 linear_111" [id=2360, type=linear]; -"2361 view_99" [id=2361, type=view]; -"2362 index_18" [id=2362, type=index]; -"2363 view_100" [id=2363, type=view]; -"2364 permute_82" [id=2364, type=permute]; -"2365 contiguous_34" [id=2365, type=contiguous]; -"2366 unsqueeze_54" [id=2366, type=unsqueeze]; -"2367 sigmoid_18" [id=2367, type=sigmoid]; -"2368 mul_36" [id=2368, type=mul]; -"2369 pad_20" [id=2369, type=pad]; -"2370 view_101" [id=2370, type=view]; -"2371 permute_83" [id=2371, type=permute]; -"2372 reshape_81" [id=2372, type=reshape]; -"2373 clone_18" [id=2373, type=clone]; -"2374 reshape_81_0_0_nncf_smooth_quant_0" [id=2374, type=call_module]; -"2375 quantize_per_tensor_default_111" [id=2375, type=quantize_per_tensor]; -"2376 dequantize_per_tensor_default_111" [id=2376, type=dequantize_per_tensor]; -"2377 linear_112_scale_0" [id=2377, type=get_attr]; -"2378 linear_112_zero_point_0" [id=2378, type=get_attr]; -"2379 _frozen_param113" [id=2379, type=get_attr]; -"2380 dequantize_per_channel_default_113" [id=2380, type=dequantize_per_channel]; -"2381 linear_112" [id=2381, type=linear]; -"2382 reshape_82" [id=2382, type=reshape]; -"2383 permute_84" [id=2383, type=permute]; -"2384 select_54" [id=2384, type=select]; -"2385 select_55" [id=2385, type=select]; -"2386 select_56" [id=2386, type=select]; -"2387 linalg_vector_norm_36" [id=2387, type=linalg_vector_norm]; -"2388 clamp_min_36" [id=2388, type=clamp_min]; -"2389 expand_as_36" [id=2389, type=expand_as]; -"2390 div_36" [id=2390, type=div]; -"2391 quantize_per_tensor_default_112" [id=2391, type=quantize_per_tensor]; -"2392 dequantize_per_tensor_default_112" [id=2392, type=dequantize_per_tensor]; -"2393 linalg_vector_norm_37" [id=2393, type=linalg_vector_norm]; -"2394 clamp_min_37" [id=2394, type=clamp_min]; -"2395 expand_as_37" [id=2395, type=expand_as]; -"2396 div_37" [id=2396, type=div]; -"2397 quantize_per_tensor_default_113" [id=2397, type=quantize_per_tensor]; -"2398 dequantize_per_tensor_default_113" [id=2398, type=dequantize_per_tensor]; -"2399 transpose_36" [id=2399, type=transpose]; -"2400 matmul_36" [id=2400, type=matmul]; -"2401 clamp_18" [id=2401, type=clamp]; -"2402 exp_18" [id=2402, type=exp]; -"2403 mul_37" [id=2403, type=mul]; -"2404 add_63" [id=2404, type=add]; -"2405 softmax_18" [id=2405, type=softmax]; -"2406 dropout_72" [id=2406, type=dropout]; -"2407 matmul_37" [id=2407, type=matmul]; -"2408 transpose_37" [id=2408, type=transpose]; -"2409 reshape_83" [id=2409, type=reshape]; -"2410 reshape_83_0_0_nncf_smooth_quant_0" [id=2410, type=call_module]; -"2411 quantize_per_tensor_default_114" [id=2411, type=quantize_per_tensor]; -"2412 dequantize_per_tensor_default_114" [id=2412, type=dequantize_per_tensor]; -"2413 linear_113_scale_0" [id=2413, type=get_attr]; -"2414 linear_113_zero_point_0" [id=2414, type=get_attr]; -"2415 _frozen_param114" [id=2415, type=get_attr]; -"2416 dequantize_per_channel_default_114" [id=2416, type=dequantize_per_channel]; -"2417 linear_113" [id=2417, type=linear]; -"2418 dropout_73" [id=2418, type=dropout]; -"2419 view_102" [id=2419, type=view]; -"2420 permute_85" [id=2420, type=permute]; -"2421 reshape_84" [id=2421, type=reshape]; -"2422 slice_274" [id=2422, type=slice]; -"2423 slice_275" [id=2423, type=slice]; -"2424 slice_276" [id=2424, type=slice]; -"2425 slice_277" [id=2425, type=slice]; -"2426 contiguous_35" [id=2426, type=contiguous]; -"2427 layer_norm_39" [id=2427, type=layer_norm]; -"2428 add_64" [id=2428, type=add]; -"2429 add_64_0_0_nncf_smooth_quant_0" [id=2429, type=call_module]; -"2430 quantize_per_tensor_default_115" [id=2430, type=quantize_per_tensor]; -"2431 dequantize_per_tensor_default_115" [id=2431, type=dequantize_per_tensor]; -"2432 linear_114_scale_0" [id=2432, type=get_attr]; -"2433 linear_114_zero_point_0" [id=2433, type=get_attr]; -"2434 _frozen_param115" [id=2434, type=get_attr]; -"2435 dequantize_per_channel_default_115" [id=2435, type=dequantize_per_channel]; -"2436 linear_114" [id=2436, type=linear]; -"2437 gelu_18" [id=2437, type=gelu]; -"2438 dropout_74" [id=2438, type=dropout]; -"2439 dropout_74_0_0_nncf_smooth_quant_0" [id=2439, type=call_module]; -"2440 quantize_per_tensor_default_116" [id=2440, type=quantize_per_tensor]; -"2441 dequantize_per_tensor_default_116" [id=2441, type=dequantize_per_tensor]; -"2442 linear_115_scale_0" [id=2442, type=get_attr]; -"2443 linear_115_zero_point_0" [id=2443, type=get_attr]; -"2444 _frozen_param116" [id=2444, type=get_attr]; -"2445 dequantize_per_channel_default_116" [id=2445, type=dequantize_per_channel]; -"2446 linear_115" [id=2446, type=linear]; -"2447 dropout_75" [id=2447, type=dropout]; -"2448 layer_norm_40" [id=2448, type=layer_norm]; -"2449 add_65" [id=2449, type=add]; -"2450 features_5_15_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=2450, type=call_module]; -"2451 linear_116_scale_0" [id=2451, type=get_attr]; -"2452 linear_116_zero_point_0" [id=2452, type=get_attr]; -"2453 _frozen_param117" [id=2453, type=get_attr]; -"2454 dequantize_per_channel_default_117" [id=2454, type=dequantize_per_channel]; -"2455 linear_116" [id=2455, type=linear]; -"2456 relu__19" [id=2456, type=relu_]; -"2457 relu__19_0_0_nncf_smooth_quant_0" [id=2457, type=call_module]; -"2458 linear_117_scale_0" [id=2458, type=get_attr]; -"2459 linear_117_zero_point_0" [id=2459, type=get_attr]; -"2460 _frozen_param118" [id=2460, type=get_attr]; -"2461 dequantize_per_channel_default_118" [id=2461, type=dequantize_per_channel]; -"2462 linear_117" [id=2462, type=linear]; -"2463 view_103" [id=2463, type=view]; -"2464 index_19" [id=2464, type=index]; -"2465 view_104" [id=2465, type=view]; -"2466 permute_86" [id=2466, type=permute]; -"2467 contiguous_36" [id=2467, type=contiguous]; -"2468 unsqueeze_55" [id=2468, type=unsqueeze]; -"2469 sigmoid_19" [id=2469, type=sigmoid]; -"2470 mul_38" [id=2470, type=mul]; -"2471 pad_21" [id=2471, type=pad]; -"2472 roll_18" [id=2472, type=roll]; -"2473 view_105" [id=2473, type=view]; -"2474 permute_87" [id=2474, type=permute]; -"2475 reshape_85" [id=2475, type=reshape]; -"2476 clone_19" [id=2476, type=clone]; -"2477 reshape_85_0_0_nncf_smooth_quant_0" [id=2477, type=call_module]; -"2478 quantize_per_tensor_default_117" [id=2478, type=quantize_per_tensor]; -"2479 dequantize_per_tensor_default_117" [id=2479, type=dequantize_per_tensor]; -"2480 linear_118_scale_0" [id=2480, type=get_attr]; -"2481 linear_118_zero_point_0" [id=2481, type=get_attr]; -"2482 _frozen_param119" [id=2482, type=get_attr]; -"2483 dequantize_per_channel_default_119" [id=2483, type=dequantize_per_channel]; -"2484 linear_118" [id=2484, type=linear]; -"2485 reshape_86" [id=2485, type=reshape]; -"2486 permute_88" [id=2486, type=permute]; -"2487 select_57" [id=2487, type=select]; -"2488 select_58" [id=2488, type=select]; -"2489 select_59" [id=2489, type=select]; -"2490 linalg_vector_norm_38" [id=2490, type=linalg_vector_norm]; -"2491 clamp_min_38" [id=2491, type=clamp_min]; -"2492 expand_as_38" [id=2492, type=expand_as]; -"2493 div_38" [id=2493, type=div]; -"2494 quantize_per_tensor_default_118" [id=2494, type=quantize_per_tensor]; -"2495 dequantize_per_tensor_default_118" [id=2495, type=dequantize_per_tensor]; -"2496 linalg_vector_norm_39" [id=2496, type=linalg_vector_norm]; -"2497 clamp_min_39" [id=2497, type=clamp_min]; -"2498 expand_as_39" [id=2498, type=expand_as]; -"2499 div_39" [id=2499, type=div]; -"2500 quantize_per_tensor_default_119" [id=2500, type=quantize_per_tensor]; -"2501 dequantize_per_tensor_default_119" [id=2501, type=dequantize_per_tensor]; -"2502 transpose_38" [id=2502, type=transpose]; -"2503 matmul_38" [id=2503, type=matmul]; -"2504 clamp_19" [id=2504, type=clamp]; -"2505 exp_19" [id=2505, type=exp]; -"2506 mul_39" [id=2506, type=mul]; -"2507 add_66" [id=2507, type=add]; -"2508 new_zeros_9" [id=2508, type=new_zeros]; -"2509 view_106" [id=2509, type=view]; -"2510 permute_89" [id=2510, type=permute]; -"2511 reshape_87" [id=2511, type=reshape]; -"2512 unsqueeze_56" [id=2512, type=unsqueeze]; -"2513 unsqueeze_57" [id=2513, type=unsqueeze]; -"2514 sub_9" [id=2514, type=sub]; -"2515 ne_9" [id=2515, type=ne]; -"2516 masked_fill_18" [id=2516, type=masked_fill]; -"2517 eq_9" [id=2517, type=eq]; -"2518 masked_fill_19" [id=2518, type=masked_fill]; -"2519 view_107" [id=2519, type=view]; -"2520 unsqueeze_58" [id=2520, type=unsqueeze]; -"2521 unsqueeze_59" [id=2521, type=unsqueeze]; -"2522 add_67" [id=2522, type=add]; -"2523 view_108" [id=2523, type=view]; -"2524 softmax_19" [id=2524, type=softmax]; -"2525 dropout_76" [id=2525, type=dropout]; -"2526 matmul_39" [id=2526, type=matmul]; -"2527 transpose_39" [id=2527, type=transpose]; -"2528 reshape_88" [id=2528, type=reshape]; -"2529 reshape_88_0_0_nncf_smooth_quant_0" [id=2529, type=call_module]; -"2530 quantize_per_tensor_default_120" [id=2530, type=quantize_per_tensor]; -"2531 dequantize_per_tensor_default_120" [id=2531, type=dequantize_per_tensor]; -"2532 linear_119_scale_0" [id=2532, type=get_attr]; -"2533 linear_119_zero_point_0" [id=2533, type=get_attr]; -"2534 _frozen_param120" [id=2534, type=get_attr]; -"2535 dequantize_per_channel_default_120" [id=2535, type=dequantize_per_channel]; -"2536 linear_119" [id=2536, type=linear]; -"2537 dropout_77" [id=2537, type=dropout]; -"2538 view_109" [id=2538, type=view]; -"2539 permute_90" [id=2539, type=permute]; -"2540 reshape_89" [id=2540, type=reshape]; -"2541 roll_19" [id=2541, type=roll]; -"2542 slice_297" [id=2542, type=slice]; -"2543 slice_298" [id=2543, type=slice]; -"2544 slice_299" [id=2544, type=slice]; -"2545 slice_300" [id=2545, type=slice]; -"2546 contiguous_37" [id=2546, type=contiguous]; -"2547 layer_norm_41" [id=2547, type=layer_norm]; -"2548 add_68" [id=2548, type=add]; -"2549 add_68_0_0_nncf_smooth_quant_0" [id=2549, type=call_module]; -"2550 quantize_per_tensor_default_121" [id=2550, type=quantize_per_tensor]; -"2551 dequantize_per_tensor_default_121" [id=2551, type=dequantize_per_tensor]; -"2552 linear_120_scale_0" [id=2552, type=get_attr]; -"2553 linear_120_zero_point_0" [id=2553, type=get_attr]; -"2554 _frozen_param121" [id=2554, type=get_attr]; -"2555 dequantize_per_channel_default_121" [id=2555, type=dequantize_per_channel]; -"2556 linear_120" [id=2556, type=linear]; -"2557 gelu_19" [id=2557, type=gelu]; -"2558 dropout_78" [id=2558, type=dropout]; -"2559 dropout_78_0_0_nncf_smooth_quant_0" [id=2559, type=call_module]; -"2560 quantize_per_tensor_default_122" [id=2560, type=quantize_per_tensor]; -"2561 dequantize_per_tensor_default_122" [id=2561, type=dequantize_per_tensor]; -"2562 linear_121_scale_0" [id=2562, type=get_attr]; -"2563 linear_121_zero_point_0" [id=2563, type=get_attr]; -"2564 _frozen_param122" [id=2564, type=get_attr]; -"2565 dequantize_per_channel_default_122" [id=2565, type=dequantize_per_channel]; -"2566 linear_121" [id=2566, type=linear]; -"2567 dropout_79" [id=2567, type=dropout]; -"2568 layer_norm_42" [id=2568, type=layer_norm]; -"2569 add_69" [id=2569, type=add]; -"2570 features_5_16_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=2570, type=call_module]; -"2571 linear_122_scale_0" [id=2571, type=get_attr]; -"2572 linear_122_zero_point_0" [id=2572, type=get_attr]; -"2573 _frozen_param123" [id=2573, type=get_attr]; -"2574 dequantize_per_channel_default_123" [id=2574, type=dequantize_per_channel]; -"2575 linear_122" [id=2575, type=linear]; -"2576 relu__20" [id=2576, type=relu_]; -"2577 relu__20_0_0_nncf_smooth_quant_0" [id=2577, type=call_module]; -"2578 linear_123_scale_0" [id=2578, type=get_attr]; -"2579 linear_123_zero_point_0" [id=2579, type=get_attr]; -"2580 _frozen_param124" [id=2580, type=get_attr]; -"2581 dequantize_per_channel_default_124" [id=2581, type=dequantize_per_channel]; -"2582 linear_123" [id=2582, type=linear]; -"2583 view_110" [id=2583, type=view]; -"2584 index_20" [id=2584, type=index]; -"2585 view_111" [id=2585, type=view]; -"2586 permute_91" [id=2586, type=permute]; -"2587 contiguous_38" [id=2587, type=contiguous]; -"2588 unsqueeze_60" [id=2588, type=unsqueeze]; -"2589 sigmoid_20" [id=2589, type=sigmoid]; -"2590 mul_40" [id=2590, type=mul]; -"2591 pad_22" [id=2591, type=pad]; -"2592 view_112" [id=2592, type=view]; -"2593 permute_92" [id=2593, type=permute]; -"2594 reshape_90" [id=2594, type=reshape]; -"2595 clone_20" [id=2595, type=clone]; -"2596 reshape_90_0_0_nncf_smooth_quant_0" [id=2596, type=call_module]; -"2597 quantize_per_tensor_default_123" [id=2597, type=quantize_per_tensor]; -"2598 dequantize_per_tensor_default_123" [id=2598, type=dequantize_per_tensor]; -"2599 linear_124_scale_0" [id=2599, type=get_attr]; -"2600 linear_124_zero_point_0" [id=2600, type=get_attr]; -"2601 _frozen_param125" [id=2601, type=get_attr]; -"2602 dequantize_per_channel_default_125" [id=2602, type=dequantize_per_channel]; -"2603 linear_124" [id=2603, type=linear]; -"2604 reshape_91" [id=2604, type=reshape]; -"2605 permute_93" [id=2605, type=permute]; -"2606 select_60" [id=2606, type=select]; -"2607 select_61" [id=2607, type=select]; -"2608 select_62" [id=2608, type=select]; -"2609 linalg_vector_norm_40" [id=2609, type=linalg_vector_norm]; -"2610 clamp_min_40" [id=2610, type=clamp_min]; -"2611 expand_as_40" [id=2611, type=expand_as]; -"2612 div_40" [id=2612, type=div]; -"2613 quantize_per_tensor_default_124" [id=2613, type=quantize_per_tensor]; -"2614 dequantize_per_tensor_default_124" [id=2614, type=dequantize_per_tensor]; -"2615 linalg_vector_norm_41" [id=2615, type=linalg_vector_norm]; -"2616 clamp_min_41" [id=2616, type=clamp_min]; -"2617 expand_as_41" [id=2617, type=expand_as]; -"2618 div_41" [id=2618, type=div]; -"2619 quantize_per_tensor_default_125" [id=2619, type=quantize_per_tensor]; -"2620 dequantize_per_tensor_default_125" [id=2620, type=dequantize_per_tensor]; -"2621 transpose_40" [id=2621, type=transpose]; -"2622 matmul_40" [id=2622, type=matmul]; -"2623 clamp_20" [id=2623, type=clamp]; -"2624 exp_20" [id=2624, type=exp]; -"2625 mul_41" [id=2625, type=mul]; -"2626 add_70" [id=2626, type=add]; -"2627 softmax_20" [id=2627, type=softmax]; -"2628 dropout_80" [id=2628, type=dropout]; -"2629 matmul_41" [id=2629, type=matmul]; -"2630 transpose_41" [id=2630, type=transpose]; -"2631 reshape_92" [id=2631, type=reshape]; -"2632 reshape_92_0_0_nncf_smooth_quant_0" [id=2632, type=call_module]; -"2633 quantize_per_tensor_default_126" [id=2633, type=quantize_per_tensor]; -"2634 dequantize_per_tensor_default_126" [id=2634, type=dequantize_per_tensor]; -"2635 linear_125_scale_0" [id=2635, type=get_attr]; -"2636 linear_125_zero_point_0" [id=2636, type=get_attr]; -"2637 _frozen_param126" [id=2637, type=get_attr]; -"2638 dequantize_per_channel_default_126" [id=2638, type=dequantize_per_channel]; -"2639 linear_125" [id=2639, type=linear]; -"2640 dropout_81" [id=2640, type=dropout]; -"2641 view_113" [id=2641, type=view]; -"2642 permute_94" [id=2642, type=permute]; -"2643 reshape_93" [id=2643, type=reshape]; -"2644 slice_302" [id=2644, type=slice]; -"2645 slice_303" [id=2645, type=slice]; -"2646 slice_304" [id=2646, type=slice]; -"2647 slice_305" [id=2647, type=slice]; -"2648 contiguous_39" [id=2648, type=contiguous]; -"2649 layer_norm_43" [id=2649, type=layer_norm]; -"2650 add_71" [id=2650, type=add]; -"2651 add_71_0_0_nncf_smooth_quant_0" [id=2651, type=call_module]; -"2652 quantize_per_tensor_default_127" [id=2652, type=quantize_per_tensor]; -"2653 dequantize_per_tensor_default_127" [id=2653, type=dequantize_per_tensor]; -"2654 linear_126_scale_0" [id=2654, type=get_attr]; -"2655 linear_126_zero_point_0" [id=2655, type=get_attr]; -"2656 _frozen_param127" [id=2656, type=get_attr]; -"2657 dequantize_per_channel_default_127" [id=2657, type=dequantize_per_channel]; -"2658 linear_126" [id=2658, type=linear]; -"2659 gelu_20" [id=2659, type=gelu]; -"2660 dropout_82" [id=2660, type=dropout]; -"2661 dropout_82_0_0_nncf_smooth_quant_0" [id=2661, type=call_module]; -"2662 quantize_per_tensor_default_128" [id=2662, type=quantize_per_tensor]; -"2663 dequantize_per_tensor_default_128" [id=2663, type=dequantize_per_tensor]; -"2664 linear_127_scale_0" [id=2664, type=get_attr]; -"2665 linear_127_zero_point_0" [id=2665, type=get_attr]; -"2666 _frozen_param128" [id=2666, type=get_attr]; -"2667 dequantize_per_channel_default_128" [id=2667, type=dequantize_per_channel]; -"2668 linear_127" [id=2668, type=linear]; -"2669 dropout_83" [id=2669, type=dropout]; -"2670 layer_norm_44" [id=2670, type=layer_norm]; -"2671 add_72" [id=2671, type=add]; -"2672 features_5_17_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=2672, type=call_module]; -"2673 linear_128_scale_0" [id=2673, type=get_attr]; -"2674 linear_128_zero_point_0" [id=2674, type=get_attr]; -"2675 _frozen_param129" [id=2675, type=get_attr]; -"2676 dequantize_per_channel_default_129" [id=2676, type=dequantize_per_channel]; -"2677 linear_128" [id=2677, type=linear]; -"2678 relu__21" [id=2678, type=relu_]; -"2679 relu__21_0_0_nncf_smooth_quant_0" [id=2679, type=call_module]; -"2680 linear_129_scale_0" [id=2680, type=get_attr]; -"2681 linear_129_zero_point_0" [id=2681, type=get_attr]; -"2682 _frozen_param130" [id=2682, type=get_attr]; -"2683 dequantize_per_channel_default_130" [id=2683, type=dequantize_per_channel]; -"2684 linear_129" [id=2684, type=linear]; -"2685 view_114" [id=2685, type=view]; -"2686 index_21" [id=2686, type=index]; -"2687 view_115" [id=2687, type=view]; -"2688 permute_95" [id=2688, type=permute]; -"2689 contiguous_40" [id=2689, type=contiguous]; -"2690 unsqueeze_61" [id=2690, type=unsqueeze]; -"2691 sigmoid_21" [id=2691, type=sigmoid]; -"2692 mul_42" [id=2692, type=mul]; -"2693 pad_23" [id=2693, type=pad]; -"2694 roll_20" [id=2694, type=roll]; -"2695 view_116" [id=2695, type=view]; -"2696 permute_96" [id=2696, type=permute]; -"2697 reshape_94" [id=2697, type=reshape]; -"2698 clone_21" [id=2698, type=clone]; -"2699 reshape_94_0_0_nncf_smooth_quant_0" [id=2699, type=call_module]; -"2700 quantize_per_tensor_default_129" [id=2700, type=quantize_per_tensor]; -"2701 dequantize_per_tensor_default_129" [id=2701, type=dequantize_per_tensor]; -"2702 linear_130_scale_0" [id=2702, type=get_attr]; -"2703 linear_130_zero_point_0" [id=2703, type=get_attr]; -"2704 _frozen_param131" [id=2704, type=get_attr]; -"2705 dequantize_per_channel_default_131" [id=2705, type=dequantize_per_channel]; -"2706 linear_130" [id=2706, type=linear]; -"2707 reshape_95" [id=2707, type=reshape]; -"2708 permute_97" [id=2708, type=permute]; -"2709 select_63" [id=2709, type=select]; -"2710 select_64" [id=2710, type=select]; -"2711 select_65" [id=2711, type=select]; -"2712 linalg_vector_norm_42" [id=2712, type=linalg_vector_norm]; -"2713 clamp_min_42" [id=2713, type=clamp_min]; -"2714 expand_as_42" [id=2714, type=expand_as]; -"2715 div_42" [id=2715, type=div]; -"2716 quantize_per_tensor_default_130" [id=2716, type=quantize_per_tensor]; -"2717 dequantize_per_tensor_default_130" [id=2717, type=dequantize_per_tensor]; -"2718 linalg_vector_norm_43" [id=2718, type=linalg_vector_norm]; -"2719 clamp_min_43" [id=2719, type=clamp_min]; -"2720 expand_as_43" [id=2720, type=expand_as]; -"2721 div_43" [id=2721, type=div]; -"2722 quantize_per_tensor_default_131" [id=2722, type=quantize_per_tensor]; -"2723 dequantize_per_tensor_default_131" [id=2723, type=dequantize_per_tensor]; -"2724 transpose_42" [id=2724, type=transpose]; -"2725 matmul_42" [id=2725, type=matmul]; -"2726 clamp_21" [id=2726, type=clamp]; -"2727 exp_21" [id=2727, type=exp]; -"2728 mul_43" [id=2728, type=mul]; -"2729 add_73" [id=2729, type=add]; -"2730 new_zeros_10" [id=2730, type=new_zeros]; -"2731 view_117" [id=2731, type=view]; -"2732 permute_98" [id=2732, type=permute]; -"2733 reshape_96" [id=2733, type=reshape]; -"2734 unsqueeze_62" [id=2734, type=unsqueeze]; -"2735 unsqueeze_63" [id=2735, type=unsqueeze]; -"2736 sub_10" [id=2736, type=sub]; -"2737 ne_10" [id=2737, type=ne]; -"2738 masked_fill_20" [id=2738, type=masked_fill]; -"2739 eq_10" [id=2739, type=eq]; -"2740 masked_fill_21" [id=2740, type=masked_fill]; -"2741 view_118" [id=2741, type=view]; -"2742 unsqueeze_64" [id=2742, type=unsqueeze]; -"2743 unsqueeze_65" [id=2743, type=unsqueeze]; -"2744 add_74" [id=2744, type=add]; -"2745 view_119" [id=2745, type=view]; -"2746 softmax_21" [id=2746, type=softmax]; -"2747 dropout_84" [id=2747, type=dropout]; -"2748 matmul_43" [id=2748, type=matmul]; -"2749 transpose_43" [id=2749, type=transpose]; -"2750 reshape_97" [id=2750, type=reshape]; -"2751 reshape_97_0_0_nncf_smooth_quant_0" [id=2751, type=call_module]; -"2752 quantize_per_tensor_default_132" [id=2752, type=quantize_per_tensor]; -"2753 dequantize_per_tensor_default_132" [id=2753, type=dequantize_per_tensor]; -"2754 linear_131_scale_0" [id=2754, type=get_attr]; -"2755 linear_131_zero_point_0" [id=2755, type=get_attr]; -"2756 _frozen_param132" [id=2756, type=get_attr]; -"2757 dequantize_per_channel_default_132" [id=2757, type=dequantize_per_channel]; -"2758 linear_131" [id=2758, type=linear]; -"2759 dropout_85" [id=2759, type=dropout]; -"2760 view_120" [id=2760, type=view]; -"2761 permute_99" [id=2761, type=permute]; -"2762 reshape_98" [id=2762, type=reshape]; -"2763 roll_21" [id=2763, type=roll]; -"2764 slice_325" [id=2764, type=slice]; -"2765 slice_326" [id=2765, type=slice]; -"2766 slice_327" [id=2766, type=slice]; -"2767 slice_328" [id=2767, type=slice]; -"2768 contiguous_41" [id=2768, type=contiguous]; -"2769 layer_norm_45" [id=2769, type=layer_norm]; -"2770 add_75" [id=2770, type=add]; -"2771 add_75_0_0_nncf_smooth_quant_0" [id=2771, type=call_module]; -"2772 quantize_per_tensor_default_133" [id=2772, type=quantize_per_tensor]; -"2773 dequantize_per_tensor_default_133" [id=2773, type=dequantize_per_tensor]; -"2774 linear_132_scale_0" [id=2774, type=get_attr]; -"2775 linear_132_zero_point_0" [id=2775, type=get_attr]; -"2776 _frozen_param133" [id=2776, type=get_attr]; -"2777 dequantize_per_channel_default_133" [id=2777, type=dequantize_per_channel]; -"2778 linear_132" [id=2778, type=linear]; -"2779 gelu_21" [id=2779, type=gelu]; -"2780 dropout_86" [id=2780, type=dropout]; -"2781 dropout_86_0_0_nncf_smooth_quant_0" [id=2781, type=call_module]; -"2782 quantize_per_tensor_default_134" [id=2782, type=quantize_per_tensor]; -"2783 dequantize_per_tensor_default_134" [id=2783, type=dequantize_per_tensor]; -"2784 linear_133_scale_0" [id=2784, type=get_attr]; -"2785 linear_133_zero_point_0" [id=2785, type=get_attr]; -"2786 _frozen_param134" [id=2786, type=get_attr]; -"2787 dequantize_per_channel_default_134" [id=2787, type=dequantize_per_channel]; -"2788 linear_133" [id=2788, type=linear]; -"2789 dropout_87" [id=2789, type=dropout]; -"2790 layer_norm_46" [id=2790, type=layer_norm]; -"2791 add_76" [id=2791, type=add]; -"2792 pad_24" [id=2792, type=pad]; -"2793 slice_329" [id=2793, type=slice]; -"2794 slice_330" [id=2794, type=slice]; -"2795 slice_331" [id=2795, type=slice]; -"2796 slice_332" [id=2796, type=slice]; -"2797 slice_333" [id=2797, type=slice]; -"2798 slice_334" [id=2798, type=slice]; -"2799 slice_335" [id=2799, type=slice]; -"2800 slice_336" [id=2800, type=slice]; -"2801 slice_337" [id=2801, type=slice]; -"2802 slice_338" [id=2802, type=slice]; -"2803 slice_339" [id=2803, type=slice]; -"2804 slice_340" [id=2804, type=slice]; -"2805 cat_2" [id=2805, type=cat]; -"2806 cat_2_0_0_nncf_smooth_quant_0" [id=2806, type=call_module]; -"2807 quantize_per_tensor_default_135" [id=2807, type=quantize_per_tensor]; -"2808 dequantize_per_tensor_default_135" [id=2808, type=dequantize_per_tensor]; -"2809 linear_134_scale_0" [id=2809, type=get_attr]; -"2810 linear_134_zero_point_0" [id=2810, type=get_attr]; -"2811 _frozen_param135" [id=2811, type=get_attr]; -"2812 dequantize_per_channel_default_135" [id=2812, type=dequantize_per_channel]; -"2813 linear_134" [id=2813, type=linear]; -"2814 layer_norm_47" [id=2814, type=layer_norm]; -"2815 features_7_0_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=2815, type=call_module]; -"2816 linear_135_scale_0" [id=2816, type=get_attr]; -"2817 linear_135_zero_point_0" [id=2817, type=get_attr]; -"2818 _frozen_param136" [id=2818, type=get_attr]; -"2819 dequantize_per_channel_default_136" [id=2819, type=dequantize_per_channel]; -"2820 linear_135" [id=2820, type=linear]; -"2821 relu__22" [id=2821, type=relu_]; -"2822 relu__22_0_0_nncf_smooth_quant_0" [id=2822, type=call_module]; -"2823 linear_136_scale_0" [id=2823, type=get_attr]; -"2824 linear_136_zero_point_0" [id=2824, type=get_attr]; -"2825 _frozen_param137" [id=2825, type=get_attr]; -"2826 dequantize_per_channel_default_137" [id=2826, type=dequantize_per_channel]; -"2827 linear_136" [id=2827, type=linear]; -"2828 view_121" [id=2828, type=view]; -"2829 index_22" [id=2829, type=index]; -"2830 view_122" [id=2830, type=view]; -"2831 permute_100" [id=2831, type=permute]; -"2832 contiguous_42" [id=2832, type=contiguous]; -"2833 unsqueeze_66" [id=2833, type=unsqueeze]; -"2834 sigmoid_22" [id=2834, type=sigmoid]; -"2835 mul_44" [id=2835, type=mul]; -"2836 pad_25" [id=2836, type=pad]; -"2837 view_123" [id=2837, type=view]; -"2838 permute_101" [id=2838, type=permute]; -"2839 reshape_99" [id=2839, type=reshape]; -"2840 clone_22" [id=2840, type=clone]; -"2841 reshape_99_0_0_nncf_smooth_quant_0" [id=2841, type=call_module]; -"2842 quantize_per_tensor_default_136" [id=2842, type=quantize_per_tensor]; -"2843 dequantize_per_tensor_default_136" [id=2843, type=dequantize_per_tensor]; -"2844 linear_137_scale_0" [id=2844, type=get_attr]; -"2845 linear_137_zero_point_0" [id=2845, type=get_attr]; -"2846 _frozen_param138" [id=2846, type=get_attr]; -"2847 dequantize_per_channel_default_138" [id=2847, type=dequantize_per_channel]; -"2848 linear_137" [id=2848, type=linear]; -"2849 reshape_100" [id=2849, type=reshape]; -"2850 permute_102" [id=2850, type=permute]; -"2851 select_66" [id=2851, type=select]; -"2852 select_67" [id=2852, type=select]; -"2853 select_68" [id=2853, type=select]; -"2854 linalg_vector_norm_44" [id=2854, type=linalg_vector_norm]; -"2855 clamp_min_44" [id=2855, type=clamp_min]; -"2856 expand_as_44" [id=2856, type=expand_as]; -"2857 div_44" [id=2857, type=div]; -"2858 quantize_per_tensor_default_137" [id=2858, type=quantize_per_tensor]; -"2859 dequantize_per_tensor_default_137" [id=2859, type=dequantize_per_tensor]; -"2860 linalg_vector_norm_45" [id=2860, type=linalg_vector_norm]; -"2861 clamp_min_45" [id=2861, type=clamp_min]; -"2862 expand_as_45" [id=2862, type=expand_as]; -"2863 div_45" [id=2863, type=div]; -"2864 quantize_per_tensor_default_138" [id=2864, type=quantize_per_tensor]; -"2865 dequantize_per_tensor_default_138" [id=2865, type=dequantize_per_tensor]; -"2866 transpose_44" [id=2866, type=transpose]; -"2867 matmul_44" [id=2867, type=matmul]; -"2868 clamp_22" [id=2868, type=clamp]; -"2869 exp_22" [id=2869, type=exp]; -"2870 mul_45" [id=2870, type=mul]; -"2871 add_77" [id=2871, type=add]; -"2872 softmax_22" [id=2872, type=softmax]; -"2873 dropout_88" [id=2873, type=dropout]; -"2874 matmul_45" [id=2874, type=matmul]; -"2875 transpose_45" [id=2875, type=transpose]; -"2876 reshape_101" [id=2876, type=reshape]; -"2877 reshape_101_0_0_nncf_smooth_quant_0" [id=2877, type=call_module]; -"2878 quantize_per_tensor_default_139" [id=2878, type=quantize_per_tensor]; -"2879 dequantize_per_tensor_default_139" [id=2879, type=dequantize_per_tensor]; -"2880 linear_138_scale_0" [id=2880, type=get_attr]; -"2881 linear_138_zero_point_0" [id=2881, type=get_attr]; -"2882 _frozen_param139" [id=2882, type=get_attr]; -"2883 dequantize_per_channel_default_139" [id=2883, type=dequantize_per_channel]; -"2884 linear_138" [id=2884, type=linear]; -"2885 dropout_89" [id=2885, type=dropout]; -"2886 view_124" [id=2886, type=view]; -"2887 permute_103" [id=2887, type=permute]; -"2888 reshape_102" [id=2888, type=reshape]; -"2889 slice_342" [id=2889, type=slice]; -"2890 slice_343" [id=2890, type=slice]; -"2891 slice_344" [id=2891, type=slice]; -"2892 slice_345" [id=2892, type=slice]; -"2893 contiguous_43" [id=2893, type=contiguous]; -"2894 layer_norm_48" [id=2894, type=layer_norm]; -"2895 add_78" [id=2895, type=add]; -"2896 add_78_0_0_nncf_smooth_quant_0" [id=2896, type=call_module]; -"2897 quantize_per_tensor_default_140" [id=2897, type=quantize_per_tensor]; -"2898 dequantize_per_tensor_default_140" [id=2898, type=dequantize_per_tensor]; -"2899 linear_139_scale_0" [id=2899, type=get_attr]; -"2900 linear_139_zero_point_0" [id=2900, type=get_attr]; -"2901 _frozen_param140" [id=2901, type=get_attr]; -"2902 dequantize_per_channel_default_140" [id=2902, type=dequantize_per_channel]; -"2903 linear_139" [id=2903, type=linear]; -"2904 gelu_22" [id=2904, type=gelu]; -"2905 dropout_90" [id=2905, type=dropout]; -"2906 dropout_90_0_0_nncf_smooth_quant_0" [id=2906, type=call_module]; -"2907 quantize_per_tensor_default_141" [id=2907, type=quantize_per_tensor]; -"2908 dequantize_per_tensor_default_141" [id=2908, type=dequantize_per_tensor]; -"2909 linear_140_scale_0" [id=2909, type=get_attr]; -"2910 linear_140_zero_point_0" [id=2910, type=get_attr]; -"2911 _frozen_param141" [id=2911, type=get_attr]; -"2912 dequantize_per_channel_default_141" [id=2912, type=dequantize_per_channel]; -"2913 linear_140" [id=2913, type=linear]; -"2914 dropout_91" [id=2914, type=dropout]; -"2915 layer_norm_49" [id=2915, type=layer_norm]; -"2916 add_79" [id=2916, type=add]; -"2917 features_7_1_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [id=2917, type=call_module]; -"2918 linear_141_scale_0" [id=2918, type=get_attr]; -"2919 linear_141_zero_point_0" [id=2919, type=get_attr]; -"2920 _frozen_param142" [id=2920, type=get_attr]; -"2921 dequantize_per_channel_default_142" [id=2921, type=dequantize_per_channel]; -"2922 linear_141" [id=2922, type=linear]; -"2923 relu__23" [id=2923, type=relu_]; -"2924 relu__23_0_0_nncf_smooth_quant_0" [id=2924, type=call_module]; -"2925 linear_142_scale_0" [id=2925, type=get_attr]; -"2926 linear_142_zero_point_0" [id=2926, type=get_attr]; -"2927 _frozen_param143" [id=2927, type=get_attr]; -"2928 dequantize_per_channel_default_143" [id=2928, type=dequantize_per_channel]; -"2929 linear_142" [id=2929, type=linear]; -"2930 view_125" [id=2930, type=view]; -"2931 index_23" [id=2931, type=index]; -"2932 view_126" [id=2932, type=view]; -"2933 permute_104" [id=2933, type=permute]; -"2934 contiguous_44" [id=2934, type=contiguous]; -"2935 unsqueeze_67" [id=2935, type=unsqueeze]; -"2936 sigmoid_23" [id=2936, type=sigmoid]; -"2937 mul_46" [id=2937, type=mul]; -"2938 pad_26" [id=2938, type=pad]; -"2939 view_127" [id=2939, type=view]; -"2940 permute_105" [id=2940, type=permute]; -"2941 reshape_103" [id=2941, type=reshape]; -"2942 clone_23" [id=2942, type=clone]; -"2943 reshape_103_0_0_nncf_smooth_quant_0" [id=2943, type=call_module]; -"2944 quantize_per_tensor_default_142" [id=2944, type=quantize_per_tensor]; -"2945 dequantize_per_tensor_default_142" [id=2945, type=dequantize_per_tensor]; -"2946 linear_143_scale_0" [id=2946, type=get_attr]; -"2947 linear_143_zero_point_0" [id=2947, type=get_attr]; -"2948 _frozen_param144" [id=2948, type=get_attr]; -"2949 dequantize_per_channel_default_144" [id=2949, type=dequantize_per_channel]; -"2950 linear_143" [id=2950, type=linear]; -"2951 reshape_104" [id=2951, type=reshape]; -"2952 permute_106" [id=2952, type=permute]; -"2953 select_69" [id=2953, type=select]; -"2954 select_70" [id=2954, type=select]; -"2955 select_71" [id=2955, type=select]; -"2956 linalg_vector_norm_46" [id=2956, type=linalg_vector_norm]; -"2957 clamp_min_46" [id=2957, type=clamp_min]; -"2958 expand_as_46" [id=2958, type=expand_as]; -"2959 div_46" [id=2959, type=div]; -"2960 quantize_per_tensor_default_143" [id=2960, type=quantize_per_tensor]; -"2961 dequantize_per_tensor_default_143" [id=2961, type=dequantize_per_tensor]; -"2962 linalg_vector_norm_47" [id=2962, type=linalg_vector_norm]; -"2963 clamp_min_47" [id=2963, type=clamp_min]; -"2964 expand_as_47" [id=2964, type=expand_as]; -"2965 div_47" [id=2965, type=div]; -"2966 quantize_per_tensor_default_144" [id=2966, type=quantize_per_tensor]; -"2967 dequantize_per_tensor_default_144" [id=2967, type=dequantize_per_tensor]; -"2968 transpose_46" [id=2968, type=transpose]; -"2969 matmul_46" [id=2969, type=matmul]; -"2970 clamp_23" [id=2970, type=clamp]; -"2971 exp_23" [id=2971, type=exp]; -"2972 mul_47" [id=2972, type=mul]; -"2973 add_80" [id=2973, type=add]; -"2974 softmax_23" [id=2974, type=softmax]; -"2975 dropout_92" [id=2975, type=dropout]; -"2976 matmul_47" [id=2976, type=matmul]; -"2977 transpose_47" [id=2977, type=transpose]; -"2978 reshape_105" [id=2978, type=reshape]; -"2979 reshape_105_0_0_nncf_smooth_quant_0" [id=2979, type=call_module]; -"2980 quantize_per_tensor_default_145" [id=2980, type=quantize_per_tensor]; -"2981 dequantize_per_tensor_default_145" [id=2981, type=dequantize_per_tensor]; -"2982 linear_144_scale_0" [id=2982, type=get_attr]; -"2983 linear_144_zero_point_0" [id=2983, type=get_attr]; -"2984 _frozen_param145" [id=2984, type=get_attr]; -"2985 dequantize_per_channel_default_145" [id=2985, type=dequantize_per_channel]; -"2986 linear_144" [id=2986, type=linear]; -"2987 dropout_93" [id=2987, type=dropout]; -"2988 view_128" [id=2988, type=view]; -"2989 permute_107" [id=2989, type=permute]; -"2990 reshape_106" [id=2990, type=reshape]; -"2991 slice_347" [id=2991, type=slice]; -"2992 slice_348" [id=2992, type=slice]; -"2993 slice_349" [id=2993, type=slice]; -"2994 slice_350" [id=2994, type=slice]; -"2995 contiguous_45" [id=2995, type=contiguous]; -"2996 layer_norm_50" [id=2996, type=layer_norm]; -"2997 add_81" [id=2997, type=add]; -"2998 add_81_0_0_nncf_smooth_quant_0" [id=2998, type=call_module]; -"2999 quantize_per_tensor_default_146" [id=2999, type=quantize_per_tensor]; -"3000 dequantize_per_tensor_default_146" [id=3000, type=dequantize_per_tensor]; -"3001 linear_145_scale_0" [id=3001, type=get_attr]; -"3002 linear_145_zero_point_0" [id=3002, type=get_attr]; -"3003 _frozen_param146" [id=3003, type=get_attr]; -"3004 dequantize_per_channel_default_146" [id=3004, type=dequantize_per_channel]; -"3005 linear_145" [id=3005, type=linear]; -"3006 gelu_23" [id=3006, type=gelu]; -"3007 dropout_94" [id=3007, type=dropout]; -"3008 dropout_94_0_0_nncf_smooth_quant_0" [id=3008, type=call_module]; -"3009 quantize_per_tensor_default_147" [id=3009, type=quantize_per_tensor]; -"3010 dequantize_per_tensor_default_147" [id=3010, type=dequantize_per_tensor]; -"3011 linear_146_scale_0" [id=3011, type=get_attr]; -"3012 linear_146_zero_point_0" [id=3012, type=get_attr]; -"3013 _frozen_param147" [id=3013, type=get_attr]; -"3014 dequantize_per_channel_default_147" [id=3014, type=dequantize_per_channel]; -"3015 linear_146" [id=3015, type=linear]; -"3016 dropout_95" [id=3016, type=dropout]; -"3017 layer_norm_51" [id=3017, type=layer_norm]; -"3018 add_82" [id=3018, type=add]; -"3019 layer_norm_52" [id=3019, type=layer_norm]; -"3020 permute_108" [id=3020, type=permute]; -"3021 adaptive_avg_pool2d" [id=3021, type=adaptive_avg_pool2d]; -"3022 flatten" [id=3022, type=flatten]; -"3023 flatten_0_0_nncf_smooth_quant_0" [id=3023, type=call_module]; -"3024 quantize_per_tensor_default_148" [id=3024, type=quantize_per_tensor]; -"3025 dequantize_per_tensor_default_148" [id=3025, type=dequantize_per_tensor]; -"3026 linear_147_scale_0" [id=3026, type=get_attr]; -"3027 linear_147_zero_point_0" [id=3027, type=get_attr]; -"3028 _frozen_param148" [id=3028, type=get_attr]; -"3029 dequantize_per_channel_default_148" [id=3029, type=dequantize_per_channel]; -"3030 linear_147" [id=3030, type=linear]; -"3031 output" [id=3031, type=output]; -"0 features_1_0_attn_qkv_bias" -> "335 clone" [label="(288,)", style=solid]; -"1 features_1_0_attn_proj_bias" -> "379 linear_3" [label="(96,)", style=solid]; -"2 features_1_0_attn_logit_scale" -> "363 clamp" [label="(3, 1, 1)", style=solid]; -"3 features_1_1_attn_qkv_bias" -> "435 clone_1" [label="(288,)", style=solid]; -"4 features_1_1_attn_proj_bias" -> "495 linear_9" [label="(96,)", style=solid]; -"5 features_1_1_attn_logit_scale" -> "463 clamp_1" [label="(3, 1, 1)", style=solid]; -"6 features_3_0_attn_qkv_bias" -> "574 clone_2" [label="(576,)", style=solid]; -"7 features_3_0_attn_proj_bias" -> "618 linear_16" [label="(192,)", style=solid]; -"8 features_3_0_attn_logit_scale" -> "602 clamp_2" [label="(6, 1, 1)", style=solid]; -"9 features_3_1_attn_qkv_bias" -> "677 clone_3" [label="(576,)", style=solid]; -"10 features_3_1_attn_proj_bias" -> "737 linear_22" [label="(192,)", style=solid]; -"11 features_3_1_attn_logit_scale" -> "705 clamp_3" [label="(6, 1, 1)", style=solid]; -"12 features_5_0_attn_qkv_bias" -> "819 clone_4" [label="(1152,)", style=solid]; -"13 features_5_0_attn_proj_bias" -> "863 linear_29" [label="(384,)", style=solid]; -"14 features_5_0_attn_logit_scale" -> "847 clamp_4" [label="(12, 1, 1)", style=solid]; -"15 features_5_1_attn_qkv_bias" -> "922 clone_5" [label="(1152,)", style=solid]; -"16 features_5_1_attn_proj_bias" -> "982 linear_35" [label="(384,)", style=solid]; -"17 features_5_1_attn_logit_scale" -> "950 clamp_5" [label="(12, 1, 1)", style=solid]; -"18 features_5_2_attn_qkv_bias" -> "1041 clone_6" [label="(1152,)", style=solid]; -"19 features_5_2_attn_proj_bias" -> "1085 linear_41" [label="(384,)", style=solid]; -"20 features_5_2_attn_logit_scale" -> "1069 clamp_6" [label="(12, 1, 1)", style=solid]; -"21 features_5_3_attn_qkv_bias" -> "1144 clone_7" [label="(1152,)", style=solid]; -"22 features_5_3_attn_proj_bias" -> "1204 linear_47" [label="(384,)", style=solid]; -"23 features_5_3_attn_logit_scale" -> "1172 clamp_7" [label="(12, 1, 1)", style=solid]; -"24 features_5_4_attn_qkv_bias" -> "1263 clone_8" [label="(1152,)", style=solid]; -"25 features_5_4_attn_proj_bias" -> "1307 linear_53" [label="(384,)", style=solid]; -"26 features_5_4_attn_logit_scale" -> "1291 clamp_8" [label="(12, 1, 1)", style=solid]; -"27 features_5_5_attn_qkv_bias" -> "1366 clone_9" [label="(1152,)", style=solid]; -"28 features_5_5_attn_proj_bias" -> "1426 linear_59" [label="(384,)", style=solid]; -"29 features_5_5_attn_logit_scale" -> "1394 clamp_9" [label="(12, 1, 1)", style=solid]; -"30 features_5_6_attn_qkv_bias" -> "1485 clone_10" [label="(1152,)", style=solid]; -"31 features_5_6_attn_proj_bias" -> "1529 linear_65" [label="(384,)", style=solid]; -"32 features_5_6_attn_logit_scale" -> "1513 clamp_10" [label="(12, 1, 1)", style=solid]; -"33 features_5_7_attn_qkv_bias" -> "1588 clone_11" [label="(1152,)", style=solid]; -"34 features_5_7_attn_proj_bias" -> "1648 linear_71" [label="(384,)", style=solid]; -"35 features_5_7_attn_logit_scale" -> "1616 clamp_11" [label="(12, 1, 1)", style=solid]; -"36 features_5_8_attn_qkv_bias" -> "1707 clone_12" [label="(1152,)", style=solid]; -"37 features_5_8_attn_proj_bias" -> "1751 linear_77" [label="(384,)", style=solid]; -"38 features_5_8_attn_logit_scale" -> "1735 clamp_12" [label="(12, 1, 1)", style=solid]; -"39 features_5_9_attn_qkv_bias" -> "1810 clone_13" [label="(1152,)", style=solid]; -"40 features_5_9_attn_proj_bias" -> "1870 linear_83" [label="(384,)", style=solid]; -"41 features_5_9_attn_logit_scale" -> "1838 clamp_13" [label="(12, 1, 1)", style=solid]; -"42 features_5_10_attn_qkv_bias" -> "1929 clone_14" [label="(1152,)", style=solid]; -"43 features_5_10_attn_proj_bias" -> "1973 linear_89" [label="(384,)", style=solid]; -"44 features_5_10_attn_logit_scale" -> "1957 clamp_14" [label="(12, 1, 1)", style=solid]; -"45 features_5_11_attn_qkv_bias" -> "2032 clone_15" [label="(1152,)", style=solid]; -"46 features_5_11_attn_proj_bias" -> "2092 linear_95" [label="(384,)", style=solid]; -"47 features_5_11_attn_logit_scale" -> "2060 clamp_15" [label="(12, 1, 1)", style=solid]; -"48 features_5_12_attn_qkv_bias" -> "2151 clone_16" [label="(1152,)", style=solid]; -"49 features_5_12_attn_proj_bias" -> "2195 linear_101" [label="(384,)", style=solid]; -"50 features_5_12_attn_logit_scale" -> "2179 clamp_16" [label="(12, 1, 1)", style=solid]; -"51 features_5_13_attn_qkv_bias" -> "2254 clone_17" [label="(1152,)", style=solid]; -"52 features_5_13_attn_proj_bias" -> "2314 linear_107" [label="(384,)", style=solid]; -"53 features_5_13_attn_logit_scale" -> "2282 clamp_17" [label="(12, 1, 1)", style=solid]; -"54 features_5_14_attn_qkv_bias" -> "2373 clone_18" [label="(1152,)", style=solid]; -"55 features_5_14_attn_proj_bias" -> "2417 linear_113" [label="(384,)", style=solid]; -"56 features_5_14_attn_logit_scale" -> "2401 clamp_18" [label="(12, 1, 1)", style=solid]; -"57 features_5_15_attn_qkv_bias" -> "2476 clone_19" [label="(1152,)", style=solid]; -"58 features_5_15_attn_proj_bias" -> "2536 linear_119" [label="(384,)", style=solid]; -"59 features_5_15_attn_logit_scale" -> "2504 clamp_19" [label="(12, 1, 1)", style=solid]; -"60 features_5_16_attn_qkv_bias" -> "2595 clone_20" [label="(1152,)", style=solid]; -"61 features_5_16_attn_proj_bias" -> "2639 linear_125" [label="(384,)", style=solid]; -"62 features_5_16_attn_logit_scale" -> "2623 clamp_20" [label="(12, 1, 1)", style=solid]; -"63 features_5_17_attn_qkv_bias" -> "2698 clone_21" [label="(1152,)", style=solid]; -"64 features_5_17_attn_proj_bias" -> "2758 linear_131" [label="(384,)", style=solid]; -"65 features_5_17_attn_logit_scale" -> "2726 clamp_21" [label="(12, 1, 1)", style=solid]; -"66 features_7_0_attn_qkv_bias" -> "2840 clone_22" [label="(2304,)", style=solid]; -"67 features_7_0_attn_proj_bias" -> "2884 linear_138" [label="(768,)", style=solid]; -"68 features_7_0_attn_logit_scale" -> "2868 clamp_22" [label="(24, 1, 1)", style=solid]; -"69 features_7_1_attn_qkv_bias" -> "2942 clone_23" [label="(2304,)", style=solid]; -"70 features_7_1_attn_proj_bias" -> "2986 linear_144" [label="(768,)", style=solid]; -"71 features_7_1_attn_logit_scale" -> "2970 clamp_23" [label="(24, 1, 1)", style=solid]; -"72 features_0_0_bias" -> "307 conv2d" [label="(96,)", style=solid]; -"73 features_0_2_weight" -> "309 layer_norm" [label="(96,)", style=solid]; -"74 features_0_2_bias" -> "309 layer_norm" [label="(96,)", style=solid]; -"75 features_1_0_attn_cpb_mlp_0_bias" -> "315 linear" [label="(512,)", style=solid]; -"76 features_1_0_norm1_weight" -> "386 layer_norm_1" [label="(96,)", style=solid]; -"77 features_1_0_norm1_bias" -> "386 layer_norm_1" [label="(96,)", style=solid]; -"78 features_1_0_mlp_0_bias" -> "395 linear_4" [label="(384,)", style=solid]; -"79 features_1_0_mlp_3_bias" -> "405 linear_5" [label="(96,)", style=solid]; -"80 features_1_0_norm2_weight" -> "407 layer_norm_2" [label="(96,)", style=solid]; -"81 features_1_0_norm2_bias" -> "407 layer_norm_2" [label="(96,)", style=solid]; -"82 features_1_1_attn_cpb_mlp_0_bias" -> "414 linear_6" [label="(512,)", style=solid]; -"83 features_1_1_norm1_weight" -> "503 layer_norm_3" [label="(96,)", style=solid]; -"84 features_1_1_norm1_bias" -> "503 layer_norm_3" [label="(96,)", style=solid]; -"85 features_1_1_mlp_0_bias" -> "512 linear_10" [label="(384,)", style=solid]; -"86 features_1_1_mlp_3_bias" -> "522 linear_11" [label="(96,)", style=solid]; -"87 features_1_1_norm2_weight" -> "524 layer_norm_4" [label="(96,)", style=solid]; -"88 features_1_1_norm2_bias" -> "524 layer_norm_4" [label="(96,)", style=solid]; -"89 features_2_norm_weight" -> "548 layer_norm_5" [label="(192,)", style=solid]; -"90 features_2_norm_bias" -> "548 layer_norm_5" [label="(192,)", style=solid]; -"91 features_3_0_attn_cpb_mlp_0_bias" -> "554 linear_13" [label="(512,)", style=solid]; -"92 features_3_0_norm1_weight" -> "628 layer_norm_6" [label="(192,)", style=solid]; -"93 features_3_0_norm1_bias" -> "628 layer_norm_6" [label="(192,)", style=solid]; -"94 features_3_0_mlp_0_bias" -> "637 linear_17" [label="(768,)", style=solid]; -"95 features_3_0_mlp_3_bias" -> "647 linear_18" [label="(192,)", style=solid]; -"96 features_3_0_norm2_weight" -> "649 layer_norm_7" [label="(192,)", style=solid]; -"97 features_3_0_norm2_bias" -> "649 layer_norm_7" [label="(192,)", style=solid]; -"98 features_3_1_attn_cpb_mlp_0_bias" -> "656 linear_19" [label="(512,)", style=solid]; -"99 features_3_1_norm1_weight" -> "748 layer_norm_8" [label="(192,)", style=solid]; -"100 features_3_1_norm1_bias" -> "748 layer_norm_8" [label="(192,)", style=solid]; -"101 features_3_1_mlp_0_bias" -> "757 linear_23" [label="(768,)", style=solid]; -"102 features_3_1_mlp_3_bias" -> "767 linear_24" [label="(192,)", style=solid]; -"103 features_3_1_norm2_weight" -> "769 layer_norm_9" [label="(192,)", style=solid]; -"104 features_3_1_norm2_bias" -> "769 layer_norm_9" [label="(192,)", style=solid]; -"105 features_4_norm_weight" -> "793 layer_norm_10" [label="(384,)", style=solid]; -"106 features_4_norm_bias" -> "793 layer_norm_10" [label="(384,)", style=solid]; -"107 features_5_0_attn_cpb_mlp_0_bias" -> "799 linear_26" [label="(512,)", style=solid]; -"108 features_5_0_norm1_weight" -> "873 layer_norm_11" [label="(384,)", style=solid]; -"109 features_5_0_norm1_bias" -> "873 layer_norm_11" [label="(384,)", style=solid]; -"110 features_5_0_mlp_0_bias" -> "882 linear_30" [label="(1536,)", style=solid]; -"111 features_5_0_mlp_3_bias" -> "892 linear_31" [label="(384,)", style=solid]; -"112 features_5_0_norm2_weight" -> "894 layer_norm_12" [label="(384,)", style=solid]; -"113 features_5_0_norm2_bias" -> "894 layer_norm_12" [label="(384,)", style=solid]; -"114 features_5_1_attn_cpb_mlp_0_bias" -> "901 linear_32" [label="(512,)", style=solid]; -"115 features_5_1_norm1_weight" -> "993 layer_norm_13" [label="(384,)", style=solid]; -"116 features_5_1_norm1_bias" -> "993 layer_norm_13" [label="(384,)", style=solid]; -"117 features_5_1_mlp_0_bias" -> "1002 linear_36" [label="(1536,)", style=solid]; -"118 features_5_1_mlp_3_bias" -> "1012 linear_37" [label="(384,)", style=solid]; -"119 features_5_1_norm2_weight" -> "1014 layer_norm_14" [label="(384,)", style=solid]; -"120 features_5_1_norm2_bias" -> "1014 layer_norm_14" [label="(384,)", style=solid]; -"121 features_5_2_attn_cpb_mlp_0_bias" -> "1021 linear_38" [label="(512,)", style=solid]; -"122 features_5_2_norm1_weight" -> "1095 layer_norm_15" [label="(384,)", style=solid]; -"123 features_5_2_norm1_bias" -> "1095 layer_norm_15" [label="(384,)", style=solid]; -"124 features_5_2_mlp_0_bias" -> "1104 linear_42" [label="(1536,)", style=solid]; -"125 features_5_2_mlp_3_bias" -> "1114 linear_43" [label="(384,)", style=solid]; -"126 features_5_2_norm2_weight" -> "1116 layer_norm_16" [label="(384,)", style=solid]; -"127 features_5_2_norm2_bias" -> "1116 layer_norm_16" [label="(384,)", style=solid]; -"128 features_5_3_attn_cpb_mlp_0_bias" -> "1123 linear_44" [label="(512,)", style=solid]; -"129 features_5_3_norm1_weight" -> "1215 layer_norm_17" [label="(384,)", style=solid]; -"130 features_5_3_norm1_bias" -> "1215 layer_norm_17" [label="(384,)", style=solid]; -"131 features_5_3_mlp_0_bias" -> "1224 linear_48" [label="(1536,)", style=solid]; -"132 features_5_3_mlp_3_bias" -> "1234 linear_49" [label="(384,)", style=solid]; -"133 features_5_3_norm2_weight" -> "1236 layer_norm_18" [label="(384,)", style=solid]; -"134 features_5_3_norm2_bias" -> "1236 layer_norm_18" [label="(384,)", style=solid]; -"135 features_5_4_attn_cpb_mlp_0_bias" -> "1243 linear_50" [label="(512,)", style=solid]; -"136 features_5_4_norm1_weight" -> "1317 layer_norm_19" [label="(384,)", style=solid]; -"137 features_5_4_norm1_bias" -> "1317 layer_norm_19" [label="(384,)", style=solid]; -"138 features_5_4_mlp_0_bias" -> "1326 linear_54" [label="(1536,)", style=solid]; -"139 features_5_4_mlp_3_bias" -> "1336 linear_55" [label="(384,)", style=solid]; -"140 features_5_4_norm2_weight" -> "1338 layer_norm_20" [label="(384,)", style=solid]; -"141 features_5_4_norm2_bias" -> "1338 layer_norm_20" [label="(384,)", style=solid]; -"142 features_5_5_attn_cpb_mlp_0_bias" -> "1345 linear_56" [label="(512,)", style=solid]; -"143 features_5_5_norm1_weight" -> "1437 layer_norm_21" [label="(384,)", style=solid]; -"144 features_5_5_norm1_bias" -> "1437 layer_norm_21" [label="(384,)", style=solid]; -"145 features_5_5_mlp_0_bias" -> "1446 linear_60" [label="(1536,)", style=solid]; -"146 features_5_5_mlp_3_bias" -> "1456 linear_61" [label="(384,)", style=solid]; -"147 features_5_5_norm2_weight" -> "1458 layer_norm_22" [label="(384,)", style=solid]; -"148 features_5_5_norm2_bias" -> "1458 layer_norm_22" [label="(384,)", style=solid]; -"149 features_5_6_attn_cpb_mlp_0_bias" -> "1465 linear_62" [label="(512,)", style=solid]; -"150 features_5_6_norm1_weight" -> "1539 layer_norm_23" [label="(384,)", style=solid]; -"151 features_5_6_norm1_bias" -> "1539 layer_norm_23" [label="(384,)", style=solid]; -"152 features_5_6_mlp_0_bias" -> "1548 linear_66" [label="(1536,)", style=solid]; -"153 features_5_6_mlp_3_bias" -> "1558 linear_67" [label="(384,)", style=solid]; -"154 features_5_6_norm2_weight" -> "1560 layer_norm_24" [label="(384,)", style=solid]; -"155 features_5_6_norm2_bias" -> "1560 layer_norm_24" [label="(384,)", style=solid]; -"156 features_5_7_attn_cpb_mlp_0_bias" -> "1567 linear_68" [label="(512,)", style=solid]; -"157 features_5_7_norm1_weight" -> "1659 layer_norm_25" [label="(384,)", style=solid]; -"158 features_5_7_norm1_bias" -> "1659 layer_norm_25" [label="(384,)", style=solid]; -"159 features_5_7_mlp_0_bias" -> "1668 linear_72" [label="(1536,)", style=solid]; -"160 features_5_7_mlp_3_bias" -> "1678 linear_73" [label="(384,)", style=solid]; -"161 features_5_7_norm2_weight" -> "1680 layer_norm_26" [label="(384,)", style=solid]; -"162 features_5_7_norm2_bias" -> "1680 layer_norm_26" [label="(384,)", style=solid]; -"163 features_5_8_attn_cpb_mlp_0_bias" -> "1687 linear_74" [label="(512,)", style=solid]; -"164 features_5_8_norm1_weight" -> "1761 layer_norm_27" [label="(384,)", style=solid]; -"165 features_5_8_norm1_bias" -> "1761 layer_norm_27" [label="(384,)", style=solid]; -"166 features_5_8_mlp_0_bias" -> "1770 linear_78" [label="(1536,)", style=solid]; -"167 features_5_8_mlp_3_bias" -> "1780 linear_79" [label="(384,)", style=solid]; -"168 features_5_8_norm2_weight" -> "1782 layer_norm_28" [label="(384,)", style=solid]; -"169 features_5_8_norm2_bias" -> "1782 layer_norm_28" [label="(384,)", style=solid]; -"170 features_5_9_attn_cpb_mlp_0_bias" -> "1789 linear_80" [label="(512,)", style=solid]; -"171 features_5_9_norm1_weight" -> "1881 layer_norm_29" [label="(384,)", style=solid]; -"172 features_5_9_norm1_bias" -> "1881 layer_norm_29" [label="(384,)", style=solid]; -"173 features_5_9_mlp_0_bias" -> "1890 linear_84" [label="(1536,)", style=solid]; -"174 features_5_9_mlp_3_bias" -> "1900 linear_85" [label="(384,)", style=solid]; -"175 features_5_9_norm2_weight" -> "1902 layer_norm_30" [label="(384,)", style=solid]; -"176 features_5_9_norm2_bias" -> "1902 layer_norm_30" [label="(384,)", style=solid]; -"177 features_5_10_attn_cpb_mlp_0_bias" -> "1909 linear_86" [label="(512,)", style=solid]; -"178 features_5_10_norm1_weight" -> "1983 layer_norm_31" [label="(384,)", style=solid]; -"179 features_5_10_norm1_bias" -> "1983 layer_norm_31" [label="(384,)", style=solid]; -"180 features_5_10_mlp_0_bias" -> "1992 linear_90" [label="(1536,)", style=solid]; -"181 features_5_10_mlp_3_bias" -> "2002 linear_91" [label="(384,)", style=solid]; -"182 features_5_10_norm2_weight" -> "2004 layer_norm_32" [label="(384,)", style=solid]; -"183 features_5_10_norm2_bias" -> "2004 layer_norm_32" [label="(384,)", style=solid]; -"184 features_5_11_attn_cpb_mlp_0_bias" -> "2011 linear_92" [label="(512,)", style=solid]; -"185 features_5_11_norm1_weight" -> "2103 layer_norm_33" [label="(384,)", style=solid]; -"186 features_5_11_norm1_bias" -> "2103 layer_norm_33" [label="(384,)", style=solid]; -"187 features_5_11_mlp_0_bias" -> "2112 linear_96" [label="(1536,)", style=solid]; -"188 features_5_11_mlp_3_bias" -> "2122 linear_97" [label="(384,)", style=solid]; -"189 features_5_11_norm2_weight" -> "2124 layer_norm_34" [label="(384,)", style=solid]; -"190 features_5_11_norm2_bias" -> "2124 layer_norm_34" [label="(384,)", style=solid]; -"191 features_5_12_attn_cpb_mlp_0_bias" -> "2131 linear_98" [label="(512,)", style=solid]; -"192 features_5_12_norm1_weight" -> "2205 layer_norm_35" [label="(384,)", style=solid]; -"193 features_5_12_norm1_bias" -> "2205 layer_norm_35" [label="(384,)", style=solid]; -"194 features_5_12_mlp_0_bias" -> "2214 linear_102" [label="(1536,)", style=solid]; -"195 features_5_12_mlp_3_bias" -> "2224 linear_103" [label="(384,)", style=solid]; -"196 features_5_12_norm2_weight" -> "2226 layer_norm_36" [label="(384,)", style=solid]; -"197 features_5_12_norm2_bias" -> "2226 layer_norm_36" [label="(384,)", style=solid]; -"198 features_5_13_attn_cpb_mlp_0_bias" -> "2233 linear_104" [label="(512,)", style=solid]; -"199 features_5_13_norm1_weight" -> "2325 layer_norm_37" [label="(384,)", style=solid]; -"200 features_5_13_norm1_bias" -> "2325 layer_norm_37" [label="(384,)", style=solid]; -"201 features_5_13_mlp_0_bias" -> "2334 linear_108" [label="(1536,)", style=solid]; -"202 features_5_13_mlp_3_bias" -> "2344 linear_109" [label="(384,)", style=solid]; -"203 features_5_13_norm2_weight" -> "2346 layer_norm_38" [label="(384,)", style=solid]; -"204 features_5_13_norm2_bias" -> "2346 layer_norm_38" [label="(384,)", style=solid]; -"205 features_5_14_attn_cpb_mlp_0_bias" -> "2353 linear_110" [label="(512,)", style=solid]; -"206 features_5_14_norm1_weight" -> "2427 layer_norm_39" [label="(384,)", style=solid]; -"207 features_5_14_norm1_bias" -> "2427 layer_norm_39" [label="(384,)", style=solid]; -"208 features_5_14_mlp_0_bias" -> "2436 linear_114" [label="(1536,)", style=solid]; -"209 features_5_14_mlp_3_bias" -> "2446 linear_115" [label="(384,)", style=solid]; -"210 features_5_14_norm2_weight" -> "2448 layer_norm_40" [label="(384,)", style=solid]; -"211 features_5_14_norm2_bias" -> "2448 layer_norm_40" [label="(384,)", style=solid]; -"212 features_5_15_attn_cpb_mlp_0_bias" -> "2455 linear_116" [label="(512,)", style=solid]; -"213 features_5_15_norm1_weight" -> "2547 layer_norm_41" [label="(384,)", style=solid]; -"214 features_5_15_norm1_bias" -> "2547 layer_norm_41" [label="(384,)", style=solid]; -"215 features_5_15_mlp_0_bias" -> "2556 linear_120" [label="(1536,)", style=solid]; -"216 features_5_15_mlp_3_bias" -> "2566 linear_121" [label="(384,)", style=solid]; -"217 features_5_15_norm2_weight" -> "2568 layer_norm_42" [label="(384,)", style=solid]; -"218 features_5_15_norm2_bias" -> "2568 layer_norm_42" [label="(384,)", style=solid]; -"219 features_5_16_attn_cpb_mlp_0_bias" -> "2575 linear_122" [label="(512,)", style=solid]; -"220 features_5_16_norm1_weight" -> "2649 layer_norm_43" [label="(384,)", style=solid]; -"221 features_5_16_norm1_bias" -> "2649 layer_norm_43" [label="(384,)", style=solid]; -"222 features_5_16_mlp_0_bias" -> "2658 linear_126" [label="(1536,)", style=solid]; -"223 features_5_16_mlp_3_bias" -> "2668 linear_127" [label="(384,)", style=solid]; -"224 features_5_16_norm2_weight" -> "2670 layer_norm_44" [label="(384,)", style=solid]; -"225 features_5_16_norm2_bias" -> "2670 layer_norm_44" [label="(384,)", style=solid]; -"226 features_5_17_attn_cpb_mlp_0_bias" -> "2677 linear_128" [label="(512,)", style=solid]; -"227 features_5_17_norm1_weight" -> "2769 layer_norm_45" [label="(384,)", style=solid]; -"228 features_5_17_norm1_bias" -> "2769 layer_norm_45" [label="(384,)", style=solid]; -"229 features_5_17_mlp_0_bias" -> "2778 linear_132" [label="(1536,)", style=solid]; -"230 features_5_17_mlp_3_bias" -> "2788 linear_133" [label="(384,)", style=solid]; -"231 features_5_17_norm2_weight" -> "2790 layer_norm_46" [label="(384,)", style=solid]; -"232 features_5_17_norm2_bias" -> "2790 layer_norm_46" [label="(384,)", style=solid]; -"233 features_6_norm_weight" -> "2814 layer_norm_47" [label="(768,)", style=solid]; -"234 features_6_norm_bias" -> "2814 layer_norm_47" [label="(768,)", style=solid]; -"235 features_7_0_attn_cpb_mlp_0_bias" -> "2820 linear_135" [label="(512,)", style=solid]; -"236 features_7_0_norm1_weight" -> "2894 layer_norm_48" [label="(768,)", style=solid]; -"237 features_7_0_norm1_bias" -> "2894 layer_norm_48" [label="(768,)", style=solid]; -"238 features_7_0_mlp_0_bias" -> "2903 linear_139" [label="(3072,)", style=solid]; -"239 features_7_0_mlp_3_bias" -> "2913 linear_140" [label="(768,)", style=solid]; -"240 features_7_0_norm2_weight" -> "2915 layer_norm_49" [label="(768,)", style=solid]; -"241 features_7_0_norm2_bias" -> "2915 layer_norm_49" [label="(768,)", style=solid]; -"242 features_7_1_attn_cpb_mlp_0_bias" -> "2922 linear_141" [label="(512,)", style=solid]; -"243 features_7_1_norm1_weight" -> "2996 layer_norm_50" [label="(768,)", style=solid]; -"244 features_7_1_norm1_bias" -> "2996 layer_norm_50" [label="(768,)", style=solid]; -"245 features_7_1_mlp_0_bias" -> "3005 linear_145" [label="(3072,)", style=solid]; -"246 features_7_1_mlp_3_bias" -> "3015 linear_146" [label="(768,)", style=solid]; -"247 features_7_1_norm2_weight" -> "3017 layer_norm_51" [label="(768,)", style=solid]; -"248 features_7_1_norm2_bias" -> "3017 layer_norm_51" [label="(768,)", style=solid]; -"249 norm_weight" -> "3019 layer_norm_52" [label="(768,)", style=solid]; -"250 norm_bias" -> "3019 layer_norm_52" [label="(768,)", style=solid]; -"251 head_bias" -> "3030 linear_147" [label="(1000,)", style=solid]; -"252 features_1_0_attn_relative_coords_table" -> "310 features_1_0_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; -"253 features_1_0_attn_relative_position_index" -> "324 index" [label="(4096,)", style=solid]; -"254 features_1_1_attn_relative_coords_table" -> "409 features_1_1_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; -"255 features_1_1_attn_relative_position_index" -> "423 index_1" [label="(4096,)", style=solid]; -"256 features_3_0_attn_relative_coords_table" -> "549 features_3_0_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; -"257 features_3_0_attn_relative_position_index" -> "563 index_2" [label="(4096,)", style=solid]; -"258 features_3_1_attn_relative_coords_table" -> "651 features_3_1_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; -"259 features_3_1_attn_relative_position_index" -> "665 index_3" [label="(4096,)", style=solid]; -"260 features_5_0_attn_relative_coords_table" -> "794 features_5_0_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; -"261 features_5_0_attn_relative_position_index" -> "808 index_4" [label="(4096,)", style=solid]; -"262 features_5_1_attn_relative_coords_table" -> "896 features_5_1_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; -"263 features_5_1_attn_relative_position_index" -> "910 index_5" [label="(4096,)", style=solid]; -"264 features_5_2_attn_relative_coords_table" -> "1016 features_5_2_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; -"265 features_5_2_attn_relative_position_index" -> "1030 index_6" [label="(4096,)", style=solid]; -"266 features_5_3_attn_relative_coords_table" -> "1118 features_5_3_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; -"267 features_5_3_attn_relative_position_index" -> "1132 index_7" [label="(4096,)", style=solid]; -"268 features_5_4_attn_relative_coords_table" -> "1238 features_5_4_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; -"269 features_5_4_attn_relative_position_index" -> "1252 index_8" [label="(4096,)", style=solid]; -"270 features_5_5_attn_relative_coords_table" -> "1340 features_5_5_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; -"271 features_5_5_attn_relative_position_index" -> "1354 index_9" [label="(4096,)", style=solid]; -"272 features_5_6_attn_relative_coords_table" -> "1460 features_5_6_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; -"273 features_5_6_attn_relative_position_index" -> "1474 index_10" [label="(4096,)", style=solid]; -"274 features_5_7_attn_relative_coords_table" -> "1562 features_5_7_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; -"275 features_5_7_attn_relative_position_index" -> "1576 index_11" [label="(4096,)", style=solid]; -"276 features_5_8_attn_relative_coords_table" -> "1682 features_5_8_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; -"277 features_5_8_attn_relative_position_index" -> "1696 index_12" [label="(4096,)", style=solid]; -"278 features_5_9_attn_relative_coords_table" -> "1784 features_5_9_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; -"279 features_5_9_attn_relative_position_index" -> "1798 index_13" [label="(4096,)", style=solid]; -"280 features_5_10_attn_relative_coords_table" -> "1904 features_5_10_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; -"281 features_5_10_attn_relative_position_index" -> "1918 index_14" [label="(4096,)", style=solid]; -"282 features_5_11_attn_relative_coords_table" -> "2006 features_5_11_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; -"283 features_5_11_attn_relative_position_index" -> "2020 index_15" [label="(4096,)", style=solid]; -"284 features_5_12_attn_relative_coords_table" -> "2126 features_5_12_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; -"285 features_5_12_attn_relative_position_index" -> "2140 index_16" [label="(4096,)", style=solid]; -"286 features_5_13_attn_relative_coords_table" -> "2228 features_5_13_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; -"287 features_5_13_attn_relative_position_index" -> "2242 index_17" [label="(4096,)", style=solid]; -"288 features_5_14_attn_relative_coords_table" -> "2348 features_5_14_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; -"289 features_5_14_attn_relative_position_index" -> "2362 index_18" [label="(4096,)", style=solid]; -"290 features_5_15_attn_relative_coords_table" -> "2450 features_5_15_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; -"291 features_5_15_attn_relative_position_index" -> "2464 index_19" [label="(4096,)", style=solid]; -"292 features_5_16_attn_relative_coords_table" -> "2570 features_5_16_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; -"293 features_5_16_attn_relative_position_index" -> "2584 index_20" [label="(4096,)", style=solid]; -"294 features_5_17_attn_relative_coords_table" -> "2672 features_5_17_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; -"295 features_5_17_attn_relative_position_index" -> "2686 index_21" [label="(4096,)", style=solid]; -"296 features_7_0_attn_relative_coords_table" -> "2815 features_7_0_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; -"297 features_7_0_attn_relative_position_index" -> "2829 index_22" [label="(4096,)", style=solid]; -"298 features_7_1_attn_relative_coords_table" -> "2917 features_7_1_attn_relative_coords_table_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 2)", style=solid]; -"299 features_7_1_attn_relative_position_index" -> "2931 index_23" [label="(4096,)", style=solid]; -"300 x" -> "301 quantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; -"301 quantize_per_tensor_default" -> "302 dequantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; -"302 dequantize_per_tensor_default" -> "307 conv2d" [label="(1, 3, 224, 224)", style=solid]; -"303 conv2d_scale_0" -> "306 dequantize_per_channel_default" [label="(96,)", style=solid]; -"304 conv2d_zero_point_0" -> "306 dequantize_per_channel_default" [label="(96,)", style=solid]; -"305 _frozen_param0" -> "306 dequantize_per_channel_default" [label="(96, 3, 4, 4)", style=solid]; -"306 dequantize_per_channel_default" -> "307 conv2d" [label="(96, 3, 4, 4)", style=solid]; -"307 conv2d" -> "308 permute" [label="(1, 96, 56, 56)", style=solid]; -"308 permute" -> "309 layer_norm" [label="(1, 56, 56, 96)", style=solid]; -"309 layer_norm" -> "331 pad" [label="(1, 56, 56, 96)", style=solid]; -"309 layer_norm" -> "387 add_1" [label="(1, 56, 56, 96)", style=solid]; -"310 features_1_0_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "315 linear" [label="(1, 15, 15, 2)", style=solid]; -"311 linear_scale_0" -> "314 dequantize_per_channel_default_1" [label="(512,)", style=solid]; -"312 linear_zero_point_0" -> "314 dequantize_per_channel_default_1" [label="(512,)", style=solid]; -"313 _frozen_param1" -> "314 dequantize_per_channel_default_1" [label="(512, 2)", style=solid]; -"314 dequantize_per_channel_default_1" -> "315 linear" [label="(512, 2)", style=solid]; -"315 linear" -> "316 relu_" [label="(1, 15, 15, 512)", style=solid]; -"316 relu_" -> "317 relu__0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; -"317 relu__0_0_nncf_smooth_quant_0" -> "322 linear_1" [label="(1, 15, 15, 512)", style=solid]; -"318 linear_1_scale_0" -> "321 dequantize_per_channel_default_2" [label="(3,)", style=solid]; -"319 linear_1_zero_point_0" -> "321 dequantize_per_channel_default_2" [label="(3,)", style=solid]; -"320 _frozen_param2" -> "321 dequantize_per_channel_default_2" [label="(3, 512)", style=solid]; -"321 dequantize_per_channel_default_2" -> "322 linear_1" [label="(3, 512)", style=solid]; -"322 linear_1" -> "323 view" [label="(1, 15, 15, 3)", style=solid]; -"323 view" -> "324 index" [label="(225, 3)", style=solid]; -"324 index" -> "325 view_1" [label="(4096, 3)", style=solid]; -"325 view_1" -> "326 permute_1" [label="(64, 64, 3)", style=solid]; -"326 permute_1" -> "327 contiguous" [label="(3, 64, 64)", style=solid]; -"327 contiguous" -> "328 unsqueeze" [label="(3, 64, 64)", style=solid]; -"328 unsqueeze" -> "329 sigmoid" [label="(1, 3, 64, 64)", style=solid]; -"329 sigmoid" -> "330 mul" [label="(1, 3, 64, 64)", style=solid]; -"330 mul" -> "366 add" [label="(1, 3, 64, 64)", style=solid]; -"331 pad" -> "332 view_2" [label="(1, 56, 56, 96)", style=solid]; -"332 view_2" -> "333 permute_2" [label="(1, 7, 8, 7, 8, 96)", style=solid]; -"333 permute_2" -> "334 reshape" [label="(1, 7, 7, 8, 8, 96)", style=solid]; -"334 reshape" -> "336 reshape_0_0_nncf_smooth_quant_0" [label="(49, 64, 96)", style=solid]; -"335 clone" -> "343 linear_2" [label="(288,)", style=solid]; -"336 reshape_0_0_nncf_smooth_quant_0" -> "337 quantize_per_tensor_default_1" [label="(49, 64, 96)", style=solid]; -"337 quantize_per_tensor_default_1" -> "338 dequantize_per_tensor_default_1" [label="(49, 64, 96)", style=solid]; -"338 dequantize_per_tensor_default_1" -> "343 linear_2" [label="(49, 64, 96)", style=solid]; -"339 linear_2_scale_0" -> "342 dequantize_per_channel_default_3" [label="(288,)", style=solid]; -"340 linear_2_zero_point_0" -> "342 dequantize_per_channel_default_3" [label="(288,)", style=solid]; -"341 _frozen_param3" -> "342 dequantize_per_channel_default_3" [label="(288, 96)", style=solid]; -"342 dequantize_per_channel_default_3" -> "343 linear_2" [label="(288, 96)", style=solid]; -"343 linear_2" -> "344 reshape_1" [label="(49, 64, 288)", style=solid]; -"344 reshape_1" -> "345 permute_3" [label="(49, 64, 3, 3, 32)", style=solid]; -"345 permute_3" -> "346 select" [label="(3, 49, 3, 64, 32)", style=solid]; -"345 permute_3" -> "347 select_1" [label="(3, 49, 3, 64, 32)", style=solid]; -"345 permute_3" -> "348 select_2" [label="(3, 49, 3, 64, 32)", style=solid]; -"346 select" -> "349 linalg_vector_norm" [label="(49, 3, 64, 32)", style=solid]; -"346 select" -> "351 expand_as" [label="(49, 3, 64, 32)", style=solid]; -"346 select" -> "352 div" [label="(49, 3, 64, 32)", style=solid]; -"347 select_1" -> "355 linalg_vector_norm_1" [label="(49, 3, 64, 32)", style=solid]; -"347 select_1" -> "357 expand_as_1" [label="(49, 3, 64, 32)", style=solid]; -"347 select_1" -> "358 div_1" [label="(49, 3, 64, 32)", style=solid]; -"348 select_2" -> "369 matmul_1" [label="(49, 3, 64, 32)", style=solid]; -"349 linalg_vector_norm" -> "350 clamp_min" [label="(49, 3, 64, 1)", style=solid]; -"350 clamp_min" -> "351 expand_as" [label="(49, 3, 64, 1)", style=solid]; -"351 expand_as" -> "352 div" [label="(49, 3, 64, 32)", style=solid]; -"352 div" -> "353 quantize_per_tensor_default_2" [label="(49, 3, 64, 32)", style=solid]; -"353 quantize_per_tensor_default_2" -> "354 dequantize_per_tensor_default_2" [label="(49, 3, 64, 32)", style=solid]; -"354 dequantize_per_tensor_default_2" -> "362 matmul" [label="(49, 3, 64, 32)", style=solid]; -"355 linalg_vector_norm_1" -> "356 clamp_min_1" [label="(49, 3, 64, 1)", style=solid]; -"356 clamp_min_1" -> "357 expand_as_1" [label="(49, 3, 64, 1)", style=solid]; -"357 expand_as_1" -> "358 div_1" [label="(49, 3, 64, 32)", style=solid]; -"358 div_1" -> "359 quantize_per_tensor_default_3" [label="(49, 3, 64, 32)", style=solid]; -"359 quantize_per_tensor_default_3" -> "360 dequantize_per_tensor_default_3" [label="(49, 3, 64, 32)", style=solid]; -"360 dequantize_per_tensor_default_3" -> "361 transpose" [label="(49, 3, 64, 32)", style=solid]; -"361 transpose" -> "362 matmul" [label="(49, 3, 32, 64)", style=solid]; -"362 matmul" -> "365 mul_1" [label="(49, 3, 64, 64)", style=solid]; -"363 clamp" -> "364 exp" [label="(3, 1, 1)", style=solid]; -"364 exp" -> "365 mul_1" [label="(3, 1, 1)", style=solid]; -"365 mul_1" -> "366 add" [label="(49, 3, 64, 64)", style=solid]; -"366 add" -> "367 softmax" [label="(49, 3, 64, 64)", style=solid]; -"367 softmax" -> "368 dropout" [label="(49, 3, 64, 64)", style=solid]; -"368 dropout" -> "369 matmul_1" [label="(49, 3, 64, 64)", style=solid]; -"369 matmul_1" -> "370 transpose_1" [label="(49, 3, 64, 32)", style=solid]; -"370 transpose_1" -> "371 reshape_2" [label="(49, 64, 3, 32)", style=solid]; -"371 reshape_2" -> "372 reshape_2_0_0_nncf_smooth_quant_0" [label="(49, 64, 96)", style=solid]; -"372 reshape_2_0_0_nncf_smooth_quant_0" -> "373 quantize_per_tensor_default_4" [label="(49, 64, 96)", style=solid]; -"373 quantize_per_tensor_default_4" -> "374 dequantize_per_tensor_default_4" [label="(49, 64, 96)", style=solid]; -"374 dequantize_per_tensor_default_4" -> "379 linear_3" [label="(49, 64, 96)", style=solid]; -"375 linear_3_scale_0" -> "378 dequantize_per_channel_default_4" [label="(96,)", style=solid]; -"376 linear_3_zero_point_0" -> "378 dequantize_per_channel_default_4" [label="(96,)", style=solid]; -"377 _frozen_param4" -> "378 dequantize_per_channel_default_4" [label="(96, 96)", style=solid]; -"378 dequantize_per_channel_default_4" -> "379 linear_3" [label="(96, 96)", style=solid]; -"379 linear_3" -> "380 dropout_1" [label="(49, 64, 96)", style=solid]; -"380 dropout_1" -> "381 view_3" [label="(49, 64, 96)", style=solid]; -"381 view_3" -> "382 permute_4" [label="(1, 7, 7, 8, 8, 96)", style=solid]; -"382 permute_4" -> "383 reshape_3" [label="(1, 7, 8, 7, 8, 96)", style=solid]; -"383 reshape_3" -> "384 slice_2" [label="(1, 56, 56, 96)", style=solid]; -"384 slice_2" -> "385 slice_3" [label="(1, 56, 56, 96)", style=solid]; -"385 slice_3" -> "386 layer_norm_1" [label="(1, 56, 56, 96)", style=solid]; -"386 layer_norm_1" -> "387 add_1" [label="(1, 56, 56, 96)", style=solid]; -"387 add_1" -> "388 add_1_0_0_nncf_smooth_quant_0" [label="(1, 56, 56, 96)", style=solid]; -"387 add_1" -> "408 add_2" [label="(1, 56, 56, 96)", style=solid]; -"388 add_1_0_0_nncf_smooth_quant_0" -> "389 quantize_per_tensor_default_5" [label="(1, 56, 56, 96)", style=solid]; -"389 quantize_per_tensor_default_5" -> "390 dequantize_per_tensor_default_5" [label="(1, 56, 56, 96)", style=solid]; -"390 dequantize_per_tensor_default_5" -> "395 linear_4" [label="(1, 56, 56, 96)", style=solid]; -"391 linear_4_scale_0" -> "394 dequantize_per_channel_default_5" [label="(384,)", style=solid]; -"392 linear_4_zero_point_0" -> "394 dequantize_per_channel_default_5" [label="(384,)", style=solid]; -"393 _frozen_param5" -> "394 dequantize_per_channel_default_5" [label="(384, 96)", style=solid]; -"394 dequantize_per_channel_default_5" -> "395 linear_4" [label="(384, 96)", style=solid]; -"395 linear_4" -> "396 gelu" [label="(1, 56, 56, 384)", style=solid]; -"396 gelu" -> "397 dropout_2" [label="(1, 56, 56, 384)", style=solid]; -"397 dropout_2" -> "398 dropout_2_0_0_nncf_smooth_quant_0" [label="(1, 56, 56, 384)", style=solid]; -"398 dropout_2_0_0_nncf_smooth_quant_0" -> "399 quantize_per_tensor_default_6" [label="(1, 56, 56, 384)", style=solid]; -"399 quantize_per_tensor_default_6" -> "400 dequantize_per_tensor_default_6" [label="(1, 56, 56, 384)", style=solid]; -"400 dequantize_per_tensor_default_6" -> "405 linear_5" [label="(1, 56, 56, 384)", style=solid]; -"401 linear_5_scale_0" -> "404 dequantize_per_channel_default_6" [label="(96,)", style=solid]; -"402 linear_5_zero_point_0" -> "404 dequantize_per_channel_default_6" [label="(96,)", style=solid]; -"403 _frozen_param6" -> "404 dequantize_per_channel_default_6" [label="(96, 384)", style=solid]; -"404 dequantize_per_channel_default_6" -> "405 linear_5" [label="(96, 384)", style=solid]; -"405 linear_5" -> "406 dropout_3" [label="(1, 56, 56, 96)", style=solid]; -"406 dropout_3" -> "407 layer_norm_2" [label="(1, 56, 56, 96)", style=solid]; -"407 layer_norm_2" -> "408 add_2" [label="(1, 56, 56, 96)", style=solid]; -"408 add_2" -> "430 pad_1" [label="(1, 56, 56, 96)", style=solid]; -"408 add_2" -> "504 add_5" [label="(1, 56, 56, 96)", style=solid]; -"409 features_1_1_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "414 linear_6" [label="(1, 15, 15, 2)", style=solid]; -"410 linear_6_scale_0" -> "413 dequantize_per_channel_default_7" [label="(512,)", style=solid]; -"411 linear_6_zero_point_0" -> "413 dequantize_per_channel_default_7" [label="(512,)", style=solid]; -"412 _frozen_param7" -> "413 dequantize_per_channel_default_7" [label="(512, 2)", style=solid]; -"413 dequantize_per_channel_default_7" -> "414 linear_6" [label="(512, 2)", style=solid]; -"414 linear_6" -> "415 relu__1" [label="(1, 15, 15, 512)", style=solid]; -"415 relu__1" -> "416 relu__1_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; -"416 relu__1_0_0_nncf_smooth_quant_0" -> "421 linear_7" [label="(1, 15, 15, 512)", style=solid]; -"417 linear_7_scale_0" -> "420 dequantize_per_channel_default_8" [label="(3,)", style=solid]; -"418 linear_7_zero_point_0" -> "420 dequantize_per_channel_default_8" [label="(3,)", style=solid]; -"419 _frozen_param8" -> "420 dequantize_per_channel_default_8" [label="(3, 512)", style=solid]; -"420 dequantize_per_channel_default_8" -> "421 linear_7" [label="(3, 512)", style=solid]; -"421 linear_7" -> "422 view_4" [label="(1, 15, 15, 3)", style=solid]; -"422 view_4" -> "423 index_1" [label="(225, 3)", style=solid]; -"423 index_1" -> "424 view_5" [label="(4096, 3)", style=solid]; -"424 view_5" -> "425 permute_5" [label="(64, 64, 3)", style=solid]; -"425 permute_5" -> "426 contiguous_1" [label="(3, 64, 64)", style=solid]; -"426 contiguous_1" -> "427 unsqueeze_1" [label="(3, 64, 64)", style=solid]; -"427 unsqueeze_1" -> "428 sigmoid_1" [label="(1, 3, 64, 64)", style=solid]; -"428 sigmoid_1" -> "429 mul_2" [label="(1, 3, 64, 64)", style=solid]; -"429 mul_2" -> "466 add_3" [label="(1, 3, 64, 64)", style=solid]; -"430 pad_1" -> "431 roll" [label="(1, 56, 56, 96)", style=solid]; -"431 roll" -> "432 view_6" [label="(1, 56, 56, 96)", style=solid]; -"432 view_6" -> "433 permute_6" [label="(1, 7, 8, 7, 8, 96)", style=solid]; -"433 permute_6" -> "434 reshape_4" [label="(1, 7, 7, 8, 8, 96)", style=solid]; -"434 reshape_4" -> "436 reshape_4_0_0_nncf_smooth_quant_0" [label="(49, 64, 96)", style=solid]; -"434 reshape_4" -> "467 new_zeros" [label="(49, 64, 96)", style=solid]; -"435 clone_1" -> "443 linear_8" [label="(288,)", style=solid]; -"436 reshape_4_0_0_nncf_smooth_quant_0" -> "437 quantize_per_tensor_default_7" [label="(49, 64, 96)", style=solid]; -"437 quantize_per_tensor_default_7" -> "438 dequantize_per_tensor_default_7" [label="(49, 64, 96)", style=solid]; -"438 dequantize_per_tensor_default_7" -> "443 linear_8" [label="(49, 64, 96)", style=solid]; -"439 linear_8_scale_0" -> "442 dequantize_per_channel_default_9" [label="(288,)", style=solid]; -"440 linear_8_zero_point_0" -> "442 dequantize_per_channel_default_9" [label="(288,)", style=solid]; -"441 _frozen_param9" -> "442 dequantize_per_channel_default_9" [label="(288, 96)", style=solid]; -"442 dequantize_per_channel_default_9" -> "443 linear_8" [label="(288, 96)", style=solid]; -"443 linear_8" -> "444 reshape_5" [label="(49, 64, 288)", style=solid]; -"444 reshape_5" -> "445 permute_7" [label="(49, 64, 3, 3, 32)", style=solid]; -"445 permute_7" -> "446 select_3" [label="(3, 49, 3, 64, 32)", style=solid]; -"445 permute_7" -> "447 select_4" [label="(3, 49, 3, 64, 32)", style=solid]; -"445 permute_7" -> "448 select_5" [label="(3, 49, 3, 64, 32)", style=solid]; -"446 select_3" -> "449 linalg_vector_norm_2" [label="(49, 3, 64, 32)", style=solid]; -"446 select_3" -> "451 expand_as_2" [label="(49, 3, 64, 32)", style=solid]; -"446 select_3" -> "452 div_2" [label="(49, 3, 64, 32)", style=solid]; -"447 select_4" -> "455 linalg_vector_norm_3" [label="(49, 3, 64, 32)", style=solid]; -"447 select_4" -> "457 expand_as_3" [label="(49, 3, 64, 32)", style=solid]; -"447 select_4" -> "458 div_3" [label="(49, 3, 64, 32)", style=solid]; -"448 select_5" -> "485 matmul_3" [label="(49, 3, 64, 32)", style=solid]; -"449 linalg_vector_norm_2" -> "450 clamp_min_2" [label="(49, 3, 64, 1)", style=solid]; -"450 clamp_min_2" -> "451 expand_as_2" [label="(49, 3, 64, 1)", style=solid]; -"451 expand_as_2" -> "452 div_2" [label="(49, 3, 64, 32)", style=solid]; -"452 div_2" -> "453 quantize_per_tensor_default_8" [label="(49, 3, 64, 32)", style=solid]; -"453 quantize_per_tensor_default_8" -> "454 dequantize_per_tensor_default_8" [label="(49, 3, 64, 32)", style=solid]; -"454 dequantize_per_tensor_default_8" -> "462 matmul_2" [label="(49, 3, 64, 32)", style=solid]; -"455 linalg_vector_norm_3" -> "456 clamp_min_3" [label="(49, 3, 64, 1)", style=solid]; -"456 clamp_min_3" -> "457 expand_as_3" [label="(49, 3, 64, 1)", style=solid]; -"457 expand_as_3" -> "458 div_3" [label="(49, 3, 64, 32)", style=solid]; -"458 div_3" -> "459 quantize_per_tensor_default_9" [label="(49, 3, 64, 32)", style=solid]; -"459 quantize_per_tensor_default_9" -> "460 dequantize_per_tensor_default_9" [label="(49, 3, 64, 32)", style=solid]; -"460 dequantize_per_tensor_default_9" -> "461 transpose_2" [label="(49, 3, 64, 32)", style=solid]; -"461 transpose_2" -> "462 matmul_2" [label="(49, 3, 32, 64)", style=solid]; -"462 matmul_2" -> "465 mul_3" [label="(49, 3, 64, 64)", style=solid]; -"463 clamp_1" -> "464 exp_1" [label="(3, 1, 1)", style=solid]; -"464 exp_1" -> "465 mul_3" [label="(3, 1, 1)", style=solid]; -"465 mul_3" -> "466 add_3" [label="(49, 3, 64, 64)", style=solid]; -"466 add_3" -> "478 view_8" [label="(49, 3, 64, 64)", style=solid]; -"467 new_zeros" -> "468 view_7" [label="(56, 56)", style=solid]; -"468 view_7" -> "469 permute_8" [label="(7, 8, 7, 8)", style=solid]; -"469 permute_8" -> "470 reshape_6" [label="(7, 7, 8, 8)", style=solid]; -"470 reshape_6" -> "471 unsqueeze_2" [label="(49, 64)", style=solid]; -"470 reshape_6" -> "472 unsqueeze_3" [label="(49, 64)", style=solid]; -"471 unsqueeze_2" -> "473 sub" [label="(49, 1, 64)", style=solid]; -"472 unsqueeze_3" -> "473 sub" [label="(49, 64, 1)", style=solid]; -"473 sub" -> "474 ne" [label="(49, 64, 64)", style=solid]; -"473 sub" -> "475 masked_fill" [label="(49, 64, 64)", style=solid]; -"473 sub" -> "476 eq" [label="(49, 64, 64)", style=solid]; -"474 ne" -> "475 masked_fill" [label="(49, 64, 64)", style=solid]; -"475 masked_fill" -> "477 masked_fill_1" [label="(49, 64, 64)", style=solid]; -"476 eq" -> "477 masked_fill_1" [label="(49, 64, 64)", style=solid]; -"477 masked_fill_1" -> "479 unsqueeze_4" [label="(49, 64, 64)", style=solid]; -"478 view_8" -> "481 add_4" [label="(1, 49, 3, 64, 64)", style=solid]; -"479 unsqueeze_4" -> "480 unsqueeze_5" [label="(49, 1, 64, 64)", style=solid]; -"480 unsqueeze_5" -> "481 add_4" [label="(1, 49, 1, 64, 64)", style=solid]; -"481 add_4" -> "482 view_9" [label="(1, 49, 3, 64, 64)", style=solid]; -"482 view_9" -> "483 softmax_1" [label="(49, 3, 64, 64)", style=solid]; -"483 softmax_1" -> "484 dropout_4" [label="(49, 3, 64, 64)", style=solid]; -"484 dropout_4" -> "485 matmul_3" [label="(49, 3, 64, 64)", style=solid]; -"485 matmul_3" -> "486 transpose_3" [label="(49, 3, 64, 32)", style=solid]; -"486 transpose_3" -> "487 reshape_7" [label="(49, 64, 3, 32)", style=solid]; -"487 reshape_7" -> "488 reshape_7_0_0_nncf_smooth_quant_0" [label="(49, 64, 96)", style=solid]; -"488 reshape_7_0_0_nncf_smooth_quant_0" -> "489 quantize_per_tensor_default_10" [label="(49, 64, 96)", style=solid]; -"489 quantize_per_tensor_default_10" -> "490 dequantize_per_tensor_default_10" [label="(49, 64, 96)", style=solid]; -"490 dequantize_per_tensor_default_10" -> "495 linear_9" [label="(49, 64, 96)", style=solid]; -"491 linear_9_scale_0" -> "494 dequantize_per_channel_default_10" [label="(96,)", style=solid]; -"492 linear_9_zero_point_0" -> "494 dequantize_per_channel_default_10" [label="(96,)", style=solid]; -"493 _frozen_param10" -> "494 dequantize_per_channel_default_10" [label="(96, 96)", style=solid]; -"494 dequantize_per_channel_default_10" -> "495 linear_9" [label="(96, 96)", style=solid]; -"495 linear_9" -> "496 dropout_5" [label="(49, 64, 96)", style=solid]; -"496 dropout_5" -> "497 view_10" [label="(49, 64, 96)", style=solid]; -"497 view_10" -> "498 permute_9" [label="(1, 7, 7, 8, 8, 96)", style=solid]; -"498 permute_9" -> "499 reshape_8" [label="(1, 7, 8, 7, 8, 96)", style=solid]; -"499 reshape_8" -> "500 roll_1" [label="(1, 56, 56, 96)", style=solid]; -"500 roll_1" -> "501 slice_23" [label="(1, 56, 56, 96)", style=solid]; -"501 slice_23" -> "502 slice_24" [label="(1, 56, 56, 96)", style=solid]; -"502 slice_24" -> "503 layer_norm_3" [label="(1, 56, 56, 96)", style=solid]; -"503 layer_norm_3" -> "504 add_5" [label="(1, 56, 56, 96)", style=solid]; -"504 add_5" -> "505 add_5_0_0_nncf_smooth_quant_0" [label="(1, 56, 56, 96)", style=solid]; -"504 add_5" -> "525 add_6" [label="(1, 56, 56, 96)", style=solid]; -"505 add_5_0_0_nncf_smooth_quant_0" -> "506 quantize_per_tensor_default_11" [label="(1, 56, 56, 96)", style=solid]; -"506 quantize_per_tensor_default_11" -> "507 dequantize_per_tensor_default_11" [label="(1, 56, 56, 96)", style=solid]; -"507 dequantize_per_tensor_default_11" -> "512 linear_10" [label="(1, 56, 56, 96)", style=solid]; -"508 linear_10_scale_0" -> "511 dequantize_per_channel_default_11" [label="(384,)", style=solid]; -"509 linear_10_zero_point_0" -> "511 dequantize_per_channel_default_11" [label="(384,)", style=solid]; -"510 _frozen_param11" -> "511 dequantize_per_channel_default_11" [label="(384, 96)", style=solid]; -"511 dequantize_per_channel_default_11" -> "512 linear_10" [label="(384, 96)", style=solid]; -"512 linear_10" -> "513 gelu_1" [label="(1, 56, 56, 384)", style=solid]; -"513 gelu_1" -> "514 dropout_6" [label="(1, 56, 56, 384)", style=solid]; -"514 dropout_6" -> "515 dropout_6_0_0_nncf_smooth_quant_0" [label="(1, 56, 56, 384)", style=solid]; -"515 dropout_6_0_0_nncf_smooth_quant_0" -> "516 quantize_per_tensor_default_12" [label="(1, 56, 56, 384)", style=solid]; -"516 quantize_per_tensor_default_12" -> "517 dequantize_per_tensor_default_12" [label="(1, 56, 56, 384)", style=solid]; -"517 dequantize_per_tensor_default_12" -> "522 linear_11" [label="(1, 56, 56, 384)", style=solid]; -"518 linear_11_scale_0" -> "521 dequantize_per_channel_default_12" [label="(96,)", style=solid]; -"519 linear_11_zero_point_0" -> "521 dequantize_per_channel_default_12" [label="(96,)", style=solid]; -"520 _frozen_param12" -> "521 dequantize_per_channel_default_12" [label="(96, 384)", style=solid]; -"521 dequantize_per_channel_default_12" -> "522 linear_11" [label="(96, 384)", style=solid]; -"522 linear_11" -> "523 dropout_7" [label="(1, 56, 56, 96)", style=solid]; -"523 dropout_7" -> "524 layer_norm_4" [label="(1, 56, 56, 96)", style=solid]; -"524 layer_norm_4" -> "525 add_6" [label="(1, 56, 56, 96)", style=solid]; -"525 add_6" -> "526 pad_2" [label="(1, 56, 56, 96)", style=solid]; -"526 pad_2" -> "527 slice_25" [label="(1, 56, 56, 96)", style=solid]; -"526 pad_2" -> "530 slice_28" [label="(1, 56, 56, 96)", style=solid]; -"526 pad_2" -> "533 slice_31" [label="(1, 56, 56, 96)", style=solid]; -"526 pad_2" -> "536 slice_34" [label="(1, 56, 56, 96)", style=solid]; -"527 slice_25" -> "528 slice_26" [label="(1, 28, 56, 96)", style=solid]; -"528 slice_26" -> "529 slice_27" [label="(1, 28, 28, 96)", style=solid]; -"529 slice_27" -> "539 cat" [label="(1, 28, 28, 96)", style=solid]; -"530 slice_28" -> "531 slice_29" [label="(1, 28, 56, 96)", style=solid]; -"531 slice_29" -> "532 slice_30" [label="(1, 28, 28, 96)", style=solid]; -"532 slice_30" -> "539 cat" [label="(1, 28, 28, 96)", style=solid]; -"533 slice_31" -> "534 slice_32" [label="(1, 28, 56, 96)", style=solid]; -"534 slice_32" -> "535 slice_33" [label="(1, 28, 28, 96)", style=solid]; -"535 slice_33" -> "539 cat" [label="(1, 28, 28, 96)", style=solid]; -"536 slice_34" -> "537 slice_35" [label="(1, 28, 56, 96)", style=solid]; -"537 slice_35" -> "538 slice_36" [label="(1, 28, 28, 96)", style=solid]; -"538 slice_36" -> "539 cat" [label="(1, 28, 28, 96)", style=solid]; -"539 cat" -> "540 cat_0_0_nncf_smooth_quant_0" [label="(1, 28, 28, 384)", style=solid]; -"540 cat_0_0_nncf_smooth_quant_0" -> "541 quantize_per_tensor_default_13" [label="(1, 28, 28, 384)", style=solid]; -"541 quantize_per_tensor_default_13" -> "542 dequantize_per_tensor_default_13" [label="(1, 28, 28, 384)", style=solid]; -"542 dequantize_per_tensor_default_13" -> "547 linear_12" [label="(1, 28, 28, 384)", style=solid]; -"543 linear_12_scale_0" -> "546 dequantize_per_channel_default_13" [label="(192,)", style=solid]; -"544 linear_12_zero_point_0" -> "546 dequantize_per_channel_default_13" [label="(192,)", style=solid]; -"545 _frozen_param13" -> "546 dequantize_per_channel_default_13" [label="(192, 384)", style=solid]; -"546 dequantize_per_channel_default_13" -> "547 linear_12" [label="(192, 384)", style=solid]; -"547 linear_12" -> "548 layer_norm_5" [label="(1, 28, 28, 192)", style=solid]; -"548 layer_norm_5" -> "570 pad_3" [label="(1, 28, 28, 192)", style=solid]; -"548 layer_norm_5" -> "629 add_8" [label="(1, 28, 28, 192)", style=solid]; -"549 features_3_0_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "554 linear_13" [label="(1, 15, 15, 2)", style=solid]; -"550 linear_13_scale_0" -> "553 dequantize_per_channel_default_14" [label="(512,)", style=solid]; -"551 linear_13_zero_point_0" -> "553 dequantize_per_channel_default_14" [label="(512,)", style=solid]; -"552 _frozen_param14" -> "553 dequantize_per_channel_default_14" [label="(512, 2)", style=solid]; -"553 dequantize_per_channel_default_14" -> "554 linear_13" [label="(512, 2)", style=solid]; -"554 linear_13" -> "555 relu__2" [label="(1, 15, 15, 512)", style=solid]; -"555 relu__2" -> "556 relu__2_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; -"556 relu__2_0_0_nncf_smooth_quant_0" -> "561 linear_14" [label="(1, 15, 15, 512)", style=solid]; -"557 linear_14_scale_0" -> "560 dequantize_per_channel_default_15" [label="(6,)", style=solid]; -"558 linear_14_zero_point_0" -> "560 dequantize_per_channel_default_15" [label="(6,)", style=solid]; -"559 _frozen_param15" -> "560 dequantize_per_channel_default_15" [label="(6, 512)", style=solid]; -"560 dequantize_per_channel_default_15" -> "561 linear_14" [label="(6, 512)", style=solid]; -"561 linear_14" -> "562 view_11" [label="(1, 15, 15, 6)", style=solid]; -"562 view_11" -> "563 index_2" [label="(225, 6)", style=solid]; -"563 index_2" -> "564 view_12" [label="(4096, 6)", style=solid]; -"564 view_12" -> "565 permute_10" [label="(64, 64, 6)", style=solid]; -"565 permute_10" -> "566 contiguous_2" [label="(6, 64, 64)", style=solid]; -"566 contiguous_2" -> "567 unsqueeze_6" [label="(6, 64, 64)", style=solid]; -"567 unsqueeze_6" -> "568 sigmoid_2" [label="(1, 6, 64, 64)", style=solid]; -"568 sigmoid_2" -> "569 mul_4" [label="(1, 6, 64, 64)", style=solid]; -"569 mul_4" -> "605 add_7" [label="(1, 6, 64, 64)", style=solid]; -"570 pad_3" -> "571 view_13" [label="(1, 32, 32, 192)", style=solid]; -"571 view_13" -> "572 permute_11" [label="(1, 4, 8, 4, 8, 192)", style=solid]; -"572 permute_11" -> "573 reshape_9" [label="(1, 4, 4, 8, 8, 192)", style=solid]; -"573 reshape_9" -> "575 reshape_9_0_0_nncf_smooth_quant_0" [label="(16, 64, 192)", style=solid]; -"574 clone_2" -> "582 linear_15" [label="(576,)", style=solid]; -"575 reshape_9_0_0_nncf_smooth_quant_0" -> "576 quantize_per_tensor_default_14" [label="(16, 64, 192)", style=solid]; -"576 quantize_per_tensor_default_14" -> "577 dequantize_per_tensor_default_14" [label="(16, 64, 192)", style=solid]; -"577 dequantize_per_tensor_default_14" -> "582 linear_15" [label="(16, 64, 192)", style=solid]; -"578 linear_15_scale_0" -> "581 dequantize_per_channel_default_16" [label="(576,)", style=solid]; -"579 linear_15_zero_point_0" -> "581 dequantize_per_channel_default_16" [label="(576,)", style=solid]; -"580 _frozen_param16" -> "581 dequantize_per_channel_default_16" [label="(576, 192)", style=solid]; -"581 dequantize_per_channel_default_16" -> "582 linear_15" [label="(576, 192)", style=solid]; -"582 linear_15" -> "583 reshape_10" [label="(16, 64, 576)", style=solid]; -"583 reshape_10" -> "584 permute_12" [label="(16, 64, 3, 6, 32)", style=solid]; -"584 permute_12" -> "585 select_6" [label="(3, 16, 6, 64, 32)", style=solid]; -"584 permute_12" -> "586 select_7" [label="(3, 16, 6, 64, 32)", style=solid]; -"584 permute_12" -> "587 select_8" [label="(3, 16, 6, 64, 32)", style=solid]; -"585 select_6" -> "588 linalg_vector_norm_4" [label="(16, 6, 64, 32)", style=solid]; -"585 select_6" -> "590 expand_as_4" [label="(16, 6, 64, 32)", style=solid]; -"585 select_6" -> "591 div_4" [label="(16, 6, 64, 32)", style=solid]; -"586 select_7" -> "594 linalg_vector_norm_5" [label="(16, 6, 64, 32)", style=solid]; -"586 select_7" -> "596 expand_as_5" [label="(16, 6, 64, 32)", style=solid]; -"586 select_7" -> "597 div_5" [label="(16, 6, 64, 32)", style=solid]; -"587 select_8" -> "608 matmul_5" [label="(16, 6, 64, 32)", style=solid]; -"588 linalg_vector_norm_4" -> "589 clamp_min_4" [label="(16, 6, 64, 1)", style=solid]; -"589 clamp_min_4" -> "590 expand_as_4" [label="(16, 6, 64, 1)", style=solid]; -"590 expand_as_4" -> "591 div_4" [label="(16, 6, 64, 32)", style=solid]; -"591 div_4" -> "592 quantize_per_tensor_default_15" [label="(16, 6, 64, 32)", style=solid]; -"592 quantize_per_tensor_default_15" -> "593 dequantize_per_tensor_default_15" [label="(16, 6, 64, 32)", style=solid]; -"593 dequantize_per_tensor_default_15" -> "601 matmul_4" [label="(16, 6, 64, 32)", style=solid]; -"594 linalg_vector_norm_5" -> "595 clamp_min_5" [label="(16, 6, 64, 1)", style=solid]; -"595 clamp_min_5" -> "596 expand_as_5" [label="(16, 6, 64, 1)", style=solid]; -"596 expand_as_5" -> "597 div_5" [label="(16, 6, 64, 32)", style=solid]; -"597 div_5" -> "598 quantize_per_tensor_default_16" [label="(16, 6, 64, 32)", style=solid]; -"598 quantize_per_tensor_default_16" -> "599 dequantize_per_tensor_default_16" [label="(16, 6, 64, 32)", style=solid]; -"599 dequantize_per_tensor_default_16" -> "600 transpose_4" [label="(16, 6, 64, 32)", style=solid]; -"600 transpose_4" -> "601 matmul_4" [label="(16, 6, 32, 64)", style=solid]; -"601 matmul_4" -> "604 mul_5" [label="(16, 6, 64, 64)", style=solid]; -"602 clamp_2" -> "603 exp_2" [label="(6, 1, 1)", style=solid]; -"603 exp_2" -> "604 mul_5" [label="(6, 1, 1)", style=solid]; -"604 mul_5" -> "605 add_7" [label="(16, 6, 64, 64)", style=solid]; -"605 add_7" -> "606 softmax_2" [label="(16, 6, 64, 64)", style=solid]; -"606 softmax_2" -> "607 dropout_8" [label="(16, 6, 64, 64)", style=solid]; -"607 dropout_8" -> "608 matmul_5" [label="(16, 6, 64, 64)", style=solid]; -"608 matmul_5" -> "609 transpose_5" [label="(16, 6, 64, 32)", style=solid]; -"609 transpose_5" -> "610 reshape_11" [label="(16, 64, 6, 32)", style=solid]; -"610 reshape_11" -> "611 reshape_11_0_0_nncf_smooth_quant_0" [label="(16, 64, 192)", style=solid]; -"611 reshape_11_0_0_nncf_smooth_quant_0" -> "612 quantize_per_tensor_default_17" [label="(16, 64, 192)", style=solid]; -"612 quantize_per_tensor_default_17" -> "613 dequantize_per_tensor_default_17" [label="(16, 64, 192)", style=solid]; -"613 dequantize_per_tensor_default_17" -> "618 linear_16" [label="(16, 64, 192)", style=solid]; -"614 linear_16_scale_0" -> "617 dequantize_per_channel_default_17" [label="(192,)", style=solid]; -"615 linear_16_zero_point_0" -> "617 dequantize_per_channel_default_17" [label="(192,)", style=solid]; -"616 _frozen_param17" -> "617 dequantize_per_channel_default_17" [label="(192, 192)", style=solid]; -"617 dequantize_per_channel_default_17" -> "618 linear_16" [label="(192, 192)", style=solid]; -"618 linear_16" -> "619 dropout_9" [label="(16, 64, 192)", style=solid]; -"619 dropout_9" -> "620 view_14" [label="(16, 64, 192)", style=solid]; -"620 view_14" -> "621 permute_13" [label="(1, 4, 4, 8, 8, 192)", style=solid]; -"621 permute_13" -> "622 reshape_12" [label="(1, 4, 8, 4, 8, 192)", style=solid]; -"622 reshape_12" -> "623 slice_38" [label="(1, 32, 32, 192)", style=solid]; -"623 slice_38" -> "624 slice_39" [label="(1, 32, 32, 192)", style=solid]; -"624 slice_39" -> "625 slice_40" [label="(1, 28, 32, 192)", style=solid]; -"625 slice_40" -> "626 slice_41" [label="(1, 28, 28, 192)", style=solid]; -"626 slice_41" -> "627 contiguous_3" [label="(1, 28, 28, 192)", style=solid]; -"627 contiguous_3" -> "628 layer_norm_6" [label="(1, 28, 28, 192)", style=solid]; -"628 layer_norm_6" -> "629 add_8" [label="(1, 28, 28, 192)", style=solid]; -"629 add_8" -> "630 add_8_0_0_nncf_smooth_quant_0" [label="(1, 28, 28, 192)", style=solid]; -"629 add_8" -> "650 add_9" [label="(1, 28, 28, 192)", style=solid]; -"630 add_8_0_0_nncf_smooth_quant_0" -> "631 quantize_per_tensor_default_18" [label="(1, 28, 28, 192)", style=solid]; -"631 quantize_per_tensor_default_18" -> "632 dequantize_per_tensor_default_18" [label="(1, 28, 28, 192)", style=solid]; -"632 dequantize_per_tensor_default_18" -> "637 linear_17" [label="(1, 28, 28, 192)", style=solid]; -"633 linear_17_scale_0" -> "636 dequantize_per_channel_default_18" [label="(768,)", style=solid]; -"634 linear_17_zero_point_0" -> "636 dequantize_per_channel_default_18" [label="(768,)", style=solid]; -"635 _frozen_param18" -> "636 dequantize_per_channel_default_18" [label="(768, 192)", style=solid]; -"636 dequantize_per_channel_default_18" -> "637 linear_17" [label="(768, 192)", style=solid]; -"637 linear_17" -> "638 gelu_2" [label="(1, 28, 28, 768)", style=solid]; -"638 gelu_2" -> "639 dropout_10" [label="(1, 28, 28, 768)", style=solid]; -"639 dropout_10" -> "640 dropout_10_0_0_nncf_smooth_quant_0" [label="(1, 28, 28, 768)", style=solid]; -"640 dropout_10_0_0_nncf_smooth_quant_0" -> "641 quantize_per_tensor_default_19" [label="(1, 28, 28, 768)", style=solid]; -"641 quantize_per_tensor_default_19" -> "642 dequantize_per_tensor_default_19" [label="(1, 28, 28, 768)", style=solid]; -"642 dequantize_per_tensor_default_19" -> "647 linear_18" [label="(1, 28, 28, 768)", style=solid]; -"643 linear_18_scale_0" -> "646 dequantize_per_channel_default_19" [label="(192,)", style=solid]; -"644 linear_18_zero_point_0" -> "646 dequantize_per_channel_default_19" [label="(192,)", style=solid]; -"645 _frozen_param19" -> "646 dequantize_per_channel_default_19" [label="(192, 768)", style=solid]; -"646 dequantize_per_channel_default_19" -> "647 linear_18" [label="(192, 768)", style=solid]; -"647 linear_18" -> "648 dropout_11" [label="(1, 28, 28, 192)", style=solid]; -"648 dropout_11" -> "649 layer_norm_7" [label="(1, 28, 28, 192)", style=solid]; -"649 layer_norm_7" -> "650 add_9" [label="(1, 28, 28, 192)", style=solid]; -"650 add_9" -> "672 pad_4" [label="(1, 28, 28, 192)", style=solid]; -"650 add_9" -> "749 add_12" [label="(1, 28, 28, 192)", style=solid]; -"651 features_3_1_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "656 linear_19" [label="(1, 15, 15, 2)", style=solid]; -"652 linear_19_scale_0" -> "655 dequantize_per_channel_default_20" [label="(512,)", style=solid]; -"653 linear_19_zero_point_0" -> "655 dequantize_per_channel_default_20" [label="(512,)", style=solid]; -"654 _frozen_param20" -> "655 dequantize_per_channel_default_20" [label="(512, 2)", style=solid]; -"655 dequantize_per_channel_default_20" -> "656 linear_19" [label="(512, 2)", style=solid]; -"656 linear_19" -> "657 relu__3" [label="(1, 15, 15, 512)", style=solid]; -"657 relu__3" -> "658 relu__3_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; -"658 relu__3_0_0_nncf_smooth_quant_0" -> "663 linear_20" [label="(1, 15, 15, 512)", style=solid]; -"659 linear_20_scale_0" -> "662 dequantize_per_channel_default_21" [label="(6,)", style=solid]; -"660 linear_20_zero_point_0" -> "662 dequantize_per_channel_default_21" [label="(6,)", style=solid]; -"661 _frozen_param21" -> "662 dequantize_per_channel_default_21" [label="(6, 512)", style=solid]; -"662 dequantize_per_channel_default_21" -> "663 linear_20" [label="(6, 512)", style=solid]; -"663 linear_20" -> "664 view_15" [label="(1, 15, 15, 6)", style=solid]; -"664 view_15" -> "665 index_3" [label="(225, 6)", style=solid]; -"665 index_3" -> "666 view_16" [label="(4096, 6)", style=solid]; -"666 view_16" -> "667 permute_14" [label="(64, 64, 6)", style=solid]; -"667 permute_14" -> "668 contiguous_4" [label="(6, 64, 64)", style=solid]; -"668 contiguous_4" -> "669 unsqueeze_7" [label="(6, 64, 64)", style=solid]; -"669 unsqueeze_7" -> "670 sigmoid_3" [label="(1, 6, 64, 64)", style=solid]; -"670 sigmoid_3" -> "671 mul_6" [label="(1, 6, 64, 64)", style=solid]; -"671 mul_6" -> "708 add_10" [label="(1, 6, 64, 64)", style=solid]; -"672 pad_4" -> "673 roll_2" [label="(1, 32, 32, 192)", style=solid]; -"673 roll_2" -> "674 view_17" [label="(1, 32, 32, 192)", style=solid]; -"674 view_17" -> "675 permute_15" [label="(1, 4, 8, 4, 8, 192)", style=solid]; -"675 permute_15" -> "676 reshape_13" [label="(1, 4, 4, 8, 8, 192)", style=solid]; -"676 reshape_13" -> "678 reshape_13_0_0_nncf_smooth_quant_0" [label="(16, 64, 192)", style=solid]; -"676 reshape_13" -> "709 new_zeros_1" [label="(16, 64, 192)", style=solid]; -"677 clone_3" -> "685 linear_21" [label="(576,)", style=solid]; -"678 reshape_13_0_0_nncf_smooth_quant_0" -> "679 quantize_per_tensor_default_20" [label="(16, 64, 192)", style=solid]; -"679 quantize_per_tensor_default_20" -> "680 dequantize_per_tensor_default_20" [label="(16, 64, 192)", style=solid]; -"680 dequantize_per_tensor_default_20" -> "685 linear_21" [label="(16, 64, 192)", style=solid]; -"681 linear_21_scale_0" -> "684 dequantize_per_channel_default_22" [label="(576,)", style=solid]; -"682 linear_21_zero_point_0" -> "684 dequantize_per_channel_default_22" [label="(576,)", style=solid]; -"683 _frozen_param22" -> "684 dequantize_per_channel_default_22" [label="(576, 192)", style=solid]; -"684 dequantize_per_channel_default_22" -> "685 linear_21" [label="(576, 192)", style=solid]; -"685 linear_21" -> "686 reshape_14" [label="(16, 64, 576)", style=solid]; -"686 reshape_14" -> "687 permute_16" [label="(16, 64, 3, 6, 32)", style=solid]; -"687 permute_16" -> "688 select_9" [label="(3, 16, 6, 64, 32)", style=solid]; -"687 permute_16" -> "689 select_10" [label="(3, 16, 6, 64, 32)", style=solid]; -"687 permute_16" -> "690 select_11" [label="(3, 16, 6, 64, 32)", style=solid]; -"688 select_9" -> "691 linalg_vector_norm_6" [label="(16, 6, 64, 32)", style=solid]; -"688 select_9" -> "693 expand_as_6" [label="(16, 6, 64, 32)", style=solid]; -"688 select_9" -> "694 div_6" [label="(16, 6, 64, 32)", style=solid]; -"689 select_10" -> "697 linalg_vector_norm_7" [label="(16, 6, 64, 32)", style=solid]; -"689 select_10" -> "699 expand_as_7" [label="(16, 6, 64, 32)", style=solid]; -"689 select_10" -> "700 div_7" [label="(16, 6, 64, 32)", style=solid]; -"690 select_11" -> "727 matmul_7" [label="(16, 6, 64, 32)", style=solid]; -"691 linalg_vector_norm_6" -> "692 clamp_min_6" [label="(16, 6, 64, 1)", style=solid]; -"692 clamp_min_6" -> "693 expand_as_6" [label="(16, 6, 64, 1)", style=solid]; -"693 expand_as_6" -> "694 div_6" [label="(16, 6, 64, 32)", style=solid]; -"694 div_6" -> "695 quantize_per_tensor_default_21" [label="(16, 6, 64, 32)", style=solid]; -"695 quantize_per_tensor_default_21" -> "696 dequantize_per_tensor_default_21" [label="(16, 6, 64, 32)", style=solid]; -"696 dequantize_per_tensor_default_21" -> "704 matmul_6" [label="(16, 6, 64, 32)", style=solid]; -"697 linalg_vector_norm_7" -> "698 clamp_min_7" [label="(16, 6, 64, 1)", style=solid]; -"698 clamp_min_7" -> "699 expand_as_7" [label="(16, 6, 64, 1)", style=solid]; -"699 expand_as_7" -> "700 div_7" [label="(16, 6, 64, 32)", style=solid]; -"700 div_7" -> "701 quantize_per_tensor_default_22" [label="(16, 6, 64, 32)", style=solid]; -"701 quantize_per_tensor_default_22" -> "702 dequantize_per_tensor_default_22" [label="(16, 6, 64, 32)", style=solid]; -"702 dequantize_per_tensor_default_22" -> "703 transpose_6" [label="(16, 6, 64, 32)", style=solid]; -"703 transpose_6" -> "704 matmul_6" [label="(16, 6, 32, 64)", style=solid]; -"704 matmul_6" -> "707 mul_7" [label="(16, 6, 64, 64)", style=solid]; -"705 clamp_3" -> "706 exp_3" [label="(6, 1, 1)", style=solid]; -"706 exp_3" -> "707 mul_7" [label="(6, 1, 1)", style=solid]; -"707 mul_7" -> "708 add_10" [label="(16, 6, 64, 64)", style=solid]; -"708 add_10" -> "720 view_19" [label="(16, 6, 64, 64)", style=solid]; -"709 new_zeros_1" -> "710 view_18" [label="(32, 32)", style=solid]; -"710 view_18" -> "711 permute_17" [label="(4, 8, 4, 8)", style=solid]; -"711 permute_17" -> "712 reshape_15" [label="(4, 4, 8, 8)", style=solid]; -"712 reshape_15" -> "713 unsqueeze_8" [label="(16, 64)", style=solid]; -"712 reshape_15" -> "714 unsqueeze_9" [label="(16, 64)", style=solid]; -"713 unsqueeze_8" -> "715 sub_1" [label="(16, 1, 64)", style=solid]; -"714 unsqueeze_9" -> "715 sub_1" [label="(16, 64, 1)", style=solid]; -"715 sub_1" -> "716 ne_1" [label="(16, 64, 64)", style=solid]; -"715 sub_1" -> "717 masked_fill_2" [label="(16, 64, 64)", style=solid]; -"715 sub_1" -> "718 eq_1" [label="(16, 64, 64)", style=solid]; -"716 ne_1" -> "717 masked_fill_2" [label="(16, 64, 64)", style=solid]; -"717 masked_fill_2" -> "719 masked_fill_3" [label="(16, 64, 64)", style=solid]; -"718 eq_1" -> "719 masked_fill_3" [label="(16, 64, 64)", style=solid]; -"719 masked_fill_3" -> "721 unsqueeze_10" [label="(16, 64, 64)", style=solid]; -"720 view_19" -> "723 add_11" [label="(1, 16, 6, 64, 64)", style=solid]; -"721 unsqueeze_10" -> "722 unsqueeze_11" [label="(16, 1, 64, 64)", style=solid]; -"722 unsqueeze_11" -> "723 add_11" [label="(1, 16, 1, 64, 64)", style=solid]; -"723 add_11" -> "724 view_20" [label="(1, 16, 6, 64, 64)", style=solid]; -"724 view_20" -> "725 softmax_3" [label="(16, 6, 64, 64)", style=solid]; -"725 softmax_3" -> "726 dropout_12" [label="(16, 6, 64, 64)", style=solid]; -"726 dropout_12" -> "727 matmul_7" [label="(16, 6, 64, 64)", style=solid]; -"727 matmul_7" -> "728 transpose_7" [label="(16, 6, 64, 32)", style=solid]; -"728 transpose_7" -> "729 reshape_16" [label="(16, 64, 6, 32)", style=solid]; -"729 reshape_16" -> "730 reshape_16_0_0_nncf_smooth_quant_0" [label="(16, 64, 192)", style=solid]; -"730 reshape_16_0_0_nncf_smooth_quant_0" -> "731 quantize_per_tensor_default_23" [label="(16, 64, 192)", style=solid]; -"731 quantize_per_tensor_default_23" -> "732 dequantize_per_tensor_default_23" [label="(16, 64, 192)", style=solid]; -"732 dequantize_per_tensor_default_23" -> "737 linear_22" [label="(16, 64, 192)", style=solid]; -"733 linear_22_scale_0" -> "736 dequantize_per_channel_default_23" [label="(192,)", style=solid]; -"734 linear_22_zero_point_0" -> "736 dequantize_per_channel_default_23" [label="(192,)", style=solid]; -"735 _frozen_param23" -> "736 dequantize_per_channel_default_23" [label="(192, 192)", style=solid]; -"736 dequantize_per_channel_default_23" -> "737 linear_22" [label="(192, 192)", style=solid]; -"737 linear_22" -> "738 dropout_13" [label="(16, 64, 192)", style=solid]; -"738 dropout_13" -> "739 view_21" [label="(16, 64, 192)", style=solid]; -"739 view_21" -> "740 permute_18" [label="(1, 4, 4, 8, 8, 192)", style=solid]; -"740 permute_18" -> "741 reshape_17" [label="(1, 4, 8, 4, 8, 192)", style=solid]; -"741 reshape_17" -> "742 roll_3" [label="(1, 32, 32, 192)", style=solid]; -"742 roll_3" -> "743 slice_61" [label="(1, 32, 32, 192)", style=solid]; -"743 slice_61" -> "744 slice_62" [label="(1, 32, 32, 192)", style=solid]; -"744 slice_62" -> "745 slice_63" [label="(1, 28, 32, 192)", style=solid]; -"745 slice_63" -> "746 slice_64" [label="(1, 28, 28, 192)", style=solid]; -"746 slice_64" -> "747 contiguous_5" [label="(1, 28, 28, 192)", style=solid]; -"747 contiguous_5" -> "748 layer_norm_8" [label="(1, 28, 28, 192)", style=solid]; -"748 layer_norm_8" -> "749 add_12" [label="(1, 28, 28, 192)", style=solid]; -"749 add_12" -> "750 add_12_0_0_nncf_smooth_quant_0" [label="(1, 28, 28, 192)", style=solid]; -"749 add_12" -> "770 add_13" [label="(1, 28, 28, 192)", style=solid]; -"750 add_12_0_0_nncf_smooth_quant_0" -> "751 quantize_per_tensor_default_24" [label="(1, 28, 28, 192)", style=solid]; -"751 quantize_per_tensor_default_24" -> "752 dequantize_per_tensor_default_24" [label="(1, 28, 28, 192)", style=solid]; -"752 dequantize_per_tensor_default_24" -> "757 linear_23" [label="(1, 28, 28, 192)", style=solid]; -"753 linear_23_scale_0" -> "756 dequantize_per_channel_default_24" [label="(768,)", style=solid]; -"754 linear_23_zero_point_0" -> "756 dequantize_per_channel_default_24" [label="(768,)", style=solid]; -"755 _frozen_param24" -> "756 dequantize_per_channel_default_24" [label="(768, 192)", style=solid]; -"756 dequantize_per_channel_default_24" -> "757 linear_23" [label="(768, 192)", style=solid]; -"757 linear_23" -> "758 gelu_3" [label="(1, 28, 28, 768)", style=solid]; -"758 gelu_3" -> "759 dropout_14" [label="(1, 28, 28, 768)", style=solid]; -"759 dropout_14" -> "760 dropout_14_0_0_nncf_smooth_quant_0" [label="(1, 28, 28, 768)", style=solid]; -"760 dropout_14_0_0_nncf_smooth_quant_0" -> "761 quantize_per_tensor_default_25" [label="(1, 28, 28, 768)", style=solid]; -"761 quantize_per_tensor_default_25" -> "762 dequantize_per_tensor_default_25" [label="(1, 28, 28, 768)", style=solid]; -"762 dequantize_per_tensor_default_25" -> "767 linear_24" [label="(1, 28, 28, 768)", style=solid]; -"763 linear_24_scale_0" -> "766 dequantize_per_channel_default_25" [label="(192,)", style=solid]; -"764 linear_24_zero_point_0" -> "766 dequantize_per_channel_default_25" [label="(192,)", style=solid]; -"765 _frozen_param25" -> "766 dequantize_per_channel_default_25" [label="(192, 768)", style=solid]; -"766 dequantize_per_channel_default_25" -> "767 linear_24" [label="(192, 768)", style=solid]; -"767 linear_24" -> "768 dropout_15" [label="(1, 28, 28, 192)", style=solid]; -"768 dropout_15" -> "769 layer_norm_9" [label="(1, 28, 28, 192)", style=solid]; -"769 layer_norm_9" -> "770 add_13" [label="(1, 28, 28, 192)", style=solid]; -"770 add_13" -> "771 pad_5" [label="(1, 28, 28, 192)", style=solid]; -"771 pad_5" -> "772 slice_65" [label="(1, 28, 28, 192)", style=solid]; -"771 pad_5" -> "775 slice_68" [label="(1, 28, 28, 192)", style=solid]; -"771 pad_5" -> "778 slice_71" [label="(1, 28, 28, 192)", style=solid]; -"771 pad_5" -> "781 slice_74" [label="(1, 28, 28, 192)", style=solid]; -"772 slice_65" -> "773 slice_66" [label="(1, 14, 28, 192)", style=solid]; -"773 slice_66" -> "774 slice_67" [label="(1, 14, 14, 192)", style=solid]; -"774 slice_67" -> "784 cat_1" [label="(1, 14, 14, 192)", style=solid]; -"775 slice_68" -> "776 slice_69" [label="(1, 14, 28, 192)", style=solid]; -"776 slice_69" -> "777 slice_70" [label="(1, 14, 14, 192)", style=solid]; -"777 slice_70" -> "784 cat_1" [label="(1, 14, 14, 192)", style=solid]; -"778 slice_71" -> "779 slice_72" [label="(1, 14, 28, 192)", style=solid]; -"779 slice_72" -> "780 slice_73" [label="(1, 14, 14, 192)", style=solid]; -"780 slice_73" -> "784 cat_1" [label="(1, 14, 14, 192)", style=solid]; -"781 slice_74" -> "782 slice_75" [label="(1, 14, 28, 192)", style=solid]; -"782 slice_75" -> "783 slice_76" [label="(1, 14, 14, 192)", style=solid]; -"783 slice_76" -> "784 cat_1" [label="(1, 14, 14, 192)", style=solid]; -"784 cat_1" -> "785 cat_1_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 768)", style=solid]; -"785 cat_1_0_0_nncf_smooth_quant_0" -> "786 quantize_per_tensor_default_26" [label="(1, 14, 14, 768)", style=solid]; -"786 quantize_per_tensor_default_26" -> "787 dequantize_per_tensor_default_26" [label="(1, 14, 14, 768)", style=solid]; -"787 dequantize_per_tensor_default_26" -> "792 linear_25" [label="(1, 14, 14, 768)", style=solid]; -"788 linear_25_scale_0" -> "791 dequantize_per_channel_default_26" [label="(384,)", style=solid]; -"789 linear_25_zero_point_0" -> "791 dequantize_per_channel_default_26" [label="(384,)", style=solid]; -"790 _frozen_param26" -> "791 dequantize_per_channel_default_26" [label="(384, 768)", style=solid]; -"791 dequantize_per_channel_default_26" -> "792 linear_25" [label="(384, 768)", style=solid]; -"792 linear_25" -> "793 layer_norm_10" [label="(1, 14, 14, 384)", style=solid]; -"793 layer_norm_10" -> "815 pad_6" [label="(1, 14, 14, 384)", style=solid]; -"793 layer_norm_10" -> "874 add_15" [label="(1, 14, 14, 384)", style=solid]; -"794 features_5_0_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "799 linear_26" [label="(1, 15, 15, 2)", style=solid]; -"795 linear_26_scale_0" -> "798 dequantize_per_channel_default_27" [label="(512,)", style=solid]; -"796 linear_26_zero_point_0" -> "798 dequantize_per_channel_default_27" [label="(512,)", style=solid]; -"797 _frozen_param27" -> "798 dequantize_per_channel_default_27" [label="(512, 2)", style=solid]; -"798 dequantize_per_channel_default_27" -> "799 linear_26" [label="(512, 2)", style=solid]; -"799 linear_26" -> "800 relu__4" [label="(1, 15, 15, 512)", style=solid]; -"800 relu__4" -> "801 relu__4_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; -"801 relu__4_0_0_nncf_smooth_quant_0" -> "806 linear_27" [label="(1, 15, 15, 512)", style=solid]; -"802 linear_27_scale_0" -> "805 dequantize_per_channel_default_28" [label="(12,)", style=solid]; -"803 linear_27_zero_point_0" -> "805 dequantize_per_channel_default_28" [label="(12,)", style=solid]; -"804 _frozen_param28" -> "805 dequantize_per_channel_default_28" [label="(12, 512)", style=solid]; -"805 dequantize_per_channel_default_28" -> "806 linear_27" [label="(12, 512)", style=solid]; -"806 linear_27" -> "807 view_22" [label="(1, 15, 15, 12)", style=solid]; -"807 view_22" -> "808 index_4" [label="(225, 12)", style=solid]; -"808 index_4" -> "809 view_23" [label="(4096, 12)", style=solid]; -"809 view_23" -> "810 permute_19" [label="(64, 64, 12)", style=solid]; -"810 permute_19" -> "811 contiguous_6" [label="(12, 64, 64)", style=solid]; -"811 contiguous_6" -> "812 unsqueeze_12" [label="(12, 64, 64)", style=solid]; -"812 unsqueeze_12" -> "813 sigmoid_4" [label="(1, 12, 64, 64)", style=solid]; -"813 sigmoid_4" -> "814 mul_8" [label="(1, 12, 64, 64)", style=solid]; -"814 mul_8" -> "850 add_14" [label="(1, 12, 64, 64)", style=solid]; -"815 pad_6" -> "816 view_24" [label="(1, 16, 16, 384)", style=solid]; -"816 view_24" -> "817 permute_20" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"817 permute_20" -> "818 reshape_18" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"818 reshape_18" -> "820 reshape_18_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"819 clone_4" -> "827 linear_28" [label="(1152,)", style=solid]; -"820 reshape_18_0_0_nncf_smooth_quant_0" -> "821 quantize_per_tensor_default_27" [label="(4, 64, 384)", style=solid]; -"821 quantize_per_tensor_default_27" -> "822 dequantize_per_tensor_default_27" [label="(4, 64, 384)", style=solid]; -"822 dequantize_per_tensor_default_27" -> "827 linear_28" [label="(4, 64, 384)", style=solid]; -"823 linear_28_scale_0" -> "826 dequantize_per_channel_default_29" [label="(1152,)", style=solid]; -"824 linear_28_zero_point_0" -> "826 dequantize_per_channel_default_29" [label="(1152,)", style=solid]; -"825 _frozen_param29" -> "826 dequantize_per_channel_default_29" [label="(1152, 384)", style=solid]; -"826 dequantize_per_channel_default_29" -> "827 linear_28" [label="(1152, 384)", style=solid]; -"827 linear_28" -> "828 reshape_19" [label="(4, 64, 1152)", style=solid]; -"828 reshape_19" -> "829 permute_21" [label="(4, 64, 3, 12, 32)", style=solid]; -"829 permute_21" -> "830 select_12" [label="(3, 4, 12, 64, 32)", style=solid]; -"829 permute_21" -> "831 select_13" [label="(3, 4, 12, 64, 32)", style=solid]; -"829 permute_21" -> "832 select_14" [label="(3, 4, 12, 64, 32)", style=solid]; -"830 select_12" -> "833 linalg_vector_norm_8" [label="(4, 12, 64, 32)", style=solid]; -"830 select_12" -> "835 expand_as_8" [label="(4, 12, 64, 32)", style=solid]; -"830 select_12" -> "836 div_8" [label="(4, 12, 64, 32)", style=solid]; -"831 select_13" -> "839 linalg_vector_norm_9" [label="(4, 12, 64, 32)", style=solid]; -"831 select_13" -> "841 expand_as_9" [label="(4, 12, 64, 32)", style=solid]; -"831 select_13" -> "842 div_9" [label="(4, 12, 64, 32)", style=solid]; -"832 select_14" -> "853 matmul_9" [label="(4, 12, 64, 32)", style=solid]; -"833 linalg_vector_norm_8" -> "834 clamp_min_8" [label="(4, 12, 64, 1)", style=solid]; -"834 clamp_min_8" -> "835 expand_as_8" [label="(4, 12, 64, 1)", style=solid]; -"835 expand_as_8" -> "836 div_8" [label="(4, 12, 64, 32)", style=solid]; -"836 div_8" -> "837 quantize_per_tensor_default_28" [label="(4, 12, 64, 32)", style=solid]; -"837 quantize_per_tensor_default_28" -> "838 dequantize_per_tensor_default_28" [label="(4, 12, 64, 32)", style=solid]; -"838 dequantize_per_tensor_default_28" -> "846 matmul_8" [label="(4, 12, 64, 32)", style=solid]; -"839 linalg_vector_norm_9" -> "840 clamp_min_9" [label="(4, 12, 64, 1)", style=solid]; -"840 clamp_min_9" -> "841 expand_as_9" [label="(4, 12, 64, 1)", style=solid]; -"841 expand_as_9" -> "842 div_9" [label="(4, 12, 64, 32)", style=solid]; -"842 div_9" -> "843 quantize_per_tensor_default_29" [label="(4, 12, 64, 32)", style=solid]; -"843 quantize_per_tensor_default_29" -> "844 dequantize_per_tensor_default_29" [label="(4, 12, 64, 32)", style=solid]; -"844 dequantize_per_tensor_default_29" -> "845 transpose_8" [label="(4, 12, 64, 32)", style=solid]; -"845 transpose_8" -> "846 matmul_8" [label="(4, 12, 32, 64)", style=solid]; -"846 matmul_8" -> "849 mul_9" [label="(4, 12, 64, 64)", style=solid]; -"847 clamp_4" -> "848 exp_4" [label="(12, 1, 1)", style=solid]; -"848 exp_4" -> "849 mul_9" [label="(12, 1, 1)", style=solid]; -"849 mul_9" -> "850 add_14" [label="(4, 12, 64, 64)", style=solid]; -"850 add_14" -> "851 softmax_4" [label="(4, 12, 64, 64)", style=solid]; -"851 softmax_4" -> "852 dropout_16" [label="(4, 12, 64, 64)", style=solid]; -"852 dropout_16" -> "853 matmul_9" [label="(4, 12, 64, 64)", style=solid]; -"853 matmul_9" -> "854 transpose_9" [label="(4, 12, 64, 32)", style=solid]; -"854 transpose_9" -> "855 reshape_20" [label="(4, 64, 12, 32)", style=solid]; -"855 reshape_20" -> "856 reshape_20_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"856 reshape_20_0_0_nncf_smooth_quant_0" -> "857 quantize_per_tensor_default_30" [label="(4, 64, 384)", style=solid]; -"857 quantize_per_tensor_default_30" -> "858 dequantize_per_tensor_default_30" [label="(4, 64, 384)", style=solid]; -"858 dequantize_per_tensor_default_30" -> "863 linear_29" [label="(4, 64, 384)", style=solid]; -"859 linear_29_scale_0" -> "862 dequantize_per_channel_default_30" [label="(384,)", style=solid]; -"860 linear_29_zero_point_0" -> "862 dequantize_per_channel_default_30" [label="(384,)", style=solid]; -"861 _frozen_param30" -> "862 dequantize_per_channel_default_30" [label="(384, 384)", style=solid]; -"862 dequantize_per_channel_default_30" -> "863 linear_29" [label="(384, 384)", style=solid]; -"863 linear_29" -> "864 dropout_17" [label="(4, 64, 384)", style=solid]; -"864 dropout_17" -> "865 view_25" [label="(4, 64, 384)", style=solid]; -"865 view_25" -> "866 permute_22" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"866 permute_22" -> "867 reshape_21" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"867 reshape_21" -> "868 slice_78" [label="(1, 16, 16, 384)", style=solid]; -"868 slice_78" -> "869 slice_79" [label="(1, 16, 16, 384)", style=solid]; -"869 slice_79" -> "870 slice_80" [label="(1, 14, 16, 384)", style=solid]; -"870 slice_80" -> "871 slice_81" [label="(1, 14, 14, 384)", style=solid]; -"871 slice_81" -> "872 contiguous_7" [label="(1, 14, 14, 384)", style=solid]; -"872 contiguous_7" -> "873 layer_norm_11" [label="(1, 14, 14, 384)", style=solid]; -"873 layer_norm_11" -> "874 add_15" [label="(1, 14, 14, 384)", style=solid]; -"874 add_15" -> "875 add_15_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; -"874 add_15" -> "895 add_16" [label="(1, 14, 14, 384)", style=solid]; -"875 add_15_0_0_nncf_smooth_quant_0" -> "876 quantize_per_tensor_default_31" [label="(1, 14, 14, 384)", style=solid]; -"876 quantize_per_tensor_default_31" -> "877 dequantize_per_tensor_default_31" [label="(1, 14, 14, 384)", style=solid]; -"877 dequantize_per_tensor_default_31" -> "882 linear_30" [label="(1, 14, 14, 384)", style=solid]; -"878 linear_30_scale_0" -> "881 dequantize_per_channel_default_31" [label="(1536,)", style=solid]; -"879 linear_30_zero_point_0" -> "881 dequantize_per_channel_default_31" [label="(1536,)", style=solid]; -"880 _frozen_param31" -> "881 dequantize_per_channel_default_31" [label="(1536, 384)", style=solid]; -"881 dequantize_per_channel_default_31" -> "882 linear_30" [label="(1536, 384)", style=solid]; -"882 linear_30" -> "883 gelu_4" [label="(1, 14, 14, 1536)", style=solid]; -"883 gelu_4" -> "884 dropout_18" [label="(1, 14, 14, 1536)", style=solid]; -"884 dropout_18" -> "885 dropout_18_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; -"885 dropout_18_0_0_nncf_smooth_quant_0" -> "886 quantize_per_tensor_default_32" [label="(1, 14, 14, 1536)", style=solid]; -"886 quantize_per_tensor_default_32" -> "887 dequantize_per_tensor_default_32" [label="(1, 14, 14, 1536)", style=solid]; -"887 dequantize_per_tensor_default_32" -> "892 linear_31" [label="(1, 14, 14, 1536)", style=solid]; -"888 linear_31_scale_0" -> "891 dequantize_per_channel_default_32" [label="(384,)", style=solid]; -"889 linear_31_zero_point_0" -> "891 dequantize_per_channel_default_32" [label="(384,)", style=solid]; -"890 _frozen_param32" -> "891 dequantize_per_channel_default_32" [label="(384, 1536)", style=solid]; -"891 dequantize_per_channel_default_32" -> "892 linear_31" [label="(384, 1536)", style=solid]; -"892 linear_31" -> "893 dropout_19" [label="(1, 14, 14, 384)", style=solid]; -"893 dropout_19" -> "894 layer_norm_12" [label="(1, 14, 14, 384)", style=solid]; -"894 layer_norm_12" -> "895 add_16" [label="(1, 14, 14, 384)", style=solid]; -"895 add_16" -> "917 pad_7" [label="(1, 14, 14, 384)", style=solid]; -"895 add_16" -> "994 add_19" [label="(1, 14, 14, 384)", style=solid]; -"896 features_5_1_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "901 linear_32" [label="(1, 15, 15, 2)", style=solid]; -"897 linear_32_scale_0" -> "900 dequantize_per_channel_default_33" [label="(512,)", style=solid]; -"898 linear_32_zero_point_0" -> "900 dequantize_per_channel_default_33" [label="(512,)", style=solid]; -"899 _frozen_param33" -> "900 dequantize_per_channel_default_33" [label="(512, 2)", style=solid]; -"900 dequantize_per_channel_default_33" -> "901 linear_32" [label="(512, 2)", style=solid]; -"901 linear_32" -> "902 relu__5" [label="(1, 15, 15, 512)", style=solid]; -"902 relu__5" -> "903 relu__5_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; -"903 relu__5_0_0_nncf_smooth_quant_0" -> "908 linear_33" [label="(1, 15, 15, 512)", style=solid]; -"904 linear_33_scale_0" -> "907 dequantize_per_channel_default_34" [label="(12,)", style=solid]; -"905 linear_33_zero_point_0" -> "907 dequantize_per_channel_default_34" [label="(12,)", style=solid]; -"906 _frozen_param34" -> "907 dequantize_per_channel_default_34" [label="(12, 512)", style=solid]; -"907 dequantize_per_channel_default_34" -> "908 linear_33" [label="(12, 512)", style=solid]; -"908 linear_33" -> "909 view_26" [label="(1, 15, 15, 12)", style=solid]; -"909 view_26" -> "910 index_5" [label="(225, 12)", style=solid]; -"910 index_5" -> "911 view_27" [label="(4096, 12)", style=solid]; -"911 view_27" -> "912 permute_23" [label="(64, 64, 12)", style=solid]; -"912 permute_23" -> "913 contiguous_8" [label="(12, 64, 64)", style=solid]; -"913 contiguous_8" -> "914 unsqueeze_13" [label="(12, 64, 64)", style=solid]; -"914 unsqueeze_13" -> "915 sigmoid_5" [label="(1, 12, 64, 64)", style=solid]; -"915 sigmoid_5" -> "916 mul_10" [label="(1, 12, 64, 64)", style=solid]; -"916 mul_10" -> "953 add_17" [label="(1, 12, 64, 64)", style=solid]; -"917 pad_7" -> "918 roll_4" [label="(1, 16, 16, 384)", style=solid]; -"918 roll_4" -> "919 view_28" [label="(1, 16, 16, 384)", style=solid]; -"919 view_28" -> "920 permute_24" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"920 permute_24" -> "921 reshape_22" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"921 reshape_22" -> "923 reshape_22_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"921 reshape_22" -> "954 new_zeros_2" [label="(4, 64, 384)", style=solid]; -"922 clone_5" -> "930 linear_34" [label="(1152,)", style=solid]; -"923 reshape_22_0_0_nncf_smooth_quant_0" -> "924 quantize_per_tensor_default_33" [label="(4, 64, 384)", style=solid]; -"924 quantize_per_tensor_default_33" -> "925 dequantize_per_tensor_default_33" [label="(4, 64, 384)", style=solid]; -"925 dequantize_per_tensor_default_33" -> "930 linear_34" [label="(4, 64, 384)", style=solid]; -"926 linear_34_scale_0" -> "929 dequantize_per_channel_default_35" [label="(1152,)", style=solid]; -"927 linear_34_zero_point_0" -> "929 dequantize_per_channel_default_35" [label="(1152,)", style=solid]; -"928 _frozen_param35" -> "929 dequantize_per_channel_default_35" [label="(1152, 384)", style=solid]; -"929 dequantize_per_channel_default_35" -> "930 linear_34" [label="(1152, 384)", style=solid]; -"930 linear_34" -> "931 reshape_23" [label="(4, 64, 1152)", style=solid]; -"931 reshape_23" -> "932 permute_25" [label="(4, 64, 3, 12, 32)", style=solid]; -"932 permute_25" -> "933 select_15" [label="(3, 4, 12, 64, 32)", style=solid]; -"932 permute_25" -> "934 select_16" [label="(3, 4, 12, 64, 32)", style=solid]; -"932 permute_25" -> "935 select_17" [label="(3, 4, 12, 64, 32)", style=solid]; -"933 select_15" -> "936 linalg_vector_norm_10" [label="(4, 12, 64, 32)", style=solid]; -"933 select_15" -> "938 expand_as_10" [label="(4, 12, 64, 32)", style=solid]; -"933 select_15" -> "939 div_10" [label="(4, 12, 64, 32)", style=solid]; -"934 select_16" -> "942 linalg_vector_norm_11" [label="(4, 12, 64, 32)", style=solid]; -"934 select_16" -> "944 expand_as_11" [label="(4, 12, 64, 32)", style=solid]; -"934 select_16" -> "945 div_11" [label="(4, 12, 64, 32)", style=solid]; -"935 select_17" -> "972 matmul_11" [label="(4, 12, 64, 32)", style=solid]; -"936 linalg_vector_norm_10" -> "937 clamp_min_10" [label="(4, 12, 64, 1)", style=solid]; -"937 clamp_min_10" -> "938 expand_as_10" [label="(4, 12, 64, 1)", style=solid]; -"938 expand_as_10" -> "939 div_10" [label="(4, 12, 64, 32)", style=solid]; -"939 div_10" -> "940 quantize_per_tensor_default_34" [label="(4, 12, 64, 32)", style=solid]; -"940 quantize_per_tensor_default_34" -> "941 dequantize_per_tensor_default_34" [label="(4, 12, 64, 32)", style=solid]; -"941 dequantize_per_tensor_default_34" -> "949 matmul_10" [label="(4, 12, 64, 32)", style=solid]; -"942 linalg_vector_norm_11" -> "943 clamp_min_11" [label="(4, 12, 64, 1)", style=solid]; -"943 clamp_min_11" -> "944 expand_as_11" [label="(4, 12, 64, 1)", style=solid]; -"944 expand_as_11" -> "945 div_11" [label="(4, 12, 64, 32)", style=solid]; -"945 div_11" -> "946 quantize_per_tensor_default_35" [label="(4, 12, 64, 32)", style=solid]; -"946 quantize_per_tensor_default_35" -> "947 dequantize_per_tensor_default_35" [label="(4, 12, 64, 32)", style=solid]; -"947 dequantize_per_tensor_default_35" -> "948 transpose_10" [label="(4, 12, 64, 32)", style=solid]; -"948 transpose_10" -> "949 matmul_10" [label="(4, 12, 32, 64)", style=solid]; -"949 matmul_10" -> "952 mul_11" [label="(4, 12, 64, 64)", style=solid]; -"950 clamp_5" -> "951 exp_5" [label="(12, 1, 1)", style=solid]; -"951 exp_5" -> "952 mul_11" [label="(12, 1, 1)", style=solid]; -"952 mul_11" -> "953 add_17" [label="(4, 12, 64, 64)", style=solid]; -"953 add_17" -> "965 view_30" [label="(4, 12, 64, 64)", style=solid]; -"954 new_zeros_2" -> "955 view_29" [label="(16, 16)", style=solid]; -"955 view_29" -> "956 permute_26" [label="(2, 8, 2, 8)", style=solid]; -"956 permute_26" -> "957 reshape_24" [label="(2, 2, 8, 8)", style=solid]; -"957 reshape_24" -> "958 unsqueeze_14" [label="(4, 64)", style=solid]; -"957 reshape_24" -> "959 unsqueeze_15" [label="(4, 64)", style=solid]; -"958 unsqueeze_14" -> "960 sub_2" [label="(4, 1, 64)", style=solid]; -"959 unsqueeze_15" -> "960 sub_2" [label="(4, 64, 1)", style=solid]; -"960 sub_2" -> "961 ne_2" [label="(4, 64, 64)", style=solid]; -"960 sub_2" -> "962 masked_fill_4" [label="(4, 64, 64)", style=solid]; -"960 sub_2" -> "963 eq_2" [label="(4, 64, 64)", style=solid]; -"961 ne_2" -> "962 masked_fill_4" [label="(4, 64, 64)", style=solid]; -"962 masked_fill_4" -> "964 masked_fill_5" [label="(4, 64, 64)", style=solid]; -"963 eq_2" -> "964 masked_fill_5" [label="(4, 64, 64)", style=solid]; -"964 masked_fill_5" -> "966 unsqueeze_16" [label="(4, 64, 64)", style=solid]; -"965 view_30" -> "968 add_18" [label="(1, 4, 12, 64, 64)", style=solid]; -"966 unsqueeze_16" -> "967 unsqueeze_17" [label="(4, 1, 64, 64)", style=solid]; -"967 unsqueeze_17" -> "968 add_18" [label="(1, 4, 1, 64, 64)", style=solid]; -"968 add_18" -> "969 view_31" [label="(1, 4, 12, 64, 64)", style=solid]; -"969 view_31" -> "970 softmax_5" [label="(4, 12, 64, 64)", style=solid]; -"970 softmax_5" -> "971 dropout_20" [label="(4, 12, 64, 64)", style=solid]; -"971 dropout_20" -> "972 matmul_11" [label="(4, 12, 64, 64)", style=solid]; -"972 matmul_11" -> "973 transpose_11" [label="(4, 12, 64, 32)", style=solid]; -"973 transpose_11" -> "974 reshape_25" [label="(4, 64, 12, 32)", style=solid]; -"974 reshape_25" -> "975 reshape_25_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"975 reshape_25_0_0_nncf_smooth_quant_0" -> "976 quantize_per_tensor_default_36" [label="(4, 64, 384)", style=solid]; -"976 quantize_per_tensor_default_36" -> "977 dequantize_per_tensor_default_36" [label="(4, 64, 384)", style=solid]; -"977 dequantize_per_tensor_default_36" -> "982 linear_35" [label="(4, 64, 384)", style=solid]; -"978 linear_35_scale_0" -> "981 dequantize_per_channel_default_36" [label="(384,)", style=solid]; -"979 linear_35_zero_point_0" -> "981 dequantize_per_channel_default_36" [label="(384,)", style=solid]; -"980 _frozen_param36" -> "981 dequantize_per_channel_default_36" [label="(384, 384)", style=solid]; -"981 dequantize_per_channel_default_36" -> "982 linear_35" [label="(384, 384)", style=solid]; -"982 linear_35" -> "983 dropout_21" [label="(4, 64, 384)", style=solid]; -"983 dropout_21" -> "984 view_32" [label="(4, 64, 384)", style=solid]; -"984 view_32" -> "985 permute_27" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"985 permute_27" -> "986 reshape_26" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"986 reshape_26" -> "987 roll_5" [label="(1, 16, 16, 384)", style=solid]; -"987 roll_5" -> "988 slice_101" [label="(1, 16, 16, 384)", style=solid]; -"988 slice_101" -> "989 slice_102" [label="(1, 16, 16, 384)", style=solid]; -"989 slice_102" -> "990 slice_103" [label="(1, 14, 16, 384)", style=solid]; -"990 slice_103" -> "991 slice_104" [label="(1, 14, 14, 384)", style=solid]; -"991 slice_104" -> "992 contiguous_9" [label="(1, 14, 14, 384)", style=solid]; -"992 contiguous_9" -> "993 layer_norm_13" [label="(1, 14, 14, 384)", style=solid]; -"993 layer_norm_13" -> "994 add_19" [label="(1, 14, 14, 384)", style=solid]; -"994 add_19" -> "995 add_19_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; -"994 add_19" -> "1015 add_20" [label="(1, 14, 14, 384)", style=solid]; -"995 add_19_0_0_nncf_smooth_quant_0" -> "996 quantize_per_tensor_default_37" [label="(1, 14, 14, 384)", style=solid]; -"996 quantize_per_tensor_default_37" -> "997 dequantize_per_tensor_default_37" [label="(1, 14, 14, 384)", style=solid]; -"997 dequantize_per_tensor_default_37" -> "1002 linear_36" [label="(1, 14, 14, 384)", style=solid]; -"998 linear_36_scale_0" -> "1001 dequantize_per_channel_default_37" [label="(1536,)", style=solid]; -"999 linear_36_zero_point_0" -> "1001 dequantize_per_channel_default_37" [label="(1536,)", style=solid]; -"1000 _frozen_param37" -> "1001 dequantize_per_channel_default_37" [label="(1536, 384)", style=solid]; -"1001 dequantize_per_channel_default_37" -> "1002 linear_36" [label="(1536, 384)", style=solid]; -"1002 linear_36" -> "1003 gelu_5" [label="(1, 14, 14, 1536)", style=solid]; -"1003 gelu_5" -> "1004 dropout_22" [label="(1, 14, 14, 1536)", style=solid]; -"1004 dropout_22" -> "1005 dropout_22_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; -"1005 dropout_22_0_0_nncf_smooth_quant_0" -> "1006 quantize_per_tensor_default_38" [label="(1, 14, 14, 1536)", style=solid]; -"1006 quantize_per_tensor_default_38" -> "1007 dequantize_per_tensor_default_38" [label="(1, 14, 14, 1536)", style=solid]; -"1007 dequantize_per_tensor_default_38" -> "1012 linear_37" [label="(1, 14, 14, 1536)", style=solid]; -"1008 linear_37_scale_0" -> "1011 dequantize_per_channel_default_38" [label="(384,)", style=solid]; -"1009 linear_37_zero_point_0" -> "1011 dequantize_per_channel_default_38" [label="(384,)", style=solid]; -"1010 _frozen_param38" -> "1011 dequantize_per_channel_default_38" [label="(384, 1536)", style=solid]; -"1011 dequantize_per_channel_default_38" -> "1012 linear_37" [label="(384, 1536)", style=solid]; -"1012 linear_37" -> "1013 dropout_23" [label="(1, 14, 14, 384)", style=solid]; -"1013 dropout_23" -> "1014 layer_norm_14" [label="(1, 14, 14, 384)", style=solid]; -"1014 layer_norm_14" -> "1015 add_20" [label="(1, 14, 14, 384)", style=solid]; -"1015 add_20" -> "1037 pad_8" [label="(1, 14, 14, 384)", style=solid]; -"1015 add_20" -> "1096 add_22" [label="(1, 14, 14, 384)", style=solid]; -"1016 features_5_2_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "1021 linear_38" [label="(1, 15, 15, 2)", style=solid]; -"1017 linear_38_scale_0" -> "1020 dequantize_per_channel_default_39" [label="(512,)", style=solid]; -"1018 linear_38_zero_point_0" -> "1020 dequantize_per_channel_default_39" [label="(512,)", style=solid]; -"1019 _frozen_param39" -> "1020 dequantize_per_channel_default_39" [label="(512, 2)", style=solid]; -"1020 dequantize_per_channel_default_39" -> "1021 linear_38" [label="(512, 2)", style=solid]; -"1021 linear_38" -> "1022 relu__6" [label="(1, 15, 15, 512)", style=solid]; -"1022 relu__6" -> "1023 relu__6_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; -"1023 relu__6_0_0_nncf_smooth_quant_0" -> "1028 linear_39" [label="(1, 15, 15, 512)", style=solid]; -"1024 linear_39_scale_0" -> "1027 dequantize_per_channel_default_40" [label="(12,)", style=solid]; -"1025 linear_39_zero_point_0" -> "1027 dequantize_per_channel_default_40" [label="(12,)", style=solid]; -"1026 _frozen_param40" -> "1027 dequantize_per_channel_default_40" [label="(12, 512)", style=solid]; -"1027 dequantize_per_channel_default_40" -> "1028 linear_39" [label="(12, 512)", style=solid]; -"1028 linear_39" -> "1029 view_33" [label="(1, 15, 15, 12)", style=solid]; -"1029 view_33" -> "1030 index_6" [label="(225, 12)", style=solid]; -"1030 index_6" -> "1031 view_34" [label="(4096, 12)", style=solid]; -"1031 view_34" -> "1032 permute_28" [label="(64, 64, 12)", style=solid]; -"1032 permute_28" -> "1033 contiguous_10" [label="(12, 64, 64)", style=solid]; -"1033 contiguous_10" -> "1034 unsqueeze_18" [label="(12, 64, 64)", style=solid]; -"1034 unsqueeze_18" -> "1035 sigmoid_6" [label="(1, 12, 64, 64)", style=solid]; -"1035 sigmoid_6" -> "1036 mul_12" [label="(1, 12, 64, 64)", style=solid]; -"1036 mul_12" -> "1072 add_21" [label="(1, 12, 64, 64)", style=solid]; -"1037 pad_8" -> "1038 view_35" [label="(1, 16, 16, 384)", style=solid]; -"1038 view_35" -> "1039 permute_29" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"1039 permute_29" -> "1040 reshape_27" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"1040 reshape_27" -> "1042 reshape_27_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"1041 clone_6" -> "1049 linear_40" [label="(1152,)", style=solid]; -"1042 reshape_27_0_0_nncf_smooth_quant_0" -> "1043 quantize_per_tensor_default_39" [label="(4, 64, 384)", style=solid]; -"1043 quantize_per_tensor_default_39" -> "1044 dequantize_per_tensor_default_39" [label="(4, 64, 384)", style=solid]; -"1044 dequantize_per_tensor_default_39" -> "1049 linear_40" [label="(4, 64, 384)", style=solid]; -"1045 linear_40_scale_0" -> "1048 dequantize_per_channel_default_41" [label="(1152,)", style=solid]; -"1046 linear_40_zero_point_0" -> "1048 dequantize_per_channel_default_41" [label="(1152,)", style=solid]; -"1047 _frozen_param41" -> "1048 dequantize_per_channel_default_41" [label="(1152, 384)", style=solid]; -"1048 dequantize_per_channel_default_41" -> "1049 linear_40" [label="(1152, 384)", style=solid]; -"1049 linear_40" -> "1050 reshape_28" [label="(4, 64, 1152)", style=solid]; -"1050 reshape_28" -> "1051 permute_30" [label="(4, 64, 3, 12, 32)", style=solid]; -"1051 permute_30" -> "1052 select_18" [label="(3, 4, 12, 64, 32)", style=solid]; -"1051 permute_30" -> "1053 select_19" [label="(3, 4, 12, 64, 32)", style=solid]; -"1051 permute_30" -> "1054 select_20" [label="(3, 4, 12, 64, 32)", style=solid]; -"1052 select_18" -> "1055 linalg_vector_norm_12" [label="(4, 12, 64, 32)", style=solid]; -"1052 select_18" -> "1057 expand_as_12" [label="(4, 12, 64, 32)", style=solid]; -"1052 select_18" -> "1058 div_12" [label="(4, 12, 64, 32)", style=solid]; -"1053 select_19" -> "1061 linalg_vector_norm_13" [label="(4, 12, 64, 32)", style=solid]; -"1053 select_19" -> "1063 expand_as_13" [label="(4, 12, 64, 32)", style=solid]; -"1053 select_19" -> "1064 div_13" [label="(4, 12, 64, 32)", style=solid]; -"1054 select_20" -> "1075 matmul_13" [label="(4, 12, 64, 32)", style=solid]; -"1055 linalg_vector_norm_12" -> "1056 clamp_min_12" [label="(4, 12, 64, 1)", style=solid]; -"1056 clamp_min_12" -> "1057 expand_as_12" [label="(4, 12, 64, 1)", style=solid]; -"1057 expand_as_12" -> "1058 div_12" [label="(4, 12, 64, 32)", style=solid]; -"1058 div_12" -> "1059 quantize_per_tensor_default_40" [label="(4, 12, 64, 32)", style=solid]; -"1059 quantize_per_tensor_default_40" -> "1060 dequantize_per_tensor_default_40" [label="(4, 12, 64, 32)", style=solid]; -"1060 dequantize_per_tensor_default_40" -> "1068 matmul_12" [label="(4, 12, 64, 32)", style=solid]; -"1061 linalg_vector_norm_13" -> "1062 clamp_min_13" [label="(4, 12, 64, 1)", style=solid]; -"1062 clamp_min_13" -> "1063 expand_as_13" [label="(4, 12, 64, 1)", style=solid]; -"1063 expand_as_13" -> "1064 div_13" [label="(4, 12, 64, 32)", style=solid]; -"1064 div_13" -> "1065 quantize_per_tensor_default_41" [label="(4, 12, 64, 32)", style=solid]; -"1065 quantize_per_tensor_default_41" -> "1066 dequantize_per_tensor_default_41" [label="(4, 12, 64, 32)", style=solid]; -"1066 dequantize_per_tensor_default_41" -> "1067 transpose_12" [label="(4, 12, 64, 32)", style=solid]; -"1067 transpose_12" -> "1068 matmul_12" [label="(4, 12, 32, 64)", style=solid]; -"1068 matmul_12" -> "1071 mul_13" [label="(4, 12, 64, 64)", style=solid]; -"1069 clamp_6" -> "1070 exp_6" [label="(12, 1, 1)", style=solid]; -"1070 exp_6" -> "1071 mul_13" [label="(12, 1, 1)", style=solid]; -"1071 mul_13" -> "1072 add_21" [label="(4, 12, 64, 64)", style=solid]; -"1072 add_21" -> "1073 softmax_6" [label="(4, 12, 64, 64)", style=solid]; -"1073 softmax_6" -> "1074 dropout_24" [label="(4, 12, 64, 64)", style=solid]; -"1074 dropout_24" -> "1075 matmul_13" [label="(4, 12, 64, 64)", style=solid]; -"1075 matmul_13" -> "1076 transpose_13" [label="(4, 12, 64, 32)", style=solid]; -"1076 transpose_13" -> "1077 reshape_29" [label="(4, 64, 12, 32)", style=solid]; -"1077 reshape_29" -> "1078 reshape_29_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"1078 reshape_29_0_0_nncf_smooth_quant_0" -> "1079 quantize_per_tensor_default_42" [label="(4, 64, 384)", style=solid]; -"1079 quantize_per_tensor_default_42" -> "1080 dequantize_per_tensor_default_42" [label="(4, 64, 384)", style=solid]; -"1080 dequantize_per_tensor_default_42" -> "1085 linear_41" [label="(4, 64, 384)", style=solid]; -"1081 linear_41_scale_0" -> "1084 dequantize_per_channel_default_42" [label="(384,)", style=solid]; -"1082 linear_41_zero_point_0" -> "1084 dequantize_per_channel_default_42" [label="(384,)", style=solid]; -"1083 _frozen_param42" -> "1084 dequantize_per_channel_default_42" [label="(384, 384)", style=solid]; -"1084 dequantize_per_channel_default_42" -> "1085 linear_41" [label="(384, 384)", style=solid]; -"1085 linear_41" -> "1086 dropout_25" [label="(4, 64, 384)", style=solid]; -"1086 dropout_25" -> "1087 view_36" [label="(4, 64, 384)", style=solid]; -"1087 view_36" -> "1088 permute_31" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"1088 permute_31" -> "1089 reshape_30" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"1089 reshape_30" -> "1090 slice_106" [label="(1, 16, 16, 384)", style=solid]; -"1090 slice_106" -> "1091 slice_107" [label="(1, 16, 16, 384)", style=solid]; -"1091 slice_107" -> "1092 slice_108" [label="(1, 14, 16, 384)", style=solid]; -"1092 slice_108" -> "1093 slice_109" [label="(1, 14, 14, 384)", style=solid]; -"1093 slice_109" -> "1094 contiguous_11" [label="(1, 14, 14, 384)", style=solid]; -"1094 contiguous_11" -> "1095 layer_norm_15" [label="(1, 14, 14, 384)", style=solid]; -"1095 layer_norm_15" -> "1096 add_22" [label="(1, 14, 14, 384)", style=solid]; -"1096 add_22" -> "1097 add_22_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; -"1096 add_22" -> "1117 add_23" [label="(1, 14, 14, 384)", style=solid]; -"1097 add_22_0_0_nncf_smooth_quant_0" -> "1098 quantize_per_tensor_default_43" [label="(1, 14, 14, 384)", style=solid]; -"1098 quantize_per_tensor_default_43" -> "1099 dequantize_per_tensor_default_43" [label="(1, 14, 14, 384)", style=solid]; -"1099 dequantize_per_tensor_default_43" -> "1104 linear_42" [label="(1, 14, 14, 384)", style=solid]; -"1100 linear_42_scale_0" -> "1103 dequantize_per_channel_default_43" [label="(1536,)", style=solid]; -"1101 linear_42_zero_point_0" -> "1103 dequantize_per_channel_default_43" [label="(1536,)", style=solid]; -"1102 _frozen_param43" -> "1103 dequantize_per_channel_default_43" [label="(1536, 384)", style=solid]; -"1103 dequantize_per_channel_default_43" -> "1104 linear_42" [label="(1536, 384)", style=solid]; -"1104 linear_42" -> "1105 gelu_6" [label="(1, 14, 14, 1536)", style=solid]; -"1105 gelu_6" -> "1106 dropout_26" [label="(1, 14, 14, 1536)", style=solid]; -"1106 dropout_26" -> "1107 dropout_26_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; -"1107 dropout_26_0_0_nncf_smooth_quant_0" -> "1108 quantize_per_tensor_default_44" [label="(1, 14, 14, 1536)", style=solid]; -"1108 quantize_per_tensor_default_44" -> "1109 dequantize_per_tensor_default_44" [label="(1, 14, 14, 1536)", style=solid]; -"1109 dequantize_per_tensor_default_44" -> "1114 linear_43" [label="(1, 14, 14, 1536)", style=solid]; -"1110 linear_43_scale_0" -> "1113 dequantize_per_channel_default_44" [label="(384,)", style=solid]; -"1111 linear_43_zero_point_0" -> "1113 dequantize_per_channel_default_44" [label="(384,)", style=solid]; -"1112 _frozen_param44" -> "1113 dequantize_per_channel_default_44" [label="(384, 1536)", style=solid]; -"1113 dequantize_per_channel_default_44" -> "1114 linear_43" [label="(384, 1536)", style=solid]; -"1114 linear_43" -> "1115 dropout_27" [label="(1, 14, 14, 384)", style=solid]; -"1115 dropout_27" -> "1116 layer_norm_16" [label="(1, 14, 14, 384)", style=solid]; -"1116 layer_norm_16" -> "1117 add_23" [label="(1, 14, 14, 384)", style=solid]; -"1117 add_23" -> "1139 pad_9" [label="(1, 14, 14, 384)", style=solid]; -"1117 add_23" -> "1216 add_26" [label="(1, 14, 14, 384)", style=solid]; -"1118 features_5_3_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "1123 linear_44" [label="(1, 15, 15, 2)", style=solid]; -"1119 linear_44_scale_0" -> "1122 dequantize_per_channel_default_45" [label="(512,)", style=solid]; -"1120 linear_44_zero_point_0" -> "1122 dequantize_per_channel_default_45" [label="(512,)", style=solid]; -"1121 _frozen_param45" -> "1122 dequantize_per_channel_default_45" [label="(512, 2)", style=solid]; -"1122 dequantize_per_channel_default_45" -> "1123 linear_44" [label="(512, 2)", style=solid]; -"1123 linear_44" -> "1124 relu__7" [label="(1, 15, 15, 512)", style=solid]; -"1124 relu__7" -> "1125 relu__7_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; -"1125 relu__7_0_0_nncf_smooth_quant_0" -> "1130 linear_45" [label="(1, 15, 15, 512)", style=solid]; -"1126 linear_45_scale_0" -> "1129 dequantize_per_channel_default_46" [label="(12,)", style=solid]; -"1127 linear_45_zero_point_0" -> "1129 dequantize_per_channel_default_46" [label="(12,)", style=solid]; -"1128 _frozen_param46" -> "1129 dequantize_per_channel_default_46" [label="(12, 512)", style=solid]; -"1129 dequantize_per_channel_default_46" -> "1130 linear_45" [label="(12, 512)", style=solid]; -"1130 linear_45" -> "1131 view_37" [label="(1, 15, 15, 12)", style=solid]; -"1131 view_37" -> "1132 index_7" [label="(225, 12)", style=solid]; -"1132 index_7" -> "1133 view_38" [label="(4096, 12)", style=solid]; -"1133 view_38" -> "1134 permute_32" [label="(64, 64, 12)", style=solid]; -"1134 permute_32" -> "1135 contiguous_12" [label="(12, 64, 64)", style=solid]; -"1135 contiguous_12" -> "1136 unsqueeze_19" [label="(12, 64, 64)", style=solid]; -"1136 unsqueeze_19" -> "1137 sigmoid_7" [label="(1, 12, 64, 64)", style=solid]; -"1137 sigmoid_7" -> "1138 mul_14" [label="(1, 12, 64, 64)", style=solid]; -"1138 mul_14" -> "1175 add_24" [label="(1, 12, 64, 64)", style=solid]; -"1139 pad_9" -> "1140 roll_6" [label="(1, 16, 16, 384)", style=solid]; -"1140 roll_6" -> "1141 view_39" [label="(1, 16, 16, 384)", style=solid]; -"1141 view_39" -> "1142 permute_33" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"1142 permute_33" -> "1143 reshape_31" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"1143 reshape_31" -> "1145 reshape_31_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"1143 reshape_31" -> "1176 new_zeros_3" [label="(4, 64, 384)", style=solid]; -"1144 clone_7" -> "1152 linear_46" [label="(1152,)", style=solid]; -"1145 reshape_31_0_0_nncf_smooth_quant_0" -> "1146 quantize_per_tensor_default_45" [label="(4, 64, 384)", style=solid]; -"1146 quantize_per_tensor_default_45" -> "1147 dequantize_per_tensor_default_45" [label="(4, 64, 384)", style=solid]; -"1147 dequantize_per_tensor_default_45" -> "1152 linear_46" [label="(4, 64, 384)", style=solid]; -"1148 linear_46_scale_0" -> "1151 dequantize_per_channel_default_47" [label="(1152,)", style=solid]; -"1149 linear_46_zero_point_0" -> "1151 dequantize_per_channel_default_47" [label="(1152,)", style=solid]; -"1150 _frozen_param47" -> "1151 dequantize_per_channel_default_47" [label="(1152, 384)", style=solid]; -"1151 dequantize_per_channel_default_47" -> "1152 linear_46" [label="(1152, 384)", style=solid]; -"1152 linear_46" -> "1153 reshape_32" [label="(4, 64, 1152)", style=solid]; -"1153 reshape_32" -> "1154 permute_34" [label="(4, 64, 3, 12, 32)", style=solid]; -"1154 permute_34" -> "1155 select_21" [label="(3, 4, 12, 64, 32)", style=solid]; -"1154 permute_34" -> "1156 select_22" [label="(3, 4, 12, 64, 32)", style=solid]; -"1154 permute_34" -> "1157 select_23" [label="(3, 4, 12, 64, 32)", style=solid]; -"1155 select_21" -> "1158 linalg_vector_norm_14" [label="(4, 12, 64, 32)", style=solid]; -"1155 select_21" -> "1160 expand_as_14" [label="(4, 12, 64, 32)", style=solid]; -"1155 select_21" -> "1161 div_14" [label="(4, 12, 64, 32)", style=solid]; -"1156 select_22" -> "1164 linalg_vector_norm_15" [label="(4, 12, 64, 32)", style=solid]; -"1156 select_22" -> "1166 expand_as_15" [label="(4, 12, 64, 32)", style=solid]; -"1156 select_22" -> "1167 div_15" [label="(4, 12, 64, 32)", style=solid]; -"1157 select_23" -> "1194 matmul_15" [label="(4, 12, 64, 32)", style=solid]; -"1158 linalg_vector_norm_14" -> "1159 clamp_min_14" [label="(4, 12, 64, 1)", style=solid]; -"1159 clamp_min_14" -> "1160 expand_as_14" [label="(4, 12, 64, 1)", style=solid]; -"1160 expand_as_14" -> "1161 div_14" [label="(4, 12, 64, 32)", style=solid]; -"1161 div_14" -> "1162 quantize_per_tensor_default_46" [label="(4, 12, 64, 32)", style=solid]; -"1162 quantize_per_tensor_default_46" -> "1163 dequantize_per_tensor_default_46" [label="(4, 12, 64, 32)", style=solid]; -"1163 dequantize_per_tensor_default_46" -> "1171 matmul_14" [label="(4, 12, 64, 32)", style=solid]; -"1164 linalg_vector_norm_15" -> "1165 clamp_min_15" [label="(4, 12, 64, 1)", style=solid]; -"1165 clamp_min_15" -> "1166 expand_as_15" [label="(4, 12, 64, 1)", style=solid]; -"1166 expand_as_15" -> "1167 div_15" [label="(4, 12, 64, 32)", style=solid]; -"1167 div_15" -> "1168 quantize_per_tensor_default_47" [label="(4, 12, 64, 32)", style=solid]; -"1168 quantize_per_tensor_default_47" -> "1169 dequantize_per_tensor_default_47" [label="(4, 12, 64, 32)", style=solid]; -"1169 dequantize_per_tensor_default_47" -> "1170 transpose_14" [label="(4, 12, 64, 32)", style=solid]; -"1170 transpose_14" -> "1171 matmul_14" [label="(4, 12, 32, 64)", style=solid]; -"1171 matmul_14" -> "1174 mul_15" [label="(4, 12, 64, 64)", style=solid]; -"1172 clamp_7" -> "1173 exp_7" [label="(12, 1, 1)", style=solid]; -"1173 exp_7" -> "1174 mul_15" [label="(12, 1, 1)", style=solid]; -"1174 mul_15" -> "1175 add_24" [label="(4, 12, 64, 64)", style=solid]; -"1175 add_24" -> "1187 view_41" [label="(4, 12, 64, 64)", style=solid]; -"1176 new_zeros_3" -> "1177 view_40" [label="(16, 16)", style=solid]; -"1177 view_40" -> "1178 permute_35" [label="(2, 8, 2, 8)", style=solid]; -"1178 permute_35" -> "1179 reshape_33" [label="(2, 2, 8, 8)", style=solid]; -"1179 reshape_33" -> "1180 unsqueeze_20" [label="(4, 64)", style=solid]; -"1179 reshape_33" -> "1181 unsqueeze_21" [label="(4, 64)", style=solid]; -"1180 unsqueeze_20" -> "1182 sub_3" [label="(4, 1, 64)", style=solid]; -"1181 unsqueeze_21" -> "1182 sub_3" [label="(4, 64, 1)", style=solid]; -"1182 sub_3" -> "1183 ne_3" [label="(4, 64, 64)", style=solid]; -"1182 sub_3" -> "1184 masked_fill_6" [label="(4, 64, 64)", style=solid]; -"1182 sub_3" -> "1185 eq_3" [label="(4, 64, 64)", style=solid]; -"1183 ne_3" -> "1184 masked_fill_6" [label="(4, 64, 64)", style=solid]; -"1184 masked_fill_6" -> "1186 masked_fill_7" [label="(4, 64, 64)", style=solid]; -"1185 eq_3" -> "1186 masked_fill_7" [label="(4, 64, 64)", style=solid]; -"1186 masked_fill_7" -> "1188 unsqueeze_22" [label="(4, 64, 64)", style=solid]; -"1187 view_41" -> "1190 add_25" [label="(1, 4, 12, 64, 64)", style=solid]; -"1188 unsqueeze_22" -> "1189 unsqueeze_23" [label="(4, 1, 64, 64)", style=solid]; -"1189 unsqueeze_23" -> "1190 add_25" [label="(1, 4, 1, 64, 64)", style=solid]; -"1190 add_25" -> "1191 view_42" [label="(1, 4, 12, 64, 64)", style=solid]; -"1191 view_42" -> "1192 softmax_7" [label="(4, 12, 64, 64)", style=solid]; -"1192 softmax_7" -> "1193 dropout_28" [label="(4, 12, 64, 64)", style=solid]; -"1193 dropout_28" -> "1194 matmul_15" [label="(4, 12, 64, 64)", style=solid]; -"1194 matmul_15" -> "1195 transpose_15" [label="(4, 12, 64, 32)", style=solid]; -"1195 transpose_15" -> "1196 reshape_34" [label="(4, 64, 12, 32)", style=solid]; -"1196 reshape_34" -> "1197 reshape_34_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"1197 reshape_34_0_0_nncf_smooth_quant_0" -> "1198 quantize_per_tensor_default_48" [label="(4, 64, 384)", style=solid]; -"1198 quantize_per_tensor_default_48" -> "1199 dequantize_per_tensor_default_48" [label="(4, 64, 384)", style=solid]; -"1199 dequantize_per_tensor_default_48" -> "1204 linear_47" [label="(4, 64, 384)", style=solid]; -"1200 linear_47_scale_0" -> "1203 dequantize_per_channel_default_48" [label="(384,)", style=solid]; -"1201 linear_47_zero_point_0" -> "1203 dequantize_per_channel_default_48" [label="(384,)", style=solid]; -"1202 _frozen_param48" -> "1203 dequantize_per_channel_default_48" [label="(384, 384)", style=solid]; -"1203 dequantize_per_channel_default_48" -> "1204 linear_47" [label="(384, 384)", style=solid]; -"1204 linear_47" -> "1205 dropout_29" [label="(4, 64, 384)", style=solid]; -"1205 dropout_29" -> "1206 view_43" [label="(4, 64, 384)", style=solid]; -"1206 view_43" -> "1207 permute_36" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"1207 permute_36" -> "1208 reshape_35" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"1208 reshape_35" -> "1209 roll_7" [label="(1, 16, 16, 384)", style=solid]; -"1209 roll_7" -> "1210 slice_129" [label="(1, 16, 16, 384)", style=solid]; -"1210 slice_129" -> "1211 slice_130" [label="(1, 16, 16, 384)", style=solid]; -"1211 slice_130" -> "1212 slice_131" [label="(1, 14, 16, 384)", style=solid]; -"1212 slice_131" -> "1213 slice_132" [label="(1, 14, 14, 384)", style=solid]; -"1213 slice_132" -> "1214 contiguous_13" [label="(1, 14, 14, 384)", style=solid]; -"1214 contiguous_13" -> "1215 layer_norm_17" [label="(1, 14, 14, 384)", style=solid]; -"1215 layer_norm_17" -> "1216 add_26" [label="(1, 14, 14, 384)", style=solid]; -"1216 add_26" -> "1217 add_26_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; -"1216 add_26" -> "1237 add_27" [label="(1, 14, 14, 384)", style=solid]; -"1217 add_26_0_0_nncf_smooth_quant_0" -> "1218 quantize_per_tensor_default_49" [label="(1, 14, 14, 384)", style=solid]; -"1218 quantize_per_tensor_default_49" -> "1219 dequantize_per_tensor_default_49" [label="(1, 14, 14, 384)", style=solid]; -"1219 dequantize_per_tensor_default_49" -> "1224 linear_48" [label="(1, 14, 14, 384)", style=solid]; -"1220 linear_48_scale_0" -> "1223 dequantize_per_channel_default_49" [label="(1536,)", style=solid]; -"1221 linear_48_zero_point_0" -> "1223 dequantize_per_channel_default_49" [label="(1536,)", style=solid]; -"1222 _frozen_param49" -> "1223 dequantize_per_channel_default_49" [label="(1536, 384)", style=solid]; -"1223 dequantize_per_channel_default_49" -> "1224 linear_48" [label="(1536, 384)", style=solid]; -"1224 linear_48" -> "1225 gelu_7" [label="(1, 14, 14, 1536)", style=solid]; -"1225 gelu_7" -> "1226 dropout_30" [label="(1, 14, 14, 1536)", style=solid]; -"1226 dropout_30" -> "1227 dropout_30_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; -"1227 dropout_30_0_0_nncf_smooth_quant_0" -> "1228 quantize_per_tensor_default_50" [label="(1, 14, 14, 1536)", style=solid]; -"1228 quantize_per_tensor_default_50" -> "1229 dequantize_per_tensor_default_50" [label="(1, 14, 14, 1536)", style=solid]; -"1229 dequantize_per_tensor_default_50" -> "1234 linear_49" [label="(1, 14, 14, 1536)", style=solid]; -"1230 linear_49_scale_0" -> "1233 dequantize_per_channel_default_50" [label="(384,)", style=solid]; -"1231 linear_49_zero_point_0" -> "1233 dequantize_per_channel_default_50" [label="(384,)", style=solid]; -"1232 _frozen_param50" -> "1233 dequantize_per_channel_default_50" [label="(384, 1536)", style=solid]; -"1233 dequantize_per_channel_default_50" -> "1234 linear_49" [label="(384, 1536)", style=solid]; -"1234 linear_49" -> "1235 dropout_31" [label="(1, 14, 14, 384)", style=solid]; -"1235 dropout_31" -> "1236 layer_norm_18" [label="(1, 14, 14, 384)", style=solid]; -"1236 layer_norm_18" -> "1237 add_27" [label="(1, 14, 14, 384)", style=solid]; -"1237 add_27" -> "1259 pad_10" [label="(1, 14, 14, 384)", style=solid]; -"1237 add_27" -> "1318 add_29" [label="(1, 14, 14, 384)", style=solid]; -"1238 features_5_4_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "1243 linear_50" [label="(1, 15, 15, 2)", style=solid]; -"1239 linear_50_scale_0" -> "1242 dequantize_per_channel_default_51" [label="(512,)", style=solid]; -"1240 linear_50_zero_point_0" -> "1242 dequantize_per_channel_default_51" [label="(512,)", style=solid]; -"1241 _frozen_param51" -> "1242 dequantize_per_channel_default_51" [label="(512, 2)", style=solid]; -"1242 dequantize_per_channel_default_51" -> "1243 linear_50" [label="(512, 2)", style=solid]; -"1243 linear_50" -> "1244 relu__8" [label="(1, 15, 15, 512)", style=solid]; -"1244 relu__8" -> "1245 relu__8_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; -"1245 relu__8_0_0_nncf_smooth_quant_0" -> "1250 linear_51" [label="(1, 15, 15, 512)", style=solid]; -"1246 linear_51_scale_0" -> "1249 dequantize_per_channel_default_52" [label="(12,)", style=solid]; -"1247 linear_51_zero_point_0" -> "1249 dequantize_per_channel_default_52" [label="(12,)", style=solid]; -"1248 _frozen_param52" -> "1249 dequantize_per_channel_default_52" [label="(12, 512)", style=solid]; -"1249 dequantize_per_channel_default_52" -> "1250 linear_51" [label="(12, 512)", style=solid]; -"1250 linear_51" -> "1251 view_44" [label="(1, 15, 15, 12)", style=solid]; -"1251 view_44" -> "1252 index_8" [label="(225, 12)", style=solid]; -"1252 index_8" -> "1253 view_45" [label="(4096, 12)", style=solid]; -"1253 view_45" -> "1254 permute_37" [label="(64, 64, 12)", style=solid]; -"1254 permute_37" -> "1255 contiguous_14" [label="(12, 64, 64)", style=solid]; -"1255 contiguous_14" -> "1256 unsqueeze_24" [label="(12, 64, 64)", style=solid]; -"1256 unsqueeze_24" -> "1257 sigmoid_8" [label="(1, 12, 64, 64)", style=solid]; -"1257 sigmoid_8" -> "1258 mul_16" [label="(1, 12, 64, 64)", style=solid]; -"1258 mul_16" -> "1294 add_28" [label="(1, 12, 64, 64)", style=solid]; -"1259 pad_10" -> "1260 view_46" [label="(1, 16, 16, 384)", style=solid]; -"1260 view_46" -> "1261 permute_38" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"1261 permute_38" -> "1262 reshape_36" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"1262 reshape_36" -> "1264 reshape_36_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"1263 clone_8" -> "1271 linear_52" [label="(1152,)", style=solid]; -"1264 reshape_36_0_0_nncf_smooth_quant_0" -> "1265 quantize_per_tensor_default_51" [label="(4, 64, 384)", style=solid]; -"1265 quantize_per_tensor_default_51" -> "1266 dequantize_per_tensor_default_51" [label="(4, 64, 384)", style=solid]; -"1266 dequantize_per_tensor_default_51" -> "1271 linear_52" [label="(4, 64, 384)", style=solid]; -"1267 linear_52_scale_0" -> "1270 dequantize_per_channel_default_53" [label="(1152,)", style=solid]; -"1268 linear_52_zero_point_0" -> "1270 dequantize_per_channel_default_53" [label="(1152,)", style=solid]; -"1269 _frozen_param53" -> "1270 dequantize_per_channel_default_53" [label="(1152, 384)", style=solid]; -"1270 dequantize_per_channel_default_53" -> "1271 linear_52" [label="(1152, 384)", style=solid]; -"1271 linear_52" -> "1272 reshape_37" [label="(4, 64, 1152)", style=solid]; -"1272 reshape_37" -> "1273 permute_39" [label="(4, 64, 3, 12, 32)", style=solid]; -"1273 permute_39" -> "1274 select_24" [label="(3, 4, 12, 64, 32)", style=solid]; -"1273 permute_39" -> "1275 select_25" [label="(3, 4, 12, 64, 32)", style=solid]; -"1273 permute_39" -> "1276 select_26" [label="(3, 4, 12, 64, 32)", style=solid]; -"1274 select_24" -> "1277 linalg_vector_norm_16" [label="(4, 12, 64, 32)", style=solid]; -"1274 select_24" -> "1279 expand_as_16" [label="(4, 12, 64, 32)", style=solid]; -"1274 select_24" -> "1280 div_16" [label="(4, 12, 64, 32)", style=solid]; -"1275 select_25" -> "1283 linalg_vector_norm_17" [label="(4, 12, 64, 32)", style=solid]; -"1275 select_25" -> "1285 expand_as_17" [label="(4, 12, 64, 32)", style=solid]; -"1275 select_25" -> "1286 div_17" [label="(4, 12, 64, 32)", style=solid]; -"1276 select_26" -> "1297 matmul_17" [label="(4, 12, 64, 32)", style=solid]; -"1277 linalg_vector_norm_16" -> "1278 clamp_min_16" [label="(4, 12, 64, 1)", style=solid]; -"1278 clamp_min_16" -> "1279 expand_as_16" [label="(4, 12, 64, 1)", style=solid]; -"1279 expand_as_16" -> "1280 div_16" [label="(4, 12, 64, 32)", style=solid]; -"1280 div_16" -> "1281 quantize_per_tensor_default_52" [label="(4, 12, 64, 32)", style=solid]; -"1281 quantize_per_tensor_default_52" -> "1282 dequantize_per_tensor_default_52" [label="(4, 12, 64, 32)", style=solid]; -"1282 dequantize_per_tensor_default_52" -> "1290 matmul_16" [label="(4, 12, 64, 32)", style=solid]; -"1283 linalg_vector_norm_17" -> "1284 clamp_min_17" [label="(4, 12, 64, 1)", style=solid]; -"1284 clamp_min_17" -> "1285 expand_as_17" [label="(4, 12, 64, 1)", style=solid]; -"1285 expand_as_17" -> "1286 div_17" [label="(4, 12, 64, 32)", style=solid]; -"1286 div_17" -> "1287 quantize_per_tensor_default_53" [label="(4, 12, 64, 32)", style=solid]; -"1287 quantize_per_tensor_default_53" -> "1288 dequantize_per_tensor_default_53" [label="(4, 12, 64, 32)", style=solid]; -"1288 dequantize_per_tensor_default_53" -> "1289 transpose_16" [label="(4, 12, 64, 32)", style=solid]; -"1289 transpose_16" -> "1290 matmul_16" [label="(4, 12, 32, 64)", style=solid]; -"1290 matmul_16" -> "1293 mul_17" [label="(4, 12, 64, 64)", style=solid]; -"1291 clamp_8" -> "1292 exp_8" [label="(12, 1, 1)", style=solid]; -"1292 exp_8" -> "1293 mul_17" [label="(12, 1, 1)", style=solid]; -"1293 mul_17" -> "1294 add_28" [label="(4, 12, 64, 64)", style=solid]; -"1294 add_28" -> "1295 softmax_8" [label="(4, 12, 64, 64)", style=solid]; -"1295 softmax_8" -> "1296 dropout_32" [label="(4, 12, 64, 64)", style=solid]; -"1296 dropout_32" -> "1297 matmul_17" [label="(4, 12, 64, 64)", style=solid]; -"1297 matmul_17" -> "1298 transpose_17" [label="(4, 12, 64, 32)", style=solid]; -"1298 transpose_17" -> "1299 reshape_38" [label="(4, 64, 12, 32)", style=solid]; -"1299 reshape_38" -> "1300 reshape_38_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"1300 reshape_38_0_0_nncf_smooth_quant_0" -> "1301 quantize_per_tensor_default_54" [label="(4, 64, 384)", style=solid]; -"1301 quantize_per_tensor_default_54" -> "1302 dequantize_per_tensor_default_54" [label="(4, 64, 384)", style=solid]; -"1302 dequantize_per_tensor_default_54" -> "1307 linear_53" [label="(4, 64, 384)", style=solid]; -"1303 linear_53_scale_0" -> "1306 dequantize_per_channel_default_54" [label="(384,)", style=solid]; -"1304 linear_53_zero_point_0" -> "1306 dequantize_per_channel_default_54" [label="(384,)", style=solid]; -"1305 _frozen_param54" -> "1306 dequantize_per_channel_default_54" [label="(384, 384)", style=solid]; -"1306 dequantize_per_channel_default_54" -> "1307 linear_53" [label="(384, 384)", style=solid]; -"1307 linear_53" -> "1308 dropout_33" [label="(4, 64, 384)", style=solid]; -"1308 dropout_33" -> "1309 view_47" [label="(4, 64, 384)", style=solid]; -"1309 view_47" -> "1310 permute_40" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"1310 permute_40" -> "1311 reshape_39" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"1311 reshape_39" -> "1312 slice_134" [label="(1, 16, 16, 384)", style=solid]; -"1312 slice_134" -> "1313 slice_135" [label="(1, 16, 16, 384)", style=solid]; -"1313 slice_135" -> "1314 slice_136" [label="(1, 14, 16, 384)", style=solid]; -"1314 slice_136" -> "1315 slice_137" [label="(1, 14, 14, 384)", style=solid]; -"1315 slice_137" -> "1316 contiguous_15" [label="(1, 14, 14, 384)", style=solid]; -"1316 contiguous_15" -> "1317 layer_norm_19" [label="(1, 14, 14, 384)", style=solid]; -"1317 layer_norm_19" -> "1318 add_29" [label="(1, 14, 14, 384)", style=solid]; -"1318 add_29" -> "1319 add_29_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; -"1318 add_29" -> "1339 add_30" [label="(1, 14, 14, 384)", style=solid]; -"1319 add_29_0_0_nncf_smooth_quant_0" -> "1320 quantize_per_tensor_default_55" [label="(1, 14, 14, 384)", style=solid]; -"1320 quantize_per_tensor_default_55" -> "1321 dequantize_per_tensor_default_55" [label="(1, 14, 14, 384)", style=solid]; -"1321 dequantize_per_tensor_default_55" -> "1326 linear_54" [label="(1, 14, 14, 384)", style=solid]; -"1322 linear_54_scale_0" -> "1325 dequantize_per_channel_default_55" [label="(1536,)", style=solid]; -"1323 linear_54_zero_point_0" -> "1325 dequantize_per_channel_default_55" [label="(1536,)", style=solid]; -"1324 _frozen_param55" -> "1325 dequantize_per_channel_default_55" [label="(1536, 384)", style=solid]; -"1325 dequantize_per_channel_default_55" -> "1326 linear_54" [label="(1536, 384)", style=solid]; -"1326 linear_54" -> "1327 gelu_8" [label="(1, 14, 14, 1536)", style=solid]; -"1327 gelu_8" -> "1328 dropout_34" [label="(1, 14, 14, 1536)", style=solid]; -"1328 dropout_34" -> "1329 dropout_34_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; -"1329 dropout_34_0_0_nncf_smooth_quant_0" -> "1330 quantize_per_tensor_default_56" [label="(1, 14, 14, 1536)", style=solid]; -"1330 quantize_per_tensor_default_56" -> "1331 dequantize_per_tensor_default_56" [label="(1, 14, 14, 1536)", style=solid]; -"1331 dequantize_per_tensor_default_56" -> "1336 linear_55" [label="(1, 14, 14, 1536)", style=solid]; -"1332 linear_55_scale_0" -> "1335 dequantize_per_channel_default_56" [label="(384,)", style=solid]; -"1333 linear_55_zero_point_0" -> "1335 dequantize_per_channel_default_56" [label="(384,)", style=solid]; -"1334 _frozen_param56" -> "1335 dequantize_per_channel_default_56" [label="(384, 1536)", style=solid]; -"1335 dequantize_per_channel_default_56" -> "1336 linear_55" [label="(384, 1536)", style=solid]; -"1336 linear_55" -> "1337 dropout_35" [label="(1, 14, 14, 384)", style=solid]; -"1337 dropout_35" -> "1338 layer_norm_20" [label="(1, 14, 14, 384)", style=solid]; -"1338 layer_norm_20" -> "1339 add_30" [label="(1, 14, 14, 384)", style=solid]; -"1339 add_30" -> "1361 pad_11" [label="(1, 14, 14, 384)", style=solid]; -"1339 add_30" -> "1438 add_33" [label="(1, 14, 14, 384)", style=solid]; -"1340 features_5_5_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "1345 linear_56" [label="(1, 15, 15, 2)", style=solid]; -"1341 linear_56_scale_0" -> "1344 dequantize_per_channel_default_57" [label="(512,)", style=solid]; -"1342 linear_56_zero_point_0" -> "1344 dequantize_per_channel_default_57" [label="(512,)", style=solid]; -"1343 _frozen_param57" -> "1344 dequantize_per_channel_default_57" [label="(512, 2)", style=solid]; -"1344 dequantize_per_channel_default_57" -> "1345 linear_56" [label="(512, 2)", style=solid]; -"1345 linear_56" -> "1346 relu__9" [label="(1, 15, 15, 512)", style=solid]; -"1346 relu__9" -> "1347 relu__9_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; -"1347 relu__9_0_0_nncf_smooth_quant_0" -> "1352 linear_57" [label="(1, 15, 15, 512)", style=solid]; -"1348 linear_57_scale_0" -> "1351 dequantize_per_channel_default_58" [label="(12,)", style=solid]; -"1349 linear_57_zero_point_0" -> "1351 dequantize_per_channel_default_58" [label="(12,)", style=solid]; -"1350 _frozen_param58" -> "1351 dequantize_per_channel_default_58" [label="(12, 512)", style=solid]; -"1351 dequantize_per_channel_default_58" -> "1352 linear_57" [label="(12, 512)", style=solid]; -"1352 linear_57" -> "1353 view_48" [label="(1, 15, 15, 12)", style=solid]; -"1353 view_48" -> "1354 index_9" [label="(225, 12)", style=solid]; -"1354 index_9" -> "1355 view_49" [label="(4096, 12)", style=solid]; -"1355 view_49" -> "1356 permute_41" [label="(64, 64, 12)", style=solid]; -"1356 permute_41" -> "1357 contiguous_16" [label="(12, 64, 64)", style=solid]; -"1357 contiguous_16" -> "1358 unsqueeze_25" [label="(12, 64, 64)", style=solid]; -"1358 unsqueeze_25" -> "1359 sigmoid_9" [label="(1, 12, 64, 64)", style=solid]; -"1359 sigmoid_9" -> "1360 mul_18" [label="(1, 12, 64, 64)", style=solid]; -"1360 mul_18" -> "1397 add_31" [label="(1, 12, 64, 64)", style=solid]; -"1361 pad_11" -> "1362 roll_8" [label="(1, 16, 16, 384)", style=solid]; -"1362 roll_8" -> "1363 view_50" [label="(1, 16, 16, 384)", style=solid]; -"1363 view_50" -> "1364 permute_42" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"1364 permute_42" -> "1365 reshape_40" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"1365 reshape_40" -> "1367 reshape_40_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"1365 reshape_40" -> "1398 new_zeros_4" [label="(4, 64, 384)", style=solid]; -"1366 clone_9" -> "1374 linear_58" [label="(1152,)", style=solid]; -"1367 reshape_40_0_0_nncf_smooth_quant_0" -> "1368 quantize_per_tensor_default_57" [label="(4, 64, 384)", style=solid]; -"1368 quantize_per_tensor_default_57" -> "1369 dequantize_per_tensor_default_57" [label="(4, 64, 384)", style=solid]; -"1369 dequantize_per_tensor_default_57" -> "1374 linear_58" [label="(4, 64, 384)", style=solid]; -"1370 linear_58_scale_0" -> "1373 dequantize_per_channel_default_59" [label="(1152,)", style=solid]; -"1371 linear_58_zero_point_0" -> "1373 dequantize_per_channel_default_59" [label="(1152,)", style=solid]; -"1372 _frozen_param59" -> "1373 dequantize_per_channel_default_59" [label="(1152, 384)", style=solid]; -"1373 dequantize_per_channel_default_59" -> "1374 linear_58" [label="(1152, 384)", style=solid]; -"1374 linear_58" -> "1375 reshape_41" [label="(4, 64, 1152)", style=solid]; -"1375 reshape_41" -> "1376 permute_43" [label="(4, 64, 3, 12, 32)", style=solid]; -"1376 permute_43" -> "1377 select_27" [label="(3, 4, 12, 64, 32)", style=solid]; -"1376 permute_43" -> "1378 select_28" [label="(3, 4, 12, 64, 32)", style=solid]; -"1376 permute_43" -> "1379 select_29" [label="(3, 4, 12, 64, 32)", style=solid]; -"1377 select_27" -> "1380 linalg_vector_norm_18" [label="(4, 12, 64, 32)", style=solid]; -"1377 select_27" -> "1382 expand_as_18" [label="(4, 12, 64, 32)", style=solid]; -"1377 select_27" -> "1383 div_18" [label="(4, 12, 64, 32)", style=solid]; -"1378 select_28" -> "1386 linalg_vector_norm_19" [label="(4, 12, 64, 32)", style=solid]; -"1378 select_28" -> "1388 expand_as_19" [label="(4, 12, 64, 32)", style=solid]; -"1378 select_28" -> "1389 div_19" [label="(4, 12, 64, 32)", style=solid]; -"1379 select_29" -> "1416 matmul_19" [label="(4, 12, 64, 32)", style=solid]; -"1380 linalg_vector_norm_18" -> "1381 clamp_min_18" [label="(4, 12, 64, 1)", style=solid]; -"1381 clamp_min_18" -> "1382 expand_as_18" [label="(4, 12, 64, 1)", style=solid]; -"1382 expand_as_18" -> "1383 div_18" [label="(4, 12, 64, 32)", style=solid]; -"1383 div_18" -> "1384 quantize_per_tensor_default_58" [label="(4, 12, 64, 32)", style=solid]; -"1384 quantize_per_tensor_default_58" -> "1385 dequantize_per_tensor_default_58" [label="(4, 12, 64, 32)", style=solid]; -"1385 dequantize_per_tensor_default_58" -> "1393 matmul_18" [label="(4, 12, 64, 32)", style=solid]; -"1386 linalg_vector_norm_19" -> "1387 clamp_min_19" [label="(4, 12, 64, 1)", style=solid]; -"1387 clamp_min_19" -> "1388 expand_as_19" [label="(4, 12, 64, 1)", style=solid]; -"1388 expand_as_19" -> "1389 div_19" [label="(4, 12, 64, 32)", style=solid]; -"1389 div_19" -> "1390 quantize_per_tensor_default_59" [label="(4, 12, 64, 32)", style=solid]; -"1390 quantize_per_tensor_default_59" -> "1391 dequantize_per_tensor_default_59" [label="(4, 12, 64, 32)", style=solid]; -"1391 dequantize_per_tensor_default_59" -> "1392 transpose_18" [label="(4, 12, 64, 32)", style=solid]; -"1392 transpose_18" -> "1393 matmul_18" [label="(4, 12, 32, 64)", style=solid]; -"1393 matmul_18" -> "1396 mul_19" [label="(4, 12, 64, 64)", style=solid]; -"1394 clamp_9" -> "1395 exp_9" [label="(12, 1, 1)", style=solid]; -"1395 exp_9" -> "1396 mul_19" [label="(12, 1, 1)", style=solid]; -"1396 mul_19" -> "1397 add_31" [label="(4, 12, 64, 64)", style=solid]; -"1397 add_31" -> "1409 view_52" [label="(4, 12, 64, 64)", style=solid]; -"1398 new_zeros_4" -> "1399 view_51" [label="(16, 16)", style=solid]; -"1399 view_51" -> "1400 permute_44" [label="(2, 8, 2, 8)", style=solid]; -"1400 permute_44" -> "1401 reshape_42" [label="(2, 2, 8, 8)", style=solid]; -"1401 reshape_42" -> "1402 unsqueeze_26" [label="(4, 64)", style=solid]; -"1401 reshape_42" -> "1403 unsqueeze_27" [label="(4, 64)", style=solid]; -"1402 unsqueeze_26" -> "1404 sub_4" [label="(4, 1, 64)", style=solid]; -"1403 unsqueeze_27" -> "1404 sub_4" [label="(4, 64, 1)", style=solid]; -"1404 sub_4" -> "1405 ne_4" [label="(4, 64, 64)", style=solid]; -"1404 sub_4" -> "1406 masked_fill_8" [label="(4, 64, 64)", style=solid]; -"1404 sub_4" -> "1407 eq_4" [label="(4, 64, 64)", style=solid]; -"1405 ne_4" -> "1406 masked_fill_8" [label="(4, 64, 64)", style=solid]; -"1406 masked_fill_8" -> "1408 masked_fill_9" [label="(4, 64, 64)", style=solid]; -"1407 eq_4" -> "1408 masked_fill_9" [label="(4, 64, 64)", style=solid]; -"1408 masked_fill_9" -> "1410 unsqueeze_28" [label="(4, 64, 64)", style=solid]; -"1409 view_52" -> "1412 add_32" [label="(1, 4, 12, 64, 64)", style=solid]; -"1410 unsqueeze_28" -> "1411 unsqueeze_29" [label="(4, 1, 64, 64)", style=solid]; -"1411 unsqueeze_29" -> "1412 add_32" [label="(1, 4, 1, 64, 64)", style=solid]; -"1412 add_32" -> "1413 view_53" [label="(1, 4, 12, 64, 64)", style=solid]; -"1413 view_53" -> "1414 softmax_9" [label="(4, 12, 64, 64)", style=solid]; -"1414 softmax_9" -> "1415 dropout_36" [label="(4, 12, 64, 64)", style=solid]; -"1415 dropout_36" -> "1416 matmul_19" [label="(4, 12, 64, 64)", style=solid]; -"1416 matmul_19" -> "1417 transpose_19" [label="(4, 12, 64, 32)", style=solid]; -"1417 transpose_19" -> "1418 reshape_43" [label="(4, 64, 12, 32)", style=solid]; -"1418 reshape_43" -> "1419 reshape_43_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"1419 reshape_43_0_0_nncf_smooth_quant_0" -> "1420 quantize_per_tensor_default_60" [label="(4, 64, 384)", style=solid]; -"1420 quantize_per_tensor_default_60" -> "1421 dequantize_per_tensor_default_60" [label="(4, 64, 384)", style=solid]; -"1421 dequantize_per_tensor_default_60" -> "1426 linear_59" [label="(4, 64, 384)", style=solid]; -"1422 linear_59_scale_0" -> "1425 dequantize_per_channel_default_60" [label="(384,)", style=solid]; -"1423 linear_59_zero_point_0" -> "1425 dequantize_per_channel_default_60" [label="(384,)", style=solid]; -"1424 _frozen_param60" -> "1425 dequantize_per_channel_default_60" [label="(384, 384)", style=solid]; -"1425 dequantize_per_channel_default_60" -> "1426 linear_59" [label="(384, 384)", style=solid]; -"1426 linear_59" -> "1427 dropout_37" [label="(4, 64, 384)", style=solid]; -"1427 dropout_37" -> "1428 view_54" [label="(4, 64, 384)", style=solid]; -"1428 view_54" -> "1429 permute_45" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"1429 permute_45" -> "1430 reshape_44" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"1430 reshape_44" -> "1431 roll_9" [label="(1, 16, 16, 384)", style=solid]; -"1431 roll_9" -> "1432 slice_157" [label="(1, 16, 16, 384)", style=solid]; -"1432 slice_157" -> "1433 slice_158" [label="(1, 16, 16, 384)", style=solid]; -"1433 slice_158" -> "1434 slice_159" [label="(1, 14, 16, 384)", style=solid]; -"1434 slice_159" -> "1435 slice_160" [label="(1, 14, 14, 384)", style=solid]; -"1435 slice_160" -> "1436 contiguous_17" [label="(1, 14, 14, 384)", style=solid]; -"1436 contiguous_17" -> "1437 layer_norm_21" [label="(1, 14, 14, 384)", style=solid]; -"1437 layer_norm_21" -> "1438 add_33" [label="(1, 14, 14, 384)", style=solid]; -"1438 add_33" -> "1439 add_33_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; -"1438 add_33" -> "1459 add_34" [label="(1, 14, 14, 384)", style=solid]; -"1439 add_33_0_0_nncf_smooth_quant_0" -> "1440 quantize_per_tensor_default_61" [label="(1, 14, 14, 384)", style=solid]; -"1440 quantize_per_tensor_default_61" -> "1441 dequantize_per_tensor_default_61" [label="(1, 14, 14, 384)", style=solid]; -"1441 dequantize_per_tensor_default_61" -> "1446 linear_60" [label="(1, 14, 14, 384)", style=solid]; -"1442 linear_60_scale_0" -> "1445 dequantize_per_channel_default_61" [label="(1536,)", style=solid]; -"1443 linear_60_zero_point_0" -> "1445 dequantize_per_channel_default_61" [label="(1536,)", style=solid]; -"1444 _frozen_param61" -> "1445 dequantize_per_channel_default_61" [label="(1536, 384)", style=solid]; -"1445 dequantize_per_channel_default_61" -> "1446 linear_60" [label="(1536, 384)", style=solid]; -"1446 linear_60" -> "1447 gelu_9" [label="(1, 14, 14, 1536)", style=solid]; -"1447 gelu_9" -> "1448 dropout_38" [label="(1, 14, 14, 1536)", style=solid]; -"1448 dropout_38" -> "1449 dropout_38_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; -"1449 dropout_38_0_0_nncf_smooth_quant_0" -> "1450 quantize_per_tensor_default_62" [label="(1, 14, 14, 1536)", style=solid]; -"1450 quantize_per_tensor_default_62" -> "1451 dequantize_per_tensor_default_62" [label="(1, 14, 14, 1536)", style=solid]; -"1451 dequantize_per_tensor_default_62" -> "1456 linear_61" [label="(1, 14, 14, 1536)", style=solid]; -"1452 linear_61_scale_0" -> "1455 dequantize_per_channel_default_62" [label="(384,)", style=solid]; -"1453 linear_61_zero_point_0" -> "1455 dequantize_per_channel_default_62" [label="(384,)", style=solid]; -"1454 _frozen_param62" -> "1455 dequantize_per_channel_default_62" [label="(384, 1536)", style=solid]; -"1455 dequantize_per_channel_default_62" -> "1456 linear_61" [label="(384, 1536)", style=solid]; -"1456 linear_61" -> "1457 dropout_39" [label="(1, 14, 14, 384)", style=solid]; -"1457 dropout_39" -> "1458 layer_norm_22" [label="(1, 14, 14, 384)", style=solid]; -"1458 layer_norm_22" -> "1459 add_34" [label="(1, 14, 14, 384)", style=solid]; -"1459 add_34" -> "1481 pad_12" [label="(1, 14, 14, 384)", style=solid]; -"1459 add_34" -> "1540 add_36" [label="(1, 14, 14, 384)", style=solid]; -"1460 features_5_6_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "1465 linear_62" [label="(1, 15, 15, 2)", style=solid]; -"1461 linear_62_scale_0" -> "1464 dequantize_per_channel_default_63" [label="(512,)", style=solid]; -"1462 linear_62_zero_point_0" -> "1464 dequantize_per_channel_default_63" [label="(512,)", style=solid]; -"1463 _frozen_param63" -> "1464 dequantize_per_channel_default_63" [label="(512, 2)", style=solid]; -"1464 dequantize_per_channel_default_63" -> "1465 linear_62" [label="(512, 2)", style=solid]; -"1465 linear_62" -> "1466 relu__10" [label="(1, 15, 15, 512)", style=solid]; -"1466 relu__10" -> "1467 relu__10_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; -"1467 relu__10_0_0_nncf_smooth_quant_0" -> "1472 linear_63" [label="(1, 15, 15, 512)", style=solid]; -"1468 linear_63_scale_0" -> "1471 dequantize_per_channel_default_64" [label="(12,)", style=solid]; -"1469 linear_63_zero_point_0" -> "1471 dequantize_per_channel_default_64" [label="(12,)", style=solid]; -"1470 _frozen_param64" -> "1471 dequantize_per_channel_default_64" [label="(12, 512)", style=solid]; -"1471 dequantize_per_channel_default_64" -> "1472 linear_63" [label="(12, 512)", style=solid]; -"1472 linear_63" -> "1473 view_55" [label="(1, 15, 15, 12)", style=solid]; -"1473 view_55" -> "1474 index_10" [label="(225, 12)", style=solid]; -"1474 index_10" -> "1475 view_56" [label="(4096, 12)", style=solid]; -"1475 view_56" -> "1476 permute_46" [label="(64, 64, 12)", style=solid]; -"1476 permute_46" -> "1477 contiguous_18" [label="(12, 64, 64)", style=solid]; -"1477 contiguous_18" -> "1478 unsqueeze_30" [label="(12, 64, 64)", style=solid]; -"1478 unsqueeze_30" -> "1479 sigmoid_10" [label="(1, 12, 64, 64)", style=solid]; -"1479 sigmoid_10" -> "1480 mul_20" [label="(1, 12, 64, 64)", style=solid]; -"1480 mul_20" -> "1516 add_35" [label="(1, 12, 64, 64)", style=solid]; -"1481 pad_12" -> "1482 view_57" [label="(1, 16, 16, 384)", style=solid]; -"1482 view_57" -> "1483 permute_47" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"1483 permute_47" -> "1484 reshape_45" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"1484 reshape_45" -> "1486 reshape_45_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"1485 clone_10" -> "1493 linear_64" [label="(1152,)", style=solid]; -"1486 reshape_45_0_0_nncf_smooth_quant_0" -> "1487 quantize_per_tensor_default_63" [label="(4, 64, 384)", style=solid]; -"1487 quantize_per_tensor_default_63" -> "1488 dequantize_per_tensor_default_63" [label="(4, 64, 384)", style=solid]; -"1488 dequantize_per_tensor_default_63" -> "1493 linear_64" [label="(4, 64, 384)", style=solid]; -"1489 linear_64_scale_0" -> "1492 dequantize_per_channel_default_65" [label="(1152,)", style=solid]; -"1490 linear_64_zero_point_0" -> "1492 dequantize_per_channel_default_65" [label="(1152,)", style=solid]; -"1491 _frozen_param65" -> "1492 dequantize_per_channel_default_65" [label="(1152, 384)", style=solid]; -"1492 dequantize_per_channel_default_65" -> "1493 linear_64" [label="(1152, 384)", style=solid]; -"1493 linear_64" -> "1494 reshape_46" [label="(4, 64, 1152)", style=solid]; -"1494 reshape_46" -> "1495 permute_48" [label="(4, 64, 3, 12, 32)", style=solid]; -"1495 permute_48" -> "1496 select_30" [label="(3, 4, 12, 64, 32)", style=solid]; -"1495 permute_48" -> "1497 select_31" [label="(3, 4, 12, 64, 32)", style=solid]; -"1495 permute_48" -> "1498 select_32" [label="(3, 4, 12, 64, 32)", style=solid]; -"1496 select_30" -> "1499 linalg_vector_norm_20" [label="(4, 12, 64, 32)", style=solid]; -"1496 select_30" -> "1501 expand_as_20" [label="(4, 12, 64, 32)", style=solid]; -"1496 select_30" -> "1502 div_20" [label="(4, 12, 64, 32)", style=solid]; -"1497 select_31" -> "1505 linalg_vector_norm_21" [label="(4, 12, 64, 32)", style=solid]; -"1497 select_31" -> "1507 expand_as_21" [label="(4, 12, 64, 32)", style=solid]; -"1497 select_31" -> "1508 div_21" [label="(4, 12, 64, 32)", style=solid]; -"1498 select_32" -> "1519 matmul_21" [label="(4, 12, 64, 32)", style=solid]; -"1499 linalg_vector_norm_20" -> "1500 clamp_min_20" [label="(4, 12, 64, 1)", style=solid]; -"1500 clamp_min_20" -> "1501 expand_as_20" [label="(4, 12, 64, 1)", style=solid]; -"1501 expand_as_20" -> "1502 div_20" [label="(4, 12, 64, 32)", style=solid]; -"1502 div_20" -> "1503 quantize_per_tensor_default_64" [label="(4, 12, 64, 32)", style=solid]; -"1503 quantize_per_tensor_default_64" -> "1504 dequantize_per_tensor_default_64" [label="(4, 12, 64, 32)", style=solid]; -"1504 dequantize_per_tensor_default_64" -> "1512 matmul_20" [label="(4, 12, 64, 32)", style=solid]; -"1505 linalg_vector_norm_21" -> "1506 clamp_min_21" [label="(4, 12, 64, 1)", style=solid]; -"1506 clamp_min_21" -> "1507 expand_as_21" [label="(4, 12, 64, 1)", style=solid]; -"1507 expand_as_21" -> "1508 div_21" [label="(4, 12, 64, 32)", style=solid]; -"1508 div_21" -> "1509 quantize_per_tensor_default_65" [label="(4, 12, 64, 32)", style=solid]; -"1509 quantize_per_tensor_default_65" -> "1510 dequantize_per_tensor_default_65" [label="(4, 12, 64, 32)", style=solid]; -"1510 dequantize_per_tensor_default_65" -> "1511 transpose_20" [label="(4, 12, 64, 32)", style=solid]; -"1511 transpose_20" -> "1512 matmul_20" [label="(4, 12, 32, 64)", style=solid]; -"1512 matmul_20" -> "1515 mul_21" [label="(4, 12, 64, 64)", style=solid]; -"1513 clamp_10" -> "1514 exp_10" [label="(12, 1, 1)", style=solid]; -"1514 exp_10" -> "1515 mul_21" [label="(12, 1, 1)", style=solid]; -"1515 mul_21" -> "1516 add_35" [label="(4, 12, 64, 64)", style=solid]; -"1516 add_35" -> "1517 softmax_10" [label="(4, 12, 64, 64)", style=solid]; -"1517 softmax_10" -> "1518 dropout_40" [label="(4, 12, 64, 64)", style=solid]; -"1518 dropout_40" -> "1519 matmul_21" [label="(4, 12, 64, 64)", style=solid]; -"1519 matmul_21" -> "1520 transpose_21" [label="(4, 12, 64, 32)", style=solid]; -"1520 transpose_21" -> "1521 reshape_47" [label="(4, 64, 12, 32)", style=solid]; -"1521 reshape_47" -> "1522 reshape_47_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"1522 reshape_47_0_0_nncf_smooth_quant_0" -> "1523 quantize_per_tensor_default_66" [label="(4, 64, 384)", style=solid]; -"1523 quantize_per_tensor_default_66" -> "1524 dequantize_per_tensor_default_66" [label="(4, 64, 384)", style=solid]; -"1524 dequantize_per_tensor_default_66" -> "1529 linear_65" [label="(4, 64, 384)", style=solid]; -"1525 linear_65_scale_0" -> "1528 dequantize_per_channel_default_66" [label="(384,)", style=solid]; -"1526 linear_65_zero_point_0" -> "1528 dequantize_per_channel_default_66" [label="(384,)", style=solid]; -"1527 _frozen_param66" -> "1528 dequantize_per_channel_default_66" [label="(384, 384)", style=solid]; -"1528 dequantize_per_channel_default_66" -> "1529 linear_65" [label="(384, 384)", style=solid]; -"1529 linear_65" -> "1530 dropout_41" [label="(4, 64, 384)", style=solid]; -"1530 dropout_41" -> "1531 view_58" [label="(4, 64, 384)", style=solid]; -"1531 view_58" -> "1532 permute_49" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"1532 permute_49" -> "1533 reshape_48" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"1533 reshape_48" -> "1534 slice_162" [label="(1, 16, 16, 384)", style=solid]; -"1534 slice_162" -> "1535 slice_163" [label="(1, 16, 16, 384)", style=solid]; -"1535 slice_163" -> "1536 slice_164" [label="(1, 14, 16, 384)", style=solid]; -"1536 slice_164" -> "1537 slice_165" [label="(1, 14, 14, 384)", style=solid]; -"1537 slice_165" -> "1538 contiguous_19" [label="(1, 14, 14, 384)", style=solid]; -"1538 contiguous_19" -> "1539 layer_norm_23" [label="(1, 14, 14, 384)", style=solid]; -"1539 layer_norm_23" -> "1540 add_36" [label="(1, 14, 14, 384)", style=solid]; -"1540 add_36" -> "1541 add_36_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; -"1540 add_36" -> "1561 add_37" [label="(1, 14, 14, 384)", style=solid]; -"1541 add_36_0_0_nncf_smooth_quant_0" -> "1542 quantize_per_tensor_default_67" [label="(1, 14, 14, 384)", style=solid]; -"1542 quantize_per_tensor_default_67" -> "1543 dequantize_per_tensor_default_67" [label="(1, 14, 14, 384)", style=solid]; -"1543 dequantize_per_tensor_default_67" -> "1548 linear_66" [label="(1, 14, 14, 384)", style=solid]; -"1544 linear_66_scale_0" -> "1547 dequantize_per_channel_default_67" [label="(1536,)", style=solid]; -"1545 linear_66_zero_point_0" -> "1547 dequantize_per_channel_default_67" [label="(1536,)", style=solid]; -"1546 _frozen_param67" -> "1547 dequantize_per_channel_default_67" [label="(1536, 384)", style=solid]; -"1547 dequantize_per_channel_default_67" -> "1548 linear_66" [label="(1536, 384)", style=solid]; -"1548 linear_66" -> "1549 gelu_10" [label="(1, 14, 14, 1536)", style=solid]; -"1549 gelu_10" -> "1550 dropout_42" [label="(1, 14, 14, 1536)", style=solid]; -"1550 dropout_42" -> "1551 dropout_42_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; -"1551 dropout_42_0_0_nncf_smooth_quant_0" -> "1552 quantize_per_tensor_default_68" [label="(1, 14, 14, 1536)", style=solid]; -"1552 quantize_per_tensor_default_68" -> "1553 dequantize_per_tensor_default_68" [label="(1, 14, 14, 1536)", style=solid]; -"1553 dequantize_per_tensor_default_68" -> "1558 linear_67" [label="(1, 14, 14, 1536)", style=solid]; -"1554 linear_67_scale_0" -> "1557 dequantize_per_channel_default_68" [label="(384,)", style=solid]; -"1555 linear_67_zero_point_0" -> "1557 dequantize_per_channel_default_68" [label="(384,)", style=solid]; -"1556 _frozen_param68" -> "1557 dequantize_per_channel_default_68" [label="(384, 1536)", style=solid]; -"1557 dequantize_per_channel_default_68" -> "1558 linear_67" [label="(384, 1536)", style=solid]; -"1558 linear_67" -> "1559 dropout_43" [label="(1, 14, 14, 384)", style=solid]; -"1559 dropout_43" -> "1560 layer_norm_24" [label="(1, 14, 14, 384)", style=solid]; -"1560 layer_norm_24" -> "1561 add_37" [label="(1, 14, 14, 384)", style=solid]; -"1561 add_37" -> "1583 pad_13" [label="(1, 14, 14, 384)", style=solid]; -"1561 add_37" -> "1660 add_40" [label="(1, 14, 14, 384)", style=solid]; -"1562 features_5_7_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "1567 linear_68" [label="(1, 15, 15, 2)", style=solid]; -"1563 linear_68_scale_0" -> "1566 dequantize_per_channel_default_69" [label="(512,)", style=solid]; -"1564 linear_68_zero_point_0" -> "1566 dequantize_per_channel_default_69" [label="(512,)", style=solid]; -"1565 _frozen_param69" -> "1566 dequantize_per_channel_default_69" [label="(512, 2)", style=solid]; -"1566 dequantize_per_channel_default_69" -> "1567 linear_68" [label="(512, 2)", style=solid]; -"1567 linear_68" -> "1568 relu__11" [label="(1, 15, 15, 512)", style=solid]; -"1568 relu__11" -> "1569 relu__11_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; -"1569 relu__11_0_0_nncf_smooth_quant_0" -> "1574 linear_69" [label="(1, 15, 15, 512)", style=solid]; -"1570 linear_69_scale_0" -> "1573 dequantize_per_channel_default_70" [label="(12,)", style=solid]; -"1571 linear_69_zero_point_0" -> "1573 dequantize_per_channel_default_70" [label="(12,)", style=solid]; -"1572 _frozen_param70" -> "1573 dequantize_per_channel_default_70" [label="(12, 512)", style=solid]; -"1573 dequantize_per_channel_default_70" -> "1574 linear_69" [label="(12, 512)", style=solid]; -"1574 linear_69" -> "1575 view_59" [label="(1, 15, 15, 12)", style=solid]; -"1575 view_59" -> "1576 index_11" [label="(225, 12)", style=solid]; -"1576 index_11" -> "1577 view_60" [label="(4096, 12)", style=solid]; -"1577 view_60" -> "1578 permute_50" [label="(64, 64, 12)", style=solid]; -"1578 permute_50" -> "1579 contiguous_20" [label="(12, 64, 64)", style=solid]; -"1579 contiguous_20" -> "1580 unsqueeze_31" [label="(12, 64, 64)", style=solid]; -"1580 unsqueeze_31" -> "1581 sigmoid_11" [label="(1, 12, 64, 64)", style=solid]; -"1581 sigmoid_11" -> "1582 mul_22" [label="(1, 12, 64, 64)", style=solid]; -"1582 mul_22" -> "1619 add_38" [label="(1, 12, 64, 64)", style=solid]; -"1583 pad_13" -> "1584 roll_10" [label="(1, 16, 16, 384)", style=solid]; -"1584 roll_10" -> "1585 view_61" [label="(1, 16, 16, 384)", style=solid]; -"1585 view_61" -> "1586 permute_51" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"1586 permute_51" -> "1587 reshape_49" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"1587 reshape_49" -> "1589 reshape_49_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"1587 reshape_49" -> "1620 new_zeros_5" [label="(4, 64, 384)", style=solid]; -"1588 clone_11" -> "1596 linear_70" [label="(1152,)", style=solid]; -"1589 reshape_49_0_0_nncf_smooth_quant_0" -> "1590 quantize_per_tensor_default_69" [label="(4, 64, 384)", style=solid]; -"1590 quantize_per_tensor_default_69" -> "1591 dequantize_per_tensor_default_69" [label="(4, 64, 384)", style=solid]; -"1591 dequantize_per_tensor_default_69" -> "1596 linear_70" [label="(4, 64, 384)", style=solid]; -"1592 linear_70_scale_0" -> "1595 dequantize_per_channel_default_71" [label="(1152,)", style=solid]; -"1593 linear_70_zero_point_0" -> "1595 dequantize_per_channel_default_71" [label="(1152,)", style=solid]; -"1594 _frozen_param71" -> "1595 dequantize_per_channel_default_71" [label="(1152, 384)", style=solid]; -"1595 dequantize_per_channel_default_71" -> "1596 linear_70" [label="(1152, 384)", style=solid]; -"1596 linear_70" -> "1597 reshape_50" [label="(4, 64, 1152)", style=solid]; -"1597 reshape_50" -> "1598 permute_52" [label="(4, 64, 3, 12, 32)", style=solid]; -"1598 permute_52" -> "1599 select_33" [label="(3, 4, 12, 64, 32)", style=solid]; -"1598 permute_52" -> "1600 select_34" [label="(3, 4, 12, 64, 32)", style=solid]; -"1598 permute_52" -> "1601 select_35" [label="(3, 4, 12, 64, 32)", style=solid]; -"1599 select_33" -> "1602 linalg_vector_norm_22" [label="(4, 12, 64, 32)", style=solid]; -"1599 select_33" -> "1604 expand_as_22" [label="(4, 12, 64, 32)", style=solid]; -"1599 select_33" -> "1605 div_22" [label="(4, 12, 64, 32)", style=solid]; -"1600 select_34" -> "1608 linalg_vector_norm_23" [label="(4, 12, 64, 32)", style=solid]; -"1600 select_34" -> "1610 expand_as_23" [label="(4, 12, 64, 32)", style=solid]; -"1600 select_34" -> "1611 div_23" [label="(4, 12, 64, 32)", style=solid]; -"1601 select_35" -> "1638 matmul_23" [label="(4, 12, 64, 32)", style=solid]; -"1602 linalg_vector_norm_22" -> "1603 clamp_min_22" [label="(4, 12, 64, 1)", style=solid]; -"1603 clamp_min_22" -> "1604 expand_as_22" [label="(4, 12, 64, 1)", style=solid]; -"1604 expand_as_22" -> "1605 div_22" [label="(4, 12, 64, 32)", style=solid]; -"1605 div_22" -> "1606 quantize_per_tensor_default_70" [label="(4, 12, 64, 32)", style=solid]; -"1606 quantize_per_tensor_default_70" -> "1607 dequantize_per_tensor_default_70" [label="(4, 12, 64, 32)", style=solid]; -"1607 dequantize_per_tensor_default_70" -> "1615 matmul_22" [label="(4, 12, 64, 32)", style=solid]; -"1608 linalg_vector_norm_23" -> "1609 clamp_min_23" [label="(4, 12, 64, 1)", style=solid]; -"1609 clamp_min_23" -> "1610 expand_as_23" [label="(4, 12, 64, 1)", style=solid]; -"1610 expand_as_23" -> "1611 div_23" [label="(4, 12, 64, 32)", style=solid]; -"1611 div_23" -> "1612 quantize_per_tensor_default_71" [label="(4, 12, 64, 32)", style=solid]; -"1612 quantize_per_tensor_default_71" -> "1613 dequantize_per_tensor_default_71" [label="(4, 12, 64, 32)", style=solid]; -"1613 dequantize_per_tensor_default_71" -> "1614 transpose_22" [label="(4, 12, 64, 32)", style=solid]; -"1614 transpose_22" -> "1615 matmul_22" [label="(4, 12, 32, 64)", style=solid]; -"1615 matmul_22" -> "1618 mul_23" [label="(4, 12, 64, 64)", style=solid]; -"1616 clamp_11" -> "1617 exp_11" [label="(12, 1, 1)", style=solid]; -"1617 exp_11" -> "1618 mul_23" [label="(12, 1, 1)", style=solid]; -"1618 mul_23" -> "1619 add_38" [label="(4, 12, 64, 64)", style=solid]; -"1619 add_38" -> "1631 view_63" [label="(4, 12, 64, 64)", style=solid]; -"1620 new_zeros_5" -> "1621 view_62" [label="(16, 16)", style=solid]; -"1621 view_62" -> "1622 permute_53" [label="(2, 8, 2, 8)", style=solid]; -"1622 permute_53" -> "1623 reshape_51" [label="(2, 2, 8, 8)", style=solid]; -"1623 reshape_51" -> "1624 unsqueeze_32" [label="(4, 64)", style=solid]; -"1623 reshape_51" -> "1625 unsqueeze_33" [label="(4, 64)", style=solid]; -"1624 unsqueeze_32" -> "1626 sub_5" [label="(4, 1, 64)", style=solid]; -"1625 unsqueeze_33" -> "1626 sub_5" [label="(4, 64, 1)", style=solid]; -"1626 sub_5" -> "1627 ne_5" [label="(4, 64, 64)", style=solid]; -"1626 sub_5" -> "1628 masked_fill_10" [label="(4, 64, 64)", style=solid]; -"1626 sub_5" -> "1629 eq_5" [label="(4, 64, 64)", style=solid]; -"1627 ne_5" -> "1628 masked_fill_10" [label="(4, 64, 64)", style=solid]; -"1628 masked_fill_10" -> "1630 masked_fill_11" [label="(4, 64, 64)", style=solid]; -"1629 eq_5" -> "1630 masked_fill_11" [label="(4, 64, 64)", style=solid]; -"1630 masked_fill_11" -> "1632 unsqueeze_34" [label="(4, 64, 64)", style=solid]; -"1631 view_63" -> "1634 add_39" [label="(1, 4, 12, 64, 64)", style=solid]; -"1632 unsqueeze_34" -> "1633 unsqueeze_35" [label="(4, 1, 64, 64)", style=solid]; -"1633 unsqueeze_35" -> "1634 add_39" [label="(1, 4, 1, 64, 64)", style=solid]; -"1634 add_39" -> "1635 view_64" [label="(1, 4, 12, 64, 64)", style=solid]; -"1635 view_64" -> "1636 softmax_11" [label="(4, 12, 64, 64)", style=solid]; -"1636 softmax_11" -> "1637 dropout_44" [label="(4, 12, 64, 64)", style=solid]; -"1637 dropout_44" -> "1638 matmul_23" [label="(4, 12, 64, 64)", style=solid]; -"1638 matmul_23" -> "1639 transpose_23" [label="(4, 12, 64, 32)", style=solid]; -"1639 transpose_23" -> "1640 reshape_52" [label="(4, 64, 12, 32)", style=solid]; -"1640 reshape_52" -> "1641 reshape_52_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"1641 reshape_52_0_0_nncf_smooth_quant_0" -> "1642 quantize_per_tensor_default_72" [label="(4, 64, 384)", style=solid]; -"1642 quantize_per_tensor_default_72" -> "1643 dequantize_per_tensor_default_72" [label="(4, 64, 384)", style=solid]; -"1643 dequantize_per_tensor_default_72" -> "1648 linear_71" [label="(4, 64, 384)", style=solid]; -"1644 linear_71_scale_0" -> "1647 dequantize_per_channel_default_72" [label="(384,)", style=solid]; -"1645 linear_71_zero_point_0" -> "1647 dequantize_per_channel_default_72" [label="(384,)", style=solid]; -"1646 _frozen_param72" -> "1647 dequantize_per_channel_default_72" [label="(384, 384)", style=solid]; -"1647 dequantize_per_channel_default_72" -> "1648 linear_71" [label="(384, 384)", style=solid]; -"1648 linear_71" -> "1649 dropout_45" [label="(4, 64, 384)", style=solid]; -"1649 dropout_45" -> "1650 view_65" [label="(4, 64, 384)", style=solid]; -"1650 view_65" -> "1651 permute_54" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"1651 permute_54" -> "1652 reshape_53" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"1652 reshape_53" -> "1653 roll_11" [label="(1, 16, 16, 384)", style=solid]; -"1653 roll_11" -> "1654 slice_185" [label="(1, 16, 16, 384)", style=solid]; -"1654 slice_185" -> "1655 slice_186" [label="(1, 16, 16, 384)", style=solid]; -"1655 slice_186" -> "1656 slice_187" [label="(1, 14, 16, 384)", style=solid]; -"1656 slice_187" -> "1657 slice_188" [label="(1, 14, 14, 384)", style=solid]; -"1657 slice_188" -> "1658 contiguous_21" [label="(1, 14, 14, 384)", style=solid]; -"1658 contiguous_21" -> "1659 layer_norm_25" [label="(1, 14, 14, 384)", style=solid]; -"1659 layer_norm_25" -> "1660 add_40" [label="(1, 14, 14, 384)", style=solid]; -"1660 add_40" -> "1661 add_40_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; -"1660 add_40" -> "1681 add_41" [label="(1, 14, 14, 384)", style=solid]; -"1661 add_40_0_0_nncf_smooth_quant_0" -> "1662 quantize_per_tensor_default_73" [label="(1, 14, 14, 384)", style=solid]; -"1662 quantize_per_tensor_default_73" -> "1663 dequantize_per_tensor_default_73" [label="(1, 14, 14, 384)", style=solid]; -"1663 dequantize_per_tensor_default_73" -> "1668 linear_72" [label="(1, 14, 14, 384)", style=solid]; -"1664 linear_72_scale_0" -> "1667 dequantize_per_channel_default_73" [label="(1536,)", style=solid]; -"1665 linear_72_zero_point_0" -> "1667 dequantize_per_channel_default_73" [label="(1536,)", style=solid]; -"1666 _frozen_param73" -> "1667 dequantize_per_channel_default_73" [label="(1536, 384)", style=solid]; -"1667 dequantize_per_channel_default_73" -> "1668 linear_72" [label="(1536, 384)", style=solid]; -"1668 linear_72" -> "1669 gelu_11" [label="(1, 14, 14, 1536)", style=solid]; -"1669 gelu_11" -> "1670 dropout_46" [label="(1, 14, 14, 1536)", style=solid]; -"1670 dropout_46" -> "1671 dropout_46_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; -"1671 dropout_46_0_0_nncf_smooth_quant_0" -> "1672 quantize_per_tensor_default_74" [label="(1, 14, 14, 1536)", style=solid]; -"1672 quantize_per_tensor_default_74" -> "1673 dequantize_per_tensor_default_74" [label="(1, 14, 14, 1536)", style=solid]; -"1673 dequantize_per_tensor_default_74" -> "1678 linear_73" [label="(1, 14, 14, 1536)", style=solid]; -"1674 linear_73_scale_0" -> "1677 dequantize_per_channel_default_74" [label="(384,)", style=solid]; -"1675 linear_73_zero_point_0" -> "1677 dequantize_per_channel_default_74" [label="(384,)", style=solid]; -"1676 _frozen_param74" -> "1677 dequantize_per_channel_default_74" [label="(384, 1536)", style=solid]; -"1677 dequantize_per_channel_default_74" -> "1678 linear_73" [label="(384, 1536)", style=solid]; -"1678 linear_73" -> "1679 dropout_47" [label="(1, 14, 14, 384)", style=solid]; -"1679 dropout_47" -> "1680 layer_norm_26" [label="(1, 14, 14, 384)", style=solid]; -"1680 layer_norm_26" -> "1681 add_41" [label="(1, 14, 14, 384)", style=solid]; -"1681 add_41" -> "1703 pad_14" [label="(1, 14, 14, 384)", style=solid]; -"1681 add_41" -> "1762 add_43" [label="(1, 14, 14, 384)", style=solid]; -"1682 features_5_8_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "1687 linear_74" [label="(1, 15, 15, 2)", style=solid]; -"1683 linear_74_scale_0" -> "1686 dequantize_per_channel_default_75" [label="(512,)", style=solid]; -"1684 linear_74_zero_point_0" -> "1686 dequantize_per_channel_default_75" [label="(512,)", style=solid]; -"1685 _frozen_param75" -> "1686 dequantize_per_channel_default_75" [label="(512, 2)", style=solid]; -"1686 dequantize_per_channel_default_75" -> "1687 linear_74" [label="(512, 2)", style=solid]; -"1687 linear_74" -> "1688 relu__12" [label="(1, 15, 15, 512)", style=solid]; -"1688 relu__12" -> "1689 relu__12_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; -"1689 relu__12_0_0_nncf_smooth_quant_0" -> "1694 linear_75" [label="(1, 15, 15, 512)", style=solid]; -"1690 linear_75_scale_0" -> "1693 dequantize_per_channel_default_76" [label="(12,)", style=solid]; -"1691 linear_75_zero_point_0" -> "1693 dequantize_per_channel_default_76" [label="(12,)", style=solid]; -"1692 _frozen_param76" -> "1693 dequantize_per_channel_default_76" [label="(12, 512)", style=solid]; -"1693 dequantize_per_channel_default_76" -> "1694 linear_75" [label="(12, 512)", style=solid]; -"1694 linear_75" -> "1695 view_66" [label="(1, 15, 15, 12)", style=solid]; -"1695 view_66" -> "1696 index_12" [label="(225, 12)", style=solid]; -"1696 index_12" -> "1697 view_67" [label="(4096, 12)", style=solid]; -"1697 view_67" -> "1698 permute_55" [label="(64, 64, 12)", style=solid]; -"1698 permute_55" -> "1699 contiguous_22" [label="(12, 64, 64)", style=solid]; -"1699 contiguous_22" -> "1700 unsqueeze_36" [label="(12, 64, 64)", style=solid]; -"1700 unsqueeze_36" -> "1701 sigmoid_12" [label="(1, 12, 64, 64)", style=solid]; -"1701 sigmoid_12" -> "1702 mul_24" [label="(1, 12, 64, 64)", style=solid]; -"1702 mul_24" -> "1738 add_42" [label="(1, 12, 64, 64)", style=solid]; -"1703 pad_14" -> "1704 view_68" [label="(1, 16, 16, 384)", style=solid]; -"1704 view_68" -> "1705 permute_56" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"1705 permute_56" -> "1706 reshape_54" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"1706 reshape_54" -> "1708 reshape_54_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"1707 clone_12" -> "1715 linear_76" [label="(1152,)", style=solid]; -"1708 reshape_54_0_0_nncf_smooth_quant_0" -> "1709 quantize_per_tensor_default_75" [label="(4, 64, 384)", style=solid]; -"1709 quantize_per_tensor_default_75" -> "1710 dequantize_per_tensor_default_75" [label="(4, 64, 384)", style=solid]; -"1710 dequantize_per_tensor_default_75" -> "1715 linear_76" [label="(4, 64, 384)", style=solid]; -"1711 linear_76_scale_0" -> "1714 dequantize_per_channel_default_77" [label="(1152,)", style=solid]; -"1712 linear_76_zero_point_0" -> "1714 dequantize_per_channel_default_77" [label="(1152,)", style=solid]; -"1713 _frozen_param77" -> "1714 dequantize_per_channel_default_77" [label="(1152, 384)", style=solid]; -"1714 dequantize_per_channel_default_77" -> "1715 linear_76" [label="(1152, 384)", style=solid]; -"1715 linear_76" -> "1716 reshape_55" [label="(4, 64, 1152)", style=solid]; -"1716 reshape_55" -> "1717 permute_57" [label="(4, 64, 3, 12, 32)", style=solid]; -"1717 permute_57" -> "1718 select_36" [label="(3, 4, 12, 64, 32)", style=solid]; -"1717 permute_57" -> "1719 select_37" [label="(3, 4, 12, 64, 32)", style=solid]; -"1717 permute_57" -> "1720 select_38" [label="(3, 4, 12, 64, 32)", style=solid]; -"1718 select_36" -> "1721 linalg_vector_norm_24" [label="(4, 12, 64, 32)", style=solid]; -"1718 select_36" -> "1723 expand_as_24" [label="(4, 12, 64, 32)", style=solid]; -"1718 select_36" -> "1724 div_24" [label="(4, 12, 64, 32)", style=solid]; -"1719 select_37" -> "1727 linalg_vector_norm_25" [label="(4, 12, 64, 32)", style=solid]; -"1719 select_37" -> "1729 expand_as_25" [label="(4, 12, 64, 32)", style=solid]; -"1719 select_37" -> "1730 div_25" [label="(4, 12, 64, 32)", style=solid]; -"1720 select_38" -> "1741 matmul_25" [label="(4, 12, 64, 32)", style=solid]; -"1721 linalg_vector_norm_24" -> "1722 clamp_min_24" [label="(4, 12, 64, 1)", style=solid]; -"1722 clamp_min_24" -> "1723 expand_as_24" [label="(4, 12, 64, 1)", style=solid]; -"1723 expand_as_24" -> "1724 div_24" [label="(4, 12, 64, 32)", style=solid]; -"1724 div_24" -> "1725 quantize_per_tensor_default_76" [label="(4, 12, 64, 32)", style=solid]; -"1725 quantize_per_tensor_default_76" -> "1726 dequantize_per_tensor_default_76" [label="(4, 12, 64, 32)", style=solid]; -"1726 dequantize_per_tensor_default_76" -> "1734 matmul_24" [label="(4, 12, 64, 32)", style=solid]; -"1727 linalg_vector_norm_25" -> "1728 clamp_min_25" [label="(4, 12, 64, 1)", style=solid]; -"1728 clamp_min_25" -> "1729 expand_as_25" [label="(4, 12, 64, 1)", style=solid]; -"1729 expand_as_25" -> "1730 div_25" [label="(4, 12, 64, 32)", style=solid]; -"1730 div_25" -> "1731 quantize_per_tensor_default_77" [label="(4, 12, 64, 32)", style=solid]; -"1731 quantize_per_tensor_default_77" -> "1732 dequantize_per_tensor_default_77" [label="(4, 12, 64, 32)", style=solid]; -"1732 dequantize_per_tensor_default_77" -> "1733 transpose_24" [label="(4, 12, 64, 32)", style=solid]; -"1733 transpose_24" -> "1734 matmul_24" [label="(4, 12, 32, 64)", style=solid]; -"1734 matmul_24" -> "1737 mul_25" [label="(4, 12, 64, 64)", style=solid]; -"1735 clamp_12" -> "1736 exp_12" [label="(12, 1, 1)", style=solid]; -"1736 exp_12" -> "1737 mul_25" [label="(12, 1, 1)", style=solid]; -"1737 mul_25" -> "1738 add_42" [label="(4, 12, 64, 64)", style=solid]; -"1738 add_42" -> "1739 softmax_12" [label="(4, 12, 64, 64)", style=solid]; -"1739 softmax_12" -> "1740 dropout_48" [label="(4, 12, 64, 64)", style=solid]; -"1740 dropout_48" -> "1741 matmul_25" [label="(4, 12, 64, 64)", style=solid]; -"1741 matmul_25" -> "1742 transpose_25" [label="(4, 12, 64, 32)", style=solid]; -"1742 transpose_25" -> "1743 reshape_56" [label="(4, 64, 12, 32)", style=solid]; -"1743 reshape_56" -> "1744 reshape_56_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"1744 reshape_56_0_0_nncf_smooth_quant_0" -> "1745 quantize_per_tensor_default_78" [label="(4, 64, 384)", style=solid]; -"1745 quantize_per_tensor_default_78" -> "1746 dequantize_per_tensor_default_78" [label="(4, 64, 384)", style=solid]; -"1746 dequantize_per_tensor_default_78" -> "1751 linear_77" [label="(4, 64, 384)", style=solid]; -"1747 linear_77_scale_0" -> "1750 dequantize_per_channel_default_78" [label="(384,)", style=solid]; -"1748 linear_77_zero_point_0" -> "1750 dequantize_per_channel_default_78" [label="(384,)", style=solid]; -"1749 _frozen_param78" -> "1750 dequantize_per_channel_default_78" [label="(384, 384)", style=solid]; -"1750 dequantize_per_channel_default_78" -> "1751 linear_77" [label="(384, 384)", style=solid]; -"1751 linear_77" -> "1752 dropout_49" [label="(4, 64, 384)", style=solid]; -"1752 dropout_49" -> "1753 view_69" [label="(4, 64, 384)", style=solid]; -"1753 view_69" -> "1754 permute_58" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"1754 permute_58" -> "1755 reshape_57" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"1755 reshape_57" -> "1756 slice_190" [label="(1, 16, 16, 384)", style=solid]; -"1756 slice_190" -> "1757 slice_191" [label="(1, 16, 16, 384)", style=solid]; -"1757 slice_191" -> "1758 slice_192" [label="(1, 14, 16, 384)", style=solid]; -"1758 slice_192" -> "1759 slice_193" [label="(1, 14, 14, 384)", style=solid]; -"1759 slice_193" -> "1760 contiguous_23" [label="(1, 14, 14, 384)", style=solid]; -"1760 contiguous_23" -> "1761 layer_norm_27" [label="(1, 14, 14, 384)", style=solid]; -"1761 layer_norm_27" -> "1762 add_43" [label="(1, 14, 14, 384)", style=solid]; -"1762 add_43" -> "1763 add_43_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; -"1762 add_43" -> "1783 add_44" [label="(1, 14, 14, 384)", style=solid]; -"1763 add_43_0_0_nncf_smooth_quant_0" -> "1764 quantize_per_tensor_default_79" [label="(1, 14, 14, 384)", style=solid]; -"1764 quantize_per_tensor_default_79" -> "1765 dequantize_per_tensor_default_79" [label="(1, 14, 14, 384)", style=solid]; -"1765 dequantize_per_tensor_default_79" -> "1770 linear_78" [label="(1, 14, 14, 384)", style=solid]; -"1766 linear_78_scale_0" -> "1769 dequantize_per_channel_default_79" [label="(1536,)", style=solid]; -"1767 linear_78_zero_point_0" -> "1769 dequantize_per_channel_default_79" [label="(1536,)", style=solid]; -"1768 _frozen_param79" -> "1769 dequantize_per_channel_default_79" [label="(1536, 384)", style=solid]; -"1769 dequantize_per_channel_default_79" -> "1770 linear_78" [label="(1536, 384)", style=solid]; -"1770 linear_78" -> "1771 gelu_12" [label="(1, 14, 14, 1536)", style=solid]; -"1771 gelu_12" -> "1772 dropout_50" [label="(1, 14, 14, 1536)", style=solid]; -"1772 dropout_50" -> "1773 dropout_50_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; -"1773 dropout_50_0_0_nncf_smooth_quant_0" -> "1774 quantize_per_tensor_default_80" [label="(1, 14, 14, 1536)", style=solid]; -"1774 quantize_per_tensor_default_80" -> "1775 dequantize_per_tensor_default_80" [label="(1, 14, 14, 1536)", style=solid]; -"1775 dequantize_per_tensor_default_80" -> "1780 linear_79" [label="(1, 14, 14, 1536)", style=solid]; -"1776 linear_79_scale_0" -> "1779 dequantize_per_channel_default_80" [label="(384,)", style=solid]; -"1777 linear_79_zero_point_0" -> "1779 dequantize_per_channel_default_80" [label="(384,)", style=solid]; -"1778 _frozen_param80" -> "1779 dequantize_per_channel_default_80" [label="(384, 1536)", style=solid]; -"1779 dequantize_per_channel_default_80" -> "1780 linear_79" [label="(384, 1536)", style=solid]; -"1780 linear_79" -> "1781 dropout_51" [label="(1, 14, 14, 384)", style=solid]; -"1781 dropout_51" -> "1782 layer_norm_28" [label="(1, 14, 14, 384)", style=solid]; -"1782 layer_norm_28" -> "1783 add_44" [label="(1, 14, 14, 384)", style=solid]; -"1783 add_44" -> "1805 pad_15" [label="(1, 14, 14, 384)", style=solid]; -"1783 add_44" -> "1882 add_47" [label="(1, 14, 14, 384)", style=solid]; -"1784 features_5_9_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "1789 linear_80" [label="(1, 15, 15, 2)", style=solid]; -"1785 linear_80_scale_0" -> "1788 dequantize_per_channel_default_81" [label="(512,)", style=solid]; -"1786 linear_80_zero_point_0" -> "1788 dequantize_per_channel_default_81" [label="(512,)", style=solid]; -"1787 _frozen_param81" -> "1788 dequantize_per_channel_default_81" [label="(512, 2)", style=solid]; -"1788 dequantize_per_channel_default_81" -> "1789 linear_80" [label="(512, 2)", style=solid]; -"1789 linear_80" -> "1790 relu__13" [label="(1, 15, 15, 512)", style=solid]; -"1790 relu__13" -> "1791 relu__13_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; -"1791 relu__13_0_0_nncf_smooth_quant_0" -> "1796 linear_81" [label="(1, 15, 15, 512)", style=solid]; -"1792 linear_81_scale_0" -> "1795 dequantize_per_channel_default_82" [label="(12,)", style=solid]; -"1793 linear_81_zero_point_0" -> "1795 dequantize_per_channel_default_82" [label="(12,)", style=solid]; -"1794 _frozen_param82" -> "1795 dequantize_per_channel_default_82" [label="(12, 512)", style=solid]; -"1795 dequantize_per_channel_default_82" -> "1796 linear_81" [label="(12, 512)", style=solid]; -"1796 linear_81" -> "1797 view_70" [label="(1, 15, 15, 12)", style=solid]; -"1797 view_70" -> "1798 index_13" [label="(225, 12)", style=solid]; -"1798 index_13" -> "1799 view_71" [label="(4096, 12)", style=solid]; -"1799 view_71" -> "1800 permute_59" [label="(64, 64, 12)", style=solid]; -"1800 permute_59" -> "1801 contiguous_24" [label="(12, 64, 64)", style=solid]; -"1801 contiguous_24" -> "1802 unsqueeze_37" [label="(12, 64, 64)", style=solid]; -"1802 unsqueeze_37" -> "1803 sigmoid_13" [label="(1, 12, 64, 64)", style=solid]; -"1803 sigmoid_13" -> "1804 mul_26" [label="(1, 12, 64, 64)", style=solid]; -"1804 mul_26" -> "1841 add_45" [label="(1, 12, 64, 64)", style=solid]; -"1805 pad_15" -> "1806 roll_12" [label="(1, 16, 16, 384)", style=solid]; -"1806 roll_12" -> "1807 view_72" [label="(1, 16, 16, 384)", style=solid]; -"1807 view_72" -> "1808 permute_60" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"1808 permute_60" -> "1809 reshape_58" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"1809 reshape_58" -> "1811 reshape_58_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"1809 reshape_58" -> "1842 new_zeros_6" [label="(4, 64, 384)", style=solid]; -"1810 clone_13" -> "1818 linear_82" [label="(1152,)", style=solid]; -"1811 reshape_58_0_0_nncf_smooth_quant_0" -> "1812 quantize_per_tensor_default_81" [label="(4, 64, 384)", style=solid]; -"1812 quantize_per_tensor_default_81" -> "1813 dequantize_per_tensor_default_81" [label="(4, 64, 384)", style=solid]; -"1813 dequantize_per_tensor_default_81" -> "1818 linear_82" [label="(4, 64, 384)", style=solid]; -"1814 linear_82_scale_0" -> "1817 dequantize_per_channel_default_83" [label="(1152,)", style=solid]; -"1815 linear_82_zero_point_0" -> "1817 dequantize_per_channel_default_83" [label="(1152,)", style=solid]; -"1816 _frozen_param83" -> "1817 dequantize_per_channel_default_83" [label="(1152, 384)", style=solid]; -"1817 dequantize_per_channel_default_83" -> "1818 linear_82" [label="(1152, 384)", style=solid]; -"1818 linear_82" -> "1819 reshape_59" [label="(4, 64, 1152)", style=solid]; -"1819 reshape_59" -> "1820 permute_61" [label="(4, 64, 3, 12, 32)", style=solid]; -"1820 permute_61" -> "1821 select_39" [label="(3, 4, 12, 64, 32)", style=solid]; -"1820 permute_61" -> "1822 select_40" [label="(3, 4, 12, 64, 32)", style=solid]; -"1820 permute_61" -> "1823 select_41" [label="(3, 4, 12, 64, 32)", style=solid]; -"1821 select_39" -> "1824 linalg_vector_norm_26" [label="(4, 12, 64, 32)", style=solid]; -"1821 select_39" -> "1826 expand_as_26" [label="(4, 12, 64, 32)", style=solid]; -"1821 select_39" -> "1827 div_26" [label="(4, 12, 64, 32)", style=solid]; -"1822 select_40" -> "1830 linalg_vector_norm_27" [label="(4, 12, 64, 32)", style=solid]; -"1822 select_40" -> "1832 expand_as_27" [label="(4, 12, 64, 32)", style=solid]; -"1822 select_40" -> "1833 div_27" [label="(4, 12, 64, 32)", style=solid]; -"1823 select_41" -> "1860 matmul_27" [label="(4, 12, 64, 32)", style=solid]; -"1824 linalg_vector_norm_26" -> "1825 clamp_min_26" [label="(4, 12, 64, 1)", style=solid]; -"1825 clamp_min_26" -> "1826 expand_as_26" [label="(4, 12, 64, 1)", style=solid]; -"1826 expand_as_26" -> "1827 div_26" [label="(4, 12, 64, 32)", style=solid]; -"1827 div_26" -> "1828 quantize_per_tensor_default_82" [label="(4, 12, 64, 32)", style=solid]; -"1828 quantize_per_tensor_default_82" -> "1829 dequantize_per_tensor_default_82" [label="(4, 12, 64, 32)", style=solid]; -"1829 dequantize_per_tensor_default_82" -> "1837 matmul_26" [label="(4, 12, 64, 32)", style=solid]; -"1830 linalg_vector_norm_27" -> "1831 clamp_min_27" [label="(4, 12, 64, 1)", style=solid]; -"1831 clamp_min_27" -> "1832 expand_as_27" [label="(4, 12, 64, 1)", style=solid]; -"1832 expand_as_27" -> "1833 div_27" [label="(4, 12, 64, 32)", style=solid]; -"1833 div_27" -> "1834 quantize_per_tensor_default_83" [label="(4, 12, 64, 32)", style=solid]; -"1834 quantize_per_tensor_default_83" -> "1835 dequantize_per_tensor_default_83" [label="(4, 12, 64, 32)", style=solid]; -"1835 dequantize_per_tensor_default_83" -> "1836 transpose_26" [label="(4, 12, 64, 32)", style=solid]; -"1836 transpose_26" -> "1837 matmul_26" [label="(4, 12, 32, 64)", style=solid]; -"1837 matmul_26" -> "1840 mul_27" [label="(4, 12, 64, 64)", style=solid]; -"1838 clamp_13" -> "1839 exp_13" [label="(12, 1, 1)", style=solid]; -"1839 exp_13" -> "1840 mul_27" [label="(12, 1, 1)", style=solid]; -"1840 mul_27" -> "1841 add_45" [label="(4, 12, 64, 64)", style=solid]; -"1841 add_45" -> "1853 view_74" [label="(4, 12, 64, 64)", style=solid]; -"1842 new_zeros_6" -> "1843 view_73" [label="(16, 16)", style=solid]; -"1843 view_73" -> "1844 permute_62" [label="(2, 8, 2, 8)", style=solid]; -"1844 permute_62" -> "1845 reshape_60" [label="(2, 2, 8, 8)", style=solid]; -"1845 reshape_60" -> "1846 unsqueeze_38" [label="(4, 64)", style=solid]; -"1845 reshape_60" -> "1847 unsqueeze_39" [label="(4, 64)", style=solid]; -"1846 unsqueeze_38" -> "1848 sub_6" [label="(4, 1, 64)", style=solid]; -"1847 unsqueeze_39" -> "1848 sub_6" [label="(4, 64, 1)", style=solid]; -"1848 sub_6" -> "1849 ne_6" [label="(4, 64, 64)", style=solid]; -"1848 sub_6" -> "1850 masked_fill_12" [label="(4, 64, 64)", style=solid]; -"1848 sub_6" -> "1851 eq_6" [label="(4, 64, 64)", style=solid]; -"1849 ne_6" -> "1850 masked_fill_12" [label="(4, 64, 64)", style=solid]; -"1850 masked_fill_12" -> "1852 masked_fill_13" [label="(4, 64, 64)", style=solid]; -"1851 eq_6" -> "1852 masked_fill_13" [label="(4, 64, 64)", style=solid]; -"1852 masked_fill_13" -> "1854 unsqueeze_40" [label="(4, 64, 64)", style=solid]; -"1853 view_74" -> "1856 add_46" [label="(1, 4, 12, 64, 64)", style=solid]; -"1854 unsqueeze_40" -> "1855 unsqueeze_41" [label="(4, 1, 64, 64)", style=solid]; -"1855 unsqueeze_41" -> "1856 add_46" [label="(1, 4, 1, 64, 64)", style=solid]; -"1856 add_46" -> "1857 view_75" [label="(1, 4, 12, 64, 64)", style=solid]; -"1857 view_75" -> "1858 softmax_13" [label="(4, 12, 64, 64)", style=solid]; -"1858 softmax_13" -> "1859 dropout_52" [label="(4, 12, 64, 64)", style=solid]; -"1859 dropout_52" -> "1860 matmul_27" [label="(4, 12, 64, 64)", style=solid]; -"1860 matmul_27" -> "1861 transpose_27" [label="(4, 12, 64, 32)", style=solid]; -"1861 transpose_27" -> "1862 reshape_61" [label="(4, 64, 12, 32)", style=solid]; -"1862 reshape_61" -> "1863 reshape_61_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"1863 reshape_61_0_0_nncf_smooth_quant_0" -> "1864 quantize_per_tensor_default_84" [label="(4, 64, 384)", style=solid]; -"1864 quantize_per_tensor_default_84" -> "1865 dequantize_per_tensor_default_84" [label="(4, 64, 384)", style=solid]; -"1865 dequantize_per_tensor_default_84" -> "1870 linear_83" [label="(4, 64, 384)", style=solid]; -"1866 linear_83_scale_0" -> "1869 dequantize_per_channel_default_84" [label="(384,)", style=solid]; -"1867 linear_83_zero_point_0" -> "1869 dequantize_per_channel_default_84" [label="(384,)", style=solid]; -"1868 _frozen_param84" -> "1869 dequantize_per_channel_default_84" [label="(384, 384)", style=solid]; -"1869 dequantize_per_channel_default_84" -> "1870 linear_83" [label="(384, 384)", style=solid]; -"1870 linear_83" -> "1871 dropout_53" [label="(4, 64, 384)", style=solid]; -"1871 dropout_53" -> "1872 view_76" [label="(4, 64, 384)", style=solid]; -"1872 view_76" -> "1873 permute_63" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"1873 permute_63" -> "1874 reshape_62" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"1874 reshape_62" -> "1875 roll_13" [label="(1, 16, 16, 384)", style=solid]; -"1875 roll_13" -> "1876 slice_213" [label="(1, 16, 16, 384)", style=solid]; -"1876 slice_213" -> "1877 slice_214" [label="(1, 16, 16, 384)", style=solid]; -"1877 slice_214" -> "1878 slice_215" [label="(1, 14, 16, 384)", style=solid]; -"1878 slice_215" -> "1879 slice_216" [label="(1, 14, 14, 384)", style=solid]; -"1879 slice_216" -> "1880 contiguous_25" [label="(1, 14, 14, 384)", style=solid]; -"1880 contiguous_25" -> "1881 layer_norm_29" [label="(1, 14, 14, 384)", style=solid]; -"1881 layer_norm_29" -> "1882 add_47" [label="(1, 14, 14, 384)", style=solid]; -"1882 add_47" -> "1883 add_47_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; -"1882 add_47" -> "1903 add_48" [label="(1, 14, 14, 384)", style=solid]; -"1883 add_47_0_0_nncf_smooth_quant_0" -> "1884 quantize_per_tensor_default_85" [label="(1, 14, 14, 384)", style=solid]; -"1884 quantize_per_tensor_default_85" -> "1885 dequantize_per_tensor_default_85" [label="(1, 14, 14, 384)", style=solid]; -"1885 dequantize_per_tensor_default_85" -> "1890 linear_84" [label="(1, 14, 14, 384)", style=solid]; -"1886 linear_84_scale_0" -> "1889 dequantize_per_channel_default_85" [label="(1536,)", style=solid]; -"1887 linear_84_zero_point_0" -> "1889 dequantize_per_channel_default_85" [label="(1536,)", style=solid]; -"1888 _frozen_param85" -> "1889 dequantize_per_channel_default_85" [label="(1536, 384)", style=solid]; -"1889 dequantize_per_channel_default_85" -> "1890 linear_84" [label="(1536, 384)", style=solid]; -"1890 linear_84" -> "1891 gelu_13" [label="(1, 14, 14, 1536)", style=solid]; -"1891 gelu_13" -> "1892 dropout_54" [label="(1, 14, 14, 1536)", style=solid]; -"1892 dropout_54" -> "1893 dropout_54_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; -"1893 dropout_54_0_0_nncf_smooth_quant_0" -> "1894 quantize_per_tensor_default_86" [label="(1, 14, 14, 1536)", style=solid]; -"1894 quantize_per_tensor_default_86" -> "1895 dequantize_per_tensor_default_86" [label="(1, 14, 14, 1536)", style=solid]; -"1895 dequantize_per_tensor_default_86" -> "1900 linear_85" [label="(1, 14, 14, 1536)", style=solid]; -"1896 linear_85_scale_0" -> "1899 dequantize_per_channel_default_86" [label="(384,)", style=solid]; -"1897 linear_85_zero_point_0" -> "1899 dequantize_per_channel_default_86" [label="(384,)", style=solid]; -"1898 _frozen_param86" -> "1899 dequantize_per_channel_default_86" [label="(384, 1536)", style=solid]; -"1899 dequantize_per_channel_default_86" -> "1900 linear_85" [label="(384, 1536)", style=solid]; -"1900 linear_85" -> "1901 dropout_55" [label="(1, 14, 14, 384)", style=solid]; -"1901 dropout_55" -> "1902 layer_norm_30" [label="(1, 14, 14, 384)", style=solid]; -"1902 layer_norm_30" -> "1903 add_48" [label="(1, 14, 14, 384)", style=solid]; -"1903 add_48" -> "1925 pad_16" [label="(1, 14, 14, 384)", style=solid]; -"1903 add_48" -> "1984 add_50" [label="(1, 14, 14, 384)", style=solid]; -"1904 features_5_10_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "1909 linear_86" [label="(1, 15, 15, 2)", style=solid]; -"1905 linear_86_scale_0" -> "1908 dequantize_per_channel_default_87" [label="(512,)", style=solid]; -"1906 linear_86_zero_point_0" -> "1908 dequantize_per_channel_default_87" [label="(512,)", style=solid]; -"1907 _frozen_param87" -> "1908 dequantize_per_channel_default_87" [label="(512, 2)", style=solid]; -"1908 dequantize_per_channel_default_87" -> "1909 linear_86" [label="(512, 2)", style=solid]; -"1909 linear_86" -> "1910 relu__14" [label="(1, 15, 15, 512)", style=solid]; -"1910 relu__14" -> "1911 relu__14_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; -"1911 relu__14_0_0_nncf_smooth_quant_0" -> "1916 linear_87" [label="(1, 15, 15, 512)", style=solid]; -"1912 linear_87_scale_0" -> "1915 dequantize_per_channel_default_88" [label="(12,)", style=solid]; -"1913 linear_87_zero_point_0" -> "1915 dequantize_per_channel_default_88" [label="(12,)", style=solid]; -"1914 _frozen_param88" -> "1915 dequantize_per_channel_default_88" [label="(12, 512)", style=solid]; -"1915 dequantize_per_channel_default_88" -> "1916 linear_87" [label="(12, 512)", style=solid]; -"1916 linear_87" -> "1917 view_77" [label="(1, 15, 15, 12)", style=solid]; -"1917 view_77" -> "1918 index_14" [label="(225, 12)", style=solid]; -"1918 index_14" -> "1919 view_78" [label="(4096, 12)", style=solid]; -"1919 view_78" -> "1920 permute_64" [label="(64, 64, 12)", style=solid]; -"1920 permute_64" -> "1921 contiguous_26" [label="(12, 64, 64)", style=solid]; -"1921 contiguous_26" -> "1922 unsqueeze_42" [label="(12, 64, 64)", style=solid]; -"1922 unsqueeze_42" -> "1923 sigmoid_14" [label="(1, 12, 64, 64)", style=solid]; -"1923 sigmoid_14" -> "1924 mul_28" [label="(1, 12, 64, 64)", style=solid]; -"1924 mul_28" -> "1960 add_49" [label="(1, 12, 64, 64)", style=solid]; -"1925 pad_16" -> "1926 view_79" [label="(1, 16, 16, 384)", style=solid]; -"1926 view_79" -> "1927 permute_65" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"1927 permute_65" -> "1928 reshape_63" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"1928 reshape_63" -> "1930 reshape_63_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"1929 clone_14" -> "1937 linear_88" [label="(1152,)", style=solid]; -"1930 reshape_63_0_0_nncf_smooth_quant_0" -> "1931 quantize_per_tensor_default_87" [label="(4, 64, 384)", style=solid]; -"1931 quantize_per_tensor_default_87" -> "1932 dequantize_per_tensor_default_87" [label="(4, 64, 384)", style=solid]; -"1932 dequantize_per_tensor_default_87" -> "1937 linear_88" [label="(4, 64, 384)", style=solid]; -"1933 linear_88_scale_0" -> "1936 dequantize_per_channel_default_89" [label="(1152,)", style=solid]; -"1934 linear_88_zero_point_0" -> "1936 dequantize_per_channel_default_89" [label="(1152,)", style=solid]; -"1935 _frozen_param89" -> "1936 dequantize_per_channel_default_89" [label="(1152, 384)", style=solid]; -"1936 dequantize_per_channel_default_89" -> "1937 linear_88" [label="(1152, 384)", style=solid]; -"1937 linear_88" -> "1938 reshape_64" [label="(4, 64, 1152)", style=solid]; -"1938 reshape_64" -> "1939 permute_66" [label="(4, 64, 3, 12, 32)", style=solid]; -"1939 permute_66" -> "1940 select_42" [label="(3, 4, 12, 64, 32)", style=solid]; -"1939 permute_66" -> "1941 select_43" [label="(3, 4, 12, 64, 32)", style=solid]; -"1939 permute_66" -> "1942 select_44" [label="(3, 4, 12, 64, 32)", style=solid]; -"1940 select_42" -> "1943 linalg_vector_norm_28" [label="(4, 12, 64, 32)", style=solid]; -"1940 select_42" -> "1945 expand_as_28" [label="(4, 12, 64, 32)", style=solid]; -"1940 select_42" -> "1946 div_28" [label="(4, 12, 64, 32)", style=solid]; -"1941 select_43" -> "1949 linalg_vector_norm_29" [label="(4, 12, 64, 32)", style=solid]; -"1941 select_43" -> "1951 expand_as_29" [label="(4, 12, 64, 32)", style=solid]; -"1941 select_43" -> "1952 div_29" [label="(4, 12, 64, 32)", style=solid]; -"1942 select_44" -> "1963 matmul_29" [label="(4, 12, 64, 32)", style=solid]; -"1943 linalg_vector_norm_28" -> "1944 clamp_min_28" [label="(4, 12, 64, 1)", style=solid]; -"1944 clamp_min_28" -> "1945 expand_as_28" [label="(4, 12, 64, 1)", style=solid]; -"1945 expand_as_28" -> "1946 div_28" [label="(4, 12, 64, 32)", style=solid]; -"1946 div_28" -> "1947 quantize_per_tensor_default_88" [label="(4, 12, 64, 32)", style=solid]; -"1947 quantize_per_tensor_default_88" -> "1948 dequantize_per_tensor_default_88" [label="(4, 12, 64, 32)", style=solid]; -"1948 dequantize_per_tensor_default_88" -> "1956 matmul_28" [label="(4, 12, 64, 32)", style=solid]; -"1949 linalg_vector_norm_29" -> "1950 clamp_min_29" [label="(4, 12, 64, 1)", style=solid]; -"1950 clamp_min_29" -> "1951 expand_as_29" [label="(4, 12, 64, 1)", style=solid]; -"1951 expand_as_29" -> "1952 div_29" [label="(4, 12, 64, 32)", style=solid]; -"1952 div_29" -> "1953 quantize_per_tensor_default_89" [label="(4, 12, 64, 32)", style=solid]; -"1953 quantize_per_tensor_default_89" -> "1954 dequantize_per_tensor_default_89" [label="(4, 12, 64, 32)", style=solid]; -"1954 dequantize_per_tensor_default_89" -> "1955 transpose_28" [label="(4, 12, 64, 32)", style=solid]; -"1955 transpose_28" -> "1956 matmul_28" [label="(4, 12, 32, 64)", style=solid]; -"1956 matmul_28" -> "1959 mul_29" [label="(4, 12, 64, 64)", style=solid]; -"1957 clamp_14" -> "1958 exp_14" [label="(12, 1, 1)", style=solid]; -"1958 exp_14" -> "1959 mul_29" [label="(12, 1, 1)", style=solid]; -"1959 mul_29" -> "1960 add_49" [label="(4, 12, 64, 64)", style=solid]; -"1960 add_49" -> "1961 softmax_14" [label="(4, 12, 64, 64)", style=solid]; -"1961 softmax_14" -> "1962 dropout_56" [label="(4, 12, 64, 64)", style=solid]; -"1962 dropout_56" -> "1963 matmul_29" [label="(4, 12, 64, 64)", style=solid]; -"1963 matmul_29" -> "1964 transpose_29" [label="(4, 12, 64, 32)", style=solid]; -"1964 transpose_29" -> "1965 reshape_65" [label="(4, 64, 12, 32)", style=solid]; -"1965 reshape_65" -> "1966 reshape_65_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"1966 reshape_65_0_0_nncf_smooth_quant_0" -> "1967 quantize_per_tensor_default_90" [label="(4, 64, 384)", style=solid]; -"1967 quantize_per_tensor_default_90" -> "1968 dequantize_per_tensor_default_90" [label="(4, 64, 384)", style=solid]; -"1968 dequantize_per_tensor_default_90" -> "1973 linear_89" [label="(4, 64, 384)", style=solid]; -"1969 linear_89_scale_0" -> "1972 dequantize_per_channel_default_90" [label="(384,)", style=solid]; -"1970 linear_89_zero_point_0" -> "1972 dequantize_per_channel_default_90" [label="(384,)", style=solid]; -"1971 _frozen_param90" -> "1972 dequantize_per_channel_default_90" [label="(384, 384)", style=solid]; -"1972 dequantize_per_channel_default_90" -> "1973 linear_89" [label="(384, 384)", style=solid]; -"1973 linear_89" -> "1974 dropout_57" [label="(4, 64, 384)", style=solid]; -"1974 dropout_57" -> "1975 view_80" [label="(4, 64, 384)", style=solid]; -"1975 view_80" -> "1976 permute_67" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"1976 permute_67" -> "1977 reshape_66" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"1977 reshape_66" -> "1978 slice_218" [label="(1, 16, 16, 384)", style=solid]; -"1978 slice_218" -> "1979 slice_219" [label="(1, 16, 16, 384)", style=solid]; -"1979 slice_219" -> "1980 slice_220" [label="(1, 14, 16, 384)", style=solid]; -"1980 slice_220" -> "1981 slice_221" [label="(1, 14, 14, 384)", style=solid]; -"1981 slice_221" -> "1982 contiguous_27" [label="(1, 14, 14, 384)", style=solid]; -"1982 contiguous_27" -> "1983 layer_norm_31" [label="(1, 14, 14, 384)", style=solid]; -"1983 layer_norm_31" -> "1984 add_50" [label="(1, 14, 14, 384)", style=solid]; -"1984 add_50" -> "1985 add_50_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; -"1984 add_50" -> "2005 add_51" [label="(1, 14, 14, 384)", style=solid]; -"1985 add_50_0_0_nncf_smooth_quant_0" -> "1986 quantize_per_tensor_default_91" [label="(1, 14, 14, 384)", style=solid]; -"1986 quantize_per_tensor_default_91" -> "1987 dequantize_per_tensor_default_91" [label="(1, 14, 14, 384)", style=solid]; -"1987 dequantize_per_tensor_default_91" -> "1992 linear_90" [label="(1, 14, 14, 384)", style=solid]; -"1988 linear_90_scale_0" -> "1991 dequantize_per_channel_default_91" [label="(1536,)", style=solid]; -"1989 linear_90_zero_point_0" -> "1991 dequantize_per_channel_default_91" [label="(1536,)", style=solid]; -"1990 _frozen_param91" -> "1991 dequantize_per_channel_default_91" [label="(1536, 384)", style=solid]; -"1991 dequantize_per_channel_default_91" -> "1992 linear_90" [label="(1536, 384)", style=solid]; -"1992 linear_90" -> "1993 gelu_14" [label="(1, 14, 14, 1536)", style=solid]; -"1993 gelu_14" -> "1994 dropout_58" [label="(1, 14, 14, 1536)", style=solid]; -"1994 dropout_58" -> "1995 dropout_58_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; -"1995 dropout_58_0_0_nncf_smooth_quant_0" -> "1996 quantize_per_tensor_default_92" [label="(1, 14, 14, 1536)", style=solid]; -"1996 quantize_per_tensor_default_92" -> "1997 dequantize_per_tensor_default_92" [label="(1, 14, 14, 1536)", style=solid]; -"1997 dequantize_per_tensor_default_92" -> "2002 linear_91" [label="(1, 14, 14, 1536)", style=solid]; -"1998 linear_91_scale_0" -> "2001 dequantize_per_channel_default_92" [label="(384,)", style=solid]; -"1999 linear_91_zero_point_0" -> "2001 dequantize_per_channel_default_92" [label="(384,)", style=solid]; -"2000 _frozen_param92" -> "2001 dequantize_per_channel_default_92" [label="(384, 1536)", style=solid]; -"2001 dequantize_per_channel_default_92" -> "2002 linear_91" [label="(384, 1536)", style=solid]; -"2002 linear_91" -> "2003 dropout_59" [label="(1, 14, 14, 384)", style=solid]; -"2003 dropout_59" -> "2004 layer_norm_32" [label="(1, 14, 14, 384)", style=solid]; -"2004 layer_norm_32" -> "2005 add_51" [label="(1, 14, 14, 384)", style=solid]; -"2005 add_51" -> "2027 pad_17" [label="(1, 14, 14, 384)", style=solid]; -"2005 add_51" -> "2104 add_54" [label="(1, 14, 14, 384)", style=solid]; -"2006 features_5_11_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "2011 linear_92" [label="(1, 15, 15, 2)", style=solid]; -"2007 linear_92_scale_0" -> "2010 dequantize_per_channel_default_93" [label="(512,)", style=solid]; -"2008 linear_92_zero_point_0" -> "2010 dequantize_per_channel_default_93" [label="(512,)", style=solid]; -"2009 _frozen_param93" -> "2010 dequantize_per_channel_default_93" [label="(512, 2)", style=solid]; -"2010 dequantize_per_channel_default_93" -> "2011 linear_92" [label="(512, 2)", style=solid]; -"2011 linear_92" -> "2012 relu__15" [label="(1, 15, 15, 512)", style=solid]; -"2012 relu__15" -> "2013 relu__15_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; -"2013 relu__15_0_0_nncf_smooth_quant_0" -> "2018 linear_93" [label="(1, 15, 15, 512)", style=solid]; -"2014 linear_93_scale_0" -> "2017 dequantize_per_channel_default_94" [label="(12,)", style=solid]; -"2015 linear_93_zero_point_0" -> "2017 dequantize_per_channel_default_94" [label="(12,)", style=solid]; -"2016 _frozen_param94" -> "2017 dequantize_per_channel_default_94" [label="(12, 512)", style=solid]; -"2017 dequantize_per_channel_default_94" -> "2018 linear_93" [label="(12, 512)", style=solid]; -"2018 linear_93" -> "2019 view_81" [label="(1, 15, 15, 12)", style=solid]; -"2019 view_81" -> "2020 index_15" [label="(225, 12)", style=solid]; -"2020 index_15" -> "2021 view_82" [label="(4096, 12)", style=solid]; -"2021 view_82" -> "2022 permute_68" [label="(64, 64, 12)", style=solid]; -"2022 permute_68" -> "2023 contiguous_28" [label="(12, 64, 64)", style=solid]; -"2023 contiguous_28" -> "2024 unsqueeze_43" [label="(12, 64, 64)", style=solid]; -"2024 unsqueeze_43" -> "2025 sigmoid_15" [label="(1, 12, 64, 64)", style=solid]; -"2025 sigmoid_15" -> "2026 mul_30" [label="(1, 12, 64, 64)", style=solid]; -"2026 mul_30" -> "2063 add_52" [label="(1, 12, 64, 64)", style=solid]; -"2027 pad_17" -> "2028 roll_14" [label="(1, 16, 16, 384)", style=solid]; -"2028 roll_14" -> "2029 view_83" [label="(1, 16, 16, 384)", style=solid]; -"2029 view_83" -> "2030 permute_69" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2030 permute_69" -> "2031 reshape_67" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2031 reshape_67" -> "2033 reshape_67_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"2031 reshape_67" -> "2064 new_zeros_7" [label="(4, 64, 384)", style=solid]; -"2032 clone_15" -> "2040 linear_94" [label="(1152,)", style=solid]; -"2033 reshape_67_0_0_nncf_smooth_quant_0" -> "2034 quantize_per_tensor_default_93" [label="(4, 64, 384)", style=solid]; -"2034 quantize_per_tensor_default_93" -> "2035 dequantize_per_tensor_default_93" [label="(4, 64, 384)", style=solid]; -"2035 dequantize_per_tensor_default_93" -> "2040 linear_94" [label="(4, 64, 384)", style=solid]; -"2036 linear_94_scale_0" -> "2039 dequantize_per_channel_default_95" [label="(1152,)", style=solid]; -"2037 linear_94_zero_point_0" -> "2039 dequantize_per_channel_default_95" [label="(1152,)", style=solid]; -"2038 _frozen_param95" -> "2039 dequantize_per_channel_default_95" [label="(1152, 384)", style=solid]; -"2039 dequantize_per_channel_default_95" -> "2040 linear_94" [label="(1152, 384)", style=solid]; -"2040 linear_94" -> "2041 reshape_68" [label="(4, 64, 1152)", style=solid]; -"2041 reshape_68" -> "2042 permute_70" [label="(4, 64, 3, 12, 32)", style=solid]; -"2042 permute_70" -> "2043 select_45" [label="(3, 4, 12, 64, 32)", style=solid]; -"2042 permute_70" -> "2044 select_46" [label="(3, 4, 12, 64, 32)", style=solid]; -"2042 permute_70" -> "2045 select_47" [label="(3, 4, 12, 64, 32)", style=solid]; -"2043 select_45" -> "2046 linalg_vector_norm_30" [label="(4, 12, 64, 32)", style=solid]; -"2043 select_45" -> "2048 expand_as_30" [label="(4, 12, 64, 32)", style=solid]; -"2043 select_45" -> "2049 div_30" [label="(4, 12, 64, 32)", style=solid]; -"2044 select_46" -> "2052 linalg_vector_norm_31" [label="(4, 12, 64, 32)", style=solid]; -"2044 select_46" -> "2054 expand_as_31" [label="(4, 12, 64, 32)", style=solid]; -"2044 select_46" -> "2055 div_31" [label="(4, 12, 64, 32)", style=solid]; -"2045 select_47" -> "2082 matmul_31" [label="(4, 12, 64, 32)", style=solid]; -"2046 linalg_vector_norm_30" -> "2047 clamp_min_30" [label="(4, 12, 64, 1)", style=solid]; -"2047 clamp_min_30" -> "2048 expand_as_30" [label="(4, 12, 64, 1)", style=solid]; -"2048 expand_as_30" -> "2049 div_30" [label="(4, 12, 64, 32)", style=solid]; -"2049 div_30" -> "2050 quantize_per_tensor_default_94" [label="(4, 12, 64, 32)", style=solid]; -"2050 quantize_per_tensor_default_94" -> "2051 dequantize_per_tensor_default_94" [label="(4, 12, 64, 32)", style=solid]; -"2051 dequantize_per_tensor_default_94" -> "2059 matmul_30" [label="(4, 12, 64, 32)", style=solid]; -"2052 linalg_vector_norm_31" -> "2053 clamp_min_31" [label="(4, 12, 64, 1)", style=solid]; -"2053 clamp_min_31" -> "2054 expand_as_31" [label="(4, 12, 64, 1)", style=solid]; -"2054 expand_as_31" -> "2055 div_31" [label="(4, 12, 64, 32)", style=solid]; -"2055 div_31" -> "2056 quantize_per_tensor_default_95" [label="(4, 12, 64, 32)", style=solid]; -"2056 quantize_per_tensor_default_95" -> "2057 dequantize_per_tensor_default_95" [label="(4, 12, 64, 32)", style=solid]; -"2057 dequantize_per_tensor_default_95" -> "2058 transpose_30" [label="(4, 12, 64, 32)", style=solid]; -"2058 transpose_30" -> "2059 matmul_30" [label="(4, 12, 32, 64)", style=solid]; -"2059 matmul_30" -> "2062 mul_31" [label="(4, 12, 64, 64)", style=solid]; -"2060 clamp_15" -> "2061 exp_15" [label="(12, 1, 1)", style=solid]; -"2061 exp_15" -> "2062 mul_31" [label="(12, 1, 1)", style=solid]; -"2062 mul_31" -> "2063 add_52" [label="(4, 12, 64, 64)", style=solid]; -"2063 add_52" -> "2075 view_85" [label="(4, 12, 64, 64)", style=solid]; -"2064 new_zeros_7" -> "2065 view_84" [label="(16, 16)", style=solid]; -"2065 view_84" -> "2066 permute_71" [label="(2, 8, 2, 8)", style=solid]; -"2066 permute_71" -> "2067 reshape_69" [label="(2, 2, 8, 8)", style=solid]; -"2067 reshape_69" -> "2068 unsqueeze_44" [label="(4, 64)", style=solid]; -"2067 reshape_69" -> "2069 unsqueeze_45" [label="(4, 64)", style=solid]; -"2068 unsqueeze_44" -> "2070 sub_7" [label="(4, 1, 64)", style=solid]; -"2069 unsqueeze_45" -> "2070 sub_7" [label="(4, 64, 1)", style=solid]; -"2070 sub_7" -> "2071 ne_7" [label="(4, 64, 64)", style=solid]; -"2070 sub_7" -> "2072 masked_fill_14" [label="(4, 64, 64)", style=solid]; -"2070 sub_7" -> "2073 eq_7" [label="(4, 64, 64)", style=solid]; -"2071 ne_7" -> "2072 masked_fill_14" [label="(4, 64, 64)", style=solid]; -"2072 masked_fill_14" -> "2074 masked_fill_15" [label="(4, 64, 64)", style=solid]; -"2073 eq_7" -> "2074 masked_fill_15" [label="(4, 64, 64)", style=solid]; -"2074 masked_fill_15" -> "2076 unsqueeze_46" [label="(4, 64, 64)", style=solid]; -"2075 view_85" -> "2078 add_53" [label="(1, 4, 12, 64, 64)", style=solid]; -"2076 unsqueeze_46" -> "2077 unsqueeze_47" [label="(4, 1, 64, 64)", style=solid]; -"2077 unsqueeze_47" -> "2078 add_53" [label="(1, 4, 1, 64, 64)", style=solid]; -"2078 add_53" -> "2079 view_86" [label="(1, 4, 12, 64, 64)", style=solid]; -"2079 view_86" -> "2080 softmax_15" [label="(4, 12, 64, 64)", style=solid]; -"2080 softmax_15" -> "2081 dropout_60" [label="(4, 12, 64, 64)", style=solid]; -"2081 dropout_60" -> "2082 matmul_31" [label="(4, 12, 64, 64)", style=solid]; -"2082 matmul_31" -> "2083 transpose_31" [label="(4, 12, 64, 32)", style=solid]; -"2083 transpose_31" -> "2084 reshape_70" [label="(4, 64, 12, 32)", style=solid]; -"2084 reshape_70" -> "2085 reshape_70_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"2085 reshape_70_0_0_nncf_smooth_quant_0" -> "2086 quantize_per_tensor_default_96" [label="(4, 64, 384)", style=solid]; -"2086 quantize_per_tensor_default_96" -> "2087 dequantize_per_tensor_default_96" [label="(4, 64, 384)", style=solid]; -"2087 dequantize_per_tensor_default_96" -> "2092 linear_95" [label="(4, 64, 384)", style=solid]; -"2088 linear_95_scale_0" -> "2091 dequantize_per_channel_default_96" [label="(384,)", style=solid]; -"2089 linear_95_zero_point_0" -> "2091 dequantize_per_channel_default_96" [label="(384,)", style=solid]; -"2090 _frozen_param96" -> "2091 dequantize_per_channel_default_96" [label="(384, 384)", style=solid]; -"2091 dequantize_per_channel_default_96" -> "2092 linear_95" [label="(384, 384)", style=solid]; -"2092 linear_95" -> "2093 dropout_61" [label="(4, 64, 384)", style=solid]; -"2093 dropout_61" -> "2094 view_87" [label="(4, 64, 384)", style=solid]; -"2094 view_87" -> "2095 permute_72" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2095 permute_72" -> "2096 reshape_71" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2096 reshape_71" -> "2097 roll_15" [label="(1, 16, 16, 384)", style=solid]; -"2097 roll_15" -> "2098 slice_241" [label="(1, 16, 16, 384)", style=solid]; -"2098 slice_241" -> "2099 slice_242" [label="(1, 16, 16, 384)", style=solid]; -"2099 slice_242" -> "2100 slice_243" [label="(1, 14, 16, 384)", style=solid]; -"2100 slice_243" -> "2101 slice_244" [label="(1, 14, 14, 384)", style=solid]; -"2101 slice_244" -> "2102 contiguous_29" [label="(1, 14, 14, 384)", style=solid]; -"2102 contiguous_29" -> "2103 layer_norm_33" [label="(1, 14, 14, 384)", style=solid]; -"2103 layer_norm_33" -> "2104 add_54" [label="(1, 14, 14, 384)", style=solid]; -"2104 add_54" -> "2105 add_54_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; -"2104 add_54" -> "2125 add_55" [label="(1, 14, 14, 384)", style=solid]; -"2105 add_54_0_0_nncf_smooth_quant_0" -> "2106 quantize_per_tensor_default_97" [label="(1, 14, 14, 384)", style=solid]; -"2106 quantize_per_tensor_default_97" -> "2107 dequantize_per_tensor_default_97" [label="(1, 14, 14, 384)", style=solid]; -"2107 dequantize_per_tensor_default_97" -> "2112 linear_96" [label="(1, 14, 14, 384)", style=solid]; -"2108 linear_96_scale_0" -> "2111 dequantize_per_channel_default_97" [label="(1536,)", style=solid]; -"2109 linear_96_zero_point_0" -> "2111 dequantize_per_channel_default_97" [label="(1536,)", style=solid]; -"2110 _frozen_param97" -> "2111 dequantize_per_channel_default_97" [label="(1536, 384)", style=solid]; -"2111 dequantize_per_channel_default_97" -> "2112 linear_96" [label="(1536, 384)", style=solid]; -"2112 linear_96" -> "2113 gelu_15" [label="(1, 14, 14, 1536)", style=solid]; -"2113 gelu_15" -> "2114 dropout_62" [label="(1, 14, 14, 1536)", style=solid]; -"2114 dropout_62" -> "2115 dropout_62_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; -"2115 dropout_62_0_0_nncf_smooth_quant_0" -> "2116 quantize_per_tensor_default_98" [label="(1, 14, 14, 1536)", style=solid]; -"2116 quantize_per_tensor_default_98" -> "2117 dequantize_per_tensor_default_98" [label="(1, 14, 14, 1536)", style=solid]; -"2117 dequantize_per_tensor_default_98" -> "2122 linear_97" [label="(1, 14, 14, 1536)", style=solid]; -"2118 linear_97_scale_0" -> "2121 dequantize_per_channel_default_98" [label="(384,)", style=solid]; -"2119 linear_97_zero_point_0" -> "2121 dequantize_per_channel_default_98" [label="(384,)", style=solid]; -"2120 _frozen_param98" -> "2121 dequantize_per_channel_default_98" [label="(384, 1536)", style=solid]; -"2121 dequantize_per_channel_default_98" -> "2122 linear_97" [label="(384, 1536)", style=solid]; -"2122 linear_97" -> "2123 dropout_63" [label="(1, 14, 14, 384)", style=solid]; -"2123 dropout_63" -> "2124 layer_norm_34" [label="(1, 14, 14, 384)", style=solid]; -"2124 layer_norm_34" -> "2125 add_55" [label="(1, 14, 14, 384)", style=solid]; -"2125 add_55" -> "2147 pad_18" [label="(1, 14, 14, 384)", style=solid]; -"2125 add_55" -> "2206 add_57" [label="(1, 14, 14, 384)", style=solid]; -"2126 features_5_12_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "2131 linear_98" [label="(1, 15, 15, 2)", style=solid]; -"2127 linear_98_scale_0" -> "2130 dequantize_per_channel_default_99" [label="(512,)", style=solid]; -"2128 linear_98_zero_point_0" -> "2130 dequantize_per_channel_default_99" [label="(512,)", style=solid]; -"2129 _frozen_param99" -> "2130 dequantize_per_channel_default_99" [label="(512, 2)", style=solid]; -"2130 dequantize_per_channel_default_99" -> "2131 linear_98" [label="(512, 2)", style=solid]; -"2131 linear_98" -> "2132 relu__16" [label="(1, 15, 15, 512)", style=solid]; -"2132 relu__16" -> "2133 relu__16_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; -"2133 relu__16_0_0_nncf_smooth_quant_0" -> "2138 linear_99" [label="(1, 15, 15, 512)", style=solid]; -"2134 linear_99_scale_0" -> "2137 dequantize_per_channel_default_100" [label="(12,)", style=solid]; -"2135 linear_99_zero_point_0" -> "2137 dequantize_per_channel_default_100" [label="(12,)", style=solid]; -"2136 _frozen_param100" -> "2137 dequantize_per_channel_default_100" [label="(12, 512)", style=solid]; -"2137 dequantize_per_channel_default_100" -> "2138 linear_99" [label="(12, 512)", style=solid]; -"2138 linear_99" -> "2139 view_88" [label="(1, 15, 15, 12)", style=solid]; -"2139 view_88" -> "2140 index_16" [label="(225, 12)", style=solid]; -"2140 index_16" -> "2141 view_89" [label="(4096, 12)", style=solid]; -"2141 view_89" -> "2142 permute_73" [label="(64, 64, 12)", style=solid]; -"2142 permute_73" -> "2143 contiguous_30" [label="(12, 64, 64)", style=solid]; -"2143 contiguous_30" -> "2144 unsqueeze_48" [label="(12, 64, 64)", style=solid]; -"2144 unsqueeze_48" -> "2145 sigmoid_16" [label="(1, 12, 64, 64)", style=solid]; -"2145 sigmoid_16" -> "2146 mul_32" [label="(1, 12, 64, 64)", style=solid]; -"2146 mul_32" -> "2182 add_56" [label="(1, 12, 64, 64)", style=solid]; -"2147 pad_18" -> "2148 view_90" [label="(1, 16, 16, 384)", style=solid]; -"2148 view_90" -> "2149 permute_74" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2149 permute_74" -> "2150 reshape_72" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2150 reshape_72" -> "2152 reshape_72_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"2151 clone_16" -> "2159 linear_100" [label="(1152,)", style=solid]; -"2152 reshape_72_0_0_nncf_smooth_quant_0" -> "2153 quantize_per_tensor_default_99" [label="(4, 64, 384)", style=solid]; -"2153 quantize_per_tensor_default_99" -> "2154 dequantize_per_tensor_default_99" [label="(4, 64, 384)", style=solid]; -"2154 dequantize_per_tensor_default_99" -> "2159 linear_100" [label="(4, 64, 384)", style=solid]; -"2155 linear_100_scale_0" -> "2158 dequantize_per_channel_default_101" [label="(1152,)", style=solid]; -"2156 linear_100_zero_point_0" -> "2158 dequantize_per_channel_default_101" [label="(1152,)", style=solid]; -"2157 _frozen_param101" -> "2158 dequantize_per_channel_default_101" [label="(1152, 384)", style=solid]; -"2158 dequantize_per_channel_default_101" -> "2159 linear_100" [label="(1152, 384)", style=solid]; -"2159 linear_100" -> "2160 reshape_73" [label="(4, 64, 1152)", style=solid]; -"2160 reshape_73" -> "2161 permute_75" [label="(4, 64, 3, 12, 32)", style=solid]; -"2161 permute_75" -> "2162 select_48" [label="(3, 4, 12, 64, 32)", style=solid]; -"2161 permute_75" -> "2163 select_49" [label="(3, 4, 12, 64, 32)", style=solid]; -"2161 permute_75" -> "2164 select_50" [label="(3, 4, 12, 64, 32)", style=solid]; -"2162 select_48" -> "2165 linalg_vector_norm_32" [label="(4, 12, 64, 32)", style=solid]; -"2162 select_48" -> "2167 expand_as_32" [label="(4, 12, 64, 32)", style=solid]; -"2162 select_48" -> "2168 div_32" [label="(4, 12, 64, 32)", style=solid]; -"2163 select_49" -> "2171 linalg_vector_norm_33" [label="(4, 12, 64, 32)", style=solid]; -"2163 select_49" -> "2173 expand_as_33" [label="(4, 12, 64, 32)", style=solid]; -"2163 select_49" -> "2174 div_33" [label="(4, 12, 64, 32)", style=solid]; -"2164 select_50" -> "2185 matmul_33" [label="(4, 12, 64, 32)", style=solid]; -"2165 linalg_vector_norm_32" -> "2166 clamp_min_32" [label="(4, 12, 64, 1)", style=solid]; -"2166 clamp_min_32" -> "2167 expand_as_32" [label="(4, 12, 64, 1)", style=solid]; -"2167 expand_as_32" -> "2168 div_32" [label="(4, 12, 64, 32)", style=solid]; -"2168 div_32" -> "2169 quantize_per_tensor_default_100" [label="(4, 12, 64, 32)", style=solid]; -"2169 quantize_per_tensor_default_100" -> "2170 dequantize_per_tensor_default_100" [label="(4, 12, 64, 32)", style=solid]; -"2170 dequantize_per_tensor_default_100" -> "2178 matmul_32" [label="(4, 12, 64, 32)", style=solid]; -"2171 linalg_vector_norm_33" -> "2172 clamp_min_33" [label="(4, 12, 64, 1)", style=solid]; -"2172 clamp_min_33" -> "2173 expand_as_33" [label="(4, 12, 64, 1)", style=solid]; -"2173 expand_as_33" -> "2174 div_33" [label="(4, 12, 64, 32)", style=solid]; -"2174 div_33" -> "2175 quantize_per_tensor_default_101" [label="(4, 12, 64, 32)", style=solid]; -"2175 quantize_per_tensor_default_101" -> "2176 dequantize_per_tensor_default_101" [label="(4, 12, 64, 32)", style=solid]; -"2176 dequantize_per_tensor_default_101" -> "2177 transpose_32" [label="(4, 12, 64, 32)", style=solid]; -"2177 transpose_32" -> "2178 matmul_32" [label="(4, 12, 32, 64)", style=solid]; -"2178 matmul_32" -> "2181 mul_33" [label="(4, 12, 64, 64)", style=solid]; -"2179 clamp_16" -> "2180 exp_16" [label="(12, 1, 1)", style=solid]; -"2180 exp_16" -> "2181 mul_33" [label="(12, 1, 1)", style=solid]; -"2181 mul_33" -> "2182 add_56" [label="(4, 12, 64, 64)", style=solid]; -"2182 add_56" -> "2183 softmax_16" [label="(4, 12, 64, 64)", style=solid]; -"2183 softmax_16" -> "2184 dropout_64" [label="(4, 12, 64, 64)", style=solid]; -"2184 dropout_64" -> "2185 matmul_33" [label="(4, 12, 64, 64)", style=solid]; -"2185 matmul_33" -> "2186 transpose_33" [label="(4, 12, 64, 32)", style=solid]; -"2186 transpose_33" -> "2187 reshape_74" [label="(4, 64, 12, 32)", style=solid]; -"2187 reshape_74" -> "2188 reshape_74_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"2188 reshape_74_0_0_nncf_smooth_quant_0" -> "2189 quantize_per_tensor_default_102" [label="(4, 64, 384)", style=solid]; -"2189 quantize_per_tensor_default_102" -> "2190 dequantize_per_tensor_default_102" [label="(4, 64, 384)", style=solid]; -"2190 dequantize_per_tensor_default_102" -> "2195 linear_101" [label="(4, 64, 384)", style=solid]; -"2191 linear_101_scale_0" -> "2194 dequantize_per_channel_default_102" [label="(384,)", style=solid]; -"2192 linear_101_zero_point_0" -> "2194 dequantize_per_channel_default_102" [label="(384,)", style=solid]; -"2193 _frozen_param102" -> "2194 dequantize_per_channel_default_102" [label="(384, 384)", style=solid]; -"2194 dequantize_per_channel_default_102" -> "2195 linear_101" [label="(384, 384)", style=solid]; -"2195 linear_101" -> "2196 dropout_65" [label="(4, 64, 384)", style=solid]; -"2196 dropout_65" -> "2197 view_91" [label="(4, 64, 384)", style=solid]; -"2197 view_91" -> "2198 permute_76" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2198 permute_76" -> "2199 reshape_75" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2199 reshape_75" -> "2200 slice_246" [label="(1, 16, 16, 384)", style=solid]; -"2200 slice_246" -> "2201 slice_247" [label="(1, 16, 16, 384)", style=solid]; -"2201 slice_247" -> "2202 slice_248" [label="(1, 14, 16, 384)", style=solid]; -"2202 slice_248" -> "2203 slice_249" [label="(1, 14, 14, 384)", style=solid]; -"2203 slice_249" -> "2204 contiguous_31" [label="(1, 14, 14, 384)", style=solid]; -"2204 contiguous_31" -> "2205 layer_norm_35" [label="(1, 14, 14, 384)", style=solid]; -"2205 layer_norm_35" -> "2206 add_57" [label="(1, 14, 14, 384)", style=solid]; -"2206 add_57" -> "2207 add_57_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; -"2206 add_57" -> "2227 add_58" [label="(1, 14, 14, 384)", style=solid]; -"2207 add_57_0_0_nncf_smooth_quant_0" -> "2208 quantize_per_tensor_default_103" [label="(1, 14, 14, 384)", style=solid]; -"2208 quantize_per_tensor_default_103" -> "2209 dequantize_per_tensor_default_103" [label="(1, 14, 14, 384)", style=solid]; -"2209 dequantize_per_tensor_default_103" -> "2214 linear_102" [label="(1, 14, 14, 384)", style=solid]; -"2210 linear_102_scale_0" -> "2213 dequantize_per_channel_default_103" [label="(1536,)", style=solid]; -"2211 linear_102_zero_point_0" -> "2213 dequantize_per_channel_default_103" [label="(1536,)", style=solid]; -"2212 _frozen_param103" -> "2213 dequantize_per_channel_default_103" [label="(1536, 384)", style=solid]; -"2213 dequantize_per_channel_default_103" -> "2214 linear_102" [label="(1536, 384)", style=solid]; -"2214 linear_102" -> "2215 gelu_16" [label="(1, 14, 14, 1536)", style=solid]; -"2215 gelu_16" -> "2216 dropout_66" [label="(1, 14, 14, 1536)", style=solid]; -"2216 dropout_66" -> "2217 dropout_66_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; -"2217 dropout_66_0_0_nncf_smooth_quant_0" -> "2218 quantize_per_tensor_default_104" [label="(1, 14, 14, 1536)", style=solid]; -"2218 quantize_per_tensor_default_104" -> "2219 dequantize_per_tensor_default_104" [label="(1, 14, 14, 1536)", style=solid]; -"2219 dequantize_per_tensor_default_104" -> "2224 linear_103" [label="(1, 14, 14, 1536)", style=solid]; -"2220 linear_103_scale_0" -> "2223 dequantize_per_channel_default_104" [label="(384,)", style=solid]; -"2221 linear_103_zero_point_0" -> "2223 dequantize_per_channel_default_104" [label="(384,)", style=solid]; -"2222 _frozen_param104" -> "2223 dequantize_per_channel_default_104" [label="(384, 1536)", style=solid]; -"2223 dequantize_per_channel_default_104" -> "2224 linear_103" [label="(384, 1536)", style=solid]; -"2224 linear_103" -> "2225 dropout_67" [label="(1, 14, 14, 384)", style=solid]; -"2225 dropout_67" -> "2226 layer_norm_36" [label="(1, 14, 14, 384)", style=solid]; -"2226 layer_norm_36" -> "2227 add_58" [label="(1, 14, 14, 384)", style=solid]; -"2227 add_58" -> "2249 pad_19" [label="(1, 14, 14, 384)", style=solid]; -"2227 add_58" -> "2326 add_61" [label="(1, 14, 14, 384)", style=solid]; -"2228 features_5_13_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "2233 linear_104" [label="(1, 15, 15, 2)", style=solid]; -"2229 linear_104_scale_0" -> "2232 dequantize_per_channel_default_105" [label="(512,)", style=solid]; -"2230 linear_104_zero_point_0" -> "2232 dequantize_per_channel_default_105" [label="(512,)", style=solid]; -"2231 _frozen_param105" -> "2232 dequantize_per_channel_default_105" [label="(512, 2)", style=solid]; -"2232 dequantize_per_channel_default_105" -> "2233 linear_104" [label="(512, 2)", style=solid]; -"2233 linear_104" -> "2234 relu__17" [label="(1, 15, 15, 512)", style=solid]; -"2234 relu__17" -> "2235 relu__17_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; -"2235 relu__17_0_0_nncf_smooth_quant_0" -> "2240 linear_105" [label="(1, 15, 15, 512)", style=solid]; -"2236 linear_105_scale_0" -> "2239 dequantize_per_channel_default_106" [label="(12,)", style=solid]; -"2237 linear_105_zero_point_0" -> "2239 dequantize_per_channel_default_106" [label="(12,)", style=solid]; -"2238 _frozen_param106" -> "2239 dequantize_per_channel_default_106" [label="(12, 512)", style=solid]; -"2239 dequantize_per_channel_default_106" -> "2240 linear_105" [label="(12, 512)", style=solid]; -"2240 linear_105" -> "2241 view_92" [label="(1, 15, 15, 12)", style=solid]; -"2241 view_92" -> "2242 index_17" [label="(225, 12)", style=solid]; -"2242 index_17" -> "2243 view_93" [label="(4096, 12)", style=solid]; -"2243 view_93" -> "2244 permute_77" [label="(64, 64, 12)", style=solid]; -"2244 permute_77" -> "2245 contiguous_32" [label="(12, 64, 64)", style=solid]; -"2245 contiguous_32" -> "2246 unsqueeze_49" [label="(12, 64, 64)", style=solid]; -"2246 unsqueeze_49" -> "2247 sigmoid_17" [label="(1, 12, 64, 64)", style=solid]; -"2247 sigmoid_17" -> "2248 mul_34" [label="(1, 12, 64, 64)", style=solid]; -"2248 mul_34" -> "2285 add_59" [label="(1, 12, 64, 64)", style=solid]; -"2249 pad_19" -> "2250 roll_16" [label="(1, 16, 16, 384)", style=solid]; -"2250 roll_16" -> "2251 view_94" [label="(1, 16, 16, 384)", style=solid]; -"2251 view_94" -> "2252 permute_78" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2252 permute_78" -> "2253 reshape_76" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2253 reshape_76" -> "2255 reshape_76_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"2253 reshape_76" -> "2286 new_zeros_8" [label="(4, 64, 384)", style=solid]; -"2254 clone_17" -> "2262 linear_106" [label="(1152,)", style=solid]; -"2255 reshape_76_0_0_nncf_smooth_quant_0" -> "2256 quantize_per_tensor_default_105" [label="(4, 64, 384)", style=solid]; -"2256 quantize_per_tensor_default_105" -> "2257 dequantize_per_tensor_default_105" [label="(4, 64, 384)", style=solid]; -"2257 dequantize_per_tensor_default_105" -> "2262 linear_106" [label="(4, 64, 384)", style=solid]; -"2258 linear_106_scale_0" -> "2261 dequantize_per_channel_default_107" [label="(1152,)", style=solid]; -"2259 linear_106_zero_point_0" -> "2261 dequantize_per_channel_default_107" [label="(1152,)", style=solid]; -"2260 _frozen_param107" -> "2261 dequantize_per_channel_default_107" [label="(1152, 384)", style=solid]; -"2261 dequantize_per_channel_default_107" -> "2262 linear_106" [label="(1152, 384)", style=solid]; -"2262 linear_106" -> "2263 reshape_77" [label="(4, 64, 1152)", style=solid]; -"2263 reshape_77" -> "2264 permute_79" [label="(4, 64, 3, 12, 32)", style=solid]; -"2264 permute_79" -> "2265 select_51" [label="(3, 4, 12, 64, 32)", style=solid]; -"2264 permute_79" -> "2266 select_52" [label="(3, 4, 12, 64, 32)", style=solid]; -"2264 permute_79" -> "2267 select_53" [label="(3, 4, 12, 64, 32)", style=solid]; -"2265 select_51" -> "2268 linalg_vector_norm_34" [label="(4, 12, 64, 32)", style=solid]; -"2265 select_51" -> "2270 expand_as_34" [label="(4, 12, 64, 32)", style=solid]; -"2265 select_51" -> "2271 div_34" [label="(4, 12, 64, 32)", style=solid]; -"2266 select_52" -> "2274 linalg_vector_norm_35" [label="(4, 12, 64, 32)", style=solid]; -"2266 select_52" -> "2276 expand_as_35" [label="(4, 12, 64, 32)", style=solid]; -"2266 select_52" -> "2277 div_35" [label="(4, 12, 64, 32)", style=solid]; -"2267 select_53" -> "2304 matmul_35" [label="(4, 12, 64, 32)", style=solid]; -"2268 linalg_vector_norm_34" -> "2269 clamp_min_34" [label="(4, 12, 64, 1)", style=solid]; -"2269 clamp_min_34" -> "2270 expand_as_34" [label="(4, 12, 64, 1)", style=solid]; -"2270 expand_as_34" -> "2271 div_34" [label="(4, 12, 64, 32)", style=solid]; -"2271 div_34" -> "2272 quantize_per_tensor_default_106" [label="(4, 12, 64, 32)", style=solid]; -"2272 quantize_per_tensor_default_106" -> "2273 dequantize_per_tensor_default_106" [label="(4, 12, 64, 32)", style=solid]; -"2273 dequantize_per_tensor_default_106" -> "2281 matmul_34" [label="(4, 12, 64, 32)", style=solid]; -"2274 linalg_vector_norm_35" -> "2275 clamp_min_35" [label="(4, 12, 64, 1)", style=solid]; -"2275 clamp_min_35" -> "2276 expand_as_35" [label="(4, 12, 64, 1)", style=solid]; -"2276 expand_as_35" -> "2277 div_35" [label="(4, 12, 64, 32)", style=solid]; -"2277 div_35" -> "2278 quantize_per_tensor_default_107" [label="(4, 12, 64, 32)", style=solid]; -"2278 quantize_per_tensor_default_107" -> "2279 dequantize_per_tensor_default_107" [label="(4, 12, 64, 32)", style=solid]; -"2279 dequantize_per_tensor_default_107" -> "2280 transpose_34" [label="(4, 12, 64, 32)", style=solid]; -"2280 transpose_34" -> "2281 matmul_34" [label="(4, 12, 32, 64)", style=solid]; -"2281 matmul_34" -> "2284 mul_35" [label="(4, 12, 64, 64)", style=solid]; -"2282 clamp_17" -> "2283 exp_17" [label="(12, 1, 1)", style=solid]; -"2283 exp_17" -> "2284 mul_35" [label="(12, 1, 1)", style=solid]; -"2284 mul_35" -> "2285 add_59" [label="(4, 12, 64, 64)", style=solid]; -"2285 add_59" -> "2297 view_96" [label="(4, 12, 64, 64)", style=solid]; -"2286 new_zeros_8" -> "2287 view_95" [label="(16, 16)", style=solid]; -"2287 view_95" -> "2288 permute_80" [label="(2, 8, 2, 8)", style=solid]; -"2288 permute_80" -> "2289 reshape_78" [label="(2, 2, 8, 8)", style=solid]; -"2289 reshape_78" -> "2290 unsqueeze_50" [label="(4, 64)", style=solid]; -"2289 reshape_78" -> "2291 unsqueeze_51" [label="(4, 64)", style=solid]; -"2290 unsqueeze_50" -> "2292 sub_8" [label="(4, 1, 64)", style=solid]; -"2291 unsqueeze_51" -> "2292 sub_8" [label="(4, 64, 1)", style=solid]; -"2292 sub_8" -> "2293 ne_8" [label="(4, 64, 64)", style=solid]; -"2292 sub_8" -> "2294 masked_fill_16" [label="(4, 64, 64)", style=solid]; -"2292 sub_8" -> "2295 eq_8" [label="(4, 64, 64)", style=solid]; -"2293 ne_8" -> "2294 masked_fill_16" [label="(4, 64, 64)", style=solid]; -"2294 masked_fill_16" -> "2296 masked_fill_17" [label="(4, 64, 64)", style=solid]; -"2295 eq_8" -> "2296 masked_fill_17" [label="(4, 64, 64)", style=solid]; -"2296 masked_fill_17" -> "2298 unsqueeze_52" [label="(4, 64, 64)", style=solid]; -"2297 view_96" -> "2300 add_60" [label="(1, 4, 12, 64, 64)", style=solid]; -"2298 unsqueeze_52" -> "2299 unsqueeze_53" [label="(4, 1, 64, 64)", style=solid]; -"2299 unsqueeze_53" -> "2300 add_60" [label="(1, 4, 1, 64, 64)", style=solid]; -"2300 add_60" -> "2301 view_97" [label="(1, 4, 12, 64, 64)", style=solid]; -"2301 view_97" -> "2302 softmax_17" [label="(4, 12, 64, 64)", style=solid]; -"2302 softmax_17" -> "2303 dropout_68" [label="(4, 12, 64, 64)", style=solid]; -"2303 dropout_68" -> "2304 matmul_35" [label="(4, 12, 64, 64)", style=solid]; -"2304 matmul_35" -> "2305 transpose_35" [label="(4, 12, 64, 32)", style=solid]; -"2305 transpose_35" -> "2306 reshape_79" [label="(4, 64, 12, 32)", style=solid]; -"2306 reshape_79" -> "2307 reshape_79_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"2307 reshape_79_0_0_nncf_smooth_quant_0" -> "2308 quantize_per_tensor_default_108" [label="(4, 64, 384)", style=solid]; -"2308 quantize_per_tensor_default_108" -> "2309 dequantize_per_tensor_default_108" [label="(4, 64, 384)", style=solid]; -"2309 dequantize_per_tensor_default_108" -> "2314 linear_107" [label="(4, 64, 384)", style=solid]; -"2310 linear_107_scale_0" -> "2313 dequantize_per_channel_default_108" [label="(384,)", style=solid]; -"2311 linear_107_zero_point_0" -> "2313 dequantize_per_channel_default_108" [label="(384,)", style=solid]; -"2312 _frozen_param108" -> "2313 dequantize_per_channel_default_108" [label="(384, 384)", style=solid]; -"2313 dequantize_per_channel_default_108" -> "2314 linear_107" [label="(384, 384)", style=solid]; -"2314 linear_107" -> "2315 dropout_69" [label="(4, 64, 384)", style=solid]; -"2315 dropout_69" -> "2316 view_98" [label="(4, 64, 384)", style=solid]; -"2316 view_98" -> "2317 permute_81" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2317 permute_81" -> "2318 reshape_80" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2318 reshape_80" -> "2319 roll_17" [label="(1, 16, 16, 384)", style=solid]; -"2319 roll_17" -> "2320 slice_269" [label="(1, 16, 16, 384)", style=solid]; -"2320 slice_269" -> "2321 slice_270" [label="(1, 16, 16, 384)", style=solid]; -"2321 slice_270" -> "2322 slice_271" [label="(1, 14, 16, 384)", style=solid]; -"2322 slice_271" -> "2323 slice_272" [label="(1, 14, 14, 384)", style=solid]; -"2323 slice_272" -> "2324 contiguous_33" [label="(1, 14, 14, 384)", style=solid]; -"2324 contiguous_33" -> "2325 layer_norm_37" [label="(1, 14, 14, 384)", style=solid]; -"2325 layer_norm_37" -> "2326 add_61" [label="(1, 14, 14, 384)", style=solid]; -"2326 add_61" -> "2327 add_61_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; -"2326 add_61" -> "2347 add_62" [label="(1, 14, 14, 384)", style=solid]; -"2327 add_61_0_0_nncf_smooth_quant_0" -> "2328 quantize_per_tensor_default_109" [label="(1, 14, 14, 384)", style=solid]; -"2328 quantize_per_tensor_default_109" -> "2329 dequantize_per_tensor_default_109" [label="(1, 14, 14, 384)", style=solid]; -"2329 dequantize_per_tensor_default_109" -> "2334 linear_108" [label="(1, 14, 14, 384)", style=solid]; -"2330 linear_108_scale_0" -> "2333 dequantize_per_channel_default_109" [label="(1536,)", style=solid]; -"2331 linear_108_zero_point_0" -> "2333 dequantize_per_channel_default_109" [label="(1536,)", style=solid]; -"2332 _frozen_param109" -> "2333 dequantize_per_channel_default_109" [label="(1536, 384)", style=solid]; -"2333 dequantize_per_channel_default_109" -> "2334 linear_108" [label="(1536, 384)", style=solid]; -"2334 linear_108" -> "2335 gelu_17" [label="(1, 14, 14, 1536)", style=solid]; -"2335 gelu_17" -> "2336 dropout_70" [label="(1, 14, 14, 1536)", style=solid]; -"2336 dropout_70" -> "2337 dropout_70_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; -"2337 dropout_70_0_0_nncf_smooth_quant_0" -> "2338 quantize_per_tensor_default_110" [label="(1, 14, 14, 1536)", style=solid]; -"2338 quantize_per_tensor_default_110" -> "2339 dequantize_per_tensor_default_110" [label="(1, 14, 14, 1536)", style=solid]; -"2339 dequantize_per_tensor_default_110" -> "2344 linear_109" [label="(1, 14, 14, 1536)", style=solid]; -"2340 linear_109_scale_0" -> "2343 dequantize_per_channel_default_110" [label="(384,)", style=solid]; -"2341 linear_109_zero_point_0" -> "2343 dequantize_per_channel_default_110" [label="(384,)", style=solid]; -"2342 _frozen_param110" -> "2343 dequantize_per_channel_default_110" [label="(384, 1536)", style=solid]; -"2343 dequantize_per_channel_default_110" -> "2344 linear_109" [label="(384, 1536)", style=solid]; -"2344 linear_109" -> "2345 dropout_71" [label="(1, 14, 14, 384)", style=solid]; -"2345 dropout_71" -> "2346 layer_norm_38" [label="(1, 14, 14, 384)", style=solid]; -"2346 layer_norm_38" -> "2347 add_62" [label="(1, 14, 14, 384)", style=solid]; -"2347 add_62" -> "2369 pad_20" [label="(1, 14, 14, 384)", style=solid]; -"2347 add_62" -> "2428 add_64" [label="(1, 14, 14, 384)", style=solid]; -"2348 features_5_14_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "2353 linear_110" [label="(1, 15, 15, 2)", style=solid]; -"2349 linear_110_scale_0" -> "2352 dequantize_per_channel_default_111" [label="(512,)", style=solid]; -"2350 linear_110_zero_point_0" -> "2352 dequantize_per_channel_default_111" [label="(512,)", style=solid]; -"2351 _frozen_param111" -> "2352 dequantize_per_channel_default_111" [label="(512, 2)", style=solid]; -"2352 dequantize_per_channel_default_111" -> "2353 linear_110" [label="(512, 2)", style=solid]; -"2353 linear_110" -> "2354 relu__18" [label="(1, 15, 15, 512)", style=solid]; -"2354 relu__18" -> "2355 relu__18_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; -"2355 relu__18_0_0_nncf_smooth_quant_0" -> "2360 linear_111" [label="(1, 15, 15, 512)", style=solid]; -"2356 linear_111_scale_0" -> "2359 dequantize_per_channel_default_112" [label="(12,)", style=solid]; -"2357 linear_111_zero_point_0" -> "2359 dequantize_per_channel_default_112" [label="(12,)", style=solid]; -"2358 _frozen_param112" -> "2359 dequantize_per_channel_default_112" [label="(12, 512)", style=solid]; -"2359 dequantize_per_channel_default_112" -> "2360 linear_111" [label="(12, 512)", style=solid]; -"2360 linear_111" -> "2361 view_99" [label="(1, 15, 15, 12)", style=solid]; -"2361 view_99" -> "2362 index_18" [label="(225, 12)", style=solid]; -"2362 index_18" -> "2363 view_100" [label="(4096, 12)", style=solid]; -"2363 view_100" -> "2364 permute_82" [label="(64, 64, 12)", style=solid]; -"2364 permute_82" -> "2365 contiguous_34" [label="(12, 64, 64)", style=solid]; -"2365 contiguous_34" -> "2366 unsqueeze_54" [label="(12, 64, 64)", style=solid]; -"2366 unsqueeze_54" -> "2367 sigmoid_18" [label="(1, 12, 64, 64)", style=solid]; -"2367 sigmoid_18" -> "2368 mul_36" [label="(1, 12, 64, 64)", style=solid]; -"2368 mul_36" -> "2404 add_63" [label="(1, 12, 64, 64)", style=solid]; -"2369 pad_20" -> "2370 view_101" [label="(1, 16, 16, 384)", style=solid]; -"2370 view_101" -> "2371 permute_83" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2371 permute_83" -> "2372 reshape_81" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2372 reshape_81" -> "2374 reshape_81_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"2373 clone_18" -> "2381 linear_112" [label="(1152,)", style=solid]; -"2374 reshape_81_0_0_nncf_smooth_quant_0" -> "2375 quantize_per_tensor_default_111" [label="(4, 64, 384)", style=solid]; -"2375 quantize_per_tensor_default_111" -> "2376 dequantize_per_tensor_default_111" [label="(4, 64, 384)", style=solid]; -"2376 dequantize_per_tensor_default_111" -> "2381 linear_112" [label="(4, 64, 384)", style=solid]; -"2377 linear_112_scale_0" -> "2380 dequantize_per_channel_default_113" [label="(1152,)", style=solid]; -"2378 linear_112_zero_point_0" -> "2380 dequantize_per_channel_default_113" [label="(1152,)", style=solid]; -"2379 _frozen_param113" -> "2380 dequantize_per_channel_default_113" [label="(1152, 384)", style=solid]; -"2380 dequantize_per_channel_default_113" -> "2381 linear_112" [label="(1152, 384)", style=solid]; -"2381 linear_112" -> "2382 reshape_82" [label="(4, 64, 1152)", style=solid]; -"2382 reshape_82" -> "2383 permute_84" [label="(4, 64, 3, 12, 32)", style=solid]; -"2383 permute_84" -> "2384 select_54" [label="(3, 4, 12, 64, 32)", style=solid]; -"2383 permute_84" -> "2385 select_55" [label="(3, 4, 12, 64, 32)", style=solid]; -"2383 permute_84" -> "2386 select_56" [label="(3, 4, 12, 64, 32)", style=solid]; -"2384 select_54" -> "2387 linalg_vector_norm_36" [label="(4, 12, 64, 32)", style=solid]; -"2384 select_54" -> "2389 expand_as_36" [label="(4, 12, 64, 32)", style=solid]; -"2384 select_54" -> "2390 div_36" [label="(4, 12, 64, 32)", style=solid]; -"2385 select_55" -> "2393 linalg_vector_norm_37" [label="(4, 12, 64, 32)", style=solid]; -"2385 select_55" -> "2395 expand_as_37" [label="(4, 12, 64, 32)", style=solid]; -"2385 select_55" -> "2396 div_37" [label="(4, 12, 64, 32)", style=solid]; -"2386 select_56" -> "2407 matmul_37" [label="(4, 12, 64, 32)", style=solid]; -"2387 linalg_vector_norm_36" -> "2388 clamp_min_36" [label="(4, 12, 64, 1)", style=solid]; -"2388 clamp_min_36" -> "2389 expand_as_36" [label="(4, 12, 64, 1)", style=solid]; -"2389 expand_as_36" -> "2390 div_36" [label="(4, 12, 64, 32)", style=solid]; -"2390 div_36" -> "2391 quantize_per_tensor_default_112" [label="(4, 12, 64, 32)", style=solid]; -"2391 quantize_per_tensor_default_112" -> "2392 dequantize_per_tensor_default_112" [label="(4, 12, 64, 32)", style=solid]; -"2392 dequantize_per_tensor_default_112" -> "2400 matmul_36" [label="(4, 12, 64, 32)", style=solid]; -"2393 linalg_vector_norm_37" -> "2394 clamp_min_37" [label="(4, 12, 64, 1)", style=solid]; -"2394 clamp_min_37" -> "2395 expand_as_37" [label="(4, 12, 64, 1)", style=solid]; -"2395 expand_as_37" -> "2396 div_37" [label="(4, 12, 64, 32)", style=solid]; -"2396 div_37" -> "2397 quantize_per_tensor_default_113" [label="(4, 12, 64, 32)", style=solid]; -"2397 quantize_per_tensor_default_113" -> "2398 dequantize_per_tensor_default_113" [label="(4, 12, 64, 32)", style=solid]; -"2398 dequantize_per_tensor_default_113" -> "2399 transpose_36" [label="(4, 12, 64, 32)", style=solid]; -"2399 transpose_36" -> "2400 matmul_36" [label="(4, 12, 32, 64)", style=solid]; -"2400 matmul_36" -> "2403 mul_37" [label="(4, 12, 64, 64)", style=solid]; -"2401 clamp_18" -> "2402 exp_18" [label="(12, 1, 1)", style=solid]; -"2402 exp_18" -> "2403 mul_37" [label="(12, 1, 1)", style=solid]; -"2403 mul_37" -> "2404 add_63" [label="(4, 12, 64, 64)", style=solid]; -"2404 add_63" -> "2405 softmax_18" [label="(4, 12, 64, 64)", style=solid]; -"2405 softmax_18" -> "2406 dropout_72" [label="(4, 12, 64, 64)", style=solid]; -"2406 dropout_72" -> "2407 matmul_37" [label="(4, 12, 64, 64)", style=solid]; -"2407 matmul_37" -> "2408 transpose_37" [label="(4, 12, 64, 32)", style=solid]; -"2408 transpose_37" -> "2409 reshape_83" [label="(4, 64, 12, 32)", style=solid]; -"2409 reshape_83" -> "2410 reshape_83_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"2410 reshape_83_0_0_nncf_smooth_quant_0" -> "2411 quantize_per_tensor_default_114" [label="(4, 64, 384)", style=solid]; -"2411 quantize_per_tensor_default_114" -> "2412 dequantize_per_tensor_default_114" [label="(4, 64, 384)", style=solid]; -"2412 dequantize_per_tensor_default_114" -> "2417 linear_113" [label="(4, 64, 384)", style=solid]; -"2413 linear_113_scale_0" -> "2416 dequantize_per_channel_default_114" [label="(384,)", style=solid]; -"2414 linear_113_zero_point_0" -> "2416 dequantize_per_channel_default_114" [label="(384,)", style=solid]; -"2415 _frozen_param114" -> "2416 dequantize_per_channel_default_114" [label="(384, 384)", style=solid]; -"2416 dequantize_per_channel_default_114" -> "2417 linear_113" [label="(384, 384)", style=solid]; -"2417 linear_113" -> "2418 dropout_73" [label="(4, 64, 384)", style=solid]; -"2418 dropout_73" -> "2419 view_102" [label="(4, 64, 384)", style=solid]; -"2419 view_102" -> "2420 permute_85" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2420 permute_85" -> "2421 reshape_84" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2421 reshape_84" -> "2422 slice_274" [label="(1, 16, 16, 384)", style=solid]; -"2422 slice_274" -> "2423 slice_275" [label="(1, 16, 16, 384)", style=solid]; -"2423 slice_275" -> "2424 slice_276" [label="(1, 14, 16, 384)", style=solid]; -"2424 slice_276" -> "2425 slice_277" [label="(1, 14, 14, 384)", style=solid]; -"2425 slice_277" -> "2426 contiguous_35" [label="(1, 14, 14, 384)", style=solid]; -"2426 contiguous_35" -> "2427 layer_norm_39" [label="(1, 14, 14, 384)", style=solid]; -"2427 layer_norm_39" -> "2428 add_64" [label="(1, 14, 14, 384)", style=solid]; -"2428 add_64" -> "2429 add_64_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; -"2428 add_64" -> "2449 add_65" [label="(1, 14, 14, 384)", style=solid]; -"2429 add_64_0_0_nncf_smooth_quant_0" -> "2430 quantize_per_tensor_default_115" [label="(1, 14, 14, 384)", style=solid]; -"2430 quantize_per_tensor_default_115" -> "2431 dequantize_per_tensor_default_115" [label="(1, 14, 14, 384)", style=solid]; -"2431 dequantize_per_tensor_default_115" -> "2436 linear_114" [label="(1, 14, 14, 384)", style=solid]; -"2432 linear_114_scale_0" -> "2435 dequantize_per_channel_default_115" [label="(1536,)", style=solid]; -"2433 linear_114_zero_point_0" -> "2435 dequantize_per_channel_default_115" [label="(1536,)", style=solid]; -"2434 _frozen_param115" -> "2435 dequantize_per_channel_default_115" [label="(1536, 384)", style=solid]; -"2435 dequantize_per_channel_default_115" -> "2436 linear_114" [label="(1536, 384)", style=solid]; -"2436 linear_114" -> "2437 gelu_18" [label="(1, 14, 14, 1536)", style=solid]; -"2437 gelu_18" -> "2438 dropout_74" [label="(1, 14, 14, 1536)", style=solid]; -"2438 dropout_74" -> "2439 dropout_74_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; -"2439 dropout_74_0_0_nncf_smooth_quant_0" -> "2440 quantize_per_tensor_default_116" [label="(1, 14, 14, 1536)", style=solid]; -"2440 quantize_per_tensor_default_116" -> "2441 dequantize_per_tensor_default_116" [label="(1, 14, 14, 1536)", style=solid]; -"2441 dequantize_per_tensor_default_116" -> "2446 linear_115" [label="(1, 14, 14, 1536)", style=solid]; -"2442 linear_115_scale_0" -> "2445 dequantize_per_channel_default_116" [label="(384,)", style=solid]; -"2443 linear_115_zero_point_0" -> "2445 dequantize_per_channel_default_116" [label="(384,)", style=solid]; -"2444 _frozen_param116" -> "2445 dequantize_per_channel_default_116" [label="(384, 1536)", style=solid]; -"2445 dequantize_per_channel_default_116" -> "2446 linear_115" [label="(384, 1536)", style=solid]; -"2446 linear_115" -> "2447 dropout_75" [label="(1, 14, 14, 384)", style=solid]; -"2447 dropout_75" -> "2448 layer_norm_40" [label="(1, 14, 14, 384)", style=solid]; -"2448 layer_norm_40" -> "2449 add_65" [label="(1, 14, 14, 384)", style=solid]; -"2449 add_65" -> "2471 pad_21" [label="(1, 14, 14, 384)", style=solid]; -"2449 add_65" -> "2548 add_68" [label="(1, 14, 14, 384)", style=solid]; -"2450 features_5_15_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "2455 linear_116" [label="(1, 15, 15, 2)", style=solid]; -"2451 linear_116_scale_0" -> "2454 dequantize_per_channel_default_117" [label="(512,)", style=solid]; -"2452 linear_116_zero_point_0" -> "2454 dequantize_per_channel_default_117" [label="(512,)", style=solid]; -"2453 _frozen_param117" -> "2454 dequantize_per_channel_default_117" [label="(512, 2)", style=solid]; -"2454 dequantize_per_channel_default_117" -> "2455 linear_116" [label="(512, 2)", style=solid]; -"2455 linear_116" -> "2456 relu__19" [label="(1, 15, 15, 512)", style=solid]; -"2456 relu__19" -> "2457 relu__19_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; -"2457 relu__19_0_0_nncf_smooth_quant_0" -> "2462 linear_117" [label="(1, 15, 15, 512)", style=solid]; -"2458 linear_117_scale_0" -> "2461 dequantize_per_channel_default_118" [label="(12,)", style=solid]; -"2459 linear_117_zero_point_0" -> "2461 dequantize_per_channel_default_118" [label="(12,)", style=solid]; -"2460 _frozen_param118" -> "2461 dequantize_per_channel_default_118" [label="(12, 512)", style=solid]; -"2461 dequantize_per_channel_default_118" -> "2462 linear_117" [label="(12, 512)", style=solid]; -"2462 linear_117" -> "2463 view_103" [label="(1, 15, 15, 12)", style=solid]; -"2463 view_103" -> "2464 index_19" [label="(225, 12)", style=solid]; -"2464 index_19" -> "2465 view_104" [label="(4096, 12)", style=solid]; -"2465 view_104" -> "2466 permute_86" [label="(64, 64, 12)", style=solid]; -"2466 permute_86" -> "2467 contiguous_36" [label="(12, 64, 64)", style=solid]; -"2467 contiguous_36" -> "2468 unsqueeze_55" [label="(12, 64, 64)", style=solid]; -"2468 unsqueeze_55" -> "2469 sigmoid_19" [label="(1, 12, 64, 64)", style=solid]; -"2469 sigmoid_19" -> "2470 mul_38" [label="(1, 12, 64, 64)", style=solid]; -"2470 mul_38" -> "2507 add_66" [label="(1, 12, 64, 64)", style=solid]; -"2471 pad_21" -> "2472 roll_18" [label="(1, 16, 16, 384)", style=solid]; -"2472 roll_18" -> "2473 view_105" [label="(1, 16, 16, 384)", style=solid]; -"2473 view_105" -> "2474 permute_87" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2474 permute_87" -> "2475 reshape_85" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2475 reshape_85" -> "2477 reshape_85_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"2475 reshape_85" -> "2508 new_zeros_9" [label="(4, 64, 384)", style=solid]; -"2476 clone_19" -> "2484 linear_118" [label="(1152,)", style=solid]; -"2477 reshape_85_0_0_nncf_smooth_quant_0" -> "2478 quantize_per_tensor_default_117" [label="(4, 64, 384)", style=solid]; -"2478 quantize_per_tensor_default_117" -> "2479 dequantize_per_tensor_default_117" [label="(4, 64, 384)", style=solid]; -"2479 dequantize_per_tensor_default_117" -> "2484 linear_118" [label="(4, 64, 384)", style=solid]; -"2480 linear_118_scale_0" -> "2483 dequantize_per_channel_default_119" [label="(1152,)", style=solid]; -"2481 linear_118_zero_point_0" -> "2483 dequantize_per_channel_default_119" [label="(1152,)", style=solid]; -"2482 _frozen_param119" -> "2483 dequantize_per_channel_default_119" [label="(1152, 384)", style=solid]; -"2483 dequantize_per_channel_default_119" -> "2484 linear_118" [label="(1152, 384)", style=solid]; -"2484 linear_118" -> "2485 reshape_86" [label="(4, 64, 1152)", style=solid]; -"2485 reshape_86" -> "2486 permute_88" [label="(4, 64, 3, 12, 32)", style=solid]; -"2486 permute_88" -> "2487 select_57" [label="(3, 4, 12, 64, 32)", style=solid]; -"2486 permute_88" -> "2488 select_58" [label="(3, 4, 12, 64, 32)", style=solid]; -"2486 permute_88" -> "2489 select_59" [label="(3, 4, 12, 64, 32)", style=solid]; -"2487 select_57" -> "2490 linalg_vector_norm_38" [label="(4, 12, 64, 32)", style=solid]; -"2487 select_57" -> "2492 expand_as_38" [label="(4, 12, 64, 32)", style=solid]; -"2487 select_57" -> "2493 div_38" [label="(4, 12, 64, 32)", style=solid]; -"2488 select_58" -> "2496 linalg_vector_norm_39" [label="(4, 12, 64, 32)", style=solid]; -"2488 select_58" -> "2498 expand_as_39" [label="(4, 12, 64, 32)", style=solid]; -"2488 select_58" -> "2499 div_39" [label="(4, 12, 64, 32)", style=solid]; -"2489 select_59" -> "2526 matmul_39" [label="(4, 12, 64, 32)", style=solid]; -"2490 linalg_vector_norm_38" -> "2491 clamp_min_38" [label="(4, 12, 64, 1)", style=solid]; -"2491 clamp_min_38" -> "2492 expand_as_38" [label="(4, 12, 64, 1)", style=solid]; -"2492 expand_as_38" -> "2493 div_38" [label="(4, 12, 64, 32)", style=solid]; -"2493 div_38" -> "2494 quantize_per_tensor_default_118" [label="(4, 12, 64, 32)", style=solid]; -"2494 quantize_per_tensor_default_118" -> "2495 dequantize_per_tensor_default_118" [label="(4, 12, 64, 32)", style=solid]; -"2495 dequantize_per_tensor_default_118" -> "2503 matmul_38" [label="(4, 12, 64, 32)", style=solid]; -"2496 linalg_vector_norm_39" -> "2497 clamp_min_39" [label="(4, 12, 64, 1)", style=solid]; -"2497 clamp_min_39" -> "2498 expand_as_39" [label="(4, 12, 64, 1)", style=solid]; -"2498 expand_as_39" -> "2499 div_39" [label="(4, 12, 64, 32)", style=solid]; -"2499 div_39" -> "2500 quantize_per_tensor_default_119" [label="(4, 12, 64, 32)", style=solid]; -"2500 quantize_per_tensor_default_119" -> "2501 dequantize_per_tensor_default_119" [label="(4, 12, 64, 32)", style=solid]; -"2501 dequantize_per_tensor_default_119" -> "2502 transpose_38" [label="(4, 12, 64, 32)", style=solid]; -"2502 transpose_38" -> "2503 matmul_38" [label="(4, 12, 32, 64)", style=solid]; -"2503 matmul_38" -> "2506 mul_39" [label="(4, 12, 64, 64)", style=solid]; -"2504 clamp_19" -> "2505 exp_19" [label="(12, 1, 1)", style=solid]; -"2505 exp_19" -> "2506 mul_39" [label="(12, 1, 1)", style=solid]; -"2506 mul_39" -> "2507 add_66" [label="(4, 12, 64, 64)", style=solid]; -"2507 add_66" -> "2519 view_107" [label="(4, 12, 64, 64)", style=solid]; -"2508 new_zeros_9" -> "2509 view_106" [label="(16, 16)", style=solid]; -"2509 view_106" -> "2510 permute_89" [label="(2, 8, 2, 8)", style=solid]; -"2510 permute_89" -> "2511 reshape_87" [label="(2, 2, 8, 8)", style=solid]; -"2511 reshape_87" -> "2512 unsqueeze_56" [label="(4, 64)", style=solid]; -"2511 reshape_87" -> "2513 unsqueeze_57" [label="(4, 64)", style=solid]; -"2512 unsqueeze_56" -> "2514 sub_9" [label="(4, 1, 64)", style=solid]; -"2513 unsqueeze_57" -> "2514 sub_9" [label="(4, 64, 1)", style=solid]; -"2514 sub_9" -> "2515 ne_9" [label="(4, 64, 64)", style=solid]; -"2514 sub_9" -> "2516 masked_fill_18" [label="(4, 64, 64)", style=solid]; -"2514 sub_9" -> "2517 eq_9" [label="(4, 64, 64)", style=solid]; -"2515 ne_9" -> "2516 masked_fill_18" [label="(4, 64, 64)", style=solid]; -"2516 masked_fill_18" -> "2518 masked_fill_19" [label="(4, 64, 64)", style=solid]; -"2517 eq_9" -> "2518 masked_fill_19" [label="(4, 64, 64)", style=solid]; -"2518 masked_fill_19" -> "2520 unsqueeze_58" [label="(4, 64, 64)", style=solid]; -"2519 view_107" -> "2522 add_67" [label="(1, 4, 12, 64, 64)", style=solid]; -"2520 unsqueeze_58" -> "2521 unsqueeze_59" [label="(4, 1, 64, 64)", style=solid]; -"2521 unsqueeze_59" -> "2522 add_67" [label="(1, 4, 1, 64, 64)", style=solid]; -"2522 add_67" -> "2523 view_108" [label="(1, 4, 12, 64, 64)", style=solid]; -"2523 view_108" -> "2524 softmax_19" [label="(4, 12, 64, 64)", style=solid]; -"2524 softmax_19" -> "2525 dropout_76" [label="(4, 12, 64, 64)", style=solid]; -"2525 dropout_76" -> "2526 matmul_39" [label="(4, 12, 64, 64)", style=solid]; -"2526 matmul_39" -> "2527 transpose_39" [label="(4, 12, 64, 32)", style=solid]; -"2527 transpose_39" -> "2528 reshape_88" [label="(4, 64, 12, 32)", style=solid]; -"2528 reshape_88" -> "2529 reshape_88_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"2529 reshape_88_0_0_nncf_smooth_quant_0" -> "2530 quantize_per_tensor_default_120" [label="(4, 64, 384)", style=solid]; -"2530 quantize_per_tensor_default_120" -> "2531 dequantize_per_tensor_default_120" [label="(4, 64, 384)", style=solid]; -"2531 dequantize_per_tensor_default_120" -> "2536 linear_119" [label="(4, 64, 384)", style=solid]; -"2532 linear_119_scale_0" -> "2535 dequantize_per_channel_default_120" [label="(384,)", style=solid]; -"2533 linear_119_zero_point_0" -> "2535 dequantize_per_channel_default_120" [label="(384,)", style=solid]; -"2534 _frozen_param120" -> "2535 dequantize_per_channel_default_120" [label="(384, 384)", style=solid]; -"2535 dequantize_per_channel_default_120" -> "2536 linear_119" [label="(384, 384)", style=solid]; -"2536 linear_119" -> "2537 dropout_77" [label="(4, 64, 384)", style=solid]; -"2537 dropout_77" -> "2538 view_109" [label="(4, 64, 384)", style=solid]; -"2538 view_109" -> "2539 permute_90" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2539 permute_90" -> "2540 reshape_89" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2540 reshape_89" -> "2541 roll_19" [label="(1, 16, 16, 384)", style=solid]; -"2541 roll_19" -> "2542 slice_297" [label="(1, 16, 16, 384)", style=solid]; -"2542 slice_297" -> "2543 slice_298" [label="(1, 16, 16, 384)", style=solid]; -"2543 slice_298" -> "2544 slice_299" [label="(1, 14, 16, 384)", style=solid]; -"2544 slice_299" -> "2545 slice_300" [label="(1, 14, 14, 384)", style=solid]; -"2545 slice_300" -> "2546 contiguous_37" [label="(1, 14, 14, 384)", style=solid]; -"2546 contiguous_37" -> "2547 layer_norm_41" [label="(1, 14, 14, 384)", style=solid]; -"2547 layer_norm_41" -> "2548 add_68" [label="(1, 14, 14, 384)", style=solid]; -"2548 add_68" -> "2549 add_68_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; -"2548 add_68" -> "2569 add_69" [label="(1, 14, 14, 384)", style=solid]; -"2549 add_68_0_0_nncf_smooth_quant_0" -> "2550 quantize_per_tensor_default_121" [label="(1, 14, 14, 384)", style=solid]; -"2550 quantize_per_tensor_default_121" -> "2551 dequantize_per_tensor_default_121" [label="(1, 14, 14, 384)", style=solid]; -"2551 dequantize_per_tensor_default_121" -> "2556 linear_120" [label="(1, 14, 14, 384)", style=solid]; -"2552 linear_120_scale_0" -> "2555 dequantize_per_channel_default_121" [label="(1536,)", style=solid]; -"2553 linear_120_zero_point_0" -> "2555 dequantize_per_channel_default_121" [label="(1536,)", style=solid]; -"2554 _frozen_param121" -> "2555 dequantize_per_channel_default_121" [label="(1536, 384)", style=solid]; -"2555 dequantize_per_channel_default_121" -> "2556 linear_120" [label="(1536, 384)", style=solid]; -"2556 linear_120" -> "2557 gelu_19" [label="(1, 14, 14, 1536)", style=solid]; -"2557 gelu_19" -> "2558 dropout_78" [label="(1, 14, 14, 1536)", style=solid]; -"2558 dropout_78" -> "2559 dropout_78_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; -"2559 dropout_78_0_0_nncf_smooth_quant_0" -> "2560 quantize_per_tensor_default_122" [label="(1, 14, 14, 1536)", style=solid]; -"2560 quantize_per_tensor_default_122" -> "2561 dequantize_per_tensor_default_122" [label="(1, 14, 14, 1536)", style=solid]; -"2561 dequantize_per_tensor_default_122" -> "2566 linear_121" [label="(1, 14, 14, 1536)", style=solid]; -"2562 linear_121_scale_0" -> "2565 dequantize_per_channel_default_122" [label="(384,)", style=solid]; -"2563 linear_121_zero_point_0" -> "2565 dequantize_per_channel_default_122" [label="(384,)", style=solid]; -"2564 _frozen_param122" -> "2565 dequantize_per_channel_default_122" [label="(384, 1536)", style=solid]; -"2565 dequantize_per_channel_default_122" -> "2566 linear_121" [label="(384, 1536)", style=solid]; -"2566 linear_121" -> "2567 dropout_79" [label="(1, 14, 14, 384)", style=solid]; -"2567 dropout_79" -> "2568 layer_norm_42" [label="(1, 14, 14, 384)", style=solid]; -"2568 layer_norm_42" -> "2569 add_69" [label="(1, 14, 14, 384)", style=solid]; -"2569 add_69" -> "2591 pad_22" [label="(1, 14, 14, 384)", style=solid]; -"2569 add_69" -> "2650 add_71" [label="(1, 14, 14, 384)", style=solid]; -"2570 features_5_16_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "2575 linear_122" [label="(1, 15, 15, 2)", style=solid]; -"2571 linear_122_scale_0" -> "2574 dequantize_per_channel_default_123" [label="(512,)", style=solid]; -"2572 linear_122_zero_point_0" -> "2574 dequantize_per_channel_default_123" [label="(512,)", style=solid]; -"2573 _frozen_param123" -> "2574 dequantize_per_channel_default_123" [label="(512, 2)", style=solid]; -"2574 dequantize_per_channel_default_123" -> "2575 linear_122" [label="(512, 2)", style=solid]; -"2575 linear_122" -> "2576 relu__20" [label="(1, 15, 15, 512)", style=solid]; -"2576 relu__20" -> "2577 relu__20_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; -"2577 relu__20_0_0_nncf_smooth_quant_0" -> "2582 linear_123" [label="(1, 15, 15, 512)", style=solid]; -"2578 linear_123_scale_0" -> "2581 dequantize_per_channel_default_124" [label="(12,)", style=solid]; -"2579 linear_123_zero_point_0" -> "2581 dequantize_per_channel_default_124" [label="(12,)", style=solid]; -"2580 _frozen_param124" -> "2581 dequantize_per_channel_default_124" [label="(12, 512)", style=solid]; -"2581 dequantize_per_channel_default_124" -> "2582 linear_123" [label="(12, 512)", style=solid]; -"2582 linear_123" -> "2583 view_110" [label="(1, 15, 15, 12)", style=solid]; -"2583 view_110" -> "2584 index_20" [label="(225, 12)", style=solid]; -"2584 index_20" -> "2585 view_111" [label="(4096, 12)", style=solid]; -"2585 view_111" -> "2586 permute_91" [label="(64, 64, 12)", style=solid]; -"2586 permute_91" -> "2587 contiguous_38" [label="(12, 64, 64)", style=solid]; -"2587 contiguous_38" -> "2588 unsqueeze_60" [label="(12, 64, 64)", style=solid]; -"2588 unsqueeze_60" -> "2589 sigmoid_20" [label="(1, 12, 64, 64)", style=solid]; -"2589 sigmoid_20" -> "2590 mul_40" [label="(1, 12, 64, 64)", style=solid]; -"2590 mul_40" -> "2626 add_70" [label="(1, 12, 64, 64)", style=solid]; -"2591 pad_22" -> "2592 view_112" [label="(1, 16, 16, 384)", style=solid]; -"2592 view_112" -> "2593 permute_92" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2593 permute_92" -> "2594 reshape_90" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2594 reshape_90" -> "2596 reshape_90_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"2595 clone_20" -> "2603 linear_124" [label="(1152,)", style=solid]; -"2596 reshape_90_0_0_nncf_smooth_quant_0" -> "2597 quantize_per_tensor_default_123" [label="(4, 64, 384)", style=solid]; -"2597 quantize_per_tensor_default_123" -> "2598 dequantize_per_tensor_default_123" [label="(4, 64, 384)", style=solid]; -"2598 dequantize_per_tensor_default_123" -> "2603 linear_124" [label="(4, 64, 384)", style=solid]; -"2599 linear_124_scale_0" -> "2602 dequantize_per_channel_default_125" [label="(1152,)", style=solid]; -"2600 linear_124_zero_point_0" -> "2602 dequantize_per_channel_default_125" [label="(1152,)", style=solid]; -"2601 _frozen_param125" -> "2602 dequantize_per_channel_default_125" [label="(1152, 384)", style=solid]; -"2602 dequantize_per_channel_default_125" -> "2603 linear_124" [label="(1152, 384)", style=solid]; -"2603 linear_124" -> "2604 reshape_91" [label="(4, 64, 1152)", style=solid]; -"2604 reshape_91" -> "2605 permute_93" [label="(4, 64, 3, 12, 32)", style=solid]; -"2605 permute_93" -> "2606 select_60" [label="(3, 4, 12, 64, 32)", style=solid]; -"2605 permute_93" -> "2607 select_61" [label="(3, 4, 12, 64, 32)", style=solid]; -"2605 permute_93" -> "2608 select_62" [label="(3, 4, 12, 64, 32)", style=solid]; -"2606 select_60" -> "2609 linalg_vector_norm_40" [label="(4, 12, 64, 32)", style=solid]; -"2606 select_60" -> "2611 expand_as_40" [label="(4, 12, 64, 32)", style=solid]; -"2606 select_60" -> "2612 div_40" [label="(4, 12, 64, 32)", style=solid]; -"2607 select_61" -> "2615 linalg_vector_norm_41" [label="(4, 12, 64, 32)", style=solid]; -"2607 select_61" -> "2617 expand_as_41" [label="(4, 12, 64, 32)", style=solid]; -"2607 select_61" -> "2618 div_41" [label="(4, 12, 64, 32)", style=solid]; -"2608 select_62" -> "2629 matmul_41" [label="(4, 12, 64, 32)", style=solid]; -"2609 linalg_vector_norm_40" -> "2610 clamp_min_40" [label="(4, 12, 64, 1)", style=solid]; -"2610 clamp_min_40" -> "2611 expand_as_40" [label="(4, 12, 64, 1)", style=solid]; -"2611 expand_as_40" -> "2612 div_40" [label="(4, 12, 64, 32)", style=solid]; -"2612 div_40" -> "2613 quantize_per_tensor_default_124" [label="(4, 12, 64, 32)", style=solid]; -"2613 quantize_per_tensor_default_124" -> "2614 dequantize_per_tensor_default_124" [label="(4, 12, 64, 32)", style=solid]; -"2614 dequantize_per_tensor_default_124" -> "2622 matmul_40" [label="(4, 12, 64, 32)", style=solid]; -"2615 linalg_vector_norm_41" -> "2616 clamp_min_41" [label="(4, 12, 64, 1)", style=solid]; -"2616 clamp_min_41" -> "2617 expand_as_41" [label="(4, 12, 64, 1)", style=solid]; -"2617 expand_as_41" -> "2618 div_41" [label="(4, 12, 64, 32)", style=solid]; -"2618 div_41" -> "2619 quantize_per_tensor_default_125" [label="(4, 12, 64, 32)", style=solid]; -"2619 quantize_per_tensor_default_125" -> "2620 dequantize_per_tensor_default_125" [label="(4, 12, 64, 32)", style=solid]; -"2620 dequantize_per_tensor_default_125" -> "2621 transpose_40" [label="(4, 12, 64, 32)", style=solid]; -"2621 transpose_40" -> "2622 matmul_40" [label="(4, 12, 32, 64)", style=solid]; -"2622 matmul_40" -> "2625 mul_41" [label="(4, 12, 64, 64)", style=solid]; -"2623 clamp_20" -> "2624 exp_20" [label="(12, 1, 1)", style=solid]; -"2624 exp_20" -> "2625 mul_41" [label="(12, 1, 1)", style=solid]; -"2625 mul_41" -> "2626 add_70" [label="(4, 12, 64, 64)", style=solid]; -"2626 add_70" -> "2627 softmax_20" [label="(4, 12, 64, 64)", style=solid]; -"2627 softmax_20" -> "2628 dropout_80" [label="(4, 12, 64, 64)", style=solid]; -"2628 dropout_80" -> "2629 matmul_41" [label="(4, 12, 64, 64)", style=solid]; -"2629 matmul_41" -> "2630 transpose_41" [label="(4, 12, 64, 32)", style=solid]; -"2630 transpose_41" -> "2631 reshape_92" [label="(4, 64, 12, 32)", style=solid]; -"2631 reshape_92" -> "2632 reshape_92_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"2632 reshape_92_0_0_nncf_smooth_quant_0" -> "2633 quantize_per_tensor_default_126" [label="(4, 64, 384)", style=solid]; -"2633 quantize_per_tensor_default_126" -> "2634 dequantize_per_tensor_default_126" [label="(4, 64, 384)", style=solid]; -"2634 dequantize_per_tensor_default_126" -> "2639 linear_125" [label="(4, 64, 384)", style=solid]; -"2635 linear_125_scale_0" -> "2638 dequantize_per_channel_default_126" [label="(384,)", style=solid]; -"2636 linear_125_zero_point_0" -> "2638 dequantize_per_channel_default_126" [label="(384,)", style=solid]; -"2637 _frozen_param126" -> "2638 dequantize_per_channel_default_126" [label="(384, 384)", style=solid]; -"2638 dequantize_per_channel_default_126" -> "2639 linear_125" [label="(384, 384)", style=solid]; -"2639 linear_125" -> "2640 dropout_81" [label="(4, 64, 384)", style=solid]; -"2640 dropout_81" -> "2641 view_113" [label="(4, 64, 384)", style=solid]; -"2641 view_113" -> "2642 permute_94" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2642 permute_94" -> "2643 reshape_93" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2643 reshape_93" -> "2644 slice_302" [label="(1, 16, 16, 384)", style=solid]; -"2644 slice_302" -> "2645 slice_303" [label="(1, 16, 16, 384)", style=solid]; -"2645 slice_303" -> "2646 slice_304" [label="(1, 14, 16, 384)", style=solid]; -"2646 slice_304" -> "2647 slice_305" [label="(1, 14, 14, 384)", style=solid]; -"2647 slice_305" -> "2648 contiguous_39" [label="(1, 14, 14, 384)", style=solid]; -"2648 contiguous_39" -> "2649 layer_norm_43" [label="(1, 14, 14, 384)", style=solid]; -"2649 layer_norm_43" -> "2650 add_71" [label="(1, 14, 14, 384)", style=solid]; -"2650 add_71" -> "2651 add_71_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; -"2650 add_71" -> "2671 add_72" [label="(1, 14, 14, 384)", style=solid]; -"2651 add_71_0_0_nncf_smooth_quant_0" -> "2652 quantize_per_tensor_default_127" [label="(1, 14, 14, 384)", style=solid]; -"2652 quantize_per_tensor_default_127" -> "2653 dequantize_per_tensor_default_127" [label="(1, 14, 14, 384)", style=solid]; -"2653 dequantize_per_tensor_default_127" -> "2658 linear_126" [label="(1, 14, 14, 384)", style=solid]; -"2654 linear_126_scale_0" -> "2657 dequantize_per_channel_default_127" [label="(1536,)", style=solid]; -"2655 linear_126_zero_point_0" -> "2657 dequantize_per_channel_default_127" [label="(1536,)", style=solid]; -"2656 _frozen_param127" -> "2657 dequantize_per_channel_default_127" [label="(1536, 384)", style=solid]; -"2657 dequantize_per_channel_default_127" -> "2658 linear_126" [label="(1536, 384)", style=solid]; -"2658 linear_126" -> "2659 gelu_20" [label="(1, 14, 14, 1536)", style=solid]; -"2659 gelu_20" -> "2660 dropout_82" [label="(1, 14, 14, 1536)", style=solid]; -"2660 dropout_82" -> "2661 dropout_82_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; -"2661 dropout_82_0_0_nncf_smooth_quant_0" -> "2662 quantize_per_tensor_default_128" [label="(1, 14, 14, 1536)", style=solid]; -"2662 quantize_per_tensor_default_128" -> "2663 dequantize_per_tensor_default_128" [label="(1, 14, 14, 1536)", style=solid]; -"2663 dequantize_per_tensor_default_128" -> "2668 linear_127" [label="(1, 14, 14, 1536)", style=solid]; -"2664 linear_127_scale_0" -> "2667 dequantize_per_channel_default_128" [label="(384,)", style=solid]; -"2665 linear_127_zero_point_0" -> "2667 dequantize_per_channel_default_128" [label="(384,)", style=solid]; -"2666 _frozen_param128" -> "2667 dequantize_per_channel_default_128" [label="(384, 1536)", style=solid]; -"2667 dequantize_per_channel_default_128" -> "2668 linear_127" [label="(384, 1536)", style=solid]; -"2668 linear_127" -> "2669 dropout_83" [label="(1, 14, 14, 384)", style=solid]; -"2669 dropout_83" -> "2670 layer_norm_44" [label="(1, 14, 14, 384)", style=solid]; -"2670 layer_norm_44" -> "2671 add_72" [label="(1, 14, 14, 384)", style=solid]; -"2671 add_72" -> "2693 pad_23" [label="(1, 14, 14, 384)", style=solid]; -"2671 add_72" -> "2770 add_75" [label="(1, 14, 14, 384)", style=solid]; -"2672 features_5_17_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "2677 linear_128" [label="(1, 15, 15, 2)", style=solid]; -"2673 linear_128_scale_0" -> "2676 dequantize_per_channel_default_129" [label="(512,)", style=solid]; -"2674 linear_128_zero_point_0" -> "2676 dequantize_per_channel_default_129" [label="(512,)", style=solid]; -"2675 _frozen_param129" -> "2676 dequantize_per_channel_default_129" [label="(512, 2)", style=solid]; -"2676 dequantize_per_channel_default_129" -> "2677 linear_128" [label="(512, 2)", style=solid]; -"2677 linear_128" -> "2678 relu__21" [label="(1, 15, 15, 512)", style=solid]; -"2678 relu__21" -> "2679 relu__21_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; -"2679 relu__21_0_0_nncf_smooth_quant_0" -> "2684 linear_129" [label="(1, 15, 15, 512)", style=solid]; -"2680 linear_129_scale_0" -> "2683 dequantize_per_channel_default_130" [label="(12,)", style=solid]; -"2681 linear_129_zero_point_0" -> "2683 dequantize_per_channel_default_130" [label="(12,)", style=solid]; -"2682 _frozen_param130" -> "2683 dequantize_per_channel_default_130" [label="(12, 512)", style=solid]; -"2683 dequantize_per_channel_default_130" -> "2684 linear_129" [label="(12, 512)", style=solid]; -"2684 linear_129" -> "2685 view_114" [label="(1, 15, 15, 12)", style=solid]; -"2685 view_114" -> "2686 index_21" [label="(225, 12)", style=solid]; -"2686 index_21" -> "2687 view_115" [label="(4096, 12)", style=solid]; -"2687 view_115" -> "2688 permute_95" [label="(64, 64, 12)", style=solid]; -"2688 permute_95" -> "2689 contiguous_40" [label="(12, 64, 64)", style=solid]; -"2689 contiguous_40" -> "2690 unsqueeze_61" [label="(12, 64, 64)", style=solid]; -"2690 unsqueeze_61" -> "2691 sigmoid_21" [label="(1, 12, 64, 64)", style=solid]; -"2691 sigmoid_21" -> "2692 mul_42" [label="(1, 12, 64, 64)", style=solid]; -"2692 mul_42" -> "2729 add_73" [label="(1, 12, 64, 64)", style=solid]; -"2693 pad_23" -> "2694 roll_20" [label="(1, 16, 16, 384)", style=solid]; -"2694 roll_20" -> "2695 view_116" [label="(1, 16, 16, 384)", style=solid]; -"2695 view_116" -> "2696 permute_96" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2696 permute_96" -> "2697 reshape_94" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2697 reshape_94" -> "2699 reshape_94_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"2697 reshape_94" -> "2730 new_zeros_10" [label="(4, 64, 384)", style=solid]; -"2698 clone_21" -> "2706 linear_130" [label="(1152,)", style=solid]; -"2699 reshape_94_0_0_nncf_smooth_quant_0" -> "2700 quantize_per_tensor_default_129" [label="(4, 64, 384)", style=solid]; -"2700 quantize_per_tensor_default_129" -> "2701 dequantize_per_tensor_default_129" [label="(4, 64, 384)", style=solid]; -"2701 dequantize_per_tensor_default_129" -> "2706 linear_130" [label="(4, 64, 384)", style=solid]; -"2702 linear_130_scale_0" -> "2705 dequantize_per_channel_default_131" [label="(1152,)", style=solid]; -"2703 linear_130_zero_point_0" -> "2705 dequantize_per_channel_default_131" [label="(1152,)", style=solid]; -"2704 _frozen_param131" -> "2705 dequantize_per_channel_default_131" [label="(1152, 384)", style=solid]; -"2705 dequantize_per_channel_default_131" -> "2706 linear_130" [label="(1152, 384)", style=solid]; -"2706 linear_130" -> "2707 reshape_95" [label="(4, 64, 1152)", style=solid]; -"2707 reshape_95" -> "2708 permute_97" [label="(4, 64, 3, 12, 32)", style=solid]; -"2708 permute_97" -> "2709 select_63" [label="(3, 4, 12, 64, 32)", style=solid]; -"2708 permute_97" -> "2710 select_64" [label="(3, 4, 12, 64, 32)", style=solid]; -"2708 permute_97" -> "2711 select_65" [label="(3, 4, 12, 64, 32)", style=solid]; -"2709 select_63" -> "2712 linalg_vector_norm_42" [label="(4, 12, 64, 32)", style=solid]; -"2709 select_63" -> "2714 expand_as_42" [label="(4, 12, 64, 32)", style=solid]; -"2709 select_63" -> "2715 div_42" [label="(4, 12, 64, 32)", style=solid]; -"2710 select_64" -> "2718 linalg_vector_norm_43" [label="(4, 12, 64, 32)", style=solid]; -"2710 select_64" -> "2720 expand_as_43" [label="(4, 12, 64, 32)", style=solid]; -"2710 select_64" -> "2721 div_43" [label="(4, 12, 64, 32)", style=solid]; -"2711 select_65" -> "2748 matmul_43" [label="(4, 12, 64, 32)", style=solid]; -"2712 linalg_vector_norm_42" -> "2713 clamp_min_42" [label="(4, 12, 64, 1)", style=solid]; -"2713 clamp_min_42" -> "2714 expand_as_42" [label="(4, 12, 64, 1)", style=solid]; -"2714 expand_as_42" -> "2715 div_42" [label="(4, 12, 64, 32)", style=solid]; -"2715 div_42" -> "2716 quantize_per_tensor_default_130" [label="(4, 12, 64, 32)", style=solid]; -"2716 quantize_per_tensor_default_130" -> "2717 dequantize_per_tensor_default_130" [label="(4, 12, 64, 32)", style=solid]; -"2717 dequantize_per_tensor_default_130" -> "2725 matmul_42" [label="(4, 12, 64, 32)", style=solid]; -"2718 linalg_vector_norm_43" -> "2719 clamp_min_43" [label="(4, 12, 64, 1)", style=solid]; -"2719 clamp_min_43" -> "2720 expand_as_43" [label="(4, 12, 64, 1)", style=solid]; -"2720 expand_as_43" -> "2721 div_43" [label="(4, 12, 64, 32)", style=solid]; -"2721 div_43" -> "2722 quantize_per_tensor_default_131" [label="(4, 12, 64, 32)", style=solid]; -"2722 quantize_per_tensor_default_131" -> "2723 dequantize_per_tensor_default_131" [label="(4, 12, 64, 32)", style=solid]; -"2723 dequantize_per_tensor_default_131" -> "2724 transpose_42" [label="(4, 12, 64, 32)", style=solid]; -"2724 transpose_42" -> "2725 matmul_42" [label="(4, 12, 32, 64)", style=solid]; -"2725 matmul_42" -> "2728 mul_43" [label="(4, 12, 64, 64)", style=solid]; -"2726 clamp_21" -> "2727 exp_21" [label="(12, 1, 1)", style=solid]; -"2727 exp_21" -> "2728 mul_43" [label="(12, 1, 1)", style=solid]; -"2728 mul_43" -> "2729 add_73" [label="(4, 12, 64, 64)", style=solid]; -"2729 add_73" -> "2741 view_118" [label="(4, 12, 64, 64)", style=solid]; -"2730 new_zeros_10" -> "2731 view_117" [label="(16, 16)", style=solid]; -"2731 view_117" -> "2732 permute_98" [label="(2, 8, 2, 8)", style=solid]; -"2732 permute_98" -> "2733 reshape_96" [label="(2, 2, 8, 8)", style=solid]; -"2733 reshape_96" -> "2734 unsqueeze_62" [label="(4, 64)", style=solid]; -"2733 reshape_96" -> "2735 unsqueeze_63" [label="(4, 64)", style=solid]; -"2734 unsqueeze_62" -> "2736 sub_10" [label="(4, 1, 64)", style=solid]; -"2735 unsqueeze_63" -> "2736 sub_10" [label="(4, 64, 1)", style=solid]; -"2736 sub_10" -> "2737 ne_10" [label="(4, 64, 64)", style=solid]; -"2736 sub_10" -> "2738 masked_fill_20" [label="(4, 64, 64)", style=solid]; -"2736 sub_10" -> "2739 eq_10" [label="(4, 64, 64)", style=solid]; -"2737 ne_10" -> "2738 masked_fill_20" [label="(4, 64, 64)", style=solid]; -"2738 masked_fill_20" -> "2740 masked_fill_21" [label="(4, 64, 64)", style=solid]; -"2739 eq_10" -> "2740 masked_fill_21" [label="(4, 64, 64)", style=solid]; -"2740 masked_fill_21" -> "2742 unsqueeze_64" [label="(4, 64, 64)", style=solid]; -"2741 view_118" -> "2744 add_74" [label="(1, 4, 12, 64, 64)", style=solid]; -"2742 unsqueeze_64" -> "2743 unsqueeze_65" [label="(4, 1, 64, 64)", style=solid]; -"2743 unsqueeze_65" -> "2744 add_74" [label="(1, 4, 1, 64, 64)", style=solid]; -"2744 add_74" -> "2745 view_119" [label="(1, 4, 12, 64, 64)", style=solid]; -"2745 view_119" -> "2746 softmax_21" [label="(4, 12, 64, 64)", style=solid]; -"2746 softmax_21" -> "2747 dropout_84" [label="(4, 12, 64, 64)", style=solid]; -"2747 dropout_84" -> "2748 matmul_43" [label="(4, 12, 64, 64)", style=solid]; -"2748 matmul_43" -> "2749 transpose_43" [label="(4, 12, 64, 32)", style=solid]; -"2749 transpose_43" -> "2750 reshape_97" [label="(4, 64, 12, 32)", style=solid]; -"2750 reshape_97" -> "2751 reshape_97_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; -"2751 reshape_97_0_0_nncf_smooth_quant_0" -> "2752 quantize_per_tensor_default_132" [label="(4, 64, 384)", style=solid]; -"2752 quantize_per_tensor_default_132" -> "2753 dequantize_per_tensor_default_132" [label="(4, 64, 384)", style=solid]; -"2753 dequantize_per_tensor_default_132" -> "2758 linear_131" [label="(4, 64, 384)", style=solid]; -"2754 linear_131_scale_0" -> "2757 dequantize_per_channel_default_132" [label="(384,)", style=solid]; -"2755 linear_131_zero_point_0" -> "2757 dequantize_per_channel_default_132" [label="(384,)", style=solid]; -"2756 _frozen_param132" -> "2757 dequantize_per_channel_default_132" [label="(384, 384)", style=solid]; -"2757 dequantize_per_channel_default_132" -> "2758 linear_131" [label="(384, 384)", style=solid]; -"2758 linear_131" -> "2759 dropout_85" [label="(4, 64, 384)", style=solid]; -"2759 dropout_85" -> "2760 view_120" [label="(4, 64, 384)", style=solid]; -"2760 view_120" -> "2761 permute_99" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2761 permute_99" -> "2762 reshape_98" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2762 reshape_98" -> "2763 roll_21" [label="(1, 16, 16, 384)", style=solid]; -"2763 roll_21" -> "2764 slice_325" [label="(1, 16, 16, 384)", style=solid]; -"2764 slice_325" -> "2765 slice_326" [label="(1, 16, 16, 384)", style=solid]; -"2765 slice_326" -> "2766 slice_327" [label="(1, 14, 16, 384)", style=solid]; -"2766 slice_327" -> "2767 slice_328" [label="(1, 14, 14, 384)", style=solid]; -"2767 slice_328" -> "2768 contiguous_41" [label="(1, 14, 14, 384)", style=solid]; -"2768 contiguous_41" -> "2769 layer_norm_45" [label="(1, 14, 14, 384)", style=solid]; -"2769 layer_norm_45" -> "2770 add_75" [label="(1, 14, 14, 384)", style=solid]; -"2770 add_75" -> "2771 add_75_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; -"2770 add_75" -> "2791 add_76" [label="(1, 14, 14, 384)", style=solid]; -"2771 add_75_0_0_nncf_smooth_quant_0" -> "2772 quantize_per_tensor_default_133" [label="(1, 14, 14, 384)", style=solid]; -"2772 quantize_per_tensor_default_133" -> "2773 dequantize_per_tensor_default_133" [label="(1, 14, 14, 384)", style=solid]; -"2773 dequantize_per_tensor_default_133" -> "2778 linear_132" [label="(1, 14, 14, 384)", style=solid]; -"2774 linear_132_scale_0" -> "2777 dequantize_per_channel_default_133" [label="(1536,)", style=solid]; -"2775 linear_132_zero_point_0" -> "2777 dequantize_per_channel_default_133" [label="(1536,)", style=solid]; -"2776 _frozen_param133" -> "2777 dequantize_per_channel_default_133" [label="(1536, 384)", style=solid]; -"2777 dequantize_per_channel_default_133" -> "2778 linear_132" [label="(1536, 384)", style=solid]; -"2778 linear_132" -> "2779 gelu_21" [label="(1, 14, 14, 1536)", style=solid]; -"2779 gelu_21" -> "2780 dropout_86" [label="(1, 14, 14, 1536)", style=solid]; -"2780 dropout_86" -> "2781 dropout_86_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; -"2781 dropout_86_0_0_nncf_smooth_quant_0" -> "2782 quantize_per_tensor_default_134" [label="(1, 14, 14, 1536)", style=solid]; -"2782 quantize_per_tensor_default_134" -> "2783 dequantize_per_tensor_default_134" [label="(1, 14, 14, 1536)", style=solid]; -"2783 dequantize_per_tensor_default_134" -> "2788 linear_133" [label="(1, 14, 14, 1536)", style=solid]; -"2784 linear_133_scale_0" -> "2787 dequantize_per_channel_default_134" [label="(384,)", style=solid]; -"2785 linear_133_zero_point_0" -> "2787 dequantize_per_channel_default_134" [label="(384,)", style=solid]; -"2786 _frozen_param134" -> "2787 dequantize_per_channel_default_134" [label="(384, 1536)", style=solid]; -"2787 dequantize_per_channel_default_134" -> "2788 linear_133" [label="(384, 1536)", style=solid]; -"2788 linear_133" -> "2789 dropout_87" [label="(1, 14, 14, 384)", style=solid]; -"2789 dropout_87" -> "2790 layer_norm_46" [label="(1, 14, 14, 384)", style=solid]; -"2790 layer_norm_46" -> "2791 add_76" [label="(1, 14, 14, 384)", style=solid]; -"2791 add_76" -> "2792 pad_24" [label="(1, 14, 14, 384)", style=solid]; -"2792 pad_24" -> "2793 slice_329" [label="(1, 14, 14, 384)", style=solid]; -"2792 pad_24" -> "2796 slice_332" [label="(1, 14, 14, 384)", style=solid]; -"2792 pad_24" -> "2799 slice_335" [label="(1, 14, 14, 384)", style=solid]; -"2792 pad_24" -> "2802 slice_338" [label="(1, 14, 14, 384)", style=solid]; -"2793 slice_329" -> "2794 slice_330" [label="(1, 7, 14, 384)", style=solid]; -"2794 slice_330" -> "2795 slice_331" [label="(1, 7, 7, 384)", style=solid]; -"2795 slice_331" -> "2805 cat_2" [label="(1, 7, 7, 384)", style=solid]; -"2796 slice_332" -> "2797 slice_333" [label="(1, 7, 14, 384)", style=solid]; -"2797 slice_333" -> "2798 slice_334" [label="(1, 7, 7, 384)", style=solid]; -"2798 slice_334" -> "2805 cat_2" [label="(1, 7, 7, 384)", style=solid]; -"2799 slice_335" -> "2800 slice_336" [label="(1, 7, 14, 384)", style=solid]; -"2800 slice_336" -> "2801 slice_337" [label="(1, 7, 7, 384)", style=solid]; -"2801 slice_337" -> "2805 cat_2" [label="(1, 7, 7, 384)", style=solid]; -"2802 slice_338" -> "2803 slice_339" [label="(1, 7, 14, 384)", style=solid]; -"2803 slice_339" -> "2804 slice_340" [label="(1, 7, 7, 384)", style=solid]; -"2804 slice_340" -> "2805 cat_2" [label="(1, 7, 7, 384)", style=solid]; -"2805 cat_2" -> "2806 cat_2_0_0_nncf_smooth_quant_0" [label="(1, 7, 7, 1536)", style=solid]; -"2806 cat_2_0_0_nncf_smooth_quant_0" -> "2807 quantize_per_tensor_default_135" [label="(1, 7, 7, 1536)", style=solid]; -"2807 quantize_per_tensor_default_135" -> "2808 dequantize_per_tensor_default_135" [label="(1, 7, 7, 1536)", style=solid]; -"2808 dequantize_per_tensor_default_135" -> "2813 linear_134" [label="(1, 7, 7, 1536)", style=solid]; -"2809 linear_134_scale_0" -> "2812 dequantize_per_channel_default_135" [label="(768,)", style=solid]; -"2810 linear_134_zero_point_0" -> "2812 dequantize_per_channel_default_135" [label="(768,)", style=solid]; -"2811 _frozen_param135" -> "2812 dequantize_per_channel_default_135" [label="(768, 1536)", style=solid]; -"2812 dequantize_per_channel_default_135" -> "2813 linear_134" [label="(768, 1536)", style=solid]; -"2813 linear_134" -> "2814 layer_norm_47" [label="(1, 7, 7, 768)", style=solid]; -"2814 layer_norm_47" -> "2836 pad_25" [label="(1, 7, 7, 768)", style=solid]; -"2814 layer_norm_47" -> "2895 add_78" [label="(1, 7, 7, 768)", style=solid]; -"2815 features_7_0_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "2820 linear_135" [label="(1, 15, 15, 2)", style=solid]; -"2816 linear_135_scale_0" -> "2819 dequantize_per_channel_default_136" [label="(512,)", style=solid]; -"2817 linear_135_zero_point_0" -> "2819 dequantize_per_channel_default_136" [label="(512,)", style=solid]; -"2818 _frozen_param136" -> "2819 dequantize_per_channel_default_136" [label="(512, 2)", style=solid]; -"2819 dequantize_per_channel_default_136" -> "2820 linear_135" [label="(512, 2)", style=solid]; -"2820 linear_135" -> "2821 relu__22" [label="(1, 15, 15, 512)", style=solid]; -"2821 relu__22" -> "2822 relu__22_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; -"2822 relu__22_0_0_nncf_smooth_quant_0" -> "2827 linear_136" [label="(1, 15, 15, 512)", style=solid]; -"2823 linear_136_scale_0" -> "2826 dequantize_per_channel_default_137" [label="(24,)", style=solid]; -"2824 linear_136_zero_point_0" -> "2826 dequantize_per_channel_default_137" [label="(24,)", style=solid]; -"2825 _frozen_param137" -> "2826 dequantize_per_channel_default_137" [label="(24, 512)", style=solid]; -"2826 dequantize_per_channel_default_137" -> "2827 linear_136" [label="(24, 512)", style=solid]; -"2827 linear_136" -> "2828 view_121" [label="(1, 15, 15, 24)", style=solid]; -"2828 view_121" -> "2829 index_22" [label="(225, 24)", style=solid]; -"2829 index_22" -> "2830 view_122" [label="(4096, 24)", style=solid]; -"2830 view_122" -> "2831 permute_100" [label="(64, 64, 24)", style=solid]; -"2831 permute_100" -> "2832 contiguous_42" [label="(24, 64, 64)", style=solid]; -"2832 contiguous_42" -> "2833 unsqueeze_66" [label="(24, 64, 64)", style=solid]; -"2833 unsqueeze_66" -> "2834 sigmoid_22" [label="(1, 24, 64, 64)", style=solid]; -"2834 sigmoid_22" -> "2835 mul_44" [label="(1, 24, 64, 64)", style=solid]; -"2835 mul_44" -> "2871 add_77" [label="(1, 24, 64, 64)", style=solid]; -"2836 pad_25" -> "2837 view_123" [label="(1, 8, 8, 768)", style=solid]; -"2837 view_123" -> "2838 permute_101" [label="(1, 1, 8, 1, 8, 768)", style=solid]; -"2838 permute_101" -> "2839 reshape_99" [label="(1, 1, 1, 8, 8, 768)", style=solid]; -"2839 reshape_99" -> "2841 reshape_99_0_0_nncf_smooth_quant_0" [label="(1, 64, 768)", style=solid]; -"2840 clone_22" -> "2848 linear_137" [label="(2304,)", style=solid]; -"2841 reshape_99_0_0_nncf_smooth_quant_0" -> "2842 quantize_per_tensor_default_136" [label="(1, 64, 768)", style=solid]; -"2842 quantize_per_tensor_default_136" -> "2843 dequantize_per_tensor_default_136" [label="(1, 64, 768)", style=solid]; -"2843 dequantize_per_tensor_default_136" -> "2848 linear_137" [label="(1, 64, 768)", style=solid]; -"2844 linear_137_scale_0" -> "2847 dequantize_per_channel_default_138" [label="(2304,)", style=solid]; -"2845 linear_137_zero_point_0" -> "2847 dequantize_per_channel_default_138" [label="(2304,)", style=solid]; -"2846 _frozen_param138" -> "2847 dequantize_per_channel_default_138" [label="(2304, 768)", style=solid]; -"2847 dequantize_per_channel_default_138" -> "2848 linear_137" [label="(2304, 768)", style=solid]; -"2848 linear_137" -> "2849 reshape_100" [label="(1, 64, 2304)", style=solid]; -"2849 reshape_100" -> "2850 permute_102" [label="(1, 64, 3, 24, 32)", style=solid]; -"2850 permute_102" -> "2851 select_66" [label="(3, 1, 24, 64, 32)", style=solid]; -"2850 permute_102" -> "2852 select_67" [label="(3, 1, 24, 64, 32)", style=solid]; -"2850 permute_102" -> "2853 select_68" [label="(3, 1, 24, 64, 32)", style=solid]; -"2851 select_66" -> "2854 linalg_vector_norm_44" [label="(1, 24, 64, 32)", style=solid]; -"2851 select_66" -> "2856 expand_as_44" [label="(1, 24, 64, 32)", style=solid]; -"2851 select_66" -> "2857 div_44" [label="(1, 24, 64, 32)", style=solid]; -"2852 select_67" -> "2860 linalg_vector_norm_45" [label="(1, 24, 64, 32)", style=solid]; -"2852 select_67" -> "2862 expand_as_45" [label="(1, 24, 64, 32)", style=solid]; -"2852 select_67" -> "2863 div_45" [label="(1, 24, 64, 32)", style=solid]; -"2853 select_68" -> "2874 matmul_45" [label="(1, 24, 64, 32)", style=solid]; -"2854 linalg_vector_norm_44" -> "2855 clamp_min_44" [label="(1, 24, 64, 1)", style=solid]; -"2855 clamp_min_44" -> "2856 expand_as_44" [label="(1, 24, 64, 1)", style=solid]; -"2856 expand_as_44" -> "2857 div_44" [label="(1, 24, 64, 32)", style=solid]; -"2857 div_44" -> "2858 quantize_per_tensor_default_137" [label="(1, 24, 64, 32)", style=solid]; -"2858 quantize_per_tensor_default_137" -> "2859 dequantize_per_tensor_default_137" [label="(1, 24, 64, 32)", style=solid]; -"2859 dequantize_per_tensor_default_137" -> "2867 matmul_44" [label="(1, 24, 64, 32)", style=solid]; -"2860 linalg_vector_norm_45" -> "2861 clamp_min_45" [label="(1, 24, 64, 1)", style=solid]; -"2861 clamp_min_45" -> "2862 expand_as_45" [label="(1, 24, 64, 1)", style=solid]; -"2862 expand_as_45" -> "2863 div_45" [label="(1, 24, 64, 32)", style=solid]; -"2863 div_45" -> "2864 quantize_per_tensor_default_138" [label="(1, 24, 64, 32)", style=solid]; -"2864 quantize_per_tensor_default_138" -> "2865 dequantize_per_tensor_default_138" [label="(1, 24, 64, 32)", style=solid]; -"2865 dequantize_per_tensor_default_138" -> "2866 transpose_44" [label="(1, 24, 64, 32)", style=solid]; -"2866 transpose_44" -> "2867 matmul_44" [label="(1, 24, 32, 64)", style=solid]; -"2867 matmul_44" -> "2870 mul_45" [label="(1, 24, 64, 64)", style=solid]; -"2868 clamp_22" -> "2869 exp_22" [label="(24, 1, 1)", style=solid]; -"2869 exp_22" -> "2870 mul_45" [label="(24, 1, 1)", style=solid]; -"2870 mul_45" -> "2871 add_77" [label="(1, 24, 64, 64)", style=solid]; -"2871 add_77" -> "2872 softmax_22" [label="(1, 24, 64, 64)", style=solid]; -"2872 softmax_22" -> "2873 dropout_88" [label="(1, 24, 64, 64)", style=solid]; -"2873 dropout_88" -> "2874 matmul_45" [label="(1, 24, 64, 64)", style=solid]; -"2874 matmul_45" -> "2875 transpose_45" [label="(1, 24, 64, 32)", style=solid]; -"2875 transpose_45" -> "2876 reshape_101" [label="(1, 64, 24, 32)", style=solid]; -"2876 reshape_101" -> "2877 reshape_101_0_0_nncf_smooth_quant_0" [label="(1, 64, 768)", style=solid]; -"2877 reshape_101_0_0_nncf_smooth_quant_0" -> "2878 quantize_per_tensor_default_139" [label="(1, 64, 768)", style=solid]; -"2878 quantize_per_tensor_default_139" -> "2879 dequantize_per_tensor_default_139" [label="(1, 64, 768)", style=solid]; -"2879 dequantize_per_tensor_default_139" -> "2884 linear_138" [label="(1, 64, 768)", style=solid]; -"2880 linear_138_scale_0" -> "2883 dequantize_per_channel_default_139" [label="(768,)", style=solid]; -"2881 linear_138_zero_point_0" -> "2883 dequantize_per_channel_default_139" [label="(768,)", style=solid]; -"2882 _frozen_param139" -> "2883 dequantize_per_channel_default_139" [label="(768, 768)", style=solid]; -"2883 dequantize_per_channel_default_139" -> "2884 linear_138" [label="(768, 768)", style=solid]; -"2884 linear_138" -> "2885 dropout_89" [label="(1, 64, 768)", style=solid]; -"2885 dropout_89" -> "2886 view_124" [label="(1, 64, 768)", style=solid]; -"2886 view_124" -> "2887 permute_103" [label="(1, 1, 1, 8, 8, 768)", style=solid]; -"2887 permute_103" -> "2888 reshape_102" [label="(1, 1, 8, 1, 8, 768)", style=solid]; -"2888 reshape_102" -> "2889 slice_342" [label="(1, 8, 8, 768)", style=solid]; -"2889 slice_342" -> "2890 slice_343" [label="(1, 8, 8, 768)", style=solid]; -"2890 slice_343" -> "2891 slice_344" [label="(1, 7, 8, 768)", style=solid]; -"2891 slice_344" -> "2892 slice_345" [label="(1, 7, 7, 768)", style=solid]; -"2892 slice_345" -> "2893 contiguous_43" [label="(1, 7, 7, 768)", style=solid]; -"2893 contiguous_43" -> "2894 layer_norm_48" [label="(1, 7, 7, 768)", style=solid]; -"2894 layer_norm_48" -> "2895 add_78" [label="(1, 7, 7, 768)", style=solid]; -"2895 add_78" -> "2896 add_78_0_0_nncf_smooth_quant_0" [label="(1, 7, 7, 768)", style=solid]; -"2895 add_78" -> "2916 add_79" [label="(1, 7, 7, 768)", style=solid]; -"2896 add_78_0_0_nncf_smooth_quant_0" -> "2897 quantize_per_tensor_default_140" [label="(1, 7, 7, 768)", style=solid]; -"2897 quantize_per_tensor_default_140" -> "2898 dequantize_per_tensor_default_140" [label="(1, 7, 7, 768)", style=solid]; -"2898 dequantize_per_tensor_default_140" -> "2903 linear_139" [label="(1, 7, 7, 768)", style=solid]; -"2899 linear_139_scale_0" -> "2902 dequantize_per_channel_default_140" [label="(3072,)", style=solid]; -"2900 linear_139_zero_point_0" -> "2902 dequantize_per_channel_default_140" [label="(3072,)", style=solid]; -"2901 _frozen_param140" -> "2902 dequantize_per_channel_default_140" [label="(3072, 768)", style=solid]; -"2902 dequantize_per_channel_default_140" -> "2903 linear_139" [label="(3072, 768)", style=solid]; -"2903 linear_139" -> "2904 gelu_22" [label="(1, 7, 7, 3072)", style=solid]; -"2904 gelu_22" -> "2905 dropout_90" [label="(1, 7, 7, 3072)", style=solid]; -"2905 dropout_90" -> "2906 dropout_90_0_0_nncf_smooth_quant_0" [label="(1, 7, 7, 3072)", style=solid]; -"2906 dropout_90_0_0_nncf_smooth_quant_0" -> "2907 quantize_per_tensor_default_141" [label="(1, 7, 7, 3072)", style=solid]; -"2907 quantize_per_tensor_default_141" -> "2908 dequantize_per_tensor_default_141" [label="(1, 7, 7, 3072)", style=solid]; -"2908 dequantize_per_tensor_default_141" -> "2913 linear_140" [label="(1, 7, 7, 3072)", style=solid]; -"2909 linear_140_scale_0" -> "2912 dequantize_per_channel_default_141" [label="(768,)", style=solid]; -"2910 linear_140_zero_point_0" -> "2912 dequantize_per_channel_default_141" [label="(768,)", style=solid]; -"2911 _frozen_param141" -> "2912 dequantize_per_channel_default_141" [label="(768, 3072)", style=solid]; -"2912 dequantize_per_channel_default_141" -> "2913 linear_140" [label="(768, 3072)", style=solid]; -"2913 linear_140" -> "2914 dropout_91" [label="(1, 7, 7, 768)", style=solid]; -"2914 dropout_91" -> "2915 layer_norm_49" [label="(1, 7, 7, 768)", style=solid]; -"2915 layer_norm_49" -> "2916 add_79" [label="(1, 7, 7, 768)", style=solid]; -"2916 add_79" -> "2938 pad_26" [label="(1, 7, 7, 768)", style=solid]; -"2916 add_79" -> "2997 add_81" [label="(1, 7, 7, 768)", style=solid]; -"2917 features_7_1_attn_relative_coords_table_0_0_nncf_smooth_quant_0" -> "2922 linear_141" [label="(1, 15, 15, 2)", style=solid]; -"2918 linear_141_scale_0" -> "2921 dequantize_per_channel_default_142" [label="(512,)", style=solid]; -"2919 linear_141_zero_point_0" -> "2921 dequantize_per_channel_default_142" [label="(512,)", style=solid]; -"2920 _frozen_param142" -> "2921 dequantize_per_channel_default_142" [label="(512, 2)", style=solid]; -"2921 dequantize_per_channel_default_142" -> "2922 linear_141" [label="(512, 2)", style=solid]; -"2922 linear_141" -> "2923 relu__23" [label="(1, 15, 15, 512)", style=solid]; -"2923 relu__23" -> "2924 relu__23_0_0_nncf_smooth_quant_0" [label="(1, 15, 15, 512)", style=solid]; -"2924 relu__23_0_0_nncf_smooth_quant_0" -> "2929 linear_142" [label="(1, 15, 15, 512)", style=solid]; -"2925 linear_142_scale_0" -> "2928 dequantize_per_channel_default_143" [label="(24,)", style=solid]; -"2926 linear_142_zero_point_0" -> "2928 dequantize_per_channel_default_143" [label="(24,)", style=solid]; -"2927 _frozen_param143" -> "2928 dequantize_per_channel_default_143" [label="(24, 512)", style=solid]; -"2928 dequantize_per_channel_default_143" -> "2929 linear_142" [label="(24, 512)", style=solid]; -"2929 linear_142" -> "2930 view_125" [label="(1, 15, 15, 24)", style=solid]; -"2930 view_125" -> "2931 index_23" [label="(225, 24)", style=solid]; -"2931 index_23" -> "2932 view_126" [label="(4096, 24)", style=solid]; -"2932 view_126" -> "2933 permute_104" [label="(64, 64, 24)", style=solid]; -"2933 permute_104" -> "2934 contiguous_44" [label="(24, 64, 64)", style=solid]; -"2934 contiguous_44" -> "2935 unsqueeze_67" [label="(24, 64, 64)", style=solid]; -"2935 unsqueeze_67" -> "2936 sigmoid_23" [label="(1, 24, 64, 64)", style=solid]; -"2936 sigmoid_23" -> "2937 mul_46" [label="(1, 24, 64, 64)", style=solid]; -"2937 mul_46" -> "2973 add_80" [label="(1, 24, 64, 64)", style=solid]; -"2938 pad_26" -> "2939 view_127" [label="(1, 8, 8, 768)", style=solid]; -"2939 view_127" -> "2940 permute_105" [label="(1, 1, 8, 1, 8, 768)", style=solid]; -"2940 permute_105" -> "2941 reshape_103" [label="(1, 1, 1, 8, 8, 768)", style=solid]; -"2941 reshape_103" -> "2943 reshape_103_0_0_nncf_smooth_quant_0" [label="(1, 64, 768)", style=solid]; -"2942 clone_23" -> "2950 linear_143" [label="(2304,)", style=solid]; -"2943 reshape_103_0_0_nncf_smooth_quant_0" -> "2944 quantize_per_tensor_default_142" [label="(1, 64, 768)", style=solid]; -"2944 quantize_per_tensor_default_142" -> "2945 dequantize_per_tensor_default_142" [label="(1, 64, 768)", style=solid]; -"2945 dequantize_per_tensor_default_142" -> "2950 linear_143" [label="(1, 64, 768)", style=solid]; -"2946 linear_143_scale_0" -> "2949 dequantize_per_channel_default_144" [label="(2304,)", style=solid]; -"2947 linear_143_zero_point_0" -> "2949 dequantize_per_channel_default_144" [label="(2304,)", style=solid]; -"2948 _frozen_param144" -> "2949 dequantize_per_channel_default_144" [label="(2304, 768)", style=solid]; -"2949 dequantize_per_channel_default_144" -> "2950 linear_143" [label="(2304, 768)", style=solid]; -"2950 linear_143" -> "2951 reshape_104" [label="(1, 64, 2304)", style=solid]; -"2951 reshape_104" -> "2952 permute_106" [label="(1, 64, 3, 24, 32)", style=solid]; -"2952 permute_106" -> "2953 select_69" [label="(3, 1, 24, 64, 32)", style=solid]; -"2952 permute_106" -> "2954 select_70" [label="(3, 1, 24, 64, 32)", style=solid]; -"2952 permute_106" -> "2955 select_71" [label="(3, 1, 24, 64, 32)", style=solid]; -"2953 select_69" -> "2956 linalg_vector_norm_46" [label="(1, 24, 64, 32)", style=solid]; -"2953 select_69" -> "2958 expand_as_46" [label="(1, 24, 64, 32)", style=solid]; -"2953 select_69" -> "2959 div_46" [label="(1, 24, 64, 32)", style=solid]; -"2954 select_70" -> "2962 linalg_vector_norm_47" [label="(1, 24, 64, 32)", style=solid]; -"2954 select_70" -> "2964 expand_as_47" [label="(1, 24, 64, 32)", style=solid]; -"2954 select_70" -> "2965 div_47" [label="(1, 24, 64, 32)", style=solid]; -"2955 select_71" -> "2976 matmul_47" [label="(1, 24, 64, 32)", style=solid]; -"2956 linalg_vector_norm_46" -> "2957 clamp_min_46" [label="(1, 24, 64, 1)", style=solid]; -"2957 clamp_min_46" -> "2958 expand_as_46" [label="(1, 24, 64, 1)", style=solid]; -"2958 expand_as_46" -> "2959 div_46" [label="(1, 24, 64, 32)", style=solid]; -"2959 div_46" -> "2960 quantize_per_tensor_default_143" [label="(1, 24, 64, 32)", style=solid]; -"2960 quantize_per_tensor_default_143" -> "2961 dequantize_per_tensor_default_143" [label="(1, 24, 64, 32)", style=solid]; -"2961 dequantize_per_tensor_default_143" -> "2969 matmul_46" [label="(1, 24, 64, 32)", style=solid]; -"2962 linalg_vector_norm_47" -> "2963 clamp_min_47" [label="(1, 24, 64, 1)", style=solid]; -"2963 clamp_min_47" -> "2964 expand_as_47" [label="(1, 24, 64, 1)", style=solid]; -"2964 expand_as_47" -> "2965 div_47" [label="(1, 24, 64, 32)", style=solid]; -"2965 div_47" -> "2966 quantize_per_tensor_default_144" [label="(1, 24, 64, 32)", style=solid]; -"2966 quantize_per_tensor_default_144" -> "2967 dequantize_per_tensor_default_144" [label="(1, 24, 64, 32)", style=solid]; -"2967 dequantize_per_tensor_default_144" -> "2968 transpose_46" [label="(1, 24, 64, 32)", style=solid]; -"2968 transpose_46" -> "2969 matmul_46" [label="(1, 24, 32, 64)", style=solid]; -"2969 matmul_46" -> "2972 mul_47" [label="(1, 24, 64, 64)", style=solid]; -"2970 clamp_23" -> "2971 exp_23" [label="(24, 1, 1)", style=solid]; -"2971 exp_23" -> "2972 mul_47" [label="(24, 1, 1)", style=solid]; -"2972 mul_47" -> "2973 add_80" [label="(1, 24, 64, 64)", style=solid]; -"2973 add_80" -> "2974 softmax_23" [label="(1, 24, 64, 64)", style=solid]; -"2974 softmax_23" -> "2975 dropout_92" [label="(1, 24, 64, 64)", style=solid]; -"2975 dropout_92" -> "2976 matmul_47" [label="(1, 24, 64, 64)", style=solid]; -"2976 matmul_47" -> "2977 transpose_47" [label="(1, 24, 64, 32)", style=solid]; -"2977 transpose_47" -> "2978 reshape_105" [label="(1, 64, 24, 32)", style=solid]; -"2978 reshape_105" -> "2979 reshape_105_0_0_nncf_smooth_quant_0" [label="(1, 64, 768)", style=solid]; -"2979 reshape_105_0_0_nncf_smooth_quant_0" -> "2980 quantize_per_tensor_default_145" [label="(1, 64, 768)", style=solid]; -"2980 quantize_per_tensor_default_145" -> "2981 dequantize_per_tensor_default_145" [label="(1, 64, 768)", style=solid]; -"2981 dequantize_per_tensor_default_145" -> "2986 linear_144" [label="(1, 64, 768)", style=solid]; -"2982 linear_144_scale_0" -> "2985 dequantize_per_channel_default_145" [label="(768,)", style=solid]; -"2983 linear_144_zero_point_0" -> "2985 dequantize_per_channel_default_145" [label="(768,)", style=solid]; -"2984 _frozen_param145" -> "2985 dequantize_per_channel_default_145" [label="(768, 768)", style=solid]; -"2985 dequantize_per_channel_default_145" -> "2986 linear_144" [label="(768, 768)", style=solid]; -"2986 linear_144" -> "2987 dropout_93" [label="(1, 64, 768)", style=solid]; -"2987 dropout_93" -> "2988 view_128" [label="(1, 64, 768)", style=solid]; -"2988 view_128" -> "2989 permute_107" [label="(1, 1, 1, 8, 8, 768)", style=solid]; -"2989 permute_107" -> "2990 reshape_106" [label="(1, 1, 8, 1, 8, 768)", style=solid]; -"2990 reshape_106" -> "2991 slice_347" [label="(1, 8, 8, 768)", style=solid]; -"2991 slice_347" -> "2992 slice_348" [label="(1, 8, 8, 768)", style=solid]; -"2992 slice_348" -> "2993 slice_349" [label="(1, 7, 8, 768)", style=solid]; -"2993 slice_349" -> "2994 slice_350" [label="(1, 7, 7, 768)", style=solid]; -"2994 slice_350" -> "2995 contiguous_45" [label="(1, 7, 7, 768)", style=solid]; -"2995 contiguous_45" -> "2996 layer_norm_50" [label="(1, 7, 7, 768)", style=solid]; -"2996 layer_norm_50" -> "2997 add_81" [label="(1, 7, 7, 768)", style=solid]; -"2997 add_81" -> "2998 add_81_0_0_nncf_smooth_quant_0" [label="(1, 7, 7, 768)", style=solid]; -"2997 add_81" -> "3018 add_82" [label="(1, 7, 7, 768)", style=solid]; -"2998 add_81_0_0_nncf_smooth_quant_0" -> "2999 quantize_per_tensor_default_146" [label="(1, 7, 7, 768)", style=solid]; -"2999 quantize_per_tensor_default_146" -> "3000 dequantize_per_tensor_default_146" [label="(1, 7, 7, 768)", style=solid]; -"3000 dequantize_per_tensor_default_146" -> "3005 linear_145" [label="(1, 7, 7, 768)", style=solid]; -"3001 linear_145_scale_0" -> "3004 dequantize_per_channel_default_146" [label="(3072,)", style=solid]; -"3002 linear_145_zero_point_0" -> "3004 dequantize_per_channel_default_146" [label="(3072,)", style=solid]; -"3003 _frozen_param146" -> "3004 dequantize_per_channel_default_146" [label="(3072, 768)", style=solid]; -"3004 dequantize_per_channel_default_146" -> "3005 linear_145" [label="(3072, 768)", style=solid]; -"3005 linear_145" -> "3006 gelu_23" [label="(1, 7, 7, 3072)", style=solid]; -"3006 gelu_23" -> "3007 dropout_94" [label="(1, 7, 7, 3072)", style=solid]; -"3007 dropout_94" -> "3008 dropout_94_0_0_nncf_smooth_quant_0" [label="(1, 7, 7, 3072)", style=solid]; -"3008 dropout_94_0_0_nncf_smooth_quant_0" -> "3009 quantize_per_tensor_default_147" [label="(1, 7, 7, 3072)", style=solid]; -"3009 quantize_per_tensor_default_147" -> "3010 dequantize_per_tensor_default_147" [label="(1, 7, 7, 3072)", style=solid]; -"3010 dequantize_per_tensor_default_147" -> "3015 linear_146" [label="(1, 7, 7, 3072)", style=solid]; -"3011 linear_146_scale_0" -> "3014 dequantize_per_channel_default_147" [label="(768,)", style=solid]; -"3012 linear_146_zero_point_0" -> "3014 dequantize_per_channel_default_147" [label="(768,)", style=solid]; -"3013 _frozen_param147" -> "3014 dequantize_per_channel_default_147" [label="(768, 3072)", style=solid]; -"3014 dequantize_per_channel_default_147" -> "3015 linear_146" [label="(768, 3072)", style=solid]; -"3015 linear_146" -> "3016 dropout_95" [label="(1, 7, 7, 768)", style=solid]; -"3016 dropout_95" -> "3017 layer_norm_51" [label="(1, 7, 7, 768)", style=solid]; -"3017 layer_norm_51" -> "3018 add_82" [label="(1, 7, 7, 768)", style=solid]; -"3018 add_82" -> "3019 layer_norm_52" [label="(1, 7, 7, 768)", style=solid]; -"3019 layer_norm_52" -> "3020 permute_108" [label="(1, 7, 7, 768)", style=solid]; -"3020 permute_108" -> "3021 adaptive_avg_pool2d" [label="(1, 768, 7, 7)", style=solid]; -"3021 adaptive_avg_pool2d" -> "3022 flatten" [label="(1, 768, 1, 1)", style=solid]; -"3022 flatten" -> "3023 flatten_0_0_nncf_smooth_quant_0" [label="(1, 768)", style=solid]; -"3023 flatten_0_0_nncf_smooth_quant_0" -> "3024 quantize_per_tensor_default_148" [label="(1, 768)", style=solid]; -"3024 quantize_per_tensor_default_148" -> "3025 dequantize_per_tensor_default_148" [label="(1, 768)", style=solid]; -"3025 dequantize_per_tensor_default_148" -> "3030 linear_147" [label="(1, 768)", style=solid]; -"3026 linear_147_scale_0" -> "3029 dequantize_per_channel_default_148" [label="(1000,)", style=solid]; -"3027 linear_147_zero_point_0" -> "3029 dequantize_per_channel_default_148" [label="(1000,)", style=solid]; -"3028 _frozen_param148" -> "3029 dequantize_per_channel_default_148" [label="(1000, 768)", style=solid]; -"3029 dequantize_per_channel_default_148" -> "3030 linear_147" [label="(1000, 768)", style=solid]; -"3030 linear_147" -> "3031 output" [label="(1, 1000)", style=solid]; +"0 features_1_0_attn_proj_bias" [id=0, type=get_attr]; +"1 features_1_1_attn_proj_bias" [id=1, type=get_attr]; +"2 features_3_0_attn_proj_bias" [id=2, type=get_attr]; +"3 features_3_1_attn_proj_bias" [id=3, type=get_attr]; +"4 features_5_0_attn_proj_bias" [id=4, type=get_attr]; +"5 features_5_1_attn_proj_bias" [id=5, type=get_attr]; +"6 features_5_2_attn_proj_bias" [id=6, type=get_attr]; +"7 features_5_3_attn_proj_bias" [id=7, type=get_attr]; +"8 features_5_4_attn_proj_bias" [id=8, type=get_attr]; +"9 features_5_5_attn_proj_bias" [id=9, type=get_attr]; +"10 features_5_6_attn_proj_bias" [id=10, type=get_attr]; +"11 features_5_7_attn_proj_bias" [id=11, type=get_attr]; +"12 features_5_8_attn_proj_bias" [id=12, type=get_attr]; +"13 features_5_9_attn_proj_bias" [id=13, type=get_attr]; +"14 features_5_10_attn_proj_bias" [id=14, type=get_attr]; +"15 features_5_11_attn_proj_bias" [id=15, type=get_attr]; +"16 features_5_12_attn_proj_bias" [id=16, type=get_attr]; +"17 features_5_13_attn_proj_bias" [id=17, type=get_attr]; +"18 features_5_14_attn_proj_bias" [id=18, type=get_attr]; +"19 features_5_15_attn_proj_bias" [id=19, type=get_attr]; +"20 features_5_16_attn_proj_bias" [id=20, type=get_attr]; +"21 features_5_17_attn_proj_bias" [id=21, type=get_attr]; +"22 features_7_0_attn_proj_bias" [id=22, type=get_attr]; +"23 features_7_1_attn_proj_bias" [id=23, type=get_attr]; +"24 features_0_0_bias" [id=24, type=get_attr]; +"25 features_0_2_weight" [id=25, type=get_attr]; +"26 features_0_2_bias" [id=26, type=get_attr]; +"27 features_1_0_norm1_weight" [id=27, type=get_attr]; +"28 features_1_0_norm1_bias" [id=28, type=get_attr]; +"29 features_1_0_mlp_0_bias" [id=29, type=get_attr]; +"30 features_1_0_mlp_3_bias" [id=30, type=get_attr]; +"31 features_1_0_norm2_weight" [id=31, type=get_attr]; +"32 features_1_0_norm2_bias" [id=32, type=get_attr]; +"33 features_1_1_norm1_weight" [id=33, type=get_attr]; +"34 features_1_1_norm1_bias" [id=34, type=get_attr]; +"35 features_1_1_mlp_0_bias" [id=35, type=get_attr]; +"36 features_1_1_mlp_3_bias" [id=36, type=get_attr]; +"37 features_1_1_norm2_weight" [id=37, type=get_attr]; +"38 features_1_1_norm2_bias" [id=38, type=get_attr]; +"39 features_2_norm_weight" [id=39, type=get_attr]; +"40 features_2_norm_bias" [id=40, type=get_attr]; +"41 features_3_0_norm1_weight" [id=41, type=get_attr]; +"42 features_3_0_norm1_bias" [id=42, type=get_attr]; +"43 features_3_0_mlp_0_bias" [id=43, type=get_attr]; +"44 features_3_0_mlp_3_bias" [id=44, type=get_attr]; +"45 features_3_0_norm2_weight" [id=45, type=get_attr]; +"46 features_3_0_norm2_bias" [id=46, type=get_attr]; +"47 features_3_1_norm1_weight" [id=47, type=get_attr]; +"48 features_3_1_norm1_bias" [id=48, type=get_attr]; +"49 features_3_1_mlp_0_bias" [id=49, type=get_attr]; +"50 features_3_1_mlp_3_bias" [id=50, type=get_attr]; +"51 features_3_1_norm2_weight" [id=51, type=get_attr]; +"52 features_3_1_norm2_bias" [id=52, type=get_attr]; +"53 features_4_norm_weight" [id=53, type=get_attr]; +"54 features_4_norm_bias" [id=54, type=get_attr]; +"55 features_5_0_norm1_weight" [id=55, type=get_attr]; +"56 features_5_0_norm1_bias" [id=56, type=get_attr]; +"57 features_5_0_mlp_0_bias" [id=57, type=get_attr]; +"58 features_5_0_mlp_3_bias" [id=58, type=get_attr]; +"59 features_5_0_norm2_weight" [id=59, type=get_attr]; +"60 features_5_0_norm2_bias" [id=60, type=get_attr]; +"61 features_5_1_norm1_weight" [id=61, type=get_attr]; +"62 features_5_1_norm1_bias" [id=62, type=get_attr]; +"63 features_5_1_mlp_0_bias" [id=63, type=get_attr]; +"64 features_5_1_mlp_3_bias" [id=64, type=get_attr]; +"65 features_5_1_norm2_weight" [id=65, type=get_attr]; +"66 features_5_1_norm2_bias" [id=66, type=get_attr]; +"67 features_5_2_norm1_weight" [id=67, type=get_attr]; +"68 features_5_2_norm1_bias" [id=68, type=get_attr]; +"69 features_5_2_mlp_0_bias" [id=69, type=get_attr]; +"70 features_5_2_mlp_3_bias" [id=70, type=get_attr]; +"71 features_5_2_norm2_weight" [id=71, type=get_attr]; +"72 features_5_2_norm2_bias" [id=72, type=get_attr]; +"73 features_5_3_norm1_weight" [id=73, type=get_attr]; +"74 features_5_3_norm1_bias" [id=74, type=get_attr]; +"75 features_5_3_mlp_0_bias" [id=75, type=get_attr]; +"76 features_5_3_mlp_3_bias" [id=76, type=get_attr]; +"77 features_5_3_norm2_weight" [id=77, type=get_attr]; +"78 features_5_3_norm2_bias" [id=78, type=get_attr]; +"79 features_5_4_norm1_weight" [id=79, type=get_attr]; +"80 features_5_4_norm1_bias" [id=80, type=get_attr]; +"81 features_5_4_mlp_0_bias" [id=81, type=get_attr]; +"82 features_5_4_mlp_3_bias" [id=82, type=get_attr]; +"83 features_5_4_norm2_weight" [id=83, type=get_attr]; +"84 features_5_4_norm2_bias" [id=84, type=get_attr]; +"85 features_5_5_norm1_weight" [id=85, type=get_attr]; +"86 features_5_5_norm1_bias" [id=86, type=get_attr]; +"87 features_5_5_mlp_0_bias" [id=87, type=get_attr]; +"88 features_5_5_mlp_3_bias" [id=88, type=get_attr]; +"89 features_5_5_norm2_weight" [id=89, type=get_attr]; +"90 features_5_5_norm2_bias" [id=90, type=get_attr]; +"91 features_5_6_norm1_weight" [id=91, type=get_attr]; +"92 features_5_6_norm1_bias" [id=92, type=get_attr]; +"93 features_5_6_mlp_0_bias" [id=93, type=get_attr]; +"94 features_5_6_mlp_3_bias" [id=94, type=get_attr]; +"95 features_5_6_norm2_weight" [id=95, type=get_attr]; +"96 features_5_6_norm2_bias" [id=96, type=get_attr]; +"97 features_5_7_norm1_weight" [id=97, type=get_attr]; +"98 features_5_7_norm1_bias" [id=98, type=get_attr]; +"99 features_5_7_mlp_0_bias" [id=99, type=get_attr]; +"100 features_5_7_mlp_3_bias" [id=100, type=get_attr]; +"101 features_5_7_norm2_weight" [id=101, type=get_attr]; +"102 features_5_7_norm2_bias" [id=102, type=get_attr]; +"103 features_5_8_norm1_weight" [id=103, type=get_attr]; +"104 features_5_8_norm1_bias" [id=104, type=get_attr]; +"105 features_5_8_mlp_0_bias" [id=105, type=get_attr]; +"106 features_5_8_mlp_3_bias" [id=106, type=get_attr]; +"107 features_5_8_norm2_weight" [id=107, type=get_attr]; +"108 features_5_8_norm2_bias" [id=108, type=get_attr]; +"109 features_5_9_norm1_weight" [id=109, type=get_attr]; +"110 features_5_9_norm1_bias" [id=110, type=get_attr]; +"111 features_5_9_mlp_0_bias" [id=111, type=get_attr]; +"112 features_5_9_mlp_3_bias" [id=112, type=get_attr]; +"113 features_5_9_norm2_weight" [id=113, type=get_attr]; +"114 features_5_9_norm2_bias" [id=114, type=get_attr]; +"115 features_5_10_norm1_weight" [id=115, type=get_attr]; +"116 features_5_10_norm1_bias" [id=116, type=get_attr]; +"117 features_5_10_mlp_0_bias" [id=117, type=get_attr]; +"118 features_5_10_mlp_3_bias" [id=118, type=get_attr]; +"119 features_5_10_norm2_weight" [id=119, type=get_attr]; +"120 features_5_10_norm2_bias" [id=120, type=get_attr]; +"121 features_5_11_norm1_weight" [id=121, type=get_attr]; +"122 features_5_11_norm1_bias" [id=122, type=get_attr]; +"123 features_5_11_mlp_0_bias" [id=123, type=get_attr]; +"124 features_5_11_mlp_3_bias" [id=124, type=get_attr]; +"125 features_5_11_norm2_weight" [id=125, type=get_attr]; +"126 features_5_11_norm2_bias" [id=126, type=get_attr]; +"127 features_5_12_norm1_weight" [id=127, type=get_attr]; +"128 features_5_12_norm1_bias" [id=128, type=get_attr]; +"129 features_5_12_mlp_0_bias" [id=129, type=get_attr]; +"130 features_5_12_mlp_3_bias" [id=130, type=get_attr]; +"131 features_5_12_norm2_weight" [id=131, type=get_attr]; +"132 features_5_12_norm2_bias" [id=132, type=get_attr]; +"133 features_5_13_norm1_weight" [id=133, type=get_attr]; +"134 features_5_13_norm1_bias" [id=134, type=get_attr]; +"135 features_5_13_mlp_0_bias" [id=135, type=get_attr]; +"136 features_5_13_mlp_3_bias" [id=136, type=get_attr]; +"137 features_5_13_norm2_weight" [id=137, type=get_attr]; +"138 features_5_13_norm2_bias" [id=138, type=get_attr]; +"139 features_5_14_norm1_weight" [id=139, type=get_attr]; +"140 features_5_14_norm1_bias" [id=140, type=get_attr]; +"141 features_5_14_mlp_0_bias" [id=141, type=get_attr]; +"142 features_5_14_mlp_3_bias" [id=142, type=get_attr]; +"143 features_5_14_norm2_weight" [id=143, type=get_attr]; +"144 features_5_14_norm2_bias" [id=144, type=get_attr]; +"145 features_5_15_norm1_weight" [id=145, type=get_attr]; +"146 features_5_15_norm1_bias" [id=146, type=get_attr]; +"147 features_5_15_mlp_0_bias" [id=147, type=get_attr]; +"148 features_5_15_mlp_3_bias" [id=148, type=get_attr]; +"149 features_5_15_norm2_weight" [id=149, type=get_attr]; +"150 features_5_15_norm2_bias" [id=150, type=get_attr]; +"151 features_5_16_norm1_weight" [id=151, type=get_attr]; +"152 features_5_16_norm1_bias" [id=152, type=get_attr]; +"153 features_5_16_mlp_0_bias" [id=153, type=get_attr]; +"154 features_5_16_mlp_3_bias" [id=154, type=get_attr]; +"155 features_5_16_norm2_weight" [id=155, type=get_attr]; +"156 features_5_16_norm2_bias" [id=156, type=get_attr]; +"157 features_5_17_norm1_weight" [id=157, type=get_attr]; +"158 features_5_17_norm1_bias" [id=158, type=get_attr]; +"159 features_5_17_mlp_0_bias" [id=159, type=get_attr]; +"160 features_5_17_mlp_3_bias" [id=160, type=get_attr]; +"161 features_5_17_norm2_weight" [id=161, type=get_attr]; +"162 features_5_17_norm2_bias" [id=162, type=get_attr]; +"163 features_6_norm_weight" [id=163, type=get_attr]; +"164 features_6_norm_bias" [id=164, type=get_attr]; +"165 features_7_0_norm1_weight" [id=165, type=get_attr]; +"166 features_7_0_norm1_bias" [id=166, type=get_attr]; +"167 features_7_0_mlp_0_bias" [id=167, type=get_attr]; +"168 features_7_0_mlp_3_bias" [id=168, type=get_attr]; +"169 features_7_0_norm2_weight" [id=169, type=get_attr]; +"170 features_7_0_norm2_bias" [id=170, type=get_attr]; +"171 features_7_1_norm1_weight" [id=171, type=get_attr]; +"172 features_7_1_norm1_bias" [id=172, type=get_attr]; +"173 features_7_1_mlp_0_bias" [id=173, type=get_attr]; +"174 features_7_1_mlp_3_bias" [id=174, type=get_attr]; +"175 features_7_1_norm2_weight" [id=175, type=get_attr]; +"176 features_7_1_norm2_bias" [id=176, type=get_attr]; +"177 norm_weight" [id=177, type=get_attr]; +"178 norm_bias" [id=178, type=get_attr]; +"179 head_bias" [id=179, type=get_attr]; +"180 x" [id=180, type=input]; +"181 quantize_per_tensor_default" [id=181, type=quantize_per_tensor]; +"182 dequantize_per_tensor_default" [id=182, type=dequantize_per_tensor]; +"183 conv2d_scale_0" [id=183, type=get_attr]; +"184 conv2d_zero_point_0" [id=184, type=get_attr]; +"185 _frozen_param2_1" [id=185, type=get_attr]; +"186 dequantize_per_channel_default" [id=186, type=dequantize_per_channel]; +"187 conv2d" [id=187, type=conv2d]; +"188 permute" [id=188, type=permute]; +"189 layer_norm" [id=189, type=layer_norm]; +"190 _frozen_param0" [id=190, type=get_attr]; +"191 pad" [id=191, type=pad]; +"192 view_2" [id=192, type=view]; +"193 permute_2" [id=193, type=permute]; +"194 reshape" [id=194, type=reshape]; +"195 _frozen_param1" [id=195, type=get_attr]; +"196 reshape_0_0_nncf_smooth_quant_0" [id=196, type=call_module]; +"197 quantize_per_tensor_default_1" [id=197, type=quantize_per_tensor]; +"198 dequantize_per_tensor_default_1" [id=198, type=dequantize_per_tensor]; +"199 linear_2_scale_0" [id=199, type=get_attr]; +"200 linear_2_zero_point_0" [id=200, type=get_attr]; +"201 _frozen_param6_1" [id=201, type=get_attr]; +"202 dequantize_per_channel_default_1" [id=202, type=dequantize_per_channel]; +"203 linear_2" [id=203, type=linear]; +"204 reshape_1" [id=204, type=reshape]; +"205 permute_3" [id=205, type=permute]; +"206 select" [id=206, type=select]; +"207 select_1" [id=207, type=select]; +"208 select_2" [id=208, type=select]; +"209 linalg_vector_norm" [id=209, type=linalg_vector_norm]; +"210 clamp_min" [id=210, type=clamp_min]; +"211 expand_as" [id=211, type=expand_as]; +"212 div" [id=212, type=div]; +"213 quantize_per_tensor_default_2" [id=213, type=quantize_per_tensor]; +"214 dequantize_per_tensor_default_2" [id=214, type=dequantize_per_tensor]; +"215 linalg_vector_norm_1" [id=215, type=linalg_vector_norm]; +"216 clamp_min_1" [id=216, type=clamp_min]; +"217 expand_as_1" [id=217, type=expand_as]; +"218 div_1" [id=218, type=div]; +"219 quantize_per_tensor_default_3" [id=219, type=quantize_per_tensor]; +"220 dequantize_per_tensor_default_3" [id=220, type=dequantize_per_tensor]; +"221 transpose" [id=221, type=transpose]; +"222 matmul" [id=222, type=matmul]; +"223 _frozen_param3" [id=223, type=get_attr]; +"224 mul_1" [id=224, type=mul]; +"225 add" [id=225, type=add]; +"226 softmax" [id=226, type=softmax]; +"227 dropout" [id=227, type=dropout]; +"228 matmul_1" [id=228, type=matmul]; +"229 transpose_1" [id=229, type=transpose]; +"230 reshape_2" [id=230, type=reshape]; +"231 reshape_2_0_0_nncf_smooth_quant_0" [id=231, type=call_module]; +"232 quantize_per_tensor_default_4" [id=232, type=quantize_per_tensor]; +"233 dequantize_per_tensor_default_4" [id=233, type=dequantize_per_tensor]; +"234 linear_3_scale_0" [id=234, type=get_attr]; +"235 linear_3_zero_point_0" [id=235, type=get_attr]; +"236 _frozen_param8_1" [id=236, type=get_attr]; +"237 dequantize_per_channel_default_2" [id=237, type=dequantize_per_channel]; +"238 linear_3" [id=238, type=linear]; +"239 dropout_1" [id=239, type=dropout]; +"240 view_3" [id=240, type=view]; +"241 permute_4" [id=241, type=permute]; +"242 reshape_3" [id=242, type=reshape]; +"243 slice_2" [id=243, type=slice]; +"244 slice_3" [id=244, type=slice]; +"245 layer_norm_1" [id=245, type=layer_norm]; +"246 add_1" [id=246, type=add]; +"247 add_1_0_0_nncf_smooth_quant_0" [id=247, type=call_module]; +"248 quantize_per_tensor_default_5" [id=248, type=quantize_per_tensor]; +"249 dequantize_per_tensor_default_5" [id=249, type=dequantize_per_tensor]; +"250 linear_4_scale_0" [id=250, type=get_attr]; +"251 linear_4_zero_point_0" [id=251, type=get_attr]; +"252 _frozen_param9_1" [id=252, type=get_attr]; +"253 dequantize_per_channel_default_3" [id=253, type=dequantize_per_channel]; +"254 linear_4" [id=254, type=linear]; +"255 gelu" [id=255, type=gelu]; +"256 dropout_2" [id=256, type=dropout]; +"257 dropout_2_0_0_nncf_smooth_quant_0" [id=257, type=call_module]; +"258 quantize_per_tensor_default_6" [id=258, type=quantize_per_tensor]; +"259 dequantize_per_tensor_default_6" [id=259, type=dequantize_per_tensor]; +"260 linear_5_scale_0" [id=260, type=get_attr]; +"261 linear_5_zero_point_0" [id=261, type=get_attr]; +"262 _frozen_param10_1" [id=262, type=get_attr]; +"263 dequantize_per_channel_default_4" [id=263, type=dequantize_per_channel]; +"264 linear_5" [id=264, type=linear]; +"265 dropout_3" [id=265, type=dropout]; +"266 layer_norm_2" [id=266, type=layer_norm]; +"267 add_2" [id=267, type=add]; +"268 _frozen_param4" [id=268, type=get_attr]; +"269 pad_1" [id=269, type=pad]; +"270 roll" [id=270, type=roll]; +"271 view_6" [id=271, type=view]; +"272 permute_6" [id=272, type=permute]; +"273 reshape_4" [id=273, type=reshape]; +"274 _frozen_param5" [id=274, type=get_attr]; +"275 reshape_4_0_0_nncf_smooth_quant_0" [id=275, type=call_module]; +"276 quantize_per_tensor_default_7" [id=276, type=quantize_per_tensor]; +"277 dequantize_per_tensor_default_7" [id=277, type=dequantize_per_tensor]; +"278 linear_8_scale_0" [id=278, type=get_attr]; +"279 linear_8_zero_point_0" [id=279, type=get_attr]; +"280 _frozen_param11_1" [id=280, type=get_attr]; +"281 dequantize_per_channel_default_5" [id=281, type=dequantize_per_channel]; +"282 linear_8" [id=282, type=linear]; +"283 reshape_5" [id=283, type=reshape]; +"284 permute_7" [id=284, type=permute]; +"285 select_3" [id=285, type=select]; +"286 select_4" [id=286, type=select]; +"287 select_5" [id=287, type=select]; +"288 linalg_vector_norm_2" [id=288, type=linalg_vector_norm]; +"289 clamp_min_2" [id=289, type=clamp_min]; +"290 expand_as_2" [id=290, type=expand_as]; +"291 div_2" [id=291, type=div]; +"292 quantize_per_tensor_default_8" [id=292, type=quantize_per_tensor]; +"293 dequantize_per_tensor_default_8" [id=293, type=dequantize_per_tensor]; +"294 linalg_vector_norm_3" [id=294, type=linalg_vector_norm]; +"295 clamp_min_3" [id=295, type=clamp_min]; +"296 expand_as_3" [id=296, type=expand_as]; +"297 div_3" [id=297, type=div]; +"298 quantize_per_tensor_default_9" [id=298, type=quantize_per_tensor]; +"299 dequantize_per_tensor_default_9" [id=299, type=dequantize_per_tensor]; +"300 transpose_2" [id=300, type=transpose]; +"301 matmul_2" [id=301, type=matmul]; +"302 _frozen_param7" [id=302, type=get_attr]; +"303 mul_3" [id=303, type=mul]; +"304 add_3" [id=304, type=add]; +"305 new_zeros" [id=305, type=new_zeros]; +"306 view_7" [id=306, type=view]; +"307 permute_8" [id=307, type=permute]; +"308 reshape_6" [id=308, type=reshape]; +"309 unsqueeze_2" [id=309, type=unsqueeze]; +"310 unsqueeze_3" [id=310, type=unsqueeze]; +"311 sub" [id=311, type=sub]; +"312 ne" [id=312, type=ne]; +"313 masked_fill" [id=313, type=masked_fill]; +"314 eq" [id=314, type=eq]; +"315 masked_fill_1" [id=315, type=masked_fill]; +"316 view_8" [id=316, type=view]; +"317 unsqueeze_4" [id=317, type=unsqueeze]; +"318 unsqueeze_5" [id=318, type=unsqueeze]; +"319 add_4" [id=319, type=add]; +"320 view_9" [id=320, type=view]; +"321 softmax_1" [id=321, type=softmax]; +"322 dropout_4" [id=322, type=dropout]; +"323 matmul_3" [id=323, type=matmul]; +"324 transpose_3" [id=324, type=transpose]; +"325 reshape_7" [id=325, type=reshape]; +"326 reshape_7_0_0_nncf_smooth_quant_0" [id=326, type=call_module]; +"327 quantize_per_tensor_default_10" [id=327, type=quantize_per_tensor]; +"328 dequantize_per_tensor_default_10" [id=328, type=dequantize_per_tensor]; +"329 linear_9_scale_0" [id=329, type=get_attr]; +"330 linear_9_zero_point_0" [id=330, type=get_attr]; +"331 _frozen_param12_1" [id=331, type=get_attr]; +"332 dequantize_per_channel_default_6" [id=332, type=dequantize_per_channel]; +"333 linear_9" [id=333, type=linear]; +"334 dropout_5" [id=334, type=dropout]; +"335 view_10" [id=335, type=view]; +"336 permute_9" [id=336, type=permute]; +"337 reshape_8" [id=337, type=reshape]; +"338 roll_1" [id=338, type=roll]; +"339 slice_23" [id=339, type=slice]; +"340 slice_24" [id=340, type=slice]; +"341 layer_norm_3" [id=341, type=layer_norm]; +"342 add_5" [id=342, type=add]; +"343 add_5_0_0_nncf_smooth_quant_0" [id=343, type=call_module]; +"344 quantize_per_tensor_default_11" [id=344, type=quantize_per_tensor]; +"345 dequantize_per_tensor_default_11" [id=345, type=dequantize_per_tensor]; +"346 linear_10_scale_0" [id=346, type=get_attr]; +"347 linear_10_zero_point_0" [id=347, type=get_attr]; +"348 _frozen_param13_1" [id=348, type=get_attr]; +"349 dequantize_per_channel_default_7" [id=349, type=dequantize_per_channel]; +"350 linear_10" [id=350, type=linear]; +"351 gelu_1" [id=351, type=gelu]; +"352 dropout_6" [id=352, type=dropout]; +"353 dropout_6_0_0_nncf_smooth_quant_0" [id=353, type=call_module]; +"354 quantize_per_tensor_default_12" [id=354, type=quantize_per_tensor]; +"355 dequantize_per_tensor_default_12" [id=355, type=dequantize_per_tensor]; +"356 linear_11_scale_0" [id=356, type=get_attr]; +"357 linear_11_zero_point_0" [id=357, type=get_attr]; +"358 _frozen_param14_1" [id=358, type=get_attr]; +"359 dequantize_per_channel_default_8" [id=359, type=dequantize_per_channel]; +"360 linear_11" [id=360, type=linear]; +"361 dropout_7" [id=361, type=dropout]; +"362 layer_norm_4" [id=362, type=layer_norm]; +"363 add_6" [id=363, type=add]; +"364 pad_2" [id=364, type=pad]; +"365 slice_25" [id=365, type=slice]; +"366 slice_26" [id=366, type=slice]; +"367 slice_27" [id=367, type=slice]; +"368 slice_28" [id=368, type=slice]; +"369 slice_29" [id=369, type=slice]; +"370 slice_30" [id=370, type=slice]; +"371 slice_31" [id=371, type=slice]; +"372 slice_32" [id=372, type=slice]; +"373 slice_33" [id=373, type=slice]; +"374 slice_34" [id=374, type=slice]; +"375 slice_35" [id=375, type=slice]; +"376 slice_36" [id=376, type=slice]; +"377 cat" [id=377, type=cat]; +"378 cat_0_0_nncf_smooth_quant_0" [id=378, type=call_module]; +"379 quantize_per_tensor_default_13" [id=379, type=quantize_per_tensor]; +"380 dequantize_per_tensor_default_13" [id=380, type=dequantize_per_tensor]; +"381 linear_12_scale_0" [id=381, type=get_attr]; +"382 linear_12_zero_point_0" [id=382, type=get_attr]; +"383 _frozen_param15_1" [id=383, type=get_attr]; +"384 dequantize_per_channel_default_9" [id=384, type=dequantize_per_channel]; +"385 linear_12" [id=385, type=linear]; +"386 layer_norm_5" [id=386, type=layer_norm]; +"387 _frozen_param17" [id=387, type=get_attr]; +"388 pad_3" [id=388, type=pad]; +"389 view_13" [id=389, type=view]; +"390 permute_11" [id=390, type=permute]; +"391 reshape_9" [id=391, type=reshape]; +"392 _frozen_param18" [id=392, type=get_attr]; +"393 reshape_9_0_0_nncf_smooth_quant_0" [id=393, type=call_module]; +"394 quantize_per_tensor_default_14" [id=394, type=quantize_per_tensor]; +"395 dequantize_per_tensor_default_14" [id=395, type=dequantize_per_tensor]; +"396 linear_15_scale_0" [id=396, type=get_attr]; +"397 linear_15_zero_point_0" [id=397, type=get_attr]; +"398 _frozen_param16_1" [id=398, type=get_attr]; +"399 dequantize_per_channel_default_10" [id=399, type=dequantize_per_channel]; +"400 linear_15" [id=400, type=linear]; +"401 reshape_10" [id=401, type=reshape]; +"402 permute_12" [id=402, type=permute]; +"403 select_6" [id=403, type=select]; +"404 select_7" [id=404, type=select]; +"405 select_8" [id=405, type=select]; +"406 linalg_vector_norm_4" [id=406, type=linalg_vector_norm]; +"407 clamp_min_4" [id=407, type=clamp_min]; +"408 expand_as_4" [id=408, type=expand_as]; +"409 div_4" [id=409, type=div]; +"410 quantize_per_tensor_default_15" [id=410, type=quantize_per_tensor]; +"411 dequantize_per_tensor_default_15" [id=411, type=dequantize_per_tensor]; +"412 linalg_vector_norm_5" [id=412, type=linalg_vector_norm]; +"413 clamp_min_5" [id=413, type=clamp_min]; +"414 expand_as_5" [id=414, type=expand_as]; +"415 div_5" [id=415, type=div]; +"416 quantize_per_tensor_default_16" [id=416, type=quantize_per_tensor]; +"417 dequantize_per_tensor_default_16" [id=417, type=dequantize_per_tensor]; +"418 transpose_4" [id=418, type=transpose]; +"419 matmul_4" [id=419, type=matmul]; +"420 _frozen_param20" [id=420, type=get_attr]; +"421 mul_5" [id=421, type=mul]; +"422 add_7" [id=422, type=add]; +"423 softmax_2" [id=423, type=softmax]; +"424 dropout_8" [id=424, type=dropout]; +"425 matmul_5" [id=425, type=matmul]; +"426 transpose_5" [id=426, type=transpose]; +"427 reshape_11" [id=427, type=reshape]; +"428 reshape_11_0_0_nncf_smooth_quant_0" [id=428, type=call_module]; +"429 quantize_per_tensor_default_17" [id=429, type=quantize_per_tensor]; +"430 dequantize_per_tensor_default_17" [id=430, type=dequantize_per_tensor]; +"431 linear_16_scale_0" [id=431, type=get_attr]; +"432 linear_16_zero_point_0" [id=432, type=get_attr]; +"433 _frozen_param19_1" [id=433, type=get_attr]; +"434 dequantize_per_channel_default_11" [id=434, type=dequantize_per_channel]; +"435 linear_16" [id=435, type=linear]; +"436 dropout_9" [id=436, type=dropout]; +"437 view_14" [id=437, type=view]; +"438 permute_13" [id=438, type=permute]; +"439 reshape_12" [id=439, type=reshape]; +"440 slice_38" [id=440, type=slice]; +"441 slice_39" [id=441, type=slice]; +"442 slice_40" [id=442, type=slice]; +"443 slice_41" [id=443, type=slice]; +"444 contiguous_3" [id=444, type=contiguous]; +"445 layer_norm_6" [id=445, type=layer_norm]; +"446 add_8" [id=446, type=add]; +"447 add_8_0_0_nncf_smooth_quant_0" [id=447, type=call_module]; +"448 quantize_per_tensor_default_18" [id=448, type=quantize_per_tensor]; +"449 dequantize_per_tensor_default_18" [id=449, type=dequantize_per_tensor]; +"450 linear_17_scale_0" [id=450, type=get_attr]; +"451 linear_17_zero_point_0" [id=451, type=get_attr]; +"452 _frozen_param23_1" [id=452, type=get_attr]; +"453 dequantize_per_channel_default_12" [id=453, type=dequantize_per_channel]; +"454 linear_17" [id=454, type=linear]; +"455 gelu_2" [id=455, type=gelu]; +"456 dropout_10" [id=456, type=dropout]; +"457 dropout_10_0_0_nncf_smooth_quant_0" [id=457, type=call_module]; +"458 quantize_per_tensor_default_19" [id=458, type=quantize_per_tensor]; +"459 dequantize_per_tensor_default_19" [id=459, type=dequantize_per_tensor]; +"460 linear_18_scale_0" [id=460, type=get_attr]; +"461 linear_18_zero_point_0" [id=461, type=get_attr]; +"462 _frozen_param25_1" [id=462, type=get_attr]; +"463 dequantize_per_channel_default_13" [id=463, type=dequantize_per_channel]; +"464 linear_18" [id=464, type=linear]; +"465 dropout_11" [id=465, type=dropout]; +"466 layer_norm_7" [id=466, type=layer_norm]; +"467 add_9" [id=467, type=add]; +"468 _frozen_param21" [id=468, type=get_attr]; +"469 pad_4" [id=469, type=pad]; +"470 roll_2" [id=470, type=roll]; +"471 view_17" [id=471, type=view]; +"472 permute_15" [id=472, type=permute]; +"473 reshape_13" [id=473, type=reshape]; +"474 _frozen_param22" [id=474, type=get_attr]; +"475 reshape_13_0_0_nncf_smooth_quant_0" [id=475, type=call_module]; +"476 quantize_per_tensor_default_20" [id=476, type=quantize_per_tensor]; +"477 dequantize_per_tensor_default_20" [id=477, type=dequantize_per_tensor]; +"478 linear_21_scale_0" [id=478, type=get_attr]; +"479 linear_21_zero_point_0" [id=479, type=get_attr]; +"480 _frozen_param26_1" [id=480, type=get_attr]; +"481 dequantize_per_channel_default_14" [id=481, type=dequantize_per_channel]; +"482 linear_21" [id=482, type=linear]; +"483 reshape_14" [id=483, type=reshape]; +"484 permute_16" [id=484, type=permute]; +"485 select_9" [id=485, type=select]; +"486 select_10" [id=486, type=select]; +"487 select_11" [id=487, type=select]; +"488 linalg_vector_norm_6" [id=488, type=linalg_vector_norm]; +"489 clamp_min_6" [id=489, type=clamp_min]; +"490 expand_as_6" [id=490, type=expand_as]; +"491 div_6" [id=491, type=div]; +"492 quantize_per_tensor_default_21" [id=492, type=quantize_per_tensor]; +"493 dequantize_per_tensor_default_21" [id=493, type=dequantize_per_tensor]; +"494 linalg_vector_norm_7" [id=494, type=linalg_vector_norm]; +"495 clamp_min_7" [id=495, type=clamp_min]; +"496 expand_as_7" [id=496, type=expand_as]; +"497 div_7" [id=497, type=div]; +"498 quantize_per_tensor_default_22" [id=498, type=quantize_per_tensor]; +"499 dequantize_per_tensor_default_22" [id=499, type=dequantize_per_tensor]; +"500 transpose_6" [id=500, type=transpose]; +"501 matmul_6" [id=501, type=matmul]; +"502 _frozen_param24" [id=502, type=get_attr]; +"503 mul_7" [id=503, type=mul]; +"504 add_10" [id=504, type=add]; +"505 new_zeros_1" [id=505, type=new_zeros]; +"506 view_18" [id=506, type=view]; +"507 permute_17" [id=507, type=permute]; +"508 reshape_15" [id=508, type=reshape]; +"509 unsqueeze_8" [id=509, type=unsqueeze]; +"510 unsqueeze_9" [id=510, type=unsqueeze]; +"511 sub_1" [id=511, type=sub]; +"512 ne_1" [id=512, type=ne]; +"513 masked_fill_2" [id=513, type=masked_fill]; +"514 eq_1" [id=514, type=eq]; +"515 masked_fill_3" [id=515, type=masked_fill]; +"516 view_19" [id=516, type=view]; +"517 unsqueeze_10" [id=517, type=unsqueeze]; +"518 unsqueeze_11" [id=518, type=unsqueeze]; +"519 add_11" [id=519, type=add]; +"520 view_20" [id=520, type=view]; +"521 softmax_3" [id=521, type=softmax]; +"522 dropout_12" [id=522, type=dropout]; +"523 matmul_7" [id=523, type=matmul]; +"524 transpose_7" [id=524, type=transpose]; +"525 reshape_16" [id=525, type=reshape]; +"526 reshape_16_0_0_nncf_smooth_quant_0" [id=526, type=call_module]; +"527 quantize_per_tensor_default_23" [id=527, type=quantize_per_tensor]; +"528 dequantize_per_tensor_default_23" [id=528, type=dequantize_per_tensor]; +"529 linear_22_scale_0" [id=529, type=get_attr]; +"530 linear_22_zero_point_0" [id=530, type=get_attr]; +"531 _frozen_param27_1" [id=531, type=get_attr]; +"532 dequantize_per_channel_default_15" [id=532, type=dequantize_per_channel]; +"533 linear_22" [id=533, type=linear]; +"534 dropout_13" [id=534, type=dropout]; +"535 view_21" [id=535, type=view]; +"536 permute_18" [id=536, type=permute]; +"537 reshape_17" [id=537, type=reshape]; +"538 roll_3" [id=538, type=roll]; +"539 slice_61" [id=539, type=slice]; +"540 slice_62" [id=540, type=slice]; +"541 slice_63" [id=541, type=slice]; +"542 slice_64" [id=542, type=slice]; +"543 contiguous_5" [id=543, type=contiguous]; +"544 layer_norm_8" [id=544, type=layer_norm]; +"545 add_12" [id=545, type=add]; +"546 add_12_0_0_nncf_smooth_quant_0" [id=546, type=call_module]; +"547 quantize_per_tensor_default_24" [id=547, type=quantize_per_tensor]; +"548 dequantize_per_tensor_default_24" [id=548, type=dequantize_per_tensor]; +"549 linear_23_scale_0" [id=549, type=get_attr]; +"550 linear_23_zero_point_0" [id=550, type=get_attr]; +"551 _frozen_param28_1" [id=551, type=get_attr]; +"552 dequantize_per_channel_default_16" [id=552, type=dequantize_per_channel]; +"553 linear_23" [id=553, type=linear]; +"554 gelu_3" [id=554, type=gelu]; +"555 dropout_14" [id=555, type=dropout]; +"556 dropout_14_0_0_nncf_smooth_quant_0" [id=556, type=call_module]; +"557 quantize_per_tensor_default_25" [id=557, type=quantize_per_tensor]; +"558 dequantize_per_tensor_default_25" [id=558, type=dequantize_per_tensor]; +"559 linear_24_scale_0" [id=559, type=get_attr]; +"560 linear_24_zero_point_0" [id=560, type=get_attr]; +"561 _frozen_param29_1" [id=561, type=get_attr]; +"562 dequantize_per_channel_default_17" [id=562, type=dequantize_per_channel]; +"563 linear_24" [id=563, type=linear]; +"564 dropout_15" [id=564, type=dropout]; +"565 layer_norm_9" [id=565, type=layer_norm]; +"566 add_13" [id=566, type=add]; +"567 pad_5" [id=567, type=pad]; +"568 slice_65" [id=568, type=slice]; +"569 slice_66" [id=569, type=slice]; +"570 slice_67" [id=570, type=slice]; +"571 slice_68" [id=571, type=slice]; +"572 slice_69" [id=572, type=slice]; +"573 slice_70" [id=573, type=slice]; +"574 slice_71" [id=574, type=slice]; +"575 slice_72" [id=575, type=slice]; +"576 slice_73" [id=576, type=slice]; +"577 slice_74" [id=577, type=slice]; +"578 slice_75" [id=578, type=slice]; +"579 slice_76" [id=579, type=slice]; +"580 cat_1" [id=580, type=cat]; +"581 cat_1_0_0_nncf_smooth_quant_0" [id=581, type=call_module]; +"582 quantize_per_tensor_default_26" [id=582, type=quantize_per_tensor]; +"583 dequantize_per_tensor_default_26" [id=583, type=dequantize_per_tensor]; +"584 linear_25_scale_0" [id=584, type=get_attr]; +"585 linear_25_zero_point_0" [id=585, type=get_attr]; +"586 _frozen_param30_1" [id=586, type=get_attr]; +"587 dequantize_per_channel_default_18" [id=587, type=dequantize_per_channel]; +"588 linear_25" [id=588, type=linear]; +"589 layer_norm_10" [id=589, type=layer_norm]; +"590 _frozen_param34" [id=590, type=get_attr]; +"591 pad_6" [id=591, type=pad]; +"592 view_24" [id=592, type=view]; +"593 permute_20" [id=593, type=permute]; +"594 reshape_18" [id=594, type=reshape]; +"595 _frozen_param35" [id=595, type=get_attr]; +"596 reshape_18_0_0_nncf_smooth_quant_0" [id=596, type=call_module]; +"597 quantize_per_tensor_default_27" [id=597, type=quantize_per_tensor]; +"598 dequantize_per_tensor_default_27" [id=598, type=dequantize_per_tensor]; +"599 linear_28_scale_0" [id=599, type=get_attr]; +"600 linear_28_zero_point_0" [id=600, type=get_attr]; +"601 _frozen_param31_1" [id=601, type=get_attr]; +"602 dequantize_per_channel_default_19" [id=602, type=dequantize_per_channel]; +"603 linear_28" [id=603, type=linear]; +"604 reshape_19" [id=604, type=reshape]; +"605 permute_21" [id=605, type=permute]; +"606 select_12" [id=606, type=select]; +"607 select_13" [id=607, type=select]; +"608 select_14" [id=608, type=select]; +"609 linalg_vector_norm_8" [id=609, type=linalg_vector_norm]; +"610 clamp_min_8" [id=610, type=clamp_min]; +"611 expand_as_8" [id=611, type=expand_as]; +"612 div_8" [id=612, type=div]; +"613 quantize_per_tensor_default_28" [id=613, type=quantize_per_tensor]; +"614 dequantize_per_tensor_default_28" [id=614, type=dequantize_per_tensor]; +"615 linalg_vector_norm_9" [id=615, type=linalg_vector_norm]; +"616 clamp_min_9" [id=616, type=clamp_min]; +"617 expand_as_9" [id=617, type=expand_as]; +"618 div_9" [id=618, type=div]; +"619 quantize_per_tensor_default_29" [id=619, type=quantize_per_tensor]; +"620 dequantize_per_tensor_default_29" [id=620, type=dequantize_per_tensor]; +"621 transpose_8" [id=621, type=transpose]; +"622 matmul_8" [id=622, type=matmul]; +"623 _frozen_param37" [id=623, type=get_attr]; +"624 mul_9" [id=624, type=mul]; +"625 add_14" [id=625, type=add]; +"626 softmax_4" [id=626, type=softmax]; +"627 dropout_16" [id=627, type=dropout]; +"628 matmul_9" [id=628, type=matmul]; +"629 transpose_9" [id=629, type=transpose]; +"630 reshape_20" [id=630, type=reshape]; +"631 reshape_20_0_0_nncf_smooth_quant_0" [id=631, type=call_module]; +"632 quantize_per_tensor_default_30" [id=632, type=quantize_per_tensor]; +"633 dequantize_per_tensor_default_30" [id=633, type=dequantize_per_tensor]; +"634 linear_29_scale_0" [id=634, type=get_attr]; +"635 linear_29_zero_point_0" [id=635, type=get_attr]; +"636 _frozen_param32_1" [id=636, type=get_attr]; +"637 dequantize_per_channel_default_20" [id=637, type=dequantize_per_channel]; +"638 linear_29" [id=638, type=linear]; +"639 dropout_17" [id=639, type=dropout]; +"640 view_25" [id=640, type=view]; +"641 permute_22" [id=641, type=permute]; +"642 reshape_21" [id=642, type=reshape]; +"643 slice_78" [id=643, type=slice]; +"644 slice_79" [id=644, type=slice]; +"645 slice_80" [id=645, type=slice]; +"646 slice_81" [id=646, type=slice]; +"647 contiguous_7" [id=647, type=contiguous]; +"648 layer_norm_11" [id=648, type=layer_norm]; +"649 add_15" [id=649, type=add]; +"650 add_15_0_0_nncf_smooth_quant_0" [id=650, type=call_module]; +"651 quantize_per_tensor_default_31" [id=651, type=quantize_per_tensor]; +"652 dequantize_per_tensor_default_31" [id=652, type=dequantize_per_tensor]; +"653 linear_30_scale_0" [id=653, type=get_attr]; +"654 linear_30_zero_point_0" [id=654, type=get_attr]; +"655 _frozen_param33_1" [id=655, type=get_attr]; +"656 dequantize_per_channel_default_21" [id=656, type=dequantize_per_channel]; +"657 linear_30" [id=657, type=linear]; +"658 gelu_4" [id=658, type=gelu]; +"659 dropout_18" [id=659, type=dropout]; +"660 dropout_18_0_0_nncf_smooth_quant_0" [id=660, type=call_module]; +"661 quantize_per_tensor_default_32" [id=661, type=quantize_per_tensor]; +"662 dequantize_per_tensor_default_32" [id=662, type=dequantize_per_tensor]; +"663 linear_31_scale_0" [id=663, type=get_attr]; +"664 linear_31_zero_point_0" [id=664, type=get_attr]; +"665 _frozen_param36_1" [id=665, type=get_attr]; +"666 dequantize_per_channel_default_22" [id=666, type=dequantize_per_channel]; +"667 linear_31" [id=667, type=linear]; +"668 dropout_19" [id=668, type=dropout]; +"669 layer_norm_12" [id=669, type=layer_norm]; +"670 add_16" [id=670, type=add]; +"671 _frozen_param38" [id=671, type=get_attr]; +"672 pad_7" [id=672, type=pad]; +"673 roll_4" [id=673, type=roll]; +"674 view_28" [id=674, type=view]; +"675 permute_24" [id=675, type=permute]; +"676 reshape_22" [id=676, type=reshape]; +"677 _frozen_param39" [id=677, type=get_attr]; +"678 reshape_22_0_0_nncf_smooth_quant_0" [id=678, type=call_module]; +"679 quantize_per_tensor_default_33" [id=679, type=quantize_per_tensor]; +"680 dequantize_per_tensor_default_33" [id=680, type=dequantize_per_tensor]; +"681 linear_34_scale_0" [id=681, type=get_attr]; +"682 linear_34_zero_point_0" [id=682, type=get_attr]; +"683 _frozen_param40_1" [id=683, type=get_attr]; +"684 dequantize_per_channel_default_23" [id=684, type=dequantize_per_channel]; +"685 linear_34" [id=685, type=linear]; +"686 reshape_23" [id=686, type=reshape]; +"687 permute_25" [id=687, type=permute]; +"688 select_15" [id=688, type=select]; +"689 select_16" [id=689, type=select]; +"690 select_17" [id=690, type=select]; +"691 linalg_vector_norm_10" [id=691, type=linalg_vector_norm]; +"692 clamp_min_10" [id=692, type=clamp_min]; +"693 expand_as_10" [id=693, type=expand_as]; +"694 div_10" [id=694, type=div]; +"695 quantize_per_tensor_default_34" [id=695, type=quantize_per_tensor]; +"696 dequantize_per_tensor_default_34" [id=696, type=dequantize_per_tensor]; +"697 linalg_vector_norm_11" [id=697, type=linalg_vector_norm]; +"698 clamp_min_11" [id=698, type=clamp_min]; +"699 expand_as_11" [id=699, type=expand_as]; +"700 div_11" [id=700, type=div]; +"701 quantize_per_tensor_default_35" [id=701, type=quantize_per_tensor]; +"702 dequantize_per_tensor_default_35" [id=702, type=dequantize_per_tensor]; +"703 transpose_10" [id=703, type=transpose]; +"704 matmul_10" [id=704, type=matmul]; +"705 _frozen_param41" [id=705, type=get_attr]; +"706 mul_11" [id=706, type=mul]; +"707 add_17" [id=707, type=add]; +"708 new_zeros_2" [id=708, type=new_zeros]; +"709 view_29" [id=709, type=view]; +"710 permute_26" [id=710, type=permute]; +"711 reshape_24" [id=711, type=reshape]; +"712 unsqueeze_14" [id=712, type=unsqueeze]; +"713 unsqueeze_15" [id=713, type=unsqueeze]; +"714 sub_2" [id=714, type=sub]; +"715 ne_2" [id=715, type=ne]; +"716 masked_fill_4" [id=716, type=masked_fill]; +"717 eq_2" [id=717, type=eq]; +"718 masked_fill_5" [id=718, type=masked_fill]; +"719 view_30" [id=719, type=view]; +"720 unsqueeze_16" [id=720, type=unsqueeze]; +"721 unsqueeze_17" [id=721, type=unsqueeze]; +"722 add_18" [id=722, type=add]; +"723 view_31" [id=723, type=view]; +"724 softmax_5" [id=724, type=softmax]; +"725 dropout_20" [id=725, type=dropout]; +"726 matmul_11" [id=726, type=matmul]; +"727 transpose_11" [id=727, type=transpose]; +"728 reshape_25" [id=728, type=reshape]; +"729 reshape_25_0_0_nncf_smooth_quant_0" [id=729, type=call_module]; +"730 quantize_per_tensor_default_36" [id=730, type=quantize_per_tensor]; +"731 dequantize_per_tensor_default_36" [id=731, type=dequantize_per_tensor]; +"732 linear_35_scale_0" [id=732, type=get_attr]; +"733 linear_35_zero_point_0" [id=733, type=get_attr]; +"734 _frozen_param42_1" [id=734, type=get_attr]; +"735 dequantize_per_channel_default_24" [id=735, type=dequantize_per_channel]; +"736 linear_35" [id=736, type=linear]; +"737 dropout_21" [id=737, type=dropout]; +"738 view_32" [id=738, type=view]; +"739 permute_27" [id=739, type=permute]; +"740 reshape_26" [id=740, type=reshape]; +"741 roll_5" [id=741, type=roll]; +"742 slice_101" [id=742, type=slice]; +"743 slice_102" [id=743, type=slice]; +"744 slice_103" [id=744, type=slice]; +"745 slice_104" [id=745, type=slice]; +"746 contiguous_9" [id=746, type=contiguous]; +"747 layer_norm_13" [id=747, type=layer_norm]; +"748 add_19" [id=748, type=add]; +"749 add_19_0_0_nncf_smooth_quant_0" [id=749, type=call_module]; +"750 quantize_per_tensor_default_37" [id=750, type=quantize_per_tensor]; +"751 dequantize_per_tensor_default_37" [id=751, type=dequantize_per_tensor]; +"752 linear_36_scale_0" [id=752, type=get_attr]; +"753 linear_36_zero_point_0" [id=753, type=get_attr]; +"754 _frozen_param43_1" [id=754, type=get_attr]; +"755 dequantize_per_channel_default_25" [id=755, type=dequantize_per_channel]; +"756 linear_36" [id=756, type=linear]; +"757 gelu_5" [id=757, type=gelu]; +"758 dropout_22" [id=758, type=dropout]; +"759 dropout_22_0_0_nncf_smooth_quant_0" [id=759, type=call_module]; +"760 quantize_per_tensor_default_38" [id=760, type=quantize_per_tensor]; +"761 dequantize_per_tensor_default_38" [id=761, type=dequantize_per_tensor]; +"762 linear_37_scale_0" [id=762, type=get_attr]; +"763 linear_37_zero_point_0" [id=763, type=get_attr]; +"764 _frozen_param44_1" [id=764, type=get_attr]; +"765 dequantize_per_channel_default_26" [id=765, type=dequantize_per_channel]; +"766 linear_37" [id=766, type=linear]; +"767 dropout_23" [id=767, type=dropout]; +"768 layer_norm_14" [id=768, type=layer_norm]; +"769 add_20" [id=769, type=add]; +"770 _frozen_param51" [id=770, type=get_attr]; +"771 pad_8" [id=771, type=pad]; +"772 view_35" [id=772, type=view]; +"773 permute_29" [id=773, type=permute]; +"774 reshape_27" [id=774, type=reshape]; +"775 _frozen_param52" [id=775, type=get_attr]; +"776 reshape_27_0_0_nncf_smooth_quant_0" [id=776, type=call_module]; +"777 quantize_per_tensor_default_39" [id=777, type=quantize_per_tensor]; +"778 dequantize_per_tensor_default_39" [id=778, type=dequantize_per_tensor]; +"779 linear_40_scale_0" [id=779, type=get_attr]; +"780 linear_40_zero_point_0" [id=780, type=get_attr]; +"781 _frozen_param45_1" [id=781, type=get_attr]; +"782 dequantize_per_channel_default_27" [id=782, type=dequantize_per_channel]; +"783 linear_40" [id=783, type=linear]; +"784 reshape_28" [id=784, type=reshape]; +"785 permute_30" [id=785, type=permute]; +"786 select_18" [id=786, type=select]; +"787 select_19" [id=787, type=select]; +"788 select_20" [id=788, type=select]; +"789 linalg_vector_norm_12" [id=789, type=linalg_vector_norm]; +"790 clamp_min_12" [id=790, type=clamp_min]; +"791 expand_as_12" [id=791, type=expand_as]; +"792 div_12" [id=792, type=div]; +"793 quantize_per_tensor_default_40" [id=793, type=quantize_per_tensor]; +"794 dequantize_per_tensor_default_40" [id=794, type=dequantize_per_tensor]; +"795 linalg_vector_norm_13" [id=795, type=linalg_vector_norm]; +"796 clamp_min_13" [id=796, type=clamp_min]; +"797 expand_as_13" [id=797, type=expand_as]; +"798 div_13" [id=798, type=div]; +"799 quantize_per_tensor_default_41" [id=799, type=quantize_per_tensor]; +"800 dequantize_per_tensor_default_41" [id=800, type=dequantize_per_tensor]; +"801 transpose_12" [id=801, type=transpose]; +"802 matmul_12" [id=802, type=matmul]; +"803 _frozen_param54" [id=803, type=get_attr]; +"804 mul_13" [id=804, type=mul]; +"805 add_21" [id=805, type=add]; +"806 softmax_6" [id=806, type=softmax]; +"807 dropout_24" [id=807, type=dropout]; +"808 matmul_13" [id=808, type=matmul]; +"809 transpose_13" [id=809, type=transpose]; +"810 reshape_29" [id=810, type=reshape]; +"811 reshape_29_0_0_nncf_smooth_quant_0" [id=811, type=call_module]; +"812 quantize_per_tensor_default_42" [id=812, type=quantize_per_tensor]; +"813 dequantize_per_tensor_default_42" [id=813, type=dequantize_per_tensor]; +"814 linear_41_scale_0" [id=814, type=get_attr]; +"815 linear_41_zero_point_0" [id=815, type=get_attr]; +"816 _frozen_param46_1" [id=816, type=get_attr]; +"817 dequantize_per_channel_default_28" [id=817, type=dequantize_per_channel]; +"818 linear_41" [id=818, type=linear]; +"819 dropout_25" [id=819, type=dropout]; +"820 view_36" [id=820, type=view]; +"821 permute_31" [id=821, type=permute]; +"822 reshape_30" [id=822, type=reshape]; +"823 slice_106" [id=823, type=slice]; +"824 slice_107" [id=824, type=slice]; +"825 slice_108" [id=825, type=slice]; +"826 slice_109" [id=826, type=slice]; +"827 contiguous_11" [id=827, type=contiguous]; +"828 layer_norm_15" [id=828, type=layer_norm]; +"829 add_22" [id=829, type=add]; +"830 add_22_0_0_nncf_smooth_quant_0" [id=830, type=call_module]; +"831 quantize_per_tensor_default_43" [id=831, type=quantize_per_tensor]; +"832 dequantize_per_tensor_default_43" [id=832, type=dequantize_per_tensor]; +"833 linear_42_scale_0" [id=833, type=get_attr]; +"834 linear_42_zero_point_0" [id=834, type=get_attr]; +"835 _frozen_param47_1" [id=835, type=get_attr]; +"836 dequantize_per_channel_default_29" [id=836, type=dequantize_per_channel]; +"837 linear_42" [id=837, type=linear]; +"838 gelu_6" [id=838, type=gelu]; +"839 dropout_26" [id=839, type=dropout]; +"840 dropout_26_0_0_nncf_smooth_quant_0" [id=840, type=call_module]; +"841 quantize_per_tensor_default_44" [id=841, type=quantize_per_tensor]; +"842 dequantize_per_tensor_default_44" [id=842, type=dequantize_per_tensor]; +"843 linear_43_scale_0" [id=843, type=get_attr]; +"844 linear_43_zero_point_0" [id=844, type=get_attr]; +"845 _frozen_param48_1" [id=845, type=get_attr]; +"846 dequantize_per_channel_default_30" [id=846, type=dequantize_per_channel]; +"847 linear_43" [id=847, type=linear]; +"848 dropout_27" [id=848, type=dropout]; +"849 layer_norm_16" [id=849, type=layer_norm]; +"850 add_23" [id=850, type=add]; +"851 _frozen_param55" [id=851, type=get_attr]; +"852 pad_9" [id=852, type=pad]; +"853 roll_6" [id=853, type=roll]; +"854 view_39" [id=854, type=view]; +"855 permute_33" [id=855, type=permute]; +"856 reshape_31" [id=856, type=reshape]; +"857 _frozen_param56" [id=857, type=get_attr]; +"858 reshape_31_0_0_nncf_smooth_quant_0" [id=858, type=call_module]; +"859 quantize_per_tensor_default_45" [id=859, type=quantize_per_tensor]; +"860 dequantize_per_tensor_default_45" [id=860, type=dequantize_per_tensor]; +"861 linear_46_scale_0" [id=861, type=get_attr]; +"862 linear_46_zero_point_0" [id=862, type=get_attr]; +"863 _frozen_param49_1" [id=863, type=get_attr]; +"864 dequantize_per_channel_default_31" [id=864, type=dequantize_per_channel]; +"865 linear_46" [id=865, type=linear]; +"866 reshape_32" [id=866, type=reshape]; +"867 permute_34" [id=867, type=permute]; +"868 select_21" [id=868, type=select]; +"869 select_22" [id=869, type=select]; +"870 select_23" [id=870, type=select]; +"871 linalg_vector_norm_14" [id=871, type=linalg_vector_norm]; +"872 clamp_min_14" [id=872, type=clamp_min]; +"873 expand_as_14" [id=873, type=expand_as]; +"874 div_14" [id=874, type=div]; +"875 quantize_per_tensor_default_46" [id=875, type=quantize_per_tensor]; +"876 dequantize_per_tensor_default_46" [id=876, type=dequantize_per_tensor]; +"877 linalg_vector_norm_15" [id=877, type=linalg_vector_norm]; +"878 clamp_min_15" [id=878, type=clamp_min]; +"879 expand_as_15" [id=879, type=expand_as]; +"880 div_15" [id=880, type=div]; +"881 quantize_per_tensor_default_47" [id=881, type=quantize_per_tensor]; +"882 dequantize_per_tensor_default_47" [id=882, type=dequantize_per_tensor]; +"883 transpose_14" [id=883, type=transpose]; +"884 matmul_14" [id=884, type=matmul]; +"885 _frozen_param58" [id=885, type=get_attr]; +"886 mul_15" [id=886, type=mul]; +"887 add_24" [id=887, type=add]; +"888 new_zeros_3" [id=888, type=new_zeros]; +"889 view_40" [id=889, type=view]; +"890 permute_35" [id=890, type=permute]; +"891 reshape_33" [id=891, type=reshape]; +"892 unsqueeze_20" [id=892, type=unsqueeze]; +"893 unsqueeze_21" [id=893, type=unsqueeze]; +"894 sub_3" [id=894, type=sub]; +"895 ne_3" [id=895, type=ne]; +"896 masked_fill_6" [id=896, type=masked_fill]; +"897 eq_3" [id=897, type=eq]; +"898 masked_fill_7" [id=898, type=masked_fill]; +"899 view_41" [id=899, type=view]; +"900 unsqueeze_22" [id=900, type=unsqueeze]; +"901 unsqueeze_23" [id=901, type=unsqueeze]; +"902 add_25" [id=902, type=add]; +"903 view_42" [id=903, type=view]; +"904 softmax_7" [id=904, type=softmax]; +"905 dropout_28" [id=905, type=dropout]; +"906 matmul_15" [id=906, type=matmul]; +"907 transpose_15" [id=907, type=transpose]; +"908 reshape_34" [id=908, type=reshape]; +"909 reshape_34_0_0_nncf_smooth_quant_0" [id=909, type=call_module]; +"910 quantize_per_tensor_default_48" [id=910, type=quantize_per_tensor]; +"911 dequantize_per_tensor_default_48" [id=911, type=dequantize_per_tensor]; +"912 linear_47_scale_0" [id=912, type=get_attr]; +"913 linear_47_zero_point_0" [id=913, type=get_attr]; +"914 _frozen_param50_1" [id=914, type=get_attr]; +"915 dequantize_per_channel_default_32" [id=915, type=dequantize_per_channel]; +"916 linear_47" [id=916, type=linear]; +"917 dropout_29" [id=917, type=dropout]; +"918 view_43" [id=918, type=view]; +"919 permute_36" [id=919, type=permute]; +"920 reshape_35" [id=920, type=reshape]; +"921 roll_7" [id=921, type=roll]; +"922 slice_129" [id=922, type=slice]; +"923 slice_130" [id=923, type=slice]; +"924 slice_131" [id=924, type=slice]; +"925 slice_132" [id=925, type=slice]; +"926 contiguous_13" [id=926, type=contiguous]; +"927 layer_norm_17" [id=927, type=layer_norm]; +"928 add_26" [id=928, type=add]; +"929 add_26_0_0_nncf_smooth_quant_0" [id=929, type=call_module]; +"930 quantize_per_tensor_default_49" [id=930, type=quantize_per_tensor]; +"931 dequantize_per_tensor_default_49" [id=931, type=dequantize_per_tensor]; +"932 linear_48_scale_0" [id=932, type=get_attr]; +"933 linear_48_zero_point_0" [id=933, type=get_attr]; +"934 _frozen_param53_1" [id=934, type=get_attr]; +"935 dequantize_per_channel_default_33" [id=935, type=dequantize_per_channel]; +"936 linear_48" [id=936, type=linear]; +"937 gelu_7" [id=937, type=gelu]; +"938 dropout_30" [id=938, type=dropout]; +"939 dropout_30_0_0_nncf_smooth_quant_0" [id=939, type=call_module]; +"940 quantize_per_tensor_default_50" [id=940, type=quantize_per_tensor]; +"941 dequantize_per_tensor_default_50" [id=941, type=dequantize_per_tensor]; +"942 linear_49_scale_0" [id=942, type=get_attr]; +"943 linear_49_zero_point_0" [id=943, type=get_attr]; +"944 _frozen_param57_1" [id=944, type=get_attr]; +"945 dequantize_per_channel_default_34" [id=945, type=dequantize_per_channel]; +"946 linear_49" [id=946, type=linear]; +"947 dropout_31" [id=947, type=dropout]; +"948 layer_norm_18" [id=948, type=layer_norm]; +"949 add_27" [id=949, type=add]; +"950 _frozen_param68" [id=950, type=get_attr]; +"951 pad_10" [id=951, type=pad]; +"952 view_46" [id=952, type=view]; +"953 permute_38" [id=953, type=permute]; +"954 reshape_36" [id=954, type=reshape]; +"955 _frozen_param69" [id=955, type=get_attr]; +"956 reshape_36_0_0_nncf_smooth_quant_0" [id=956, type=call_module]; +"957 quantize_per_tensor_default_51" [id=957, type=quantize_per_tensor]; +"958 dequantize_per_tensor_default_51" [id=958, type=dequantize_per_tensor]; +"959 linear_52_scale_0" [id=959, type=get_attr]; +"960 linear_52_zero_point_0" [id=960, type=get_attr]; +"961 _frozen_param59_1" [id=961, type=get_attr]; +"962 dequantize_per_channel_default_35" [id=962, type=dequantize_per_channel]; +"963 linear_52" [id=963, type=linear]; +"964 reshape_37" [id=964, type=reshape]; +"965 permute_39" [id=965, type=permute]; +"966 select_24" [id=966, type=select]; +"967 select_25" [id=967, type=select]; +"968 select_26" [id=968, type=select]; +"969 linalg_vector_norm_16" [id=969, type=linalg_vector_norm]; +"970 clamp_min_16" [id=970, type=clamp_min]; +"971 expand_as_16" [id=971, type=expand_as]; +"972 div_16" [id=972, type=div]; +"973 quantize_per_tensor_default_52" [id=973, type=quantize_per_tensor]; +"974 dequantize_per_tensor_default_52" [id=974, type=dequantize_per_tensor]; +"975 linalg_vector_norm_17" [id=975, type=linalg_vector_norm]; +"976 clamp_min_17" [id=976, type=clamp_min]; +"977 expand_as_17" [id=977, type=expand_as]; +"978 div_17" [id=978, type=div]; +"979 quantize_per_tensor_default_53" [id=979, type=quantize_per_tensor]; +"980 dequantize_per_tensor_default_53" [id=980, type=dequantize_per_tensor]; +"981 transpose_16" [id=981, type=transpose]; +"982 matmul_16" [id=982, type=matmul]; +"983 _frozen_param71" [id=983, type=get_attr]; +"984 mul_17" [id=984, type=mul]; +"985 add_28" [id=985, type=add]; +"986 softmax_8" [id=986, type=softmax]; +"987 dropout_32" [id=987, type=dropout]; +"988 matmul_17" [id=988, type=matmul]; +"989 transpose_17" [id=989, type=transpose]; +"990 reshape_38" [id=990, type=reshape]; +"991 reshape_38_0_0_nncf_smooth_quant_0" [id=991, type=call_module]; +"992 quantize_per_tensor_default_54" [id=992, type=quantize_per_tensor]; +"993 dequantize_per_tensor_default_54" [id=993, type=dequantize_per_tensor]; +"994 linear_53_scale_0" [id=994, type=get_attr]; +"995 linear_53_zero_point_0" [id=995, type=get_attr]; +"996 _frozen_param60_1" [id=996, type=get_attr]; +"997 dequantize_per_channel_default_36" [id=997, type=dequantize_per_channel]; +"998 linear_53" [id=998, type=linear]; +"999 dropout_33" [id=999, type=dropout]; +"1000 view_47" [id=1000, type=view]; +"1001 permute_40" [id=1001, type=permute]; +"1002 reshape_39" [id=1002, type=reshape]; +"1003 slice_134" [id=1003, type=slice]; +"1004 slice_135" [id=1004, type=slice]; +"1005 slice_136" [id=1005, type=slice]; +"1006 slice_137" [id=1006, type=slice]; +"1007 contiguous_15" [id=1007, type=contiguous]; +"1008 layer_norm_19" [id=1008, type=layer_norm]; +"1009 add_29" [id=1009, type=add]; +"1010 add_29_0_0_nncf_smooth_quant_0" [id=1010, type=call_module]; +"1011 quantize_per_tensor_default_55" [id=1011, type=quantize_per_tensor]; +"1012 dequantize_per_tensor_default_55" [id=1012, type=dequantize_per_tensor]; +"1013 linear_54_scale_0" [id=1013, type=get_attr]; +"1014 linear_54_zero_point_0" [id=1014, type=get_attr]; +"1015 _frozen_param61_1" [id=1015, type=get_attr]; +"1016 dequantize_per_channel_default_37" [id=1016, type=dequantize_per_channel]; +"1017 linear_54" [id=1017, type=linear]; +"1018 gelu_8" [id=1018, type=gelu]; +"1019 dropout_34" [id=1019, type=dropout]; +"1020 dropout_34_0_0_nncf_smooth_quant_0" [id=1020, type=call_module]; +"1021 quantize_per_tensor_default_56" [id=1021, type=quantize_per_tensor]; +"1022 dequantize_per_tensor_default_56" [id=1022, type=dequantize_per_tensor]; +"1023 linear_55_scale_0" [id=1023, type=get_attr]; +"1024 linear_55_zero_point_0" [id=1024, type=get_attr]; +"1025 _frozen_param62_1" [id=1025, type=get_attr]; +"1026 dequantize_per_channel_default_38" [id=1026, type=dequantize_per_channel]; +"1027 linear_55" [id=1027, type=linear]; +"1028 dropout_35" [id=1028, type=dropout]; +"1029 layer_norm_20" [id=1029, type=layer_norm]; +"1030 add_30" [id=1030, type=add]; +"1031 _frozen_param72" [id=1031, type=get_attr]; +"1032 pad_11" [id=1032, type=pad]; +"1033 roll_8" [id=1033, type=roll]; +"1034 view_50" [id=1034, type=view]; +"1035 permute_42" [id=1035, type=permute]; +"1036 reshape_40" [id=1036, type=reshape]; +"1037 _frozen_param73" [id=1037, type=get_attr]; +"1038 reshape_40_0_0_nncf_smooth_quant_0" [id=1038, type=call_module]; +"1039 quantize_per_tensor_default_57" [id=1039, type=quantize_per_tensor]; +"1040 dequantize_per_tensor_default_57" [id=1040, type=dequantize_per_tensor]; +"1041 linear_58_scale_0" [id=1041, type=get_attr]; +"1042 linear_58_zero_point_0" [id=1042, type=get_attr]; +"1043 _frozen_param63_1" [id=1043, type=get_attr]; +"1044 dequantize_per_channel_default_39" [id=1044, type=dequantize_per_channel]; +"1045 linear_58" [id=1045, type=linear]; +"1046 reshape_41" [id=1046, type=reshape]; +"1047 permute_43" [id=1047, type=permute]; +"1048 select_27" [id=1048, type=select]; +"1049 select_28" [id=1049, type=select]; +"1050 select_29" [id=1050, type=select]; +"1051 linalg_vector_norm_18" [id=1051, type=linalg_vector_norm]; +"1052 clamp_min_18" [id=1052, type=clamp_min]; +"1053 expand_as_18" [id=1053, type=expand_as]; +"1054 div_18" [id=1054, type=div]; +"1055 quantize_per_tensor_default_58" [id=1055, type=quantize_per_tensor]; +"1056 dequantize_per_tensor_default_58" [id=1056, type=dequantize_per_tensor]; +"1057 linalg_vector_norm_19" [id=1057, type=linalg_vector_norm]; +"1058 clamp_min_19" [id=1058, type=clamp_min]; +"1059 expand_as_19" [id=1059, type=expand_as]; +"1060 div_19" [id=1060, type=div]; +"1061 quantize_per_tensor_default_59" [id=1061, type=quantize_per_tensor]; +"1062 dequantize_per_tensor_default_59" [id=1062, type=dequantize_per_tensor]; +"1063 transpose_18" [id=1063, type=transpose]; +"1064 matmul_18" [id=1064, type=matmul]; +"1065 _frozen_param75" [id=1065, type=get_attr]; +"1066 mul_19" [id=1066, type=mul]; +"1067 add_31" [id=1067, type=add]; +"1068 new_zeros_4" [id=1068, type=new_zeros]; +"1069 view_51" [id=1069, type=view]; +"1070 permute_44" [id=1070, type=permute]; +"1071 reshape_42" [id=1071, type=reshape]; +"1072 unsqueeze_26" [id=1072, type=unsqueeze]; +"1073 unsqueeze_27" [id=1073, type=unsqueeze]; +"1074 sub_4" [id=1074, type=sub]; +"1075 ne_4" [id=1075, type=ne]; +"1076 masked_fill_8" [id=1076, type=masked_fill]; +"1077 eq_4" [id=1077, type=eq]; +"1078 masked_fill_9" [id=1078, type=masked_fill]; +"1079 view_52" [id=1079, type=view]; +"1080 unsqueeze_28" [id=1080, type=unsqueeze]; +"1081 unsqueeze_29" [id=1081, type=unsqueeze]; +"1082 add_32" [id=1082, type=add]; +"1083 view_53" [id=1083, type=view]; +"1084 softmax_9" [id=1084, type=softmax]; +"1085 dropout_36" [id=1085, type=dropout]; +"1086 matmul_19" [id=1086, type=matmul]; +"1087 transpose_19" [id=1087, type=transpose]; +"1088 reshape_43" [id=1088, type=reshape]; +"1089 reshape_43_0_0_nncf_smooth_quant_0" [id=1089, type=call_module]; +"1090 quantize_per_tensor_default_60" [id=1090, type=quantize_per_tensor]; +"1091 dequantize_per_tensor_default_60" [id=1091, type=dequantize_per_tensor]; +"1092 linear_59_scale_0" [id=1092, type=get_attr]; +"1093 linear_59_zero_point_0" [id=1093, type=get_attr]; +"1094 _frozen_param64_1" [id=1094, type=get_attr]; +"1095 dequantize_per_channel_default_40" [id=1095, type=dequantize_per_channel]; +"1096 linear_59" [id=1096, type=linear]; +"1097 dropout_37" [id=1097, type=dropout]; +"1098 view_54" [id=1098, type=view]; +"1099 permute_45" [id=1099, type=permute]; +"1100 reshape_44" [id=1100, type=reshape]; +"1101 roll_9" [id=1101, type=roll]; +"1102 slice_157" [id=1102, type=slice]; +"1103 slice_158" [id=1103, type=slice]; +"1104 slice_159" [id=1104, type=slice]; +"1105 slice_160" [id=1105, type=slice]; +"1106 contiguous_17" [id=1106, type=contiguous]; +"1107 layer_norm_21" [id=1107, type=layer_norm]; +"1108 add_33" [id=1108, type=add]; +"1109 add_33_0_0_nncf_smooth_quant_0" [id=1109, type=call_module]; +"1110 quantize_per_tensor_default_61" [id=1110, type=quantize_per_tensor]; +"1111 dequantize_per_tensor_default_61" [id=1111, type=dequantize_per_tensor]; +"1112 linear_60_scale_0" [id=1112, type=get_attr]; +"1113 linear_60_zero_point_0" [id=1113, type=get_attr]; +"1114 _frozen_param65_1" [id=1114, type=get_attr]; +"1115 dequantize_per_channel_default_41" [id=1115, type=dequantize_per_channel]; +"1116 linear_60" [id=1116, type=linear]; +"1117 gelu_9" [id=1117, type=gelu]; +"1118 dropout_38" [id=1118, type=dropout]; +"1119 dropout_38_0_0_nncf_smooth_quant_0" [id=1119, type=call_module]; +"1120 quantize_per_tensor_default_62" [id=1120, type=quantize_per_tensor]; +"1121 dequantize_per_tensor_default_62" [id=1121, type=dequantize_per_tensor]; +"1122 linear_61_scale_0" [id=1122, type=get_attr]; +"1123 linear_61_zero_point_0" [id=1123, type=get_attr]; +"1124 _frozen_param66_1" [id=1124, type=get_attr]; +"1125 dequantize_per_channel_default_42" [id=1125, type=dequantize_per_channel]; +"1126 linear_61" [id=1126, type=linear]; +"1127 dropout_39" [id=1127, type=dropout]; +"1128 layer_norm_22" [id=1128, type=layer_norm]; +"1129 add_34" [id=1129, type=add]; +"1130 _frozen_param85" [id=1130, type=get_attr]; +"1131 pad_12" [id=1131, type=pad]; +"1132 view_57" [id=1132, type=view]; +"1133 permute_47" [id=1133, type=permute]; +"1134 reshape_45" [id=1134, type=reshape]; +"1135 _frozen_param86" [id=1135, type=get_attr]; +"1136 reshape_45_0_0_nncf_smooth_quant_0" [id=1136, type=call_module]; +"1137 quantize_per_tensor_default_63" [id=1137, type=quantize_per_tensor]; +"1138 dequantize_per_tensor_default_63" [id=1138, type=dequantize_per_tensor]; +"1139 linear_64_scale_0" [id=1139, type=get_attr]; +"1140 linear_64_zero_point_0" [id=1140, type=get_attr]; +"1141 _frozen_param67_1" [id=1141, type=get_attr]; +"1142 dequantize_per_channel_default_43" [id=1142, type=dequantize_per_channel]; +"1143 linear_64" [id=1143, type=linear]; +"1144 reshape_46" [id=1144, type=reshape]; +"1145 permute_48" [id=1145, type=permute]; +"1146 select_30" [id=1146, type=select]; +"1147 select_31" [id=1147, type=select]; +"1148 select_32" [id=1148, type=select]; +"1149 linalg_vector_norm_20" [id=1149, type=linalg_vector_norm]; +"1150 clamp_min_20" [id=1150, type=clamp_min]; +"1151 expand_as_20" [id=1151, type=expand_as]; +"1152 div_20" [id=1152, type=div]; +"1153 quantize_per_tensor_default_64" [id=1153, type=quantize_per_tensor]; +"1154 dequantize_per_tensor_default_64" [id=1154, type=dequantize_per_tensor]; +"1155 linalg_vector_norm_21" [id=1155, type=linalg_vector_norm]; +"1156 clamp_min_21" [id=1156, type=clamp_min]; +"1157 expand_as_21" [id=1157, type=expand_as]; +"1158 div_21" [id=1158, type=div]; +"1159 quantize_per_tensor_default_65" [id=1159, type=quantize_per_tensor]; +"1160 dequantize_per_tensor_default_65" [id=1160, type=dequantize_per_tensor]; +"1161 transpose_20" [id=1161, type=transpose]; +"1162 matmul_20" [id=1162, type=matmul]; +"1163 _frozen_param88" [id=1163, type=get_attr]; +"1164 mul_21" [id=1164, type=mul]; +"1165 add_35" [id=1165, type=add]; +"1166 softmax_10" [id=1166, type=softmax]; +"1167 dropout_40" [id=1167, type=dropout]; +"1168 matmul_21" [id=1168, type=matmul]; +"1169 transpose_21" [id=1169, type=transpose]; +"1170 reshape_47" [id=1170, type=reshape]; +"1171 reshape_47_0_0_nncf_smooth_quant_0" [id=1171, type=call_module]; +"1172 quantize_per_tensor_default_66" [id=1172, type=quantize_per_tensor]; +"1173 dequantize_per_tensor_default_66" [id=1173, type=dequantize_per_tensor]; +"1174 linear_65_scale_0" [id=1174, type=get_attr]; +"1175 linear_65_zero_point_0" [id=1175, type=get_attr]; +"1176 _frozen_param70_1" [id=1176, type=get_attr]; +"1177 dequantize_per_channel_default_44" [id=1177, type=dequantize_per_channel]; +"1178 linear_65" [id=1178, type=linear]; +"1179 dropout_41" [id=1179, type=dropout]; +"1180 view_58" [id=1180, type=view]; +"1181 permute_49" [id=1181, type=permute]; +"1182 reshape_48" [id=1182, type=reshape]; +"1183 slice_162" [id=1183, type=slice]; +"1184 slice_163" [id=1184, type=slice]; +"1185 slice_164" [id=1185, type=slice]; +"1186 slice_165" [id=1186, type=slice]; +"1187 contiguous_19" [id=1187, type=contiguous]; +"1188 layer_norm_23" [id=1188, type=layer_norm]; +"1189 add_36" [id=1189, type=add]; +"1190 add_36_0_0_nncf_smooth_quant_0" [id=1190, type=call_module]; +"1191 quantize_per_tensor_default_67" [id=1191, type=quantize_per_tensor]; +"1192 dequantize_per_tensor_default_67" [id=1192, type=dequantize_per_tensor]; +"1193 linear_66_scale_0" [id=1193, type=get_attr]; +"1194 linear_66_zero_point_0" [id=1194, type=get_attr]; +"1195 _frozen_param74_1" [id=1195, type=get_attr]; +"1196 dequantize_per_channel_default_45" [id=1196, type=dequantize_per_channel]; +"1197 linear_66" [id=1197, type=linear]; +"1198 gelu_10" [id=1198, type=gelu]; +"1199 dropout_42" [id=1199, type=dropout]; +"1200 dropout_42_0_0_nncf_smooth_quant_0" [id=1200, type=call_module]; +"1201 quantize_per_tensor_default_68" [id=1201, type=quantize_per_tensor]; +"1202 dequantize_per_tensor_default_68" [id=1202, type=dequantize_per_tensor]; +"1203 linear_67_scale_0" [id=1203, type=get_attr]; +"1204 linear_67_zero_point_0" [id=1204, type=get_attr]; +"1205 _frozen_param76_1" [id=1205, type=get_attr]; +"1206 dequantize_per_channel_default_46" [id=1206, type=dequantize_per_channel]; +"1207 linear_67" [id=1207, type=linear]; +"1208 dropout_43" [id=1208, type=dropout]; +"1209 layer_norm_24" [id=1209, type=layer_norm]; +"1210 add_37" [id=1210, type=add]; +"1211 _frozen_param89" [id=1211, type=get_attr]; +"1212 pad_13" [id=1212, type=pad]; +"1213 roll_10" [id=1213, type=roll]; +"1214 view_61" [id=1214, type=view]; +"1215 permute_51" [id=1215, type=permute]; +"1216 reshape_49" [id=1216, type=reshape]; +"1217 _frozen_param90" [id=1217, type=get_attr]; +"1218 reshape_49_0_0_nncf_smooth_quant_0" [id=1218, type=call_module]; +"1219 quantize_per_tensor_default_69" [id=1219, type=quantize_per_tensor]; +"1220 dequantize_per_tensor_default_69" [id=1220, type=dequantize_per_tensor]; +"1221 linear_70_scale_0" [id=1221, type=get_attr]; +"1222 linear_70_zero_point_0" [id=1222, type=get_attr]; +"1223 _frozen_param77_1" [id=1223, type=get_attr]; +"1224 dequantize_per_channel_default_47" [id=1224, type=dequantize_per_channel]; +"1225 linear_70" [id=1225, type=linear]; +"1226 reshape_50" [id=1226, type=reshape]; +"1227 permute_52" [id=1227, type=permute]; +"1228 select_33" [id=1228, type=select]; +"1229 select_34" [id=1229, type=select]; +"1230 select_35" [id=1230, type=select]; +"1231 linalg_vector_norm_22" [id=1231, type=linalg_vector_norm]; +"1232 clamp_min_22" [id=1232, type=clamp_min]; +"1233 expand_as_22" [id=1233, type=expand_as]; +"1234 div_22" [id=1234, type=div]; +"1235 quantize_per_tensor_default_70" [id=1235, type=quantize_per_tensor]; +"1236 dequantize_per_tensor_default_70" [id=1236, type=dequantize_per_tensor]; +"1237 linalg_vector_norm_23" [id=1237, type=linalg_vector_norm]; +"1238 clamp_min_23" [id=1238, type=clamp_min]; +"1239 expand_as_23" [id=1239, type=expand_as]; +"1240 div_23" [id=1240, type=div]; +"1241 quantize_per_tensor_default_71" [id=1241, type=quantize_per_tensor]; +"1242 dequantize_per_tensor_default_71" [id=1242, type=dequantize_per_tensor]; +"1243 transpose_22" [id=1243, type=transpose]; +"1244 matmul_22" [id=1244, type=matmul]; +"1245 _frozen_param92" [id=1245, type=get_attr]; +"1246 mul_23" [id=1246, type=mul]; +"1247 add_38" [id=1247, type=add]; +"1248 new_zeros_5" [id=1248, type=new_zeros]; +"1249 view_62" [id=1249, type=view]; +"1250 permute_53" [id=1250, type=permute]; +"1251 reshape_51" [id=1251, type=reshape]; +"1252 unsqueeze_32" [id=1252, type=unsqueeze]; +"1253 unsqueeze_33" [id=1253, type=unsqueeze]; +"1254 sub_5" [id=1254, type=sub]; +"1255 ne_5" [id=1255, type=ne]; +"1256 masked_fill_10" [id=1256, type=masked_fill]; +"1257 eq_5" [id=1257, type=eq]; +"1258 masked_fill_11" [id=1258, type=masked_fill]; +"1259 view_63" [id=1259, type=view]; +"1260 unsqueeze_34" [id=1260, type=unsqueeze]; +"1261 unsqueeze_35" [id=1261, type=unsqueeze]; +"1262 add_39" [id=1262, type=add]; +"1263 view_64" [id=1263, type=view]; +"1264 softmax_11" [id=1264, type=softmax]; +"1265 dropout_44" [id=1265, type=dropout]; +"1266 matmul_23" [id=1266, type=matmul]; +"1267 transpose_23" [id=1267, type=transpose]; +"1268 reshape_52" [id=1268, type=reshape]; +"1269 reshape_52_0_0_nncf_smooth_quant_0" [id=1269, type=call_module]; +"1270 quantize_per_tensor_default_72" [id=1270, type=quantize_per_tensor]; +"1271 dequantize_per_tensor_default_72" [id=1271, type=dequantize_per_tensor]; +"1272 linear_71_scale_0" [id=1272, type=get_attr]; +"1273 linear_71_zero_point_0" [id=1273, type=get_attr]; +"1274 _frozen_param78_1" [id=1274, type=get_attr]; +"1275 dequantize_per_channel_default_48" [id=1275, type=dequantize_per_channel]; +"1276 linear_71" [id=1276, type=linear]; +"1277 dropout_45" [id=1277, type=dropout]; +"1278 view_65" [id=1278, type=view]; +"1279 permute_54" [id=1279, type=permute]; +"1280 reshape_53" [id=1280, type=reshape]; +"1281 roll_11" [id=1281, type=roll]; +"1282 slice_185" [id=1282, type=slice]; +"1283 slice_186" [id=1283, type=slice]; +"1284 slice_187" [id=1284, type=slice]; +"1285 slice_188" [id=1285, type=slice]; +"1286 contiguous_21" [id=1286, type=contiguous]; +"1287 layer_norm_25" [id=1287, type=layer_norm]; +"1288 add_40" [id=1288, type=add]; +"1289 add_40_0_0_nncf_smooth_quant_0" [id=1289, type=call_module]; +"1290 quantize_per_tensor_default_73" [id=1290, type=quantize_per_tensor]; +"1291 dequantize_per_tensor_default_73" [id=1291, type=dequantize_per_tensor]; +"1292 linear_72_scale_0" [id=1292, type=get_attr]; +"1293 linear_72_zero_point_0" [id=1293, type=get_attr]; +"1294 _frozen_param79_1" [id=1294, type=get_attr]; +"1295 dequantize_per_channel_default_49" [id=1295, type=dequantize_per_channel]; +"1296 linear_72" [id=1296, type=linear]; +"1297 gelu_11" [id=1297, type=gelu]; +"1298 dropout_46" [id=1298, type=dropout]; +"1299 dropout_46_0_0_nncf_smooth_quant_0" [id=1299, type=call_module]; +"1300 quantize_per_tensor_default_74" [id=1300, type=quantize_per_tensor]; +"1301 dequantize_per_tensor_default_74" [id=1301, type=dequantize_per_tensor]; +"1302 linear_73_scale_0" [id=1302, type=get_attr]; +"1303 linear_73_zero_point_0" [id=1303, type=get_attr]; +"1304 _frozen_param80_1" [id=1304, type=get_attr]; +"1305 dequantize_per_channel_default_50" [id=1305, type=dequantize_per_channel]; +"1306 linear_73" [id=1306, type=linear]; +"1307 dropout_47" [id=1307, type=dropout]; +"1308 layer_norm_26" [id=1308, type=layer_norm]; +"1309 add_41" [id=1309, type=add]; +"1310 _frozen_param102" [id=1310, type=get_attr]; +"1311 pad_14" [id=1311, type=pad]; +"1312 view_68" [id=1312, type=view]; +"1313 permute_56" [id=1313, type=permute]; +"1314 reshape_54" [id=1314, type=reshape]; +"1315 _frozen_param103" [id=1315, type=get_attr]; +"1316 reshape_54_0_0_nncf_smooth_quant_0" [id=1316, type=call_module]; +"1317 quantize_per_tensor_default_75" [id=1317, type=quantize_per_tensor]; +"1318 dequantize_per_tensor_default_75" [id=1318, type=dequantize_per_tensor]; +"1319 linear_76_scale_0" [id=1319, type=get_attr]; +"1320 linear_76_zero_point_0" [id=1320, type=get_attr]; +"1321 _frozen_param81_1" [id=1321, type=get_attr]; +"1322 dequantize_per_channel_default_51" [id=1322, type=dequantize_per_channel]; +"1323 linear_76" [id=1323, type=linear]; +"1324 reshape_55" [id=1324, type=reshape]; +"1325 permute_57" [id=1325, type=permute]; +"1326 select_36" [id=1326, type=select]; +"1327 select_37" [id=1327, type=select]; +"1328 select_38" [id=1328, type=select]; +"1329 linalg_vector_norm_24" [id=1329, type=linalg_vector_norm]; +"1330 clamp_min_24" [id=1330, type=clamp_min]; +"1331 expand_as_24" [id=1331, type=expand_as]; +"1332 div_24" [id=1332, type=div]; +"1333 quantize_per_tensor_default_76" [id=1333, type=quantize_per_tensor]; +"1334 dequantize_per_tensor_default_76" [id=1334, type=dequantize_per_tensor]; +"1335 linalg_vector_norm_25" [id=1335, type=linalg_vector_norm]; +"1336 clamp_min_25" [id=1336, type=clamp_min]; +"1337 expand_as_25" [id=1337, type=expand_as]; +"1338 div_25" [id=1338, type=div]; +"1339 quantize_per_tensor_default_77" [id=1339, type=quantize_per_tensor]; +"1340 dequantize_per_tensor_default_77" [id=1340, type=dequantize_per_tensor]; +"1341 transpose_24" [id=1341, type=transpose]; +"1342 matmul_24" [id=1342, type=matmul]; +"1343 _frozen_param105" [id=1343, type=get_attr]; +"1344 mul_25" [id=1344, type=mul]; +"1345 add_42" [id=1345, type=add]; +"1346 softmax_12" [id=1346, type=softmax]; +"1347 dropout_48" [id=1347, type=dropout]; +"1348 matmul_25" [id=1348, type=matmul]; +"1349 transpose_25" [id=1349, type=transpose]; +"1350 reshape_56" [id=1350, type=reshape]; +"1351 reshape_56_0_0_nncf_smooth_quant_0" [id=1351, type=call_module]; +"1352 quantize_per_tensor_default_78" [id=1352, type=quantize_per_tensor]; +"1353 dequantize_per_tensor_default_78" [id=1353, type=dequantize_per_tensor]; +"1354 linear_77_scale_0" [id=1354, type=get_attr]; +"1355 linear_77_zero_point_0" [id=1355, type=get_attr]; +"1356 _frozen_param82_1" [id=1356, type=get_attr]; +"1357 dequantize_per_channel_default_52" [id=1357, type=dequantize_per_channel]; +"1358 linear_77" [id=1358, type=linear]; +"1359 dropout_49" [id=1359, type=dropout]; +"1360 view_69" [id=1360, type=view]; +"1361 permute_58" [id=1361, type=permute]; +"1362 reshape_57" [id=1362, type=reshape]; +"1363 slice_190" [id=1363, type=slice]; +"1364 slice_191" [id=1364, type=slice]; +"1365 slice_192" [id=1365, type=slice]; +"1366 slice_193" [id=1366, type=slice]; +"1367 contiguous_23" [id=1367, type=contiguous]; +"1368 layer_norm_27" [id=1368, type=layer_norm]; +"1369 add_43" [id=1369, type=add]; +"1370 add_43_0_0_nncf_smooth_quant_0" [id=1370, type=call_module]; +"1371 quantize_per_tensor_default_79" [id=1371, type=quantize_per_tensor]; +"1372 dequantize_per_tensor_default_79" [id=1372, type=dequantize_per_tensor]; +"1373 linear_78_scale_0" [id=1373, type=get_attr]; +"1374 linear_78_zero_point_0" [id=1374, type=get_attr]; +"1375 _frozen_param83_1" [id=1375, type=get_attr]; +"1376 dequantize_per_channel_default_53" [id=1376, type=dequantize_per_channel]; +"1377 linear_78" [id=1377, type=linear]; +"1378 gelu_12" [id=1378, type=gelu]; +"1379 dropout_50" [id=1379, type=dropout]; +"1380 dropout_50_0_0_nncf_smooth_quant_0" [id=1380, type=call_module]; +"1381 quantize_per_tensor_default_80" [id=1381, type=quantize_per_tensor]; +"1382 dequantize_per_tensor_default_80" [id=1382, type=dequantize_per_tensor]; +"1383 linear_79_scale_0" [id=1383, type=get_attr]; +"1384 linear_79_zero_point_0" [id=1384, type=get_attr]; +"1385 _frozen_param84_1" [id=1385, type=get_attr]; +"1386 dequantize_per_channel_default_54" [id=1386, type=dequantize_per_channel]; +"1387 linear_79" [id=1387, type=linear]; +"1388 dropout_51" [id=1388, type=dropout]; +"1389 layer_norm_28" [id=1389, type=layer_norm]; +"1390 add_44" [id=1390, type=add]; +"1391 _frozen_param106" [id=1391, type=get_attr]; +"1392 pad_15" [id=1392, type=pad]; +"1393 roll_12" [id=1393, type=roll]; +"1394 view_72" [id=1394, type=view]; +"1395 permute_60" [id=1395, type=permute]; +"1396 reshape_58" [id=1396, type=reshape]; +"1397 _frozen_param107" [id=1397, type=get_attr]; +"1398 reshape_58_0_0_nncf_smooth_quant_0" [id=1398, type=call_module]; +"1399 quantize_per_tensor_default_81" [id=1399, type=quantize_per_tensor]; +"1400 dequantize_per_tensor_default_81" [id=1400, type=dequantize_per_tensor]; +"1401 linear_82_scale_0" [id=1401, type=get_attr]; +"1402 linear_82_zero_point_0" [id=1402, type=get_attr]; +"1403 _frozen_param87_1" [id=1403, type=get_attr]; +"1404 dequantize_per_channel_default_55" [id=1404, type=dequantize_per_channel]; +"1405 linear_82" [id=1405, type=linear]; +"1406 reshape_59" [id=1406, type=reshape]; +"1407 permute_61" [id=1407, type=permute]; +"1408 select_39" [id=1408, type=select]; +"1409 select_40" [id=1409, type=select]; +"1410 select_41" [id=1410, type=select]; +"1411 linalg_vector_norm_26" [id=1411, type=linalg_vector_norm]; +"1412 clamp_min_26" [id=1412, type=clamp_min]; +"1413 expand_as_26" [id=1413, type=expand_as]; +"1414 div_26" [id=1414, type=div]; +"1415 quantize_per_tensor_default_82" [id=1415, type=quantize_per_tensor]; +"1416 dequantize_per_tensor_default_82" [id=1416, type=dequantize_per_tensor]; +"1417 linalg_vector_norm_27" [id=1417, type=linalg_vector_norm]; +"1418 clamp_min_27" [id=1418, type=clamp_min]; +"1419 expand_as_27" [id=1419, type=expand_as]; +"1420 div_27" [id=1420, type=div]; +"1421 quantize_per_tensor_default_83" [id=1421, type=quantize_per_tensor]; +"1422 dequantize_per_tensor_default_83" [id=1422, type=dequantize_per_tensor]; +"1423 transpose_26" [id=1423, type=transpose]; +"1424 matmul_26" [id=1424, type=matmul]; +"1425 _frozen_param109" [id=1425, type=get_attr]; +"1426 mul_27" [id=1426, type=mul]; +"1427 add_45" [id=1427, type=add]; +"1428 new_zeros_6" [id=1428, type=new_zeros]; +"1429 view_73" [id=1429, type=view]; +"1430 permute_62" [id=1430, type=permute]; +"1431 reshape_60" [id=1431, type=reshape]; +"1432 unsqueeze_38" [id=1432, type=unsqueeze]; +"1433 unsqueeze_39" [id=1433, type=unsqueeze]; +"1434 sub_6" [id=1434, type=sub]; +"1435 ne_6" [id=1435, type=ne]; +"1436 masked_fill_12" [id=1436, type=masked_fill]; +"1437 eq_6" [id=1437, type=eq]; +"1438 masked_fill_13" [id=1438, type=masked_fill]; +"1439 view_74" [id=1439, type=view]; +"1440 unsqueeze_40" [id=1440, type=unsqueeze]; +"1441 unsqueeze_41" [id=1441, type=unsqueeze]; +"1442 add_46" [id=1442, type=add]; +"1443 view_75" [id=1443, type=view]; +"1444 softmax_13" [id=1444, type=softmax]; +"1445 dropout_52" [id=1445, type=dropout]; +"1446 matmul_27" [id=1446, type=matmul]; +"1447 transpose_27" [id=1447, type=transpose]; +"1448 reshape_61" [id=1448, type=reshape]; +"1449 reshape_61_0_0_nncf_smooth_quant_0" [id=1449, type=call_module]; +"1450 quantize_per_tensor_default_84" [id=1450, type=quantize_per_tensor]; +"1451 dequantize_per_tensor_default_84" [id=1451, type=dequantize_per_tensor]; +"1452 linear_83_scale_0" [id=1452, type=get_attr]; +"1453 linear_83_zero_point_0" [id=1453, type=get_attr]; +"1454 _frozen_param91_1" [id=1454, type=get_attr]; +"1455 dequantize_per_channel_default_56" [id=1455, type=dequantize_per_channel]; +"1456 linear_83" [id=1456, type=linear]; +"1457 dropout_53" [id=1457, type=dropout]; +"1458 view_76" [id=1458, type=view]; +"1459 permute_63" [id=1459, type=permute]; +"1460 reshape_62" [id=1460, type=reshape]; +"1461 roll_13" [id=1461, type=roll]; +"1462 slice_213" [id=1462, type=slice]; +"1463 slice_214" [id=1463, type=slice]; +"1464 slice_215" [id=1464, type=slice]; +"1465 slice_216" [id=1465, type=slice]; +"1466 contiguous_25" [id=1466, type=contiguous]; +"1467 layer_norm_29" [id=1467, type=layer_norm]; +"1468 add_47" [id=1468, type=add]; +"1469 add_47_0_0_nncf_smooth_quant_0" [id=1469, type=call_module]; +"1470 quantize_per_tensor_default_85" [id=1470, type=quantize_per_tensor]; +"1471 dequantize_per_tensor_default_85" [id=1471, type=dequantize_per_tensor]; +"1472 linear_84_scale_0" [id=1472, type=get_attr]; +"1473 linear_84_zero_point_0" [id=1473, type=get_attr]; +"1474 _frozen_param93_1" [id=1474, type=get_attr]; +"1475 dequantize_per_channel_default_57" [id=1475, type=dequantize_per_channel]; +"1476 linear_84" [id=1476, type=linear]; +"1477 gelu_13" [id=1477, type=gelu]; +"1478 dropout_54" [id=1478, type=dropout]; +"1479 dropout_54_0_0_nncf_smooth_quant_0" [id=1479, type=call_module]; +"1480 quantize_per_tensor_default_86" [id=1480, type=quantize_per_tensor]; +"1481 dequantize_per_tensor_default_86" [id=1481, type=dequantize_per_tensor]; +"1482 linear_85_scale_0" [id=1482, type=get_attr]; +"1483 linear_85_zero_point_0" [id=1483, type=get_attr]; +"1484 _frozen_param94_1" [id=1484, type=get_attr]; +"1485 dequantize_per_channel_default_58" [id=1485, type=dequantize_per_channel]; +"1486 linear_85" [id=1486, type=linear]; +"1487 dropout_55" [id=1487, type=dropout]; +"1488 layer_norm_30" [id=1488, type=layer_norm]; +"1489 add_48" [id=1489, type=add]; +"1490 _frozen_param119" [id=1490, type=get_attr]; +"1491 pad_16" [id=1491, type=pad]; +"1492 view_79" [id=1492, type=view]; +"1493 permute_65" [id=1493, type=permute]; +"1494 reshape_63" [id=1494, type=reshape]; +"1495 _frozen_param120" [id=1495, type=get_attr]; +"1496 reshape_63_0_0_nncf_smooth_quant_0" [id=1496, type=call_module]; +"1497 quantize_per_tensor_default_87" [id=1497, type=quantize_per_tensor]; +"1498 dequantize_per_tensor_default_87" [id=1498, type=dequantize_per_tensor]; +"1499 linear_88_scale_0" [id=1499, type=get_attr]; +"1500 linear_88_zero_point_0" [id=1500, type=get_attr]; +"1501 _frozen_param95_1" [id=1501, type=get_attr]; +"1502 dequantize_per_channel_default_59" [id=1502, type=dequantize_per_channel]; +"1503 linear_88" [id=1503, type=linear]; +"1504 reshape_64" [id=1504, type=reshape]; +"1505 permute_66" [id=1505, type=permute]; +"1506 select_42" [id=1506, type=select]; +"1507 select_43" [id=1507, type=select]; +"1508 select_44" [id=1508, type=select]; +"1509 linalg_vector_norm_28" [id=1509, type=linalg_vector_norm]; +"1510 clamp_min_28" [id=1510, type=clamp_min]; +"1511 expand_as_28" [id=1511, type=expand_as]; +"1512 div_28" [id=1512, type=div]; +"1513 quantize_per_tensor_default_88" [id=1513, type=quantize_per_tensor]; +"1514 dequantize_per_tensor_default_88" [id=1514, type=dequantize_per_tensor]; +"1515 linalg_vector_norm_29" [id=1515, type=linalg_vector_norm]; +"1516 clamp_min_29" [id=1516, type=clamp_min]; +"1517 expand_as_29" [id=1517, type=expand_as]; +"1518 div_29" [id=1518, type=div]; +"1519 quantize_per_tensor_default_89" [id=1519, type=quantize_per_tensor]; +"1520 dequantize_per_tensor_default_89" [id=1520, type=dequantize_per_tensor]; +"1521 transpose_28" [id=1521, type=transpose]; +"1522 matmul_28" [id=1522, type=matmul]; +"1523 _frozen_param122" [id=1523, type=get_attr]; +"1524 mul_29" [id=1524, type=mul]; +"1525 add_49" [id=1525, type=add]; +"1526 softmax_14" [id=1526, type=softmax]; +"1527 dropout_56" [id=1527, type=dropout]; +"1528 matmul_29" [id=1528, type=matmul]; +"1529 transpose_29" [id=1529, type=transpose]; +"1530 reshape_65" [id=1530, type=reshape]; +"1531 reshape_65_0_0_nncf_smooth_quant_0" [id=1531, type=call_module]; +"1532 quantize_per_tensor_default_90" [id=1532, type=quantize_per_tensor]; +"1533 dequantize_per_tensor_default_90" [id=1533, type=dequantize_per_tensor]; +"1534 linear_89_scale_0" [id=1534, type=get_attr]; +"1535 linear_89_zero_point_0" [id=1535, type=get_attr]; +"1536 _frozen_param96_1" [id=1536, type=get_attr]; +"1537 dequantize_per_channel_default_60" [id=1537, type=dequantize_per_channel]; +"1538 linear_89" [id=1538, type=linear]; +"1539 dropout_57" [id=1539, type=dropout]; +"1540 view_80" [id=1540, type=view]; +"1541 permute_67" [id=1541, type=permute]; +"1542 reshape_66" [id=1542, type=reshape]; +"1543 slice_218" [id=1543, type=slice]; +"1544 slice_219" [id=1544, type=slice]; +"1545 slice_220" [id=1545, type=slice]; +"1546 slice_221" [id=1546, type=slice]; +"1547 contiguous_27" [id=1547, type=contiguous]; +"1548 layer_norm_31" [id=1548, type=layer_norm]; +"1549 add_50" [id=1549, type=add]; +"1550 add_50_0_0_nncf_smooth_quant_0" [id=1550, type=call_module]; +"1551 quantize_per_tensor_default_91" [id=1551, type=quantize_per_tensor]; +"1552 dequantize_per_tensor_default_91" [id=1552, type=dequantize_per_tensor]; +"1553 linear_90_scale_0" [id=1553, type=get_attr]; +"1554 linear_90_zero_point_0" [id=1554, type=get_attr]; +"1555 _frozen_param97_1" [id=1555, type=get_attr]; +"1556 dequantize_per_channel_default_61" [id=1556, type=dequantize_per_channel]; +"1557 linear_90" [id=1557, type=linear]; +"1558 gelu_14" [id=1558, type=gelu]; +"1559 dropout_58" [id=1559, type=dropout]; +"1560 dropout_58_0_0_nncf_smooth_quant_0" [id=1560, type=call_module]; +"1561 quantize_per_tensor_default_92" [id=1561, type=quantize_per_tensor]; +"1562 dequantize_per_tensor_default_92" [id=1562, type=dequantize_per_tensor]; +"1563 linear_91_scale_0" [id=1563, type=get_attr]; +"1564 linear_91_zero_point_0" [id=1564, type=get_attr]; +"1565 _frozen_param98_1" [id=1565, type=get_attr]; +"1566 dequantize_per_channel_default_62" [id=1566, type=dequantize_per_channel]; +"1567 linear_91" [id=1567, type=linear]; +"1568 dropout_59" [id=1568, type=dropout]; +"1569 layer_norm_32" [id=1569, type=layer_norm]; +"1570 add_51" [id=1570, type=add]; +"1571 _frozen_param123" [id=1571, type=get_attr]; +"1572 pad_17" [id=1572, type=pad]; +"1573 roll_14" [id=1573, type=roll]; +"1574 view_83" [id=1574, type=view]; +"1575 permute_69" [id=1575, type=permute]; +"1576 reshape_67" [id=1576, type=reshape]; +"1577 _frozen_param124" [id=1577, type=get_attr]; +"1578 reshape_67_0_0_nncf_smooth_quant_0" [id=1578, type=call_module]; +"1579 quantize_per_tensor_default_93" [id=1579, type=quantize_per_tensor]; +"1580 dequantize_per_tensor_default_93" [id=1580, type=dequantize_per_tensor]; +"1581 linear_94_scale_0" [id=1581, type=get_attr]; +"1582 linear_94_zero_point_0" [id=1582, type=get_attr]; +"1583 _frozen_param99_1" [id=1583, type=get_attr]; +"1584 dequantize_per_channel_default_63" [id=1584, type=dequantize_per_channel]; +"1585 linear_94" [id=1585, type=linear]; +"1586 reshape_68" [id=1586, type=reshape]; +"1587 permute_70" [id=1587, type=permute]; +"1588 select_45" [id=1588, type=select]; +"1589 select_46" [id=1589, type=select]; +"1590 select_47" [id=1590, type=select]; +"1591 linalg_vector_norm_30" [id=1591, type=linalg_vector_norm]; +"1592 clamp_min_30" [id=1592, type=clamp_min]; +"1593 expand_as_30" [id=1593, type=expand_as]; +"1594 div_30" [id=1594, type=div]; +"1595 quantize_per_tensor_default_94" [id=1595, type=quantize_per_tensor]; +"1596 dequantize_per_tensor_default_94" [id=1596, type=dequantize_per_tensor]; +"1597 linalg_vector_norm_31" [id=1597, type=linalg_vector_norm]; +"1598 clamp_min_31" [id=1598, type=clamp_min]; +"1599 expand_as_31" [id=1599, type=expand_as]; +"1600 div_31" [id=1600, type=div]; +"1601 quantize_per_tensor_default_95" [id=1601, type=quantize_per_tensor]; +"1602 dequantize_per_tensor_default_95" [id=1602, type=dequantize_per_tensor]; +"1603 transpose_30" [id=1603, type=transpose]; +"1604 matmul_30" [id=1604, type=matmul]; +"1605 _frozen_param126" [id=1605, type=get_attr]; +"1606 mul_31" [id=1606, type=mul]; +"1607 add_52" [id=1607, type=add]; +"1608 new_zeros_7" [id=1608, type=new_zeros]; +"1609 view_84" [id=1609, type=view]; +"1610 permute_71" [id=1610, type=permute]; +"1611 reshape_69" [id=1611, type=reshape]; +"1612 unsqueeze_44" [id=1612, type=unsqueeze]; +"1613 unsqueeze_45" [id=1613, type=unsqueeze]; +"1614 sub_7" [id=1614, type=sub]; +"1615 ne_7" [id=1615, type=ne]; +"1616 masked_fill_14" [id=1616, type=masked_fill]; +"1617 eq_7" [id=1617, type=eq]; +"1618 masked_fill_15" [id=1618, type=masked_fill]; +"1619 view_85" [id=1619, type=view]; +"1620 unsqueeze_46" [id=1620, type=unsqueeze]; +"1621 unsqueeze_47" [id=1621, type=unsqueeze]; +"1622 add_53" [id=1622, type=add]; +"1623 view_86" [id=1623, type=view]; +"1624 softmax_15" [id=1624, type=softmax]; +"1625 dropout_60" [id=1625, type=dropout]; +"1626 matmul_31" [id=1626, type=matmul]; +"1627 transpose_31" [id=1627, type=transpose]; +"1628 reshape_70" [id=1628, type=reshape]; +"1629 reshape_70_0_0_nncf_smooth_quant_0" [id=1629, type=call_module]; +"1630 quantize_per_tensor_default_96" [id=1630, type=quantize_per_tensor]; +"1631 dequantize_per_tensor_default_96" [id=1631, type=dequantize_per_tensor]; +"1632 linear_95_scale_0" [id=1632, type=get_attr]; +"1633 linear_95_zero_point_0" [id=1633, type=get_attr]; +"1634 _frozen_param100_1" [id=1634, type=get_attr]; +"1635 dequantize_per_channel_default_64" [id=1635, type=dequantize_per_channel]; +"1636 linear_95" [id=1636, type=linear]; +"1637 dropout_61" [id=1637, type=dropout]; +"1638 view_87" [id=1638, type=view]; +"1639 permute_72" [id=1639, type=permute]; +"1640 reshape_71" [id=1640, type=reshape]; +"1641 roll_15" [id=1641, type=roll]; +"1642 slice_241" [id=1642, type=slice]; +"1643 slice_242" [id=1643, type=slice]; +"1644 slice_243" [id=1644, type=slice]; +"1645 slice_244" [id=1645, type=slice]; +"1646 contiguous_29" [id=1646, type=contiguous]; +"1647 layer_norm_33" [id=1647, type=layer_norm]; +"1648 add_54" [id=1648, type=add]; +"1649 add_54_0_0_nncf_smooth_quant_0" [id=1649, type=call_module]; +"1650 quantize_per_tensor_default_97" [id=1650, type=quantize_per_tensor]; +"1651 dequantize_per_tensor_default_97" [id=1651, type=dequantize_per_tensor]; +"1652 linear_96_scale_0" [id=1652, type=get_attr]; +"1653 linear_96_zero_point_0" [id=1653, type=get_attr]; +"1654 _frozen_param101_1" [id=1654, type=get_attr]; +"1655 dequantize_per_channel_default_65" [id=1655, type=dequantize_per_channel]; +"1656 linear_96" [id=1656, type=linear]; +"1657 gelu_15" [id=1657, type=gelu]; +"1658 dropout_62" [id=1658, type=dropout]; +"1659 dropout_62_0_0_nncf_smooth_quant_0" [id=1659, type=call_module]; +"1660 quantize_per_tensor_default_98" [id=1660, type=quantize_per_tensor]; +"1661 dequantize_per_tensor_default_98" [id=1661, type=dequantize_per_tensor]; +"1662 linear_97_scale_0" [id=1662, type=get_attr]; +"1663 linear_97_zero_point_0" [id=1663, type=get_attr]; +"1664 _frozen_param104_1" [id=1664, type=get_attr]; +"1665 dequantize_per_channel_default_66" [id=1665, type=dequantize_per_channel]; +"1666 linear_97" [id=1666, type=linear]; +"1667 dropout_63" [id=1667, type=dropout]; +"1668 layer_norm_34" [id=1668, type=layer_norm]; +"1669 add_55" [id=1669, type=add]; +"1670 _frozen_param136" [id=1670, type=get_attr]; +"1671 pad_18" [id=1671, type=pad]; +"1672 view_90" [id=1672, type=view]; +"1673 permute_74" [id=1673, type=permute]; +"1674 reshape_72" [id=1674, type=reshape]; +"1675 _frozen_param137" [id=1675, type=get_attr]; +"1676 reshape_72_0_0_nncf_smooth_quant_0" [id=1676, type=call_module]; +"1677 quantize_per_tensor_default_99" [id=1677, type=quantize_per_tensor]; +"1678 dequantize_per_tensor_default_99" [id=1678, type=dequantize_per_tensor]; +"1679 linear_100_scale_0" [id=1679, type=get_attr]; +"1680 linear_100_zero_point_0" [id=1680, type=get_attr]; +"1681 _frozen_param108_1" [id=1681, type=get_attr]; +"1682 dequantize_per_channel_default_67" [id=1682, type=dequantize_per_channel]; +"1683 linear_100" [id=1683, type=linear]; +"1684 reshape_73" [id=1684, type=reshape]; +"1685 permute_75" [id=1685, type=permute]; +"1686 select_48" [id=1686, type=select]; +"1687 select_49" [id=1687, type=select]; +"1688 select_50" [id=1688, type=select]; +"1689 linalg_vector_norm_32" [id=1689, type=linalg_vector_norm]; +"1690 clamp_min_32" [id=1690, type=clamp_min]; +"1691 expand_as_32" [id=1691, type=expand_as]; +"1692 div_32" [id=1692, type=div]; +"1693 quantize_per_tensor_default_100" [id=1693, type=quantize_per_tensor]; +"1694 dequantize_per_tensor_default_100" [id=1694, type=dequantize_per_tensor]; +"1695 linalg_vector_norm_33" [id=1695, type=linalg_vector_norm]; +"1696 clamp_min_33" [id=1696, type=clamp_min]; +"1697 expand_as_33" [id=1697, type=expand_as]; +"1698 div_33" [id=1698, type=div]; +"1699 quantize_per_tensor_default_101" [id=1699, type=quantize_per_tensor]; +"1700 dequantize_per_tensor_default_101" [id=1700, type=dequantize_per_tensor]; +"1701 transpose_32" [id=1701, type=transpose]; +"1702 matmul_32" [id=1702, type=matmul]; +"1703 _frozen_param139" [id=1703, type=get_attr]; +"1704 mul_33" [id=1704, type=mul]; +"1705 add_56" [id=1705, type=add]; +"1706 softmax_16" [id=1706, type=softmax]; +"1707 dropout_64" [id=1707, type=dropout]; +"1708 matmul_33" [id=1708, type=matmul]; +"1709 transpose_33" [id=1709, type=transpose]; +"1710 reshape_74" [id=1710, type=reshape]; +"1711 reshape_74_0_0_nncf_smooth_quant_0" [id=1711, type=call_module]; +"1712 quantize_per_tensor_default_102" [id=1712, type=quantize_per_tensor]; +"1713 dequantize_per_tensor_default_102" [id=1713, type=dequantize_per_tensor]; +"1714 linear_101_scale_0" [id=1714, type=get_attr]; +"1715 linear_101_zero_point_0" [id=1715, type=get_attr]; +"1716 _frozen_param110_1" [id=1716, type=get_attr]; +"1717 dequantize_per_channel_default_68" [id=1717, type=dequantize_per_channel]; +"1718 linear_101" [id=1718, type=linear]; +"1719 dropout_65" [id=1719, type=dropout]; +"1720 view_91" [id=1720, type=view]; +"1721 permute_76" [id=1721, type=permute]; +"1722 reshape_75" [id=1722, type=reshape]; +"1723 slice_246" [id=1723, type=slice]; +"1724 slice_247" [id=1724, type=slice]; +"1725 slice_248" [id=1725, type=slice]; +"1726 slice_249" [id=1726, type=slice]; +"1727 contiguous_31" [id=1727, type=contiguous]; +"1728 layer_norm_35" [id=1728, type=layer_norm]; +"1729 add_57" [id=1729, type=add]; +"1730 add_57_0_0_nncf_smooth_quant_0" [id=1730, type=call_module]; +"1731 quantize_per_tensor_default_103" [id=1731, type=quantize_per_tensor]; +"1732 dequantize_per_tensor_default_103" [id=1732, type=dequantize_per_tensor]; +"1733 linear_102_scale_0" [id=1733, type=get_attr]; +"1734 linear_102_zero_point_0" [id=1734, type=get_attr]; +"1735 _frozen_param111_1" [id=1735, type=get_attr]; +"1736 dequantize_per_channel_default_69" [id=1736, type=dequantize_per_channel]; +"1737 linear_102" [id=1737, type=linear]; +"1738 gelu_16" [id=1738, type=gelu]; +"1739 dropout_66" [id=1739, type=dropout]; +"1740 dropout_66_0_0_nncf_smooth_quant_0" [id=1740, type=call_module]; +"1741 quantize_per_tensor_default_104" [id=1741, type=quantize_per_tensor]; +"1742 dequantize_per_tensor_default_104" [id=1742, type=dequantize_per_tensor]; +"1743 linear_103_scale_0" [id=1743, type=get_attr]; +"1744 linear_103_zero_point_0" [id=1744, type=get_attr]; +"1745 _frozen_param112_1" [id=1745, type=get_attr]; +"1746 dequantize_per_channel_default_70" [id=1746, type=dequantize_per_channel]; +"1747 linear_103" [id=1747, type=linear]; +"1748 dropout_67" [id=1748, type=dropout]; +"1749 layer_norm_36" [id=1749, type=layer_norm]; +"1750 add_58" [id=1750, type=add]; +"1751 _frozen_param140" [id=1751, type=get_attr]; +"1752 pad_19" [id=1752, type=pad]; +"1753 roll_16" [id=1753, type=roll]; +"1754 view_94" [id=1754, type=view]; +"1755 permute_78" [id=1755, type=permute]; +"1756 reshape_76" [id=1756, type=reshape]; +"1757 _frozen_param141" [id=1757, type=get_attr]; +"1758 reshape_76_0_0_nncf_smooth_quant_0" [id=1758, type=call_module]; +"1759 quantize_per_tensor_default_105" [id=1759, type=quantize_per_tensor]; +"1760 dequantize_per_tensor_default_105" [id=1760, type=dequantize_per_tensor]; +"1761 linear_106_scale_0" [id=1761, type=get_attr]; +"1762 linear_106_zero_point_0" [id=1762, type=get_attr]; +"1763 _frozen_param113_1" [id=1763, type=get_attr]; +"1764 dequantize_per_channel_default_71" [id=1764, type=dequantize_per_channel]; +"1765 linear_106" [id=1765, type=linear]; +"1766 reshape_77" [id=1766, type=reshape]; +"1767 permute_79" [id=1767, type=permute]; +"1768 select_51" [id=1768, type=select]; +"1769 select_52" [id=1769, type=select]; +"1770 select_53" [id=1770, type=select]; +"1771 linalg_vector_norm_34" [id=1771, type=linalg_vector_norm]; +"1772 clamp_min_34" [id=1772, type=clamp_min]; +"1773 expand_as_34" [id=1773, type=expand_as]; +"1774 div_34" [id=1774, type=div]; +"1775 quantize_per_tensor_default_106" [id=1775, type=quantize_per_tensor]; +"1776 dequantize_per_tensor_default_106" [id=1776, type=dequantize_per_tensor]; +"1777 linalg_vector_norm_35" [id=1777, type=linalg_vector_norm]; +"1778 clamp_min_35" [id=1778, type=clamp_min]; +"1779 expand_as_35" [id=1779, type=expand_as]; +"1780 div_35" [id=1780, type=div]; +"1781 quantize_per_tensor_default_107" [id=1781, type=quantize_per_tensor]; +"1782 dequantize_per_tensor_default_107" [id=1782, type=dequantize_per_tensor]; +"1783 transpose_34" [id=1783, type=transpose]; +"1784 matmul_34" [id=1784, type=matmul]; +"1785 _frozen_param143" [id=1785, type=get_attr]; +"1786 mul_35" [id=1786, type=mul]; +"1787 add_59" [id=1787, type=add]; +"1788 new_zeros_8" [id=1788, type=new_zeros]; +"1789 view_95" [id=1789, type=view]; +"1790 permute_80" [id=1790, type=permute]; +"1791 reshape_78" [id=1791, type=reshape]; +"1792 unsqueeze_50" [id=1792, type=unsqueeze]; +"1793 unsqueeze_51" [id=1793, type=unsqueeze]; +"1794 sub_8" [id=1794, type=sub]; +"1795 ne_8" [id=1795, type=ne]; +"1796 masked_fill_16" [id=1796, type=masked_fill]; +"1797 eq_8" [id=1797, type=eq]; +"1798 masked_fill_17" [id=1798, type=masked_fill]; +"1799 view_96" [id=1799, type=view]; +"1800 unsqueeze_52" [id=1800, type=unsqueeze]; +"1801 unsqueeze_53" [id=1801, type=unsqueeze]; +"1802 add_60" [id=1802, type=add]; +"1803 view_97" [id=1803, type=view]; +"1804 softmax_17" [id=1804, type=softmax]; +"1805 dropout_68" [id=1805, type=dropout]; +"1806 matmul_35" [id=1806, type=matmul]; +"1807 transpose_35" [id=1807, type=transpose]; +"1808 reshape_79" [id=1808, type=reshape]; +"1809 reshape_79_0_0_nncf_smooth_quant_0" [id=1809, type=call_module]; +"1810 quantize_per_tensor_default_108" [id=1810, type=quantize_per_tensor]; +"1811 dequantize_per_tensor_default_108" [id=1811, type=dequantize_per_tensor]; +"1812 linear_107_scale_0" [id=1812, type=get_attr]; +"1813 linear_107_zero_point_0" [id=1813, type=get_attr]; +"1814 _frozen_param114_1" [id=1814, type=get_attr]; +"1815 dequantize_per_channel_default_72" [id=1815, type=dequantize_per_channel]; +"1816 linear_107" [id=1816, type=linear]; +"1817 dropout_69" [id=1817, type=dropout]; +"1818 view_98" [id=1818, type=view]; +"1819 permute_81" [id=1819, type=permute]; +"1820 reshape_80" [id=1820, type=reshape]; +"1821 roll_17" [id=1821, type=roll]; +"1822 slice_269" [id=1822, type=slice]; +"1823 slice_270" [id=1823, type=slice]; +"1824 slice_271" [id=1824, type=slice]; +"1825 slice_272" [id=1825, type=slice]; +"1826 contiguous_33" [id=1826, type=contiguous]; +"1827 layer_norm_37" [id=1827, type=layer_norm]; +"1828 add_61" [id=1828, type=add]; +"1829 add_61_0_0_nncf_smooth_quant_0" [id=1829, type=call_module]; +"1830 quantize_per_tensor_default_109" [id=1830, type=quantize_per_tensor]; +"1831 dequantize_per_tensor_default_109" [id=1831, type=dequantize_per_tensor]; +"1832 linear_108_scale_0" [id=1832, type=get_attr]; +"1833 linear_108_zero_point_0" [id=1833, type=get_attr]; +"1834 _frozen_param115_1" [id=1834, type=get_attr]; +"1835 dequantize_per_channel_default_73" [id=1835, type=dequantize_per_channel]; +"1836 linear_108" [id=1836, type=linear]; +"1837 gelu_17" [id=1837, type=gelu]; +"1838 dropout_70" [id=1838, type=dropout]; +"1839 dropout_70_0_0_nncf_smooth_quant_0" [id=1839, type=call_module]; +"1840 quantize_per_tensor_default_110" [id=1840, type=quantize_per_tensor]; +"1841 dequantize_per_tensor_default_110" [id=1841, type=dequantize_per_tensor]; +"1842 linear_109_scale_0" [id=1842, type=get_attr]; +"1843 linear_109_zero_point_0" [id=1843, type=get_attr]; +"1844 _frozen_param116_1" [id=1844, type=get_attr]; +"1845 dequantize_per_channel_default_74" [id=1845, type=dequantize_per_channel]; +"1846 linear_109" [id=1846, type=linear]; +"1847 dropout_71" [id=1847, type=dropout]; +"1848 layer_norm_38" [id=1848, type=layer_norm]; +"1849 add_62" [id=1849, type=add]; +"1850 _frozen_param153" [id=1850, type=get_attr]; +"1851 pad_20" [id=1851, type=pad]; +"1852 view_101" [id=1852, type=view]; +"1853 permute_83" [id=1853, type=permute]; +"1854 reshape_81" [id=1854, type=reshape]; +"1855 _frozen_param154" [id=1855, type=get_attr]; +"1856 reshape_81_0_0_nncf_smooth_quant_0" [id=1856, type=call_module]; +"1857 quantize_per_tensor_default_111" [id=1857, type=quantize_per_tensor]; +"1858 dequantize_per_tensor_default_111" [id=1858, type=dequantize_per_tensor]; +"1859 linear_112_scale_0" [id=1859, type=get_attr]; +"1860 linear_112_zero_point_0" [id=1860, type=get_attr]; +"1861 _frozen_param117_1" [id=1861, type=get_attr]; +"1862 dequantize_per_channel_default_75" [id=1862, type=dequantize_per_channel]; +"1863 linear_112" [id=1863, type=linear]; +"1864 reshape_82" [id=1864, type=reshape]; +"1865 permute_84" [id=1865, type=permute]; +"1866 select_54" [id=1866, type=select]; +"1867 select_55" [id=1867, type=select]; +"1868 select_56" [id=1868, type=select]; +"1869 linalg_vector_norm_36" [id=1869, type=linalg_vector_norm]; +"1870 clamp_min_36" [id=1870, type=clamp_min]; +"1871 expand_as_36" [id=1871, type=expand_as]; +"1872 div_36" [id=1872, type=div]; +"1873 quantize_per_tensor_default_112" [id=1873, type=quantize_per_tensor]; +"1874 dequantize_per_tensor_default_112" [id=1874, type=dequantize_per_tensor]; +"1875 linalg_vector_norm_37" [id=1875, type=linalg_vector_norm]; +"1876 clamp_min_37" [id=1876, type=clamp_min]; +"1877 expand_as_37" [id=1877, type=expand_as]; +"1878 div_37" [id=1878, type=div]; +"1879 quantize_per_tensor_default_113" [id=1879, type=quantize_per_tensor]; +"1880 dequantize_per_tensor_default_113" [id=1880, type=dequantize_per_tensor]; +"1881 transpose_36" [id=1881, type=transpose]; +"1882 matmul_36" [id=1882, type=matmul]; +"1883 _frozen_param156" [id=1883, type=get_attr]; +"1884 mul_37" [id=1884, type=mul]; +"1885 add_63" [id=1885, type=add]; +"1886 softmax_18" [id=1886, type=softmax]; +"1887 dropout_72" [id=1887, type=dropout]; +"1888 matmul_37" [id=1888, type=matmul]; +"1889 transpose_37" [id=1889, type=transpose]; +"1890 reshape_83" [id=1890, type=reshape]; +"1891 reshape_83_0_0_nncf_smooth_quant_0" [id=1891, type=call_module]; +"1892 quantize_per_tensor_default_114" [id=1892, type=quantize_per_tensor]; +"1893 dequantize_per_tensor_default_114" [id=1893, type=dequantize_per_tensor]; +"1894 linear_113_scale_0" [id=1894, type=get_attr]; +"1895 linear_113_zero_point_0" [id=1895, type=get_attr]; +"1896 _frozen_param118_1" [id=1896, type=get_attr]; +"1897 dequantize_per_channel_default_76" [id=1897, type=dequantize_per_channel]; +"1898 linear_113" [id=1898, type=linear]; +"1899 dropout_73" [id=1899, type=dropout]; +"1900 view_102" [id=1900, type=view]; +"1901 permute_85" [id=1901, type=permute]; +"1902 reshape_84" [id=1902, type=reshape]; +"1903 slice_274" [id=1903, type=slice]; +"1904 slice_275" [id=1904, type=slice]; +"1905 slice_276" [id=1905, type=slice]; +"1906 slice_277" [id=1906, type=slice]; +"1907 contiguous_35" [id=1907, type=contiguous]; +"1908 layer_norm_39" [id=1908, type=layer_norm]; +"1909 add_64" [id=1909, type=add]; +"1910 add_64_0_0_nncf_smooth_quant_0" [id=1910, type=call_module]; +"1911 quantize_per_tensor_default_115" [id=1911, type=quantize_per_tensor]; +"1912 dequantize_per_tensor_default_115" [id=1912, type=dequantize_per_tensor]; +"1913 linear_114_scale_0" [id=1913, type=get_attr]; +"1914 linear_114_zero_point_0" [id=1914, type=get_attr]; +"1915 _frozen_param121_1" [id=1915, type=get_attr]; +"1916 dequantize_per_channel_default_77" [id=1916, type=dequantize_per_channel]; +"1917 linear_114" [id=1917, type=linear]; +"1918 gelu_18" [id=1918, type=gelu]; +"1919 dropout_74" [id=1919, type=dropout]; +"1920 dropout_74_0_0_nncf_smooth_quant_0" [id=1920, type=call_module]; +"1921 quantize_per_tensor_default_116" [id=1921, type=quantize_per_tensor]; +"1922 dequantize_per_tensor_default_116" [id=1922, type=dequantize_per_tensor]; +"1923 linear_115_scale_0" [id=1923, type=get_attr]; +"1924 linear_115_zero_point_0" [id=1924, type=get_attr]; +"1925 _frozen_param125_1" [id=1925, type=get_attr]; +"1926 dequantize_per_channel_default_78" [id=1926, type=dequantize_per_channel]; +"1927 linear_115" [id=1927, type=linear]; +"1928 dropout_75" [id=1928, type=dropout]; +"1929 layer_norm_40" [id=1929, type=layer_norm]; +"1930 add_65" [id=1930, type=add]; +"1931 _frozen_param157" [id=1931, type=get_attr]; +"1932 pad_21" [id=1932, type=pad]; +"1933 roll_18" [id=1933, type=roll]; +"1934 view_105" [id=1934, type=view]; +"1935 permute_87" [id=1935, type=permute]; +"1936 reshape_85" [id=1936, type=reshape]; +"1937 _frozen_param158" [id=1937, type=get_attr]; +"1938 reshape_85_0_0_nncf_smooth_quant_0" [id=1938, type=call_module]; +"1939 quantize_per_tensor_default_117" [id=1939, type=quantize_per_tensor]; +"1940 dequantize_per_tensor_default_117" [id=1940, type=dequantize_per_tensor]; +"1941 linear_118_scale_0" [id=1941, type=get_attr]; +"1942 linear_118_zero_point_0" [id=1942, type=get_attr]; +"1943 _frozen_param127_1" [id=1943, type=get_attr]; +"1944 dequantize_per_channel_default_79" [id=1944, type=dequantize_per_channel]; +"1945 linear_118" [id=1945, type=linear]; +"1946 reshape_86" [id=1946, type=reshape]; +"1947 permute_88" [id=1947, type=permute]; +"1948 select_57" [id=1948, type=select]; +"1949 select_58" [id=1949, type=select]; +"1950 select_59" [id=1950, type=select]; +"1951 linalg_vector_norm_38" [id=1951, type=linalg_vector_norm]; +"1952 clamp_min_38" [id=1952, type=clamp_min]; +"1953 expand_as_38" [id=1953, type=expand_as]; +"1954 div_38" [id=1954, type=div]; +"1955 quantize_per_tensor_default_118" [id=1955, type=quantize_per_tensor]; +"1956 dequantize_per_tensor_default_118" [id=1956, type=dequantize_per_tensor]; +"1957 linalg_vector_norm_39" [id=1957, type=linalg_vector_norm]; +"1958 clamp_min_39" [id=1958, type=clamp_min]; +"1959 expand_as_39" [id=1959, type=expand_as]; +"1960 div_39" [id=1960, type=div]; +"1961 quantize_per_tensor_default_119" [id=1961, type=quantize_per_tensor]; +"1962 dequantize_per_tensor_default_119" [id=1962, type=dequantize_per_tensor]; +"1963 transpose_38" [id=1963, type=transpose]; +"1964 matmul_38" [id=1964, type=matmul]; +"1965 _frozen_param160" [id=1965, type=get_attr]; +"1966 mul_39" [id=1966, type=mul]; +"1967 add_66" [id=1967, type=add]; +"1968 new_zeros_9" [id=1968, type=new_zeros]; +"1969 view_106" [id=1969, type=view]; +"1970 permute_89" [id=1970, type=permute]; +"1971 reshape_87" [id=1971, type=reshape]; +"1972 unsqueeze_56" [id=1972, type=unsqueeze]; +"1973 unsqueeze_57" [id=1973, type=unsqueeze]; +"1974 sub_9" [id=1974, type=sub]; +"1975 ne_9" [id=1975, type=ne]; +"1976 masked_fill_18" [id=1976, type=masked_fill]; +"1977 eq_9" [id=1977, type=eq]; +"1978 masked_fill_19" [id=1978, type=masked_fill]; +"1979 view_107" [id=1979, type=view]; +"1980 unsqueeze_58" [id=1980, type=unsqueeze]; +"1981 unsqueeze_59" [id=1981, type=unsqueeze]; +"1982 add_67" [id=1982, type=add]; +"1983 view_108" [id=1983, type=view]; +"1984 softmax_19" [id=1984, type=softmax]; +"1985 dropout_76" [id=1985, type=dropout]; +"1986 matmul_39" [id=1986, type=matmul]; +"1987 transpose_39" [id=1987, type=transpose]; +"1988 reshape_88" [id=1988, type=reshape]; +"1989 reshape_88_0_0_nncf_smooth_quant_0" [id=1989, type=call_module]; +"1990 quantize_per_tensor_default_120" [id=1990, type=quantize_per_tensor]; +"1991 dequantize_per_tensor_default_120" [id=1991, type=dequantize_per_tensor]; +"1992 linear_119_scale_0" [id=1992, type=get_attr]; +"1993 linear_119_zero_point_0" [id=1993, type=get_attr]; +"1994 _frozen_param128_1" [id=1994, type=get_attr]; +"1995 dequantize_per_channel_default_80" [id=1995, type=dequantize_per_channel]; +"1996 linear_119" [id=1996, type=linear]; +"1997 dropout_77" [id=1997, type=dropout]; +"1998 view_109" [id=1998, type=view]; +"1999 permute_90" [id=1999, type=permute]; +"2000 reshape_89" [id=2000, type=reshape]; +"2001 roll_19" [id=2001, type=roll]; +"2002 slice_297" [id=2002, type=slice]; +"2003 slice_298" [id=2003, type=slice]; +"2004 slice_299" [id=2004, type=slice]; +"2005 slice_300" [id=2005, type=slice]; +"2006 contiguous_37" [id=2006, type=contiguous]; +"2007 layer_norm_41" [id=2007, type=layer_norm]; +"2008 add_68" [id=2008, type=add]; +"2009 add_68_0_0_nncf_smooth_quant_0" [id=2009, type=call_module]; +"2010 quantize_per_tensor_default_121" [id=2010, type=quantize_per_tensor]; +"2011 dequantize_per_tensor_default_121" [id=2011, type=dequantize_per_tensor]; +"2012 linear_120_scale_0" [id=2012, type=get_attr]; +"2013 linear_120_zero_point_0" [id=2013, type=get_attr]; +"2014 _frozen_param129_1" [id=2014, type=get_attr]; +"2015 dequantize_per_channel_default_81" [id=2015, type=dequantize_per_channel]; +"2016 linear_120" [id=2016, type=linear]; +"2017 gelu_19" [id=2017, type=gelu]; +"2018 dropout_78" [id=2018, type=dropout]; +"2019 dropout_78_0_0_nncf_smooth_quant_0" [id=2019, type=call_module]; +"2020 quantize_per_tensor_default_122" [id=2020, type=quantize_per_tensor]; +"2021 dequantize_per_tensor_default_122" [id=2021, type=dequantize_per_tensor]; +"2022 linear_121_scale_0" [id=2022, type=get_attr]; +"2023 linear_121_zero_point_0" [id=2023, type=get_attr]; +"2024 _frozen_param130_1" [id=2024, type=get_attr]; +"2025 dequantize_per_channel_default_82" [id=2025, type=dequantize_per_channel]; +"2026 linear_121" [id=2026, type=linear]; +"2027 dropout_79" [id=2027, type=dropout]; +"2028 layer_norm_42" [id=2028, type=layer_norm]; +"2029 add_69" [id=2029, type=add]; +"2030 _frozen_param170" [id=2030, type=get_attr]; +"2031 pad_22" [id=2031, type=pad]; +"2032 view_112" [id=2032, type=view]; +"2033 permute_92" [id=2033, type=permute]; +"2034 reshape_90" [id=2034, type=reshape]; +"2035 _frozen_param171" [id=2035, type=get_attr]; +"2036 reshape_90_0_0_nncf_smooth_quant_0" [id=2036, type=call_module]; +"2037 quantize_per_tensor_default_123" [id=2037, type=quantize_per_tensor]; +"2038 dequantize_per_tensor_default_123" [id=2038, type=dequantize_per_tensor]; +"2039 linear_124_scale_0" [id=2039, type=get_attr]; +"2040 linear_124_zero_point_0" [id=2040, type=get_attr]; +"2041 _frozen_param131_1" [id=2041, type=get_attr]; +"2042 dequantize_per_channel_default_83" [id=2042, type=dequantize_per_channel]; +"2043 linear_124" [id=2043, type=linear]; +"2044 reshape_91" [id=2044, type=reshape]; +"2045 permute_93" [id=2045, type=permute]; +"2046 select_60" [id=2046, type=select]; +"2047 select_61" [id=2047, type=select]; +"2048 select_62" [id=2048, type=select]; +"2049 linalg_vector_norm_40" [id=2049, type=linalg_vector_norm]; +"2050 clamp_min_40" [id=2050, type=clamp_min]; +"2051 expand_as_40" [id=2051, type=expand_as]; +"2052 div_40" [id=2052, type=div]; +"2053 quantize_per_tensor_default_124" [id=2053, type=quantize_per_tensor]; +"2054 dequantize_per_tensor_default_124" [id=2054, type=dequantize_per_tensor]; +"2055 linalg_vector_norm_41" [id=2055, type=linalg_vector_norm]; +"2056 clamp_min_41" [id=2056, type=clamp_min]; +"2057 expand_as_41" [id=2057, type=expand_as]; +"2058 div_41" [id=2058, type=div]; +"2059 quantize_per_tensor_default_125" [id=2059, type=quantize_per_tensor]; +"2060 dequantize_per_tensor_default_125" [id=2060, type=dequantize_per_tensor]; +"2061 transpose_40" [id=2061, type=transpose]; +"2062 matmul_40" [id=2062, type=matmul]; +"2063 _frozen_param173" [id=2063, type=get_attr]; +"2064 mul_41" [id=2064, type=mul]; +"2065 add_70" [id=2065, type=add]; +"2066 softmax_20" [id=2066, type=softmax]; +"2067 dropout_80" [id=2067, type=dropout]; +"2068 matmul_41" [id=2068, type=matmul]; +"2069 transpose_41" [id=2069, type=transpose]; +"2070 reshape_92" [id=2070, type=reshape]; +"2071 reshape_92_0_0_nncf_smooth_quant_0" [id=2071, type=call_module]; +"2072 quantize_per_tensor_default_126" [id=2072, type=quantize_per_tensor]; +"2073 dequantize_per_tensor_default_126" [id=2073, type=dequantize_per_tensor]; +"2074 linear_125_scale_0" [id=2074, type=get_attr]; +"2075 linear_125_zero_point_0" [id=2075, type=get_attr]; +"2076 _frozen_param132_1" [id=2076, type=get_attr]; +"2077 dequantize_per_channel_default_84" [id=2077, type=dequantize_per_channel]; +"2078 linear_125" [id=2078, type=linear]; +"2079 dropout_81" [id=2079, type=dropout]; +"2080 view_113" [id=2080, type=view]; +"2081 permute_94" [id=2081, type=permute]; +"2082 reshape_93" [id=2082, type=reshape]; +"2083 slice_302" [id=2083, type=slice]; +"2084 slice_303" [id=2084, type=slice]; +"2085 slice_304" [id=2085, type=slice]; +"2086 slice_305" [id=2086, type=slice]; +"2087 contiguous_39" [id=2087, type=contiguous]; +"2088 layer_norm_43" [id=2088, type=layer_norm]; +"2089 add_71" [id=2089, type=add]; +"2090 add_71_0_0_nncf_smooth_quant_0" [id=2090, type=call_module]; +"2091 quantize_per_tensor_default_127" [id=2091, type=quantize_per_tensor]; +"2092 dequantize_per_tensor_default_127" [id=2092, type=dequantize_per_tensor]; +"2093 linear_126_scale_0" [id=2093, type=get_attr]; +"2094 linear_126_zero_point_0" [id=2094, type=get_attr]; +"2095 _frozen_param133_1" [id=2095, type=get_attr]; +"2096 dequantize_per_channel_default_85" [id=2096, type=dequantize_per_channel]; +"2097 linear_126" [id=2097, type=linear]; +"2098 gelu_20" [id=2098, type=gelu]; +"2099 dropout_82" [id=2099, type=dropout]; +"2100 dropout_82_0_0_nncf_smooth_quant_0" [id=2100, type=call_module]; +"2101 quantize_per_tensor_default_128" [id=2101, type=quantize_per_tensor]; +"2102 dequantize_per_tensor_default_128" [id=2102, type=dequantize_per_tensor]; +"2103 linear_127_scale_0" [id=2103, type=get_attr]; +"2104 linear_127_zero_point_0" [id=2104, type=get_attr]; +"2105 _frozen_param134_1" [id=2105, type=get_attr]; +"2106 dequantize_per_channel_default_86" [id=2106, type=dequantize_per_channel]; +"2107 linear_127" [id=2107, type=linear]; +"2108 dropout_83" [id=2108, type=dropout]; +"2109 layer_norm_44" [id=2109, type=layer_norm]; +"2110 add_72" [id=2110, type=add]; +"2111 _frozen_param174" [id=2111, type=get_attr]; +"2112 pad_23" [id=2112, type=pad]; +"2113 roll_20" [id=2113, type=roll]; +"2114 view_116" [id=2114, type=view]; +"2115 permute_96" [id=2115, type=permute]; +"2116 reshape_94" [id=2116, type=reshape]; +"2117 _frozen_param175" [id=2117, type=get_attr]; +"2118 reshape_94_0_0_nncf_smooth_quant_0" [id=2118, type=call_module]; +"2119 quantize_per_tensor_default_129" [id=2119, type=quantize_per_tensor]; +"2120 dequantize_per_tensor_default_129" [id=2120, type=dequantize_per_tensor]; +"2121 linear_130_scale_0" [id=2121, type=get_attr]; +"2122 linear_130_zero_point_0" [id=2122, type=get_attr]; +"2123 _frozen_param135_1" [id=2123, type=get_attr]; +"2124 dequantize_per_channel_default_87" [id=2124, type=dequantize_per_channel]; +"2125 linear_130" [id=2125, type=linear]; +"2126 reshape_95" [id=2126, type=reshape]; +"2127 permute_97" [id=2127, type=permute]; +"2128 select_63" [id=2128, type=select]; +"2129 select_64" [id=2129, type=select]; +"2130 select_65" [id=2130, type=select]; +"2131 linalg_vector_norm_42" [id=2131, type=linalg_vector_norm]; +"2132 clamp_min_42" [id=2132, type=clamp_min]; +"2133 expand_as_42" [id=2133, type=expand_as]; +"2134 div_42" [id=2134, type=div]; +"2135 quantize_per_tensor_default_130" [id=2135, type=quantize_per_tensor]; +"2136 dequantize_per_tensor_default_130" [id=2136, type=dequantize_per_tensor]; +"2137 linalg_vector_norm_43" [id=2137, type=linalg_vector_norm]; +"2138 clamp_min_43" [id=2138, type=clamp_min]; +"2139 expand_as_43" [id=2139, type=expand_as]; +"2140 div_43" [id=2140, type=div]; +"2141 quantize_per_tensor_default_131" [id=2141, type=quantize_per_tensor]; +"2142 dequantize_per_tensor_default_131" [id=2142, type=dequantize_per_tensor]; +"2143 transpose_42" [id=2143, type=transpose]; +"2144 matmul_42" [id=2144, type=matmul]; +"2145 _frozen_param177" [id=2145, type=get_attr]; +"2146 mul_43" [id=2146, type=mul]; +"2147 add_73" [id=2147, type=add]; +"2148 new_zeros_10" [id=2148, type=new_zeros]; +"2149 view_117" [id=2149, type=view]; +"2150 permute_98" [id=2150, type=permute]; +"2151 reshape_96" [id=2151, type=reshape]; +"2152 unsqueeze_62" [id=2152, type=unsqueeze]; +"2153 unsqueeze_63" [id=2153, type=unsqueeze]; +"2154 sub_10" [id=2154, type=sub]; +"2155 ne_10" [id=2155, type=ne]; +"2156 masked_fill_20" [id=2156, type=masked_fill]; +"2157 eq_10" [id=2157, type=eq]; +"2158 masked_fill_21" [id=2158, type=masked_fill]; +"2159 view_118" [id=2159, type=view]; +"2160 unsqueeze_64" [id=2160, type=unsqueeze]; +"2161 unsqueeze_65" [id=2161, type=unsqueeze]; +"2162 add_74" [id=2162, type=add]; +"2163 view_119" [id=2163, type=view]; +"2164 softmax_21" [id=2164, type=softmax]; +"2165 dropout_84" [id=2165, type=dropout]; +"2166 matmul_43" [id=2166, type=matmul]; +"2167 transpose_43" [id=2167, type=transpose]; +"2168 reshape_97" [id=2168, type=reshape]; +"2169 reshape_97_0_0_nncf_smooth_quant_0" [id=2169, type=call_module]; +"2170 quantize_per_tensor_default_132" [id=2170, type=quantize_per_tensor]; +"2171 dequantize_per_tensor_default_132" [id=2171, type=dequantize_per_tensor]; +"2172 linear_131_scale_0" [id=2172, type=get_attr]; +"2173 linear_131_zero_point_0" [id=2173, type=get_attr]; +"2174 _frozen_param138_1" [id=2174, type=get_attr]; +"2175 dequantize_per_channel_default_88" [id=2175, type=dequantize_per_channel]; +"2176 linear_131" [id=2176, type=linear]; +"2177 dropout_85" [id=2177, type=dropout]; +"2178 view_120" [id=2178, type=view]; +"2179 permute_99" [id=2179, type=permute]; +"2180 reshape_98" [id=2180, type=reshape]; +"2181 roll_21" [id=2181, type=roll]; +"2182 slice_325" [id=2182, type=slice]; +"2183 slice_326" [id=2183, type=slice]; +"2184 slice_327" [id=2184, type=slice]; +"2185 slice_328" [id=2185, type=slice]; +"2186 contiguous_41" [id=2186, type=contiguous]; +"2187 layer_norm_45" [id=2187, type=layer_norm]; +"2188 add_75" [id=2188, type=add]; +"2189 add_75_0_0_nncf_smooth_quant_0" [id=2189, type=call_module]; +"2190 quantize_per_tensor_default_133" [id=2190, type=quantize_per_tensor]; +"2191 dequantize_per_tensor_default_133" [id=2191, type=dequantize_per_tensor]; +"2192 linear_132_scale_0" [id=2192, type=get_attr]; +"2193 linear_132_zero_point_0" [id=2193, type=get_attr]; +"2194 _frozen_param142_1" [id=2194, type=get_attr]; +"2195 dequantize_per_channel_default_89" [id=2195, type=dequantize_per_channel]; +"2196 linear_132" [id=2196, type=linear]; +"2197 gelu_21" [id=2197, type=gelu]; +"2198 dropout_86" [id=2198, type=dropout]; +"2199 dropout_86_0_0_nncf_smooth_quant_0" [id=2199, type=call_module]; +"2200 quantize_per_tensor_default_134" [id=2200, type=quantize_per_tensor]; +"2201 dequantize_per_tensor_default_134" [id=2201, type=dequantize_per_tensor]; +"2202 linear_133_scale_0" [id=2202, type=get_attr]; +"2203 linear_133_zero_point_0" [id=2203, type=get_attr]; +"2204 _frozen_param144_1" [id=2204, type=get_attr]; +"2205 dequantize_per_channel_default_90" [id=2205, type=dequantize_per_channel]; +"2206 linear_133" [id=2206, type=linear]; +"2207 dropout_87" [id=2207, type=dropout]; +"2208 layer_norm_46" [id=2208, type=layer_norm]; +"2209 add_76" [id=2209, type=add]; +"2210 pad_24" [id=2210, type=pad]; +"2211 slice_329" [id=2211, type=slice]; +"2212 slice_330" [id=2212, type=slice]; +"2213 slice_331" [id=2213, type=slice]; +"2214 slice_332" [id=2214, type=slice]; +"2215 slice_333" [id=2215, type=slice]; +"2216 slice_334" [id=2216, type=slice]; +"2217 slice_335" [id=2217, type=slice]; +"2218 slice_336" [id=2218, type=slice]; +"2219 slice_337" [id=2219, type=slice]; +"2220 slice_338" [id=2220, type=slice]; +"2221 slice_339" [id=2221, type=slice]; +"2222 slice_340" [id=2222, type=slice]; +"2223 cat_2" [id=2223, type=cat]; +"2224 cat_2_0_0_nncf_smooth_quant_0" [id=2224, type=call_module]; +"2225 quantize_per_tensor_default_135" [id=2225, type=quantize_per_tensor]; +"2226 dequantize_per_tensor_default_135" [id=2226, type=dequantize_per_tensor]; +"2227 linear_134_scale_0" [id=2227, type=get_attr]; +"2228 linear_134_zero_point_0" [id=2228, type=get_attr]; +"2229 _frozen_param145_1" [id=2229, type=get_attr]; +"2230 dequantize_per_channel_default_91" [id=2230, type=dequantize_per_channel]; +"2231 linear_134" [id=2231, type=linear]; +"2232 layer_norm_47" [id=2232, type=layer_norm]; +"2233 _frozen_param187" [id=2233, type=get_attr]; +"2234 pad_25" [id=2234, type=pad]; +"2235 view_123" [id=2235, type=view]; +"2236 permute_101" [id=2236, type=permute]; +"2237 reshape_99" [id=2237, type=reshape]; +"2238 _frozen_param188" [id=2238, type=get_attr]; +"2239 reshape_99_0_0_nncf_smooth_quant_0" [id=2239, type=call_module]; +"2240 quantize_per_tensor_default_136" [id=2240, type=quantize_per_tensor]; +"2241 dequantize_per_tensor_default_136" [id=2241, type=dequantize_per_tensor]; +"2242 linear_137_scale_0" [id=2242, type=get_attr]; +"2243 linear_137_zero_point_0" [id=2243, type=get_attr]; +"2244 _frozen_param146_1" [id=2244, type=get_attr]; +"2245 dequantize_per_channel_default_92" [id=2245, type=dequantize_per_channel]; +"2246 linear_137" [id=2246, type=linear]; +"2247 reshape_100" [id=2247, type=reshape]; +"2248 permute_102" [id=2248, type=permute]; +"2249 select_66" [id=2249, type=select]; +"2250 select_67" [id=2250, type=select]; +"2251 select_68" [id=2251, type=select]; +"2252 linalg_vector_norm_44" [id=2252, type=linalg_vector_norm]; +"2253 clamp_min_44" [id=2253, type=clamp_min]; +"2254 expand_as_44" [id=2254, type=expand_as]; +"2255 div_44" [id=2255, type=div]; +"2256 quantize_per_tensor_default_137" [id=2256, type=quantize_per_tensor]; +"2257 dequantize_per_tensor_default_137" [id=2257, type=dequantize_per_tensor]; +"2258 linalg_vector_norm_45" [id=2258, type=linalg_vector_norm]; +"2259 clamp_min_45" [id=2259, type=clamp_min]; +"2260 expand_as_45" [id=2260, type=expand_as]; +"2261 div_45" [id=2261, type=div]; +"2262 quantize_per_tensor_default_138" [id=2262, type=quantize_per_tensor]; +"2263 dequantize_per_tensor_default_138" [id=2263, type=dequantize_per_tensor]; +"2264 transpose_44" [id=2264, type=transpose]; +"2265 matmul_44" [id=2265, type=matmul]; +"2266 _frozen_param190" [id=2266, type=get_attr]; +"2267 mul_45" [id=2267, type=mul]; +"2268 add_77" [id=2268, type=add]; +"2269 softmax_22" [id=2269, type=softmax]; +"2270 dropout_88" [id=2270, type=dropout]; +"2271 matmul_45" [id=2271, type=matmul]; +"2272 transpose_45" [id=2272, type=transpose]; +"2273 reshape_101" [id=2273, type=reshape]; +"2274 reshape_101_0_0_nncf_smooth_quant_0" [id=2274, type=call_module]; +"2275 quantize_per_tensor_default_139" [id=2275, type=quantize_per_tensor]; +"2276 dequantize_per_tensor_default_139" [id=2276, type=dequantize_per_tensor]; +"2277 linear_138_scale_0" [id=2277, type=get_attr]; +"2278 linear_138_zero_point_0" [id=2278, type=get_attr]; +"2279 _frozen_param147_1" [id=2279, type=get_attr]; +"2280 dequantize_per_channel_default_93" [id=2280, type=dequantize_per_channel]; +"2281 linear_138" [id=2281, type=linear]; +"2282 dropout_89" [id=2282, type=dropout]; +"2283 view_124" [id=2283, type=view]; +"2284 permute_103" [id=2284, type=permute]; +"2285 reshape_102" [id=2285, type=reshape]; +"2286 slice_342" [id=2286, type=slice]; +"2287 slice_343" [id=2287, type=slice]; +"2288 slice_344" [id=2288, type=slice]; +"2289 slice_345" [id=2289, type=slice]; +"2290 contiguous_43" [id=2290, type=contiguous]; +"2291 layer_norm_48" [id=2291, type=layer_norm]; +"2292 add_78" [id=2292, type=add]; +"2293 add_78_0_0_nncf_smooth_quant_0" [id=2293, type=call_module]; +"2294 quantize_per_tensor_default_140" [id=2294, type=quantize_per_tensor]; +"2295 dequantize_per_tensor_default_140" [id=2295, type=dequantize_per_tensor]; +"2296 linear_139_scale_0" [id=2296, type=get_attr]; +"2297 linear_139_zero_point_0" [id=2297, type=get_attr]; +"2298 _frozen_param148_1" [id=2298, type=get_attr]; +"2299 dequantize_per_channel_default_94" [id=2299, type=dequantize_per_channel]; +"2300 linear_139" [id=2300, type=linear]; +"2301 gelu_22" [id=2301, type=gelu]; +"2302 dropout_90" [id=2302, type=dropout]; +"2303 dropout_90_0_0_nncf_smooth_quant_0" [id=2303, type=call_module]; +"2304 quantize_per_tensor_default_141" [id=2304, type=quantize_per_tensor]; +"2305 dequantize_per_tensor_default_141" [id=2305, type=dequantize_per_tensor]; +"2306 linear_140_scale_0" [id=2306, type=get_attr]; +"2307 linear_140_zero_point_0" [id=2307, type=get_attr]; +"2308 _frozen_param149_1" [id=2308, type=get_attr]; +"2309 dequantize_per_channel_default_95" [id=2309, type=dequantize_per_channel]; +"2310 linear_140" [id=2310, type=linear]; +"2311 dropout_91" [id=2311, type=dropout]; +"2312 layer_norm_49" [id=2312, type=layer_norm]; +"2313 add_79" [id=2313, type=add]; +"2314 _frozen_param191" [id=2314, type=get_attr]; +"2315 pad_26" [id=2315, type=pad]; +"2316 view_127" [id=2316, type=view]; +"2317 permute_105" [id=2317, type=permute]; +"2318 reshape_103" [id=2318, type=reshape]; +"2319 _frozen_param192" [id=2319, type=get_attr]; +"2320 reshape_103_0_0_nncf_smooth_quant_0" [id=2320, type=call_module]; +"2321 quantize_per_tensor_default_142" [id=2321, type=quantize_per_tensor]; +"2322 dequantize_per_tensor_default_142" [id=2322, type=dequantize_per_tensor]; +"2323 linear_143_scale_0" [id=2323, type=get_attr]; +"2324 linear_143_zero_point_0" [id=2324, type=get_attr]; +"2325 _frozen_param150_1" [id=2325, type=get_attr]; +"2326 dequantize_per_channel_default_96" [id=2326, type=dequantize_per_channel]; +"2327 linear_143" [id=2327, type=linear]; +"2328 reshape_104" [id=2328, type=reshape]; +"2329 permute_106" [id=2329, type=permute]; +"2330 select_69" [id=2330, type=select]; +"2331 select_70" [id=2331, type=select]; +"2332 select_71" [id=2332, type=select]; +"2333 linalg_vector_norm_46" [id=2333, type=linalg_vector_norm]; +"2334 clamp_min_46" [id=2334, type=clamp_min]; +"2335 expand_as_46" [id=2335, type=expand_as]; +"2336 div_46" [id=2336, type=div]; +"2337 quantize_per_tensor_default_143" [id=2337, type=quantize_per_tensor]; +"2338 dequantize_per_tensor_default_143" [id=2338, type=dequantize_per_tensor]; +"2339 linalg_vector_norm_47" [id=2339, type=linalg_vector_norm]; +"2340 clamp_min_47" [id=2340, type=clamp_min]; +"2341 expand_as_47" [id=2341, type=expand_as]; +"2342 div_47" [id=2342, type=div]; +"2343 quantize_per_tensor_default_144" [id=2343, type=quantize_per_tensor]; +"2344 dequantize_per_tensor_default_144" [id=2344, type=dequantize_per_tensor]; +"2345 transpose_46" [id=2345, type=transpose]; +"2346 matmul_46" [id=2346, type=matmul]; +"2347 _frozen_param194" [id=2347, type=get_attr]; +"2348 mul_47" [id=2348, type=mul]; +"2349 add_80" [id=2349, type=add]; +"2350 softmax_23" [id=2350, type=softmax]; +"2351 dropout_92" [id=2351, type=dropout]; +"2352 matmul_47" [id=2352, type=matmul]; +"2353 transpose_47" [id=2353, type=transpose]; +"2354 reshape_105" [id=2354, type=reshape]; +"2355 reshape_105_0_0_nncf_smooth_quant_0" [id=2355, type=call_module]; +"2356 quantize_per_tensor_default_145" [id=2356, type=quantize_per_tensor]; +"2357 dequantize_per_tensor_default_145" [id=2357, type=dequantize_per_tensor]; +"2358 linear_144_scale_0" [id=2358, type=get_attr]; +"2359 linear_144_zero_point_0" [id=2359, type=get_attr]; +"2360 _frozen_param151_1" [id=2360, type=get_attr]; +"2361 dequantize_per_channel_default_97" [id=2361, type=dequantize_per_channel]; +"2362 linear_144" [id=2362, type=linear]; +"2363 dropout_93" [id=2363, type=dropout]; +"2364 view_128" [id=2364, type=view]; +"2365 permute_107" [id=2365, type=permute]; +"2366 reshape_106" [id=2366, type=reshape]; +"2367 slice_347" [id=2367, type=slice]; +"2368 slice_348" [id=2368, type=slice]; +"2369 slice_349" [id=2369, type=slice]; +"2370 slice_350" [id=2370, type=slice]; +"2371 contiguous_45" [id=2371, type=contiguous]; +"2372 layer_norm_50" [id=2372, type=layer_norm]; +"2373 add_81" [id=2373, type=add]; +"2374 add_81_0_0_nncf_smooth_quant_0" [id=2374, type=call_module]; +"2375 quantize_per_tensor_default_146" [id=2375, type=quantize_per_tensor]; +"2376 dequantize_per_tensor_default_146" [id=2376, type=dequantize_per_tensor]; +"2377 linear_145_scale_0" [id=2377, type=get_attr]; +"2378 linear_145_zero_point_0" [id=2378, type=get_attr]; +"2379 _frozen_param152_1" [id=2379, type=get_attr]; +"2380 dequantize_per_channel_default_98" [id=2380, type=dequantize_per_channel]; +"2381 linear_145" [id=2381, type=linear]; +"2382 gelu_23" [id=2382, type=gelu]; +"2383 dropout_94" [id=2383, type=dropout]; +"2384 dropout_94_0_0_nncf_smooth_quant_0" [id=2384, type=call_module]; +"2385 quantize_per_tensor_default_147" [id=2385, type=quantize_per_tensor]; +"2386 dequantize_per_tensor_default_147" [id=2386, type=dequantize_per_tensor]; +"2387 linear_146_scale_0" [id=2387, type=get_attr]; +"2388 linear_146_zero_point_0" [id=2388, type=get_attr]; +"2389 _frozen_param155_1" [id=2389, type=get_attr]; +"2390 dequantize_per_channel_default_99" [id=2390, type=dequantize_per_channel]; +"2391 linear_146" [id=2391, type=linear]; +"2392 dropout_95" [id=2392, type=dropout]; +"2393 layer_norm_51" [id=2393, type=layer_norm]; +"2394 add_82" [id=2394, type=add]; +"2395 layer_norm_52" [id=2395, type=layer_norm]; +"2396 permute_108" [id=2396, type=permute]; +"2397 adaptive_avg_pool2d" [id=2397, type=adaptive_avg_pool2d]; +"2398 flatten" [id=2398, type=flatten]; +"2399 flatten_0_0_nncf_smooth_quant_0" [id=2399, type=call_module]; +"2400 quantize_per_tensor_default_148" [id=2400, type=quantize_per_tensor]; +"2401 dequantize_per_tensor_default_148" [id=2401, type=dequantize_per_tensor]; +"2402 linear_147_scale_0" [id=2402, type=get_attr]; +"2403 linear_147_zero_point_0" [id=2403, type=get_attr]; +"2404 _frozen_param159_1" [id=2404, type=get_attr]; +"2405 dequantize_per_channel_default_100" [id=2405, type=dequantize_per_channel]; +"2406 linear_147" [id=2406, type=linear]; +"2407 output" [id=2407, type=output]; +"0 features_1_0_attn_proj_bias" -> "238 linear_3" [label="(96,)", style=solid]; +"1 features_1_1_attn_proj_bias" -> "333 linear_9" [label="(96,)", style=solid]; +"2 features_3_0_attn_proj_bias" -> "435 linear_16" [label="(192,)", style=solid]; +"3 features_3_1_attn_proj_bias" -> "533 linear_22" [label="(192,)", style=solid]; +"4 features_5_0_attn_proj_bias" -> "638 linear_29" [label="(384,)", style=solid]; +"5 features_5_1_attn_proj_bias" -> "736 linear_35" [label="(384,)", style=solid]; +"6 features_5_2_attn_proj_bias" -> "818 linear_41" [label="(384,)", style=solid]; +"7 features_5_3_attn_proj_bias" -> "916 linear_47" [label="(384,)", style=solid]; +"8 features_5_4_attn_proj_bias" -> "998 linear_53" [label="(384,)", style=solid]; +"9 features_5_5_attn_proj_bias" -> "1096 linear_59" [label="(384,)", style=solid]; +"10 features_5_6_attn_proj_bias" -> "1178 linear_65" [label="(384,)", style=solid]; +"11 features_5_7_attn_proj_bias" -> "1276 linear_71" [label="(384,)", style=solid]; +"12 features_5_8_attn_proj_bias" -> "1358 linear_77" [label="(384,)", style=solid]; +"13 features_5_9_attn_proj_bias" -> "1456 linear_83" [label="(384,)", style=solid]; +"14 features_5_10_attn_proj_bias" -> "1538 linear_89" [label="(384,)", style=solid]; +"15 features_5_11_attn_proj_bias" -> "1636 linear_95" [label="(384,)", style=solid]; +"16 features_5_12_attn_proj_bias" -> "1718 linear_101" [label="(384,)", style=solid]; +"17 features_5_13_attn_proj_bias" -> "1816 linear_107" [label="(384,)", style=solid]; +"18 features_5_14_attn_proj_bias" -> "1898 linear_113" [label="(384,)", style=solid]; +"19 features_5_15_attn_proj_bias" -> "1996 linear_119" [label="(384,)", style=solid]; +"20 features_5_16_attn_proj_bias" -> "2078 linear_125" [label="(384,)", style=solid]; +"21 features_5_17_attn_proj_bias" -> "2176 linear_131" [label="(384,)", style=solid]; +"22 features_7_0_attn_proj_bias" -> "2281 linear_138" [label="(768,)", style=solid]; +"23 features_7_1_attn_proj_bias" -> "2362 linear_144" [label="(768,)", style=solid]; +"24 features_0_0_bias" -> "187 conv2d" [label="(96,)", style=solid]; +"25 features_0_2_weight" -> "189 layer_norm" [label="(96,)", style=solid]; +"26 features_0_2_bias" -> "189 layer_norm" [label="(96,)", style=solid]; +"27 features_1_0_norm1_weight" -> "245 layer_norm_1" [label="(96,)", style=solid]; +"28 features_1_0_norm1_bias" -> "245 layer_norm_1" [label="(96,)", style=solid]; +"29 features_1_0_mlp_0_bias" -> "254 linear_4" [label="(384,)", style=solid]; +"30 features_1_0_mlp_3_bias" -> "264 linear_5" [label="(96,)", style=solid]; +"31 features_1_0_norm2_weight" -> "266 layer_norm_2" [label="(96,)", style=solid]; +"32 features_1_0_norm2_bias" -> "266 layer_norm_2" [label="(96,)", style=solid]; +"33 features_1_1_norm1_weight" -> "341 layer_norm_3" [label="(96,)", style=solid]; +"34 features_1_1_norm1_bias" -> "341 layer_norm_3" [label="(96,)", style=solid]; +"35 features_1_1_mlp_0_bias" -> "350 linear_10" [label="(384,)", style=solid]; +"36 features_1_1_mlp_3_bias" -> "360 linear_11" [label="(96,)", style=solid]; +"37 features_1_1_norm2_weight" -> "362 layer_norm_4" [label="(96,)", style=solid]; +"38 features_1_1_norm2_bias" -> "362 layer_norm_4" [label="(96,)", style=solid]; +"39 features_2_norm_weight" -> "386 layer_norm_5" [label="(192,)", style=solid]; +"40 features_2_norm_bias" -> "386 layer_norm_5" [label="(192,)", style=solid]; +"41 features_3_0_norm1_weight" -> "445 layer_norm_6" [label="(192,)", style=solid]; +"42 features_3_0_norm1_bias" -> "445 layer_norm_6" [label="(192,)", style=solid]; +"43 features_3_0_mlp_0_bias" -> "454 linear_17" [label="(768,)", style=solid]; +"44 features_3_0_mlp_3_bias" -> "464 linear_18" [label="(192,)", style=solid]; +"45 features_3_0_norm2_weight" -> "466 layer_norm_7" [label="(192,)", style=solid]; +"46 features_3_0_norm2_bias" -> "466 layer_norm_7" [label="(192,)", style=solid]; +"47 features_3_1_norm1_weight" -> "544 layer_norm_8" [label="(192,)", style=solid]; +"48 features_3_1_norm1_bias" -> "544 layer_norm_8" [label="(192,)", style=solid]; +"49 features_3_1_mlp_0_bias" -> "553 linear_23" [label="(768,)", style=solid]; +"50 features_3_1_mlp_3_bias" -> "563 linear_24" [label="(192,)", style=solid]; +"51 features_3_1_norm2_weight" -> "565 layer_norm_9" [label="(192,)", style=solid]; +"52 features_3_1_norm2_bias" -> "565 layer_norm_9" [label="(192,)", style=solid]; +"53 features_4_norm_weight" -> "589 layer_norm_10" [label="(384,)", style=solid]; +"54 features_4_norm_bias" -> "589 layer_norm_10" [label="(384,)", style=solid]; +"55 features_5_0_norm1_weight" -> "648 layer_norm_11" [label="(384,)", style=solid]; +"56 features_5_0_norm1_bias" -> "648 layer_norm_11" [label="(384,)", style=solid]; +"57 features_5_0_mlp_0_bias" -> "657 linear_30" [label="(1536,)", style=solid]; +"58 features_5_0_mlp_3_bias" -> "667 linear_31" [label="(384,)", style=solid]; +"59 features_5_0_norm2_weight" -> "669 layer_norm_12" [label="(384,)", style=solid]; +"60 features_5_0_norm2_bias" -> "669 layer_norm_12" [label="(384,)", style=solid]; +"61 features_5_1_norm1_weight" -> "747 layer_norm_13" [label="(384,)", style=solid]; +"62 features_5_1_norm1_bias" -> "747 layer_norm_13" [label="(384,)", style=solid]; +"63 features_5_1_mlp_0_bias" -> "756 linear_36" [label="(1536,)", style=solid]; +"64 features_5_1_mlp_3_bias" -> "766 linear_37" [label="(384,)", style=solid]; +"65 features_5_1_norm2_weight" -> "768 layer_norm_14" [label="(384,)", style=solid]; +"66 features_5_1_norm2_bias" -> "768 layer_norm_14" [label="(384,)", style=solid]; +"67 features_5_2_norm1_weight" -> "828 layer_norm_15" [label="(384,)", style=solid]; +"68 features_5_2_norm1_bias" -> "828 layer_norm_15" [label="(384,)", style=solid]; +"69 features_5_2_mlp_0_bias" -> "837 linear_42" [label="(1536,)", style=solid]; +"70 features_5_2_mlp_3_bias" -> "847 linear_43" [label="(384,)", style=solid]; +"71 features_5_2_norm2_weight" -> "849 layer_norm_16" [label="(384,)", style=solid]; +"72 features_5_2_norm2_bias" -> "849 layer_norm_16" [label="(384,)", style=solid]; +"73 features_5_3_norm1_weight" -> "927 layer_norm_17" [label="(384,)", style=solid]; +"74 features_5_3_norm1_bias" -> "927 layer_norm_17" [label="(384,)", style=solid]; +"75 features_5_3_mlp_0_bias" -> "936 linear_48" [label="(1536,)", style=solid]; +"76 features_5_3_mlp_3_bias" -> "946 linear_49" [label="(384,)", style=solid]; +"77 features_5_3_norm2_weight" -> "948 layer_norm_18" [label="(384,)", style=solid]; +"78 features_5_3_norm2_bias" -> "948 layer_norm_18" [label="(384,)", style=solid]; +"79 features_5_4_norm1_weight" -> "1008 layer_norm_19" [label="(384,)", style=solid]; +"80 features_5_4_norm1_bias" -> "1008 layer_norm_19" [label="(384,)", style=solid]; +"81 features_5_4_mlp_0_bias" -> "1017 linear_54" [label="(1536,)", style=solid]; +"82 features_5_4_mlp_3_bias" -> "1027 linear_55" [label="(384,)", style=solid]; +"83 features_5_4_norm2_weight" -> "1029 layer_norm_20" [label="(384,)", style=solid]; +"84 features_5_4_norm2_bias" -> "1029 layer_norm_20" [label="(384,)", style=solid]; +"85 features_5_5_norm1_weight" -> "1107 layer_norm_21" [label="(384,)", style=solid]; +"86 features_5_5_norm1_bias" -> "1107 layer_norm_21" [label="(384,)", style=solid]; +"87 features_5_5_mlp_0_bias" -> "1116 linear_60" [label="(1536,)", style=solid]; +"88 features_5_5_mlp_3_bias" -> "1126 linear_61" [label="(384,)", style=solid]; +"89 features_5_5_norm2_weight" -> "1128 layer_norm_22" [label="(384,)", style=solid]; +"90 features_5_5_norm2_bias" -> "1128 layer_norm_22" [label="(384,)", style=solid]; +"91 features_5_6_norm1_weight" -> "1188 layer_norm_23" [label="(384,)", style=solid]; +"92 features_5_6_norm1_bias" -> "1188 layer_norm_23" [label="(384,)", style=solid]; +"93 features_5_6_mlp_0_bias" -> "1197 linear_66" [label="(1536,)", style=solid]; +"94 features_5_6_mlp_3_bias" -> "1207 linear_67" [label="(384,)", style=solid]; +"95 features_5_6_norm2_weight" -> "1209 layer_norm_24" [label="(384,)", style=solid]; +"96 features_5_6_norm2_bias" -> "1209 layer_norm_24" [label="(384,)", style=solid]; +"97 features_5_7_norm1_weight" -> "1287 layer_norm_25" [label="(384,)", style=solid]; +"98 features_5_7_norm1_bias" -> "1287 layer_norm_25" [label="(384,)", style=solid]; +"99 features_5_7_mlp_0_bias" -> "1296 linear_72" [label="(1536,)", style=solid]; +"100 features_5_7_mlp_3_bias" -> "1306 linear_73" [label="(384,)", style=solid]; +"101 features_5_7_norm2_weight" -> "1308 layer_norm_26" [label="(384,)", style=solid]; +"102 features_5_7_norm2_bias" -> "1308 layer_norm_26" [label="(384,)", style=solid]; +"103 features_5_8_norm1_weight" -> "1368 layer_norm_27" [label="(384,)", style=solid]; +"104 features_5_8_norm1_bias" -> "1368 layer_norm_27" [label="(384,)", style=solid]; +"105 features_5_8_mlp_0_bias" -> "1377 linear_78" [label="(1536,)", style=solid]; +"106 features_5_8_mlp_3_bias" -> "1387 linear_79" [label="(384,)", style=solid]; +"107 features_5_8_norm2_weight" -> "1389 layer_norm_28" [label="(384,)", style=solid]; +"108 features_5_8_norm2_bias" -> "1389 layer_norm_28" [label="(384,)", style=solid]; +"109 features_5_9_norm1_weight" -> "1467 layer_norm_29" [label="(384,)", style=solid]; +"110 features_5_9_norm1_bias" -> "1467 layer_norm_29" [label="(384,)", style=solid]; +"111 features_5_9_mlp_0_bias" -> "1476 linear_84" [label="(1536,)", style=solid]; +"112 features_5_9_mlp_3_bias" -> "1486 linear_85" [label="(384,)", style=solid]; +"113 features_5_9_norm2_weight" -> "1488 layer_norm_30" [label="(384,)", style=solid]; +"114 features_5_9_norm2_bias" -> "1488 layer_norm_30" [label="(384,)", style=solid]; +"115 features_5_10_norm1_weight" -> "1548 layer_norm_31" [label="(384,)", style=solid]; +"116 features_5_10_norm1_bias" -> "1548 layer_norm_31" [label="(384,)", style=solid]; +"117 features_5_10_mlp_0_bias" -> "1557 linear_90" [label="(1536,)", style=solid]; +"118 features_5_10_mlp_3_bias" -> "1567 linear_91" [label="(384,)", style=solid]; +"119 features_5_10_norm2_weight" -> "1569 layer_norm_32" [label="(384,)", style=solid]; +"120 features_5_10_norm2_bias" -> "1569 layer_norm_32" [label="(384,)", style=solid]; +"121 features_5_11_norm1_weight" -> "1647 layer_norm_33" [label="(384,)", style=solid]; +"122 features_5_11_norm1_bias" -> "1647 layer_norm_33" [label="(384,)", style=solid]; +"123 features_5_11_mlp_0_bias" -> "1656 linear_96" [label="(1536,)", style=solid]; +"124 features_5_11_mlp_3_bias" -> "1666 linear_97" [label="(384,)", style=solid]; +"125 features_5_11_norm2_weight" -> "1668 layer_norm_34" [label="(384,)", style=solid]; +"126 features_5_11_norm2_bias" -> "1668 layer_norm_34" [label="(384,)", style=solid]; +"127 features_5_12_norm1_weight" -> "1728 layer_norm_35" [label="(384,)", style=solid]; +"128 features_5_12_norm1_bias" -> "1728 layer_norm_35" [label="(384,)", style=solid]; +"129 features_5_12_mlp_0_bias" -> "1737 linear_102" [label="(1536,)", style=solid]; +"130 features_5_12_mlp_3_bias" -> "1747 linear_103" [label="(384,)", style=solid]; +"131 features_5_12_norm2_weight" -> "1749 layer_norm_36" [label="(384,)", style=solid]; +"132 features_5_12_norm2_bias" -> "1749 layer_norm_36" [label="(384,)", style=solid]; +"133 features_5_13_norm1_weight" -> "1827 layer_norm_37" [label="(384,)", style=solid]; +"134 features_5_13_norm1_bias" -> "1827 layer_norm_37" [label="(384,)", style=solid]; +"135 features_5_13_mlp_0_bias" -> "1836 linear_108" [label="(1536,)", style=solid]; +"136 features_5_13_mlp_3_bias" -> "1846 linear_109" [label="(384,)", style=solid]; +"137 features_5_13_norm2_weight" -> "1848 layer_norm_38" [label="(384,)", style=solid]; +"138 features_5_13_norm2_bias" -> "1848 layer_norm_38" [label="(384,)", style=solid]; +"139 features_5_14_norm1_weight" -> "1908 layer_norm_39" [label="(384,)", style=solid]; +"140 features_5_14_norm1_bias" -> "1908 layer_norm_39" [label="(384,)", style=solid]; +"141 features_5_14_mlp_0_bias" -> "1917 linear_114" [label="(1536,)", style=solid]; +"142 features_5_14_mlp_3_bias" -> "1927 linear_115" [label="(384,)", style=solid]; +"143 features_5_14_norm2_weight" -> "1929 layer_norm_40" [label="(384,)", style=solid]; +"144 features_5_14_norm2_bias" -> "1929 layer_norm_40" [label="(384,)", style=solid]; +"145 features_5_15_norm1_weight" -> "2007 layer_norm_41" [label="(384,)", style=solid]; +"146 features_5_15_norm1_bias" -> "2007 layer_norm_41" [label="(384,)", style=solid]; +"147 features_5_15_mlp_0_bias" -> "2016 linear_120" [label="(1536,)", style=solid]; +"148 features_5_15_mlp_3_bias" -> "2026 linear_121" [label="(384,)", style=solid]; +"149 features_5_15_norm2_weight" -> "2028 layer_norm_42" [label="(384,)", style=solid]; +"150 features_5_15_norm2_bias" -> "2028 layer_norm_42" [label="(384,)", style=solid]; +"151 features_5_16_norm1_weight" -> "2088 layer_norm_43" [label="(384,)", style=solid]; +"152 features_5_16_norm1_bias" -> "2088 layer_norm_43" [label="(384,)", style=solid]; +"153 features_5_16_mlp_0_bias" -> "2097 linear_126" [label="(1536,)", style=solid]; +"154 features_5_16_mlp_3_bias" -> "2107 linear_127" [label="(384,)", style=solid]; +"155 features_5_16_norm2_weight" -> "2109 layer_norm_44" [label="(384,)", style=solid]; +"156 features_5_16_norm2_bias" -> "2109 layer_norm_44" [label="(384,)", style=solid]; +"157 features_5_17_norm1_weight" -> "2187 layer_norm_45" [label="(384,)", style=solid]; +"158 features_5_17_norm1_bias" -> "2187 layer_norm_45" [label="(384,)", style=solid]; +"159 features_5_17_mlp_0_bias" -> "2196 linear_132" [label="(1536,)", style=solid]; +"160 features_5_17_mlp_3_bias" -> "2206 linear_133" [label="(384,)", style=solid]; +"161 features_5_17_norm2_weight" -> "2208 layer_norm_46" [label="(384,)", style=solid]; +"162 features_5_17_norm2_bias" -> "2208 layer_norm_46" [label="(384,)", style=solid]; +"163 features_6_norm_weight" -> "2232 layer_norm_47" [label="(768,)", style=solid]; +"164 features_6_norm_bias" -> "2232 layer_norm_47" [label="(768,)", style=solid]; +"165 features_7_0_norm1_weight" -> "2291 layer_norm_48" [label="(768,)", style=solid]; +"166 features_7_0_norm1_bias" -> "2291 layer_norm_48" [label="(768,)", style=solid]; +"167 features_7_0_mlp_0_bias" -> "2300 linear_139" [label="(3072,)", style=solid]; +"168 features_7_0_mlp_3_bias" -> "2310 linear_140" [label="(768,)", style=solid]; +"169 features_7_0_norm2_weight" -> "2312 layer_norm_49" [label="(768,)", style=solid]; +"170 features_7_0_norm2_bias" -> "2312 layer_norm_49" [label="(768,)", style=solid]; +"171 features_7_1_norm1_weight" -> "2372 layer_norm_50" [label="(768,)", style=solid]; +"172 features_7_1_norm1_bias" -> "2372 layer_norm_50" [label="(768,)", style=solid]; +"173 features_7_1_mlp_0_bias" -> "2381 linear_145" [label="(3072,)", style=solid]; +"174 features_7_1_mlp_3_bias" -> "2391 linear_146" [label="(768,)", style=solid]; +"175 features_7_1_norm2_weight" -> "2393 layer_norm_51" [label="(768,)", style=solid]; +"176 features_7_1_norm2_bias" -> "2393 layer_norm_51" [label="(768,)", style=solid]; +"177 norm_weight" -> "2395 layer_norm_52" [label="(768,)", style=solid]; +"178 norm_bias" -> "2395 layer_norm_52" [label="(768,)", style=solid]; +"179 head_bias" -> "2406 linear_147" [label="(1000,)", style=solid]; +"180 x" -> "181 quantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; +"181 quantize_per_tensor_default" -> "182 dequantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; +"182 dequantize_per_tensor_default" -> "187 conv2d" [label="(1, 3, 224, 224)", style=solid]; +"183 conv2d_scale_0" -> "186 dequantize_per_channel_default" [label="(96,)", style=solid]; +"184 conv2d_zero_point_0" -> "186 dequantize_per_channel_default" [label="(96,)", style=solid]; +"185 _frozen_param2_1" -> "186 dequantize_per_channel_default" [label="(96, 3, 4, 4)", style=solid]; +"186 dequantize_per_channel_default" -> "187 conv2d" [label="(96, 3, 4, 4)", style=solid]; +"187 conv2d" -> "188 permute" [label="(1, 96, 56, 56)", style=solid]; +"188 permute" -> "189 layer_norm" [label="(1, 56, 56, 96)", style=solid]; +"189 layer_norm" -> "191 pad" [label="(1, 56, 56, 96)", style=solid]; +"189 layer_norm" -> "246 add_1" [label="(1, 56, 56, 96)", style=solid]; +"190 _frozen_param0" -> "225 add" [label="(1, 3, 64, 64)", style=solid]; +"191 pad" -> "192 view_2" [label="(1, 56, 56, 96)", style=solid]; +"192 view_2" -> "193 permute_2" [label="(1, 7, 8, 7, 8, 96)", style=solid]; +"193 permute_2" -> "194 reshape" [label="(1, 7, 7, 8, 8, 96)", style=solid]; +"194 reshape" -> "196 reshape_0_0_nncf_smooth_quant_0" [label="(49, 64, 96)", style=solid]; +"195 _frozen_param1" -> "203 linear_2" [label="(288,)", style=solid]; +"196 reshape_0_0_nncf_smooth_quant_0" -> "197 quantize_per_tensor_default_1" [label="(49, 64, 96)", style=solid]; +"197 quantize_per_tensor_default_1" -> "198 dequantize_per_tensor_default_1" [label="(49, 64, 96)", style=solid]; +"198 dequantize_per_tensor_default_1" -> "203 linear_2" [label="(49, 64, 96)", style=solid]; +"199 linear_2_scale_0" -> "202 dequantize_per_channel_default_1" [label="(288,)", style=solid]; +"200 linear_2_zero_point_0" -> "202 dequantize_per_channel_default_1" [label="(288,)", style=solid]; +"201 _frozen_param6_1" -> "202 dequantize_per_channel_default_1" [label="(288, 96)", style=solid]; +"202 dequantize_per_channel_default_1" -> "203 linear_2" [label="(288, 96)", style=solid]; +"203 linear_2" -> "204 reshape_1" [label="(49, 64, 288)", style=solid]; +"204 reshape_1" -> "205 permute_3" [label="(49, 64, 3, 3, 32)", style=solid]; +"205 permute_3" -> "206 select" [label="(3, 49, 3, 64, 32)", style=solid]; +"205 permute_3" -> "207 select_1" [label="(3, 49, 3, 64, 32)", style=solid]; +"205 permute_3" -> "208 select_2" [label="(3, 49, 3, 64, 32)", style=solid]; +"206 select" -> "209 linalg_vector_norm" [label="(49, 3, 64, 32)", style=solid]; +"206 select" -> "211 expand_as" [label="(49, 3, 64, 32)", style=solid]; +"206 select" -> "212 div" [label="(49, 3, 64, 32)", style=solid]; +"207 select_1" -> "215 linalg_vector_norm_1" [label="(49, 3, 64, 32)", style=solid]; +"207 select_1" -> "217 expand_as_1" [label="(49, 3, 64, 32)", style=solid]; +"207 select_1" -> "218 div_1" [label="(49, 3, 64, 32)", style=solid]; +"208 select_2" -> "228 matmul_1" [label="(49, 3, 64, 32)", style=solid]; +"209 linalg_vector_norm" -> "210 clamp_min" [label="(49, 3, 64, 1)", style=solid]; +"210 clamp_min" -> "211 expand_as" [label="(49, 3, 64, 1)", style=solid]; +"211 expand_as" -> "212 div" [label="(49, 3, 64, 32)", style=solid]; +"212 div" -> "213 quantize_per_tensor_default_2" [label="(49, 3, 64, 32)", style=solid]; +"213 quantize_per_tensor_default_2" -> "214 dequantize_per_tensor_default_2" [label="(49, 3, 64, 32)", style=solid]; +"214 dequantize_per_tensor_default_2" -> "222 matmul" [label="(49, 3, 64, 32)", style=solid]; +"215 linalg_vector_norm_1" -> "216 clamp_min_1" [label="(49, 3, 64, 1)", style=solid]; +"216 clamp_min_1" -> "217 expand_as_1" [label="(49, 3, 64, 1)", style=solid]; +"217 expand_as_1" -> "218 div_1" [label="(49, 3, 64, 32)", style=solid]; +"218 div_1" -> "219 quantize_per_tensor_default_3" [label="(49, 3, 64, 32)", style=solid]; +"219 quantize_per_tensor_default_3" -> "220 dequantize_per_tensor_default_3" [label="(49, 3, 64, 32)", style=solid]; +"220 dequantize_per_tensor_default_3" -> "221 transpose" [label="(49, 3, 64, 32)", style=solid]; +"221 transpose" -> "222 matmul" [label="(49, 3, 32, 64)", style=solid]; +"222 matmul" -> "224 mul_1" [label="(49, 3, 64, 64)", style=solid]; +"223 _frozen_param3" -> "224 mul_1" [label="(3, 1, 1)", style=solid]; +"224 mul_1" -> "225 add" [label="(49, 3, 64, 64)", style=solid]; +"225 add" -> "226 softmax" [label="(49, 3, 64, 64)", style=solid]; +"226 softmax" -> "227 dropout" [label="(49, 3, 64, 64)", style=solid]; +"227 dropout" -> "228 matmul_1" [label="(49, 3, 64, 64)", style=solid]; +"228 matmul_1" -> "229 transpose_1" [label="(49, 3, 64, 32)", style=solid]; +"229 transpose_1" -> "230 reshape_2" [label="(49, 64, 3, 32)", style=solid]; +"230 reshape_2" -> "231 reshape_2_0_0_nncf_smooth_quant_0" [label="(49, 64, 96)", style=solid]; +"231 reshape_2_0_0_nncf_smooth_quant_0" -> "232 quantize_per_tensor_default_4" [label="(49, 64, 96)", style=solid]; +"232 quantize_per_tensor_default_4" -> "233 dequantize_per_tensor_default_4" [label="(49, 64, 96)", style=solid]; +"233 dequantize_per_tensor_default_4" -> "238 linear_3" [label="(49, 64, 96)", style=solid]; +"234 linear_3_scale_0" -> "237 dequantize_per_channel_default_2" [label="(96,)", style=solid]; +"235 linear_3_zero_point_0" -> "237 dequantize_per_channel_default_2" [label="(96,)", style=solid]; +"236 _frozen_param8_1" -> "237 dequantize_per_channel_default_2" [label="(96, 96)", style=solid]; +"237 dequantize_per_channel_default_2" -> "238 linear_3" [label="(96, 96)", style=solid]; +"238 linear_3" -> "239 dropout_1" [label="(49, 64, 96)", style=solid]; +"239 dropout_1" -> "240 view_3" [label="(49, 64, 96)", style=solid]; +"240 view_3" -> "241 permute_4" [label="(1, 7, 7, 8, 8, 96)", style=solid]; +"241 permute_4" -> "242 reshape_3" [label="(1, 7, 8, 7, 8, 96)", style=solid]; +"242 reshape_3" -> "243 slice_2" [label="(1, 56, 56, 96)", style=solid]; +"243 slice_2" -> "244 slice_3" [label="(1, 56, 56, 96)", style=solid]; +"244 slice_3" -> "245 layer_norm_1" [label="(1, 56, 56, 96)", style=solid]; +"245 layer_norm_1" -> "246 add_1" [label="(1, 56, 56, 96)", style=solid]; +"246 add_1" -> "247 add_1_0_0_nncf_smooth_quant_0" [label="(1, 56, 56, 96)", style=solid]; +"246 add_1" -> "267 add_2" [label="(1, 56, 56, 96)", style=solid]; +"247 add_1_0_0_nncf_smooth_quant_0" -> "248 quantize_per_tensor_default_5" [label="(1, 56, 56, 96)", style=solid]; +"248 quantize_per_tensor_default_5" -> "249 dequantize_per_tensor_default_5" [label="(1, 56, 56, 96)", style=solid]; +"249 dequantize_per_tensor_default_5" -> "254 linear_4" [label="(1, 56, 56, 96)", style=solid]; +"250 linear_4_scale_0" -> "253 dequantize_per_channel_default_3" [label="(384,)", style=solid]; +"251 linear_4_zero_point_0" -> "253 dequantize_per_channel_default_3" [label="(384,)", style=solid]; +"252 _frozen_param9_1" -> "253 dequantize_per_channel_default_3" [label="(384, 96)", style=solid]; +"253 dequantize_per_channel_default_3" -> "254 linear_4" [label="(384, 96)", style=solid]; +"254 linear_4" -> "255 gelu" [label="(1, 56, 56, 384)", style=solid]; +"255 gelu" -> "256 dropout_2" [label="(1, 56, 56, 384)", style=solid]; +"256 dropout_2" -> "257 dropout_2_0_0_nncf_smooth_quant_0" [label="(1, 56, 56, 384)", style=solid]; +"257 dropout_2_0_0_nncf_smooth_quant_0" -> "258 quantize_per_tensor_default_6" [label="(1, 56, 56, 384)", style=solid]; +"258 quantize_per_tensor_default_6" -> "259 dequantize_per_tensor_default_6" [label="(1, 56, 56, 384)", style=solid]; +"259 dequantize_per_tensor_default_6" -> "264 linear_5" [label="(1, 56, 56, 384)", style=solid]; +"260 linear_5_scale_0" -> "263 dequantize_per_channel_default_4" [label="(96,)", style=solid]; +"261 linear_5_zero_point_0" -> "263 dequantize_per_channel_default_4" [label="(96,)", style=solid]; +"262 _frozen_param10_1" -> "263 dequantize_per_channel_default_4" [label="(96, 384)", style=solid]; +"263 dequantize_per_channel_default_4" -> "264 linear_5" [label="(96, 384)", style=solid]; +"264 linear_5" -> "265 dropout_3" [label="(1, 56, 56, 96)", style=solid]; +"265 dropout_3" -> "266 layer_norm_2" [label="(1, 56, 56, 96)", style=solid]; +"266 layer_norm_2" -> "267 add_2" [label="(1, 56, 56, 96)", style=solid]; +"267 add_2" -> "269 pad_1" [label="(1, 56, 56, 96)", style=solid]; +"267 add_2" -> "342 add_5" [label="(1, 56, 56, 96)", style=solid]; +"268 _frozen_param4" -> "304 add_3" [label="(1, 3, 64, 64)", style=solid]; +"269 pad_1" -> "270 roll" [label="(1, 56, 56, 96)", style=solid]; +"270 roll" -> "271 view_6" [label="(1, 56, 56, 96)", style=solid]; +"271 view_6" -> "272 permute_6" [label="(1, 7, 8, 7, 8, 96)", style=solid]; +"272 permute_6" -> "273 reshape_4" [label="(1, 7, 7, 8, 8, 96)", style=solid]; +"273 reshape_4" -> "275 reshape_4_0_0_nncf_smooth_quant_0" [label="(49, 64, 96)", style=solid]; +"273 reshape_4" -> "305 new_zeros" [label="(49, 64, 96)", style=solid]; +"274 _frozen_param5" -> "282 linear_8" [label="(288,)", style=solid]; +"275 reshape_4_0_0_nncf_smooth_quant_0" -> "276 quantize_per_tensor_default_7" [label="(49, 64, 96)", style=solid]; +"276 quantize_per_tensor_default_7" -> "277 dequantize_per_tensor_default_7" [label="(49, 64, 96)", style=solid]; +"277 dequantize_per_tensor_default_7" -> "282 linear_8" [label="(49, 64, 96)", style=solid]; +"278 linear_8_scale_0" -> "281 dequantize_per_channel_default_5" [label="(288,)", style=solid]; +"279 linear_8_zero_point_0" -> "281 dequantize_per_channel_default_5" [label="(288,)", style=solid]; +"280 _frozen_param11_1" -> "281 dequantize_per_channel_default_5" [label="(288, 96)", style=solid]; +"281 dequantize_per_channel_default_5" -> "282 linear_8" [label="(288, 96)", style=solid]; +"282 linear_8" -> "283 reshape_5" [label="(49, 64, 288)", style=solid]; +"283 reshape_5" -> "284 permute_7" [label="(49, 64, 3, 3, 32)", style=solid]; +"284 permute_7" -> "285 select_3" [label="(3, 49, 3, 64, 32)", style=solid]; +"284 permute_7" -> "286 select_4" [label="(3, 49, 3, 64, 32)", style=solid]; +"284 permute_7" -> "287 select_5" [label="(3, 49, 3, 64, 32)", style=solid]; +"285 select_3" -> "288 linalg_vector_norm_2" [label="(49, 3, 64, 32)", style=solid]; +"285 select_3" -> "290 expand_as_2" [label="(49, 3, 64, 32)", style=solid]; +"285 select_3" -> "291 div_2" [label="(49, 3, 64, 32)", style=solid]; +"286 select_4" -> "294 linalg_vector_norm_3" [label="(49, 3, 64, 32)", style=solid]; +"286 select_4" -> "296 expand_as_3" [label="(49, 3, 64, 32)", style=solid]; +"286 select_4" -> "297 div_3" [label="(49, 3, 64, 32)", style=solid]; +"287 select_5" -> "323 matmul_3" [label="(49, 3, 64, 32)", style=solid]; +"288 linalg_vector_norm_2" -> "289 clamp_min_2" [label="(49, 3, 64, 1)", style=solid]; +"289 clamp_min_2" -> "290 expand_as_2" [label="(49, 3, 64, 1)", style=solid]; +"290 expand_as_2" -> "291 div_2" [label="(49, 3, 64, 32)", style=solid]; +"291 div_2" -> "292 quantize_per_tensor_default_8" [label="(49, 3, 64, 32)", style=solid]; +"292 quantize_per_tensor_default_8" -> "293 dequantize_per_tensor_default_8" [label="(49, 3, 64, 32)", style=solid]; +"293 dequantize_per_tensor_default_8" -> "301 matmul_2" [label="(49, 3, 64, 32)", style=solid]; +"294 linalg_vector_norm_3" -> "295 clamp_min_3" [label="(49, 3, 64, 1)", style=solid]; +"295 clamp_min_3" -> "296 expand_as_3" [label="(49, 3, 64, 1)", style=solid]; +"296 expand_as_3" -> "297 div_3" [label="(49, 3, 64, 32)", style=solid]; +"297 div_3" -> "298 quantize_per_tensor_default_9" [label="(49, 3, 64, 32)", style=solid]; +"298 quantize_per_tensor_default_9" -> "299 dequantize_per_tensor_default_9" [label="(49, 3, 64, 32)", style=solid]; +"299 dequantize_per_tensor_default_9" -> "300 transpose_2" [label="(49, 3, 64, 32)", style=solid]; +"300 transpose_2" -> "301 matmul_2" [label="(49, 3, 32, 64)", style=solid]; +"301 matmul_2" -> "303 mul_3" [label="(49, 3, 64, 64)", style=solid]; +"302 _frozen_param7" -> "303 mul_3" [label="(3, 1, 1)", style=solid]; +"303 mul_3" -> "304 add_3" [label="(49, 3, 64, 64)", style=solid]; +"304 add_3" -> "316 view_8" [label="(49, 3, 64, 64)", style=solid]; +"305 new_zeros" -> "306 view_7" [label="(56, 56)", style=solid]; +"306 view_7" -> "307 permute_8" [label="(7, 8, 7, 8)", style=solid]; +"307 permute_8" -> "308 reshape_6" [label="(7, 7, 8, 8)", style=solid]; +"308 reshape_6" -> "309 unsqueeze_2" [label="(49, 64)", style=solid]; +"308 reshape_6" -> "310 unsqueeze_3" [label="(49, 64)", style=solid]; +"309 unsqueeze_2" -> "311 sub" [label="(49, 1, 64)", style=solid]; +"310 unsqueeze_3" -> "311 sub" [label="(49, 64, 1)", style=solid]; +"311 sub" -> "312 ne" [label="(49, 64, 64)", style=solid]; +"311 sub" -> "313 masked_fill" [label="(49, 64, 64)", style=solid]; +"311 sub" -> "314 eq" [label="(49, 64, 64)", style=solid]; +"312 ne" -> "313 masked_fill" [label="(49, 64, 64)", style=solid]; +"313 masked_fill" -> "315 masked_fill_1" [label="(49, 64, 64)", style=solid]; +"314 eq" -> "315 masked_fill_1" [label="(49, 64, 64)", style=solid]; +"315 masked_fill_1" -> "317 unsqueeze_4" [label="(49, 64, 64)", style=solid]; +"316 view_8" -> "319 add_4" [label="(1, 49, 3, 64, 64)", style=solid]; +"317 unsqueeze_4" -> "318 unsqueeze_5" [label="(49, 1, 64, 64)", style=solid]; +"318 unsqueeze_5" -> "319 add_4" [label="(1, 49, 1, 64, 64)", style=solid]; +"319 add_4" -> "320 view_9" [label="(1, 49, 3, 64, 64)", style=solid]; +"320 view_9" -> "321 softmax_1" [label="(49, 3, 64, 64)", style=solid]; +"321 softmax_1" -> "322 dropout_4" [label="(49, 3, 64, 64)", style=solid]; +"322 dropout_4" -> "323 matmul_3" [label="(49, 3, 64, 64)", style=solid]; +"323 matmul_3" -> "324 transpose_3" [label="(49, 3, 64, 32)", style=solid]; +"324 transpose_3" -> "325 reshape_7" [label="(49, 64, 3, 32)", style=solid]; +"325 reshape_7" -> "326 reshape_7_0_0_nncf_smooth_quant_0" [label="(49, 64, 96)", style=solid]; +"326 reshape_7_0_0_nncf_smooth_quant_0" -> "327 quantize_per_tensor_default_10" [label="(49, 64, 96)", style=solid]; +"327 quantize_per_tensor_default_10" -> "328 dequantize_per_tensor_default_10" [label="(49, 64, 96)", style=solid]; +"328 dequantize_per_tensor_default_10" -> "333 linear_9" [label="(49, 64, 96)", style=solid]; +"329 linear_9_scale_0" -> "332 dequantize_per_channel_default_6" [label="(96,)", style=solid]; +"330 linear_9_zero_point_0" -> "332 dequantize_per_channel_default_6" [label="(96,)", style=solid]; +"331 _frozen_param12_1" -> "332 dequantize_per_channel_default_6" [label="(96, 96)", style=solid]; +"332 dequantize_per_channel_default_6" -> "333 linear_9" [label="(96, 96)", style=solid]; +"333 linear_9" -> "334 dropout_5" [label="(49, 64, 96)", style=solid]; +"334 dropout_5" -> "335 view_10" [label="(49, 64, 96)", style=solid]; +"335 view_10" -> "336 permute_9" [label="(1, 7, 7, 8, 8, 96)", style=solid]; +"336 permute_9" -> "337 reshape_8" [label="(1, 7, 8, 7, 8, 96)", style=solid]; +"337 reshape_8" -> "338 roll_1" [label="(1, 56, 56, 96)", style=solid]; +"338 roll_1" -> "339 slice_23" [label="(1, 56, 56, 96)", style=solid]; +"339 slice_23" -> "340 slice_24" [label="(1, 56, 56, 96)", style=solid]; +"340 slice_24" -> "341 layer_norm_3" [label="(1, 56, 56, 96)", style=solid]; +"341 layer_norm_3" -> "342 add_5" [label="(1, 56, 56, 96)", style=solid]; +"342 add_5" -> "343 add_5_0_0_nncf_smooth_quant_0" [label="(1, 56, 56, 96)", style=solid]; +"342 add_5" -> "363 add_6" [label="(1, 56, 56, 96)", style=solid]; +"343 add_5_0_0_nncf_smooth_quant_0" -> "344 quantize_per_tensor_default_11" [label="(1, 56, 56, 96)", style=solid]; +"344 quantize_per_tensor_default_11" -> "345 dequantize_per_tensor_default_11" [label="(1, 56, 56, 96)", style=solid]; +"345 dequantize_per_tensor_default_11" -> "350 linear_10" [label="(1, 56, 56, 96)", style=solid]; +"346 linear_10_scale_0" -> "349 dequantize_per_channel_default_7" [label="(384,)", style=solid]; +"347 linear_10_zero_point_0" -> "349 dequantize_per_channel_default_7" [label="(384,)", style=solid]; +"348 _frozen_param13_1" -> "349 dequantize_per_channel_default_7" [label="(384, 96)", style=solid]; +"349 dequantize_per_channel_default_7" -> "350 linear_10" [label="(384, 96)", style=solid]; +"350 linear_10" -> "351 gelu_1" [label="(1, 56, 56, 384)", style=solid]; +"351 gelu_1" -> "352 dropout_6" [label="(1, 56, 56, 384)", style=solid]; +"352 dropout_6" -> "353 dropout_6_0_0_nncf_smooth_quant_0" [label="(1, 56, 56, 384)", style=solid]; +"353 dropout_6_0_0_nncf_smooth_quant_0" -> "354 quantize_per_tensor_default_12" [label="(1, 56, 56, 384)", style=solid]; +"354 quantize_per_tensor_default_12" -> "355 dequantize_per_tensor_default_12" [label="(1, 56, 56, 384)", style=solid]; +"355 dequantize_per_tensor_default_12" -> "360 linear_11" [label="(1, 56, 56, 384)", style=solid]; +"356 linear_11_scale_0" -> "359 dequantize_per_channel_default_8" [label="(96,)", style=solid]; +"357 linear_11_zero_point_0" -> "359 dequantize_per_channel_default_8" [label="(96,)", style=solid]; +"358 _frozen_param14_1" -> "359 dequantize_per_channel_default_8" [label="(96, 384)", style=solid]; +"359 dequantize_per_channel_default_8" -> "360 linear_11" [label="(96, 384)", style=solid]; +"360 linear_11" -> "361 dropout_7" [label="(1, 56, 56, 96)", style=solid]; +"361 dropout_7" -> "362 layer_norm_4" [label="(1, 56, 56, 96)", style=solid]; +"362 layer_norm_4" -> "363 add_6" [label="(1, 56, 56, 96)", style=solid]; +"363 add_6" -> "364 pad_2" [label="(1, 56, 56, 96)", style=solid]; +"364 pad_2" -> "365 slice_25" [label="(1, 56, 56, 96)", style=solid]; +"364 pad_2" -> "368 slice_28" [label="(1, 56, 56, 96)", style=solid]; +"364 pad_2" -> "371 slice_31" [label="(1, 56, 56, 96)", style=solid]; +"364 pad_2" -> "374 slice_34" [label="(1, 56, 56, 96)", style=solid]; +"365 slice_25" -> "366 slice_26" [label="(1, 28, 56, 96)", style=solid]; +"366 slice_26" -> "367 slice_27" [label="(1, 28, 28, 96)", style=solid]; +"367 slice_27" -> "377 cat" [label="(1, 28, 28, 96)", style=solid]; +"368 slice_28" -> "369 slice_29" [label="(1, 28, 56, 96)", style=solid]; +"369 slice_29" -> "370 slice_30" [label="(1, 28, 28, 96)", style=solid]; +"370 slice_30" -> "377 cat" [label="(1, 28, 28, 96)", style=solid]; +"371 slice_31" -> "372 slice_32" [label="(1, 28, 56, 96)", style=solid]; +"372 slice_32" -> "373 slice_33" [label="(1, 28, 28, 96)", style=solid]; +"373 slice_33" -> "377 cat" [label="(1, 28, 28, 96)", style=solid]; +"374 slice_34" -> "375 slice_35" [label="(1, 28, 56, 96)", style=solid]; +"375 slice_35" -> "376 slice_36" [label="(1, 28, 28, 96)", style=solid]; +"376 slice_36" -> "377 cat" [label="(1, 28, 28, 96)", style=solid]; +"377 cat" -> "378 cat_0_0_nncf_smooth_quant_0" [label="(1, 28, 28, 384)", style=solid]; +"378 cat_0_0_nncf_smooth_quant_0" -> "379 quantize_per_tensor_default_13" [label="(1, 28, 28, 384)", style=solid]; +"379 quantize_per_tensor_default_13" -> "380 dequantize_per_tensor_default_13" [label="(1, 28, 28, 384)", style=solid]; +"380 dequantize_per_tensor_default_13" -> "385 linear_12" [label="(1, 28, 28, 384)", style=solid]; +"381 linear_12_scale_0" -> "384 dequantize_per_channel_default_9" [label="(192,)", style=solid]; +"382 linear_12_zero_point_0" -> "384 dequantize_per_channel_default_9" [label="(192,)", style=solid]; +"383 _frozen_param15_1" -> "384 dequantize_per_channel_default_9" [label="(192, 384)", style=solid]; +"384 dequantize_per_channel_default_9" -> "385 linear_12" [label="(192, 384)", style=solid]; +"385 linear_12" -> "386 layer_norm_5" [label="(1, 28, 28, 192)", style=solid]; +"386 layer_norm_5" -> "388 pad_3" [label="(1, 28, 28, 192)", style=solid]; +"386 layer_norm_5" -> "446 add_8" [label="(1, 28, 28, 192)", style=solid]; +"387 _frozen_param17" -> "422 add_7" [label="(1, 6, 64, 64)", style=solid]; +"388 pad_3" -> "389 view_13" [label="(1, 32, 32, 192)", style=solid]; +"389 view_13" -> "390 permute_11" [label="(1, 4, 8, 4, 8, 192)", style=solid]; +"390 permute_11" -> "391 reshape_9" [label="(1, 4, 4, 8, 8, 192)", style=solid]; +"391 reshape_9" -> "393 reshape_9_0_0_nncf_smooth_quant_0" [label="(16, 64, 192)", style=solid]; +"392 _frozen_param18" -> "400 linear_15" [label="(576,)", style=solid]; +"393 reshape_9_0_0_nncf_smooth_quant_0" -> "394 quantize_per_tensor_default_14" [label="(16, 64, 192)", style=solid]; +"394 quantize_per_tensor_default_14" -> "395 dequantize_per_tensor_default_14" [label="(16, 64, 192)", style=solid]; +"395 dequantize_per_tensor_default_14" -> "400 linear_15" [label="(16, 64, 192)", style=solid]; +"396 linear_15_scale_0" -> "399 dequantize_per_channel_default_10" [label="(576,)", style=solid]; +"397 linear_15_zero_point_0" -> "399 dequantize_per_channel_default_10" [label="(576,)", style=solid]; +"398 _frozen_param16_1" -> "399 dequantize_per_channel_default_10" [label="(576, 192)", style=solid]; +"399 dequantize_per_channel_default_10" -> "400 linear_15" [label="(576, 192)", style=solid]; +"400 linear_15" -> "401 reshape_10" [label="(16, 64, 576)", style=solid]; +"401 reshape_10" -> "402 permute_12" [label="(16, 64, 3, 6, 32)", style=solid]; +"402 permute_12" -> "403 select_6" [label="(3, 16, 6, 64, 32)", style=solid]; +"402 permute_12" -> "404 select_7" [label="(3, 16, 6, 64, 32)", style=solid]; +"402 permute_12" -> "405 select_8" [label="(3, 16, 6, 64, 32)", style=solid]; +"403 select_6" -> "406 linalg_vector_norm_4" [label="(16, 6, 64, 32)", style=solid]; +"403 select_6" -> "408 expand_as_4" [label="(16, 6, 64, 32)", style=solid]; +"403 select_6" -> "409 div_4" [label="(16, 6, 64, 32)", style=solid]; +"404 select_7" -> "412 linalg_vector_norm_5" [label="(16, 6, 64, 32)", style=solid]; +"404 select_7" -> "414 expand_as_5" [label="(16, 6, 64, 32)", style=solid]; +"404 select_7" -> "415 div_5" [label="(16, 6, 64, 32)", style=solid]; +"405 select_8" -> "425 matmul_5" [label="(16, 6, 64, 32)", style=solid]; +"406 linalg_vector_norm_4" -> "407 clamp_min_4" [label="(16, 6, 64, 1)", style=solid]; +"407 clamp_min_4" -> "408 expand_as_4" [label="(16, 6, 64, 1)", style=solid]; +"408 expand_as_4" -> "409 div_4" [label="(16, 6, 64, 32)", style=solid]; +"409 div_4" -> "410 quantize_per_tensor_default_15" [label="(16, 6, 64, 32)", style=solid]; +"410 quantize_per_tensor_default_15" -> "411 dequantize_per_tensor_default_15" [label="(16, 6, 64, 32)", style=solid]; +"411 dequantize_per_tensor_default_15" -> "419 matmul_4" [label="(16, 6, 64, 32)", style=solid]; +"412 linalg_vector_norm_5" -> "413 clamp_min_5" [label="(16, 6, 64, 1)", style=solid]; +"413 clamp_min_5" -> "414 expand_as_5" [label="(16, 6, 64, 1)", style=solid]; +"414 expand_as_5" -> "415 div_5" [label="(16, 6, 64, 32)", style=solid]; +"415 div_5" -> "416 quantize_per_tensor_default_16" [label="(16, 6, 64, 32)", style=solid]; +"416 quantize_per_tensor_default_16" -> "417 dequantize_per_tensor_default_16" [label="(16, 6, 64, 32)", style=solid]; +"417 dequantize_per_tensor_default_16" -> "418 transpose_4" [label="(16, 6, 64, 32)", style=solid]; +"418 transpose_4" -> "419 matmul_4" [label="(16, 6, 32, 64)", style=solid]; +"419 matmul_4" -> "421 mul_5" [label="(16, 6, 64, 64)", style=solid]; +"420 _frozen_param20" -> "421 mul_5" [label="(6, 1, 1)", style=solid]; +"421 mul_5" -> "422 add_7" [label="(16, 6, 64, 64)", style=solid]; +"422 add_7" -> "423 softmax_2" [label="(16, 6, 64, 64)", style=solid]; +"423 softmax_2" -> "424 dropout_8" [label="(16, 6, 64, 64)", style=solid]; +"424 dropout_8" -> "425 matmul_5" [label="(16, 6, 64, 64)", style=solid]; +"425 matmul_5" -> "426 transpose_5" [label="(16, 6, 64, 32)", style=solid]; +"426 transpose_5" -> "427 reshape_11" [label="(16, 64, 6, 32)", style=solid]; +"427 reshape_11" -> "428 reshape_11_0_0_nncf_smooth_quant_0" [label="(16, 64, 192)", style=solid]; +"428 reshape_11_0_0_nncf_smooth_quant_0" -> "429 quantize_per_tensor_default_17" [label="(16, 64, 192)", style=solid]; +"429 quantize_per_tensor_default_17" -> "430 dequantize_per_tensor_default_17" [label="(16, 64, 192)", style=solid]; +"430 dequantize_per_tensor_default_17" -> "435 linear_16" [label="(16, 64, 192)", style=solid]; +"431 linear_16_scale_0" -> "434 dequantize_per_channel_default_11" [label="(192,)", style=solid]; +"432 linear_16_zero_point_0" -> "434 dequantize_per_channel_default_11" [label="(192,)", style=solid]; +"433 _frozen_param19_1" -> "434 dequantize_per_channel_default_11" [label="(192, 192)", style=solid]; +"434 dequantize_per_channel_default_11" -> "435 linear_16" [label="(192, 192)", style=solid]; +"435 linear_16" -> "436 dropout_9" [label="(16, 64, 192)", style=solid]; +"436 dropout_9" -> "437 view_14" [label="(16, 64, 192)", style=solid]; +"437 view_14" -> "438 permute_13" [label="(1, 4, 4, 8, 8, 192)", style=solid]; +"438 permute_13" -> "439 reshape_12" [label="(1, 4, 8, 4, 8, 192)", style=solid]; +"439 reshape_12" -> "440 slice_38" [label="(1, 32, 32, 192)", style=solid]; +"440 slice_38" -> "441 slice_39" [label="(1, 32, 32, 192)", style=solid]; +"441 slice_39" -> "442 slice_40" [label="(1, 28, 32, 192)", style=solid]; +"442 slice_40" -> "443 slice_41" [label="(1, 28, 28, 192)", style=solid]; +"443 slice_41" -> "444 contiguous_3" [label="(1, 28, 28, 192)", style=solid]; +"444 contiguous_3" -> "445 layer_norm_6" [label="(1, 28, 28, 192)", style=solid]; +"445 layer_norm_6" -> "446 add_8" [label="(1, 28, 28, 192)", style=solid]; +"446 add_8" -> "447 add_8_0_0_nncf_smooth_quant_0" [label="(1, 28, 28, 192)", style=solid]; +"446 add_8" -> "467 add_9" [label="(1, 28, 28, 192)", style=solid]; +"447 add_8_0_0_nncf_smooth_quant_0" -> "448 quantize_per_tensor_default_18" [label="(1, 28, 28, 192)", style=solid]; +"448 quantize_per_tensor_default_18" -> "449 dequantize_per_tensor_default_18" [label="(1, 28, 28, 192)", style=solid]; +"449 dequantize_per_tensor_default_18" -> "454 linear_17" [label="(1, 28, 28, 192)", style=solid]; +"450 linear_17_scale_0" -> "453 dequantize_per_channel_default_12" [label="(768,)", style=solid]; +"451 linear_17_zero_point_0" -> "453 dequantize_per_channel_default_12" [label="(768,)", style=solid]; +"452 _frozen_param23_1" -> "453 dequantize_per_channel_default_12" [label="(768, 192)", style=solid]; +"453 dequantize_per_channel_default_12" -> "454 linear_17" [label="(768, 192)", style=solid]; +"454 linear_17" -> "455 gelu_2" [label="(1, 28, 28, 768)", style=solid]; +"455 gelu_2" -> "456 dropout_10" [label="(1, 28, 28, 768)", style=solid]; +"456 dropout_10" -> "457 dropout_10_0_0_nncf_smooth_quant_0" [label="(1, 28, 28, 768)", style=solid]; +"457 dropout_10_0_0_nncf_smooth_quant_0" -> "458 quantize_per_tensor_default_19" [label="(1, 28, 28, 768)", style=solid]; +"458 quantize_per_tensor_default_19" -> "459 dequantize_per_tensor_default_19" [label="(1, 28, 28, 768)", style=solid]; +"459 dequantize_per_tensor_default_19" -> "464 linear_18" [label="(1, 28, 28, 768)", style=solid]; +"460 linear_18_scale_0" -> "463 dequantize_per_channel_default_13" [label="(192,)", style=solid]; +"461 linear_18_zero_point_0" -> "463 dequantize_per_channel_default_13" [label="(192,)", style=solid]; +"462 _frozen_param25_1" -> "463 dequantize_per_channel_default_13" [label="(192, 768)", style=solid]; +"463 dequantize_per_channel_default_13" -> "464 linear_18" [label="(192, 768)", style=solid]; +"464 linear_18" -> "465 dropout_11" [label="(1, 28, 28, 192)", style=solid]; +"465 dropout_11" -> "466 layer_norm_7" [label="(1, 28, 28, 192)", style=solid]; +"466 layer_norm_7" -> "467 add_9" [label="(1, 28, 28, 192)", style=solid]; +"467 add_9" -> "469 pad_4" [label="(1, 28, 28, 192)", style=solid]; +"467 add_9" -> "545 add_12" [label="(1, 28, 28, 192)", style=solid]; +"468 _frozen_param21" -> "504 add_10" [label="(1, 6, 64, 64)", style=solid]; +"469 pad_4" -> "470 roll_2" [label="(1, 32, 32, 192)", style=solid]; +"470 roll_2" -> "471 view_17" [label="(1, 32, 32, 192)", style=solid]; +"471 view_17" -> "472 permute_15" [label="(1, 4, 8, 4, 8, 192)", style=solid]; +"472 permute_15" -> "473 reshape_13" [label="(1, 4, 4, 8, 8, 192)", style=solid]; +"473 reshape_13" -> "475 reshape_13_0_0_nncf_smooth_quant_0" [label="(16, 64, 192)", style=solid]; +"473 reshape_13" -> "505 new_zeros_1" [label="(16, 64, 192)", style=solid]; +"474 _frozen_param22" -> "482 linear_21" [label="(576,)", style=solid]; +"475 reshape_13_0_0_nncf_smooth_quant_0" -> "476 quantize_per_tensor_default_20" [label="(16, 64, 192)", style=solid]; +"476 quantize_per_tensor_default_20" -> "477 dequantize_per_tensor_default_20" [label="(16, 64, 192)", style=solid]; +"477 dequantize_per_tensor_default_20" -> "482 linear_21" [label="(16, 64, 192)", style=solid]; +"478 linear_21_scale_0" -> "481 dequantize_per_channel_default_14" [label="(576,)", style=solid]; +"479 linear_21_zero_point_0" -> "481 dequantize_per_channel_default_14" [label="(576,)", style=solid]; +"480 _frozen_param26_1" -> "481 dequantize_per_channel_default_14" [label="(576, 192)", style=solid]; +"481 dequantize_per_channel_default_14" -> "482 linear_21" [label="(576, 192)", style=solid]; +"482 linear_21" -> "483 reshape_14" [label="(16, 64, 576)", style=solid]; +"483 reshape_14" -> "484 permute_16" [label="(16, 64, 3, 6, 32)", style=solid]; +"484 permute_16" -> "485 select_9" [label="(3, 16, 6, 64, 32)", style=solid]; +"484 permute_16" -> "486 select_10" [label="(3, 16, 6, 64, 32)", style=solid]; +"484 permute_16" -> "487 select_11" [label="(3, 16, 6, 64, 32)", style=solid]; +"485 select_9" -> "488 linalg_vector_norm_6" [label="(16, 6, 64, 32)", style=solid]; +"485 select_9" -> "490 expand_as_6" [label="(16, 6, 64, 32)", style=solid]; +"485 select_9" -> "491 div_6" [label="(16, 6, 64, 32)", style=solid]; +"486 select_10" -> "494 linalg_vector_norm_7" [label="(16, 6, 64, 32)", style=solid]; +"486 select_10" -> "496 expand_as_7" [label="(16, 6, 64, 32)", style=solid]; +"486 select_10" -> "497 div_7" [label="(16, 6, 64, 32)", style=solid]; +"487 select_11" -> "523 matmul_7" [label="(16, 6, 64, 32)", style=solid]; +"488 linalg_vector_norm_6" -> "489 clamp_min_6" [label="(16, 6, 64, 1)", style=solid]; +"489 clamp_min_6" -> "490 expand_as_6" [label="(16, 6, 64, 1)", style=solid]; +"490 expand_as_6" -> "491 div_6" [label="(16, 6, 64, 32)", style=solid]; +"491 div_6" -> "492 quantize_per_tensor_default_21" [label="(16, 6, 64, 32)", style=solid]; +"492 quantize_per_tensor_default_21" -> "493 dequantize_per_tensor_default_21" [label="(16, 6, 64, 32)", style=solid]; +"493 dequantize_per_tensor_default_21" -> "501 matmul_6" [label="(16, 6, 64, 32)", style=solid]; +"494 linalg_vector_norm_7" -> "495 clamp_min_7" [label="(16, 6, 64, 1)", style=solid]; +"495 clamp_min_7" -> "496 expand_as_7" [label="(16, 6, 64, 1)", style=solid]; +"496 expand_as_7" -> "497 div_7" [label="(16, 6, 64, 32)", style=solid]; +"497 div_7" -> "498 quantize_per_tensor_default_22" [label="(16, 6, 64, 32)", style=solid]; +"498 quantize_per_tensor_default_22" -> "499 dequantize_per_tensor_default_22" [label="(16, 6, 64, 32)", style=solid]; +"499 dequantize_per_tensor_default_22" -> "500 transpose_6" [label="(16, 6, 64, 32)", style=solid]; +"500 transpose_6" -> "501 matmul_6" [label="(16, 6, 32, 64)", style=solid]; +"501 matmul_6" -> "503 mul_7" [label="(16, 6, 64, 64)", style=solid]; +"502 _frozen_param24" -> "503 mul_7" [label="(6, 1, 1)", style=solid]; +"503 mul_7" -> "504 add_10" [label="(16, 6, 64, 64)", style=solid]; +"504 add_10" -> "516 view_19" [label="(16, 6, 64, 64)", style=solid]; +"505 new_zeros_1" -> "506 view_18" [label="(32, 32)", style=solid]; +"506 view_18" -> "507 permute_17" [label="(4, 8, 4, 8)", style=solid]; +"507 permute_17" -> "508 reshape_15" [label="(4, 4, 8, 8)", style=solid]; +"508 reshape_15" -> "509 unsqueeze_8" [label="(16, 64)", style=solid]; +"508 reshape_15" -> "510 unsqueeze_9" [label="(16, 64)", style=solid]; +"509 unsqueeze_8" -> "511 sub_1" [label="(16, 1, 64)", style=solid]; +"510 unsqueeze_9" -> "511 sub_1" [label="(16, 64, 1)", style=solid]; +"511 sub_1" -> "512 ne_1" [label="(16, 64, 64)", style=solid]; +"511 sub_1" -> "513 masked_fill_2" [label="(16, 64, 64)", style=solid]; +"511 sub_1" -> "514 eq_1" [label="(16, 64, 64)", style=solid]; +"512 ne_1" -> "513 masked_fill_2" [label="(16, 64, 64)", style=solid]; +"513 masked_fill_2" -> "515 masked_fill_3" [label="(16, 64, 64)", style=solid]; +"514 eq_1" -> "515 masked_fill_3" [label="(16, 64, 64)", style=solid]; +"515 masked_fill_3" -> "517 unsqueeze_10" [label="(16, 64, 64)", style=solid]; +"516 view_19" -> "519 add_11" [label="(1, 16, 6, 64, 64)", style=solid]; +"517 unsqueeze_10" -> "518 unsqueeze_11" [label="(16, 1, 64, 64)", style=solid]; +"518 unsqueeze_11" -> "519 add_11" [label="(1, 16, 1, 64, 64)", style=solid]; +"519 add_11" -> "520 view_20" [label="(1, 16, 6, 64, 64)", style=solid]; +"520 view_20" -> "521 softmax_3" [label="(16, 6, 64, 64)", style=solid]; +"521 softmax_3" -> "522 dropout_12" [label="(16, 6, 64, 64)", style=solid]; +"522 dropout_12" -> "523 matmul_7" [label="(16, 6, 64, 64)", style=solid]; +"523 matmul_7" -> "524 transpose_7" [label="(16, 6, 64, 32)", style=solid]; +"524 transpose_7" -> "525 reshape_16" [label="(16, 64, 6, 32)", style=solid]; +"525 reshape_16" -> "526 reshape_16_0_0_nncf_smooth_quant_0" [label="(16, 64, 192)", style=solid]; +"526 reshape_16_0_0_nncf_smooth_quant_0" -> "527 quantize_per_tensor_default_23" [label="(16, 64, 192)", style=solid]; +"527 quantize_per_tensor_default_23" -> "528 dequantize_per_tensor_default_23" [label="(16, 64, 192)", style=solid]; +"528 dequantize_per_tensor_default_23" -> "533 linear_22" [label="(16, 64, 192)", style=solid]; +"529 linear_22_scale_0" -> "532 dequantize_per_channel_default_15" [label="(192,)", style=solid]; +"530 linear_22_zero_point_0" -> "532 dequantize_per_channel_default_15" [label="(192,)", style=solid]; +"531 _frozen_param27_1" -> "532 dequantize_per_channel_default_15" [label="(192, 192)", style=solid]; +"532 dequantize_per_channel_default_15" -> "533 linear_22" [label="(192, 192)", style=solid]; +"533 linear_22" -> "534 dropout_13" [label="(16, 64, 192)", style=solid]; +"534 dropout_13" -> "535 view_21" [label="(16, 64, 192)", style=solid]; +"535 view_21" -> "536 permute_18" [label="(1, 4, 4, 8, 8, 192)", style=solid]; +"536 permute_18" -> "537 reshape_17" [label="(1, 4, 8, 4, 8, 192)", style=solid]; +"537 reshape_17" -> "538 roll_3" [label="(1, 32, 32, 192)", style=solid]; +"538 roll_3" -> "539 slice_61" [label="(1, 32, 32, 192)", style=solid]; +"539 slice_61" -> "540 slice_62" [label="(1, 32, 32, 192)", style=solid]; +"540 slice_62" -> "541 slice_63" [label="(1, 28, 32, 192)", style=solid]; +"541 slice_63" -> "542 slice_64" [label="(1, 28, 28, 192)", style=solid]; +"542 slice_64" -> "543 contiguous_5" [label="(1, 28, 28, 192)", style=solid]; +"543 contiguous_5" -> "544 layer_norm_8" [label="(1, 28, 28, 192)", style=solid]; +"544 layer_norm_8" -> "545 add_12" [label="(1, 28, 28, 192)", style=solid]; +"545 add_12" -> "546 add_12_0_0_nncf_smooth_quant_0" [label="(1, 28, 28, 192)", style=solid]; +"545 add_12" -> "566 add_13" [label="(1, 28, 28, 192)", style=solid]; +"546 add_12_0_0_nncf_smooth_quant_0" -> "547 quantize_per_tensor_default_24" [label="(1, 28, 28, 192)", style=solid]; +"547 quantize_per_tensor_default_24" -> "548 dequantize_per_tensor_default_24" [label="(1, 28, 28, 192)", style=solid]; +"548 dequantize_per_tensor_default_24" -> "553 linear_23" [label="(1, 28, 28, 192)", style=solid]; +"549 linear_23_scale_0" -> "552 dequantize_per_channel_default_16" [label="(768,)", style=solid]; +"550 linear_23_zero_point_0" -> "552 dequantize_per_channel_default_16" [label="(768,)", style=solid]; +"551 _frozen_param28_1" -> "552 dequantize_per_channel_default_16" [label="(768, 192)", style=solid]; +"552 dequantize_per_channel_default_16" -> "553 linear_23" [label="(768, 192)", style=solid]; +"553 linear_23" -> "554 gelu_3" [label="(1, 28, 28, 768)", style=solid]; +"554 gelu_3" -> "555 dropout_14" [label="(1, 28, 28, 768)", style=solid]; +"555 dropout_14" -> "556 dropout_14_0_0_nncf_smooth_quant_0" [label="(1, 28, 28, 768)", style=solid]; +"556 dropout_14_0_0_nncf_smooth_quant_0" -> "557 quantize_per_tensor_default_25" [label="(1, 28, 28, 768)", style=solid]; +"557 quantize_per_tensor_default_25" -> "558 dequantize_per_tensor_default_25" [label="(1, 28, 28, 768)", style=solid]; +"558 dequantize_per_tensor_default_25" -> "563 linear_24" [label="(1, 28, 28, 768)", style=solid]; +"559 linear_24_scale_0" -> "562 dequantize_per_channel_default_17" [label="(192,)", style=solid]; +"560 linear_24_zero_point_0" -> "562 dequantize_per_channel_default_17" [label="(192,)", style=solid]; +"561 _frozen_param29_1" -> "562 dequantize_per_channel_default_17" [label="(192, 768)", style=solid]; +"562 dequantize_per_channel_default_17" -> "563 linear_24" [label="(192, 768)", style=solid]; +"563 linear_24" -> "564 dropout_15" [label="(1, 28, 28, 192)", style=solid]; +"564 dropout_15" -> "565 layer_norm_9" [label="(1, 28, 28, 192)", style=solid]; +"565 layer_norm_9" -> "566 add_13" [label="(1, 28, 28, 192)", style=solid]; +"566 add_13" -> "567 pad_5" [label="(1, 28, 28, 192)", style=solid]; +"567 pad_5" -> "568 slice_65" [label="(1, 28, 28, 192)", style=solid]; +"567 pad_5" -> "571 slice_68" [label="(1, 28, 28, 192)", style=solid]; +"567 pad_5" -> "574 slice_71" [label="(1, 28, 28, 192)", style=solid]; +"567 pad_5" -> "577 slice_74" [label="(1, 28, 28, 192)", style=solid]; +"568 slice_65" -> "569 slice_66" [label="(1, 14, 28, 192)", style=solid]; +"569 slice_66" -> "570 slice_67" [label="(1, 14, 14, 192)", style=solid]; +"570 slice_67" -> "580 cat_1" [label="(1, 14, 14, 192)", style=solid]; +"571 slice_68" -> "572 slice_69" [label="(1, 14, 28, 192)", style=solid]; +"572 slice_69" -> "573 slice_70" [label="(1, 14, 14, 192)", style=solid]; +"573 slice_70" -> "580 cat_1" [label="(1, 14, 14, 192)", style=solid]; +"574 slice_71" -> "575 slice_72" [label="(1, 14, 28, 192)", style=solid]; +"575 slice_72" -> "576 slice_73" [label="(1, 14, 14, 192)", style=solid]; +"576 slice_73" -> "580 cat_1" [label="(1, 14, 14, 192)", style=solid]; +"577 slice_74" -> "578 slice_75" [label="(1, 14, 28, 192)", style=solid]; +"578 slice_75" -> "579 slice_76" [label="(1, 14, 14, 192)", style=solid]; +"579 slice_76" -> "580 cat_1" [label="(1, 14, 14, 192)", style=solid]; +"580 cat_1" -> "581 cat_1_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 768)", style=solid]; +"581 cat_1_0_0_nncf_smooth_quant_0" -> "582 quantize_per_tensor_default_26" [label="(1, 14, 14, 768)", style=solid]; +"582 quantize_per_tensor_default_26" -> "583 dequantize_per_tensor_default_26" [label="(1, 14, 14, 768)", style=solid]; +"583 dequantize_per_tensor_default_26" -> "588 linear_25" [label="(1, 14, 14, 768)", style=solid]; +"584 linear_25_scale_0" -> "587 dequantize_per_channel_default_18" [label="(384,)", style=solid]; +"585 linear_25_zero_point_0" -> "587 dequantize_per_channel_default_18" [label="(384,)", style=solid]; +"586 _frozen_param30_1" -> "587 dequantize_per_channel_default_18" [label="(384, 768)", style=solid]; +"587 dequantize_per_channel_default_18" -> "588 linear_25" [label="(384, 768)", style=solid]; +"588 linear_25" -> "589 layer_norm_10" [label="(1, 14, 14, 384)", style=solid]; +"589 layer_norm_10" -> "591 pad_6" [label="(1, 14, 14, 384)", style=solid]; +"589 layer_norm_10" -> "649 add_15" [label="(1, 14, 14, 384)", style=solid]; +"590 _frozen_param34" -> "625 add_14" [label="(1, 12, 64, 64)", style=solid]; +"591 pad_6" -> "592 view_24" [label="(1, 16, 16, 384)", style=solid]; +"592 view_24" -> "593 permute_20" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"593 permute_20" -> "594 reshape_18" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"594 reshape_18" -> "596 reshape_18_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"595 _frozen_param35" -> "603 linear_28" [label="(1152,)", style=solid]; +"596 reshape_18_0_0_nncf_smooth_quant_0" -> "597 quantize_per_tensor_default_27" [label="(4, 64, 384)", style=solid]; +"597 quantize_per_tensor_default_27" -> "598 dequantize_per_tensor_default_27" [label="(4, 64, 384)", style=solid]; +"598 dequantize_per_tensor_default_27" -> "603 linear_28" [label="(4, 64, 384)", style=solid]; +"599 linear_28_scale_0" -> "602 dequantize_per_channel_default_19" [label="(1152,)", style=solid]; +"600 linear_28_zero_point_0" -> "602 dequantize_per_channel_default_19" [label="(1152,)", style=solid]; +"601 _frozen_param31_1" -> "602 dequantize_per_channel_default_19" [label="(1152, 384)", style=solid]; +"602 dequantize_per_channel_default_19" -> "603 linear_28" [label="(1152, 384)", style=solid]; +"603 linear_28" -> "604 reshape_19" [label="(4, 64, 1152)", style=solid]; +"604 reshape_19" -> "605 permute_21" [label="(4, 64, 3, 12, 32)", style=solid]; +"605 permute_21" -> "606 select_12" [label="(3, 4, 12, 64, 32)", style=solid]; +"605 permute_21" -> "607 select_13" [label="(3, 4, 12, 64, 32)", style=solid]; +"605 permute_21" -> "608 select_14" [label="(3, 4, 12, 64, 32)", style=solid]; +"606 select_12" -> "609 linalg_vector_norm_8" [label="(4, 12, 64, 32)", style=solid]; +"606 select_12" -> "611 expand_as_8" [label="(4, 12, 64, 32)", style=solid]; +"606 select_12" -> "612 div_8" [label="(4, 12, 64, 32)", style=solid]; +"607 select_13" -> "615 linalg_vector_norm_9" [label="(4, 12, 64, 32)", style=solid]; +"607 select_13" -> "617 expand_as_9" [label="(4, 12, 64, 32)", style=solid]; +"607 select_13" -> "618 div_9" [label="(4, 12, 64, 32)", style=solid]; +"608 select_14" -> "628 matmul_9" [label="(4, 12, 64, 32)", style=solid]; +"609 linalg_vector_norm_8" -> "610 clamp_min_8" [label="(4, 12, 64, 1)", style=solid]; +"610 clamp_min_8" -> "611 expand_as_8" [label="(4, 12, 64, 1)", style=solid]; +"611 expand_as_8" -> "612 div_8" [label="(4, 12, 64, 32)", style=solid]; +"612 div_8" -> "613 quantize_per_tensor_default_28" [label="(4, 12, 64, 32)", style=solid]; +"613 quantize_per_tensor_default_28" -> "614 dequantize_per_tensor_default_28" [label="(4, 12, 64, 32)", style=solid]; +"614 dequantize_per_tensor_default_28" -> "622 matmul_8" [label="(4, 12, 64, 32)", style=solid]; +"615 linalg_vector_norm_9" -> "616 clamp_min_9" [label="(4, 12, 64, 1)", style=solid]; +"616 clamp_min_9" -> "617 expand_as_9" [label="(4, 12, 64, 1)", style=solid]; +"617 expand_as_9" -> "618 div_9" [label="(4, 12, 64, 32)", style=solid]; +"618 div_9" -> "619 quantize_per_tensor_default_29" [label="(4, 12, 64, 32)", style=solid]; +"619 quantize_per_tensor_default_29" -> "620 dequantize_per_tensor_default_29" [label="(4, 12, 64, 32)", style=solid]; +"620 dequantize_per_tensor_default_29" -> "621 transpose_8" [label="(4, 12, 64, 32)", style=solid]; +"621 transpose_8" -> "622 matmul_8" [label="(4, 12, 32, 64)", style=solid]; +"622 matmul_8" -> "624 mul_9" [label="(4, 12, 64, 64)", style=solid]; +"623 _frozen_param37" -> "624 mul_9" [label="(12, 1, 1)", style=solid]; +"624 mul_9" -> "625 add_14" [label="(4, 12, 64, 64)", style=solid]; +"625 add_14" -> "626 softmax_4" [label="(4, 12, 64, 64)", style=solid]; +"626 softmax_4" -> "627 dropout_16" [label="(4, 12, 64, 64)", style=solid]; +"627 dropout_16" -> "628 matmul_9" [label="(4, 12, 64, 64)", style=solid]; +"628 matmul_9" -> "629 transpose_9" [label="(4, 12, 64, 32)", style=solid]; +"629 transpose_9" -> "630 reshape_20" [label="(4, 64, 12, 32)", style=solid]; +"630 reshape_20" -> "631 reshape_20_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"631 reshape_20_0_0_nncf_smooth_quant_0" -> "632 quantize_per_tensor_default_30" [label="(4, 64, 384)", style=solid]; +"632 quantize_per_tensor_default_30" -> "633 dequantize_per_tensor_default_30" [label="(4, 64, 384)", style=solid]; +"633 dequantize_per_tensor_default_30" -> "638 linear_29" [label="(4, 64, 384)", style=solid]; +"634 linear_29_scale_0" -> "637 dequantize_per_channel_default_20" [label="(384,)", style=solid]; +"635 linear_29_zero_point_0" -> "637 dequantize_per_channel_default_20" [label="(384,)", style=solid]; +"636 _frozen_param32_1" -> "637 dequantize_per_channel_default_20" [label="(384, 384)", style=solid]; +"637 dequantize_per_channel_default_20" -> "638 linear_29" [label="(384, 384)", style=solid]; +"638 linear_29" -> "639 dropout_17" [label="(4, 64, 384)", style=solid]; +"639 dropout_17" -> "640 view_25" [label="(4, 64, 384)", style=solid]; +"640 view_25" -> "641 permute_22" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"641 permute_22" -> "642 reshape_21" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"642 reshape_21" -> "643 slice_78" [label="(1, 16, 16, 384)", style=solid]; +"643 slice_78" -> "644 slice_79" [label="(1, 16, 16, 384)", style=solid]; +"644 slice_79" -> "645 slice_80" [label="(1, 14, 16, 384)", style=solid]; +"645 slice_80" -> "646 slice_81" [label="(1, 14, 14, 384)", style=solid]; +"646 slice_81" -> "647 contiguous_7" [label="(1, 14, 14, 384)", style=solid]; +"647 contiguous_7" -> "648 layer_norm_11" [label="(1, 14, 14, 384)", style=solid]; +"648 layer_norm_11" -> "649 add_15" [label="(1, 14, 14, 384)", style=solid]; +"649 add_15" -> "650 add_15_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"649 add_15" -> "670 add_16" [label="(1, 14, 14, 384)", style=solid]; +"650 add_15_0_0_nncf_smooth_quant_0" -> "651 quantize_per_tensor_default_31" [label="(1, 14, 14, 384)", style=solid]; +"651 quantize_per_tensor_default_31" -> "652 dequantize_per_tensor_default_31" [label="(1, 14, 14, 384)", style=solid]; +"652 dequantize_per_tensor_default_31" -> "657 linear_30" [label="(1, 14, 14, 384)", style=solid]; +"653 linear_30_scale_0" -> "656 dequantize_per_channel_default_21" [label="(1536,)", style=solid]; +"654 linear_30_zero_point_0" -> "656 dequantize_per_channel_default_21" [label="(1536,)", style=solid]; +"655 _frozen_param33_1" -> "656 dequantize_per_channel_default_21" [label="(1536, 384)", style=solid]; +"656 dequantize_per_channel_default_21" -> "657 linear_30" [label="(1536, 384)", style=solid]; +"657 linear_30" -> "658 gelu_4" [label="(1, 14, 14, 1536)", style=solid]; +"658 gelu_4" -> "659 dropout_18" [label="(1, 14, 14, 1536)", style=solid]; +"659 dropout_18" -> "660 dropout_18_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"660 dropout_18_0_0_nncf_smooth_quant_0" -> "661 quantize_per_tensor_default_32" [label="(1, 14, 14, 1536)", style=solid]; +"661 quantize_per_tensor_default_32" -> "662 dequantize_per_tensor_default_32" [label="(1, 14, 14, 1536)", style=solid]; +"662 dequantize_per_tensor_default_32" -> "667 linear_31" [label="(1, 14, 14, 1536)", style=solid]; +"663 linear_31_scale_0" -> "666 dequantize_per_channel_default_22" [label="(384,)", style=solid]; +"664 linear_31_zero_point_0" -> "666 dequantize_per_channel_default_22" [label="(384,)", style=solid]; +"665 _frozen_param36_1" -> "666 dequantize_per_channel_default_22" [label="(384, 1536)", style=solid]; +"666 dequantize_per_channel_default_22" -> "667 linear_31" [label="(384, 1536)", style=solid]; +"667 linear_31" -> "668 dropout_19" [label="(1, 14, 14, 384)", style=solid]; +"668 dropout_19" -> "669 layer_norm_12" [label="(1, 14, 14, 384)", style=solid]; +"669 layer_norm_12" -> "670 add_16" [label="(1, 14, 14, 384)", style=solid]; +"670 add_16" -> "672 pad_7" [label="(1, 14, 14, 384)", style=solid]; +"670 add_16" -> "748 add_19" [label="(1, 14, 14, 384)", style=solid]; +"671 _frozen_param38" -> "707 add_17" [label="(1, 12, 64, 64)", style=solid]; +"672 pad_7" -> "673 roll_4" [label="(1, 16, 16, 384)", style=solid]; +"673 roll_4" -> "674 view_28" [label="(1, 16, 16, 384)", style=solid]; +"674 view_28" -> "675 permute_24" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"675 permute_24" -> "676 reshape_22" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"676 reshape_22" -> "678 reshape_22_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"676 reshape_22" -> "708 new_zeros_2" [label="(4, 64, 384)", style=solid]; +"677 _frozen_param39" -> "685 linear_34" [label="(1152,)", style=solid]; +"678 reshape_22_0_0_nncf_smooth_quant_0" -> "679 quantize_per_tensor_default_33" [label="(4, 64, 384)", style=solid]; +"679 quantize_per_tensor_default_33" -> "680 dequantize_per_tensor_default_33" [label="(4, 64, 384)", style=solid]; +"680 dequantize_per_tensor_default_33" -> "685 linear_34" [label="(4, 64, 384)", style=solid]; +"681 linear_34_scale_0" -> "684 dequantize_per_channel_default_23" [label="(1152,)", style=solid]; +"682 linear_34_zero_point_0" -> "684 dequantize_per_channel_default_23" [label="(1152,)", style=solid]; +"683 _frozen_param40_1" -> "684 dequantize_per_channel_default_23" [label="(1152, 384)", style=solid]; +"684 dequantize_per_channel_default_23" -> "685 linear_34" [label="(1152, 384)", style=solid]; +"685 linear_34" -> "686 reshape_23" [label="(4, 64, 1152)", style=solid]; +"686 reshape_23" -> "687 permute_25" [label="(4, 64, 3, 12, 32)", style=solid]; +"687 permute_25" -> "688 select_15" [label="(3, 4, 12, 64, 32)", style=solid]; +"687 permute_25" -> "689 select_16" [label="(3, 4, 12, 64, 32)", style=solid]; +"687 permute_25" -> "690 select_17" [label="(3, 4, 12, 64, 32)", style=solid]; +"688 select_15" -> "691 linalg_vector_norm_10" [label="(4, 12, 64, 32)", style=solid]; +"688 select_15" -> "693 expand_as_10" [label="(4, 12, 64, 32)", style=solid]; +"688 select_15" -> "694 div_10" [label="(4, 12, 64, 32)", style=solid]; +"689 select_16" -> "697 linalg_vector_norm_11" [label="(4, 12, 64, 32)", style=solid]; +"689 select_16" -> "699 expand_as_11" [label="(4, 12, 64, 32)", style=solid]; +"689 select_16" -> "700 div_11" [label="(4, 12, 64, 32)", style=solid]; +"690 select_17" -> "726 matmul_11" [label="(4, 12, 64, 32)", style=solid]; +"691 linalg_vector_norm_10" -> "692 clamp_min_10" [label="(4, 12, 64, 1)", style=solid]; +"692 clamp_min_10" -> "693 expand_as_10" [label="(4, 12, 64, 1)", style=solid]; +"693 expand_as_10" -> "694 div_10" [label="(4, 12, 64, 32)", style=solid]; +"694 div_10" -> "695 quantize_per_tensor_default_34" [label="(4, 12, 64, 32)", style=solid]; +"695 quantize_per_tensor_default_34" -> "696 dequantize_per_tensor_default_34" [label="(4, 12, 64, 32)", style=solid]; +"696 dequantize_per_tensor_default_34" -> "704 matmul_10" [label="(4, 12, 64, 32)", style=solid]; +"697 linalg_vector_norm_11" -> "698 clamp_min_11" [label="(4, 12, 64, 1)", style=solid]; +"698 clamp_min_11" -> "699 expand_as_11" [label="(4, 12, 64, 1)", style=solid]; +"699 expand_as_11" -> "700 div_11" [label="(4, 12, 64, 32)", style=solid]; +"700 div_11" -> "701 quantize_per_tensor_default_35" [label="(4, 12, 64, 32)", style=solid]; +"701 quantize_per_tensor_default_35" -> "702 dequantize_per_tensor_default_35" [label="(4, 12, 64, 32)", style=solid]; +"702 dequantize_per_tensor_default_35" -> "703 transpose_10" [label="(4, 12, 64, 32)", style=solid]; +"703 transpose_10" -> "704 matmul_10" [label="(4, 12, 32, 64)", style=solid]; +"704 matmul_10" -> "706 mul_11" [label="(4, 12, 64, 64)", style=solid]; +"705 _frozen_param41" -> "706 mul_11" [label="(12, 1, 1)", style=solid]; +"706 mul_11" -> "707 add_17" [label="(4, 12, 64, 64)", style=solid]; +"707 add_17" -> "719 view_30" [label="(4, 12, 64, 64)", style=solid]; +"708 new_zeros_2" -> "709 view_29" [label="(16, 16)", style=solid]; +"709 view_29" -> "710 permute_26" [label="(2, 8, 2, 8)", style=solid]; +"710 permute_26" -> "711 reshape_24" [label="(2, 2, 8, 8)", style=solid]; +"711 reshape_24" -> "712 unsqueeze_14" [label="(4, 64)", style=solid]; +"711 reshape_24" -> "713 unsqueeze_15" [label="(4, 64)", style=solid]; +"712 unsqueeze_14" -> "714 sub_2" [label="(4, 1, 64)", style=solid]; +"713 unsqueeze_15" -> "714 sub_2" [label="(4, 64, 1)", style=solid]; +"714 sub_2" -> "715 ne_2" [label="(4, 64, 64)", style=solid]; +"714 sub_2" -> "716 masked_fill_4" [label="(4, 64, 64)", style=solid]; +"714 sub_2" -> "717 eq_2" [label="(4, 64, 64)", style=solid]; +"715 ne_2" -> "716 masked_fill_4" [label="(4, 64, 64)", style=solid]; +"716 masked_fill_4" -> "718 masked_fill_5" [label="(4, 64, 64)", style=solid]; +"717 eq_2" -> "718 masked_fill_5" [label="(4, 64, 64)", style=solid]; +"718 masked_fill_5" -> "720 unsqueeze_16" [label="(4, 64, 64)", style=solid]; +"719 view_30" -> "722 add_18" [label="(1, 4, 12, 64, 64)", style=solid]; +"720 unsqueeze_16" -> "721 unsqueeze_17" [label="(4, 1, 64, 64)", style=solid]; +"721 unsqueeze_17" -> "722 add_18" [label="(1, 4, 1, 64, 64)", style=solid]; +"722 add_18" -> "723 view_31" [label="(1, 4, 12, 64, 64)", style=solid]; +"723 view_31" -> "724 softmax_5" [label="(4, 12, 64, 64)", style=solid]; +"724 softmax_5" -> "725 dropout_20" [label="(4, 12, 64, 64)", style=solid]; +"725 dropout_20" -> "726 matmul_11" [label="(4, 12, 64, 64)", style=solid]; +"726 matmul_11" -> "727 transpose_11" [label="(4, 12, 64, 32)", style=solid]; +"727 transpose_11" -> "728 reshape_25" [label="(4, 64, 12, 32)", style=solid]; +"728 reshape_25" -> "729 reshape_25_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"729 reshape_25_0_0_nncf_smooth_quant_0" -> "730 quantize_per_tensor_default_36" [label="(4, 64, 384)", style=solid]; +"730 quantize_per_tensor_default_36" -> "731 dequantize_per_tensor_default_36" [label="(4, 64, 384)", style=solid]; +"731 dequantize_per_tensor_default_36" -> "736 linear_35" [label="(4, 64, 384)", style=solid]; +"732 linear_35_scale_0" -> "735 dequantize_per_channel_default_24" [label="(384,)", style=solid]; +"733 linear_35_zero_point_0" -> "735 dequantize_per_channel_default_24" [label="(384,)", style=solid]; +"734 _frozen_param42_1" -> "735 dequantize_per_channel_default_24" [label="(384, 384)", style=solid]; +"735 dequantize_per_channel_default_24" -> "736 linear_35" [label="(384, 384)", style=solid]; +"736 linear_35" -> "737 dropout_21" [label="(4, 64, 384)", style=solid]; +"737 dropout_21" -> "738 view_32" [label="(4, 64, 384)", style=solid]; +"738 view_32" -> "739 permute_27" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"739 permute_27" -> "740 reshape_26" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"740 reshape_26" -> "741 roll_5" [label="(1, 16, 16, 384)", style=solid]; +"741 roll_5" -> "742 slice_101" [label="(1, 16, 16, 384)", style=solid]; +"742 slice_101" -> "743 slice_102" [label="(1, 16, 16, 384)", style=solid]; +"743 slice_102" -> "744 slice_103" [label="(1, 14, 16, 384)", style=solid]; +"744 slice_103" -> "745 slice_104" [label="(1, 14, 14, 384)", style=solid]; +"745 slice_104" -> "746 contiguous_9" [label="(1, 14, 14, 384)", style=solid]; +"746 contiguous_9" -> "747 layer_norm_13" [label="(1, 14, 14, 384)", style=solid]; +"747 layer_norm_13" -> "748 add_19" [label="(1, 14, 14, 384)", style=solid]; +"748 add_19" -> "749 add_19_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"748 add_19" -> "769 add_20" [label="(1, 14, 14, 384)", style=solid]; +"749 add_19_0_0_nncf_smooth_quant_0" -> "750 quantize_per_tensor_default_37" [label="(1, 14, 14, 384)", style=solid]; +"750 quantize_per_tensor_default_37" -> "751 dequantize_per_tensor_default_37" [label="(1, 14, 14, 384)", style=solid]; +"751 dequantize_per_tensor_default_37" -> "756 linear_36" [label="(1, 14, 14, 384)", style=solid]; +"752 linear_36_scale_0" -> "755 dequantize_per_channel_default_25" [label="(1536,)", style=solid]; +"753 linear_36_zero_point_0" -> "755 dequantize_per_channel_default_25" [label="(1536,)", style=solid]; +"754 _frozen_param43_1" -> "755 dequantize_per_channel_default_25" [label="(1536, 384)", style=solid]; +"755 dequantize_per_channel_default_25" -> "756 linear_36" [label="(1536, 384)", style=solid]; +"756 linear_36" -> "757 gelu_5" [label="(1, 14, 14, 1536)", style=solid]; +"757 gelu_5" -> "758 dropout_22" [label="(1, 14, 14, 1536)", style=solid]; +"758 dropout_22" -> "759 dropout_22_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"759 dropout_22_0_0_nncf_smooth_quant_0" -> "760 quantize_per_tensor_default_38" [label="(1, 14, 14, 1536)", style=solid]; +"760 quantize_per_tensor_default_38" -> "761 dequantize_per_tensor_default_38" [label="(1, 14, 14, 1536)", style=solid]; +"761 dequantize_per_tensor_default_38" -> "766 linear_37" [label="(1, 14, 14, 1536)", style=solid]; +"762 linear_37_scale_0" -> "765 dequantize_per_channel_default_26" [label="(384,)", style=solid]; +"763 linear_37_zero_point_0" -> "765 dequantize_per_channel_default_26" [label="(384,)", style=solid]; +"764 _frozen_param44_1" -> "765 dequantize_per_channel_default_26" [label="(384, 1536)", style=solid]; +"765 dequantize_per_channel_default_26" -> "766 linear_37" [label="(384, 1536)", style=solid]; +"766 linear_37" -> "767 dropout_23" [label="(1, 14, 14, 384)", style=solid]; +"767 dropout_23" -> "768 layer_norm_14" [label="(1, 14, 14, 384)", style=solid]; +"768 layer_norm_14" -> "769 add_20" [label="(1, 14, 14, 384)", style=solid]; +"769 add_20" -> "771 pad_8" [label="(1, 14, 14, 384)", style=solid]; +"769 add_20" -> "829 add_22" [label="(1, 14, 14, 384)", style=solid]; +"770 _frozen_param51" -> "805 add_21" [label="(1, 12, 64, 64)", style=solid]; +"771 pad_8" -> "772 view_35" [label="(1, 16, 16, 384)", style=solid]; +"772 view_35" -> "773 permute_29" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"773 permute_29" -> "774 reshape_27" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"774 reshape_27" -> "776 reshape_27_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"775 _frozen_param52" -> "783 linear_40" [label="(1152,)", style=solid]; +"776 reshape_27_0_0_nncf_smooth_quant_0" -> "777 quantize_per_tensor_default_39" [label="(4, 64, 384)", style=solid]; +"777 quantize_per_tensor_default_39" -> "778 dequantize_per_tensor_default_39" [label="(4, 64, 384)", style=solid]; +"778 dequantize_per_tensor_default_39" -> "783 linear_40" [label="(4, 64, 384)", style=solid]; +"779 linear_40_scale_0" -> "782 dequantize_per_channel_default_27" [label="(1152,)", style=solid]; +"780 linear_40_zero_point_0" -> "782 dequantize_per_channel_default_27" [label="(1152,)", style=solid]; +"781 _frozen_param45_1" -> "782 dequantize_per_channel_default_27" [label="(1152, 384)", style=solid]; +"782 dequantize_per_channel_default_27" -> "783 linear_40" [label="(1152, 384)", style=solid]; +"783 linear_40" -> "784 reshape_28" [label="(4, 64, 1152)", style=solid]; +"784 reshape_28" -> "785 permute_30" [label="(4, 64, 3, 12, 32)", style=solid]; +"785 permute_30" -> "786 select_18" [label="(3, 4, 12, 64, 32)", style=solid]; +"785 permute_30" -> "787 select_19" [label="(3, 4, 12, 64, 32)", style=solid]; +"785 permute_30" -> "788 select_20" [label="(3, 4, 12, 64, 32)", style=solid]; +"786 select_18" -> "789 linalg_vector_norm_12" [label="(4, 12, 64, 32)", style=solid]; +"786 select_18" -> "791 expand_as_12" [label="(4, 12, 64, 32)", style=solid]; +"786 select_18" -> "792 div_12" [label="(4, 12, 64, 32)", style=solid]; +"787 select_19" -> "795 linalg_vector_norm_13" [label="(4, 12, 64, 32)", style=solid]; +"787 select_19" -> "797 expand_as_13" [label="(4, 12, 64, 32)", style=solid]; +"787 select_19" -> "798 div_13" [label="(4, 12, 64, 32)", style=solid]; +"788 select_20" -> "808 matmul_13" [label="(4, 12, 64, 32)", style=solid]; +"789 linalg_vector_norm_12" -> "790 clamp_min_12" [label="(4, 12, 64, 1)", style=solid]; +"790 clamp_min_12" -> "791 expand_as_12" [label="(4, 12, 64, 1)", style=solid]; +"791 expand_as_12" -> "792 div_12" [label="(4, 12, 64, 32)", style=solid]; +"792 div_12" -> "793 quantize_per_tensor_default_40" [label="(4, 12, 64, 32)", style=solid]; +"793 quantize_per_tensor_default_40" -> "794 dequantize_per_tensor_default_40" [label="(4, 12, 64, 32)", style=solid]; +"794 dequantize_per_tensor_default_40" -> "802 matmul_12" [label="(4, 12, 64, 32)", style=solid]; +"795 linalg_vector_norm_13" -> "796 clamp_min_13" [label="(4, 12, 64, 1)", style=solid]; +"796 clamp_min_13" -> "797 expand_as_13" [label="(4, 12, 64, 1)", style=solid]; +"797 expand_as_13" -> "798 div_13" [label="(4, 12, 64, 32)", style=solid]; +"798 div_13" -> "799 quantize_per_tensor_default_41" [label="(4, 12, 64, 32)", style=solid]; +"799 quantize_per_tensor_default_41" -> "800 dequantize_per_tensor_default_41" [label="(4, 12, 64, 32)", style=solid]; +"800 dequantize_per_tensor_default_41" -> "801 transpose_12" [label="(4, 12, 64, 32)", style=solid]; +"801 transpose_12" -> "802 matmul_12" [label="(4, 12, 32, 64)", style=solid]; +"802 matmul_12" -> "804 mul_13" [label="(4, 12, 64, 64)", style=solid]; +"803 _frozen_param54" -> "804 mul_13" [label="(12, 1, 1)", style=solid]; +"804 mul_13" -> "805 add_21" [label="(4, 12, 64, 64)", style=solid]; +"805 add_21" -> "806 softmax_6" [label="(4, 12, 64, 64)", style=solid]; +"806 softmax_6" -> "807 dropout_24" [label="(4, 12, 64, 64)", style=solid]; +"807 dropout_24" -> "808 matmul_13" [label="(4, 12, 64, 64)", style=solid]; +"808 matmul_13" -> "809 transpose_13" [label="(4, 12, 64, 32)", style=solid]; +"809 transpose_13" -> "810 reshape_29" [label="(4, 64, 12, 32)", style=solid]; +"810 reshape_29" -> "811 reshape_29_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"811 reshape_29_0_0_nncf_smooth_quant_0" -> "812 quantize_per_tensor_default_42" [label="(4, 64, 384)", style=solid]; +"812 quantize_per_tensor_default_42" -> "813 dequantize_per_tensor_default_42" [label="(4, 64, 384)", style=solid]; +"813 dequantize_per_tensor_default_42" -> "818 linear_41" [label="(4, 64, 384)", style=solid]; +"814 linear_41_scale_0" -> "817 dequantize_per_channel_default_28" [label="(384,)", style=solid]; +"815 linear_41_zero_point_0" -> "817 dequantize_per_channel_default_28" [label="(384,)", style=solid]; +"816 _frozen_param46_1" -> "817 dequantize_per_channel_default_28" [label="(384, 384)", style=solid]; +"817 dequantize_per_channel_default_28" -> "818 linear_41" [label="(384, 384)", style=solid]; +"818 linear_41" -> "819 dropout_25" [label="(4, 64, 384)", style=solid]; +"819 dropout_25" -> "820 view_36" [label="(4, 64, 384)", style=solid]; +"820 view_36" -> "821 permute_31" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"821 permute_31" -> "822 reshape_30" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"822 reshape_30" -> "823 slice_106" [label="(1, 16, 16, 384)", style=solid]; +"823 slice_106" -> "824 slice_107" [label="(1, 16, 16, 384)", style=solid]; +"824 slice_107" -> "825 slice_108" [label="(1, 14, 16, 384)", style=solid]; +"825 slice_108" -> "826 slice_109" [label="(1, 14, 14, 384)", style=solid]; +"826 slice_109" -> "827 contiguous_11" [label="(1, 14, 14, 384)", style=solid]; +"827 contiguous_11" -> "828 layer_norm_15" [label="(1, 14, 14, 384)", style=solid]; +"828 layer_norm_15" -> "829 add_22" [label="(1, 14, 14, 384)", style=solid]; +"829 add_22" -> "830 add_22_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"829 add_22" -> "850 add_23" [label="(1, 14, 14, 384)", style=solid]; +"830 add_22_0_0_nncf_smooth_quant_0" -> "831 quantize_per_tensor_default_43" [label="(1, 14, 14, 384)", style=solid]; +"831 quantize_per_tensor_default_43" -> "832 dequantize_per_tensor_default_43" [label="(1, 14, 14, 384)", style=solid]; +"832 dequantize_per_tensor_default_43" -> "837 linear_42" [label="(1, 14, 14, 384)", style=solid]; +"833 linear_42_scale_0" -> "836 dequantize_per_channel_default_29" [label="(1536,)", style=solid]; +"834 linear_42_zero_point_0" -> "836 dequantize_per_channel_default_29" [label="(1536,)", style=solid]; +"835 _frozen_param47_1" -> "836 dequantize_per_channel_default_29" [label="(1536, 384)", style=solid]; +"836 dequantize_per_channel_default_29" -> "837 linear_42" [label="(1536, 384)", style=solid]; +"837 linear_42" -> "838 gelu_6" [label="(1, 14, 14, 1536)", style=solid]; +"838 gelu_6" -> "839 dropout_26" [label="(1, 14, 14, 1536)", style=solid]; +"839 dropout_26" -> "840 dropout_26_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"840 dropout_26_0_0_nncf_smooth_quant_0" -> "841 quantize_per_tensor_default_44" [label="(1, 14, 14, 1536)", style=solid]; +"841 quantize_per_tensor_default_44" -> "842 dequantize_per_tensor_default_44" [label="(1, 14, 14, 1536)", style=solid]; +"842 dequantize_per_tensor_default_44" -> "847 linear_43" [label="(1, 14, 14, 1536)", style=solid]; +"843 linear_43_scale_0" -> "846 dequantize_per_channel_default_30" [label="(384,)", style=solid]; +"844 linear_43_zero_point_0" -> "846 dequantize_per_channel_default_30" [label="(384,)", style=solid]; +"845 _frozen_param48_1" -> "846 dequantize_per_channel_default_30" [label="(384, 1536)", style=solid]; +"846 dequantize_per_channel_default_30" -> "847 linear_43" [label="(384, 1536)", style=solid]; +"847 linear_43" -> "848 dropout_27" [label="(1, 14, 14, 384)", style=solid]; +"848 dropout_27" -> "849 layer_norm_16" [label="(1, 14, 14, 384)", style=solid]; +"849 layer_norm_16" -> "850 add_23" [label="(1, 14, 14, 384)", style=solid]; +"850 add_23" -> "852 pad_9" [label="(1, 14, 14, 384)", style=solid]; +"850 add_23" -> "928 add_26" [label="(1, 14, 14, 384)", style=solid]; +"851 _frozen_param55" -> "887 add_24" [label="(1, 12, 64, 64)", style=solid]; +"852 pad_9" -> "853 roll_6" [label="(1, 16, 16, 384)", style=solid]; +"853 roll_6" -> "854 view_39" [label="(1, 16, 16, 384)", style=solid]; +"854 view_39" -> "855 permute_33" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"855 permute_33" -> "856 reshape_31" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"856 reshape_31" -> "858 reshape_31_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"856 reshape_31" -> "888 new_zeros_3" [label="(4, 64, 384)", style=solid]; +"857 _frozen_param56" -> "865 linear_46" [label="(1152,)", style=solid]; +"858 reshape_31_0_0_nncf_smooth_quant_0" -> "859 quantize_per_tensor_default_45" [label="(4, 64, 384)", style=solid]; +"859 quantize_per_tensor_default_45" -> "860 dequantize_per_tensor_default_45" [label="(4, 64, 384)", style=solid]; +"860 dequantize_per_tensor_default_45" -> "865 linear_46" [label="(4, 64, 384)", style=solid]; +"861 linear_46_scale_0" -> "864 dequantize_per_channel_default_31" [label="(1152,)", style=solid]; +"862 linear_46_zero_point_0" -> "864 dequantize_per_channel_default_31" [label="(1152,)", style=solid]; +"863 _frozen_param49_1" -> "864 dequantize_per_channel_default_31" [label="(1152, 384)", style=solid]; +"864 dequantize_per_channel_default_31" -> "865 linear_46" [label="(1152, 384)", style=solid]; +"865 linear_46" -> "866 reshape_32" [label="(4, 64, 1152)", style=solid]; +"866 reshape_32" -> "867 permute_34" [label="(4, 64, 3, 12, 32)", style=solid]; +"867 permute_34" -> "868 select_21" [label="(3, 4, 12, 64, 32)", style=solid]; +"867 permute_34" -> "869 select_22" [label="(3, 4, 12, 64, 32)", style=solid]; +"867 permute_34" -> "870 select_23" [label="(3, 4, 12, 64, 32)", style=solid]; +"868 select_21" -> "871 linalg_vector_norm_14" [label="(4, 12, 64, 32)", style=solid]; +"868 select_21" -> "873 expand_as_14" [label="(4, 12, 64, 32)", style=solid]; +"868 select_21" -> "874 div_14" [label="(4, 12, 64, 32)", style=solid]; +"869 select_22" -> "877 linalg_vector_norm_15" [label="(4, 12, 64, 32)", style=solid]; +"869 select_22" -> "879 expand_as_15" [label="(4, 12, 64, 32)", style=solid]; +"869 select_22" -> "880 div_15" [label="(4, 12, 64, 32)", style=solid]; +"870 select_23" -> "906 matmul_15" [label="(4, 12, 64, 32)", style=solid]; +"871 linalg_vector_norm_14" -> "872 clamp_min_14" [label="(4, 12, 64, 1)", style=solid]; +"872 clamp_min_14" -> "873 expand_as_14" [label="(4, 12, 64, 1)", style=solid]; +"873 expand_as_14" -> "874 div_14" [label="(4, 12, 64, 32)", style=solid]; +"874 div_14" -> "875 quantize_per_tensor_default_46" [label="(4, 12, 64, 32)", style=solid]; +"875 quantize_per_tensor_default_46" -> "876 dequantize_per_tensor_default_46" [label="(4, 12, 64, 32)", style=solid]; +"876 dequantize_per_tensor_default_46" -> "884 matmul_14" [label="(4, 12, 64, 32)", style=solid]; +"877 linalg_vector_norm_15" -> "878 clamp_min_15" [label="(4, 12, 64, 1)", style=solid]; +"878 clamp_min_15" -> "879 expand_as_15" [label="(4, 12, 64, 1)", style=solid]; +"879 expand_as_15" -> "880 div_15" [label="(4, 12, 64, 32)", style=solid]; +"880 div_15" -> "881 quantize_per_tensor_default_47" [label="(4, 12, 64, 32)", style=solid]; +"881 quantize_per_tensor_default_47" -> "882 dequantize_per_tensor_default_47" [label="(4, 12, 64, 32)", style=solid]; +"882 dequantize_per_tensor_default_47" -> "883 transpose_14" [label="(4, 12, 64, 32)", style=solid]; +"883 transpose_14" -> "884 matmul_14" [label="(4, 12, 32, 64)", style=solid]; +"884 matmul_14" -> "886 mul_15" [label="(4, 12, 64, 64)", style=solid]; +"885 _frozen_param58" -> "886 mul_15" [label="(12, 1, 1)", style=solid]; +"886 mul_15" -> "887 add_24" [label="(4, 12, 64, 64)", style=solid]; +"887 add_24" -> "899 view_41" [label="(4, 12, 64, 64)", style=solid]; +"888 new_zeros_3" -> "889 view_40" [label="(16, 16)", style=solid]; +"889 view_40" -> "890 permute_35" [label="(2, 8, 2, 8)", style=solid]; +"890 permute_35" -> "891 reshape_33" [label="(2, 2, 8, 8)", style=solid]; +"891 reshape_33" -> "892 unsqueeze_20" [label="(4, 64)", style=solid]; +"891 reshape_33" -> "893 unsqueeze_21" [label="(4, 64)", style=solid]; +"892 unsqueeze_20" -> "894 sub_3" [label="(4, 1, 64)", style=solid]; +"893 unsqueeze_21" -> "894 sub_3" [label="(4, 64, 1)", style=solid]; +"894 sub_3" -> "895 ne_3" [label="(4, 64, 64)", style=solid]; +"894 sub_3" -> "896 masked_fill_6" [label="(4, 64, 64)", style=solid]; +"894 sub_3" -> "897 eq_3" [label="(4, 64, 64)", style=solid]; +"895 ne_3" -> "896 masked_fill_6" [label="(4, 64, 64)", style=solid]; +"896 masked_fill_6" -> "898 masked_fill_7" [label="(4, 64, 64)", style=solid]; +"897 eq_3" -> "898 masked_fill_7" [label="(4, 64, 64)", style=solid]; +"898 masked_fill_7" -> "900 unsqueeze_22" [label="(4, 64, 64)", style=solid]; +"899 view_41" -> "902 add_25" [label="(1, 4, 12, 64, 64)", style=solid]; +"900 unsqueeze_22" -> "901 unsqueeze_23" [label="(4, 1, 64, 64)", style=solid]; +"901 unsqueeze_23" -> "902 add_25" [label="(1, 4, 1, 64, 64)", style=solid]; +"902 add_25" -> "903 view_42" [label="(1, 4, 12, 64, 64)", style=solid]; +"903 view_42" -> "904 softmax_7" [label="(4, 12, 64, 64)", style=solid]; +"904 softmax_7" -> "905 dropout_28" [label="(4, 12, 64, 64)", style=solid]; +"905 dropout_28" -> "906 matmul_15" [label="(4, 12, 64, 64)", style=solid]; +"906 matmul_15" -> "907 transpose_15" [label="(4, 12, 64, 32)", style=solid]; +"907 transpose_15" -> "908 reshape_34" [label="(4, 64, 12, 32)", style=solid]; +"908 reshape_34" -> "909 reshape_34_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"909 reshape_34_0_0_nncf_smooth_quant_0" -> "910 quantize_per_tensor_default_48" [label="(4, 64, 384)", style=solid]; +"910 quantize_per_tensor_default_48" -> "911 dequantize_per_tensor_default_48" [label="(4, 64, 384)", style=solid]; +"911 dequantize_per_tensor_default_48" -> "916 linear_47" [label="(4, 64, 384)", style=solid]; +"912 linear_47_scale_0" -> "915 dequantize_per_channel_default_32" [label="(384,)", style=solid]; +"913 linear_47_zero_point_0" -> "915 dequantize_per_channel_default_32" [label="(384,)", style=solid]; +"914 _frozen_param50_1" -> "915 dequantize_per_channel_default_32" [label="(384, 384)", style=solid]; +"915 dequantize_per_channel_default_32" -> "916 linear_47" [label="(384, 384)", style=solid]; +"916 linear_47" -> "917 dropout_29" [label="(4, 64, 384)", style=solid]; +"917 dropout_29" -> "918 view_43" [label="(4, 64, 384)", style=solid]; +"918 view_43" -> "919 permute_36" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"919 permute_36" -> "920 reshape_35" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"920 reshape_35" -> "921 roll_7" [label="(1, 16, 16, 384)", style=solid]; +"921 roll_7" -> "922 slice_129" [label="(1, 16, 16, 384)", style=solid]; +"922 slice_129" -> "923 slice_130" [label="(1, 16, 16, 384)", style=solid]; +"923 slice_130" -> "924 slice_131" [label="(1, 14, 16, 384)", style=solid]; +"924 slice_131" -> "925 slice_132" [label="(1, 14, 14, 384)", style=solid]; +"925 slice_132" -> "926 contiguous_13" [label="(1, 14, 14, 384)", style=solid]; +"926 contiguous_13" -> "927 layer_norm_17" [label="(1, 14, 14, 384)", style=solid]; +"927 layer_norm_17" -> "928 add_26" [label="(1, 14, 14, 384)", style=solid]; +"928 add_26" -> "929 add_26_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"928 add_26" -> "949 add_27" [label="(1, 14, 14, 384)", style=solid]; +"929 add_26_0_0_nncf_smooth_quant_0" -> "930 quantize_per_tensor_default_49" [label="(1, 14, 14, 384)", style=solid]; +"930 quantize_per_tensor_default_49" -> "931 dequantize_per_tensor_default_49" [label="(1, 14, 14, 384)", style=solid]; +"931 dequantize_per_tensor_default_49" -> "936 linear_48" [label="(1, 14, 14, 384)", style=solid]; +"932 linear_48_scale_0" -> "935 dequantize_per_channel_default_33" [label="(1536,)", style=solid]; +"933 linear_48_zero_point_0" -> "935 dequantize_per_channel_default_33" [label="(1536,)", style=solid]; +"934 _frozen_param53_1" -> "935 dequantize_per_channel_default_33" [label="(1536, 384)", style=solid]; +"935 dequantize_per_channel_default_33" -> "936 linear_48" [label="(1536, 384)", style=solid]; +"936 linear_48" -> "937 gelu_7" [label="(1, 14, 14, 1536)", style=solid]; +"937 gelu_7" -> "938 dropout_30" [label="(1, 14, 14, 1536)", style=solid]; +"938 dropout_30" -> "939 dropout_30_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"939 dropout_30_0_0_nncf_smooth_quant_0" -> "940 quantize_per_tensor_default_50" [label="(1, 14, 14, 1536)", style=solid]; +"940 quantize_per_tensor_default_50" -> "941 dequantize_per_tensor_default_50" [label="(1, 14, 14, 1536)", style=solid]; +"941 dequantize_per_tensor_default_50" -> "946 linear_49" [label="(1, 14, 14, 1536)", style=solid]; +"942 linear_49_scale_0" -> "945 dequantize_per_channel_default_34" [label="(384,)", style=solid]; +"943 linear_49_zero_point_0" -> "945 dequantize_per_channel_default_34" [label="(384,)", style=solid]; +"944 _frozen_param57_1" -> "945 dequantize_per_channel_default_34" [label="(384, 1536)", style=solid]; +"945 dequantize_per_channel_default_34" -> "946 linear_49" [label="(384, 1536)", style=solid]; +"946 linear_49" -> "947 dropout_31" [label="(1, 14, 14, 384)", style=solid]; +"947 dropout_31" -> "948 layer_norm_18" [label="(1, 14, 14, 384)", style=solid]; +"948 layer_norm_18" -> "949 add_27" [label="(1, 14, 14, 384)", style=solid]; +"949 add_27" -> "951 pad_10" [label="(1, 14, 14, 384)", style=solid]; +"949 add_27" -> "1009 add_29" [label="(1, 14, 14, 384)", style=solid]; +"950 _frozen_param68" -> "985 add_28" [label="(1, 12, 64, 64)", style=solid]; +"951 pad_10" -> "952 view_46" [label="(1, 16, 16, 384)", style=solid]; +"952 view_46" -> "953 permute_38" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"953 permute_38" -> "954 reshape_36" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"954 reshape_36" -> "956 reshape_36_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"955 _frozen_param69" -> "963 linear_52" [label="(1152,)", style=solid]; +"956 reshape_36_0_0_nncf_smooth_quant_0" -> "957 quantize_per_tensor_default_51" [label="(4, 64, 384)", style=solid]; +"957 quantize_per_tensor_default_51" -> "958 dequantize_per_tensor_default_51" [label="(4, 64, 384)", style=solid]; +"958 dequantize_per_tensor_default_51" -> "963 linear_52" [label="(4, 64, 384)", style=solid]; +"959 linear_52_scale_0" -> "962 dequantize_per_channel_default_35" [label="(1152,)", style=solid]; +"960 linear_52_zero_point_0" -> "962 dequantize_per_channel_default_35" [label="(1152,)", style=solid]; +"961 _frozen_param59_1" -> "962 dequantize_per_channel_default_35" [label="(1152, 384)", style=solid]; +"962 dequantize_per_channel_default_35" -> "963 linear_52" [label="(1152, 384)", style=solid]; +"963 linear_52" -> "964 reshape_37" [label="(4, 64, 1152)", style=solid]; +"964 reshape_37" -> "965 permute_39" [label="(4, 64, 3, 12, 32)", style=solid]; +"965 permute_39" -> "966 select_24" [label="(3, 4, 12, 64, 32)", style=solid]; +"965 permute_39" -> "967 select_25" [label="(3, 4, 12, 64, 32)", style=solid]; +"965 permute_39" -> "968 select_26" [label="(3, 4, 12, 64, 32)", style=solid]; +"966 select_24" -> "969 linalg_vector_norm_16" [label="(4, 12, 64, 32)", style=solid]; +"966 select_24" -> "971 expand_as_16" [label="(4, 12, 64, 32)", style=solid]; +"966 select_24" -> "972 div_16" [label="(4, 12, 64, 32)", style=solid]; +"967 select_25" -> "975 linalg_vector_norm_17" [label="(4, 12, 64, 32)", style=solid]; +"967 select_25" -> "977 expand_as_17" [label="(4, 12, 64, 32)", style=solid]; +"967 select_25" -> "978 div_17" [label="(4, 12, 64, 32)", style=solid]; +"968 select_26" -> "988 matmul_17" [label="(4, 12, 64, 32)", style=solid]; +"969 linalg_vector_norm_16" -> "970 clamp_min_16" [label="(4, 12, 64, 1)", style=solid]; +"970 clamp_min_16" -> "971 expand_as_16" [label="(4, 12, 64, 1)", style=solid]; +"971 expand_as_16" -> "972 div_16" [label="(4, 12, 64, 32)", style=solid]; +"972 div_16" -> "973 quantize_per_tensor_default_52" [label="(4, 12, 64, 32)", style=solid]; +"973 quantize_per_tensor_default_52" -> "974 dequantize_per_tensor_default_52" [label="(4, 12, 64, 32)", style=solid]; +"974 dequantize_per_tensor_default_52" -> "982 matmul_16" [label="(4, 12, 64, 32)", style=solid]; +"975 linalg_vector_norm_17" -> "976 clamp_min_17" [label="(4, 12, 64, 1)", style=solid]; +"976 clamp_min_17" -> "977 expand_as_17" [label="(4, 12, 64, 1)", style=solid]; +"977 expand_as_17" -> "978 div_17" [label="(4, 12, 64, 32)", style=solid]; +"978 div_17" -> "979 quantize_per_tensor_default_53" [label="(4, 12, 64, 32)", style=solid]; +"979 quantize_per_tensor_default_53" -> "980 dequantize_per_tensor_default_53" [label="(4, 12, 64, 32)", style=solid]; +"980 dequantize_per_tensor_default_53" -> "981 transpose_16" [label="(4, 12, 64, 32)", style=solid]; +"981 transpose_16" -> "982 matmul_16" [label="(4, 12, 32, 64)", style=solid]; +"982 matmul_16" -> "984 mul_17" [label="(4, 12, 64, 64)", style=solid]; +"983 _frozen_param71" -> "984 mul_17" [label="(12, 1, 1)", style=solid]; +"984 mul_17" -> "985 add_28" [label="(4, 12, 64, 64)", style=solid]; +"985 add_28" -> "986 softmax_8" [label="(4, 12, 64, 64)", style=solid]; +"986 softmax_8" -> "987 dropout_32" [label="(4, 12, 64, 64)", style=solid]; +"987 dropout_32" -> "988 matmul_17" [label="(4, 12, 64, 64)", style=solid]; +"988 matmul_17" -> "989 transpose_17" [label="(4, 12, 64, 32)", style=solid]; +"989 transpose_17" -> "990 reshape_38" [label="(4, 64, 12, 32)", style=solid]; +"990 reshape_38" -> "991 reshape_38_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"991 reshape_38_0_0_nncf_smooth_quant_0" -> "992 quantize_per_tensor_default_54" [label="(4, 64, 384)", style=solid]; +"992 quantize_per_tensor_default_54" -> "993 dequantize_per_tensor_default_54" [label="(4, 64, 384)", style=solid]; +"993 dequantize_per_tensor_default_54" -> "998 linear_53" [label="(4, 64, 384)", style=solid]; +"994 linear_53_scale_0" -> "997 dequantize_per_channel_default_36" [label="(384,)", style=solid]; +"995 linear_53_zero_point_0" -> "997 dequantize_per_channel_default_36" [label="(384,)", style=solid]; +"996 _frozen_param60_1" -> "997 dequantize_per_channel_default_36" [label="(384, 384)", style=solid]; +"997 dequantize_per_channel_default_36" -> "998 linear_53" [label="(384, 384)", style=solid]; +"998 linear_53" -> "999 dropout_33" [label="(4, 64, 384)", style=solid]; +"999 dropout_33" -> "1000 view_47" [label="(4, 64, 384)", style=solid]; +"1000 view_47" -> "1001 permute_40" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1001 permute_40" -> "1002 reshape_39" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1002 reshape_39" -> "1003 slice_134" [label="(1, 16, 16, 384)", style=solid]; +"1003 slice_134" -> "1004 slice_135" [label="(1, 16, 16, 384)", style=solid]; +"1004 slice_135" -> "1005 slice_136" [label="(1, 14, 16, 384)", style=solid]; +"1005 slice_136" -> "1006 slice_137" [label="(1, 14, 14, 384)", style=solid]; +"1006 slice_137" -> "1007 contiguous_15" [label="(1, 14, 14, 384)", style=solid]; +"1007 contiguous_15" -> "1008 layer_norm_19" [label="(1, 14, 14, 384)", style=solid]; +"1008 layer_norm_19" -> "1009 add_29" [label="(1, 14, 14, 384)", style=solid]; +"1009 add_29" -> "1010 add_29_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"1009 add_29" -> "1030 add_30" [label="(1, 14, 14, 384)", style=solid]; +"1010 add_29_0_0_nncf_smooth_quant_0" -> "1011 quantize_per_tensor_default_55" [label="(1, 14, 14, 384)", style=solid]; +"1011 quantize_per_tensor_default_55" -> "1012 dequantize_per_tensor_default_55" [label="(1, 14, 14, 384)", style=solid]; +"1012 dequantize_per_tensor_default_55" -> "1017 linear_54" [label="(1, 14, 14, 384)", style=solid]; +"1013 linear_54_scale_0" -> "1016 dequantize_per_channel_default_37" [label="(1536,)", style=solid]; +"1014 linear_54_zero_point_0" -> "1016 dequantize_per_channel_default_37" [label="(1536,)", style=solid]; +"1015 _frozen_param61_1" -> "1016 dequantize_per_channel_default_37" [label="(1536, 384)", style=solid]; +"1016 dequantize_per_channel_default_37" -> "1017 linear_54" [label="(1536, 384)", style=solid]; +"1017 linear_54" -> "1018 gelu_8" [label="(1, 14, 14, 1536)", style=solid]; +"1018 gelu_8" -> "1019 dropout_34" [label="(1, 14, 14, 1536)", style=solid]; +"1019 dropout_34" -> "1020 dropout_34_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"1020 dropout_34_0_0_nncf_smooth_quant_0" -> "1021 quantize_per_tensor_default_56" [label="(1, 14, 14, 1536)", style=solid]; +"1021 quantize_per_tensor_default_56" -> "1022 dequantize_per_tensor_default_56" [label="(1, 14, 14, 1536)", style=solid]; +"1022 dequantize_per_tensor_default_56" -> "1027 linear_55" [label="(1, 14, 14, 1536)", style=solid]; +"1023 linear_55_scale_0" -> "1026 dequantize_per_channel_default_38" [label="(384,)", style=solid]; +"1024 linear_55_zero_point_0" -> "1026 dequantize_per_channel_default_38" [label="(384,)", style=solid]; +"1025 _frozen_param62_1" -> "1026 dequantize_per_channel_default_38" [label="(384, 1536)", style=solid]; +"1026 dequantize_per_channel_default_38" -> "1027 linear_55" [label="(384, 1536)", style=solid]; +"1027 linear_55" -> "1028 dropout_35" [label="(1, 14, 14, 384)", style=solid]; +"1028 dropout_35" -> "1029 layer_norm_20" [label="(1, 14, 14, 384)", style=solid]; +"1029 layer_norm_20" -> "1030 add_30" [label="(1, 14, 14, 384)", style=solid]; +"1030 add_30" -> "1032 pad_11" [label="(1, 14, 14, 384)", style=solid]; +"1030 add_30" -> "1108 add_33" [label="(1, 14, 14, 384)", style=solid]; +"1031 _frozen_param72" -> "1067 add_31" [label="(1, 12, 64, 64)", style=solid]; +"1032 pad_11" -> "1033 roll_8" [label="(1, 16, 16, 384)", style=solid]; +"1033 roll_8" -> "1034 view_50" [label="(1, 16, 16, 384)", style=solid]; +"1034 view_50" -> "1035 permute_42" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1035 permute_42" -> "1036 reshape_40" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1036 reshape_40" -> "1038 reshape_40_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1036 reshape_40" -> "1068 new_zeros_4" [label="(4, 64, 384)", style=solid]; +"1037 _frozen_param73" -> "1045 linear_58" [label="(1152,)", style=solid]; +"1038 reshape_40_0_0_nncf_smooth_quant_0" -> "1039 quantize_per_tensor_default_57" [label="(4, 64, 384)", style=solid]; +"1039 quantize_per_tensor_default_57" -> "1040 dequantize_per_tensor_default_57" [label="(4, 64, 384)", style=solid]; +"1040 dequantize_per_tensor_default_57" -> "1045 linear_58" [label="(4, 64, 384)", style=solid]; +"1041 linear_58_scale_0" -> "1044 dequantize_per_channel_default_39" [label="(1152,)", style=solid]; +"1042 linear_58_zero_point_0" -> "1044 dequantize_per_channel_default_39" [label="(1152,)", style=solid]; +"1043 _frozen_param63_1" -> "1044 dequantize_per_channel_default_39" [label="(1152, 384)", style=solid]; +"1044 dequantize_per_channel_default_39" -> "1045 linear_58" [label="(1152, 384)", style=solid]; +"1045 linear_58" -> "1046 reshape_41" [label="(4, 64, 1152)", style=solid]; +"1046 reshape_41" -> "1047 permute_43" [label="(4, 64, 3, 12, 32)", style=solid]; +"1047 permute_43" -> "1048 select_27" [label="(3, 4, 12, 64, 32)", style=solid]; +"1047 permute_43" -> "1049 select_28" [label="(3, 4, 12, 64, 32)", style=solid]; +"1047 permute_43" -> "1050 select_29" [label="(3, 4, 12, 64, 32)", style=solid]; +"1048 select_27" -> "1051 linalg_vector_norm_18" [label="(4, 12, 64, 32)", style=solid]; +"1048 select_27" -> "1053 expand_as_18" [label="(4, 12, 64, 32)", style=solid]; +"1048 select_27" -> "1054 div_18" [label="(4, 12, 64, 32)", style=solid]; +"1049 select_28" -> "1057 linalg_vector_norm_19" [label="(4, 12, 64, 32)", style=solid]; +"1049 select_28" -> "1059 expand_as_19" [label="(4, 12, 64, 32)", style=solid]; +"1049 select_28" -> "1060 div_19" [label="(4, 12, 64, 32)", style=solid]; +"1050 select_29" -> "1086 matmul_19" [label="(4, 12, 64, 32)", style=solid]; +"1051 linalg_vector_norm_18" -> "1052 clamp_min_18" [label="(4, 12, 64, 1)", style=solid]; +"1052 clamp_min_18" -> "1053 expand_as_18" [label="(4, 12, 64, 1)", style=solid]; +"1053 expand_as_18" -> "1054 div_18" [label="(4, 12, 64, 32)", style=solid]; +"1054 div_18" -> "1055 quantize_per_tensor_default_58" [label="(4, 12, 64, 32)", style=solid]; +"1055 quantize_per_tensor_default_58" -> "1056 dequantize_per_tensor_default_58" [label="(4, 12, 64, 32)", style=solid]; +"1056 dequantize_per_tensor_default_58" -> "1064 matmul_18" [label="(4, 12, 64, 32)", style=solid]; +"1057 linalg_vector_norm_19" -> "1058 clamp_min_19" [label="(4, 12, 64, 1)", style=solid]; +"1058 clamp_min_19" -> "1059 expand_as_19" [label="(4, 12, 64, 1)", style=solid]; +"1059 expand_as_19" -> "1060 div_19" [label="(4, 12, 64, 32)", style=solid]; +"1060 div_19" -> "1061 quantize_per_tensor_default_59" [label="(4, 12, 64, 32)", style=solid]; +"1061 quantize_per_tensor_default_59" -> "1062 dequantize_per_tensor_default_59" [label="(4, 12, 64, 32)", style=solid]; +"1062 dequantize_per_tensor_default_59" -> "1063 transpose_18" [label="(4, 12, 64, 32)", style=solid]; +"1063 transpose_18" -> "1064 matmul_18" [label="(4, 12, 32, 64)", style=solid]; +"1064 matmul_18" -> "1066 mul_19" [label="(4, 12, 64, 64)", style=solid]; +"1065 _frozen_param75" -> "1066 mul_19" [label="(12, 1, 1)", style=solid]; +"1066 mul_19" -> "1067 add_31" [label="(4, 12, 64, 64)", style=solid]; +"1067 add_31" -> "1079 view_52" [label="(4, 12, 64, 64)", style=solid]; +"1068 new_zeros_4" -> "1069 view_51" [label="(16, 16)", style=solid]; +"1069 view_51" -> "1070 permute_44" [label="(2, 8, 2, 8)", style=solid]; +"1070 permute_44" -> "1071 reshape_42" [label="(2, 2, 8, 8)", style=solid]; +"1071 reshape_42" -> "1072 unsqueeze_26" [label="(4, 64)", style=solid]; +"1071 reshape_42" -> "1073 unsqueeze_27" [label="(4, 64)", style=solid]; +"1072 unsqueeze_26" -> "1074 sub_4" [label="(4, 1, 64)", style=solid]; +"1073 unsqueeze_27" -> "1074 sub_4" [label="(4, 64, 1)", style=solid]; +"1074 sub_4" -> "1075 ne_4" [label="(4, 64, 64)", style=solid]; +"1074 sub_4" -> "1076 masked_fill_8" [label="(4, 64, 64)", style=solid]; +"1074 sub_4" -> "1077 eq_4" [label="(4, 64, 64)", style=solid]; +"1075 ne_4" -> "1076 masked_fill_8" [label="(4, 64, 64)", style=solid]; +"1076 masked_fill_8" -> "1078 masked_fill_9" [label="(4, 64, 64)", style=solid]; +"1077 eq_4" -> "1078 masked_fill_9" [label="(4, 64, 64)", style=solid]; +"1078 masked_fill_9" -> "1080 unsqueeze_28" [label="(4, 64, 64)", style=solid]; +"1079 view_52" -> "1082 add_32" [label="(1, 4, 12, 64, 64)", style=solid]; +"1080 unsqueeze_28" -> "1081 unsqueeze_29" [label="(4, 1, 64, 64)", style=solid]; +"1081 unsqueeze_29" -> "1082 add_32" [label="(1, 4, 1, 64, 64)", style=solid]; +"1082 add_32" -> "1083 view_53" [label="(1, 4, 12, 64, 64)", style=solid]; +"1083 view_53" -> "1084 softmax_9" [label="(4, 12, 64, 64)", style=solid]; +"1084 softmax_9" -> "1085 dropout_36" [label="(4, 12, 64, 64)", style=solid]; +"1085 dropout_36" -> "1086 matmul_19" [label="(4, 12, 64, 64)", style=solid]; +"1086 matmul_19" -> "1087 transpose_19" [label="(4, 12, 64, 32)", style=solid]; +"1087 transpose_19" -> "1088 reshape_43" [label="(4, 64, 12, 32)", style=solid]; +"1088 reshape_43" -> "1089 reshape_43_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1089 reshape_43_0_0_nncf_smooth_quant_0" -> "1090 quantize_per_tensor_default_60" [label="(4, 64, 384)", style=solid]; +"1090 quantize_per_tensor_default_60" -> "1091 dequantize_per_tensor_default_60" [label="(4, 64, 384)", style=solid]; +"1091 dequantize_per_tensor_default_60" -> "1096 linear_59" [label="(4, 64, 384)", style=solid]; +"1092 linear_59_scale_0" -> "1095 dequantize_per_channel_default_40" [label="(384,)", style=solid]; +"1093 linear_59_zero_point_0" -> "1095 dequantize_per_channel_default_40" [label="(384,)", style=solid]; +"1094 _frozen_param64_1" -> "1095 dequantize_per_channel_default_40" [label="(384, 384)", style=solid]; +"1095 dequantize_per_channel_default_40" -> "1096 linear_59" [label="(384, 384)", style=solid]; +"1096 linear_59" -> "1097 dropout_37" [label="(4, 64, 384)", style=solid]; +"1097 dropout_37" -> "1098 view_54" [label="(4, 64, 384)", style=solid]; +"1098 view_54" -> "1099 permute_45" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1099 permute_45" -> "1100 reshape_44" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1100 reshape_44" -> "1101 roll_9" [label="(1, 16, 16, 384)", style=solid]; +"1101 roll_9" -> "1102 slice_157" [label="(1, 16, 16, 384)", style=solid]; +"1102 slice_157" -> "1103 slice_158" [label="(1, 16, 16, 384)", style=solid]; +"1103 slice_158" -> "1104 slice_159" [label="(1, 14, 16, 384)", style=solid]; +"1104 slice_159" -> "1105 slice_160" [label="(1, 14, 14, 384)", style=solid]; +"1105 slice_160" -> "1106 contiguous_17" [label="(1, 14, 14, 384)", style=solid]; +"1106 contiguous_17" -> "1107 layer_norm_21" [label="(1, 14, 14, 384)", style=solid]; +"1107 layer_norm_21" -> "1108 add_33" [label="(1, 14, 14, 384)", style=solid]; +"1108 add_33" -> "1109 add_33_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"1108 add_33" -> "1129 add_34" [label="(1, 14, 14, 384)", style=solid]; +"1109 add_33_0_0_nncf_smooth_quant_0" -> "1110 quantize_per_tensor_default_61" [label="(1, 14, 14, 384)", style=solid]; +"1110 quantize_per_tensor_default_61" -> "1111 dequantize_per_tensor_default_61" [label="(1, 14, 14, 384)", style=solid]; +"1111 dequantize_per_tensor_default_61" -> "1116 linear_60" [label="(1, 14, 14, 384)", style=solid]; +"1112 linear_60_scale_0" -> "1115 dequantize_per_channel_default_41" [label="(1536,)", style=solid]; +"1113 linear_60_zero_point_0" -> "1115 dequantize_per_channel_default_41" [label="(1536,)", style=solid]; +"1114 _frozen_param65_1" -> "1115 dequantize_per_channel_default_41" [label="(1536, 384)", style=solid]; +"1115 dequantize_per_channel_default_41" -> "1116 linear_60" [label="(1536, 384)", style=solid]; +"1116 linear_60" -> "1117 gelu_9" [label="(1, 14, 14, 1536)", style=solid]; +"1117 gelu_9" -> "1118 dropout_38" [label="(1, 14, 14, 1536)", style=solid]; +"1118 dropout_38" -> "1119 dropout_38_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"1119 dropout_38_0_0_nncf_smooth_quant_0" -> "1120 quantize_per_tensor_default_62" [label="(1, 14, 14, 1536)", style=solid]; +"1120 quantize_per_tensor_default_62" -> "1121 dequantize_per_tensor_default_62" [label="(1, 14, 14, 1536)", style=solid]; +"1121 dequantize_per_tensor_default_62" -> "1126 linear_61" [label="(1, 14, 14, 1536)", style=solid]; +"1122 linear_61_scale_0" -> "1125 dequantize_per_channel_default_42" [label="(384,)", style=solid]; +"1123 linear_61_zero_point_0" -> "1125 dequantize_per_channel_default_42" [label="(384,)", style=solid]; +"1124 _frozen_param66_1" -> "1125 dequantize_per_channel_default_42" [label="(384, 1536)", style=solid]; +"1125 dequantize_per_channel_default_42" -> "1126 linear_61" [label="(384, 1536)", style=solid]; +"1126 linear_61" -> "1127 dropout_39" [label="(1, 14, 14, 384)", style=solid]; +"1127 dropout_39" -> "1128 layer_norm_22" [label="(1, 14, 14, 384)", style=solid]; +"1128 layer_norm_22" -> "1129 add_34" [label="(1, 14, 14, 384)", style=solid]; +"1129 add_34" -> "1131 pad_12" [label="(1, 14, 14, 384)", style=solid]; +"1129 add_34" -> "1189 add_36" [label="(1, 14, 14, 384)", style=solid]; +"1130 _frozen_param85" -> "1165 add_35" [label="(1, 12, 64, 64)", style=solid]; +"1131 pad_12" -> "1132 view_57" [label="(1, 16, 16, 384)", style=solid]; +"1132 view_57" -> "1133 permute_47" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1133 permute_47" -> "1134 reshape_45" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1134 reshape_45" -> "1136 reshape_45_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1135 _frozen_param86" -> "1143 linear_64" [label="(1152,)", style=solid]; +"1136 reshape_45_0_0_nncf_smooth_quant_0" -> "1137 quantize_per_tensor_default_63" [label="(4, 64, 384)", style=solid]; +"1137 quantize_per_tensor_default_63" -> "1138 dequantize_per_tensor_default_63" [label="(4, 64, 384)", style=solid]; +"1138 dequantize_per_tensor_default_63" -> "1143 linear_64" [label="(4, 64, 384)", style=solid]; +"1139 linear_64_scale_0" -> "1142 dequantize_per_channel_default_43" [label="(1152,)", style=solid]; +"1140 linear_64_zero_point_0" -> "1142 dequantize_per_channel_default_43" [label="(1152,)", style=solid]; +"1141 _frozen_param67_1" -> "1142 dequantize_per_channel_default_43" [label="(1152, 384)", style=solid]; +"1142 dequantize_per_channel_default_43" -> "1143 linear_64" [label="(1152, 384)", style=solid]; +"1143 linear_64" -> "1144 reshape_46" [label="(4, 64, 1152)", style=solid]; +"1144 reshape_46" -> "1145 permute_48" [label="(4, 64, 3, 12, 32)", style=solid]; +"1145 permute_48" -> "1146 select_30" [label="(3, 4, 12, 64, 32)", style=solid]; +"1145 permute_48" -> "1147 select_31" [label="(3, 4, 12, 64, 32)", style=solid]; +"1145 permute_48" -> "1148 select_32" [label="(3, 4, 12, 64, 32)", style=solid]; +"1146 select_30" -> "1149 linalg_vector_norm_20" [label="(4, 12, 64, 32)", style=solid]; +"1146 select_30" -> "1151 expand_as_20" [label="(4, 12, 64, 32)", style=solid]; +"1146 select_30" -> "1152 div_20" [label="(4, 12, 64, 32)", style=solid]; +"1147 select_31" -> "1155 linalg_vector_norm_21" [label="(4, 12, 64, 32)", style=solid]; +"1147 select_31" -> "1157 expand_as_21" [label="(4, 12, 64, 32)", style=solid]; +"1147 select_31" -> "1158 div_21" [label="(4, 12, 64, 32)", style=solid]; +"1148 select_32" -> "1168 matmul_21" [label="(4, 12, 64, 32)", style=solid]; +"1149 linalg_vector_norm_20" -> "1150 clamp_min_20" [label="(4, 12, 64, 1)", style=solid]; +"1150 clamp_min_20" -> "1151 expand_as_20" [label="(4, 12, 64, 1)", style=solid]; +"1151 expand_as_20" -> "1152 div_20" [label="(4, 12, 64, 32)", style=solid]; +"1152 div_20" -> "1153 quantize_per_tensor_default_64" [label="(4, 12, 64, 32)", style=solid]; +"1153 quantize_per_tensor_default_64" -> "1154 dequantize_per_tensor_default_64" [label="(4, 12, 64, 32)", style=solid]; +"1154 dequantize_per_tensor_default_64" -> "1162 matmul_20" [label="(4, 12, 64, 32)", style=solid]; +"1155 linalg_vector_norm_21" -> "1156 clamp_min_21" [label="(4, 12, 64, 1)", style=solid]; +"1156 clamp_min_21" -> "1157 expand_as_21" [label="(4, 12, 64, 1)", style=solid]; +"1157 expand_as_21" -> "1158 div_21" [label="(4, 12, 64, 32)", style=solid]; +"1158 div_21" -> "1159 quantize_per_tensor_default_65" [label="(4, 12, 64, 32)", style=solid]; +"1159 quantize_per_tensor_default_65" -> "1160 dequantize_per_tensor_default_65" [label="(4, 12, 64, 32)", style=solid]; +"1160 dequantize_per_tensor_default_65" -> "1161 transpose_20" [label="(4, 12, 64, 32)", style=solid]; +"1161 transpose_20" -> "1162 matmul_20" [label="(4, 12, 32, 64)", style=solid]; +"1162 matmul_20" -> "1164 mul_21" [label="(4, 12, 64, 64)", style=solid]; +"1163 _frozen_param88" -> "1164 mul_21" [label="(12, 1, 1)", style=solid]; +"1164 mul_21" -> "1165 add_35" [label="(4, 12, 64, 64)", style=solid]; +"1165 add_35" -> "1166 softmax_10" [label="(4, 12, 64, 64)", style=solid]; +"1166 softmax_10" -> "1167 dropout_40" [label="(4, 12, 64, 64)", style=solid]; +"1167 dropout_40" -> "1168 matmul_21" [label="(4, 12, 64, 64)", style=solid]; +"1168 matmul_21" -> "1169 transpose_21" [label="(4, 12, 64, 32)", style=solid]; +"1169 transpose_21" -> "1170 reshape_47" [label="(4, 64, 12, 32)", style=solid]; +"1170 reshape_47" -> "1171 reshape_47_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1171 reshape_47_0_0_nncf_smooth_quant_0" -> "1172 quantize_per_tensor_default_66" [label="(4, 64, 384)", style=solid]; +"1172 quantize_per_tensor_default_66" -> "1173 dequantize_per_tensor_default_66" [label="(4, 64, 384)", style=solid]; +"1173 dequantize_per_tensor_default_66" -> "1178 linear_65" [label="(4, 64, 384)", style=solid]; +"1174 linear_65_scale_0" -> "1177 dequantize_per_channel_default_44" [label="(384,)", style=solid]; +"1175 linear_65_zero_point_0" -> "1177 dequantize_per_channel_default_44" [label="(384,)", style=solid]; +"1176 _frozen_param70_1" -> "1177 dequantize_per_channel_default_44" [label="(384, 384)", style=solid]; +"1177 dequantize_per_channel_default_44" -> "1178 linear_65" [label="(384, 384)", style=solid]; +"1178 linear_65" -> "1179 dropout_41" [label="(4, 64, 384)", style=solid]; +"1179 dropout_41" -> "1180 view_58" [label="(4, 64, 384)", style=solid]; +"1180 view_58" -> "1181 permute_49" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1181 permute_49" -> "1182 reshape_48" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1182 reshape_48" -> "1183 slice_162" [label="(1, 16, 16, 384)", style=solid]; +"1183 slice_162" -> "1184 slice_163" [label="(1, 16, 16, 384)", style=solid]; +"1184 slice_163" -> "1185 slice_164" [label="(1, 14, 16, 384)", style=solid]; +"1185 slice_164" -> "1186 slice_165" [label="(1, 14, 14, 384)", style=solid]; +"1186 slice_165" -> "1187 contiguous_19" [label="(1, 14, 14, 384)", style=solid]; +"1187 contiguous_19" -> "1188 layer_norm_23" [label="(1, 14, 14, 384)", style=solid]; +"1188 layer_norm_23" -> "1189 add_36" [label="(1, 14, 14, 384)", style=solid]; +"1189 add_36" -> "1190 add_36_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"1189 add_36" -> "1210 add_37" [label="(1, 14, 14, 384)", style=solid]; +"1190 add_36_0_0_nncf_smooth_quant_0" -> "1191 quantize_per_tensor_default_67" [label="(1, 14, 14, 384)", style=solid]; +"1191 quantize_per_tensor_default_67" -> "1192 dequantize_per_tensor_default_67" [label="(1, 14, 14, 384)", style=solid]; +"1192 dequantize_per_tensor_default_67" -> "1197 linear_66" [label="(1, 14, 14, 384)", style=solid]; +"1193 linear_66_scale_0" -> "1196 dequantize_per_channel_default_45" [label="(1536,)", style=solid]; +"1194 linear_66_zero_point_0" -> "1196 dequantize_per_channel_default_45" [label="(1536,)", style=solid]; +"1195 _frozen_param74_1" -> "1196 dequantize_per_channel_default_45" [label="(1536, 384)", style=solid]; +"1196 dequantize_per_channel_default_45" -> "1197 linear_66" [label="(1536, 384)", style=solid]; +"1197 linear_66" -> "1198 gelu_10" [label="(1, 14, 14, 1536)", style=solid]; +"1198 gelu_10" -> "1199 dropout_42" [label="(1, 14, 14, 1536)", style=solid]; +"1199 dropout_42" -> "1200 dropout_42_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"1200 dropout_42_0_0_nncf_smooth_quant_0" -> "1201 quantize_per_tensor_default_68" [label="(1, 14, 14, 1536)", style=solid]; +"1201 quantize_per_tensor_default_68" -> "1202 dequantize_per_tensor_default_68" [label="(1, 14, 14, 1536)", style=solid]; +"1202 dequantize_per_tensor_default_68" -> "1207 linear_67" [label="(1, 14, 14, 1536)", style=solid]; +"1203 linear_67_scale_0" -> "1206 dequantize_per_channel_default_46" [label="(384,)", style=solid]; +"1204 linear_67_zero_point_0" -> "1206 dequantize_per_channel_default_46" [label="(384,)", style=solid]; +"1205 _frozen_param76_1" -> "1206 dequantize_per_channel_default_46" [label="(384, 1536)", style=solid]; +"1206 dequantize_per_channel_default_46" -> "1207 linear_67" [label="(384, 1536)", style=solid]; +"1207 linear_67" -> "1208 dropout_43" [label="(1, 14, 14, 384)", style=solid]; +"1208 dropout_43" -> "1209 layer_norm_24" [label="(1, 14, 14, 384)", style=solid]; +"1209 layer_norm_24" -> "1210 add_37" [label="(1, 14, 14, 384)", style=solid]; +"1210 add_37" -> "1212 pad_13" [label="(1, 14, 14, 384)", style=solid]; +"1210 add_37" -> "1288 add_40" [label="(1, 14, 14, 384)", style=solid]; +"1211 _frozen_param89" -> "1247 add_38" [label="(1, 12, 64, 64)", style=solid]; +"1212 pad_13" -> "1213 roll_10" [label="(1, 16, 16, 384)", style=solid]; +"1213 roll_10" -> "1214 view_61" [label="(1, 16, 16, 384)", style=solid]; +"1214 view_61" -> "1215 permute_51" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1215 permute_51" -> "1216 reshape_49" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1216 reshape_49" -> "1218 reshape_49_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1216 reshape_49" -> "1248 new_zeros_5" [label="(4, 64, 384)", style=solid]; +"1217 _frozen_param90" -> "1225 linear_70" [label="(1152,)", style=solid]; +"1218 reshape_49_0_0_nncf_smooth_quant_0" -> "1219 quantize_per_tensor_default_69" [label="(4, 64, 384)", style=solid]; +"1219 quantize_per_tensor_default_69" -> "1220 dequantize_per_tensor_default_69" [label="(4, 64, 384)", style=solid]; +"1220 dequantize_per_tensor_default_69" -> "1225 linear_70" [label="(4, 64, 384)", style=solid]; +"1221 linear_70_scale_0" -> "1224 dequantize_per_channel_default_47" [label="(1152,)", style=solid]; +"1222 linear_70_zero_point_0" -> "1224 dequantize_per_channel_default_47" [label="(1152,)", style=solid]; +"1223 _frozen_param77_1" -> "1224 dequantize_per_channel_default_47" [label="(1152, 384)", style=solid]; +"1224 dequantize_per_channel_default_47" -> "1225 linear_70" [label="(1152, 384)", style=solid]; +"1225 linear_70" -> "1226 reshape_50" [label="(4, 64, 1152)", style=solid]; +"1226 reshape_50" -> "1227 permute_52" [label="(4, 64, 3, 12, 32)", style=solid]; +"1227 permute_52" -> "1228 select_33" [label="(3, 4, 12, 64, 32)", style=solid]; +"1227 permute_52" -> "1229 select_34" [label="(3, 4, 12, 64, 32)", style=solid]; +"1227 permute_52" -> "1230 select_35" [label="(3, 4, 12, 64, 32)", style=solid]; +"1228 select_33" -> "1231 linalg_vector_norm_22" [label="(4, 12, 64, 32)", style=solid]; +"1228 select_33" -> "1233 expand_as_22" [label="(4, 12, 64, 32)", style=solid]; +"1228 select_33" -> "1234 div_22" [label="(4, 12, 64, 32)", style=solid]; +"1229 select_34" -> "1237 linalg_vector_norm_23" [label="(4, 12, 64, 32)", style=solid]; +"1229 select_34" -> "1239 expand_as_23" [label="(4, 12, 64, 32)", style=solid]; +"1229 select_34" -> "1240 div_23" [label="(4, 12, 64, 32)", style=solid]; +"1230 select_35" -> "1266 matmul_23" [label="(4, 12, 64, 32)", style=solid]; +"1231 linalg_vector_norm_22" -> "1232 clamp_min_22" [label="(4, 12, 64, 1)", style=solid]; +"1232 clamp_min_22" -> "1233 expand_as_22" [label="(4, 12, 64, 1)", style=solid]; +"1233 expand_as_22" -> "1234 div_22" [label="(4, 12, 64, 32)", style=solid]; +"1234 div_22" -> "1235 quantize_per_tensor_default_70" [label="(4, 12, 64, 32)", style=solid]; +"1235 quantize_per_tensor_default_70" -> "1236 dequantize_per_tensor_default_70" [label="(4, 12, 64, 32)", style=solid]; +"1236 dequantize_per_tensor_default_70" -> "1244 matmul_22" [label="(4, 12, 64, 32)", style=solid]; +"1237 linalg_vector_norm_23" -> "1238 clamp_min_23" [label="(4, 12, 64, 1)", style=solid]; +"1238 clamp_min_23" -> "1239 expand_as_23" [label="(4, 12, 64, 1)", style=solid]; +"1239 expand_as_23" -> "1240 div_23" [label="(4, 12, 64, 32)", style=solid]; +"1240 div_23" -> "1241 quantize_per_tensor_default_71" [label="(4, 12, 64, 32)", style=solid]; +"1241 quantize_per_tensor_default_71" -> "1242 dequantize_per_tensor_default_71" [label="(4, 12, 64, 32)", style=solid]; +"1242 dequantize_per_tensor_default_71" -> "1243 transpose_22" [label="(4, 12, 64, 32)", style=solid]; +"1243 transpose_22" -> "1244 matmul_22" [label="(4, 12, 32, 64)", style=solid]; +"1244 matmul_22" -> "1246 mul_23" [label="(4, 12, 64, 64)", style=solid]; +"1245 _frozen_param92" -> "1246 mul_23" [label="(12, 1, 1)", style=solid]; +"1246 mul_23" -> "1247 add_38" [label="(4, 12, 64, 64)", style=solid]; +"1247 add_38" -> "1259 view_63" [label="(4, 12, 64, 64)", style=solid]; +"1248 new_zeros_5" -> "1249 view_62" [label="(16, 16)", style=solid]; +"1249 view_62" -> "1250 permute_53" [label="(2, 8, 2, 8)", style=solid]; +"1250 permute_53" -> "1251 reshape_51" [label="(2, 2, 8, 8)", style=solid]; +"1251 reshape_51" -> "1252 unsqueeze_32" [label="(4, 64)", style=solid]; +"1251 reshape_51" -> "1253 unsqueeze_33" [label="(4, 64)", style=solid]; +"1252 unsqueeze_32" -> "1254 sub_5" [label="(4, 1, 64)", style=solid]; +"1253 unsqueeze_33" -> "1254 sub_5" [label="(4, 64, 1)", style=solid]; +"1254 sub_5" -> "1255 ne_5" [label="(4, 64, 64)", style=solid]; +"1254 sub_5" -> "1256 masked_fill_10" [label="(4, 64, 64)", style=solid]; +"1254 sub_5" -> "1257 eq_5" [label="(4, 64, 64)", style=solid]; +"1255 ne_5" -> "1256 masked_fill_10" [label="(4, 64, 64)", style=solid]; +"1256 masked_fill_10" -> "1258 masked_fill_11" [label="(4, 64, 64)", style=solid]; +"1257 eq_5" -> "1258 masked_fill_11" [label="(4, 64, 64)", style=solid]; +"1258 masked_fill_11" -> "1260 unsqueeze_34" [label="(4, 64, 64)", style=solid]; +"1259 view_63" -> "1262 add_39" [label="(1, 4, 12, 64, 64)", style=solid]; +"1260 unsqueeze_34" -> "1261 unsqueeze_35" [label="(4, 1, 64, 64)", style=solid]; +"1261 unsqueeze_35" -> "1262 add_39" [label="(1, 4, 1, 64, 64)", style=solid]; +"1262 add_39" -> "1263 view_64" [label="(1, 4, 12, 64, 64)", style=solid]; +"1263 view_64" -> "1264 softmax_11" [label="(4, 12, 64, 64)", style=solid]; +"1264 softmax_11" -> "1265 dropout_44" [label="(4, 12, 64, 64)", style=solid]; +"1265 dropout_44" -> "1266 matmul_23" [label="(4, 12, 64, 64)", style=solid]; +"1266 matmul_23" -> "1267 transpose_23" [label="(4, 12, 64, 32)", style=solid]; +"1267 transpose_23" -> "1268 reshape_52" [label="(4, 64, 12, 32)", style=solid]; +"1268 reshape_52" -> "1269 reshape_52_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1269 reshape_52_0_0_nncf_smooth_quant_0" -> "1270 quantize_per_tensor_default_72" [label="(4, 64, 384)", style=solid]; +"1270 quantize_per_tensor_default_72" -> "1271 dequantize_per_tensor_default_72" [label="(4, 64, 384)", style=solid]; +"1271 dequantize_per_tensor_default_72" -> "1276 linear_71" [label="(4, 64, 384)", style=solid]; +"1272 linear_71_scale_0" -> "1275 dequantize_per_channel_default_48" [label="(384,)", style=solid]; +"1273 linear_71_zero_point_0" -> "1275 dequantize_per_channel_default_48" [label="(384,)", style=solid]; +"1274 _frozen_param78_1" -> "1275 dequantize_per_channel_default_48" [label="(384, 384)", style=solid]; +"1275 dequantize_per_channel_default_48" -> "1276 linear_71" [label="(384, 384)", style=solid]; +"1276 linear_71" -> "1277 dropout_45" [label="(4, 64, 384)", style=solid]; +"1277 dropout_45" -> "1278 view_65" [label="(4, 64, 384)", style=solid]; +"1278 view_65" -> "1279 permute_54" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1279 permute_54" -> "1280 reshape_53" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1280 reshape_53" -> "1281 roll_11" [label="(1, 16, 16, 384)", style=solid]; +"1281 roll_11" -> "1282 slice_185" [label="(1, 16, 16, 384)", style=solid]; +"1282 slice_185" -> "1283 slice_186" [label="(1, 16, 16, 384)", style=solid]; +"1283 slice_186" -> "1284 slice_187" [label="(1, 14, 16, 384)", style=solid]; +"1284 slice_187" -> "1285 slice_188" [label="(1, 14, 14, 384)", style=solid]; +"1285 slice_188" -> "1286 contiguous_21" [label="(1, 14, 14, 384)", style=solid]; +"1286 contiguous_21" -> "1287 layer_norm_25" [label="(1, 14, 14, 384)", style=solid]; +"1287 layer_norm_25" -> "1288 add_40" [label="(1, 14, 14, 384)", style=solid]; +"1288 add_40" -> "1289 add_40_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"1288 add_40" -> "1309 add_41" [label="(1, 14, 14, 384)", style=solid]; +"1289 add_40_0_0_nncf_smooth_quant_0" -> "1290 quantize_per_tensor_default_73" [label="(1, 14, 14, 384)", style=solid]; +"1290 quantize_per_tensor_default_73" -> "1291 dequantize_per_tensor_default_73" [label="(1, 14, 14, 384)", style=solid]; +"1291 dequantize_per_tensor_default_73" -> "1296 linear_72" [label="(1, 14, 14, 384)", style=solid]; +"1292 linear_72_scale_0" -> "1295 dequantize_per_channel_default_49" [label="(1536,)", style=solid]; +"1293 linear_72_zero_point_0" -> "1295 dequantize_per_channel_default_49" [label="(1536,)", style=solid]; +"1294 _frozen_param79_1" -> "1295 dequantize_per_channel_default_49" [label="(1536, 384)", style=solid]; +"1295 dequantize_per_channel_default_49" -> "1296 linear_72" [label="(1536, 384)", style=solid]; +"1296 linear_72" -> "1297 gelu_11" [label="(1, 14, 14, 1536)", style=solid]; +"1297 gelu_11" -> "1298 dropout_46" [label="(1, 14, 14, 1536)", style=solid]; +"1298 dropout_46" -> "1299 dropout_46_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"1299 dropout_46_0_0_nncf_smooth_quant_0" -> "1300 quantize_per_tensor_default_74" [label="(1, 14, 14, 1536)", style=solid]; +"1300 quantize_per_tensor_default_74" -> "1301 dequantize_per_tensor_default_74" [label="(1, 14, 14, 1536)", style=solid]; +"1301 dequantize_per_tensor_default_74" -> "1306 linear_73" [label="(1, 14, 14, 1536)", style=solid]; +"1302 linear_73_scale_0" -> "1305 dequantize_per_channel_default_50" [label="(384,)", style=solid]; +"1303 linear_73_zero_point_0" -> "1305 dequantize_per_channel_default_50" [label="(384,)", style=solid]; +"1304 _frozen_param80_1" -> "1305 dequantize_per_channel_default_50" [label="(384, 1536)", style=solid]; +"1305 dequantize_per_channel_default_50" -> "1306 linear_73" [label="(384, 1536)", style=solid]; +"1306 linear_73" -> "1307 dropout_47" [label="(1, 14, 14, 384)", style=solid]; +"1307 dropout_47" -> "1308 layer_norm_26" [label="(1, 14, 14, 384)", style=solid]; +"1308 layer_norm_26" -> "1309 add_41" [label="(1, 14, 14, 384)", style=solid]; +"1309 add_41" -> "1311 pad_14" [label="(1, 14, 14, 384)", style=solid]; +"1309 add_41" -> "1369 add_43" [label="(1, 14, 14, 384)", style=solid]; +"1310 _frozen_param102" -> "1345 add_42" [label="(1, 12, 64, 64)", style=solid]; +"1311 pad_14" -> "1312 view_68" [label="(1, 16, 16, 384)", style=solid]; +"1312 view_68" -> "1313 permute_56" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1313 permute_56" -> "1314 reshape_54" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1314 reshape_54" -> "1316 reshape_54_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1315 _frozen_param103" -> "1323 linear_76" [label="(1152,)", style=solid]; +"1316 reshape_54_0_0_nncf_smooth_quant_0" -> "1317 quantize_per_tensor_default_75" [label="(4, 64, 384)", style=solid]; +"1317 quantize_per_tensor_default_75" -> "1318 dequantize_per_tensor_default_75" [label="(4, 64, 384)", style=solid]; +"1318 dequantize_per_tensor_default_75" -> "1323 linear_76" [label="(4, 64, 384)", style=solid]; +"1319 linear_76_scale_0" -> "1322 dequantize_per_channel_default_51" [label="(1152,)", style=solid]; +"1320 linear_76_zero_point_0" -> "1322 dequantize_per_channel_default_51" [label="(1152,)", style=solid]; +"1321 _frozen_param81_1" -> "1322 dequantize_per_channel_default_51" [label="(1152, 384)", style=solid]; +"1322 dequantize_per_channel_default_51" -> "1323 linear_76" [label="(1152, 384)", style=solid]; +"1323 linear_76" -> "1324 reshape_55" [label="(4, 64, 1152)", style=solid]; +"1324 reshape_55" -> "1325 permute_57" [label="(4, 64, 3, 12, 32)", style=solid]; +"1325 permute_57" -> "1326 select_36" [label="(3, 4, 12, 64, 32)", style=solid]; +"1325 permute_57" -> "1327 select_37" [label="(3, 4, 12, 64, 32)", style=solid]; +"1325 permute_57" -> "1328 select_38" [label="(3, 4, 12, 64, 32)", style=solid]; +"1326 select_36" -> "1329 linalg_vector_norm_24" [label="(4, 12, 64, 32)", style=solid]; +"1326 select_36" -> "1331 expand_as_24" [label="(4, 12, 64, 32)", style=solid]; +"1326 select_36" -> "1332 div_24" [label="(4, 12, 64, 32)", style=solid]; +"1327 select_37" -> "1335 linalg_vector_norm_25" [label="(4, 12, 64, 32)", style=solid]; +"1327 select_37" -> "1337 expand_as_25" [label="(4, 12, 64, 32)", style=solid]; +"1327 select_37" -> "1338 div_25" [label="(4, 12, 64, 32)", style=solid]; +"1328 select_38" -> "1348 matmul_25" [label="(4, 12, 64, 32)", style=solid]; +"1329 linalg_vector_norm_24" -> "1330 clamp_min_24" [label="(4, 12, 64, 1)", style=solid]; +"1330 clamp_min_24" -> "1331 expand_as_24" [label="(4, 12, 64, 1)", style=solid]; +"1331 expand_as_24" -> "1332 div_24" [label="(4, 12, 64, 32)", style=solid]; +"1332 div_24" -> "1333 quantize_per_tensor_default_76" [label="(4, 12, 64, 32)", style=solid]; +"1333 quantize_per_tensor_default_76" -> "1334 dequantize_per_tensor_default_76" [label="(4, 12, 64, 32)", style=solid]; +"1334 dequantize_per_tensor_default_76" -> "1342 matmul_24" [label="(4, 12, 64, 32)", style=solid]; +"1335 linalg_vector_norm_25" -> "1336 clamp_min_25" [label="(4, 12, 64, 1)", style=solid]; +"1336 clamp_min_25" -> "1337 expand_as_25" [label="(4, 12, 64, 1)", style=solid]; +"1337 expand_as_25" -> "1338 div_25" [label="(4, 12, 64, 32)", style=solid]; +"1338 div_25" -> "1339 quantize_per_tensor_default_77" [label="(4, 12, 64, 32)", style=solid]; +"1339 quantize_per_tensor_default_77" -> "1340 dequantize_per_tensor_default_77" [label="(4, 12, 64, 32)", style=solid]; +"1340 dequantize_per_tensor_default_77" -> "1341 transpose_24" [label="(4, 12, 64, 32)", style=solid]; +"1341 transpose_24" -> "1342 matmul_24" [label="(4, 12, 32, 64)", style=solid]; +"1342 matmul_24" -> "1344 mul_25" [label="(4, 12, 64, 64)", style=solid]; +"1343 _frozen_param105" -> "1344 mul_25" [label="(12, 1, 1)", style=solid]; +"1344 mul_25" -> "1345 add_42" [label="(4, 12, 64, 64)", style=solid]; +"1345 add_42" -> "1346 softmax_12" [label="(4, 12, 64, 64)", style=solid]; +"1346 softmax_12" -> "1347 dropout_48" [label="(4, 12, 64, 64)", style=solid]; +"1347 dropout_48" -> "1348 matmul_25" [label="(4, 12, 64, 64)", style=solid]; +"1348 matmul_25" -> "1349 transpose_25" [label="(4, 12, 64, 32)", style=solid]; +"1349 transpose_25" -> "1350 reshape_56" [label="(4, 64, 12, 32)", style=solid]; +"1350 reshape_56" -> "1351 reshape_56_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1351 reshape_56_0_0_nncf_smooth_quant_0" -> "1352 quantize_per_tensor_default_78" [label="(4, 64, 384)", style=solid]; +"1352 quantize_per_tensor_default_78" -> "1353 dequantize_per_tensor_default_78" [label="(4, 64, 384)", style=solid]; +"1353 dequantize_per_tensor_default_78" -> "1358 linear_77" [label="(4, 64, 384)", style=solid]; +"1354 linear_77_scale_0" -> "1357 dequantize_per_channel_default_52" [label="(384,)", style=solid]; +"1355 linear_77_zero_point_0" -> "1357 dequantize_per_channel_default_52" [label="(384,)", style=solid]; +"1356 _frozen_param82_1" -> "1357 dequantize_per_channel_default_52" [label="(384, 384)", style=solid]; +"1357 dequantize_per_channel_default_52" -> "1358 linear_77" [label="(384, 384)", style=solid]; +"1358 linear_77" -> "1359 dropout_49" [label="(4, 64, 384)", style=solid]; +"1359 dropout_49" -> "1360 view_69" [label="(4, 64, 384)", style=solid]; +"1360 view_69" -> "1361 permute_58" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1361 permute_58" -> "1362 reshape_57" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1362 reshape_57" -> "1363 slice_190" [label="(1, 16, 16, 384)", style=solid]; +"1363 slice_190" -> "1364 slice_191" [label="(1, 16, 16, 384)", style=solid]; +"1364 slice_191" -> "1365 slice_192" [label="(1, 14, 16, 384)", style=solid]; +"1365 slice_192" -> "1366 slice_193" [label="(1, 14, 14, 384)", style=solid]; +"1366 slice_193" -> "1367 contiguous_23" [label="(1, 14, 14, 384)", style=solid]; +"1367 contiguous_23" -> "1368 layer_norm_27" [label="(1, 14, 14, 384)", style=solid]; +"1368 layer_norm_27" -> "1369 add_43" [label="(1, 14, 14, 384)", style=solid]; +"1369 add_43" -> "1370 add_43_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"1369 add_43" -> "1390 add_44" [label="(1, 14, 14, 384)", style=solid]; +"1370 add_43_0_0_nncf_smooth_quant_0" -> "1371 quantize_per_tensor_default_79" [label="(1, 14, 14, 384)", style=solid]; +"1371 quantize_per_tensor_default_79" -> "1372 dequantize_per_tensor_default_79" [label="(1, 14, 14, 384)", style=solid]; +"1372 dequantize_per_tensor_default_79" -> "1377 linear_78" [label="(1, 14, 14, 384)", style=solid]; +"1373 linear_78_scale_0" -> "1376 dequantize_per_channel_default_53" [label="(1536,)", style=solid]; +"1374 linear_78_zero_point_0" -> "1376 dequantize_per_channel_default_53" [label="(1536,)", style=solid]; +"1375 _frozen_param83_1" -> "1376 dequantize_per_channel_default_53" [label="(1536, 384)", style=solid]; +"1376 dequantize_per_channel_default_53" -> "1377 linear_78" [label="(1536, 384)", style=solid]; +"1377 linear_78" -> "1378 gelu_12" [label="(1, 14, 14, 1536)", style=solid]; +"1378 gelu_12" -> "1379 dropout_50" [label="(1, 14, 14, 1536)", style=solid]; +"1379 dropout_50" -> "1380 dropout_50_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"1380 dropout_50_0_0_nncf_smooth_quant_0" -> "1381 quantize_per_tensor_default_80" [label="(1, 14, 14, 1536)", style=solid]; +"1381 quantize_per_tensor_default_80" -> "1382 dequantize_per_tensor_default_80" [label="(1, 14, 14, 1536)", style=solid]; +"1382 dequantize_per_tensor_default_80" -> "1387 linear_79" [label="(1, 14, 14, 1536)", style=solid]; +"1383 linear_79_scale_0" -> "1386 dequantize_per_channel_default_54" [label="(384,)", style=solid]; +"1384 linear_79_zero_point_0" -> "1386 dequantize_per_channel_default_54" [label="(384,)", style=solid]; +"1385 _frozen_param84_1" -> "1386 dequantize_per_channel_default_54" [label="(384, 1536)", style=solid]; +"1386 dequantize_per_channel_default_54" -> "1387 linear_79" [label="(384, 1536)", style=solid]; +"1387 linear_79" -> "1388 dropout_51" [label="(1, 14, 14, 384)", style=solid]; +"1388 dropout_51" -> "1389 layer_norm_28" [label="(1, 14, 14, 384)", style=solid]; +"1389 layer_norm_28" -> "1390 add_44" [label="(1, 14, 14, 384)", style=solid]; +"1390 add_44" -> "1392 pad_15" [label="(1, 14, 14, 384)", style=solid]; +"1390 add_44" -> "1468 add_47" [label="(1, 14, 14, 384)", style=solid]; +"1391 _frozen_param106" -> "1427 add_45" [label="(1, 12, 64, 64)", style=solid]; +"1392 pad_15" -> "1393 roll_12" [label="(1, 16, 16, 384)", style=solid]; +"1393 roll_12" -> "1394 view_72" [label="(1, 16, 16, 384)", style=solid]; +"1394 view_72" -> "1395 permute_60" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1395 permute_60" -> "1396 reshape_58" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1396 reshape_58" -> "1398 reshape_58_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1396 reshape_58" -> "1428 new_zeros_6" [label="(4, 64, 384)", style=solid]; +"1397 _frozen_param107" -> "1405 linear_82" [label="(1152,)", style=solid]; +"1398 reshape_58_0_0_nncf_smooth_quant_0" -> "1399 quantize_per_tensor_default_81" [label="(4, 64, 384)", style=solid]; +"1399 quantize_per_tensor_default_81" -> "1400 dequantize_per_tensor_default_81" [label="(4, 64, 384)", style=solid]; +"1400 dequantize_per_tensor_default_81" -> "1405 linear_82" [label="(4, 64, 384)", style=solid]; +"1401 linear_82_scale_0" -> "1404 dequantize_per_channel_default_55" [label="(1152,)", style=solid]; +"1402 linear_82_zero_point_0" -> "1404 dequantize_per_channel_default_55" [label="(1152,)", style=solid]; +"1403 _frozen_param87_1" -> "1404 dequantize_per_channel_default_55" [label="(1152, 384)", style=solid]; +"1404 dequantize_per_channel_default_55" -> "1405 linear_82" [label="(1152, 384)", style=solid]; +"1405 linear_82" -> "1406 reshape_59" [label="(4, 64, 1152)", style=solid]; +"1406 reshape_59" -> "1407 permute_61" [label="(4, 64, 3, 12, 32)", style=solid]; +"1407 permute_61" -> "1408 select_39" [label="(3, 4, 12, 64, 32)", style=solid]; +"1407 permute_61" -> "1409 select_40" [label="(3, 4, 12, 64, 32)", style=solid]; +"1407 permute_61" -> "1410 select_41" [label="(3, 4, 12, 64, 32)", style=solid]; +"1408 select_39" -> "1411 linalg_vector_norm_26" [label="(4, 12, 64, 32)", style=solid]; +"1408 select_39" -> "1413 expand_as_26" [label="(4, 12, 64, 32)", style=solid]; +"1408 select_39" -> "1414 div_26" [label="(4, 12, 64, 32)", style=solid]; +"1409 select_40" -> "1417 linalg_vector_norm_27" [label="(4, 12, 64, 32)", style=solid]; +"1409 select_40" -> "1419 expand_as_27" [label="(4, 12, 64, 32)", style=solid]; +"1409 select_40" -> "1420 div_27" [label="(4, 12, 64, 32)", style=solid]; +"1410 select_41" -> "1446 matmul_27" [label="(4, 12, 64, 32)", style=solid]; +"1411 linalg_vector_norm_26" -> "1412 clamp_min_26" [label="(4, 12, 64, 1)", style=solid]; +"1412 clamp_min_26" -> "1413 expand_as_26" [label="(4, 12, 64, 1)", style=solid]; +"1413 expand_as_26" -> "1414 div_26" [label="(4, 12, 64, 32)", style=solid]; +"1414 div_26" -> "1415 quantize_per_tensor_default_82" [label="(4, 12, 64, 32)", style=solid]; +"1415 quantize_per_tensor_default_82" -> "1416 dequantize_per_tensor_default_82" [label="(4, 12, 64, 32)", style=solid]; +"1416 dequantize_per_tensor_default_82" -> "1424 matmul_26" [label="(4, 12, 64, 32)", style=solid]; +"1417 linalg_vector_norm_27" -> "1418 clamp_min_27" [label="(4, 12, 64, 1)", style=solid]; +"1418 clamp_min_27" -> "1419 expand_as_27" [label="(4, 12, 64, 1)", style=solid]; +"1419 expand_as_27" -> "1420 div_27" [label="(4, 12, 64, 32)", style=solid]; +"1420 div_27" -> "1421 quantize_per_tensor_default_83" [label="(4, 12, 64, 32)", style=solid]; +"1421 quantize_per_tensor_default_83" -> "1422 dequantize_per_tensor_default_83" [label="(4, 12, 64, 32)", style=solid]; +"1422 dequantize_per_tensor_default_83" -> "1423 transpose_26" [label="(4, 12, 64, 32)", style=solid]; +"1423 transpose_26" -> "1424 matmul_26" [label="(4, 12, 32, 64)", style=solid]; +"1424 matmul_26" -> "1426 mul_27" [label="(4, 12, 64, 64)", style=solid]; +"1425 _frozen_param109" -> "1426 mul_27" [label="(12, 1, 1)", style=solid]; +"1426 mul_27" -> "1427 add_45" [label="(4, 12, 64, 64)", style=solid]; +"1427 add_45" -> "1439 view_74" [label="(4, 12, 64, 64)", style=solid]; +"1428 new_zeros_6" -> "1429 view_73" [label="(16, 16)", style=solid]; +"1429 view_73" -> "1430 permute_62" [label="(2, 8, 2, 8)", style=solid]; +"1430 permute_62" -> "1431 reshape_60" [label="(2, 2, 8, 8)", style=solid]; +"1431 reshape_60" -> "1432 unsqueeze_38" [label="(4, 64)", style=solid]; +"1431 reshape_60" -> "1433 unsqueeze_39" [label="(4, 64)", style=solid]; +"1432 unsqueeze_38" -> "1434 sub_6" [label="(4, 1, 64)", style=solid]; +"1433 unsqueeze_39" -> "1434 sub_6" [label="(4, 64, 1)", style=solid]; +"1434 sub_6" -> "1435 ne_6" [label="(4, 64, 64)", style=solid]; +"1434 sub_6" -> "1436 masked_fill_12" [label="(4, 64, 64)", style=solid]; +"1434 sub_6" -> "1437 eq_6" [label="(4, 64, 64)", style=solid]; +"1435 ne_6" -> "1436 masked_fill_12" [label="(4, 64, 64)", style=solid]; +"1436 masked_fill_12" -> "1438 masked_fill_13" [label="(4, 64, 64)", style=solid]; +"1437 eq_6" -> "1438 masked_fill_13" [label="(4, 64, 64)", style=solid]; +"1438 masked_fill_13" -> "1440 unsqueeze_40" [label="(4, 64, 64)", style=solid]; +"1439 view_74" -> "1442 add_46" [label="(1, 4, 12, 64, 64)", style=solid]; +"1440 unsqueeze_40" -> "1441 unsqueeze_41" [label="(4, 1, 64, 64)", style=solid]; +"1441 unsqueeze_41" -> "1442 add_46" [label="(1, 4, 1, 64, 64)", style=solid]; +"1442 add_46" -> "1443 view_75" [label="(1, 4, 12, 64, 64)", style=solid]; +"1443 view_75" -> "1444 softmax_13" [label="(4, 12, 64, 64)", style=solid]; +"1444 softmax_13" -> "1445 dropout_52" [label="(4, 12, 64, 64)", style=solid]; +"1445 dropout_52" -> "1446 matmul_27" [label="(4, 12, 64, 64)", style=solid]; +"1446 matmul_27" -> "1447 transpose_27" [label="(4, 12, 64, 32)", style=solid]; +"1447 transpose_27" -> "1448 reshape_61" [label="(4, 64, 12, 32)", style=solid]; +"1448 reshape_61" -> "1449 reshape_61_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1449 reshape_61_0_0_nncf_smooth_quant_0" -> "1450 quantize_per_tensor_default_84" [label="(4, 64, 384)", style=solid]; +"1450 quantize_per_tensor_default_84" -> "1451 dequantize_per_tensor_default_84" [label="(4, 64, 384)", style=solid]; +"1451 dequantize_per_tensor_default_84" -> "1456 linear_83" [label="(4, 64, 384)", style=solid]; +"1452 linear_83_scale_0" -> "1455 dequantize_per_channel_default_56" [label="(384,)", style=solid]; +"1453 linear_83_zero_point_0" -> "1455 dequantize_per_channel_default_56" [label="(384,)", style=solid]; +"1454 _frozen_param91_1" -> "1455 dequantize_per_channel_default_56" [label="(384, 384)", style=solid]; +"1455 dequantize_per_channel_default_56" -> "1456 linear_83" [label="(384, 384)", style=solid]; +"1456 linear_83" -> "1457 dropout_53" [label="(4, 64, 384)", style=solid]; +"1457 dropout_53" -> "1458 view_76" [label="(4, 64, 384)", style=solid]; +"1458 view_76" -> "1459 permute_63" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1459 permute_63" -> "1460 reshape_62" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1460 reshape_62" -> "1461 roll_13" [label="(1, 16, 16, 384)", style=solid]; +"1461 roll_13" -> "1462 slice_213" [label="(1, 16, 16, 384)", style=solid]; +"1462 slice_213" -> "1463 slice_214" [label="(1, 16, 16, 384)", style=solid]; +"1463 slice_214" -> "1464 slice_215" [label="(1, 14, 16, 384)", style=solid]; +"1464 slice_215" -> "1465 slice_216" [label="(1, 14, 14, 384)", style=solid]; +"1465 slice_216" -> "1466 contiguous_25" [label="(1, 14, 14, 384)", style=solid]; +"1466 contiguous_25" -> "1467 layer_norm_29" [label="(1, 14, 14, 384)", style=solid]; +"1467 layer_norm_29" -> "1468 add_47" [label="(1, 14, 14, 384)", style=solid]; +"1468 add_47" -> "1469 add_47_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"1468 add_47" -> "1489 add_48" [label="(1, 14, 14, 384)", style=solid]; +"1469 add_47_0_0_nncf_smooth_quant_0" -> "1470 quantize_per_tensor_default_85" [label="(1, 14, 14, 384)", style=solid]; +"1470 quantize_per_tensor_default_85" -> "1471 dequantize_per_tensor_default_85" [label="(1, 14, 14, 384)", style=solid]; +"1471 dequantize_per_tensor_default_85" -> "1476 linear_84" [label="(1, 14, 14, 384)", style=solid]; +"1472 linear_84_scale_0" -> "1475 dequantize_per_channel_default_57" [label="(1536,)", style=solid]; +"1473 linear_84_zero_point_0" -> "1475 dequantize_per_channel_default_57" [label="(1536,)", style=solid]; +"1474 _frozen_param93_1" -> "1475 dequantize_per_channel_default_57" [label="(1536, 384)", style=solid]; +"1475 dequantize_per_channel_default_57" -> "1476 linear_84" [label="(1536, 384)", style=solid]; +"1476 linear_84" -> "1477 gelu_13" [label="(1, 14, 14, 1536)", style=solid]; +"1477 gelu_13" -> "1478 dropout_54" [label="(1, 14, 14, 1536)", style=solid]; +"1478 dropout_54" -> "1479 dropout_54_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"1479 dropout_54_0_0_nncf_smooth_quant_0" -> "1480 quantize_per_tensor_default_86" [label="(1, 14, 14, 1536)", style=solid]; +"1480 quantize_per_tensor_default_86" -> "1481 dequantize_per_tensor_default_86" [label="(1, 14, 14, 1536)", style=solid]; +"1481 dequantize_per_tensor_default_86" -> "1486 linear_85" [label="(1, 14, 14, 1536)", style=solid]; +"1482 linear_85_scale_0" -> "1485 dequantize_per_channel_default_58" [label="(384,)", style=solid]; +"1483 linear_85_zero_point_0" -> "1485 dequantize_per_channel_default_58" [label="(384,)", style=solid]; +"1484 _frozen_param94_1" -> "1485 dequantize_per_channel_default_58" [label="(384, 1536)", style=solid]; +"1485 dequantize_per_channel_default_58" -> "1486 linear_85" [label="(384, 1536)", style=solid]; +"1486 linear_85" -> "1487 dropout_55" [label="(1, 14, 14, 384)", style=solid]; +"1487 dropout_55" -> "1488 layer_norm_30" [label="(1, 14, 14, 384)", style=solid]; +"1488 layer_norm_30" -> "1489 add_48" [label="(1, 14, 14, 384)", style=solid]; +"1489 add_48" -> "1491 pad_16" [label="(1, 14, 14, 384)", style=solid]; +"1489 add_48" -> "1549 add_50" [label="(1, 14, 14, 384)", style=solid]; +"1490 _frozen_param119" -> "1525 add_49" [label="(1, 12, 64, 64)", style=solid]; +"1491 pad_16" -> "1492 view_79" [label="(1, 16, 16, 384)", style=solid]; +"1492 view_79" -> "1493 permute_65" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1493 permute_65" -> "1494 reshape_63" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1494 reshape_63" -> "1496 reshape_63_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1495 _frozen_param120" -> "1503 linear_88" [label="(1152,)", style=solid]; +"1496 reshape_63_0_0_nncf_smooth_quant_0" -> "1497 quantize_per_tensor_default_87" [label="(4, 64, 384)", style=solid]; +"1497 quantize_per_tensor_default_87" -> "1498 dequantize_per_tensor_default_87" [label="(4, 64, 384)", style=solid]; +"1498 dequantize_per_tensor_default_87" -> "1503 linear_88" [label="(4, 64, 384)", style=solid]; +"1499 linear_88_scale_0" -> "1502 dequantize_per_channel_default_59" [label="(1152,)", style=solid]; +"1500 linear_88_zero_point_0" -> "1502 dequantize_per_channel_default_59" [label="(1152,)", style=solid]; +"1501 _frozen_param95_1" -> "1502 dequantize_per_channel_default_59" [label="(1152, 384)", style=solid]; +"1502 dequantize_per_channel_default_59" -> "1503 linear_88" [label="(1152, 384)", style=solid]; +"1503 linear_88" -> "1504 reshape_64" [label="(4, 64, 1152)", style=solid]; +"1504 reshape_64" -> "1505 permute_66" [label="(4, 64, 3, 12, 32)", style=solid]; +"1505 permute_66" -> "1506 select_42" [label="(3, 4, 12, 64, 32)", style=solid]; +"1505 permute_66" -> "1507 select_43" [label="(3, 4, 12, 64, 32)", style=solid]; +"1505 permute_66" -> "1508 select_44" [label="(3, 4, 12, 64, 32)", style=solid]; +"1506 select_42" -> "1509 linalg_vector_norm_28" [label="(4, 12, 64, 32)", style=solid]; +"1506 select_42" -> "1511 expand_as_28" [label="(4, 12, 64, 32)", style=solid]; +"1506 select_42" -> "1512 div_28" [label="(4, 12, 64, 32)", style=solid]; +"1507 select_43" -> "1515 linalg_vector_norm_29" [label="(4, 12, 64, 32)", style=solid]; +"1507 select_43" -> "1517 expand_as_29" [label="(4, 12, 64, 32)", style=solid]; +"1507 select_43" -> "1518 div_29" [label="(4, 12, 64, 32)", style=solid]; +"1508 select_44" -> "1528 matmul_29" [label="(4, 12, 64, 32)", style=solid]; +"1509 linalg_vector_norm_28" -> "1510 clamp_min_28" [label="(4, 12, 64, 1)", style=solid]; +"1510 clamp_min_28" -> "1511 expand_as_28" [label="(4, 12, 64, 1)", style=solid]; +"1511 expand_as_28" -> "1512 div_28" [label="(4, 12, 64, 32)", style=solid]; +"1512 div_28" -> "1513 quantize_per_tensor_default_88" [label="(4, 12, 64, 32)", style=solid]; +"1513 quantize_per_tensor_default_88" -> "1514 dequantize_per_tensor_default_88" [label="(4, 12, 64, 32)", style=solid]; +"1514 dequantize_per_tensor_default_88" -> "1522 matmul_28" [label="(4, 12, 64, 32)", style=solid]; +"1515 linalg_vector_norm_29" -> "1516 clamp_min_29" [label="(4, 12, 64, 1)", style=solid]; +"1516 clamp_min_29" -> "1517 expand_as_29" [label="(4, 12, 64, 1)", style=solid]; +"1517 expand_as_29" -> "1518 div_29" [label="(4, 12, 64, 32)", style=solid]; +"1518 div_29" -> "1519 quantize_per_tensor_default_89" [label="(4, 12, 64, 32)", style=solid]; +"1519 quantize_per_tensor_default_89" -> "1520 dequantize_per_tensor_default_89" [label="(4, 12, 64, 32)", style=solid]; +"1520 dequantize_per_tensor_default_89" -> "1521 transpose_28" [label="(4, 12, 64, 32)", style=solid]; +"1521 transpose_28" -> "1522 matmul_28" [label="(4, 12, 32, 64)", style=solid]; +"1522 matmul_28" -> "1524 mul_29" [label="(4, 12, 64, 64)", style=solid]; +"1523 _frozen_param122" -> "1524 mul_29" [label="(12, 1, 1)", style=solid]; +"1524 mul_29" -> "1525 add_49" [label="(4, 12, 64, 64)", style=solid]; +"1525 add_49" -> "1526 softmax_14" [label="(4, 12, 64, 64)", style=solid]; +"1526 softmax_14" -> "1527 dropout_56" [label="(4, 12, 64, 64)", style=solid]; +"1527 dropout_56" -> "1528 matmul_29" [label="(4, 12, 64, 64)", style=solid]; +"1528 matmul_29" -> "1529 transpose_29" [label="(4, 12, 64, 32)", style=solid]; +"1529 transpose_29" -> "1530 reshape_65" [label="(4, 64, 12, 32)", style=solid]; +"1530 reshape_65" -> "1531 reshape_65_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1531 reshape_65_0_0_nncf_smooth_quant_0" -> "1532 quantize_per_tensor_default_90" [label="(4, 64, 384)", style=solid]; +"1532 quantize_per_tensor_default_90" -> "1533 dequantize_per_tensor_default_90" [label="(4, 64, 384)", style=solid]; +"1533 dequantize_per_tensor_default_90" -> "1538 linear_89" [label="(4, 64, 384)", style=solid]; +"1534 linear_89_scale_0" -> "1537 dequantize_per_channel_default_60" [label="(384,)", style=solid]; +"1535 linear_89_zero_point_0" -> "1537 dequantize_per_channel_default_60" [label="(384,)", style=solid]; +"1536 _frozen_param96_1" -> "1537 dequantize_per_channel_default_60" [label="(384, 384)", style=solid]; +"1537 dequantize_per_channel_default_60" -> "1538 linear_89" [label="(384, 384)", style=solid]; +"1538 linear_89" -> "1539 dropout_57" [label="(4, 64, 384)", style=solid]; +"1539 dropout_57" -> "1540 view_80" [label="(4, 64, 384)", style=solid]; +"1540 view_80" -> "1541 permute_67" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1541 permute_67" -> "1542 reshape_66" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1542 reshape_66" -> "1543 slice_218" [label="(1, 16, 16, 384)", style=solid]; +"1543 slice_218" -> "1544 slice_219" [label="(1, 16, 16, 384)", style=solid]; +"1544 slice_219" -> "1545 slice_220" [label="(1, 14, 16, 384)", style=solid]; +"1545 slice_220" -> "1546 slice_221" [label="(1, 14, 14, 384)", style=solid]; +"1546 slice_221" -> "1547 contiguous_27" [label="(1, 14, 14, 384)", style=solid]; +"1547 contiguous_27" -> "1548 layer_norm_31" [label="(1, 14, 14, 384)", style=solid]; +"1548 layer_norm_31" -> "1549 add_50" [label="(1, 14, 14, 384)", style=solid]; +"1549 add_50" -> "1550 add_50_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"1549 add_50" -> "1570 add_51" [label="(1, 14, 14, 384)", style=solid]; +"1550 add_50_0_0_nncf_smooth_quant_0" -> "1551 quantize_per_tensor_default_91" [label="(1, 14, 14, 384)", style=solid]; +"1551 quantize_per_tensor_default_91" -> "1552 dequantize_per_tensor_default_91" [label="(1, 14, 14, 384)", style=solid]; +"1552 dequantize_per_tensor_default_91" -> "1557 linear_90" [label="(1, 14, 14, 384)", style=solid]; +"1553 linear_90_scale_0" -> "1556 dequantize_per_channel_default_61" [label="(1536,)", style=solid]; +"1554 linear_90_zero_point_0" -> "1556 dequantize_per_channel_default_61" [label="(1536,)", style=solid]; +"1555 _frozen_param97_1" -> "1556 dequantize_per_channel_default_61" [label="(1536, 384)", style=solid]; +"1556 dequantize_per_channel_default_61" -> "1557 linear_90" [label="(1536, 384)", style=solid]; +"1557 linear_90" -> "1558 gelu_14" [label="(1, 14, 14, 1536)", style=solid]; +"1558 gelu_14" -> "1559 dropout_58" [label="(1, 14, 14, 1536)", style=solid]; +"1559 dropout_58" -> "1560 dropout_58_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"1560 dropout_58_0_0_nncf_smooth_quant_0" -> "1561 quantize_per_tensor_default_92" [label="(1, 14, 14, 1536)", style=solid]; +"1561 quantize_per_tensor_default_92" -> "1562 dequantize_per_tensor_default_92" [label="(1, 14, 14, 1536)", style=solid]; +"1562 dequantize_per_tensor_default_92" -> "1567 linear_91" [label="(1, 14, 14, 1536)", style=solid]; +"1563 linear_91_scale_0" -> "1566 dequantize_per_channel_default_62" [label="(384,)", style=solid]; +"1564 linear_91_zero_point_0" -> "1566 dequantize_per_channel_default_62" [label="(384,)", style=solid]; +"1565 _frozen_param98_1" -> "1566 dequantize_per_channel_default_62" [label="(384, 1536)", style=solid]; +"1566 dequantize_per_channel_default_62" -> "1567 linear_91" [label="(384, 1536)", style=solid]; +"1567 linear_91" -> "1568 dropout_59" [label="(1, 14, 14, 384)", style=solid]; +"1568 dropout_59" -> "1569 layer_norm_32" [label="(1, 14, 14, 384)", style=solid]; +"1569 layer_norm_32" -> "1570 add_51" [label="(1, 14, 14, 384)", style=solid]; +"1570 add_51" -> "1572 pad_17" [label="(1, 14, 14, 384)", style=solid]; +"1570 add_51" -> "1648 add_54" [label="(1, 14, 14, 384)", style=solid]; +"1571 _frozen_param123" -> "1607 add_52" [label="(1, 12, 64, 64)", style=solid]; +"1572 pad_17" -> "1573 roll_14" [label="(1, 16, 16, 384)", style=solid]; +"1573 roll_14" -> "1574 view_83" [label="(1, 16, 16, 384)", style=solid]; +"1574 view_83" -> "1575 permute_69" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1575 permute_69" -> "1576 reshape_67" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1576 reshape_67" -> "1578 reshape_67_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1576 reshape_67" -> "1608 new_zeros_7" [label="(4, 64, 384)", style=solid]; +"1577 _frozen_param124" -> "1585 linear_94" [label="(1152,)", style=solid]; +"1578 reshape_67_0_0_nncf_smooth_quant_0" -> "1579 quantize_per_tensor_default_93" [label="(4, 64, 384)", style=solid]; +"1579 quantize_per_tensor_default_93" -> "1580 dequantize_per_tensor_default_93" [label="(4, 64, 384)", style=solid]; +"1580 dequantize_per_tensor_default_93" -> "1585 linear_94" [label="(4, 64, 384)", style=solid]; +"1581 linear_94_scale_0" -> "1584 dequantize_per_channel_default_63" [label="(1152,)", style=solid]; +"1582 linear_94_zero_point_0" -> "1584 dequantize_per_channel_default_63" [label="(1152,)", style=solid]; +"1583 _frozen_param99_1" -> "1584 dequantize_per_channel_default_63" [label="(1152, 384)", style=solid]; +"1584 dequantize_per_channel_default_63" -> "1585 linear_94" [label="(1152, 384)", style=solid]; +"1585 linear_94" -> "1586 reshape_68" [label="(4, 64, 1152)", style=solid]; +"1586 reshape_68" -> "1587 permute_70" [label="(4, 64, 3, 12, 32)", style=solid]; +"1587 permute_70" -> "1588 select_45" [label="(3, 4, 12, 64, 32)", style=solid]; +"1587 permute_70" -> "1589 select_46" [label="(3, 4, 12, 64, 32)", style=solid]; +"1587 permute_70" -> "1590 select_47" [label="(3, 4, 12, 64, 32)", style=solid]; +"1588 select_45" -> "1591 linalg_vector_norm_30" [label="(4, 12, 64, 32)", style=solid]; +"1588 select_45" -> "1593 expand_as_30" [label="(4, 12, 64, 32)", style=solid]; +"1588 select_45" -> "1594 div_30" [label="(4, 12, 64, 32)", style=solid]; +"1589 select_46" -> "1597 linalg_vector_norm_31" [label="(4, 12, 64, 32)", style=solid]; +"1589 select_46" -> "1599 expand_as_31" [label="(4, 12, 64, 32)", style=solid]; +"1589 select_46" -> "1600 div_31" [label="(4, 12, 64, 32)", style=solid]; +"1590 select_47" -> "1626 matmul_31" [label="(4, 12, 64, 32)", style=solid]; +"1591 linalg_vector_norm_30" -> "1592 clamp_min_30" [label="(4, 12, 64, 1)", style=solid]; +"1592 clamp_min_30" -> "1593 expand_as_30" [label="(4, 12, 64, 1)", style=solid]; +"1593 expand_as_30" -> "1594 div_30" [label="(4, 12, 64, 32)", style=solid]; +"1594 div_30" -> "1595 quantize_per_tensor_default_94" [label="(4, 12, 64, 32)", style=solid]; +"1595 quantize_per_tensor_default_94" -> "1596 dequantize_per_tensor_default_94" [label="(4, 12, 64, 32)", style=solid]; +"1596 dequantize_per_tensor_default_94" -> "1604 matmul_30" [label="(4, 12, 64, 32)", style=solid]; +"1597 linalg_vector_norm_31" -> "1598 clamp_min_31" [label="(4, 12, 64, 1)", style=solid]; +"1598 clamp_min_31" -> "1599 expand_as_31" [label="(4, 12, 64, 1)", style=solid]; +"1599 expand_as_31" -> "1600 div_31" [label="(4, 12, 64, 32)", style=solid]; +"1600 div_31" -> "1601 quantize_per_tensor_default_95" [label="(4, 12, 64, 32)", style=solid]; +"1601 quantize_per_tensor_default_95" -> "1602 dequantize_per_tensor_default_95" [label="(4, 12, 64, 32)", style=solid]; +"1602 dequantize_per_tensor_default_95" -> "1603 transpose_30" [label="(4, 12, 64, 32)", style=solid]; +"1603 transpose_30" -> "1604 matmul_30" [label="(4, 12, 32, 64)", style=solid]; +"1604 matmul_30" -> "1606 mul_31" [label="(4, 12, 64, 64)", style=solid]; +"1605 _frozen_param126" -> "1606 mul_31" [label="(12, 1, 1)", style=solid]; +"1606 mul_31" -> "1607 add_52" [label="(4, 12, 64, 64)", style=solid]; +"1607 add_52" -> "1619 view_85" [label="(4, 12, 64, 64)", style=solid]; +"1608 new_zeros_7" -> "1609 view_84" [label="(16, 16)", style=solid]; +"1609 view_84" -> "1610 permute_71" [label="(2, 8, 2, 8)", style=solid]; +"1610 permute_71" -> "1611 reshape_69" [label="(2, 2, 8, 8)", style=solid]; +"1611 reshape_69" -> "1612 unsqueeze_44" [label="(4, 64)", style=solid]; +"1611 reshape_69" -> "1613 unsqueeze_45" [label="(4, 64)", style=solid]; +"1612 unsqueeze_44" -> "1614 sub_7" [label="(4, 1, 64)", style=solid]; +"1613 unsqueeze_45" -> "1614 sub_7" [label="(4, 64, 1)", style=solid]; +"1614 sub_7" -> "1615 ne_7" [label="(4, 64, 64)", style=solid]; +"1614 sub_7" -> "1616 masked_fill_14" [label="(4, 64, 64)", style=solid]; +"1614 sub_7" -> "1617 eq_7" [label="(4, 64, 64)", style=solid]; +"1615 ne_7" -> "1616 masked_fill_14" [label="(4, 64, 64)", style=solid]; +"1616 masked_fill_14" -> "1618 masked_fill_15" [label="(4, 64, 64)", style=solid]; +"1617 eq_7" -> "1618 masked_fill_15" [label="(4, 64, 64)", style=solid]; +"1618 masked_fill_15" -> "1620 unsqueeze_46" [label="(4, 64, 64)", style=solid]; +"1619 view_85" -> "1622 add_53" [label="(1, 4, 12, 64, 64)", style=solid]; +"1620 unsqueeze_46" -> "1621 unsqueeze_47" [label="(4, 1, 64, 64)", style=solid]; +"1621 unsqueeze_47" -> "1622 add_53" [label="(1, 4, 1, 64, 64)", style=solid]; +"1622 add_53" -> "1623 view_86" [label="(1, 4, 12, 64, 64)", style=solid]; +"1623 view_86" -> "1624 softmax_15" [label="(4, 12, 64, 64)", style=solid]; +"1624 softmax_15" -> "1625 dropout_60" [label="(4, 12, 64, 64)", style=solid]; +"1625 dropout_60" -> "1626 matmul_31" [label="(4, 12, 64, 64)", style=solid]; +"1626 matmul_31" -> "1627 transpose_31" [label="(4, 12, 64, 32)", style=solid]; +"1627 transpose_31" -> "1628 reshape_70" [label="(4, 64, 12, 32)", style=solid]; +"1628 reshape_70" -> "1629 reshape_70_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1629 reshape_70_0_0_nncf_smooth_quant_0" -> "1630 quantize_per_tensor_default_96" [label="(4, 64, 384)", style=solid]; +"1630 quantize_per_tensor_default_96" -> "1631 dequantize_per_tensor_default_96" [label="(4, 64, 384)", style=solid]; +"1631 dequantize_per_tensor_default_96" -> "1636 linear_95" [label="(4, 64, 384)", style=solid]; +"1632 linear_95_scale_0" -> "1635 dequantize_per_channel_default_64" [label="(384,)", style=solid]; +"1633 linear_95_zero_point_0" -> "1635 dequantize_per_channel_default_64" [label="(384,)", style=solid]; +"1634 _frozen_param100_1" -> "1635 dequantize_per_channel_default_64" [label="(384, 384)", style=solid]; +"1635 dequantize_per_channel_default_64" -> "1636 linear_95" [label="(384, 384)", style=solid]; +"1636 linear_95" -> "1637 dropout_61" [label="(4, 64, 384)", style=solid]; +"1637 dropout_61" -> "1638 view_87" [label="(4, 64, 384)", style=solid]; +"1638 view_87" -> "1639 permute_72" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1639 permute_72" -> "1640 reshape_71" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1640 reshape_71" -> "1641 roll_15" [label="(1, 16, 16, 384)", style=solid]; +"1641 roll_15" -> "1642 slice_241" [label="(1, 16, 16, 384)", style=solid]; +"1642 slice_241" -> "1643 slice_242" [label="(1, 16, 16, 384)", style=solid]; +"1643 slice_242" -> "1644 slice_243" [label="(1, 14, 16, 384)", style=solid]; +"1644 slice_243" -> "1645 slice_244" [label="(1, 14, 14, 384)", style=solid]; +"1645 slice_244" -> "1646 contiguous_29" [label="(1, 14, 14, 384)", style=solid]; +"1646 contiguous_29" -> "1647 layer_norm_33" [label="(1, 14, 14, 384)", style=solid]; +"1647 layer_norm_33" -> "1648 add_54" [label="(1, 14, 14, 384)", style=solid]; +"1648 add_54" -> "1649 add_54_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"1648 add_54" -> "1669 add_55" [label="(1, 14, 14, 384)", style=solid]; +"1649 add_54_0_0_nncf_smooth_quant_0" -> "1650 quantize_per_tensor_default_97" [label="(1, 14, 14, 384)", style=solid]; +"1650 quantize_per_tensor_default_97" -> "1651 dequantize_per_tensor_default_97" [label="(1, 14, 14, 384)", style=solid]; +"1651 dequantize_per_tensor_default_97" -> "1656 linear_96" [label="(1, 14, 14, 384)", style=solid]; +"1652 linear_96_scale_0" -> "1655 dequantize_per_channel_default_65" [label="(1536,)", style=solid]; +"1653 linear_96_zero_point_0" -> "1655 dequantize_per_channel_default_65" [label="(1536,)", style=solid]; +"1654 _frozen_param101_1" -> "1655 dequantize_per_channel_default_65" [label="(1536, 384)", style=solid]; +"1655 dequantize_per_channel_default_65" -> "1656 linear_96" [label="(1536, 384)", style=solid]; +"1656 linear_96" -> "1657 gelu_15" [label="(1, 14, 14, 1536)", style=solid]; +"1657 gelu_15" -> "1658 dropout_62" [label="(1, 14, 14, 1536)", style=solid]; +"1658 dropout_62" -> "1659 dropout_62_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"1659 dropout_62_0_0_nncf_smooth_quant_0" -> "1660 quantize_per_tensor_default_98" [label="(1, 14, 14, 1536)", style=solid]; +"1660 quantize_per_tensor_default_98" -> "1661 dequantize_per_tensor_default_98" [label="(1, 14, 14, 1536)", style=solid]; +"1661 dequantize_per_tensor_default_98" -> "1666 linear_97" [label="(1, 14, 14, 1536)", style=solid]; +"1662 linear_97_scale_0" -> "1665 dequantize_per_channel_default_66" [label="(384,)", style=solid]; +"1663 linear_97_zero_point_0" -> "1665 dequantize_per_channel_default_66" [label="(384,)", style=solid]; +"1664 _frozen_param104_1" -> "1665 dequantize_per_channel_default_66" [label="(384, 1536)", style=solid]; +"1665 dequantize_per_channel_default_66" -> "1666 linear_97" [label="(384, 1536)", style=solid]; +"1666 linear_97" -> "1667 dropout_63" [label="(1, 14, 14, 384)", style=solid]; +"1667 dropout_63" -> "1668 layer_norm_34" [label="(1, 14, 14, 384)", style=solid]; +"1668 layer_norm_34" -> "1669 add_55" [label="(1, 14, 14, 384)", style=solid]; +"1669 add_55" -> "1671 pad_18" [label="(1, 14, 14, 384)", style=solid]; +"1669 add_55" -> "1729 add_57" [label="(1, 14, 14, 384)", style=solid]; +"1670 _frozen_param136" -> "1705 add_56" [label="(1, 12, 64, 64)", style=solid]; +"1671 pad_18" -> "1672 view_90" [label="(1, 16, 16, 384)", style=solid]; +"1672 view_90" -> "1673 permute_74" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1673 permute_74" -> "1674 reshape_72" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1674 reshape_72" -> "1676 reshape_72_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1675 _frozen_param137" -> "1683 linear_100" [label="(1152,)", style=solid]; +"1676 reshape_72_0_0_nncf_smooth_quant_0" -> "1677 quantize_per_tensor_default_99" [label="(4, 64, 384)", style=solid]; +"1677 quantize_per_tensor_default_99" -> "1678 dequantize_per_tensor_default_99" [label="(4, 64, 384)", style=solid]; +"1678 dequantize_per_tensor_default_99" -> "1683 linear_100" [label="(4, 64, 384)", style=solid]; +"1679 linear_100_scale_0" -> "1682 dequantize_per_channel_default_67" [label="(1152,)", style=solid]; +"1680 linear_100_zero_point_0" -> "1682 dequantize_per_channel_default_67" [label="(1152,)", style=solid]; +"1681 _frozen_param108_1" -> "1682 dequantize_per_channel_default_67" [label="(1152, 384)", style=solid]; +"1682 dequantize_per_channel_default_67" -> "1683 linear_100" [label="(1152, 384)", style=solid]; +"1683 linear_100" -> "1684 reshape_73" [label="(4, 64, 1152)", style=solid]; +"1684 reshape_73" -> "1685 permute_75" [label="(4, 64, 3, 12, 32)", style=solid]; +"1685 permute_75" -> "1686 select_48" [label="(3, 4, 12, 64, 32)", style=solid]; +"1685 permute_75" -> "1687 select_49" [label="(3, 4, 12, 64, 32)", style=solid]; +"1685 permute_75" -> "1688 select_50" [label="(3, 4, 12, 64, 32)", style=solid]; +"1686 select_48" -> "1689 linalg_vector_norm_32" [label="(4, 12, 64, 32)", style=solid]; +"1686 select_48" -> "1691 expand_as_32" [label="(4, 12, 64, 32)", style=solid]; +"1686 select_48" -> "1692 div_32" [label="(4, 12, 64, 32)", style=solid]; +"1687 select_49" -> "1695 linalg_vector_norm_33" [label="(4, 12, 64, 32)", style=solid]; +"1687 select_49" -> "1697 expand_as_33" [label="(4, 12, 64, 32)", style=solid]; +"1687 select_49" -> "1698 div_33" [label="(4, 12, 64, 32)", style=solid]; +"1688 select_50" -> "1708 matmul_33" [label="(4, 12, 64, 32)", style=solid]; +"1689 linalg_vector_norm_32" -> "1690 clamp_min_32" [label="(4, 12, 64, 1)", style=solid]; +"1690 clamp_min_32" -> "1691 expand_as_32" [label="(4, 12, 64, 1)", style=solid]; +"1691 expand_as_32" -> "1692 div_32" [label="(4, 12, 64, 32)", style=solid]; +"1692 div_32" -> "1693 quantize_per_tensor_default_100" [label="(4, 12, 64, 32)", style=solid]; +"1693 quantize_per_tensor_default_100" -> "1694 dequantize_per_tensor_default_100" [label="(4, 12, 64, 32)", style=solid]; +"1694 dequantize_per_tensor_default_100" -> "1702 matmul_32" [label="(4, 12, 64, 32)", style=solid]; +"1695 linalg_vector_norm_33" -> "1696 clamp_min_33" [label="(4, 12, 64, 1)", style=solid]; +"1696 clamp_min_33" -> "1697 expand_as_33" [label="(4, 12, 64, 1)", style=solid]; +"1697 expand_as_33" -> "1698 div_33" [label="(4, 12, 64, 32)", style=solid]; +"1698 div_33" -> "1699 quantize_per_tensor_default_101" [label="(4, 12, 64, 32)", style=solid]; +"1699 quantize_per_tensor_default_101" -> "1700 dequantize_per_tensor_default_101" [label="(4, 12, 64, 32)", style=solid]; +"1700 dequantize_per_tensor_default_101" -> "1701 transpose_32" [label="(4, 12, 64, 32)", style=solid]; +"1701 transpose_32" -> "1702 matmul_32" [label="(4, 12, 32, 64)", style=solid]; +"1702 matmul_32" -> "1704 mul_33" [label="(4, 12, 64, 64)", style=solid]; +"1703 _frozen_param139" -> "1704 mul_33" [label="(12, 1, 1)", style=solid]; +"1704 mul_33" -> "1705 add_56" [label="(4, 12, 64, 64)", style=solid]; +"1705 add_56" -> "1706 softmax_16" [label="(4, 12, 64, 64)", style=solid]; +"1706 softmax_16" -> "1707 dropout_64" [label="(4, 12, 64, 64)", style=solid]; +"1707 dropout_64" -> "1708 matmul_33" [label="(4, 12, 64, 64)", style=solid]; +"1708 matmul_33" -> "1709 transpose_33" [label="(4, 12, 64, 32)", style=solid]; +"1709 transpose_33" -> "1710 reshape_74" [label="(4, 64, 12, 32)", style=solid]; +"1710 reshape_74" -> "1711 reshape_74_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1711 reshape_74_0_0_nncf_smooth_quant_0" -> "1712 quantize_per_tensor_default_102" [label="(4, 64, 384)", style=solid]; +"1712 quantize_per_tensor_default_102" -> "1713 dequantize_per_tensor_default_102" [label="(4, 64, 384)", style=solid]; +"1713 dequantize_per_tensor_default_102" -> "1718 linear_101" [label="(4, 64, 384)", style=solid]; +"1714 linear_101_scale_0" -> "1717 dequantize_per_channel_default_68" [label="(384,)", style=solid]; +"1715 linear_101_zero_point_0" -> "1717 dequantize_per_channel_default_68" [label="(384,)", style=solid]; +"1716 _frozen_param110_1" -> "1717 dequantize_per_channel_default_68" [label="(384, 384)", style=solid]; +"1717 dequantize_per_channel_default_68" -> "1718 linear_101" [label="(384, 384)", style=solid]; +"1718 linear_101" -> "1719 dropout_65" [label="(4, 64, 384)", style=solid]; +"1719 dropout_65" -> "1720 view_91" [label="(4, 64, 384)", style=solid]; +"1720 view_91" -> "1721 permute_76" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1721 permute_76" -> "1722 reshape_75" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1722 reshape_75" -> "1723 slice_246" [label="(1, 16, 16, 384)", style=solid]; +"1723 slice_246" -> "1724 slice_247" [label="(1, 16, 16, 384)", style=solid]; +"1724 slice_247" -> "1725 slice_248" [label="(1, 14, 16, 384)", style=solid]; +"1725 slice_248" -> "1726 slice_249" [label="(1, 14, 14, 384)", style=solid]; +"1726 slice_249" -> "1727 contiguous_31" [label="(1, 14, 14, 384)", style=solid]; +"1727 contiguous_31" -> "1728 layer_norm_35" [label="(1, 14, 14, 384)", style=solid]; +"1728 layer_norm_35" -> "1729 add_57" [label="(1, 14, 14, 384)", style=solid]; +"1729 add_57" -> "1730 add_57_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"1729 add_57" -> "1750 add_58" [label="(1, 14, 14, 384)", style=solid]; +"1730 add_57_0_0_nncf_smooth_quant_0" -> "1731 quantize_per_tensor_default_103" [label="(1, 14, 14, 384)", style=solid]; +"1731 quantize_per_tensor_default_103" -> "1732 dequantize_per_tensor_default_103" [label="(1, 14, 14, 384)", style=solid]; +"1732 dequantize_per_tensor_default_103" -> "1737 linear_102" [label="(1, 14, 14, 384)", style=solid]; +"1733 linear_102_scale_0" -> "1736 dequantize_per_channel_default_69" [label="(1536,)", style=solid]; +"1734 linear_102_zero_point_0" -> "1736 dequantize_per_channel_default_69" [label="(1536,)", style=solid]; +"1735 _frozen_param111_1" -> "1736 dequantize_per_channel_default_69" [label="(1536, 384)", style=solid]; +"1736 dequantize_per_channel_default_69" -> "1737 linear_102" [label="(1536, 384)", style=solid]; +"1737 linear_102" -> "1738 gelu_16" [label="(1, 14, 14, 1536)", style=solid]; +"1738 gelu_16" -> "1739 dropout_66" [label="(1, 14, 14, 1536)", style=solid]; +"1739 dropout_66" -> "1740 dropout_66_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"1740 dropout_66_0_0_nncf_smooth_quant_0" -> "1741 quantize_per_tensor_default_104" [label="(1, 14, 14, 1536)", style=solid]; +"1741 quantize_per_tensor_default_104" -> "1742 dequantize_per_tensor_default_104" [label="(1, 14, 14, 1536)", style=solid]; +"1742 dequantize_per_tensor_default_104" -> "1747 linear_103" [label="(1, 14, 14, 1536)", style=solid]; +"1743 linear_103_scale_0" -> "1746 dequantize_per_channel_default_70" [label="(384,)", style=solid]; +"1744 linear_103_zero_point_0" -> "1746 dequantize_per_channel_default_70" [label="(384,)", style=solid]; +"1745 _frozen_param112_1" -> "1746 dequantize_per_channel_default_70" [label="(384, 1536)", style=solid]; +"1746 dequantize_per_channel_default_70" -> "1747 linear_103" [label="(384, 1536)", style=solid]; +"1747 linear_103" -> "1748 dropout_67" [label="(1, 14, 14, 384)", style=solid]; +"1748 dropout_67" -> "1749 layer_norm_36" [label="(1, 14, 14, 384)", style=solid]; +"1749 layer_norm_36" -> "1750 add_58" [label="(1, 14, 14, 384)", style=solid]; +"1750 add_58" -> "1752 pad_19" [label="(1, 14, 14, 384)", style=solid]; +"1750 add_58" -> "1828 add_61" [label="(1, 14, 14, 384)", style=solid]; +"1751 _frozen_param140" -> "1787 add_59" [label="(1, 12, 64, 64)", style=solid]; +"1752 pad_19" -> "1753 roll_16" [label="(1, 16, 16, 384)", style=solid]; +"1753 roll_16" -> "1754 view_94" [label="(1, 16, 16, 384)", style=solid]; +"1754 view_94" -> "1755 permute_78" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1755 permute_78" -> "1756 reshape_76" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1756 reshape_76" -> "1758 reshape_76_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1756 reshape_76" -> "1788 new_zeros_8" [label="(4, 64, 384)", style=solid]; +"1757 _frozen_param141" -> "1765 linear_106" [label="(1152,)", style=solid]; +"1758 reshape_76_0_0_nncf_smooth_quant_0" -> "1759 quantize_per_tensor_default_105" [label="(4, 64, 384)", style=solid]; +"1759 quantize_per_tensor_default_105" -> "1760 dequantize_per_tensor_default_105" [label="(4, 64, 384)", style=solid]; +"1760 dequantize_per_tensor_default_105" -> "1765 linear_106" [label="(4, 64, 384)", style=solid]; +"1761 linear_106_scale_0" -> "1764 dequantize_per_channel_default_71" [label="(1152,)", style=solid]; +"1762 linear_106_zero_point_0" -> "1764 dequantize_per_channel_default_71" [label="(1152,)", style=solid]; +"1763 _frozen_param113_1" -> "1764 dequantize_per_channel_default_71" [label="(1152, 384)", style=solid]; +"1764 dequantize_per_channel_default_71" -> "1765 linear_106" [label="(1152, 384)", style=solid]; +"1765 linear_106" -> "1766 reshape_77" [label="(4, 64, 1152)", style=solid]; +"1766 reshape_77" -> "1767 permute_79" [label="(4, 64, 3, 12, 32)", style=solid]; +"1767 permute_79" -> "1768 select_51" [label="(3, 4, 12, 64, 32)", style=solid]; +"1767 permute_79" -> "1769 select_52" [label="(3, 4, 12, 64, 32)", style=solid]; +"1767 permute_79" -> "1770 select_53" [label="(3, 4, 12, 64, 32)", style=solid]; +"1768 select_51" -> "1771 linalg_vector_norm_34" [label="(4, 12, 64, 32)", style=solid]; +"1768 select_51" -> "1773 expand_as_34" [label="(4, 12, 64, 32)", style=solid]; +"1768 select_51" -> "1774 div_34" [label="(4, 12, 64, 32)", style=solid]; +"1769 select_52" -> "1777 linalg_vector_norm_35" [label="(4, 12, 64, 32)", style=solid]; +"1769 select_52" -> "1779 expand_as_35" [label="(4, 12, 64, 32)", style=solid]; +"1769 select_52" -> "1780 div_35" [label="(4, 12, 64, 32)", style=solid]; +"1770 select_53" -> "1806 matmul_35" [label="(4, 12, 64, 32)", style=solid]; +"1771 linalg_vector_norm_34" -> "1772 clamp_min_34" [label="(4, 12, 64, 1)", style=solid]; +"1772 clamp_min_34" -> "1773 expand_as_34" [label="(4, 12, 64, 1)", style=solid]; +"1773 expand_as_34" -> "1774 div_34" [label="(4, 12, 64, 32)", style=solid]; +"1774 div_34" -> "1775 quantize_per_tensor_default_106" [label="(4, 12, 64, 32)", style=solid]; +"1775 quantize_per_tensor_default_106" -> "1776 dequantize_per_tensor_default_106" [label="(4, 12, 64, 32)", style=solid]; +"1776 dequantize_per_tensor_default_106" -> "1784 matmul_34" [label="(4, 12, 64, 32)", style=solid]; +"1777 linalg_vector_norm_35" -> "1778 clamp_min_35" [label="(4, 12, 64, 1)", style=solid]; +"1778 clamp_min_35" -> "1779 expand_as_35" [label="(4, 12, 64, 1)", style=solid]; +"1779 expand_as_35" -> "1780 div_35" [label="(4, 12, 64, 32)", style=solid]; +"1780 div_35" -> "1781 quantize_per_tensor_default_107" [label="(4, 12, 64, 32)", style=solid]; +"1781 quantize_per_tensor_default_107" -> "1782 dequantize_per_tensor_default_107" [label="(4, 12, 64, 32)", style=solid]; +"1782 dequantize_per_tensor_default_107" -> "1783 transpose_34" [label="(4, 12, 64, 32)", style=solid]; +"1783 transpose_34" -> "1784 matmul_34" [label="(4, 12, 32, 64)", style=solid]; +"1784 matmul_34" -> "1786 mul_35" [label="(4, 12, 64, 64)", style=solid]; +"1785 _frozen_param143" -> "1786 mul_35" [label="(12, 1, 1)", style=solid]; +"1786 mul_35" -> "1787 add_59" [label="(4, 12, 64, 64)", style=solid]; +"1787 add_59" -> "1799 view_96" [label="(4, 12, 64, 64)", style=solid]; +"1788 new_zeros_8" -> "1789 view_95" [label="(16, 16)", style=solid]; +"1789 view_95" -> "1790 permute_80" [label="(2, 8, 2, 8)", style=solid]; +"1790 permute_80" -> "1791 reshape_78" [label="(2, 2, 8, 8)", style=solid]; +"1791 reshape_78" -> "1792 unsqueeze_50" [label="(4, 64)", style=solid]; +"1791 reshape_78" -> "1793 unsqueeze_51" [label="(4, 64)", style=solid]; +"1792 unsqueeze_50" -> "1794 sub_8" [label="(4, 1, 64)", style=solid]; +"1793 unsqueeze_51" -> "1794 sub_8" [label="(4, 64, 1)", style=solid]; +"1794 sub_8" -> "1795 ne_8" [label="(4, 64, 64)", style=solid]; +"1794 sub_8" -> "1796 masked_fill_16" [label="(4, 64, 64)", style=solid]; +"1794 sub_8" -> "1797 eq_8" [label="(4, 64, 64)", style=solid]; +"1795 ne_8" -> "1796 masked_fill_16" [label="(4, 64, 64)", style=solid]; +"1796 masked_fill_16" -> "1798 masked_fill_17" [label="(4, 64, 64)", style=solid]; +"1797 eq_8" -> "1798 masked_fill_17" [label="(4, 64, 64)", style=solid]; +"1798 masked_fill_17" -> "1800 unsqueeze_52" [label="(4, 64, 64)", style=solid]; +"1799 view_96" -> "1802 add_60" [label="(1, 4, 12, 64, 64)", style=solid]; +"1800 unsqueeze_52" -> "1801 unsqueeze_53" [label="(4, 1, 64, 64)", style=solid]; +"1801 unsqueeze_53" -> "1802 add_60" [label="(1, 4, 1, 64, 64)", style=solid]; +"1802 add_60" -> "1803 view_97" [label="(1, 4, 12, 64, 64)", style=solid]; +"1803 view_97" -> "1804 softmax_17" [label="(4, 12, 64, 64)", style=solid]; +"1804 softmax_17" -> "1805 dropout_68" [label="(4, 12, 64, 64)", style=solid]; +"1805 dropout_68" -> "1806 matmul_35" [label="(4, 12, 64, 64)", style=solid]; +"1806 matmul_35" -> "1807 transpose_35" [label="(4, 12, 64, 32)", style=solid]; +"1807 transpose_35" -> "1808 reshape_79" [label="(4, 64, 12, 32)", style=solid]; +"1808 reshape_79" -> "1809 reshape_79_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1809 reshape_79_0_0_nncf_smooth_quant_0" -> "1810 quantize_per_tensor_default_108" [label="(4, 64, 384)", style=solid]; +"1810 quantize_per_tensor_default_108" -> "1811 dequantize_per_tensor_default_108" [label="(4, 64, 384)", style=solid]; +"1811 dequantize_per_tensor_default_108" -> "1816 linear_107" [label="(4, 64, 384)", style=solid]; +"1812 linear_107_scale_0" -> "1815 dequantize_per_channel_default_72" [label="(384,)", style=solid]; +"1813 linear_107_zero_point_0" -> "1815 dequantize_per_channel_default_72" [label="(384,)", style=solid]; +"1814 _frozen_param114_1" -> "1815 dequantize_per_channel_default_72" [label="(384, 384)", style=solid]; +"1815 dequantize_per_channel_default_72" -> "1816 linear_107" [label="(384, 384)", style=solid]; +"1816 linear_107" -> "1817 dropout_69" [label="(4, 64, 384)", style=solid]; +"1817 dropout_69" -> "1818 view_98" [label="(4, 64, 384)", style=solid]; +"1818 view_98" -> "1819 permute_81" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1819 permute_81" -> "1820 reshape_80" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1820 reshape_80" -> "1821 roll_17" [label="(1, 16, 16, 384)", style=solid]; +"1821 roll_17" -> "1822 slice_269" [label="(1, 16, 16, 384)", style=solid]; +"1822 slice_269" -> "1823 slice_270" [label="(1, 16, 16, 384)", style=solid]; +"1823 slice_270" -> "1824 slice_271" [label="(1, 14, 16, 384)", style=solid]; +"1824 slice_271" -> "1825 slice_272" [label="(1, 14, 14, 384)", style=solid]; +"1825 slice_272" -> "1826 contiguous_33" [label="(1, 14, 14, 384)", style=solid]; +"1826 contiguous_33" -> "1827 layer_norm_37" [label="(1, 14, 14, 384)", style=solid]; +"1827 layer_norm_37" -> "1828 add_61" [label="(1, 14, 14, 384)", style=solid]; +"1828 add_61" -> "1829 add_61_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"1828 add_61" -> "1849 add_62" [label="(1, 14, 14, 384)", style=solid]; +"1829 add_61_0_0_nncf_smooth_quant_0" -> "1830 quantize_per_tensor_default_109" [label="(1, 14, 14, 384)", style=solid]; +"1830 quantize_per_tensor_default_109" -> "1831 dequantize_per_tensor_default_109" [label="(1, 14, 14, 384)", style=solid]; +"1831 dequantize_per_tensor_default_109" -> "1836 linear_108" [label="(1, 14, 14, 384)", style=solid]; +"1832 linear_108_scale_0" -> "1835 dequantize_per_channel_default_73" [label="(1536,)", style=solid]; +"1833 linear_108_zero_point_0" -> "1835 dequantize_per_channel_default_73" [label="(1536,)", style=solid]; +"1834 _frozen_param115_1" -> "1835 dequantize_per_channel_default_73" [label="(1536, 384)", style=solid]; +"1835 dequantize_per_channel_default_73" -> "1836 linear_108" [label="(1536, 384)", style=solid]; +"1836 linear_108" -> "1837 gelu_17" [label="(1, 14, 14, 1536)", style=solid]; +"1837 gelu_17" -> "1838 dropout_70" [label="(1, 14, 14, 1536)", style=solid]; +"1838 dropout_70" -> "1839 dropout_70_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"1839 dropout_70_0_0_nncf_smooth_quant_0" -> "1840 quantize_per_tensor_default_110" [label="(1, 14, 14, 1536)", style=solid]; +"1840 quantize_per_tensor_default_110" -> "1841 dequantize_per_tensor_default_110" [label="(1, 14, 14, 1536)", style=solid]; +"1841 dequantize_per_tensor_default_110" -> "1846 linear_109" [label="(1, 14, 14, 1536)", style=solid]; +"1842 linear_109_scale_0" -> "1845 dequantize_per_channel_default_74" [label="(384,)", style=solid]; +"1843 linear_109_zero_point_0" -> "1845 dequantize_per_channel_default_74" [label="(384,)", style=solid]; +"1844 _frozen_param116_1" -> "1845 dequantize_per_channel_default_74" [label="(384, 1536)", style=solid]; +"1845 dequantize_per_channel_default_74" -> "1846 linear_109" [label="(384, 1536)", style=solid]; +"1846 linear_109" -> "1847 dropout_71" [label="(1, 14, 14, 384)", style=solid]; +"1847 dropout_71" -> "1848 layer_norm_38" [label="(1, 14, 14, 384)", style=solid]; +"1848 layer_norm_38" -> "1849 add_62" [label="(1, 14, 14, 384)", style=solid]; +"1849 add_62" -> "1851 pad_20" [label="(1, 14, 14, 384)", style=solid]; +"1849 add_62" -> "1909 add_64" [label="(1, 14, 14, 384)", style=solid]; +"1850 _frozen_param153" -> "1885 add_63" [label="(1, 12, 64, 64)", style=solid]; +"1851 pad_20" -> "1852 view_101" [label="(1, 16, 16, 384)", style=solid]; +"1852 view_101" -> "1853 permute_83" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1853 permute_83" -> "1854 reshape_81" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1854 reshape_81" -> "1856 reshape_81_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1855 _frozen_param154" -> "1863 linear_112" [label="(1152,)", style=solid]; +"1856 reshape_81_0_0_nncf_smooth_quant_0" -> "1857 quantize_per_tensor_default_111" [label="(4, 64, 384)", style=solid]; +"1857 quantize_per_tensor_default_111" -> "1858 dequantize_per_tensor_default_111" [label="(4, 64, 384)", style=solid]; +"1858 dequantize_per_tensor_default_111" -> "1863 linear_112" [label="(4, 64, 384)", style=solid]; +"1859 linear_112_scale_0" -> "1862 dequantize_per_channel_default_75" [label="(1152,)", style=solid]; +"1860 linear_112_zero_point_0" -> "1862 dequantize_per_channel_default_75" [label="(1152,)", style=solid]; +"1861 _frozen_param117_1" -> "1862 dequantize_per_channel_default_75" [label="(1152, 384)", style=solid]; +"1862 dequantize_per_channel_default_75" -> "1863 linear_112" [label="(1152, 384)", style=solid]; +"1863 linear_112" -> "1864 reshape_82" [label="(4, 64, 1152)", style=solid]; +"1864 reshape_82" -> "1865 permute_84" [label="(4, 64, 3, 12, 32)", style=solid]; +"1865 permute_84" -> "1866 select_54" [label="(3, 4, 12, 64, 32)", style=solid]; +"1865 permute_84" -> "1867 select_55" [label="(3, 4, 12, 64, 32)", style=solid]; +"1865 permute_84" -> "1868 select_56" [label="(3, 4, 12, 64, 32)", style=solid]; +"1866 select_54" -> "1869 linalg_vector_norm_36" [label="(4, 12, 64, 32)", style=solid]; +"1866 select_54" -> "1871 expand_as_36" [label="(4, 12, 64, 32)", style=solid]; +"1866 select_54" -> "1872 div_36" [label="(4, 12, 64, 32)", style=solid]; +"1867 select_55" -> "1875 linalg_vector_norm_37" [label="(4, 12, 64, 32)", style=solid]; +"1867 select_55" -> "1877 expand_as_37" [label="(4, 12, 64, 32)", style=solid]; +"1867 select_55" -> "1878 div_37" [label="(4, 12, 64, 32)", style=solid]; +"1868 select_56" -> "1888 matmul_37" [label="(4, 12, 64, 32)", style=solid]; +"1869 linalg_vector_norm_36" -> "1870 clamp_min_36" [label="(4, 12, 64, 1)", style=solid]; +"1870 clamp_min_36" -> "1871 expand_as_36" [label="(4, 12, 64, 1)", style=solid]; +"1871 expand_as_36" -> "1872 div_36" [label="(4, 12, 64, 32)", style=solid]; +"1872 div_36" -> "1873 quantize_per_tensor_default_112" [label="(4, 12, 64, 32)", style=solid]; +"1873 quantize_per_tensor_default_112" -> "1874 dequantize_per_tensor_default_112" [label="(4, 12, 64, 32)", style=solid]; +"1874 dequantize_per_tensor_default_112" -> "1882 matmul_36" [label="(4, 12, 64, 32)", style=solid]; +"1875 linalg_vector_norm_37" -> "1876 clamp_min_37" [label="(4, 12, 64, 1)", style=solid]; +"1876 clamp_min_37" -> "1877 expand_as_37" [label="(4, 12, 64, 1)", style=solid]; +"1877 expand_as_37" -> "1878 div_37" [label="(4, 12, 64, 32)", style=solid]; +"1878 div_37" -> "1879 quantize_per_tensor_default_113" [label="(4, 12, 64, 32)", style=solid]; +"1879 quantize_per_tensor_default_113" -> "1880 dequantize_per_tensor_default_113" [label="(4, 12, 64, 32)", style=solid]; +"1880 dequantize_per_tensor_default_113" -> "1881 transpose_36" [label="(4, 12, 64, 32)", style=solid]; +"1881 transpose_36" -> "1882 matmul_36" [label="(4, 12, 32, 64)", style=solid]; +"1882 matmul_36" -> "1884 mul_37" [label="(4, 12, 64, 64)", style=solid]; +"1883 _frozen_param156" -> "1884 mul_37" [label="(12, 1, 1)", style=solid]; +"1884 mul_37" -> "1885 add_63" [label="(4, 12, 64, 64)", style=solid]; +"1885 add_63" -> "1886 softmax_18" [label="(4, 12, 64, 64)", style=solid]; +"1886 softmax_18" -> "1887 dropout_72" [label="(4, 12, 64, 64)", style=solid]; +"1887 dropout_72" -> "1888 matmul_37" [label="(4, 12, 64, 64)", style=solid]; +"1888 matmul_37" -> "1889 transpose_37" [label="(4, 12, 64, 32)", style=solid]; +"1889 transpose_37" -> "1890 reshape_83" [label="(4, 64, 12, 32)", style=solid]; +"1890 reshape_83" -> "1891 reshape_83_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1891 reshape_83_0_0_nncf_smooth_quant_0" -> "1892 quantize_per_tensor_default_114" [label="(4, 64, 384)", style=solid]; +"1892 quantize_per_tensor_default_114" -> "1893 dequantize_per_tensor_default_114" [label="(4, 64, 384)", style=solid]; +"1893 dequantize_per_tensor_default_114" -> "1898 linear_113" [label="(4, 64, 384)", style=solid]; +"1894 linear_113_scale_0" -> "1897 dequantize_per_channel_default_76" [label="(384,)", style=solid]; +"1895 linear_113_zero_point_0" -> "1897 dequantize_per_channel_default_76" [label="(384,)", style=solid]; +"1896 _frozen_param118_1" -> "1897 dequantize_per_channel_default_76" [label="(384, 384)", style=solid]; +"1897 dequantize_per_channel_default_76" -> "1898 linear_113" [label="(384, 384)", style=solid]; +"1898 linear_113" -> "1899 dropout_73" [label="(4, 64, 384)", style=solid]; +"1899 dropout_73" -> "1900 view_102" [label="(4, 64, 384)", style=solid]; +"1900 view_102" -> "1901 permute_85" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1901 permute_85" -> "1902 reshape_84" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1902 reshape_84" -> "1903 slice_274" [label="(1, 16, 16, 384)", style=solid]; +"1903 slice_274" -> "1904 slice_275" [label="(1, 16, 16, 384)", style=solid]; +"1904 slice_275" -> "1905 slice_276" [label="(1, 14, 16, 384)", style=solid]; +"1905 slice_276" -> "1906 slice_277" [label="(1, 14, 14, 384)", style=solid]; +"1906 slice_277" -> "1907 contiguous_35" [label="(1, 14, 14, 384)", style=solid]; +"1907 contiguous_35" -> "1908 layer_norm_39" [label="(1, 14, 14, 384)", style=solid]; +"1908 layer_norm_39" -> "1909 add_64" [label="(1, 14, 14, 384)", style=solid]; +"1909 add_64" -> "1910 add_64_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"1909 add_64" -> "1930 add_65" [label="(1, 14, 14, 384)", style=solid]; +"1910 add_64_0_0_nncf_smooth_quant_0" -> "1911 quantize_per_tensor_default_115" [label="(1, 14, 14, 384)", style=solid]; +"1911 quantize_per_tensor_default_115" -> "1912 dequantize_per_tensor_default_115" [label="(1, 14, 14, 384)", style=solid]; +"1912 dequantize_per_tensor_default_115" -> "1917 linear_114" [label="(1, 14, 14, 384)", style=solid]; +"1913 linear_114_scale_0" -> "1916 dequantize_per_channel_default_77" [label="(1536,)", style=solid]; +"1914 linear_114_zero_point_0" -> "1916 dequantize_per_channel_default_77" [label="(1536,)", style=solid]; +"1915 _frozen_param121_1" -> "1916 dequantize_per_channel_default_77" [label="(1536, 384)", style=solid]; +"1916 dequantize_per_channel_default_77" -> "1917 linear_114" [label="(1536, 384)", style=solid]; +"1917 linear_114" -> "1918 gelu_18" [label="(1, 14, 14, 1536)", style=solid]; +"1918 gelu_18" -> "1919 dropout_74" [label="(1, 14, 14, 1536)", style=solid]; +"1919 dropout_74" -> "1920 dropout_74_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"1920 dropout_74_0_0_nncf_smooth_quant_0" -> "1921 quantize_per_tensor_default_116" [label="(1, 14, 14, 1536)", style=solid]; +"1921 quantize_per_tensor_default_116" -> "1922 dequantize_per_tensor_default_116" [label="(1, 14, 14, 1536)", style=solid]; +"1922 dequantize_per_tensor_default_116" -> "1927 linear_115" [label="(1, 14, 14, 1536)", style=solid]; +"1923 linear_115_scale_0" -> "1926 dequantize_per_channel_default_78" [label="(384,)", style=solid]; +"1924 linear_115_zero_point_0" -> "1926 dequantize_per_channel_default_78" [label="(384,)", style=solid]; +"1925 _frozen_param125_1" -> "1926 dequantize_per_channel_default_78" [label="(384, 1536)", style=solid]; +"1926 dequantize_per_channel_default_78" -> "1927 linear_115" [label="(384, 1536)", style=solid]; +"1927 linear_115" -> "1928 dropout_75" [label="(1, 14, 14, 384)", style=solid]; +"1928 dropout_75" -> "1929 layer_norm_40" [label="(1, 14, 14, 384)", style=solid]; +"1929 layer_norm_40" -> "1930 add_65" [label="(1, 14, 14, 384)", style=solid]; +"1930 add_65" -> "1932 pad_21" [label="(1, 14, 14, 384)", style=solid]; +"1930 add_65" -> "2008 add_68" [label="(1, 14, 14, 384)", style=solid]; +"1931 _frozen_param157" -> "1967 add_66" [label="(1, 12, 64, 64)", style=solid]; +"1932 pad_21" -> "1933 roll_18" [label="(1, 16, 16, 384)", style=solid]; +"1933 roll_18" -> "1934 view_105" [label="(1, 16, 16, 384)", style=solid]; +"1934 view_105" -> "1935 permute_87" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1935 permute_87" -> "1936 reshape_85" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1936 reshape_85" -> "1938 reshape_85_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1936 reshape_85" -> "1968 new_zeros_9" [label="(4, 64, 384)", style=solid]; +"1937 _frozen_param158" -> "1945 linear_118" [label="(1152,)", style=solid]; +"1938 reshape_85_0_0_nncf_smooth_quant_0" -> "1939 quantize_per_tensor_default_117" [label="(4, 64, 384)", style=solid]; +"1939 quantize_per_tensor_default_117" -> "1940 dequantize_per_tensor_default_117" [label="(4, 64, 384)", style=solid]; +"1940 dequantize_per_tensor_default_117" -> "1945 linear_118" [label="(4, 64, 384)", style=solid]; +"1941 linear_118_scale_0" -> "1944 dequantize_per_channel_default_79" [label="(1152,)", style=solid]; +"1942 linear_118_zero_point_0" -> "1944 dequantize_per_channel_default_79" [label="(1152,)", style=solid]; +"1943 _frozen_param127_1" -> "1944 dequantize_per_channel_default_79" [label="(1152, 384)", style=solid]; +"1944 dequantize_per_channel_default_79" -> "1945 linear_118" [label="(1152, 384)", style=solid]; +"1945 linear_118" -> "1946 reshape_86" [label="(4, 64, 1152)", style=solid]; +"1946 reshape_86" -> "1947 permute_88" [label="(4, 64, 3, 12, 32)", style=solid]; +"1947 permute_88" -> "1948 select_57" [label="(3, 4, 12, 64, 32)", style=solid]; +"1947 permute_88" -> "1949 select_58" [label="(3, 4, 12, 64, 32)", style=solid]; +"1947 permute_88" -> "1950 select_59" [label="(3, 4, 12, 64, 32)", style=solid]; +"1948 select_57" -> "1951 linalg_vector_norm_38" [label="(4, 12, 64, 32)", style=solid]; +"1948 select_57" -> "1953 expand_as_38" [label="(4, 12, 64, 32)", style=solid]; +"1948 select_57" -> "1954 div_38" [label="(4, 12, 64, 32)", style=solid]; +"1949 select_58" -> "1957 linalg_vector_norm_39" [label="(4, 12, 64, 32)", style=solid]; +"1949 select_58" -> "1959 expand_as_39" [label="(4, 12, 64, 32)", style=solid]; +"1949 select_58" -> "1960 div_39" [label="(4, 12, 64, 32)", style=solid]; +"1950 select_59" -> "1986 matmul_39" [label="(4, 12, 64, 32)", style=solid]; +"1951 linalg_vector_norm_38" -> "1952 clamp_min_38" [label="(4, 12, 64, 1)", style=solid]; +"1952 clamp_min_38" -> "1953 expand_as_38" [label="(4, 12, 64, 1)", style=solid]; +"1953 expand_as_38" -> "1954 div_38" [label="(4, 12, 64, 32)", style=solid]; +"1954 div_38" -> "1955 quantize_per_tensor_default_118" [label="(4, 12, 64, 32)", style=solid]; +"1955 quantize_per_tensor_default_118" -> "1956 dequantize_per_tensor_default_118" [label="(4, 12, 64, 32)", style=solid]; +"1956 dequantize_per_tensor_default_118" -> "1964 matmul_38" [label="(4, 12, 64, 32)", style=solid]; +"1957 linalg_vector_norm_39" -> "1958 clamp_min_39" [label="(4, 12, 64, 1)", style=solid]; +"1958 clamp_min_39" -> "1959 expand_as_39" [label="(4, 12, 64, 1)", style=solid]; +"1959 expand_as_39" -> "1960 div_39" [label="(4, 12, 64, 32)", style=solid]; +"1960 div_39" -> "1961 quantize_per_tensor_default_119" [label="(4, 12, 64, 32)", style=solid]; +"1961 quantize_per_tensor_default_119" -> "1962 dequantize_per_tensor_default_119" [label="(4, 12, 64, 32)", style=solid]; +"1962 dequantize_per_tensor_default_119" -> "1963 transpose_38" [label="(4, 12, 64, 32)", style=solid]; +"1963 transpose_38" -> "1964 matmul_38" [label="(4, 12, 32, 64)", style=solid]; +"1964 matmul_38" -> "1966 mul_39" [label="(4, 12, 64, 64)", style=solid]; +"1965 _frozen_param160" -> "1966 mul_39" [label="(12, 1, 1)", style=solid]; +"1966 mul_39" -> "1967 add_66" [label="(4, 12, 64, 64)", style=solid]; +"1967 add_66" -> "1979 view_107" [label="(4, 12, 64, 64)", style=solid]; +"1968 new_zeros_9" -> "1969 view_106" [label="(16, 16)", style=solid]; +"1969 view_106" -> "1970 permute_89" [label="(2, 8, 2, 8)", style=solid]; +"1970 permute_89" -> "1971 reshape_87" [label="(2, 2, 8, 8)", style=solid]; +"1971 reshape_87" -> "1972 unsqueeze_56" [label="(4, 64)", style=solid]; +"1971 reshape_87" -> "1973 unsqueeze_57" [label="(4, 64)", style=solid]; +"1972 unsqueeze_56" -> "1974 sub_9" [label="(4, 1, 64)", style=solid]; +"1973 unsqueeze_57" -> "1974 sub_9" [label="(4, 64, 1)", style=solid]; +"1974 sub_9" -> "1975 ne_9" [label="(4, 64, 64)", style=solid]; +"1974 sub_9" -> "1976 masked_fill_18" [label="(4, 64, 64)", style=solid]; +"1974 sub_9" -> "1977 eq_9" [label="(4, 64, 64)", style=solid]; +"1975 ne_9" -> "1976 masked_fill_18" [label="(4, 64, 64)", style=solid]; +"1976 masked_fill_18" -> "1978 masked_fill_19" [label="(4, 64, 64)", style=solid]; +"1977 eq_9" -> "1978 masked_fill_19" [label="(4, 64, 64)", style=solid]; +"1978 masked_fill_19" -> "1980 unsqueeze_58" [label="(4, 64, 64)", style=solid]; +"1979 view_107" -> "1982 add_67" [label="(1, 4, 12, 64, 64)", style=solid]; +"1980 unsqueeze_58" -> "1981 unsqueeze_59" [label="(4, 1, 64, 64)", style=solid]; +"1981 unsqueeze_59" -> "1982 add_67" [label="(1, 4, 1, 64, 64)", style=solid]; +"1982 add_67" -> "1983 view_108" [label="(1, 4, 12, 64, 64)", style=solid]; +"1983 view_108" -> "1984 softmax_19" [label="(4, 12, 64, 64)", style=solid]; +"1984 softmax_19" -> "1985 dropout_76" [label="(4, 12, 64, 64)", style=solid]; +"1985 dropout_76" -> "1986 matmul_39" [label="(4, 12, 64, 64)", style=solid]; +"1986 matmul_39" -> "1987 transpose_39" [label="(4, 12, 64, 32)", style=solid]; +"1987 transpose_39" -> "1988 reshape_88" [label="(4, 64, 12, 32)", style=solid]; +"1988 reshape_88" -> "1989 reshape_88_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"1989 reshape_88_0_0_nncf_smooth_quant_0" -> "1990 quantize_per_tensor_default_120" [label="(4, 64, 384)", style=solid]; +"1990 quantize_per_tensor_default_120" -> "1991 dequantize_per_tensor_default_120" [label="(4, 64, 384)", style=solid]; +"1991 dequantize_per_tensor_default_120" -> "1996 linear_119" [label="(4, 64, 384)", style=solid]; +"1992 linear_119_scale_0" -> "1995 dequantize_per_channel_default_80" [label="(384,)", style=solid]; +"1993 linear_119_zero_point_0" -> "1995 dequantize_per_channel_default_80" [label="(384,)", style=solid]; +"1994 _frozen_param128_1" -> "1995 dequantize_per_channel_default_80" [label="(384, 384)", style=solid]; +"1995 dequantize_per_channel_default_80" -> "1996 linear_119" [label="(384, 384)", style=solid]; +"1996 linear_119" -> "1997 dropout_77" [label="(4, 64, 384)", style=solid]; +"1997 dropout_77" -> "1998 view_109" [label="(4, 64, 384)", style=solid]; +"1998 view_109" -> "1999 permute_90" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1999 permute_90" -> "2000 reshape_89" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2000 reshape_89" -> "2001 roll_19" [label="(1, 16, 16, 384)", style=solid]; +"2001 roll_19" -> "2002 slice_297" [label="(1, 16, 16, 384)", style=solid]; +"2002 slice_297" -> "2003 slice_298" [label="(1, 16, 16, 384)", style=solid]; +"2003 slice_298" -> "2004 slice_299" [label="(1, 14, 16, 384)", style=solid]; +"2004 slice_299" -> "2005 slice_300" [label="(1, 14, 14, 384)", style=solid]; +"2005 slice_300" -> "2006 contiguous_37" [label="(1, 14, 14, 384)", style=solid]; +"2006 contiguous_37" -> "2007 layer_norm_41" [label="(1, 14, 14, 384)", style=solid]; +"2007 layer_norm_41" -> "2008 add_68" [label="(1, 14, 14, 384)", style=solid]; +"2008 add_68" -> "2009 add_68_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"2008 add_68" -> "2029 add_69" [label="(1, 14, 14, 384)", style=solid]; +"2009 add_68_0_0_nncf_smooth_quant_0" -> "2010 quantize_per_tensor_default_121" [label="(1, 14, 14, 384)", style=solid]; +"2010 quantize_per_tensor_default_121" -> "2011 dequantize_per_tensor_default_121" [label="(1, 14, 14, 384)", style=solid]; +"2011 dequantize_per_tensor_default_121" -> "2016 linear_120" [label="(1, 14, 14, 384)", style=solid]; +"2012 linear_120_scale_0" -> "2015 dequantize_per_channel_default_81" [label="(1536,)", style=solid]; +"2013 linear_120_zero_point_0" -> "2015 dequantize_per_channel_default_81" [label="(1536,)", style=solid]; +"2014 _frozen_param129_1" -> "2015 dequantize_per_channel_default_81" [label="(1536, 384)", style=solid]; +"2015 dequantize_per_channel_default_81" -> "2016 linear_120" [label="(1536, 384)", style=solid]; +"2016 linear_120" -> "2017 gelu_19" [label="(1, 14, 14, 1536)", style=solid]; +"2017 gelu_19" -> "2018 dropout_78" [label="(1, 14, 14, 1536)", style=solid]; +"2018 dropout_78" -> "2019 dropout_78_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"2019 dropout_78_0_0_nncf_smooth_quant_0" -> "2020 quantize_per_tensor_default_122" [label="(1, 14, 14, 1536)", style=solid]; +"2020 quantize_per_tensor_default_122" -> "2021 dequantize_per_tensor_default_122" [label="(1, 14, 14, 1536)", style=solid]; +"2021 dequantize_per_tensor_default_122" -> "2026 linear_121" [label="(1, 14, 14, 1536)", style=solid]; +"2022 linear_121_scale_0" -> "2025 dequantize_per_channel_default_82" [label="(384,)", style=solid]; +"2023 linear_121_zero_point_0" -> "2025 dequantize_per_channel_default_82" [label="(384,)", style=solid]; +"2024 _frozen_param130_1" -> "2025 dequantize_per_channel_default_82" [label="(384, 1536)", style=solid]; +"2025 dequantize_per_channel_default_82" -> "2026 linear_121" [label="(384, 1536)", style=solid]; +"2026 linear_121" -> "2027 dropout_79" [label="(1, 14, 14, 384)", style=solid]; +"2027 dropout_79" -> "2028 layer_norm_42" [label="(1, 14, 14, 384)", style=solid]; +"2028 layer_norm_42" -> "2029 add_69" [label="(1, 14, 14, 384)", style=solid]; +"2029 add_69" -> "2031 pad_22" [label="(1, 14, 14, 384)", style=solid]; +"2029 add_69" -> "2089 add_71" [label="(1, 14, 14, 384)", style=solid]; +"2030 _frozen_param170" -> "2065 add_70" [label="(1, 12, 64, 64)", style=solid]; +"2031 pad_22" -> "2032 view_112" [label="(1, 16, 16, 384)", style=solid]; +"2032 view_112" -> "2033 permute_92" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2033 permute_92" -> "2034 reshape_90" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2034 reshape_90" -> "2036 reshape_90_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"2035 _frozen_param171" -> "2043 linear_124" [label="(1152,)", style=solid]; +"2036 reshape_90_0_0_nncf_smooth_quant_0" -> "2037 quantize_per_tensor_default_123" [label="(4, 64, 384)", style=solid]; +"2037 quantize_per_tensor_default_123" -> "2038 dequantize_per_tensor_default_123" [label="(4, 64, 384)", style=solid]; +"2038 dequantize_per_tensor_default_123" -> "2043 linear_124" [label="(4, 64, 384)", style=solid]; +"2039 linear_124_scale_0" -> "2042 dequantize_per_channel_default_83" [label="(1152,)", style=solid]; +"2040 linear_124_zero_point_0" -> "2042 dequantize_per_channel_default_83" [label="(1152,)", style=solid]; +"2041 _frozen_param131_1" -> "2042 dequantize_per_channel_default_83" [label="(1152, 384)", style=solid]; +"2042 dequantize_per_channel_default_83" -> "2043 linear_124" [label="(1152, 384)", style=solid]; +"2043 linear_124" -> "2044 reshape_91" [label="(4, 64, 1152)", style=solid]; +"2044 reshape_91" -> "2045 permute_93" [label="(4, 64, 3, 12, 32)", style=solid]; +"2045 permute_93" -> "2046 select_60" [label="(3, 4, 12, 64, 32)", style=solid]; +"2045 permute_93" -> "2047 select_61" [label="(3, 4, 12, 64, 32)", style=solid]; +"2045 permute_93" -> "2048 select_62" [label="(3, 4, 12, 64, 32)", style=solid]; +"2046 select_60" -> "2049 linalg_vector_norm_40" [label="(4, 12, 64, 32)", style=solid]; +"2046 select_60" -> "2051 expand_as_40" [label="(4, 12, 64, 32)", style=solid]; +"2046 select_60" -> "2052 div_40" [label="(4, 12, 64, 32)", style=solid]; +"2047 select_61" -> "2055 linalg_vector_norm_41" [label="(4, 12, 64, 32)", style=solid]; +"2047 select_61" -> "2057 expand_as_41" [label="(4, 12, 64, 32)", style=solid]; +"2047 select_61" -> "2058 div_41" [label="(4, 12, 64, 32)", style=solid]; +"2048 select_62" -> "2068 matmul_41" [label="(4, 12, 64, 32)", style=solid]; +"2049 linalg_vector_norm_40" -> "2050 clamp_min_40" [label="(4, 12, 64, 1)", style=solid]; +"2050 clamp_min_40" -> "2051 expand_as_40" [label="(4, 12, 64, 1)", style=solid]; +"2051 expand_as_40" -> "2052 div_40" [label="(4, 12, 64, 32)", style=solid]; +"2052 div_40" -> "2053 quantize_per_tensor_default_124" [label="(4, 12, 64, 32)", style=solid]; +"2053 quantize_per_tensor_default_124" -> "2054 dequantize_per_tensor_default_124" [label="(4, 12, 64, 32)", style=solid]; +"2054 dequantize_per_tensor_default_124" -> "2062 matmul_40" [label="(4, 12, 64, 32)", style=solid]; +"2055 linalg_vector_norm_41" -> "2056 clamp_min_41" [label="(4, 12, 64, 1)", style=solid]; +"2056 clamp_min_41" -> "2057 expand_as_41" [label="(4, 12, 64, 1)", style=solid]; +"2057 expand_as_41" -> "2058 div_41" [label="(4, 12, 64, 32)", style=solid]; +"2058 div_41" -> "2059 quantize_per_tensor_default_125" [label="(4, 12, 64, 32)", style=solid]; +"2059 quantize_per_tensor_default_125" -> "2060 dequantize_per_tensor_default_125" [label="(4, 12, 64, 32)", style=solid]; +"2060 dequantize_per_tensor_default_125" -> "2061 transpose_40" [label="(4, 12, 64, 32)", style=solid]; +"2061 transpose_40" -> "2062 matmul_40" [label="(4, 12, 32, 64)", style=solid]; +"2062 matmul_40" -> "2064 mul_41" [label="(4, 12, 64, 64)", style=solid]; +"2063 _frozen_param173" -> "2064 mul_41" [label="(12, 1, 1)", style=solid]; +"2064 mul_41" -> "2065 add_70" [label="(4, 12, 64, 64)", style=solid]; +"2065 add_70" -> "2066 softmax_20" [label="(4, 12, 64, 64)", style=solid]; +"2066 softmax_20" -> "2067 dropout_80" [label="(4, 12, 64, 64)", style=solid]; +"2067 dropout_80" -> "2068 matmul_41" [label="(4, 12, 64, 64)", style=solid]; +"2068 matmul_41" -> "2069 transpose_41" [label="(4, 12, 64, 32)", style=solid]; +"2069 transpose_41" -> "2070 reshape_92" [label="(4, 64, 12, 32)", style=solid]; +"2070 reshape_92" -> "2071 reshape_92_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"2071 reshape_92_0_0_nncf_smooth_quant_0" -> "2072 quantize_per_tensor_default_126" [label="(4, 64, 384)", style=solid]; +"2072 quantize_per_tensor_default_126" -> "2073 dequantize_per_tensor_default_126" [label="(4, 64, 384)", style=solid]; +"2073 dequantize_per_tensor_default_126" -> "2078 linear_125" [label="(4, 64, 384)", style=solid]; +"2074 linear_125_scale_0" -> "2077 dequantize_per_channel_default_84" [label="(384,)", style=solid]; +"2075 linear_125_zero_point_0" -> "2077 dequantize_per_channel_default_84" [label="(384,)", style=solid]; +"2076 _frozen_param132_1" -> "2077 dequantize_per_channel_default_84" [label="(384, 384)", style=solid]; +"2077 dequantize_per_channel_default_84" -> "2078 linear_125" [label="(384, 384)", style=solid]; +"2078 linear_125" -> "2079 dropout_81" [label="(4, 64, 384)", style=solid]; +"2079 dropout_81" -> "2080 view_113" [label="(4, 64, 384)", style=solid]; +"2080 view_113" -> "2081 permute_94" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2081 permute_94" -> "2082 reshape_93" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2082 reshape_93" -> "2083 slice_302" [label="(1, 16, 16, 384)", style=solid]; +"2083 slice_302" -> "2084 slice_303" [label="(1, 16, 16, 384)", style=solid]; +"2084 slice_303" -> "2085 slice_304" [label="(1, 14, 16, 384)", style=solid]; +"2085 slice_304" -> "2086 slice_305" [label="(1, 14, 14, 384)", style=solid]; +"2086 slice_305" -> "2087 contiguous_39" [label="(1, 14, 14, 384)", style=solid]; +"2087 contiguous_39" -> "2088 layer_norm_43" [label="(1, 14, 14, 384)", style=solid]; +"2088 layer_norm_43" -> "2089 add_71" [label="(1, 14, 14, 384)", style=solid]; +"2089 add_71" -> "2090 add_71_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"2089 add_71" -> "2110 add_72" [label="(1, 14, 14, 384)", style=solid]; +"2090 add_71_0_0_nncf_smooth_quant_0" -> "2091 quantize_per_tensor_default_127" [label="(1, 14, 14, 384)", style=solid]; +"2091 quantize_per_tensor_default_127" -> "2092 dequantize_per_tensor_default_127" [label="(1, 14, 14, 384)", style=solid]; +"2092 dequantize_per_tensor_default_127" -> "2097 linear_126" [label="(1, 14, 14, 384)", style=solid]; +"2093 linear_126_scale_0" -> "2096 dequantize_per_channel_default_85" [label="(1536,)", style=solid]; +"2094 linear_126_zero_point_0" -> "2096 dequantize_per_channel_default_85" [label="(1536,)", style=solid]; +"2095 _frozen_param133_1" -> "2096 dequantize_per_channel_default_85" [label="(1536, 384)", style=solid]; +"2096 dequantize_per_channel_default_85" -> "2097 linear_126" [label="(1536, 384)", style=solid]; +"2097 linear_126" -> "2098 gelu_20" [label="(1, 14, 14, 1536)", style=solid]; +"2098 gelu_20" -> "2099 dropout_82" [label="(1, 14, 14, 1536)", style=solid]; +"2099 dropout_82" -> "2100 dropout_82_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"2100 dropout_82_0_0_nncf_smooth_quant_0" -> "2101 quantize_per_tensor_default_128" [label="(1, 14, 14, 1536)", style=solid]; +"2101 quantize_per_tensor_default_128" -> "2102 dequantize_per_tensor_default_128" [label="(1, 14, 14, 1536)", style=solid]; +"2102 dequantize_per_tensor_default_128" -> "2107 linear_127" [label="(1, 14, 14, 1536)", style=solid]; +"2103 linear_127_scale_0" -> "2106 dequantize_per_channel_default_86" [label="(384,)", style=solid]; +"2104 linear_127_zero_point_0" -> "2106 dequantize_per_channel_default_86" [label="(384,)", style=solid]; +"2105 _frozen_param134_1" -> "2106 dequantize_per_channel_default_86" [label="(384, 1536)", style=solid]; +"2106 dequantize_per_channel_default_86" -> "2107 linear_127" [label="(384, 1536)", style=solid]; +"2107 linear_127" -> "2108 dropout_83" [label="(1, 14, 14, 384)", style=solid]; +"2108 dropout_83" -> "2109 layer_norm_44" [label="(1, 14, 14, 384)", style=solid]; +"2109 layer_norm_44" -> "2110 add_72" [label="(1, 14, 14, 384)", style=solid]; +"2110 add_72" -> "2112 pad_23" [label="(1, 14, 14, 384)", style=solid]; +"2110 add_72" -> "2188 add_75" [label="(1, 14, 14, 384)", style=solid]; +"2111 _frozen_param174" -> "2147 add_73" [label="(1, 12, 64, 64)", style=solid]; +"2112 pad_23" -> "2113 roll_20" [label="(1, 16, 16, 384)", style=solid]; +"2113 roll_20" -> "2114 view_116" [label="(1, 16, 16, 384)", style=solid]; +"2114 view_116" -> "2115 permute_96" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2115 permute_96" -> "2116 reshape_94" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2116 reshape_94" -> "2118 reshape_94_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"2116 reshape_94" -> "2148 new_zeros_10" [label="(4, 64, 384)", style=solid]; +"2117 _frozen_param175" -> "2125 linear_130" [label="(1152,)", style=solid]; +"2118 reshape_94_0_0_nncf_smooth_quant_0" -> "2119 quantize_per_tensor_default_129" [label="(4, 64, 384)", style=solid]; +"2119 quantize_per_tensor_default_129" -> "2120 dequantize_per_tensor_default_129" [label="(4, 64, 384)", style=solid]; +"2120 dequantize_per_tensor_default_129" -> "2125 linear_130" [label="(4, 64, 384)", style=solid]; +"2121 linear_130_scale_0" -> "2124 dequantize_per_channel_default_87" [label="(1152,)", style=solid]; +"2122 linear_130_zero_point_0" -> "2124 dequantize_per_channel_default_87" [label="(1152,)", style=solid]; +"2123 _frozen_param135_1" -> "2124 dequantize_per_channel_default_87" [label="(1152, 384)", style=solid]; +"2124 dequantize_per_channel_default_87" -> "2125 linear_130" [label="(1152, 384)", style=solid]; +"2125 linear_130" -> "2126 reshape_95" [label="(4, 64, 1152)", style=solid]; +"2126 reshape_95" -> "2127 permute_97" [label="(4, 64, 3, 12, 32)", style=solid]; +"2127 permute_97" -> "2128 select_63" [label="(3, 4, 12, 64, 32)", style=solid]; +"2127 permute_97" -> "2129 select_64" [label="(3, 4, 12, 64, 32)", style=solid]; +"2127 permute_97" -> "2130 select_65" [label="(3, 4, 12, 64, 32)", style=solid]; +"2128 select_63" -> "2131 linalg_vector_norm_42" [label="(4, 12, 64, 32)", style=solid]; +"2128 select_63" -> "2133 expand_as_42" [label="(4, 12, 64, 32)", style=solid]; +"2128 select_63" -> "2134 div_42" [label="(4, 12, 64, 32)", style=solid]; +"2129 select_64" -> "2137 linalg_vector_norm_43" [label="(4, 12, 64, 32)", style=solid]; +"2129 select_64" -> "2139 expand_as_43" [label="(4, 12, 64, 32)", style=solid]; +"2129 select_64" -> "2140 div_43" [label="(4, 12, 64, 32)", style=solid]; +"2130 select_65" -> "2166 matmul_43" [label="(4, 12, 64, 32)", style=solid]; +"2131 linalg_vector_norm_42" -> "2132 clamp_min_42" [label="(4, 12, 64, 1)", style=solid]; +"2132 clamp_min_42" -> "2133 expand_as_42" [label="(4, 12, 64, 1)", style=solid]; +"2133 expand_as_42" -> "2134 div_42" [label="(4, 12, 64, 32)", style=solid]; +"2134 div_42" -> "2135 quantize_per_tensor_default_130" [label="(4, 12, 64, 32)", style=solid]; +"2135 quantize_per_tensor_default_130" -> "2136 dequantize_per_tensor_default_130" [label="(4, 12, 64, 32)", style=solid]; +"2136 dequantize_per_tensor_default_130" -> "2144 matmul_42" [label="(4, 12, 64, 32)", style=solid]; +"2137 linalg_vector_norm_43" -> "2138 clamp_min_43" [label="(4, 12, 64, 1)", style=solid]; +"2138 clamp_min_43" -> "2139 expand_as_43" [label="(4, 12, 64, 1)", style=solid]; +"2139 expand_as_43" -> "2140 div_43" [label="(4, 12, 64, 32)", style=solid]; +"2140 div_43" -> "2141 quantize_per_tensor_default_131" [label="(4, 12, 64, 32)", style=solid]; +"2141 quantize_per_tensor_default_131" -> "2142 dequantize_per_tensor_default_131" [label="(4, 12, 64, 32)", style=solid]; +"2142 dequantize_per_tensor_default_131" -> "2143 transpose_42" [label="(4, 12, 64, 32)", style=solid]; +"2143 transpose_42" -> "2144 matmul_42" [label="(4, 12, 32, 64)", style=solid]; +"2144 matmul_42" -> "2146 mul_43" [label="(4, 12, 64, 64)", style=solid]; +"2145 _frozen_param177" -> "2146 mul_43" [label="(12, 1, 1)", style=solid]; +"2146 mul_43" -> "2147 add_73" [label="(4, 12, 64, 64)", style=solid]; +"2147 add_73" -> "2159 view_118" [label="(4, 12, 64, 64)", style=solid]; +"2148 new_zeros_10" -> "2149 view_117" [label="(16, 16)", style=solid]; +"2149 view_117" -> "2150 permute_98" [label="(2, 8, 2, 8)", style=solid]; +"2150 permute_98" -> "2151 reshape_96" [label="(2, 2, 8, 8)", style=solid]; +"2151 reshape_96" -> "2152 unsqueeze_62" [label="(4, 64)", style=solid]; +"2151 reshape_96" -> "2153 unsqueeze_63" [label="(4, 64)", style=solid]; +"2152 unsqueeze_62" -> "2154 sub_10" [label="(4, 1, 64)", style=solid]; +"2153 unsqueeze_63" -> "2154 sub_10" [label="(4, 64, 1)", style=solid]; +"2154 sub_10" -> "2155 ne_10" [label="(4, 64, 64)", style=solid]; +"2154 sub_10" -> "2156 masked_fill_20" [label="(4, 64, 64)", style=solid]; +"2154 sub_10" -> "2157 eq_10" [label="(4, 64, 64)", style=solid]; +"2155 ne_10" -> "2156 masked_fill_20" [label="(4, 64, 64)", style=solid]; +"2156 masked_fill_20" -> "2158 masked_fill_21" [label="(4, 64, 64)", style=solid]; +"2157 eq_10" -> "2158 masked_fill_21" [label="(4, 64, 64)", style=solid]; +"2158 masked_fill_21" -> "2160 unsqueeze_64" [label="(4, 64, 64)", style=solid]; +"2159 view_118" -> "2162 add_74" [label="(1, 4, 12, 64, 64)", style=solid]; +"2160 unsqueeze_64" -> "2161 unsqueeze_65" [label="(4, 1, 64, 64)", style=solid]; +"2161 unsqueeze_65" -> "2162 add_74" [label="(1, 4, 1, 64, 64)", style=solid]; +"2162 add_74" -> "2163 view_119" [label="(1, 4, 12, 64, 64)", style=solid]; +"2163 view_119" -> "2164 softmax_21" [label="(4, 12, 64, 64)", style=solid]; +"2164 softmax_21" -> "2165 dropout_84" [label="(4, 12, 64, 64)", style=solid]; +"2165 dropout_84" -> "2166 matmul_43" [label="(4, 12, 64, 64)", style=solid]; +"2166 matmul_43" -> "2167 transpose_43" [label="(4, 12, 64, 32)", style=solid]; +"2167 transpose_43" -> "2168 reshape_97" [label="(4, 64, 12, 32)", style=solid]; +"2168 reshape_97" -> "2169 reshape_97_0_0_nncf_smooth_quant_0" [label="(4, 64, 384)", style=solid]; +"2169 reshape_97_0_0_nncf_smooth_quant_0" -> "2170 quantize_per_tensor_default_132" [label="(4, 64, 384)", style=solid]; +"2170 quantize_per_tensor_default_132" -> "2171 dequantize_per_tensor_default_132" [label="(4, 64, 384)", style=solid]; +"2171 dequantize_per_tensor_default_132" -> "2176 linear_131" [label="(4, 64, 384)", style=solid]; +"2172 linear_131_scale_0" -> "2175 dequantize_per_channel_default_88" [label="(384,)", style=solid]; +"2173 linear_131_zero_point_0" -> "2175 dequantize_per_channel_default_88" [label="(384,)", style=solid]; +"2174 _frozen_param138_1" -> "2175 dequantize_per_channel_default_88" [label="(384, 384)", style=solid]; +"2175 dequantize_per_channel_default_88" -> "2176 linear_131" [label="(384, 384)", style=solid]; +"2176 linear_131" -> "2177 dropout_85" [label="(4, 64, 384)", style=solid]; +"2177 dropout_85" -> "2178 view_120" [label="(4, 64, 384)", style=solid]; +"2178 view_120" -> "2179 permute_99" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2179 permute_99" -> "2180 reshape_98" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2180 reshape_98" -> "2181 roll_21" [label="(1, 16, 16, 384)", style=solid]; +"2181 roll_21" -> "2182 slice_325" [label="(1, 16, 16, 384)", style=solid]; +"2182 slice_325" -> "2183 slice_326" [label="(1, 16, 16, 384)", style=solid]; +"2183 slice_326" -> "2184 slice_327" [label="(1, 14, 16, 384)", style=solid]; +"2184 slice_327" -> "2185 slice_328" [label="(1, 14, 14, 384)", style=solid]; +"2185 slice_328" -> "2186 contiguous_41" [label="(1, 14, 14, 384)", style=solid]; +"2186 contiguous_41" -> "2187 layer_norm_45" [label="(1, 14, 14, 384)", style=solid]; +"2187 layer_norm_45" -> "2188 add_75" [label="(1, 14, 14, 384)", style=solid]; +"2188 add_75" -> "2189 add_75_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 384)", style=solid]; +"2188 add_75" -> "2209 add_76" [label="(1, 14, 14, 384)", style=solid]; +"2189 add_75_0_0_nncf_smooth_quant_0" -> "2190 quantize_per_tensor_default_133" [label="(1, 14, 14, 384)", style=solid]; +"2190 quantize_per_tensor_default_133" -> "2191 dequantize_per_tensor_default_133" [label="(1, 14, 14, 384)", style=solid]; +"2191 dequantize_per_tensor_default_133" -> "2196 linear_132" [label="(1, 14, 14, 384)", style=solid]; +"2192 linear_132_scale_0" -> "2195 dequantize_per_channel_default_89" [label="(1536,)", style=solid]; +"2193 linear_132_zero_point_0" -> "2195 dequantize_per_channel_default_89" [label="(1536,)", style=solid]; +"2194 _frozen_param142_1" -> "2195 dequantize_per_channel_default_89" [label="(1536, 384)", style=solid]; +"2195 dequantize_per_channel_default_89" -> "2196 linear_132" [label="(1536, 384)", style=solid]; +"2196 linear_132" -> "2197 gelu_21" [label="(1, 14, 14, 1536)", style=solid]; +"2197 gelu_21" -> "2198 dropout_86" [label="(1, 14, 14, 1536)", style=solid]; +"2198 dropout_86" -> "2199 dropout_86_0_0_nncf_smooth_quant_0" [label="(1, 14, 14, 1536)", style=solid]; +"2199 dropout_86_0_0_nncf_smooth_quant_0" -> "2200 quantize_per_tensor_default_134" [label="(1, 14, 14, 1536)", style=solid]; +"2200 quantize_per_tensor_default_134" -> "2201 dequantize_per_tensor_default_134" [label="(1, 14, 14, 1536)", style=solid]; +"2201 dequantize_per_tensor_default_134" -> "2206 linear_133" [label="(1, 14, 14, 1536)", style=solid]; +"2202 linear_133_scale_0" -> "2205 dequantize_per_channel_default_90" [label="(384,)", style=solid]; +"2203 linear_133_zero_point_0" -> "2205 dequantize_per_channel_default_90" [label="(384,)", style=solid]; +"2204 _frozen_param144_1" -> "2205 dequantize_per_channel_default_90" [label="(384, 1536)", style=solid]; +"2205 dequantize_per_channel_default_90" -> "2206 linear_133" [label="(384, 1536)", style=solid]; +"2206 linear_133" -> "2207 dropout_87" [label="(1, 14, 14, 384)", style=solid]; +"2207 dropout_87" -> "2208 layer_norm_46" [label="(1, 14, 14, 384)", style=solid]; +"2208 layer_norm_46" -> "2209 add_76" [label="(1, 14, 14, 384)", style=solid]; +"2209 add_76" -> "2210 pad_24" [label="(1, 14, 14, 384)", style=solid]; +"2210 pad_24" -> "2211 slice_329" [label="(1, 14, 14, 384)", style=solid]; +"2210 pad_24" -> "2214 slice_332" [label="(1, 14, 14, 384)", style=solid]; +"2210 pad_24" -> "2217 slice_335" [label="(1, 14, 14, 384)", style=solid]; +"2210 pad_24" -> "2220 slice_338" [label="(1, 14, 14, 384)", style=solid]; +"2211 slice_329" -> "2212 slice_330" [label="(1, 7, 14, 384)", style=solid]; +"2212 slice_330" -> "2213 slice_331" [label="(1, 7, 7, 384)", style=solid]; +"2213 slice_331" -> "2223 cat_2" [label="(1, 7, 7, 384)", style=solid]; +"2214 slice_332" -> "2215 slice_333" [label="(1, 7, 14, 384)", style=solid]; +"2215 slice_333" -> "2216 slice_334" [label="(1, 7, 7, 384)", style=solid]; +"2216 slice_334" -> "2223 cat_2" [label="(1, 7, 7, 384)", style=solid]; +"2217 slice_335" -> "2218 slice_336" [label="(1, 7, 14, 384)", style=solid]; +"2218 slice_336" -> "2219 slice_337" [label="(1, 7, 7, 384)", style=solid]; +"2219 slice_337" -> "2223 cat_2" [label="(1, 7, 7, 384)", style=solid]; +"2220 slice_338" -> "2221 slice_339" [label="(1, 7, 14, 384)", style=solid]; +"2221 slice_339" -> "2222 slice_340" [label="(1, 7, 7, 384)", style=solid]; +"2222 slice_340" -> "2223 cat_2" [label="(1, 7, 7, 384)", style=solid]; +"2223 cat_2" -> "2224 cat_2_0_0_nncf_smooth_quant_0" [label="(1, 7, 7, 1536)", style=solid]; +"2224 cat_2_0_0_nncf_smooth_quant_0" -> "2225 quantize_per_tensor_default_135" [label="(1, 7, 7, 1536)", style=solid]; +"2225 quantize_per_tensor_default_135" -> "2226 dequantize_per_tensor_default_135" [label="(1, 7, 7, 1536)", style=solid]; +"2226 dequantize_per_tensor_default_135" -> "2231 linear_134" [label="(1, 7, 7, 1536)", style=solid]; +"2227 linear_134_scale_0" -> "2230 dequantize_per_channel_default_91" [label="(768,)", style=solid]; +"2228 linear_134_zero_point_0" -> "2230 dequantize_per_channel_default_91" [label="(768,)", style=solid]; +"2229 _frozen_param145_1" -> "2230 dequantize_per_channel_default_91" [label="(768, 1536)", style=solid]; +"2230 dequantize_per_channel_default_91" -> "2231 linear_134" [label="(768, 1536)", style=solid]; +"2231 linear_134" -> "2232 layer_norm_47" [label="(1, 7, 7, 768)", style=solid]; +"2232 layer_norm_47" -> "2234 pad_25" [label="(1, 7, 7, 768)", style=solid]; +"2232 layer_norm_47" -> "2292 add_78" [label="(1, 7, 7, 768)", style=solid]; +"2233 _frozen_param187" -> "2268 add_77" [label="(1, 24, 64, 64)", style=solid]; +"2234 pad_25" -> "2235 view_123" [label="(1, 8, 8, 768)", style=solid]; +"2235 view_123" -> "2236 permute_101" [label="(1, 1, 8, 1, 8, 768)", style=solid]; +"2236 permute_101" -> "2237 reshape_99" [label="(1, 1, 1, 8, 8, 768)", style=solid]; +"2237 reshape_99" -> "2239 reshape_99_0_0_nncf_smooth_quant_0" [label="(1, 64, 768)", style=solid]; +"2238 _frozen_param188" -> "2246 linear_137" [label="(2304,)", style=solid]; +"2239 reshape_99_0_0_nncf_smooth_quant_0" -> "2240 quantize_per_tensor_default_136" [label="(1, 64, 768)", style=solid]; +"2240 quantize_per_tensor_default_136" -> "2241 dequantize_per_tensor_default_136" [label="(1, 64, 768)", style=solid]; +"2241 dequantize_per_tensor_default_136" -> "2246 linear_137" [label="(1, 64, 768)", style=solid]; +"2242 linear_137_scale_0" -> "2245 dequantize_per_channel_default_92" [label="(2304,)", style=solid]; +"2243 linear_137_zero_point_0" -> "2245 dequantize_per_channel_default_92" [label="(2304,)", style=solid]; +"2244 _frozen_param146_1" -> "2245 dequantize_per_channel_default_92" [label="(2304, 768)", style=solid]; +"2245 dequantize_per_channel_default_92" -> "2246 linear_137" [label="(2304, 768)", style=solid]; +"2246 linear_137" -> "2247 reshape_100" [label="(1, 64, 2304)", style=solid]; +"2247 reshape_100" -> "2248 permute_102" [label="(1, 64, 3, 24, 32)", style=solid]; +"2248 permute_102" -> "2249 select_66" [label="(3, 1, 24, 64, 32)", style=solid]; +"2248 permute_102" -> "2250 select_67" [label="(3, 1, 24, 64, 32)", style=solid]; +"2248 permute_102" -> "2251 select_68" [label="(3, 1, 24, 64, 32)", style=solid]; +"2249 select_66" -> "2252 linalg_vector_norm_44" [label="(1, 24, 64, 32)", style=solid]; +"2249 select_66" -> "2254 expand_as_44" [label="(1, 24, 64, 32)", style=solid]; +"2249 select_66" -> "2255 div_44" [label="(1, 24, 64, 32)", style=solid]; +"2250 select_67" -> "2258 linalg_vector_norm_45" [label="(1, 24, 64, 32)", style=solid]; +"2250 select_67" -> "2260 expand_as_45" [label="(1, 24, 64, 32)", style=solid]; +"2250 select_67" -> "2261 div_45" [label="(1, 24, 64, 32)", style=solid]; +"2251 select_68" -> "2271 matmul_45" [label="(1, 24, 64, 32)", style=solid]; +"2252 linalg_vector_norm_44" -> "2253 clamp_min_44" [label="(1, 24, 64, 1)", style=solid]; +"2253 clamp_min_44" -> "2254 expand_as_44" [label="(1, 24, 64, 1)", style=solid]; +"2254 expand_as_44" -> "2255 div_44" [label="(1, 24, 64, 32)", style=solid]; +"2255 div_44" -> "2256 quantize_per_tensor_default_137" [label="(1, 24, 64, 32)", style=solid]; +"2256 quantize_per_tensor_default_137" -> "2257 dequantize_per_tensor_default_137" [label="(1, 24, 64, 32)", style=solid]; +"2257 dequantize_per_tensor_default_137" -> "2265 matmul_44" [label="(1, 24, 64, 32)", style=solid]; +"2258 linalg_vector_norm_45" -> "2259 clamp_min_45" [label="(1, 24, 64, 1)", style=solid]; +"2259 clamp_min_45" -> "2260 expand_as_45" [label="(1, 24, 64, 1)", style=solid]; +"2260 expand_as_45" -> "2261 div_45" [label="(1, 24, 64, 32)", style=solid]; +"2261 div_45" -> "2262 quantize_per_tensor_default_138" [label="(1, 24, 64, 32)", style=solid]; +"2262 quantize_per_tensor_default_138" -> "2263 dequantize_per_tensor_default_138" [label="(1, 24, 64, 32)", style=solid]; +"2263 dequantize_per_tensor_default_138" -> "2264 transpose_44" [label="(1, 24, 64, 32)", style=solid]; +"2264 transpose_44" -> "2265 matmul_44" [label="(1, 24, 32, 64)", style=solid]; +"2265 matmul_44" -> "2267 mul_45" [label="(1, 24, 64, 64)", style=solid]; +"2266 _frozen_param190" -> "2267 mul_45" [label="(24, 1, 1)", style=solid]; +"2267 mul_45" -> "2268 add_77" [label="(1, 24, 64, 64)", style=solid]; +"2268 add_77" -> "2269 softmax_22" [label="(1, 24, 64, 64)", style=solid]; +"2269 softmax_22" -> "2270 dropout_88" [label="(1, 24, 64, 64)", style=solid]; +"2270 dropout_88" -> "2271 matmul_45" [label="(1, 24, 64, 64)", style=solid]; +"2271 matmul_45" -> "2272 transpose_45" [label="(1, 24, 64, 32)", style=solid]; +"2272 transpose_45" -> "2273 reshape_101" [label="(1, 64, 24, 32)", style=solid]; +"2273 reshape_101" -> "2274 reshape_101_0_0_nncf_smooth_quant_0" [label="(1, 64, 768)", style=solid]; +"2274 reshape_101_0_0_nncf_smooth_quant_0" -> "2275 quantize_per_tensor_default_139" [label="(1, 64, 768)", style=solid]; +"2275 quantize_per_tensor_default_139" -> "2276 dequantize_per_tensor_default_139" [label="(1, 64, 768)", style=solid]; +"2276 dequantize_per_tensor_default_139" -> "2281 linear_138" [label="(1, 64, 768)", style=solid]; +"2277 linear_138_scale_0" -> "2280 dequantize_per_channel_default_93" [label="(768,)", style=solid]; +"2278 linear_138_zero_point_0" -> "2280 dequantize_per_channel_default_93" [label="(768,)", style=solid]; +"2279 _frozen_param147_1" -> "2280 dequantize_per_channel_default_93" [label="(768, 768)", style=solid]; +"2280 dequantize_per_channel_default_93" -> "2281 linear_138" [label="(768, 768)", style=solid]; +"2281 linear_138" -> "2282 dropout_89" [label="(1, 64, 768)", style=solid]; +"2282 dropout_89" -> "2283 view_124" [label="(1, 64, 768)", style=solid]; +"2283 view_124" -> "2284 permute_103" [label="(1, 1, 1, 8, 8, 768)", style=solid]; +"2284 permute_103" -> "2285 reshape_102" [label="(1, 1, 8, 1, 8, 768)", style=solid]; +"2285 reshape_102" -> "2286 slice_342" [label="(1, 8, 8, 768)", style=solid]; +"2286 slice_342" -> "2287 slice_343" [label="(1, 8, 8, 768)", style=solid]; +"2287 slice_343" -> "2288 slice_344" [label="(1, 7, 8, 768)", style=solid]; +"2288 slice_344" -> "2289 slice_345" [label="(1, 7, 7, 768)", style=solid]; +"2289 slice_345" -> "2290 contiguous_43" [label="(1, 7, 7, 768)", style=solid]; +"2290 contiguous_43" -> "2291 layer_norm_48" [label="(1, 7, 7, 768)", style=solid]; +"2291 layer_norm_48" -> "2292 add_78" [label="(1, 7, 7, 768)", style=solid]; +"2292 add_78" -> "2293 add_78_0_0_nncf_smooth_quant_0" [label="(1, 7, 7, 768)", style=solid]; +"2292 add_78" -> "2313 add_79" [label="(1, 7, 7, 768)", style=solid]; +"2293 add_78_0_0_nncf_smooth_quant_0" -> "2294 quantize_per_tensor_default_140" [label="(1, 7, 7, 768)", style=solid]; +"2294 quantize_per_tensor_default_140" -> "2295 dequantize_per_tensor_default_140" [label="(1, 7, 7, 768)", style=solid]; +"2295 dequantize_per_tensor_default_140" -> "2300 linear_139" [label="(1, 7, 7, 768)", style=solid]; +"2296 linear_139_scale_0" -> "2299 dequantize_per_channel_default_94" [label="(3072,)", style=solid]; +"2297 linear_139_zero_point_0" -> "2299 dequantize_per_channel_default_94" [label="(3072,)", style=solid]; +"2298 _frozen_param148_1" -> "2299 dequantize_per_channel_default_94" [label="(3072, 768)", style=solid]; +"2299 dequantize_per_channel_default_94" -> "2300 linear_139" [label="(3072, 768)", style=solid]; +"2300 linear_139" -> "2301 gelu_22" [label="(1, 7, 7, 3072)", style=solid]; +"2301 gelu_22" -> "2302 dropout_90" [label="(1, 7, 7, 3072)", style=solid]; +"2302 dropout_90" -> "2303 dropout_90_0_0_nncf_smooth_quant_0" [label="(1, 7, 7, 3072)", style=solid]; +"2303 dropout_90_0_0_nncf_smooth_quant_0" -> "2304 quantize_per_tensor_default_141" [label="(1, 7, 7, 3072)", style=solid]; +"2304 quantize_per_tensor_default_141" -> "2305 dequantize_per_tensor_default_141" [label="(1, 7, 7, 3072)", style=solid]; +"2305 dequantize_per_tensor_default_141" -> "2310 linear_140" [label="(1, 7, 7, 3072)", style=solid]; +"2306 linear_140_scale_0" -> "2309 dequantize_per_channel_default_95" [label="(768,)", style=solid]; +"2307 linear_140_zero_point_0" -> "2309 dequantize_per_channel_default_95" [label="(768,)", style=solid]; +"2308 _frozen_param149_1" -> "2309 dequantize_per_channel_default_95" [label="(768, 3072)", style=solid]; +"2309 dequantize_per_channel_default_95" -> "2310 linear_140" [label="(768, 3072)", style=solid]; +"2310 linear_140" -> "2311 dropout_91" [label="(1, 7, 7, 768)", style=solid]; +"2311 dropout_91" -> "2312 layer_norm_49" [label="(1, 7, 7, 768)", style=solid]; +"2312 layer_norm_49" -> "2313 add_79" [label="(1, 7, 7, 768)", style=solid]; +"2313 add_79" -> "2315 pad_26" [label="(1, 7, 7, 768)", style=solid]; +"2313 add_79" -> "2373 add_81" [label="(1, 7, 7, 768)", style=solid]; +"2314 _frozen_param191" -> "2349 add_80" [label="(1, 24, 64, 64)", style=solid]; +"2315 pad_26" -> "2316 view_127" [label="(1, 8, 8, 768)", style=solid]; +"2316 view_127" -> "2317 permute_105" [label="(1, 1, 8, 1, 8, 768)", style=solid]; +"2317 permute_105" -> "2318 reshape_103" [label="(1, 1, 1, 8, 8, 768)", style=solid]; +"2318 reshape_103" -> "2320 reshape_103_0_0_nncf_smooth_quant_0" [label="(1, 64, 768)", style=solid]; +"2319 _frozen_param192" -> "2327 linear_143" [label="(2304,)", style=solid]; +"2320 reshape_103_0_0_nncf_smooth_quant_0" -> "2321 quantize_per_tensor_default_142" [label="(1, 64, 768)", style=solid]; +"2321 quantize_per_tensor_default_142" -> "2322 dequantize_per_tensor_default_142" [label="(1, 64, 768)", style=solid]; +"2322 dequantize_per_tensor_default_142" -> "2327 linear_143" [label="(1, 64, 768)", style=solid]; +"2323 linear_143_scale_0" -> "2326 dequantize_per_channel_default_96" [label="(2304,)", style=solid]; +"2324 linear_143_zero_point_0" -> "2326 dequantize_per_channel_default_96" [label="(2304,)", style=solid]; +"2325 _frozen_param150_1" -> "2326 dequantize_per_channel_default_96" [label="(2304, 768)", style=solid]; +"2326 dequantize_per_channel_default_96" -> "2327 linear_143" [label="(2304, 768)", style=solid]; +"2327 linear_143" -> "2328 reshape_104" [label="(1, 64, 2304)", style=solid]; +"2328 reshape_104" -> "2329 permute_106" [label="(1, 64, 3, 24, 32)", style=solid]; +"2329 permute_106" -> "2330 select_69" [label="(3, 1, 24, 64, 32)", style=solid]; +"2329 permute_106" -> "2331 select_70" [label="(3, 1, 24, 64, 32)", style=solid]; +"2329 permute_106" -> "2332 select_71" [label="(3, 1, 24, 64, 32)", style=solid]; +"2330 select_69" -> "2333 linalg_vector_norm_46" [label="(1, 24, 64, 32)", style=solid]; +"2330 select_69" -> "2335 expand_as_46" [label="(1, 24, 64, 32)", style=solid]; +"2330 select_69" -> "2336 div_46" [label="(1, 24, 64, 32)", style=solid]; +"2331 select_70" -> "2339 linalg_vector_norm_47" [label="(1, 24, 64, 32)", style=solid]; +"2331 select_70" -> "2341 expand_as_47" [label="(1, 24, 64, 32)", style=solid]; +"2331 select_70" -> "2342 div_47" [label="(1, 24, 64, 32)", style=solid]; +"2332 select_71" -> "2352 matmul_47" [label="(1, 24, 64, 32)", style=solid]; +"2333 linalg_vector_norm_46" -> "2334 clamp_min_46" [label="(1, 24, 64, 1)", style=solid]; +"2334 clamp_min_46" -> "2335 expand_as_46" [label="(1, 24, 64, 1)", style=solid]; +"2335 expand_as_46" -> "2336 div_46" [label="(1, 24, 64, 32)", style=solid]; +"2336 div_46" -> "2337 quantize_per_tensor_default_143" [label="(1, 24, 64, 32)", style=solid]; +"2337 quantize_per_tensor_default_143" -> "2338 dequantize_per_tensor_default_143" [label="(1, 24, 64, 32)", style=solid]; +"2338 dequantize_per_tensor_default_143" -> "2346 matmul_46" [label="(1, 24, 64, 32)", style=solid]; +"2339 linalg_vector_norm_47" -> "2340 clamp_min_47" [label="(1, 24, 64, 1)", style=solid]; +"2340 clamp_min_47" -> "2341 expand_as_47" [label="(1, 24, 64, 1)", style=solid]; +"2341 expand_as_47" -> "2342 div_47" [label="(1, 24, 64, 32)", style=solid]; +"2342 div_47" -> "2343 quantize_per_tensor_default_144" [label="(1, 24, 64, 32)", style=solid]; +"2343 quantize_per_tensor_default_144" -> "2344 dequantize_per_tensor_default_144" [label="(1, 24, 64, 32)", style=solid]; +"2344 dequantize_per_tensor_default_144" -> "2345 transpose_46" [label="(1, 24, 64, 32)", style=solid]; +"2345 transpose_46" -> "2346 matmul_46" [label="(1, 24, 32, 64)", style=solid]; +"2346 matmul_46" -> "2348 mul_47" [label="(1, 24, 64, 64)", style=solid]; +"2347 _frozen_param194" -> "2348 mul_47" [label="(24, 1, 1)", style=solid]; +"2348 mul_47" -> "2349 add_80" [label="(1, 24, 64, 64)", style=solid]; +"2349 add_80" -> "2350 softmax_23" [label="(1, 24, 64, 64)", style=solid]; +"2350 softmax_23" -> "2351 dropout_92" [label="(1, 24, 64, 64)", style=solid]; +"2351 dropout_92" -> "2352 matmul_47" [label="(1, 24, 64, 64)", style=solid]; +"2352 matmul_47" -> "2353 transpose_47" [label="(1, 24, 64, 32)", style=solid]; +"2353 transpose_47" -> "2354 reshape_105" [label="(1, 64, 24, 32)", style=solid]; +"2354 reshape_105" -> "2355 reshape_105_0_0_nncf_smooth_quant_0" [label="(1, 64, 768)", style=solid]; +"2355 reshape_105_0_0_nncf_smooth_quant_0" -> "2356 quantize_per_tensor_default_145" [label="(1, 64, 768)", style=solid]; +"2356 quantize_per_tensor_default_145" -> "2357 dequantize_per_tensor_default_145" [label="(1, 64, 768)", style=solid]; +"2357 dequantize_per_tensor_default_145" -> "2362 linear_144" [label="(1, 64, 768)", style=solid]; +"2358 linear_144_scale_0" -> "2361 dequantize_per_channel_default_97" [label="(768,)", style=solid]; +"2359 linear_144_zero_point_0" -> "2361 dequantize_per_channel_default_97" [label="(768,)", style=solid]; +"2360 _frozen_param151_1" -> "2361 dequantize_per_channel_default_97" [label="(768, 768)", style=solid]; +"2361 dequantize_per_channel_default_97" -> "2362 linear_144" [label="(768, 768)", style=solid]; +"2362 linear_144" -> "2363 dropout_93" [label="(1, 64, 768)", style=solid]; +"2363 dropout_93" -> "2364 view_128" [label="(1, 64, 768)", style=solid]; +"2364 view_128" -> "2365 permute_107" [label="(1, 1, 1, 8, 8, 768)", style=solid]; +"2365 permute_107" -> "2366 reshape_106" [label="(1, 1, 8, 1, 8, 768)", style=solid]; +"2366 reshape_106" -> "2367 slice_347" [label="(1, 8, 8, 768)", style=solid]; +"2367 slice_347" -> "2368 slice_348" [label="(1, 8, 8, 768)", style=solid]; +"2368 slice_348" -> "2369 slice_349" [label="(1, 7, 8, 768)", style=solid]; +"2369 slice_349" -> "2370 slice_350" [label="(1, 7, 7, 768)", style=solid]; +"2370 slice_350" -> "2371 contiguous_45" [label="(1, 7, 7, 768)", style=solid]; +"2371 contiguous_45" -> "2372 layer_norm_50" [label="(1, 7, 7, 768)", style=solid]; +"2372 layer_norm_50" -> "2373 add_81" [label="(1, 7, 7, 768)", style=solid]; +"2373 add_81" -> "2374 add_81_0_0_nncf_smooth_quant_0" [label="(1, 7, 7, 768)", style=solid]; +"2373 add_81" -> "2394 add_82" [label="(1, 7, 7, 768)", style=solid]; +"2374 add_81_0_0_nncf_smooth_quant_0" -> "2375 quantize_per_tensor_default_146" [label="(1, 7, 7, 768)", style=solid]; +"2375 quantize_per_tensor_default_146" -> "2376 dequantize_per_tensor_default_146" [label="(1, 7, 7, 768)", style=solid]; +"2376 dequantize_per_tensor_default_146" -> "2381 linear_145" [label="(1, 7, 7, 768)", style=solid]; +"2377 linear_145_scale_0" -> "2380 dequantize_per_channel_default_98" [label="(3072,)", style=solid]; +"2378 linear_145_zero_point_0" -> "2380 dequantize_per_channel_default_98" [label="(3072,)", style=solid]; +"2379 _frozen_param152_1" -> "2380 dequantize_per_channel_default_98" [label="(3072, 768)", style=solid]; +"2380 dequantize_per_channel_default_98" -> "2381 linear_145" [label="(3072, 768)", style=solid]; +"2381 linear_145" -> "2382 gelu_23" [label="(1, 7, 7, 3072)", style=solid]; +"2382 gelu_23" -> "2383 dropout_94" [label="(1, 7, 7, 3072)", style=solid]; +"2383 dropout_94" -> "2384 dropout_94_0_0_nncf_smooth_quant_0" [label="(1, 7, 7, 3072)", style=solid]; +"2384 dropout_94_0_0_nncf_smooth_quant_0" -> "2385 quantize_per_tensor_default_147" [label="(1, 7, 7, 3072)", style=solid]; +"2385 quantize_per_tensor_default_147" -> "2386 dequantize_per_tensor_default_147" [label="(1, 7, 7, 3072)", style=solid]; +"2386 dequantize_per_tensor_default_147" -> "2391 linear_146" [label="(1, 7, 7, 3072)", style=solid]; +"2387 linear_146_scale_0" -> "2390 dequantize_per_channel_default_99" [label="(768,)", style=solid]; +"2388 linear_146_zero_point_0" -> "2390 dequantize_per_channel_default_99" [label="(768,)", style=solid]; +"2389 _frozen_param155_1" -> "2390 dequantize_per_channel_default_99" [label="(768, 3072)", style=solid]; +"2390 dequantize_per_channel_default_99" -> "2391 linear_146" [label="(768, 3072)", style=solid]; +"2391 linear_146" -> "2392 dropout_95" [label="(1, 7, 7, 768)", style=solid]; +"2392 dropout_95" -> "2393 layer_norm_51" [label="(1, 7, 7, 768)", style=solid]; +"2393 layer_norm_51" -> "2394 add_82" [label="(1, 7, 7, 768)", style=solid]; +"2394 add_82" -> "2395 layer_norm_52" [label="(1, 7, 7, 768)", style=solid]; +"2395 layer_norm_52" -> "2396 permute_108" [label="(1, 7, 7, 768)", style=solid]; +"2396 permute_108" -> "2397 adaptive_avg_pool2d" [label="(1, 768, 7, 7)", style=solid]; +"2397 adaptive_avg_pool2d" -> "2398 flatten" [label="(1, 768, 1, 1)", style=solid]; +"2398 flatten" -> "2399 flatten_0_0_nncf_smooth_quant_0" [label="(1, 768)", style=solid]; +"2399 flatten_0_0_nncf_smooth_quant_0" -> "2400 quantize_per_tensor_default_148" [label="(1, 768)", style=solid]; +"2400 quantize_per_tensor_default_148" -> "2401 dequantize_per_tensor_default_148" [label="(1, 768)", style=solid]; +"2401 dequantize_per_tensor_default_148" -> "2406 linear_147" [label="(1, 768)", style=solid]; +"2402 linear_147_scale_0" -> "2405 dequantize_per_channel_default_100" [label="(1000,)", style=solid]; +"2403 linear_147_zero_point_0" -> "2405 dequantize_per_channel_default_100" [label="(1000,)", style=solid]; +"2404 _frozen_param159_1" -> "2405 dequantize_per_channel_default_100" [label="(1000, 768)", style=solid]; +"2405 dequantize_per_channel_default_100" -> "2406 linear_147" [label="(1000, 768)", style=solid]; +"2406 linear_147" -> "2407 output" [label="(1, 1000)", style=solid]; } diff --git a/tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/vit_b_16.dot b/tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/vit_b_16.dot index 6209b772176..f4214d12db9 100644 --- a/tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/vit_b_16.dot +++ b/tests/torch/data/reference_graphs/fx/experimental/OpenVINOQuantizer/vit_b_16.dot @@ -1,1913 +1,1911 @@ strict digraph { -"0 class_token" [id=0, type=get_attr]; -"1 encoder_pos_embedding" [id=1, type=get_attr]; -"2 conv_proj_bias" [id=2, type=get_attr]; -"3 encoder_layers_encoder_layer_0_ln_1_weight" [id=3, type=get_attr]; -"4 encoder_layers_encoder_layer_0_ln_1_bias" [id=4, type=get_attr]; -"5 encoder_layers_encoder_layer_0_self_attention_in_proj_bias" [id=5, type=get_attr]; -"6 encoder_layers_encoder_layer_0_self_attention_out_proj_bias" [id=6, type=get_attr]; -"7 encoder_layers_encoder_layer_0_ln_2_weight" [id=7, type=get_attr]; -"8 encoder_layers_encoder_layer_0_ln_2_bias" [id=8, type=get_attr]; -"9 encoder_layers_encoder_layer_0_mlp_0_bias" [id=9, type=get_attr]; -"10 encoder_layers_encoder_layer_0_mlp_3_bias" [id=10, type=get_attr]; -"11 encoder_layers_encoder_layer_1_ln_1_weight" [id=11, type=get_attr]; -"12 encoder_layers_encoder_layer_1_ln_1_bias" [id=12, type=get_attr]; -"13 encoder_layers_encoder_layer_1_self_attention_in_proj_bias" [id=13, type=get_attr]; -"14 encoder_layers_encoder_layer_1_self_attention_out_proj_bias" [id=14, type=get_attr]; -"15 encoder_layers_encoder_layer_1_ln_2_weight" [id=15, type=get_attr]; -"16 encoder_layers_encoder_layer_1_ln_2_bias" [id=16, type=get_attr]; -"17 encoder_layers_encoder_layer_1_mlp_0_bias" [id=17, type=get_attr]; -"18 encoder_layers_encoder_layer_1_mlp_3_bias" [id=18, type=get_attr]; -"19 encoder_layers_encoder_layer_2_ln_1_weight" [id=19, type=get_attr]; -"20 encoder_layers_encoder_layer_2_ln_1_bias" [id=20, type=get_attr]; -"21 encoder_layers_encoder_layer_2_self_attention_in_proj_bias" [id=21, type=get_attr]; -"22 encoder_layers_encoder_layer_2_self_attention_out_proj_bias" [id=22, type=get_attr]; -"23 encoder_layers_encoder_layer_2_ln_2_weight" [id=23, type=get_attr]; -"24 encoder_layers_encoder_layer_2_ln_2_bias" [id=24, type=get_attr]; -"25 encoder_layers_encoder_layer_2_mlp_0_bias" [id=25, type=get_attr]; -"26 encoder_layers_encoder_layer_2_mlp_3_bias" [id=26, type=get_attr]; -"27 encoder_layers_encoder_layer_3_ln_1_weight" [id=27, type=get_attr]; -"28 encoder_layers_encoder_layer_3_ln_1_bias" [id=28, type=get_attr]; -"29 encoder_layers_encoder_layer_3_self_attention_in_proj_bias" [id=29, type=get_attr]; -"30 encoder_layers_encoder_layer_3_self_attention_out_proj_bias" [id=30, type=get_attr]; -"31 encoder_layers_encoder_layer_3_ln_2_weight" [id=31, type=get_attr]; -"32 encoder_layers_encoder_layer_3_ln_2_bias" [id=32, type=get_attr]; -"33 encoder_layers_encoder_layer_3_mlp_0_bias" [id=33, type=get_attr]; -"34 encoder_layers_encoder_layer_3_mlp_3_bias" [id=34, type=get_attr]; -"35 encoder_layers_encoder_layer_4_ln_1_weight" [id=35, type=get_attr]; -"36 encoder_layers_encoder_layer_4_ln_1_bias" [id=36, type=get_attr]; -"37 encoder_layers_encoder_layer_4_self_attention_in_proj_bias" [id=37, type=get_attr]; -"38 encoder_layers_encoder_layer_4_self_attention_out_proj_bias" [id=38, type=get_attr]; -"39 encoder_layers_encoder_layer_4_ln_2_weight" [id=39, type=get_attr]; -"40 encoder_layers_encoder_layer_4_ln_2_bias" [id=40, type=get_attr]; -"41 encoder_layers_encoder_layer_4_mlp_0_bias" [id=41, type=get_attr]; -"42 encoder_layers_encoder_layer_4_mlp_3_bias" [id=42, type=get_attr]; -"43 encoder_layers_encoder_layer_5_ln_1_weight" [id=43, type=get_attr]; -"44 encoder_layers_encoder_layer_5_ln_1_bias" [id=44, type=get_attr]; -"45 encoder_layers_encoder_layer_5_self_attention_in_proj_bias" [id=45, type=get_attr]; -"46 encoder_layers_encoder_layer_5_self_attention_out_proj_bias" [id=46, type=get_attr]; -"47 encoder_layers_encoder_layer_5_ln_2_weight" [id=47, type=get_attr]; -"48 encoder_layers_encoder_layer_5_ln_2_bias" [id=48, type=get_attr]; -"49 encoder_layers_encoder_layer_5_mlp_0_bias" [id=49, type=get_attr]; -"50 encoder_layers_encoder_layer_5_mlp_3_bias" [id=50, type=get_attr]; -"51 encoder_layers_encoder_layer_6_ln_1_weight" [id=51, type=get_attr]; -"52 encoder_layers_encoder_layer_6_ln_1_bias" [id=52, type=get_attr]; -"53 encoder_layers_encoder_layer_6_self_attention_in_proj_bias" [id=53, type=get_attr]; -"54 encoder_layers_encoder_layer_6_self_attention_out_proj_bias" [id=54, type=get_attr]; -"55 encoder_layers_encoder_layer_6_ln_2_weight" [id=55, type=get_attr]; -"56 encoder_layers_encoder_layer_6_ln_2_bias" [id=56, type=get_attr]; -"57 encoder_layers_encoder_layer_6_mlp_0_bias" [id=57, type=get_attr]; -"58 encoder_layers_encoder_layer_6_mlp_3_bias" [id=58, type=get_attr]; -"59 encoder_layers_encoder_layer_7_ln_1_weight" [id=59, type=get_attr]; -"60 encoder_layers_encoder_layer_7_ln_1_bias" [id=60, type=get_attr]; -"61 encoder_layers_encoder_layer_7_self_attention_in_proj_bias" [id=61, type=get_attr]; -"62 encoder_layers_encoder_layer_7_self_attention_out_proj_bias" [id=62, type=get_attr]; -"63 encoder_layers_encoder_layer_7_ln_2_weight" [id=63, type=get_attr]; -"64 encoder_layers_encoder_layer_7_ln_2_bias" [id=64, type=get_attr]; -"65 encoder_layers_encoder_layer_7_mlp_0_bias" [id=65, type=get_attr]; -"66 encoder_layers_encoder_layer_7_mlp_3_bias" [id=66, type=get_attr]; -"67 encoder_layers_encoder_layer_8_ln_1_weight" [id=67, type=get_attr]; -"68 encoder_layers_encoder_layer_8_ln_1_bias" [id=68, type=get_attr]; -"69 encoder_layers_encoder_layer_8_self_attention_in_proj_bias" [id=69, type=get_attr]; -"70 encoder_layers_encoder_layer_8_self_attention_out_proj_bias" [id=70, type=get_attr]; -"71 encoder_layers_encoder_layer_8_ln_2_weight" [id=71, type=get_attr]; -"72 encoder_layers_encoder_layer_8_ln_2_bias" [id=72, type=get_attr]; -"73 encoder_layers_encoder_layer_8_mlp_0_bias" [id=73, type=get_attr]; -"74 encoder_layers_encoder_layer_8_mlp_3_bias" [id=74, type=get_attr]; -"75 encoder_layers_encoder_layer_9_ln_1_weight" [id=75, type=get_attr]; -"76 encoder_layers_encoder_layer_9_ln_1_bias" [id=76, type=get_attr]; -"77 encoder_layers_encoder_layer_9_self_attention_in_proj_bias" [id=77, type=get_attr]; -"78 encoder_layers_encoder_layer_9_self_attention_out_proj_bias" [id=78, type=get_attr]; -"79 encoder_layers_encoder_layer_9_ln_2_weight" [id=79, type=get_attr]; -"80 encoder_layers_encoder_layer_9_ln_2_bias" [id=80, type=get_attr]; -"81 encoder_layers_encoder_layer_9_mlp_0_bias" [id=81, type=get_attr]; -"82 encoder_layers_encoder_layer_9_mlp_3_bias" [id=82, type=get_attr]; -"83 encoder_layers_encoder_layer_10_ln_1_weight" [id=83, type=get_attr]; -"84 encoder_layers_encoder_layer_10_ln_1_bias" [id=84, type=get_attr]; -"85 encoder_layers_encoder_layer_10_self_attention_in_proj_bias" [id=85, type=get_attr]; -"86 encoder_layers_encoder_layer_10_self_attention_out_proj_bias" [id=86, type=get_attr]; -"87 encoder_layers_encoder_layer_10_ln_2_weight" [id=87, type=get_attr]; -"88 encoder_layers_encoder_layer_10_ln_2_bias" [id=88, type=get_attr]; -"89 encoder_layers_encoder_layer_10_mlp_0_bias" [id=89, type=get_attr]; -"90 encoder_layers_encoder_layer_10_mlp_3_bias" [id=90, type=get_attr]; -"91 encoder_layers_encoder_layer_11_ln_1_weight" [id=91, type=get_attr]; -"92 encoder_layers_encoder_layer_11_ln_1_bias" [id=92, type=get_attr]; -"93 encoder_layers_encoder_layer_11_self_attention_in_proj_bias" [id=93, type=get_attr]; -"94 encoder_layers_encoder_layer_11_self_attention_out_proj_bias" [id=94, type=get_attr]; -"95 encoder_layers_encoder_layer_11_ln_2_weight" [id=95, type=get_attr]; -"96 encoder_layers_encoder_layer_11_ln_2_bias" [id=96, type=get_attr]; -"97 encoder_layers_encoder_layer_11_mlp_0_bias" [id=97, type=get_attr]; -"98 encoder_layers_encoder_layer_11_mlp_3_bias" [id=98, type=get_attr]; -"99 encoder_ln_weight" [id=99, type=get_attr]; -"100 encoder_ln_bias" [id=100, type=get_attr]; -"101 heads_head_bias" [id=101, type=get_attr]; -"102 x" [id=102, type=input]; -"103 quantize_per_tensor_default" [id=103, type=quantize_per_tensor]; -"104 dequantize_per_tensor_default" [id=104, type=dequantize_per_tensor]; -"105 conv2d_scale_0" [id=105, type=get_attr]; -"106 conv2d_zero_point_0" [id=106, type=get_attr]; -"107 _frozen_param0" [id=107, type=get_attr]; -"108 dequantize_per_channel_default" [id=108, type=dequantize_per_channel]; -"109 conv2d" [id=109, type=conv2d]; -"110 reshape" [id=110, type=reshape]; -"111 permute" [id=111, type=permute]; -"112 expand" [id=112, type=expand]; -"113 cat" [id=113, type=cat]; -"114 add" [id=114, type=add]; -"115 dropout" [id=115, type=dropout]; -"116 layer_norm" [id=116, type=layer_norm]; -"117 transpose" [id=117, type=transpose]; -"118 transpose_0_0_nncf_smooth_quant_0" [id=118, type=call_module]; -"119 quantize_per_tensor_default_1" [id=119, type=quantize_per_tensor]; -"120 dequantize_per_tensor_default_1" [id=120, type=dequantize_per_tensor]; -"121 linear_scale_0" [id=121, type=get_attr]; -"122 linear_zero_point_0" [id=122, type=get_attr]; -"123 _frozen_param1" [id=123, type=get_attr]; -"124 dequantize_per_channel_default_1" [id=124, type=dequantize_per_channel]; -"125 linear" [id=125, type=linear]; -"126 unflatten" [id=126, type=unflatten]; -"127 unsqueeze" [id=127, type=unsqueeze]; -"128 transpose_1" [id=128, type=transpose]; -"129 squeeze" [id=129, type=squeeze]; -"130 contiguous" [id=130, type=contiguous]; -"131 quantize_per_tensor_default_2" [id=131, type=quantize_per_tensor]; -"132 dequantize_per_tensor_default_2" [id=132, type=dequantize_per_tensor]; -"133 select" [id=133, type=select]; -"134 quantize_per_tensor_default_3" [id=134, type=quantize_per_tensor]; -"135 dequantize_per_tensor_default_3" [id=135, type=dequantize_per_tensor]; -"136 select_1" [id=136, type=select]; -"137 select_2" [id=137, type=select]; -"138 view" [id=138, type=view]; -"139 transpose_2" [id=139, type=transpose]; -"140 view_1" [id=140, type=view]; -"141 transpose_3" [id=141, type=transpose]; -"142 view_2" [id=142, type=view]; -"143 transpose_4" [id=143, type=transpose]; -"144 view_3" [id=144, type=view]; -"145 view_4" [id=145, type=view]; -"146 view_5" [id=146, type=view]; -"147 scaled_dot_product_attention" [id=147, type=scaled_dot_product_attention]; -"148 permute_1" [id=148, type=permute]; -"149 view_6" [id=149, type=view]; -"150 view_6_0_0_nncf_smooth_quant_0" [id=150, type=call_module]; -"151 quantize_per_tensor_default_4" [id=151, type=quantize_per_tensor]; -"152 dequantize_per_tensor_default_4" [id=152, type=dequantize_per_tensor]; -"153 linear_1_scale_0" [id=153, type=get_attr]; -"154 linear_1_zero_point_0" [id=154, type=get_attr]; -"155 _frozen_param2" [id=155, type=get_attr]; -"156 dequantize_per_channel_default_2" [id=156, type=dequantize_per_channel]; -"157 linear_1" [id=157, type=linear]; -"158 view_7" [id=158, type=view]; -"159 transpose_5" [id=159, type=transpose]; -"160 dropout_1" [id=160, type=dropout]; -"161 add_1" [id=161, type=add]; -"162 layer_norm_1" [id=162, type=layer_norm]; -"163 layer_norm_1_0_0_nncf_smooth_quant_0" [id=163, type=call_module]; -"164 quantize_per_tensor_default_5" [id=164, type=quantize_per_tensor]; -"165 dequantize_per_tensor_default_5" [id=165, type=dequantize_per_tensor]; -"166 linear_2_scale_0" [id=166, type=get_attr]; -"167 linear_2_zero_point_0" [id=167, type=get_attr]; -"168 _frozen_param3" [id=168, type=get_attr]; -"169 dequantize_per_channel_default_3" [id=169, type=dequantize_per_channel]; -"170 linear_2" [id=170, type=linear]; -"171 gelu" [id=171, type=gelu]; -"172 dropout_2" [id=172, type=dropout]; -"173 dropout_2_0_0_nncf_smooth_quant_0" [id=173, type=call_module]; -"174 quantize_per_tensor_default_6" [id=174, type=quantize_per_tensor]; -"175 dequantize_per_tensor_default_6" [id=175, type=dequantize_per_tensor]; -"176 linear_3_scale_0" [id=176, type=get_attr]; -"177 linear_3_zero_point_0" [id=177, type=get_attr]; -"178 _frozen_param4" [id=178, type=get_attr]; -"179 dequantize_per_channel_default_4" [id=179, type=dequantize_per_channel]; -"180 linear_3" [id=180, type=linear]; -"181 dropout_3" [id=181, type=dropout]; -"182 add_2" [id=182, type=add]; -"183 layer_norm_2" [id=183, type=layer_norm]; -"184 transpose_6" [id=184, type=transpose]; -"185 transpose_6_0_0_nncf_smooth_quant_0" [id=185, type=call_module]; -"186 quantize_per_tensor_default_7" [id=186, type=quantize_per_tensor]; -"187 dequantize_per_tensor_default_7" [id=187, type=dequantize_per_tensor]; -"188 linear_4_scale_0" [id=188, type=get_attr]; -"189 linear_4_zero_point_0" [id=189, type=get_attr]; -"190 _frozen_param5" [id=190, type=get_attr]; -"191 dequantize_per_channel_default_5" [id=191, type=dequantize_per_channel]; -"192 linear_4" [id=192, type=linear]; -"193 unflatten_1" [id=193, type=unflatten]; -"194 unsqueeze_1" [id=194, type=unsqueeze]; -"195 transpose_7" [id=195, type=transpose]; -"196 squeeze_1" [id=196, type=squeeze]; -"197 contiguous_1" [id=197, type=contiguous]; -"198 quantize_per_tensor_default_8" [id=198, type=quantize_per_tensor]; -"199 dequantize_per_tensor_default_8" [id=199, type=dequantize_per_tensor]; -"200 select_3" [id=200, type=select]; -"201 quantize_per_tensor_default_9" [id=201, type=quantize_per_tensor]; -"202 dequantize_per_tensor_default_9" [id=202, type=dequantize_per_tensor]; -"203 select_4" [id=203, type=select]; -"204 select_5" [id=204, type=select]; -"205 view_8" [id=205, type=view]; -"206 transpose_8" [id=206, type=transpose]; -"207 view_9" [id=207, type=view]; -"208 transpose_9" [id=208, type=transpose]; -"209 view_10" [id=209, type=view]; -"210 transpose_10" [id=210, type=transpose]; -"211 view_11" [id=211, type=view]; -"212 view_12" [id=212, type=view]; -"213 view_13" [id=213, type=view]; -"214 scaled_dot_product_attention_1" [id=214, type=scaled_dot_product_attention]; -"215 permute_2" [id=215, type=permute]; -"216 view_14" [id=216, type=view]; -"217 view_14_0_0_nncf_smooth_quant_0" [id=217, type=call_module]; -"218 quantize_per_tensor_default_10" [id=218, type=quantize_per_tensor]; -"219 dequantize_per_tensor_default_10" [id=219, type=dequantize_per_tensor]; -"220 linear_5_scale_0" [id=220, type=get_attr]; -"221 linear_5_zero_point_0" [id=221, type=get_attr]; -"222 _frozen_param6" [id=222, type=get_attr]; -"223 dequantize_per_channel_default_6" [id=223, type=dequantize_per_channel]; -"224 linear_5" [id=224, type=linear]; -"225 view_15" [id=225, type=view]; -"226 transpose_11" [id=226, type=transpose]; -"227 dropout_4" [id=227, type=dropout]; -"228 add_3" [id=228, type=add]; -"229 layer_norm_3" [id=229, type=layer_norm]; -"230 layer_norm_3_0_0_nncf_smooth_quant_0" [id=230, type=call_module]; -"231 quantize_per_tensor_default_11" [id=231, type=quantize_per_tensor]; -"232 dequantize_per_tensor_default_11" [id=232, type=dequantize_per_tensor]; -"233 linear_6_scale_0" [id=233, type=get_attr]; -"234 linear_6_zero_point_0" [id=234, type=get_attr]; -"235 _frozen_param7" [id=235, type=get_attr]; -"236 dequantize_per_channel_default_7" [id=236, type=dequantize_per_channel]; -"237 linear_6" [id=237, type=linear]; -"238 gelu_1" [id=238, type=gelu]; -"239 dropout_5" [id=239, type=dropout]; -"240 dropout_5_0_0_nncf_smooth_quant_0" [id=240, type=call_module]; -"241 quantize_per_tensor_default_12" [id=241, type=quantize_per_tensor]; -"242 dequantize_per_tensor_default_12" [id=242, type=dequantize_per_tensor]; -"243 linear_7_scale_0" [id=243, type=get_attr]; -"244 linear_7_zero_point_0" [id=244, type=get_attr]; -"245 _frozen_param8" [id=245, type=get_attr]; -"246 dequantize_per_channel_default_8" [id=246, type=dequantize_per_channel]; -"247 linear_7" [id=247, type=linear]; -"248 dropout_6" [id=248, type=dropout]; -"249 add_4" [id=249, type=add]; -"250 layer_norm_4" [id=250, type=layer_norm]; -"251 transpose_12" [id=251, type=transpose]; -"252 transpose_12_0_0_nncf_smooth_quant_0" [id=252, type=call_module]; -"253 quantize_per_tensor_default_13" [id=253, type=quantize_per_tensor]; -"254 dequantize_per_tensor_default_13" [id=254, type=dequantize_per_tensor]; -"255 linear_8_scale_0" [id=255, type=get_attr]; -"256 linear_8_zero_point_0" [id=256, type=get_attr]; -"257 _frozen_param9" [id=257, type=get_attr]; -"258 dequantize_per_channel_default_9" [id=258, type=dequantize_per_channel]; -"259 linear_8" [id=259, type=linear]; -"260 unflatten_2" [id=260, type=unflatten]; -"261 unsqueeze_2" [id=261, type=unsqueeze]; -"262 transpose_13" [id=262, type=transpose]; -"263 squeeze_2" [id=263, type=squeeze]; -"264 contiguous_2" [id=264, type=contiguous]; -"265 quantize_per_tensor_default_14" [id=265, type=quantize_per_tensor]; -"266 dequantize_per_tensor_default_14" [id=266, type=dequantize_per_tensor]; -"267 select_6" [id=267, type=select]; -"268 quantize_per_tensor_default_15" [id=268, type=quantize_per_tensor]; -"269 dequantize_per_tensor_default_15" [id=269, type=dequantize_per_tensor]; -"270 select_7" [id=270, type=select]; -"271 select_8" [id=271, type=select]; -"272 view_16" [id=272, type=view]; -"273 transpose_14" [id=273, type=transpose]; -"274 view_17" [id=274, type=view]; -"275 transpose_15" [id=275, type=transpose]; -"276 view_18" [id=276, type=view]; -"277 transpose_16" [id=277, type=transpose]; -"278 view_19" [id=278, type=view]; -"279 view_20" [id=279, type=view]; -"280 view_21" [id=280, type=view]; -"281 scaled_dot_product_attention_2" [id=281, type=scaled_dot_product_attention]; -"282 permute_3" [id=282, type=permute]; -"283 view_22" [id=283, type=view]; -"284 view_22_0_0_nncf_smooth_quant_0" [id=284, type=call_module]; -"285 quantize_per_tensor_default_16" [id=285, type=quantize_per_tensor]; -"286 dequantize_per_tensor_default_16" [id=286, type=dequantize_per_tensor]; -"287 linear_9_scale_0" [id=287, type=get_attr]; -"288 linear_9_zero_point_0" [id=288, type=get_attr]; -"289 _frozen_param10" [id=289, type=get_attr]; -"290 dequantize_per_channel_default_10" [id=290, type=dequantize_per_channel]; -"291 linear_9" [id=291, type=linear]; -"292 view_23" [id=292, type=view]; -"293 transpose_17" [id=293, type=transpose]; -"294 dropout_7" [id=294, type=dropout]; -"295 add_5" [id=295, type=add]; -"296 layer_norm_5" [id=296, type=layer_norm]; -"297 layer_norm_5_0_0_nncf_smooth_quant_0" [id=297, type=call_module]; -"298 quantize_per_tensor_default_17" [id=298, type=quantize_per_tensor]; -"299 dequantize_per_tensor_default_17" [id=299, type=dequantize_per_tensor]; -"300 linear_10_scale_0" [id=300, type=get_attr]; -"301 linear_10_zero_point_0" [id=301, type=get_attr]; -"302 _frozen_param11" [id=302, type=get_attr]; -"303 dequantize_per_channel_default_11" [id=303, type=dequantize_per_channel]; -"304 linear_10" [id=304, type=linear]; -"305 gelu_2" [id=305, type=gelu]; -"306 dropout_8" [id=306, type=dropout]; -"307 dropout_8_0_0_nncf_smooth_quant_0" [id=307, type=call_module]; -"308 quantize_per_tensor_default_18" [id=308, type=quantize_per_tensor]; -"309 dequantize_per_tensor_default_18" [id=309, type=dequantize_per_tensor]; -"310 linear_11_scale_0" [id=310, type=get_attr]; -"311 linear_11_zero_point_0" [id=311, type=get_attr]; -"312 _frozen_param12" [id=312, type=get_attr]; -"313 dequantize_per_channel_default_12" [id=313, type=dequantize_per_channel]; -"314 linear_11" [id=314, type=linear]; -"315 dropout_9" [id=315, type=dropout]; -"316 add_6" [id=316, type=add]; -"317 layer_norm_6" [id=317, type=layer_norm]; -"318 transpose_18" [id=318, type=transpose]; -"319 transpose_18_0_0_nncf_smooth_quant_0" [id=319, type=call_module]; -"320 quantize_per_tensor_default_19" [id=320, type=quantize_per_tensor]; -"321 dequantize_per_tensor_default_19" [id=321, type=dequantize_per_tensor]; -"322 linear_12_scale_0" [id=322, type=get_attr]; -"323 linear_12_zero_point_0" [id=323, type=get_attr]; -"324 _frozen_param13" [id=324, type=get_attr]; -"325 dequantize_per_channel_default_13" [id=325, type=dequantize_per_channel]; -"326 linear_12" [id=326, type=linear]; -"327 unflatten_3" [id=327, type=unflatten]; -"328 unsqueeze_3" [id=328, type=unsqueeze]; -"329 transpose_19" [id=329, type=transpose]; -"330 squeeze_3" [id=330, type=squeeze]; -"331 contiguous_3" [id=331, type=contiguous]; -"332 quantize_per_tensor_default_20" [id=332, type=quantize_per_tensor]; -"333 dequantize_per_tensor_default_20" [id=333, type=dequantize_per_tensor]; -"334 select_9" [id=334, type=select]; -"335 quantize_per_tensor_default_21" [id=335, type=quantize_per_tensor]; -"336 dequantize_per_tensor_default_21" [id=336, type=dequantize_per_tensor]; -"337 select_10" [id=337, type=select]; -"338 select_11" [id=338, type=select]; -"339 view_24" [id=339, type=view]; -"340 transpose_20" [id=340, type=transpose]; -"341 view_25" [id=341, type=view]; -"342 transpose_21" [id=342, type=transpose]; -"343 view_26" [id=343, type=view]; -"344 transpose_22" [id=344, type=transpose]; -"345 view_27" [id=345, type=view]; -"346 view_28" [id=346, type=view]; -"347 view_29" [id=347, type=view]; -"348 scaled_dot_product_attention_3" [id=348, type=scaled_dot_product_attention]; -"349 permute_4" [id=349, type=permute]; -"350 view_30" [id=350, type=view]; -"351 view_30_0_0_nncf_smooth_quant_0" [id=351, type=call_module]; -"352 quantize_per_tensor_default_22" [id=352, type=quantize_per_tensor]; -"353 dequantize_per_tensor_default_22" [id=353, type=dequantize_per_tensor]; -"354 linear_13_scale_0" [id=354, type=get_attr]; -"355 linear_13_zero_point_0" [id=355, type=get_attr]; -"356 _frozen_param14" [id=356, type=get_attr]; -"357 dequantize_per_channel_default_14" [id=357, type=dequantize_per_channel]; -"358 linear_13" [id=358, type=linear]; -"359 view_31" [id=359, type=view]; -"360 transpose_23" [id=360, type=transpose]; -"361 dropout_10" [id=361, type=dropout]; -"362 add_7" [id=362, type=add]; -"363 layer_norm_7" [id=363, type=layer_norm]; -"364 layer_norm_7_0_0_nncf_smooth_quant_0" [id=364, type=call_module]; -"365 quantize_per_tensor_default_23" [id=365, type=quantize_per_tensor]; -"366 dequantize_per_tensor_default_23" [id=366, type=dequantize_per_tensor]; -"367 linear_14_scale_0" [id=367, type=get_attr]; -"368 linear_14_zero_point_0" [id=368, type=get_attr]; -"369 _frozen_param15" [id=369, type=get_attr]; -"370 dequantize_per_channel_default_15" [id=370, type=dequantize_per_channel]; -"371 linear_14" [id=371, type=linear]; -"372 gelu_3" [id=372, type=gelu]; -"373 dropout_11" [id=373, type=dropout]; -"374 dropout_11_0_0_nncf_smooth_quant_0" [id=374, type=call_module]; -"375 quantize_per_tensor_default_24" [id=375, type=quantize_per_tensor]; -"376 dequantize_per_tensor_default_24" [id=376, type=dequantize_per_tensor]; -"377 linear_15_scale_0" [id=377, type=get_attr]; -"378 linear_15_zero_point_0" [id=378, type=get_attr]; -"379 _frozen_param16" [id=379, type=get_attr]; -"380 dequantize_per_channel_default_16" [id=380, type=dequantize_per_channel]; -"381 linear_15" [id=381, type=linear]; -"382 dropout_12" [id=382, type=dropout]; -"383 add_8" [id=383, type=add]; -"384 layer_norm_8" [id=384, type=layer_norm]; -"385 transpose_24" [id=385, type=transpose]; -"386 transpose_24_0_0_nncf_smooth_quant_0" [id=386, type=call_module]; -"387 quantize_per_tensor_default_25" [id=387, type=quantize_per_tensor]; -"388 dequantize_per_tensor_default_25" [id=388, type=dequantize_per_tensor]; -"389 linear_16_scale_0" [id=389, type=get_attr]; -"390 linear_16_zero_point_0" [id=390, type=get_attr]; -"391 _frozen_param17" [id=391, type=get_attr]; -"392 dequantize_per_channel_default_17" [id=392, type=dequantize_per_channel]; -"393 linear_16" [id=393, type=linear]; -"394 unflatten_4" [id=394, type=unflatten]; -"395 unsqueeze_4" [id=395, type=unsqueeze]; -"396 transpose_25" [id=396, type=transpose]; -"397 squeeze_4" [id=397, type=squeeze]; -"398 contiguous_4" [id=398, type=contiguous]; -"399 quantize_per_tensor_default_26" [id=399, type=quantize_per_tensor]; -"400 dequantize_per_tensor_default_26" [id=400, type=dequantize_per_tensor]; -"401 select_12" [id=401, type=select]; -"402 quantize_per_tensor_default_27" [id=402, type=quantize_per_tensor]; -"403 dequantize_per_tensor_default_27" [id=403, type=dequantize_per_tensor]; -"404 select_13" [id=404, type=select]; -"405 select_14" [id=405, type=select]; -"406 view_32" [id=406, type=view]; -"407 transpose_26" [id=407, type=transpose]; -"408 view_33" [id=408, type=view]; -"409 transpose_27" [id=409, type=transpose]; -"410 view_34" [id=410, type=view]; -"411 transpose_28" [id=411, type=transpose]; -"412 view_35" [id=412, type=view]; -"413 view_36" [id=413, type=view]; -"414 view_37" [id=414, type=view]; -"415 scaled_dot_product_attention_4" [id=415, type=scaled_dot_product_attention]; -"416 permute_5" [id=416, type=permute]; -"417 view_38" [id=417, type=view]; -"418 view_38_0_0_nncf_smooth_quant_0" [id=418, type=call_module]; -"419 quantize_per_tensor_default_28" [id=419, type=quantize_per_tensor]; -"420 dequantize_per_tensor_default_28" [id=420, type=dequantize_per_tensor]; -"421 linear_17_scale_0" [id=421, type=get_attr]; -"422 linear_17_zero_point_0" [id=422, type=get_attr]; -"423 _frozen_param18" [id=423, type=get_attr]; -"424 dequantize_per_channel_default_18" [id=424, type=dequantize_per_channel]; -"425 linear_17" [id=425, type=linear]; -"426 view_39" [id=426, type=view]; -"427 transpose_29" [id=427, type=transpose]; -"428 dropout_13" [id=428, type=dropout]; -"429 add_9" [id=429, type=add]; -"430 layer_norm_9" [id=430, type=layer_norm]; -"431 layer_norm_9_0_0_nncf_smooth_quant_0" [id=431, type=call_module]; -"432 quantize_per_tensor_default_29" [id=432, type=quantize_per_tensor]; -"433 dequantize_per_tensor_default_29" [id=433, type=dequantize_per_tensor]; -"434 linear_18_scale_0" [id=434, type=get_attr]; -"435 linear_18_zero_point_0" [id=435, type=get_attr]; -"436 _frozen_param19" [id=436, type=get_attr]; -"437 dequantize_per_channel_default_19" [id=437, type=dequantize_per_channel]; -"438 linear_18" [id=438, type=linear]; -"439 gelu_4" [id=439, type=gelu]; -"440 dropout_14" [id=440, type=dropout]; -"441 dropout_14_0_0_nncf_smooth_quant_0" [id=441, type=call_module]; -"442 quantize_per_tensor_default_30" [id=442, type=quantize_per_tensor]; -"443 dequantize_per_tensor_default_30" [id=443, type=dequantize_per_tensor]; -"444 linear_19_scale_0" [id=444, type=get_attr]; -"445 linear_19_zero_point_0" [id=445, type=get_attr]; -"446 _frozen_param20" [id=446, type=get_attr]; -"447 dequantize_per_channel_default_20" [id=447, type=dequantize_per_channel]; -"448 linear_19" [id=448, type=linear]; -"449 dropout_15" [id=449, type=dropout]; -"450 add_10" [id=450, type=add]; -"451 layer_norm_10" [id=451, type=layer_norm]; -"452 transpose_30" [id=452, type=transpose]; -"453 transpose_30_0_0_nncf_smooth_quant_0" [id=453, type=call_module]; -"454 quantize_per_tensor_default_31" [id=454, type=quantize_per_tensor]; -"455 dequantize_per_tensor_default_31" [id=455, type=dequantize_per_tensor]; -"456 linear_20_scale_0" [id=456, type=get_attr]; -"457 linear_20_zero_point_0" [id=457, type=get_attr]; -"458 _frozen_param21" [id=458, type=get_attr]; -"459 dequantize_per_channel_default_21" [id=459, type=dequantize_per_channel]; -"460 linear_20" [id=460, type=linear]; -"461 unflatten_5" [id=461, type=unflatten]; -"462 unsqueeze_5" [id=462, type=unsqueeze]; -"463 transpose_31" [id=463, type=transpose]; -"464 squeeze_5" [id=464, type=squeeze]; -"465 contiguous_5" [id=465, type=contiguous]; -"466 quantize_per_tensor_default_32" [id=466, type=quantize_per_tensor]; -"467 dequantize_per_tensor_default_32" [id=467, type=dequantize_per_tensor]; -"468 select_15" [id=468, type=select]; -"469 quantize_per_tensor_default_33" [id=469, type=quantize_per_tensor]; -"470 dequantize_per_tensor_default_33" [id=470, type=dequantize_per_tensor]; -"471 select_16" [id=471, type=select]; -"472 select_17" [id=472, type=select]; -"473 view_40" [id=473, type=view]; -"474 transpose_32" [id=474, type=transpose]; -"475 view_41" [id=475, type=view]; -"476 transpose_33" [id=476, type=transpose]; -"477 view_42" [id=477, type=view]; -"478 transpose_34" [id=478, type=transpose]; -"479 view_43" [id=479, type=view]; -"480 view_44" [id=480, type=view]; -"481 view_45" [id=481, type=view]; -"482 scaled_dot_product_attention_5" [id=482, type=scaled_dot_product_attention]; -"483 permute_6" [id=483, type=permute]; -"484 view_46" [id=484, type=view]; -"485 view_46_0_0_nncf_smooth_quant_0" [id=485, type=call_module]; -"486 quantize_per_tensor_default_34" [id=486, type=quantize_per_tensor]; -"487 dequantize_per_tensor_default_34" [id=487, type=dequantize_per_tensor]; -"488 linear_21_scale_0" [id=488, type=get_attr]; -"489 linear_21_zero_point_0" [id=489, type=get_attr]; -"490 _frozen_param22" [id=490, type=get_attr]; -"491 dequantize_per_channel_default_22" [id=491, type=dequantize_per_channel]; -"492 linear_21" [id=492, type=linear]; -"493 view_47" [id=493, type=view]; -"494 transpose_35" [id=494, type=transpose]; -"495 dropout_16" [id=495, type=dropout]; -"496 add_11" [id=496, type=add]; -"497 layer_norm_11" [id=497, type=layer_norm]; -"498 layer_norm_11_0_0_nncf_smooth_quant_0" [id=498, type=call_module]; -"499 quantize_per_tensor_default_35" [id=499, type=quantize_per_tensor]; -"500 dequantize_per_tensor_default_35" [id=500, type=dequantize_per_tensor]; -"501 linear_22_scale_0" [id=501, type=get_attr]; -"502 linear_22_zero_point_0" [id=502, type=get_attr]; -"503 _frozen_param23" [id=503, type=get_attr]; -"504 dequantize_per_channel_default_23" [id=504, type=dequantize_per_channel]; -"505 linear_22" [id=505, type=linear]; -"506 gelu_5" [id=506, type=gelu]; -"507 dropout_17" [id=507, type=dropout]; -"508 dropout_17_0_0_nncf_smooth_quant_0" [id=508, type=call_module]; -"509 quantize_per_tensor_default_36" [id=509, type=quantize_per_tensor]; -"510 dequantize_per_tensor_default_36" [id=510, type=dequantize_per_tensor]; -"511 linear_23_scale_0" [id=511, type=get_attr]; -"512 linear_23_zero_point_0" [id=512, type=get_attr]; -"513 _frozen_param24" [id=513, type=get_attr]; -"514 dequantize_per_channel_default_24" [id=514, type=dequantize_per_channel]; -"515 linear_23" [id=515, type=linear]; -"516 dropout_18" [id=516, type=dropout]; -"517 add_12" [id=517, type=add]; -"518 layer_norm_12" [id=518, type=layer_norm]; -"519 transpose_36" [id=519, type=transpose]; -"520 transpose_36_0_0_nncf_smooth_quant_0" [id=520, type=call_module]; -"521 quantize_per_tensor_default_37" [id=521, type=quantize_per_tensor]; -"522 dequantize_per_tensor_default_37" [id=522, type=dequantize_per_tensor]; -"523 linear_24_scale_0" [id=523, type=get_attr]; -"524 linear_24_zero_point_0" [id=524, type=get_attr]; -"525 _frozen_param25" [id=525, type=get_attr]; -"526 dequantize_per_channel_default_25" [id=526, type=dequantize_per_channel]; -"527 linear_24" [id=527, type=linear]; -"528 unflatten_6" [id=528, type=unflatten]; -"529 unsqueeze_6" [id=529, type=unsqueeze]; -"530 transpose_37" [id=530, type=transpose]; -"531 squeeze_6" [id=531, type=squeeze]; -"532 contiguous_6" [id=532, type=contiguous]; -"533 quantize_per_tensor_default_38" [id=533, type=quantize_per_tensor]; -"534 dequantize_per_tensor_default_38" [id=534, type=dequantize_per_tensor]; -"535 select_18" [id=535, type=select]; -"536 quantize_per_tensor_default_39" [id=536, type=quantize_per_tensor]; -"537 dequantize_per_tensor_default_39" [id=537, type=dequantize_per_tensor]; -"538 select_19" [id=538, type=select]; -"539 select_20" [id=539, type=select]; -"540 view_48" [id=540, type=view]; -"541 transpose_38" [id=541, type=transpose]; -"542 view_49" [id=542, type=view]; -"543 transpose_39" [id=543, type=transpose]; -"544 view_50" [id=544, type=view]; -"545 transpose_40" [id=545, type=transpose]; -"546 view_51" [id=546, type=view]; -"547 view_52" [id=547, type=view]; -"548 view_53" [id=548, type=view]; -"549 scaled_dot_product_attention_6" [id=549, type=scaled_dot_product_attention]; -"550 permute_7" [id=550, type=permute]; -"551 view_54" [id=551, type=view]; -"552 view_54_0_0_nncf_smooth_quant_0" [id=552, type=call_module]; -"553 quantize_per_tensor_default_40" [id=553, type=quantize_per_tensor]; -"554 dequantize_per_tensor_default_40" [id=554, type=dequantize_per_tensor]; -"555 linear_25_scale_0" [id=555, type=get_attr]; -"556 linear_25_zero_point_0" [id=556, type=get_attr]; -"557 _frozen_param26" [id=557, type=get_attr]; -"558 dequantize_per_channel_default_26" [id=558, type=dequantize_per_channel]; -"559 linear_25" [id=559, type=linear]; -"560 view_55" [id=560, type=view]; -"561 transpose_41" [id=561, type=transpose]; -"562 dropout_19" [id=562, type=dropout]; -"563 add_13" [id=563, type=add]; -"564 layer_norm_13" [id=564, type=layer_norm]; -"565 layer_norm_13_0_0_nncf_smooth_quant_0" [id=565, type=call_module]; -"566 quantize_per_tensor_default_41" [id=566, type=quantize_per_tensor]; -"567 dequantize_per_tensor_default_41" [id=567, type=dequantize_per_tensor]; -"568 linear_26_scale_0" [id=568, type=get_attr]; -"569 linear_26_zero_point_0" [id=569, type=get_attr]; -"570 _frozen_param27" [id=570, type=get_attr]; -"571 dequantize_per_channel_default_27" [id=571, type=dequantize_per_channel]; -"572 linear_26" [id=572, type=linear]; -"573 gelu_6" [id=573, type=gelu]; -"574 dropout_20" [id=574, type=dropout]; -"575 dropout_20_0_0_nncf_smooth_quant_0" [id=575, type=call_module]; -"576 quantize_per_tensor_default_42" [id=576, type=quantize_per_tensor]; -"577 dequantize_per_tensor_default_42" [id=577, type=dequantize_per_tensor]; -"578 linear_27_scale_0" [id=578, type=get_attr]; -"579 linear_27_zero_point_0" [id=579, type=get_attr]; -"580 _frozen_param28" [id=580, type=get_attr]; -"581 dequantize_per_channel_default_28" [id=581, type=dequantize_per_channel]; -"582 linear_27" [id=582, type=linear]; -"583 dropout_21" [id=583, type=dropout]; -"584 add_14" [id=584, type=add]; -"585 layer_norm_14" [id=585, type=layer_norm]; -"586 transpose_42" [id=586, type=transpose]; -"587 transpose_42_0_0_nncf_smooth_quant_0" [id=587, type=call_module]; -"588 quantize_per_tensor_default_43" [id=588, type=quantize_per_tensor]; -"589 dequantize_per_tensor_default_43" [id=589, type=dequantize_per_tensor]; -"590 linear_28_scale_0" [id=590, type=get_attr]; -"591 linear_28_zero_point_0" [id=591, type=get_attr]; -"592 _frozen_param29" [id=592, type=get_attr]; -"593 dequantize_per_channel_default_29" [id=593, type=dequantize_per_channel]; -"594 linear_28" [id=594, type=linear]; -"595 unflatten_7" [id=595, type=unflatten]; -"596 unsqueeze_7" [id=596, type=unsqueeze]; -"597 transpose_43" [id=597, type=transpose]; -"598 squeeze_7" [id=598, type=squeeze]; -"599 contiguous_7" [id=599, type=contiguous]; -"600 quantize_per_tensor_default_44" [id=600, type=quantize_per_tensor]; -"601 dequantize_per_tensor_default_44" [id=601, type=dequantize_per_tensor]; -"602 select_21" [id=602, type=select]; -"603 quantize_per_tensor_default_45" [id=603, type=quantize_per_tensor]; -"604 dequantize_per_tensor_default_45" [id=604, type=dequantize_per_tensor]; -"605 select_22" [id=605, type=select]; -"606 select_23" [id=606, type=select]; -"607 view_56" [id=607, type=view]; -"608 transpose_44" [id=608, type=transpose]; -"609 view_57" [id=609, type=view]; -"610 transpose_45" [id=610, type=transpose]; -"611 view_58" [id=611, type=view]; -"612 transpose_46" [id=612, type=transpose]; -"613 view_59" [id=613, type=view]; -"614 view_60" [id=614, type=view]; -"615 view_61" [id=615, type=view]; -"616 scaled_dot_product_attention_7" [id=616, type=scaled_dot_product_attention]; -"617 permute_8" [id=617, type=permute]; -"618 view_62" [id=618, type=view]; -"619 view_62_0_0_nncf_smooth_quant_0" [id=619, type=call_module]; -"620 quantize_per_tensor_default_46" [id=620, type=quantize_per_tensor]; -"621 dequantize_per_tensor_default_46" [id=621, type=dequantize_per_tensor]; -"622 linear_29_scale_0" [id=622, type=get_attr]; -"623 linear_29_zero_point_0" [id=623, type=get_attr]; -"624 _frozen_param30" [id=624, type=get_attr]; -"625 dequantize_per_channel_default_30" [id=625, type=dequantize_per_channel]; -"626 linear_29" [id=626, type=linear]; -"627 view_63" [id=627, type=view]; -"628 transpose_47" [id=628, type=transpose]; -"629 dropout_22" [id=629, type=dropout]; -"630 add_15" [id=630, type=add]; -"631 layer_norm_15" [id=631, type=layer_norm]; -"632 layer_norm_15_0_0_nncf_smooth_quant_0" [id=632, type=call_module]; -"633 quantize_per_tensor_default_47" [id=633, type=quantize_per_tensor]; -"634 dequantize_per_tensor_default_47" [id=634, type=dequantize_per_tensor]; -"635 linear_30_scale_0" [id=635, type=get_attr]; -"636 linear_30_zero_point_0" [id=636, type=get_attr]; -"637 _frozen_param31" [id=637, type=get_attr]; -"638 dequantize_per_channel_default_31" [id=638, type=dequantize_per_channel]; -"639 linear_30" [id=639, type=linear]; -"640 gelu_7" [id=640, type=gelu]; -"641 dropout_23" [id=641, type=dropout]; -"642 dropout_23_0_0_nncf_smooth_quant_0" [id=642, type=call_module]; -"643 quantize_per_tensor_default_48" [id=643, type=quantize_per_tensor]; -"644 dequantize_per_tensor_default_48" [id=644, type=dequantize_per_tensor]; -"645 linear_31_scale_0" [id=645, type=get_attr]; -"646 linear_31_zero_point_0" [id=646, type=get_attr]; -"647 _frozen_param32" [id=647, type=get_attr]; -"648 dequantize_per_channel_default_32" [id=648, type=dequantize_per_channel]; -"649 linear_31" [id=649, type=linear]; -"650 dropout_24" [id=650, type=dropout]; -"651 add_16" [id=651, type=add]; -"652 layer_norm_16" [id=652, type=layer_norm]; -"653 transpose_48" [id=653, type=transpose]; -"654 transpose_48_0_0_nncf_smooth_quant_0" [id=654, type=call_module]; -"655 quantize_per_tensor_default_49" [id=655, type=quantize_per_tensor]; -"656 dequantize_per_tensor_default_49" [id=656, type=dequantize_per_tensor]; -"657 linear_32_scale_0" [id=657, type=get_attr]; -"658 linear_32_zero_point_0" [id=658, type=get_attr]; -"659 _frozen_param33" [id=659, type=get_attr]; -"660 dequantize_per_channel_default_33" [id=660, type=dequantize_per_channel]; -"661 linear_32" [id=661, type=linear]; -"662 unflatten_8" [id=662, type=unflatten]; -"663 unsqueeze_8" [id=663, type=unsqueeze]; -"664 transpose_49" [id=664, type=transpose]; -"665 squeeze_8" [id=665, type=squeeze]; -"666 contiguous_8" [id=666, type=contiguous]; -"667 quantize_per_tensor_default_50" [id=667, type=quantize_per_tensor]; -"668 dequantize_per_tensor_default_50" [id=668, type=dequantize_per_tensor]; -"669 select_24" [id=669, type=select]; -"670 quantize_per_tensor_default_51" [id=670, type=quantize_per_tensor]; -"671 dequantize_per_tensor_default_51" [id=671, type=dequantize_per_tensor]; -"672 select_25" [id=672, type=select]; -"673 select_26" [id=673, type=select]; -"674 view_64" [id=674, type=view]; -"675 transpose_50" [id=675, type=transpose]; -"676 view_65" [id=676, type=view]; -"677 transpose_51" [id=677, type=transpose]; -"678 view_66" [id=678, type=view]; -"679 transpose_52" [id=679, type=transpose]; -"680 view_67" [id=680, type=view]; -"681 view_68" [id=681, type=view]; -"682 view_69" [id=682, type=view]; -"683 scaled_dot_product_attention_8" [id=683, type=scaled_dot_product_attention]; -"684 permute_9" [id=684, type=permute]; -"685 view_70" [id=685, type=view]; -"686 view_70_0_0_nncf_smooth_quant_0" [id=686, type=call_module]; -"687 quantize_per_tensor_default_52" [id=687, type=quantize_per_tensor]; -"688 dequantize_per_tensor_default_52" [id=688, type=dequantize_per_tensor]; -"689 linear_33_scale_0" [id=689, type=get_attr]; -"690 linear_33_zero_point_0" [id=690, type=get_attr]; -"691 _frozen_param34" [id=691, type=get_attr]; -"692 dequantize_per_channel_default_34" [id=692, type=dequantize_per_channel]; -"693 linear_33" [id=693, type=linear]; -"694 view_71" [id=694, type=view]; -"695 transpose_53" [id=695, type=transpose]; -"696 dropout_25" [id=696, type=dropout]; -"697 add_17" [id=697, type=add]; -"698 layer_norm_17" [id=698, type=layer_norm]; -"699 layer_norm_17_0_0_nncf_smooth_quant_0" [id=699, type=call_module]; -"700 quantize_per_tensor_default_53" [id=700, type=quantize_per_tensor]; -"701 dequantize_per_tensor_default_53" [id=701, type=dequantize_per_tensor]; -"702 linear_34_scale_0" [id=702, type=get_attr]; -"703 linear_34_zero_point_0" [id=703, type=get_attr]; -"704 _frozen_param35" [id=704, type=get_attr]; -"705 dequantize_per_channel_default_35" [id=705, type=dequantize_per_channel]; -"706 linear_34" [id=706, type=linear]; -"707 gelu_8" [id=707, type=gelu]; -"708 dropout_26" [id=708, type=dropout]; -"709 dropout_26_0_0_nncf_smooth_quant_0" [id=709, type=call_module]; -"710 quantize_per_tensor_default_54" [id=710, type=quantize_per_tensor]; -"711 dequantize_per_tensor_default_54" [id=711, type=dequantize_per_tensor]; -"712 linear_35_scale_0" [id=712, type=get_attr]; -"713 linear_35_zero_point_0" [id=713, type=get_attr]; -"714 _frozen_param36" [id=714, type=get_attr]; -"715 dequantize_per_channel_default_36" [id=715, type=dequantize_per_channel]; -"716 linear_35" [id=716, type=linear]; -"717 dropout_27" [id=717, type=dropout]; -"718 add_18" [id=718, type=add]; -"719 layer_norm_18" [id=719, type=layer_norm]; -"720 transpose_54" [id=720, type=transpose]; -"721 transpose_54_0_0_nncf_smooth_quant_0" [id=721, type=call_module]; -"722 quantize_per_tensor_default_55" [id=722, type=quantize_per_tensor]; -"723 dequantize_per_tensor_default_55" [id=723, type=dequantize_per_tensor]; -"724 linear_36_scale_0" [id=724, type=get_attr]; -"725 linear_36_zero_point_0" [id=725, type=get_attr]; -"726 _frozen_param37" [id=726, type=get_attr]; -"727 dequantize_per_channel_default_37" [id=727, type=dequantize_per_channel]; -"728 linear_36" [id=728, type=linear]; -"729 unflatten_9" [id=729, type=unflatten]; -"730 unsqueeze_9" [id=730, type=unsqueeze]; -"731 transpose_55" [id=731, type=transpose]; -"732 squeeze_9" [id=732, type=squeeze]; -"733 contiguous_9" [id=733, type=contiguous]; -"734 quantize_per_tensor_default_56" [id=734, type=quantize_per_tensor]; -"735 dequantize_per_tensor_default_56" [id=735, type=dequantize_per_tensor]; -"736 select_27" [id=736, type=select]; -"737 quantize_per_tensor_default_57" [id=737, type=quantize_per_tensor]; -"738 dequantize_per_tensor_default_57" [id=738, type=dequantize_per_tensor]; -"739 select_28" [id=739, type=select]; -"740 select_29" [id=740, type=select]; -"741 view_72" [id=741, type=view]; -"742 transpose_56" [id=742, type=transpose]; -"743 view_73" [id=743, type=view]; -"744 transpose_57" [id=744, type=transpose]; -"745 view_74" [id=745, type=view]; -"746 transpose_58" [id=746, type=transpose]; -"747 view_75" [id=747, type=view]; -"748 view_76" [id=748, type=view]; -"749 view_77" [id=749, type=view]; -"750 scaled_dot_product_attention_9" [id=750, type=scaled_dot_product_attention]; -"751 permute_10" [id=751, type=permute]; -"752 view_78" [id=752, type=view]; -"753 view_78_0_0_nncf_smooth_quant_0" [id=753, type=call_module]; -"754 quantize_per_tensor_default_58" [id=754, type=quantize_per_tensor]; -"755 dequantize_per_tensor_default_58" [id=755, type=dequantize_per_tensor]; -"756 linear_37_scale_0" [id=756, type=get_attr]; -"757 linear_37_zero_point_0" [id=757, type=get_attr]; -"758 _frozen_param38" [id=758, type=get_attr]; -"759 dequantize_per_channel_default_38" [id=759, type=dequantize_per_channel]; -"760 linear_37" [id=760, type=linear]; -"761 view_79" [id=761, type=view]; -"762 transpose_59" [id=762, type=transpose]; -"763 dropout_28" [id=763, type=dropout]; -"764 add_19" [id=764, type=add]; -"765 layer_norm_19" [id=765, type=layer_norm]; -"766 layer_norm_19_0_0_nncf_smooth_quant_0" [id=766, type=call_module]; -"767 quantize_per_tensor_default_59" [id=767, type=quantize_per_tensor]; -"768 dequantize_per_tensor_default_59" [id=768, type=dequantize_per_tensor]; -"769 linear_38_scale_0" [id=769, type=get_attr]; -"770 linear_38_zero_point_0" [id=770, type=get_attr]; -"771 _frozen_param39" [id=771, type=get_attr]; -"772 dequantize_per_channel_default_39" [id=772, type=dequantize_per_channel]; -"773 linear_38" [id=773, type=linear]; -"774 gelu_9" [id=774, type=gelu]; -"775 dropout_29" [id=775, type=dropout]; -"776 dropout_29_0_0_nncf_smooth_quant_0" [id=776, type=call_module]; -"777 quantize_per_tensor_default_60" [id=777, type=quantize_per_tensor]; -"778 dequantize_per_tensor_default_60" [id=778, type=dequantize_per_tensor]; -"779 linear_39_scale_0" [id=779, type=get_attr]; -"780 linear_39_zero_point_0" [id=780, type=get_attr]; -"781 _frozen_param40" [id=781, type=get_attr]; -"782 dequantize_per_channel_default_40" [id=782, type=dequantize_per_channel]; -"783 linear_39" [id=783, type=linear]; -"784 dropout_30" [id=784, type=dropout]; -"785 add_20" [id=785, type=add]; -"786 layer_norm_20" [id=786, type=layer_norm]; -"787 transpose_60" [id=787, type=transpose]; -"788 transpose_60_0_0_nncf_smooth_quant_0" [id=788, type=call_module]; -"789 quantize_per_tensor_default_61" [id=789, type=quantize_per_tensor]; -"790 dequantize_per_tensor_default_61" [id=790, type=dequantize_per_tensor]; -"791 linear_40_scale_0" [id=791, type=get_attr]; -"792 linear_40_zero_point_0" [id=792, type=get_attr]; -"793 _frozen_param41" [id=793, type=get_attr]; -"794 dequantize_per_channel_default_41" [id=794, type=dequantize_per_channel]; -"795 linear_40" [id=795, type=linear]; -"796 unflatten_10" [id=796, type=unflatten]; -"797 unsqueeze_10" [id=797, type=unsqueeze]; -"798 transpose_61" [id=798, type=transpose]; -"799 squeeze_10" [id=799, type=squeeze]; -"800 contiguous_10" [id=800, type=contiguous]; -"801 quantize_per_tensor_default_62" [id=801, type=quantize_per_tensor]; -"802 dequantize_per_tensor_default_62" [id=802, type=dequantize_per_tensor]; -"803 select_30" [id=803, type=select]; -"804 quantize_per_tensor_default_63" [id=804, type=quantize_per_tensor]; -"805 dequantize_per_tensor_default_63" [id=805, type=dequantize_per_tensor]; -"806 select_31" [id=806, type=select]; -"807 select_32" [id=807, type=select]; -"808 view_80" [id=808, type=view]; -"809 transpose_62" [id=809, type=transpose]; -"810 view_81" [id=810, type=view]; -"811 transpose_63" [id=811, type=transpose]; -"812 view_82" [id=812, type=view]; -"813 transpose_64" [id=813, type=transpose]; -"814 view_83" [id=814, type=view]; -"815 view_84" [id=815, type=view]; -"816 view_85" [id=816, type=view]; -"817 scaled_dot_product_attention_10" [id=817, type=scaled_dot_product_attention]; -"818 permute_11" [id=818, type=permute]; -"819 view_86" [id=819, type=view]; -"820 view_86_0_0_nncf_smooth_quant_0" [id=820, type=call_module]; -"821 quantize_per_tensor_default_64" [id=821, type=quantize_per_tensor]; -"822 dequantize_per_tensor_default_64" [id=822, type=dequantize_per_tensor]; -"823 linear_41_scale_0" [id=823, type=get_attr]; -"824 linear_41_zero_point_0" [id=824, type=get_attr]; -"825 _frozen_param42" [id=825, type=get_attr]; -"826 dequantize_per_channel_default_42" [id=826, type=dequantize_per_channel]; -"827 linear_41" [id=827, type=linear]; -"828 view_87" [id=828, type=view]; -"829 transpose_65" [id=829, type=transpose]; -"830 dropout_31" [id=830, type=dropout]; -"831 add_21" [id=831, type=add]; -"832 layer_norm_21" [id=832, type=layer_norm]; -"833 layer_norm_21_0_0_nncf_smooth_quant_0" [id=833, type=call_module]; -"834 quantize_per_tensor_default_65" [id=834, type=quantize_per_tensor]; -"835 dequantize_per_tensor_default_65" [id=835, type=dequantize_per_tensor]; -"836 linear_42_scale_0" [id=836, type=get_attr]; -"837 linear_42_zero_point_0" [id=837, type=get_attr]; -"838 _frozen_param43" [id=838, type=get_attr]; -"839 dequantize_per_channel_default_43" [id=839, type=dequantize_per_channel]; -"840 linear_42" [id=840, type=linear]; -"841 gelu_10" [id=841, type=gelu]; -"842 dropout_32" [id=842, type=dropout]; -"843 dropout_32_0_0_nncf_smooth_quant_0" [id=843, type=call_module]; -"844 quantize_per_tensor_default_66" [id=844, type=quantize_per_tensor]; -"845 dequantize_per_tensor_default_66" [id=845, type=dequantize_per_tensor]; -"846 linear_43_scale_0" [id=846, type=get_attr]; -"847 linear_43_zero_point_0" [id=847, type=get_attr]; -"848 _frozen_param44" [id=848, type=get_attr]; -"849 dequantize_per_channel_default_44" [id=849, type=dequantize_per_channel]; -"850 linear_43" [id=850, type=linear]; -"851 dropout_33" [id=851, type=dropout]; -"852 add_22" [id=852, type=add]; -"853 layer_norm_22" [id=853, type=layer_norm]; -"854 transpose_66" [id=854, type=transpose]; -"855 transpose_66_0_0_nncf_smooth_quant_0" [id=855, type=call_module]; -"856 quantize_per_tensor_default_67" [id=856, type=quantize_per_tensor]; -"857 dequantize_per_tensor_default_67" [id=857, type=dequantize_per_tensor]; -"858 linear_44_scale_0" [id=858, type=get_attr]; -"859 linear_44_zero_point_0" [id=859, type=get_attr]; -"860 _frozen_param45" [id=860, type=get_attr]; -"861 dequantize_per_channel_default_45" [id=861, type=dequantize_per_channel]; -"862 linear_44" [id=862, type=linear]; -"863 unflatten_11" [id=863, type=unflatten]; -"864 unsqueeze_11" [id=864, type=unsqueeze]; -"865 transpose_67" [id=865, type=transpose]; -"866 squeeze_11" [id=866, type=squeeze]; -"867 contiguous_11" [id=867, type=contiguous]; -"868 quantize_per_tensor_default_68" [id=868, type=quantize_per_tensor]; -"869 dequantize_per_tensor_default_68" [id=869, type=dequantize_per_tensor]; -"870 select_33" [id=870, type=select]; -"871 quantize_per_tensor_default_69" [id=871, type=quantize_per_tensor]; -"872 dequantize_per_tensor_default_69" [id=872, type=dequantize_per_tensor]; -"873 select_34" [id=873, type=select]; -"874 select_35" [id=874, type=select]; -"875 view_88" [id=875, type=view]; -"876 transpose_68" [id=876, type=transpose]; -"877 view_89" [id=877, type=view]; -"878 transpose_69" [id=878, type=transpose]; -"879 view_90" [id=879, type=view]; -"880 transpose_70" [id=880, type=transpose]; -"881 view_91" [id=881, type=view]; -"882 view_92" [id=882, type=view]; -"883 view_93" [id=883, type=view]; -"884 scaled_dot_product_attention_11" [id=884, type=scaled_dot_product_attention]; -"885 permute_12" [id=885, type=permute]; -"886 view_94" [id=886, type=view]; -"887 view_94_0_0_nncf_smooth_quant_0" [id=887, type=call_module]; -"888 quantize_per_tensor_default_70" [id=888, type=quantize_per_tensor]; -"889 dequantize_per_tensor_default_70" [id=889, type=dequantize_per_tensor]; -"890 linear_45_scale_0" [id=890, type=get_attr]; -"891 linear_45_zero_point_0" [id=891, type=get_attr]; -"892 _frozen_param46" [id=892, type=get_attr]; -"893 dequantize_per_channel_default_46" [id=893, type=dequantize_per_channel]; -"894 linear_45" [id=894, type=linear]; -"895 view_95" [id=895, type=view]; -"896 transpose_71" [id=896, type=transpose]; -"897 dropout_34" [id=897, type=dropout]; -"898 add_23" [id=898, type=add]; -"899 layer_norm_23" [id=899, type=layer_norm]; -"900 layer_norm_23_0_0_nncf_smooth_quant_0" [id=900, type=call_module]; -"901 quantize_per_tensor_default_71" [id=901, type=quantize_per_tensor]; -"902 dequantize_per_tensor_default_71" [id=902, type=dequantize_per_tensor]; -"903 linear_46_scale_0" [id=903, type=get_attr]; -"904 linear_46_zero_point_0" [id=904, type=get_attr]; -"905 _frozen_param47" [id=905, type=get_attr]; -"906 dequantize_per_channel_default_47" [id=906, type=dequantize_per_channel]; -"907 linear_46" [id=907, type=linear]; -"908 gelu_11" [id=908, type=gelu]; -"909 dropout_35" [id=909, type=dropout]; -"910 dropout_35_0_0_nncf_smooth_quant_0" [id=910, type=call_module]; -"911 quantize_per_tensor_default_72" [id=911, type=quantize_per_tensor]; -"912 dequantize_per_tensor_default_72" [id=912, type=dequantize_per_tensor]; -"913 linear_47_scale_0" [id=913, type=get_attr]; -"914 linear_47_zero_point_0" [id=914, type=get_attr]; -"915 _frozen_param48" [id=915, type=get_attr]; -"916 dequantize_per_channel_default_48" [id=916, type=dequantize_per_channel]; -"917 linear_47" [id=917, type=linear]; -"918 dropout_36" [id=918, type=dropout]; -"919 add_24" [id=919, type=add]; -"920 layer_norm_24" [id=920, type=layer_norm]; -"921 slice_1" [id=921, type=slice]; -"922 select_36" [id=922, type=select]; -"923 select_36_0_0_nncf_smooth_quant_0" [id=923, type=call_module]; -"924 quantize_per_tensor_default_73" [id=924, type=quantize_per_tensor]; -"925 dequantize_per_tensor_default_73" [id=925, type=dequantize_per_tensor]; -"926 linear_48_scale_0" [id=926, type=get_attr]; -"927 linear_48_zero_point_0" [id=927, type=get_attr]; -"928 _frozen_param49" [id=928, type=get_attr]; -"929 dequantize_per_channel_default_49" [id=929, type=dequantize_per_channel]; -"930 linear_48" [id=930, type=linear]; -"931 output" [id=931, type=output]; -"0 class_token" -> "112 expand" [label="(1, 1, 768)", style=solid]; -"1 encoder_pos_embedding" -> "114 add" [label="(1, 197, 768)", style=solid]; -"2 conv_proj_bias" -> "109 conv2d" [label="(768,)", style=solid]; -"3 encoder_layers_encoder_layer_0_ln_1_weight" -> "116 layer_norm" [label="(768,)", style=solid]; -"4 encoder_layers_encoder_layer_0_ln_1_bias" -> "116 layer_norm" [label="(768,)", style=solid]; -"5 encoder_layers_encoder_layer_0_self_attention_in_proj_bias" -> "125 linear" [label="(2304,)", style=solid]; -"6 encoder_layers_encoder_layer_0_self_attention_out_proj_bias" -> "157 linear_1" [label="(768,)", style=solid]; -"7 encoder_layers_encoder_layer_0_ln_2_weight" -> "162 layer_norm_1" [label="(768,)", style=solid]; -"8 encoder_layers_encoder_layer_0_ln_2_bias" -> "162 layer_norm_1" [label="(768,)", style=solid]; -"9 encoder_layers_encoder_layer_0_mlp_0_bias" -> "170 linear_2" [label="(3072,)", style=solid]; -"10 encoder_layers_encoder_layer_0_mlp_3_bias" -> "180 linear_3" [label="(768,)", style=solid]; -"11 encoder_layers_encoder_layer_1_ln_1_weight" -> "183 layer_norm_2" [label="(768,)", style=solid]; -"12 encoder_layers_encoder_layer_1_ln_1_bias" -> "183 layer_norm_2" [label="(768,)", style=solid]; -"13 encoder_layers_encoder_layer_1_self_attention_in_proj_bias" -> "192 linear_4" [label="(2304,)", style=solid]; -"14 encoder_layers_encoder_layer_1_self_attention_out_proj_bias" -> "224 linear_5" [label="(768,)", style=solid]; -"15 encoder_layers_encoder_layer_1_ln_2_weight" -> "229 layer_norm_3" [label="(768,)", style=solid]; -"16 encoder_layers_encoder_layer_1_ln_2_bias" -> "229 layer_norm_3" [label="(768,)", style=solid]; -"17 encoder_layers_encoder_layer_1_mlp_0_bias" -> "237 linear_6" [label="(3072,)", style=solid]; -"18 encoder_layers_encoder_layer_1_mlp_3_bias" -> "247 linear_7" [label="(768,)", style=solid]; -"19 encoder_layers_encoder_layer_2_ln_1_weight" -> "250 layer_norm_4" [label="(768,)", style=solid]; -"20 encoder_layers_encoder_layer_2_ln_1_bias" -> "250 layer_norm_4" [label="(768,)", style=solid]; -"21 encoder_layers_encoder_layer_2_self_attention_in_proj_bias" -> "259 linear_8" [label="(2304,)", style=solid]; -"22 encoder_layers_encoder_layer_2_self_attention_out_proj_bias" -> "291 linear_9" [label="(768,)", style=solid]; -"23 encoder_layers_encoder_layer_2_ln_2_weight" -> "296 layer_norm_5" [label="(768,)", style=solid]; -"24 encoder_layers_encoder_layer_2_ln_2_bias" -> "296 layer_norm_5" [label="(768,)", style=solid]; -"25 encoder_layers_encoder_layer_2_mlp_0_bias" -> "304 linear_10" [label="(3072,)", style=solid]; -"26 encoder_layers_encoder_layer_2_mlp_3_bias" -> "314 linear_11" [label="(768,)", style=solid]; -"27 encoder_layers_encoder_layer_3_ln_1_weight" -> "317 layer_norm_6" [label="(768,)", style=solid]; -"28 encoder_layers_encoder_layer_3_ln_1_bias" -> "317 layer_norm_6" [label="(768,)", style=solid]; -"29 encoder_layers_encoder_layer_3_self_attention_in_proj_bias" -> "326 linear_12" [label="(2304,)", style=solid]; -"30 encoder_layers_encoder_layer_3_self_attention_out_proj_bias" -> "358 linear_13" [label="(768,)", style=solid]; -"31 encoder_layers_encoder_layer_3_ln_2_weight" -> "363 layer_norm_7" [label="(768,)", style=solid]; -"32 encoder_layers_encoder_layer_3_ln_2_bias" -> "363 layer_norm_7" [label="(768,)", style=solid]; -"33 encoder_layers_encoder_layer_3_mlp_0_bias" -> "371 linear_14" [label="(3072,)", style=solid]; -"34 encoder_layers_encoder_layer_3_mlp_3_bias" -> "381 linear_15" [label="(768,)", style=solid]; -"35 encoder_layers_encoder_layer_4_ln_1_weight" -> "384 layer_norm_8" [label="(768,)", style=solid]; -"36 encoder_layers_encoder_layer_4_ln_1_bias" -> "384 layer_norm_8" [label="(768,)", style=solid]; -"37 encoder_layers_encoder_layer_4_self_attention_in_proj_bias" -> "393 linear_16" [label="(2304,)", style=solid]; -"38 encoder_layers_encoder_layer_4_self_attention_out_proj_bias" -> "425 linear_17" [label="(768,)", style=solid]; -"39 encoder_layers_encoder_layer_4_ln_2_weight" -> "430 layer_norm_9" [label="(768,)", style=solid]; -"40 encoder_layers_encoder_layer_4_ln_2_bias" -> "430 layer_norm_9" [label="(768,)", style=solid]; -"41 encoder_layers_encoder_layer_4_mlp_0_bias" -> "438 linear_18" [label="(3072,)", style=solid]; -"42 encoder_layers_encoder_layer_4_mlp_3_bias" -> "448 linear_19" [label="(768,)", style=solid]; -"43 encoder_layers_encoder_layer_5_ln_1_weight" -> "451 layer_norm_10" [label="(768,)", style=solid]; -"44 encoder_layers_encoder_layer_5_ln_1_bias" -> "451 layer_norm_10" [label="(768,)", style=solid]; -"45 encoder_layers_encoder_layer_5_self_attention_in_proj_bias" -> "460 linear_20" [label="(2304,)", style=solid]; -"46 encoder_layers_encoder_layer_5_self_attention_out_proj_bias" -> "492 linear_21" [label="(768,)", style=solid]; -"47 encoder_layers_encoder_layer_5_ln_2_weight" -> "497 layer_norm_11" [label="(768,)", style=solid]; -"48 encoder_layers_encoder_layer_5_ln_2_bias" -> "497 layer_norm_11" [label="(768,)", style=solid]; -"49 encoder_layers_encoder_layer_5_mlp_0_bias" -> "505 linear_22" [label="(3072,)", style=solid]; -"50 encoder_layers_encoder_layer_5_mlp_3_bias" -> "515 linear_23" [label="(768,)", style=solid]; -"51 encoder_layers_encoder_layer_6_ln_1_weight" -> "518 layer_norm_12" [label="(768,)", style=solid]; -"52 encoder_layers_encoder_layer_6_ln_1_bias" -> "518 layer_norm_12" [label="(768,)", style=solid]; -"53 encoder_layers_encoder_layer_6_self_attention_in_proj_bias" -> "527 linear_24" [label="(2304,)", style=solid]; -"54 encoder_layers_encoder_layer_6_self_attention_out_proj_bias" -> "559 linear_25" [label="(768,)", style=solid]; -"55 encoder_layers_encoder_layer_6_ln_2_weight" -> "564 layer_norm_13" [label="(768,)", style=solid]; -"56 encoder_layers_encoder_layer_6_ln_2_bias" -> "564 layer_norm_13" [label="(768,)", style=solid]; -"57 encoder_layers_encoder_layer_6_mlp_0_bias" -> "572 linear_26" [label="(3072,)", style=solid]; -"58 encoder_layers_encoder_layer_6_mlp_3_bias" -> "582 linear_27" [label="(768,)", style=solid]; -"59 encoder_layers_encoder_layer_7_ln_1_weight" -> "585 layer_norm_14" [label="(768,)", style=solid]; -"60 encoder_layers_encoder_layer_7_ln_1_bias" -> "585 layer_norm_14" [label="(768,)", style=solid]; -"61 encoder_layers_encoder_layer_7_self_attention_in_proj_bias" -> "594 linear_28" [label="(2304,)", style=solid]; -"62 encoder_layers_encoder_layer_7_self_attention_out_proj_bias" -> "626 linear_29" [label="(768,)", style=solid]; -"63 encoder_layers_encoder_layer_7_ln_2_weight" -> "631 layer_norm_15" [label="(768,)", style=solid]; -"64 encoder_layers_encoder_layer_7_ln_2_bias" -> "631 layer_norm_15" [label="(768,)", style=solid]; -"65 encoder_layers_encoder_layer_7_mlp_0_bias" -> "639 linear_30" [label="(3072,)", style=solid]; -"66 encoder_layers_encoder_layer_7_mlp_3_bias" -> "649 linear_31" [label="(768,)", style=solid]; -"67 encoder_layers_encoder_layer_8_ln_1_weight" -> "652 layer_norm_16" [label="(768,)", style=solid]; -"68 encoder_layers_encoder_layer_8_ln_1_bias" -> "652 layer_norm_16" [label="(768,)", style=solid]; -"69 encoder_layers_encoder_layer_8_self_attention_in_proj_bias" -> "661 linear_32" [label="(2304,)", style=solid]; -"70 encoder_layers_encoder_layer_8_self_attention_out_proj_bias" -> "693 linear_33" [label="(768,)", style=solid]; -"71 encoder_layers_encoder_layer_8_ln_2_weight" -> "698 layer_norm_17" [label="(768,)", style=solid]; -"72 encoder_layers_encoder_layer_8_ln_2_bias" -> "698 layer_norm_17" [label="(768,)", style=solid]; -"73 encoder_layers_encoder_layer_8_mlp_0_bias" -> "706 linear_34" [label="(3072,)", style=solid]; -"74 encoder_layers_encoder_layer_8_mlp_3_bias" -> "716 linear_35" [label="(768,)", style=solid]; -"75 encoder_layers_encoder_layer_9_ln_1_weight" -> "719 layer_norm_18" [label="(768,)", style=solid]; -"76 encoder_layers_encoder_layer_9_ln_1_bias" -> "719 layer_norm_18" [label="(768,)", style=solid]; -"77 encoder_layers_encoder_layer_9_self_attention_in_proj_bias" -> "728 linear_36" [label="(2304,)", style=solid]; -"78 encoder_layers_encoder_layer_9_self_attention_out_proj_bias" -> "760 linear_37" [label="(768,)", style=solid]; -"79 encoder_layers_encoder_layer_9_ln_2_weight" -> "765 layer_norm_19" [label="(768,)", style=solid]; -"80 encoder_layers_encoder_layer_9_ln_2_bias" -> "765 layer_norm_19" [label="(768,)", style=solid]; -"81 encoder_layers_encoder_layer_9_mlp_0_bias" -> "773 linear_38" [label="(3072,)", style=solid]; -"82 encoder_layers_encoder_layer_9_mlp_3_bias" -> "783 linear_39" [label="(768,)", style=solid]; -"83 encoder_layers_encoder_layer_10_ln_1_weight" -> "786 layer_norm_20" [label="(768,)", style=solid]; -"84 encoder_layers_encoder_layer_10_ln_1_bias" -> "786 layer_norm_20" [label="(768,)", style=solid]; -"85 encoder_layers_encoder_layer_10_self_attention_in_proj_bias" -> "795 linear_40" [label="(2304,)", style=solid]; -"86 encoder_layers_encoder_layer_10_self_attention_out_proj_bias" -> "827 linear_41" [label="(768,)", style=solid]; -"87 encoder_layers_encoder_layer_10_ln_2_weight" -> "832 layer_norm_21" [label="(768,)", style=solid]; -"88 encoder_layers_encoder_layer_10_ln_2_bias" -> "832 layer_norm_21" [label="(768,)", style=solid]; -"89 encoder_layers_encoder_layer_10_mlp_0_bias" -> "840 linear_42" [label="(3072,)", style=solid]; -"90 encoder_layers_encoder_layer_10_mlp_3_bias" -> "850 linear_43" [label="(768,)", style=solid]; -"91 encoder_layers_encoder_layer_11_ln_1_weight" -> "853 layer_norm_22" [label="(768,)", style=solid]; -"92 encoder_layers_encoder_layer_11_ln_1_bias" -> "853 layer_norm_22" [label="(768,)", style=solid]; -"93 encoder_layers_encoder_layer_11_self_attention_in_proj_bias" -> "862 linear_44" [label="(2304,)", style=solid]; -"94 encoder_layers_encoder_layer_11_self_attention_out_proj_bias" -> "894 linear_45" [label="(768,)", style=solid]; -"95 encoder_layers_encoder_layer_11_ln_2_weight" -> "899 layer_norm_23" [label="(768,)", style=solid]; -"96 encoder_layers_encoder_layer_11_ln_2_bias" -> "899 layer_norm_23" [label="(768,)", style=solid]; -"97 encoder_layers_encoder_layer_11_mlp_0_bias" -> "907 linear_46" [label="(3072,)", style=solid]; -"98 encoder_layers_encoder_layer_11_mlp_3_bias" -> "917 linear_47" [label="(768,)", style=solid]; -"99 encoder_ln_weight" -> "920 layer_norm_24" [label="(768,)", style=solid]; -"100 encoder_ln_bias" -> "920 layer_norm_24" [label="(768,)", style=solid]; -"101 heads_head_bias" -> "930 linear_48" [label="(1000,)", style=solid]; -"102 x" -> "103 quantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; -"103 quantize_per_tensor_default" -> "104 dequantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; -"104 dequantize_per_tensor_default" -> "109 conv2d" [label="(1, 3, 224, 224)", style=solid]; -"105 conv2d_scale_0" -> "108 dequantize_per_channel_default" [label="(768,)", style=solid]; -"106 conv2d_zero_point_0" -> "108 dequantize_per_channel_default" [label="(768,)", style=solid]; -"107 _frozen_param0" -> "108 dequantize_per_channel_default" [label="(768, 3, 16, 16)", style=solid]; -"108 dequantize_per_channel_default" -> "109 conv2d" [label="(768, 3, 16, 16)", style=solid]; -"109 conv2d" -> "110 reshape" [label="(1, 768, 14, 14)", style=solid]; -"110 reshape" -> "111 permute" [label="(1, 768, 196)", style=solid]; -"111 permute" -> "113 cat" [label="(1, 196, 768)", style=solid]; -"112 expand" -> "113 cat" [label="(1, 1, 768)", style=solid]; -"113 cat" -> "114 add" [label="(1, 197, 768)", style=solid]; -"114 add" -> "115 dropout" [label="(1, 197, 768)", style=solid]; -"115 dropout" -> "116 layer_norm" [label="(1, 197, 768)", style=solid]; -"115 dropout" -> "161 add_1" [label="(1, 197, 768)", style=solid]; -"116 layer_norm" -> "117 transpose" [label="(1, 197, 768)", style=solid]; -"117 transpose" -> "118 transpose_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; -"118 transpose_0_0_nncf_smooth_quant_0" -> "119 quantize_per_tensor_default_1" [label="(197, 1, 768)", style=solid]; -"119 quantize_per_tensor_default_1" -> "120 dequantize_per_tensor_default_1" [label="(197, 1, 768)", style=solid]; -"120 dequantize_per_tensor_default_1" -> "125 linear" [label="(197, 1, 768)", style=solid]; -"121 linear_scale_0" -> "124 dequantize_per_channel_default_1" [label="(2304,)", style=solid]; -"122 linear_zero_point_0" -> "124 dequantize_per_channel_default_1" [label="(2304,)", style=solid]; -"123 _frozen_param1" -> "124 dequantize_per_channel_default_1" [label="(2304, 768)", style=solid]; -"124 dequantize_per_channel_default_1" -> "125 linear" [label="(2304, 768)", style=solid]; -"125 linear" -> "126 unflatten" [label="(197, 1, 2304)", style=solid]; -"126 unflatten" -> "127 unsqueeze" [label="(197, 1, 3, 768)", style=solid]; -"127 unsqueeze" -> "128 transpose_1" [label="(1, 197, 1, 3, 768)", style=solid]; -"128 transpose_1" -> "129 squeeze" [label="(3, 197, 1, 1, 768)", style=solid]; -"129 squeeze" -> "130 contiguous" [label="(3, 197, 1, 768)", style=solid]; -"130 contiguous" -> "131 quantize_per_tensor_default_2" [label="(3, 197, 1, 768)", style=solid]; -"130 contiguous" -> "134 quantize_per_tensor_default_3" [label="(3, 197, 1, 768)", style=solid]; -"130 contiguous" -> "137 select_2" [label="(3, 197, 1, 768)", style=solid]; -"131 quantize_per_tensor_default_2" -> "132 dequantize_per_tensor_default_2" [label="(3, 197, 1, 768)", style=solid]; -"132 dequantize_per_tensor_default_2" -> "133 select" [label="(3, 197, 1, 768)", style=solid]; -"133 select" -> "138 view" [label="(197, 1, 768)", style=solid]; -"134 quantize_per_tensor_default_3" -> "135 dequantize_per_tensor_default_3" [label="(3, 197, 1, 768)", style=solid]; -"135 dequantize_per_tensor_default_3" -> "136 select_1" [label="(3, 197, 1, 768)", style=solid]; -"136 select_1" -> "140 view_1" [label="(197, 1, 768)", style=solid]; -"137 select_2" -> "142 view_2" [label="(197, 1, 768)", style=solid]; -"138 view" -> "139 transpose_2" [label="(197, 12, 64)", style=solid]; -"139 transpose_2" -> "144 view_3" [label="(12, 197, 64)", style=solid]; -"140 view_1" -> "141 transpose_3" [label="(197, 12, 64)", style=solid]; -"141 transpose_3" -> "145 view_4" [label="(12, 197, 64)", style=solid]; -"142 view_2" -> "143 transpose_4" [label="(197, 12, 64)", style=solid]; -"143 transpose_4" -> "146 view_5" [label="(12, 197, 64)", style=solid]; -"144 view_3" -> "147 scaled_dot_product_attention" [label="(1, 12, 197, 64)", style=solid]; -"145 view_4" -> "147 scaled_dot_product_attention" [label="(1, 12, 197, 64)", style=solid]; -"146 view_5" -> "147 scaled_dot_product_attention" [label="(1, 12, 197, 64)", style=solid]; -"147 scaled_dot_product_attention" -> "148 permute_1" [label="(1, 12, 197, 64)", style=solid]; -"148 permute_1" -> "149 view_6" [label="(197, 1, 12, 64)", style=solid]; -"149 view_6" -> "150 view_6_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; -"150 view_6_0_0_nncf_smooth_quant_0" -> "151 quantize_per_tensor_default_4" [label="(197, 768)", style=solid]; -"151 quantize_per_tensor_default_4" -> "152 dequantize_per_tensor_default_4" [label="(197, 768)", style=solid]; -"152 dequantize_per_tensor_default_4" -> "157 linear_1" [label="(197, 768)", style=solid]; -"153 linear_1_scale_0" -> "156 dequantize_per_channel_default_2" [label="(768,)", style=solid]; -"154 linear_1_zero_point_0" -> "156 dequantize_per_channel_default_2" [label="(768,)", style=solid]; -"155 _frozen_param2" -> "156 dequantize_per_channel_default_2" [label="(768, 768)", style=solid]; -"156 dequantize_per_channel_default_2" -> "157 linear_1" [label="(768, 768)", style=solid]; -"157 linear_1" -> "158 view_7" [label="(197, 768)", style=solid]; -"158 view_7" -> "159 transpose_5" [label="(197, 1, 768)", style=solid]; -"159 transpose_5" -> "160 dropout_1" [label="(1, 197, 768)", style=solid]; -"160 dropout_1" -> "161 add_1" [label="(1, 197, 768)", style=solid]; -"161 add_1" -> "162 layer_norm_1" [label="(1, 197, 768)", style=solid]; -"161 add_1" -> "182 add_2" [label="(1, 197, 768)", style=solid]; -"162 layer_norm_1" -> "163 layer_norm_1_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; -"163 layer_norm_1_0_0_nncf_smooth_quant_0" -> "164 quantize_per_tensor_default_5" [label="(1, 197, 768)", style=solid]; -"164 quantize_per_tensor_default_5" -> "165 dequantize_per_tensor_default_5" [label="(1, 197, 768)", style=solid]; -"165 dequantize_per_tensor_default_5" -> "170 linear_2" [label="(1, 197, 768)", style=solid]; -"166 linear_2_scale_0" -> "169 dequantize_per_channel_default_3" [label="(3072,)", style=solid]; -"167 linear_2_zero_point_0" -> "169 dequantize_per_channel_default_3" [label="(3072,)", style=solid]; -"168 _frozen_param3" -> "169 dequantize_per_channel_default_3" [label="(3072, 768)", style=solid]; -"169 dequantize_per_channel_default_3" -> "170 linear_2" [label="(3072, 768)", style=solid]; -"170 linear_2" -> "171 gelu" [label="(1, 197, 3072)", style=solid]; -"171 gelu" -> "172 dropout_2" [label="(1, 197, 3072)", style=solid]; -"172 dropout_2" -> "173 dropout_2_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; -"173 dropout_2_0_0_nncf_smooth_quant_0" -> "174 quantize_per_tensor_default_6" [label="(1, 197, 3072)", style=solid]; -"174 quantize_per_tensor_default_6" -> "175 dequantize_per_tensor_default_6" [label="(1, 197, 3072)", style=solid]; -"175 dequantize_per_tensor_default_6" -> "180 linear_3" [label="(1, 197, 3072)", style=solid]; -"176 linear_3_scale_0" -> "179 dequantize_per_channel_default_4" [label="(768,)", style=solid]; -"177 linear_3_zero_point_0" -> "179 dequantize_per_channel_default_4" [label="(768,)", style=solid]; -"178 _frozen_param4" -> "179 dequantize_per_channel_default_4" [label="(768, 3072)", style=solid]; -"179 dequantize_per_channel_default_4" -> "180 linear_3" [label="(768, 3072)", style=solid]; -"180 linear_3" -> "181 dropout_3" [label="(1, 197, 768)", style=solid]; -"181 dropout_3" -> "182 add_2" [label="(1, 197, 768)", style=solid]; -"182 add_2" -> "183 layer_norm_2" [label="(1, 197, 768)", style=solid]; -"182 add_2" -> "228 add_3" [label="(1, 197, 768)", style=solid]; -"183 layer_norm_2" -> "184 transpose_6" [label="(1, 197, 768)", style=solid]; -"184 transpose_6" -> "185 transpose_6_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; -"185 transpose_6_0_0_nncf_smooth_quant_0" -> "186 quantize_per_tensor_default_7" [label="(197, 1, 768)", style=solid]; -"186 quantize_per_tensor_default_7" -> "187 dequantize_per_tensor_default_7" [label="(197, 1, 768)", style=solid]; -"187 dequantize_per_tensor_default_7" -> "192 linear_4" [label="(197, 1, 768)", style=solid]; -"188 linear_4_scale_0" -> "191 dequantize_per_channel_default_5" [label="(2304,)", style=solid]; -"189 linear_4_zero_point_0" -> "191 dequantize_per_channel_default_5" [label="(2304,)", style=solid]; -"190 _frozen_param5" -> "191 dequantize_per_channel_default_5" [label="(2304, 768)", style=solid]; -"191 dequantize_per_channel_default_5" -> "192 linear_4" [label="(2304, 768)", style=solid]; -"192 linear_4" -> "193 unflatten_1" [label="(197, 1, 2304)", style=solid]; -"193 unflatten_1" -> "194 unsqueeze_1" [label="(197, 1, 3, 768)", style=solid]; -"194 unsqueeze_1" -> "195 transpose_7" [label="(1, 197, 1, 3, 768)", style=solid]; -"195 transpose_7" -> "196 squeeze_1" [label="(3, 197, 1, 1, 768)", style=solid]; -"196 squeeze_1" -> "197 contiguous_1" [label="(3, 197, 1, 768)", style=solid]; -"197 contiguous_1" -> "198 quantize_per_tensor_default_8" [label="(3, 197, 1, 768)", style=solid]; -"197 contiguous_1" -> "201 quantize_per_tensor_default_9" [label="(3, 197, 1, 768)", style=solid]; -"197 contiguous_1" -> "204 select_5" [label="(3, 197, 1, 768)", style=solid]; -"198 quantize_per_tensor_default_8" -> "199 dequantize_per_tensor_default_8" [label="(3, 197, 1, 768)", style=solid]; -"199 dequantize_per_tensor_default_8" -> "200 select_3" [label="(3, 197, 1, 768)", style=solid]; -"200 select_3" -> "205 view_8" [label="(197, 1, 768)", style=solid]; -"201 quantize_per_tensor_default_9" -> "202 dequantize_per_tensor_default_9" [label="(3, 197, 1, 768)", style=solid]; -"202 dequantize_per_tensor_default_9" -> "203 select_4" [label="(3, 197, 1, 768)", style=solid]; -"203 select_4" -> "207 view_9" [label="(197, 1, 768)", style=solid]; -"204 select_5" -> "209 view_10" [label="(197, 1, 768)", style=solid]; -"205 view_8" -> "206 transpose_8" [label="(197, 12, 64)", style=solid]; -"206 transpose_8" -> "211 view_11" [label="(12, 197, 64)", style=solid]; -"207 view_9" -> "208 transpose_9" [label="(197, 12, 64)", style=solid]; -"208 transpose_9" -> "212 view_12" [label="(12, 197, 64)", style=solid]; -"209 view_10" -> "210 transpose_10" [label="(197, 12, 64)", style=solid]; -"210 transpose_10" -> "213 view_13" [label="(12, 197, 64)", style=solid]; -"211 view_11" -> "214 scaled_dot_product_attention_1" [label="(1, 12, 197, 64)", style=solid]; -"212 view_12" -> "214 scaled_dot_product_attention_1" [label="(1, 12, 197, 64)", style=solid]; -"213 view_13" -> "214 scaled_dot_product_attention_1" [label="(1, 12, 197, 64)", style=solid]; -"214 scaled_dot_product_attention_1" -> "215 permute_2" [label="(1, 12, 197, 64)", style=solid]; -"215 permute_2" -> "216 view_14" [label="(197, 1, 12, 64)", style=solid]; -"216 view_14" -> "217 view_14_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; -"217 view_14_0_0_nncf_smooth_quant_0" -> "218 quantize_per_tensor_default_10" [label="(197, 768)", style=solid]; -"218 quantize_per_tensor_default_10" -> "219 dequantize_per_tensor_default_10" [label="(197, 768)", style=solid]; -"219 dequantize_per_tensor_default_10" -> "224 linear_5" [label="(197, 768)", style=solid]; -"220 linear_5_scale_0" -> "223 dequantize_per_channel_default_6" [label="(768,)", style=solid]; -"221 linear_5_zero_point_0" -> "223 dequantize_per_channel_default_6" [label="(768,)", style=solid]; -"222 _frozen_param6" -> "223 dequantize_per_channel_default_6" [label="(768, 768)", style=solid]; -"223 dequantize_per_channel_default_6" -> "224 linear_5" [label="(768, 768)", style=solid]; -"224 linear_5" -> "225 view_15" [label="(197, 768)", style=solid]; -"225 view_15" -> "226 transpose_11" [label="(197, 1, 768)", style=solid]; -"226 transpose_11" -> "227 dropout_4" [label="(1, 197, 768)", style=solid]; -"227 dropout_4" -> "228 add_3" [label="(1, 197, 768)", style=solid]; -"228 add_3" -> "229 layer_norm_3" [label="(1, 197, 768)", style=solid]; -"228 add_3" -> "249 add_4" [label="(1, 197, 768)", style=solid]; -"229 layer_norm_3" -> "230 layer_norm_3_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; -"230 layer_norm_3_0_0_nncf_smooth_quant_0" -> "231 quantize_per_tensor_default_11" [label="(1, 197, 768)", style=solid]; -"231 quantize_per_tensor_default_11" -> "232 dequantize_per_tensor_default_11" [label="(1, 197, 768)", style=solid]; -"232 dequantize_per_tensor_default_11" -> "237 linear_6" [label="(1, 197, 768)", style=solid]; -"233 linear_6_scale_0" -> "236 dequantize_per_channel_default_7" [label="(3072,)", style=solid]; -"234 linear_6_zero_point_0" -> "236 dequantize_per_channel_default_7" [label="(3072,)", style=solid]; -"235 _frozen_param7" -> "236 dequantize_per_channel_default_7" [label="(3072, 768)", style=solid]; -"236 dequantize_per_channel_default_7" -> "237 linear_6" [label="(3072, 768)", style=solid]; -"237 linear_6" -> "238 gelu_1" [label="(1, 197, 3072)", style=solid]; -"238 gelu_1" -> "239 dropout_5" [label="(1, 197, 3072)", style=solid]; -"239 dropout_5" -> "240 dropout_5_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; -"240 dropout_5_0_0_nncf_smooth_quant_0" -> "241 quantize_per_tensor_default_12" [label="(1, 197, 3072)", style=solid]; -"241 quantize_per_tensor_default_12" -> "242 dequantize_per_tensor_default_12" [label="(1, 197, 3072)", style=solid]; -"242 dequantize_per_tensor_default_12" -> "247 linear_7" [label="(1, 197, 3072)", style=solid]; -"243 linear_7_scale_0" -> "246 dequantize_per_channel_default_8" [label="(768,)", style=solid]; -"244 linear_7_zero_point_0" -> "246 dequantize_per_channel_default_8" [label="(768,)", style=solid]; -"245 _frozen_param8" -> "246 dequantize_per_channel_default_8" [label="(768, 3072)", style=solid]; -"246 dequantize_per_channel_default_8" -> "247 linear_7" [label="(768, 3072)", style=solid]; -"247 linear_7" -> "248 dropout_6" [label="(1, 197, 768)", style=solid]; -"248 dropout_6" -> "249 add_4" [label="(1, 197, 768)", style=solid]; -"249 add_4" -> "250 layer_norm_4" [label="(1, 197, 768)", style=solid]; -"249 add_4" -> "295 add_5" [label="(1, 197, 768)", style=solid]; -"250 layer_norm_4" -> "251 transpose_12" [label="(1, 197, 768)", style=solid]; -"251 transpose_12" -> "252 transpose_12_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; -"252 transpose_12_0_0_nncf_smooth_quant_0" -> "253 quantize_per_tensor_default_13" [label="(197, 1, 768)", style=solid]; -"253 quantize_per_tensor_default_13" -> "254 dequantize_per_tensor_default_13" [label="(197, 1, 768)", style=solid]; -"254 dequantize_per_tensor_default_13" -> "259 linear_8" [label="(197, 1, 768)", style=solid]; -"255 linear_8_scale_0" -> "258 dequantize_per_channel_default_9" [label="(2304,)", style=solid]; -"256 linear_8_zero_point_0" -> "258 dequantize_per_channel_default_9" [label="(2304,)", style=solid]; -"257 _frozen_param9" -> "258 dequantize_per_channel_default_9" [label="(2304, 768)", style=solid]; -"258 dequantize_per_channel_default_9" -> "259 linear_8" [label="(2304, 768)", style=solid]; -"259 linear_8" -> "260 unflatten_2" [label="(197, 1, 2304)", style=solid]; -"260 unflatten_2" -> "261 unsqueeze_2" [label="(197, 1, 3, 768)", style=solid]; -"261 unsqueeze_2" -> "262 transpose_13" [label="(1, 197, 1, 3, 768)", style=solid]; -"262 transpose_13" -> "263 squeeze_2" [label="(3, 197, 1, 1, 768)", style=solid]; -"263 squeeze_2" -> "264 contiguous_2" [label="(3, 197, 1, 768)", style=solid]; -"264 contiguous_2" -> "265 quantize_per_tensor_default_14" [label="(3, 197, 1, 768)", style=solid]; -"264 contiguous_2" -> "268 quantize_per_tensor_default_15" [label="(3, 197, 1, 768)", style=solid]; -"264 contiguous_2" -> "271 select_8" [label="(3, 197, 1, 768)", style=solid]; -"265 quantize_per_tensor_default_14" -> "266 dequantize_per_tensor_default_14" [label="(3, 197, 1, 768)", style=solid]; -"266 dequantize_per_tensor_default_14" -> "267 select_6" [label="(3, 197, 1, 768)", style=solid]; -"267 select_6" -> "272 view_16" [label="(197, 1, 768)", style=solid]; -"268 quantize_per_tensor_default_15" -> "269 dequantize_per_tensor_default_15" [label="(3, 197, 1, 768)", style=solid]; -"269 dequantize_per_tensor_default_15" -> "270 select_7" [label="(3, 197, 1, 768)", style=solid]; -"270 select_7" -> "274 view_17" [label="(197, 1, 768)", style=solid]; -"271 select_8" -> "276 view_18" [label="(197, 1, 768)", style=solid]; -"272 view_16" -> "273 transpose_14" [label="(197, 12, 64)", style=solid]; -"273 transpose_14" -> "278 view_19" [label="(12, 197, 64)", style=solid]; -"274 view_17" -> "275 transpose_15" [label="(197, 12, 64)", style=solid]; -"275 transpose_15" -> "279 view_20" [label="(12, 197, 64)", style=solid]; -"276 view_18" -> "277 transpose_16" [label="(197, 12, 64)", style=solid]; -"277 transpose_16" -> "280 view_21" [label="(12, 197, 64)", style=solid]; -"278 view_19" -> "281 scaled_dot_product_attention_2" [label="(1, 12, 197, 64)", style=solid]; -"279 view_20" -> "281 scaled_dot_product_attention_2" [label="(1, 12, 197, 64)", style=solid]; -"280 view_21" -> "281 scaled_dot_product_attention_2" [label="(1, 12, 197, 64)", style=solid]; -"281 scaled_dot_product_attention_2" -> "282 permute_3" [label="(1, 12, 197, 64)", style=solid]; -"282 permute_3" -> "283 view_22" [label="(197, 1, 12, 64)", style=solid]; -"283 view_22" -> "284 view_22_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; -"284 view_22_0_0_nncf_smooth_quant_0" -> "285 quantize_per_tensor_default_16" [label="(197, 768)", style=solid]; -"285 quantize_per_tensor_default_16" -> "286 dequantize_per_tensor_default_16" [label="(197, 768)", style=solid]; -"286 dequantize_per_tensor_default_16" -> "291 linear_9" [label="(197, 768)", style=solid]; -"287 linear_9_scale_0" -> "290 dequantize_per_channel_default_10" [label="(768,)", style=solid]; -"288 linear_9_zero_point_0" -> "290 dequantize_per_channel_default_10" [label="(768,)", style=solid]; -"289 _frozen_param10" -> "290 dequantize_per_channel_default_10" [label="(768, 768)", style=solid]; -"290 dequantize_per_channel_default_10" -> "291 linear_9" [label="(768, 768)", style=solid]; -"291 linear_9" -> "292 view_23" [label="(197, 768)", style=solid]; -"292 view_23" -> "293 transpose_17" [label="(197, 1, 768)", style=solid]; -"293 transpose_17" -> "294 dropout_7" [label="(1, 197, 768)", style=solid]; -"294 dropout_7" -> "295 add_5" [label="(1, 197, 768)", style=solid]; -"295 add_5" -> "296 layer_norm_5" [label="(1, 197, 768)", style=solid]; -"295 add_5" -> "316 add_6" [label="(1, 197, 768)", style=solid]; -"296 layer_norm_5" -> "297 layer_norm_5_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; -"297 layer_norm_5_0_0_nncf_smooth_quant_0" -> "298 quantize_per_tensor_default_17" [label="(1, 197, 768)", style=solid]; -"298 quantize_per_tensor_default_17" -> "299 dequantize_per_tensor_default_17" [label="(1, 197, 768)", style=solid]; -"299 dequantize_per_tensor_default_17" -> "304 linear_10" [label="(1, 197, 768)", style=solid]; -"300 linear_10_scale_0" -> "303 dequantize_per_channel_default_11" [label="(3072,)", style=solid]; -"301 linear_10_zero_point_0" -> "303 dequantize_per_channel_default_11" [label="(3072,)", style=solid]; -"302 _frozen_param11" -> "303 dequantize_per_channel_default_11" [label="(3072, 768)", style=solid]; -"303 dequantize_per_channel_default_11" -> "304 linear_10" [label="(3072, 768)", style=solid]; -"304 linear_10" -> "305 gelu_2" [label="(1, 197, 3072)", style=solid]; -"305 gelu_2" -> "306 dropout_8" [label="(1, 197, 3072)", style=solid]; -"306 dropout_8" -> "307 dropout_8_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; -"307 dropout_8_0_0_nncf_smooth_quant_0" -> "308 quantize_per_tensor_default_18" [label="(1, 197, 3072)", style=solid]; -"308 quantize_per_tensor_default_18" -> "309 dequantize_per_tensor_default_18" [label="(1, 197, 3072)", style=solid]; -"309 dequantize_per_tensor_default_18" -> "314 linear_11" [label="(1, 197, 3072)", style=solid]; -"310 linear_11_scale_0" -> "313 dequantize_per_channel_default_12" [label="(768,)", style=solid]; -"311 linear_11_zero_point_0" -> "313 dequantize_per_channel_default_12" [label="(768,)", style=solid]; -"312 _frozen_param12" -> "313 dequantize_per_channel_default_12" [label="(768, 3072)", style=solid]; -"313 dequantize_per_channel_default_12" -> "314 linear_11" [label="(768, 3072)", style=solid]; -"314 linear_11" -> "315 dropout_9" [label="(1, 197, 768)", style=solid]; -"315 dropout_9" -> "316 add_6" [label="(1, 197, 768)", style=solid]; -"316 add_6" -> "317 layer_norm_6" [label="(1, 197, 768)", style=solid]; -"316 add_6" -> "362 add_7" [label="(1, 197, 768)", style=solid]; -"317 layer_norm_6" -> "318 transpose_18" [label="(1, 197, 768)", style=solid]; -"318 transpose_18" -> "319 transpose_18_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; -"319 transpose_18_0_0_nncf_smooth_quant_0" -> "320 quantize_per_tensor_default_19" [label="(197, 1, 768)", style=solid]; -"320 quantize_per_tensor_default_19" -> "321 dequantize_per_tensor_default_19" [label="(197, 1, 768)", style=solid]; -"321 dequantize_per_tensor_default_19" -> "326 linear_12" [label="(197, 1, 768)", style=solid]; -"322 linear_12_scale_0" -> "325 dequantize_per_channel_default_13" [label="(2304,)", style=solid]; -"323 linear_12_zero_point_0" -> "325 dequantize_per_channel_default_13" [label="(2304,)", style=solid]; -"324 _frozen_param13" -> "325 dequantize_per_channel_default_13" [label="(2304, 768)", style=solid]; -"325 dequantize_per_channel_default_13" -> "326 linear_12" [label="(2304, 768)", style=solid]; -"326 linear_12" -> "327 unflatten_3" [label="(197, 1, 2304)", style=solid]; -"327 unflatten_3" -> "328 unsqueeze_3" [label="(197, 1, 3, 768)", style=solid]; -"328 unsqueeze_3" -> "329 transpose_19" [label="(1, 197, 1, 3, 768)", style=solid]; -"329 transpose_19" -> "330 squeeze_3" [label="(3, 197, 1, 1, 768)", style=solid]; -"330 squeeze_3" -> "331 contiguous_3" [label="(3, 197, 1, 768)", style=solid]; -"331 contiguous_3" -> "332 quantize_per_tensor_default_20" [label="(3, 197, 1, 768)", style=solid]; -"331 contiguous_3" -> "335 quantize_per_tensor_default_21" [label="(3, 197, 1, 768)", style=solid]; -"331 contiguous_3" -> "338 select_11" [label="(3, 197, 1, 768)", style=solid]; -"332 quantize_per_tensor_default_20" -> "333 dequantize_per_tensor_default_20" [label="(3, 197, 1, 768)", style=solid]; -"333 dequantize_per_tensor_default_20" -> "334 select_9" [label="(3, 197, 1, 768)", style=solid]; -"334 select_9" -> "339 view_24" [label="(197, 1, 768)", style=solid]; -"335 quantize_per_tensor_default_21" -> "336 dequantize_per_tensor_default_21" [label="(3, 197, 1, 768)", style=solid]; -"336 dequantize_per_tensor_default_21" -> "337 select_10" [label="(3, 197, 1, 768)", style=solid]; -"337 select_10" -> "341 view_25" [label="(197, 1, 768)", style=solid]; -"338 select_11" -> "343 view_26" [label="(197, 1, 768)", style=solid]; -"339 view_24" -> "340 transpose_20" [label="(197, 12, 64)", style=solid]; -"340 transpose_20" -> "345 view_27" [label="(12, 197, 64)", style=solid]; -"341 view_25" -> "342 transpose_21" [label="(197, 12, 64)", style=solid]; -"342 transpose_21" -> "346 view_28" [label="(12, 197, 64)", style=solid]; -"343 view_26" -> "344 transpose_22" [label="(197, 12, 64)", style=solid]; -"344 transpose_22" -> "347 view_29" [label="(12, 197, 64)", style=solid]; -"345 view_27" -> "348 scaled_dot_product_attention_3" [label="(1, 12, 197, 64)", style=solid]; -"346 view_28" -> "348 scaled_dot_product_attention_3" [label="(1, 12, 197, 64)", style=solid]; -"347 view_29" -> "348 scaled_dot_product_attention_3" [label="(1, 12, 197, 64)", style=solid]; -"348 scaled_dot_product_attention_3" -> "349 permute_4" [label="(1, 12, 197, 64)", style=solid]; -"349 permute_4" -> "350 view_30" [label="(197, 1, 12, 64)", style=solid]; -"350 view_30" -> "351 view_30_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; -"351 view_30_0_0_nncf_smooth_quant_0" -> "352 quantize_per_tensor_default_22" [label="(197, 768)", style=solid]; -"352 quantize_per_tensor_default_22" -> "353 dequantize_per_tensor_default_22" [label="(197, 768)", style=solid]; -"353 dequantize_per_tensor_default_22" -> "358 linear_13" [label="(197, 768)", style=solid]; -"354 linear_13_scale_0" -> "357 dequantize_per_channel_default_14" [label="(768,)", style=solid]; -"355 linear_13_zero_point_0" -> "357 dequantize_per_channel_default_14" [label="(768,)", style=solid]; -"356 _frozen_param14" -> "357 dequantize_per_channel_default_14" [label="(768, 768)", style=solid]; -"357 dequantize_per_channel_default_14" -> "358 linear_13" [label="(768, 768)", style=solid]; -"358 linear_13" -> "359 view_31" [label="(197, 768)", style=solid]; -"359 view_31" -> "360 transpose_23" [label="(197, 1, 768)", style=solid]; -"360 transpose_23" -> "361 dropout_10" [label="(1, 197, 768)", style=solid]; -"361 dropout_10" -> "362 add_7" [label="(1, 197, 768)", style=solid]; -"362 add_7" -> "363 layer_norm_7" [label="(1, 197, 768)", style=solid]; -"362 add_7" -> "383 add_8" [label="(1, 197, 768)", style=solid]; -"363 layer_norm_7" -> "364 layer_norm_7_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; -"364 layer_norm_7_0_0_nncf_smooth_quant_0" -> "365 quantize_per_tensor_default_23" [label="(1, 197, 768)", style=solid]; -"365 quantize_per_tensor_default_23" -> "366 dequantize_per_tensor_default_23" [label="(1, 197, 768)", style=solid]; -"366 dequantize_per_tensor_default_23" -> "371 linear_14" [label="(1, 197, 768)", style=solid]; -"367 linear_14_scale_0" -> "370 dequantize_per_channel_default_15" [label="(3072,)", style=solid]; -"368 linear_14_zero_point_0" -> "370 dequantize_per_channel_default_15" [label="(3072,)", style=solid]; -"369 _frozen_param15" -> "370 dequantize_per_channel_default_15" [label="(3072, 768)", style=solid]; -"370 dequantize_per_channel_default_15" -> "371 linear_14" [label="(3072, 768)", style=solid]; -"371 linear_14" -> "372 gelu_3" [label="(1, 197, 3072)", style=solid]; -"372 gelu_3" -> "373 dropout_11" [label="(1, 197, 3072)", style=solid]; -"373 dropout_11" -> "374 dropout_11_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; -"374 dropout_11_0_0_nncf_smooth_quant_0" -> "375 quantize_per_tensor_default_24" [label="(1, 197, 3072)", style=solid]; -"375 quantize_per_tensor_default_24" -> "376 dequantize_per_tensor_default_24" [label="(1, 197, 3072)", style=solid]; -"376 dequantize_per_tensor_default_24" -> "381 linear_15" [label="(1, 197, 3072)", style=solid]; -"377 linear_15_scale_0" -> "380 dequantize_per_channel_default_16" [label="(768,)", style=solid]; -"378 linear_15_zero_point_0" -> "380 dequantize_per_channel_default_16" [label="(768,)", style=solid]; -"379 _frozen_param16" -> "380 dequantize_per_channel_default_16" [label="(768, 3072)", style=solid]; -"380 dequantize_per_channel_default_16" -> "381 linear_15" [label="(768, 3072)", style=solid]; -"381 linear_15" -> "382 dropout_12" [label="(1, 197, 768)", style=solid]; -"382 dropout_12" -> "383 add_8" [label="(1, 197, 768)", style=solid]; -"383 add_8" -> "384 layer_norm_8" [label="(1, 197, 768)", style=solid]; -"383 add_8" -> "429 add_9" [label="(1, 197, 768)", style=solid]; -"384 layer_norm_8" -> "385 transpose_24" [label="(1, 197, 768)", style=solid]; -"385 transpose_24" -> "386 transpose_24_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; -"386 transpose_24_0_0_nncf_smooth_quant_0" -> "387 quantize_per_tensor_default_25" [label="(197, 1, 768)", style=solid]; -"387 quantize_per_tensor_default_25" -> "388 dequantize_per_tensor_default_25" [label="(197, 1, 768)", style=solid]; -"388 dequantize_per_tensor_default_25" -> "393 linear_16" [label="(197, 1, 768)", style=solid]; -"389 linear_16_scale_0" -> "392 dequantize_per_channel_default_17" [label="(2304,)", style=solid]; -"390 linear_16_zero_point_0" -> "392 dequantize_per_channel_default_17" [label="(2304,)", style=solid]; -"391 _frozen_param17" -> "392 dequantize_per_channel_default_17" [label="(2304, 768)", style=solid]; -"392 dequantize_per_channel_default_17" -> "393 linear_16" [label="(2304, 768)", style=solid]; -"393 linear_16" -> "394 unflatten_4" [label="(197, 1, 2304)", style=solid]; -"394 unflatten_4" -> "395 unsqueeze_4" [label="(197, 1, 3, 768)", style=solid]; -"395 unsqueeze_4" -> "396 transpose_25" [label="(1, 197, 1, 3, 768)", style=solid]; -"396 transpose_25" -> "397 squeeze_4" [label="(3, 197, 1, 1, 768)", style=solid]; -"397 squeeze_4" -> "398 contiguous_4" [label="(3, 197, 1, 768)", style=solid]; -"398 contiguous_4" -> "399 quantize_per_tensor_default_26" [label="(3, 197, 1, 768)", style=solid]; -"398 contiguous_4" -> "402 quantize_per_tensor_default_27" [label="(3, 197, 1, 768)", style=solid]; -"398 contiguous_4" -> "405 select_14" [label="(3, 197, 1, 768)", style=solid]; -"399 quantize_per_tensor_default_26" -> "400 dequantize_per_tensor_default_26" [label="(3, 197, 1, 768)", style=solid]; -"400 dequantize_per_tensor_default_26" -> "401 select_12" [label="(3, 197, 1, 768)", style=solid]; -"401 select_12" -> "406 view_32" [label="(197, 1, 768)", style=solid]; -"402 quantize_per_tensor_default_27" -> "403 dequantize_per_tensor_default_27" [label="(3, 197, 1, 768)", style=solid]; -"403 dequantize_per_tensor_default_27" -> "404 select_13" [label="(3, 197, 1, 768)", style=solid]; -"404 select_13" -> "408 view_33" [label="(197, 1, 768)", style=solid]; -"405 select_14" -> "410 view_34" [label="(197, 1, 768)", style=solid]; -"406 view_32" -> "407 transpose_26" [label="(197, 12, 64)", style=solid]; -"407 transpose_26" -> "412 view_35" [label="(12, 197, 64)", style=solid]; -"408 view_33" -> "409 transpose_27" [label="(197, 12, 64)", style=solid]; -"409 transpose_27" -> "413 view_36" [label="(12, 197, 64)", style=solid]; -"410 view_34" -> "411 transpose_28" [label="(197, 12, 64)", style=solid]; -"411 transpose_28" -> "414 view_37" [label="(12, 197, 64)", style=solid]; -"412 view_35" -> "415 scaled_dot_product_attention_4" [label="(1, 12, 197, 64)", style=solid]; -"413 view_36" -> "415 scaled_dot_product_attention_4" [label="(1, 12, 197, 64)", style=solid]; -"414 view_37" -> "415 scaled_dot_product_attention_4" [label="(1, 12, 197, 64)", style=solid]; -"415 scaled_dot_product_attention_4" -> "416 permute_5" [label="(1, 12, 197, 64)", style=solid]; -"416 permute_5" -> "417 view_38" [label="(197, 1, 12, 64)", style=solid]; -"417 view_38" -> "418 view_38_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; -"418 view_38_0_0_nncf_smooth_quant_0" -> "419 quantize_per_tensor_default_28" [label="(197, 768)", style=solid]; -"419 quantize_per_tensor_default_28" -> "420 dequantize_per_tensor_default_28" [label="(197, 768)", style=solid]; -"420 dequantize_per_tensor_default_28" -> "425 linear_17" [label="(197, 768)", style=solid]; -"421 linear_17_scale_0" -> "424 dequantize_per_channel_default_18" [label="(768,)", style=solid]; -"422 linear_17_zero_point_0" -> "424 dequantize_per_channel_default_18" [label="(768,)", style=solid]; -"423 _frozen_param18" -> "424 dequantize_per_channel_default_18" [label="(768, 768)", style=solid]; -"424 dequantize_per_channel_default_18" -> "425 linear_17" [label="(768, 768)", style=solid]; -"425 linear_17" -> "426 view_39" [label="(197, 768)", style=solid]; -"426 view_39" -> "427 transpose_29" [label="(197, 1, 768)", style=solid]; -"427 transpose_29" -> "428 dropout_13" [label="(1, 197, 768)", style=solid]; -"428 dropout_13" -> "429 add_9" [label="(1, 197, 768)", style=solid]; -"429 add_9" -> "430 layer_norm_9" [label="(1, 197, 768)", style=solid]; -"429 add_9" -> "450 add_10" [label="(1, 197, 768)", style=solid]; -"430 layer_norm_9" -> "431 layer_norm_9_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; -"431 layer_norm_9_0_0_nncf_smooth_quant_0" -> "432 quantize_per_tensor_default_29" [label="(1, 197, 768)", style=solid]; -"432 quantize_per_tensor_default_29" -> "433 dequantize_per_tensor_default_29" [label="(1, 197, 768)", style=solid]; -"433 dequantize_per_tensor_default_29" -> "438 linear_18" [label="(1, 197, 768)", style=solid]; -"434 linear_18_scale_0" -> "437 dequantize_per_channel_default_19" [label="(3072,)", style=solid]; -"435 linear_18_zero_point_0" -> "437 dequantize_per_channel_default_19" [label="(3072,)", style=solid]; -"436 _frozen_param19" -> "437 dequantize_per_channel_default_19" [label="(3072, 768)", style=solid]; -"437 dequantize_per_channel_default_19" -> "438 linear_18" [label="(3072, 768)", style=solid]; -"438 linear_18" -> "439 gelu_4" [label="(1, 197, 3072)", style=solid]; -"439 gelu_4" -> "440 dropout_14" [label="(1, 197, 3072)", style=solid]; -"440 dropout_14" -> "441 dropout_14_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; -"441 dropout_14_0_0_nncf_smooth_quant_0" -> "442 quantize_per_tensor_default_30" [label="(1, 197, 3072)", style=solid]; -"442 quantize_per_tensor_default_30" -> "443 dequantize_per_tensor_default_30" [label="(1, 197, 3072)", style=solid]; -"443 dequantize_per_tensor_default_30" -> "448 linear_19" [label="(1, 197, 3072)", style=solid]; -"444 linear_19_scale_0" -> "447 dequantize_per_channel_default_20" [label="(768,)", style=solid]; -"445 linear_19_zero_point_0" -> "447 dequantize_per_channel_default_20" [label="(768,)", style=solid]; -"446 _frozen_param20" -> "447 dequantize_per_channel_default_20" [label="(768, 3072)", style=solid]; -"447 dequantize_per_channel_default_20" -> "448 linear_19" [label="(768, 3072)", style=solid]; -"448 linear_19" -> "449 dropout_15" [label="(1, 197, 768)", style=solid]; -"449 dropout_15" -> "450 add_10" [label="(1, 197, 768)", style=solid]; -"450 add_10" -> "451 layer_norm_10" [label="(1, 197, 768)", style=solid]; -"450 add_10" -> "496 add_11" [label="(1, 197, 768)", style=solid]; -"451 layer_norm_10" -> "452 transpose_30" [label="(1, 197, 768)", style=solid]; -"452 transpose_30" -> "453 transpose_30_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; -"453 transpose_30_0_0_nncf_smooth_quant_0" -> "454 quantize_per_tensor_default_31" [label="(197, 1, 768)", style=solid]; -"454 quantize_per_tensor_default_31" -> "455 dequantize_per_tensor_default_31" [label="(197, 1, 768)", style=solid]; -"455 dequantize_per_tensor_default_31" -> "460 linear_20" [label="(197, 1, 768)", style=solid]; -"456 linear_20_scale_0" -> "459 dequantize_per_channel_default_21" [label="(2304,)", style=solid]; -"457 linear_20_zero_point_0" -> "459 dequantize_per_channel_default_21" [label="(2304,)", style=solid]; -"458 _frozen_param21" -> "459 dequantize_per_channel_default_21" [label="(2304, 768)", style=solid]; -"459 dequantize_per_channel_default_21" -> "460 linear_20" [label="(2304, 768)", style=solid]; -"460 linear_20" -> "461 unflatten_5" [label="(197, 1, 2304)", style=solid]; -"461 unflatten_5" -> "462 unsqueeze_5" [label="(197, 1, 3, 768)", style=solid]; -"462 unsqueeze_5" -> "463 transpose_31" [label="(1, 197, 1, 3, 768)", style=solid]; -"463 transpose_31" -> "464 squeeze_5" [label="(3, 197, 1, 1, 768)", style=solid]; -"464 squeeze_5" -> "465 contiguous_5" [label="(3, 197, 1, 768)", style=solid]; -"465 contiguous_5" -> "466 quantize_per_tensor_default_32" [label="(3, 197, 1, 768)", style=solid]; -"465 contiguous_5" -> "469 quantize_per_tensor_default_33" [label="(3, 197, 1, 768)", style=solid]; -"465 contiguous_5" -> "472 select_17" [label="(3, 197, 1, 768)", style=solid]; -"466 quantize_per_tensor_default_32" -> "467 dequantize_per_tensor_default_32" [label="(3, 197, 1, 768)", style=solid]; -"467 dequantize_per_tensor_default_32" -> "468 select_15" [label="(3, 197, 1, 768)", style=solid]; -"468 select_15" -> "473 view_40" [label="(197, 1, 768)", style=solid]; -"469 quantize_per_tensor_default_33" -> "470 dequantize_per_tensor_default_33" [label="(3, 197, 1, 768)", style=solid]; -"470 dequantize_per_tensor_default_33" -> "471 select_16" [label="(3, 197, 1, 768)", style=solid]; -"471 select_16" -> "475 view_41" [label="(197, 1, 768)", style=solid]; -"472 select_17" -> "477 view_42" [label="(197, 1, 768)", style=solid]; -"473 view_40" -> "474 transpose_32" [label="(197, 12, 64)", style=solid]; -"474 transpose_32" -> "479 view_43" [label="(12, 197, 64)", style=solid]; -"475 view_41" -> "476 transpose_33" [label="(197, 12, 64)", style=solid]; -"476 transpose_33" -> "480 view_44" [label="(12, 197, 64)", style=solid]; -"477 view_42" -> "478 transpose_34" [label="(197, 12, 64)", style=solid]; -"478 transpose_34" -> "481 view_45" [label="(12, 197, 64)", style=solid]; -"479 view_43" -> "482 scaled_dot_product_attention_5" [label="(1, 12, 197, 64)", style=solid]; -"480 view_44" -> "482 scaled_dot_product_attention_5" [label="(1, 12, 197, 64)", style=solid]; -"481 view_45" -> "482 scaled_dot_product_attention_5" [label="(1, 12, 197, 64)", style=solid]; -"482 scaled_dot_product_attention_5" -> "483 permute_6" [label="(1, 12, 197, 64)", style=solid]; -"483 permute_6" -> "484 view_46" [label="(197, 1, 12, 64)", style=solid]; -"484 view_46" -> "485 view_46_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; -"485 view_46_0_0_nncf_smooth_quant_0" -> "486 quantize_per_tensor_default_34" [label="(197, 768)", style=solid]; -"486 quantize_per_tensor_default_34" -> "487 dequantize_per_tensor_default_34" [label="(197, 768)", style=solid]; -"487 dequantize_per_tensor_default_34" -> "492 linear_21" [label="(197, 768)", style=solid]; -"488 linear_21_scale_0" -> "491 dequantize_per_channel_default_22" [label="(768,)", style=solid]; -"489 linear_21_zero_point_0" -> "491 dequantize_per_channel_default_22" [label="(768,)", style=solid]; -"490 _frozen_param22" -> "491 dequantize_per_channel_default_22" [label="(768, 768)", style=solid]; -"491 dequantize_per_channel_default_22" -> "492 linear_21" [label="(768, 768)", style=solid]; -"492 linear_21" -> "493 view_47" [label="(197, 768)", style=solid]; -"493 view_47" -> "494 transpose_35" [label="(197, 1, 768)", style=solid]; -"494 transpose_35" -> "495 dropout_16" [label="(1, 197, 768)", style=solid]; -"495 dropout_16" -> "496 add_11" [label="(1, 197, 768)", style=solid]; -"496 add_11" -> "497 layer_norm_11" [label="(1, 197, 768)", style=solid]; -"496 add_11" -> "517 add_12" [label="(1, 197, 768)", style=solid]; -"497 layer_norm_11" -> "498 layer_norm_11_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; -"498 layer_norm_11_0_0_nncf_smooth_quant_0" -> "499 quantize_per_tensor_default_35" [label="(1, 197, 768)", style=solid]; -"499 quantize_per_tensor_default_35" -> "500 dequantize_per_tensor_default_35" [label="(1, 197, 768)", style=solid]; -"500 dequantize_per_tensor_default_35" -> "505 linear_22" [label="(1, 197, 768)", style=solid]; -"501 linear_22_scale_0" -> "504 dequantize_per_channel_default_23" [label="(3072,)", style=solid]; -"502 linear_22_zero_point_0" -> "504 dequantize_per_channel_default_23" [label="(3072,)", style=solid]; -"503 _frozen_param23" -> "504 dequantize_per_channel_default_23" [label="(3072, 768)", style=solid]; -"504 dequantize_per_channel_default_23" -> "505 linear_22" [label="(3072, 768)", style=solid]; -"505 linear_22" -> "506 gelu_5" [label="(1, 197, 3072)", style=solid]; -"506 gelu_5" -> "507 dropout_17" [label="(1, 197, 3072)", style=solid]; -"507 dropout_17" -> "508 dropout_17_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; -"508 dropout_17_0_0_nncf_smooth_quant_0" -> "509 quantize_per_tensor_default_36" [label="(1, 197, 3072)", style=solid]; -"509 quantize_per_tensor_default_36" -> "510 dequantize_per_tensor_default_36" [label="(1, 197, 3072)", style=solid]; -"510 dequantize_per_tensor_default_36" -> "515 linear_23" [label="(1, 197, 3072)", style=solid]; -"511 linear_23_scale_0" -> "514 dequantize_per_channel_default_24" [label="(768,)", style=solid]; -"512 linear_23_zero_point_0" -> "514 dequantize_per_channel_default_24" [label="(768,)", style=solid]; -"513 _frozen_param24" -> "514 dequantize_per_channel_default_24" [label="(768, 3072)", style=solid]; -"514 dequantize_per_channel_default_24" -> "515 linear_23" [label="(768, 3072)", style=solid]; -"515 linear_23" -> "516 dropout_18" [label="(1, 197, 768)", style=solid]; -"516 dropout_18" -> "517 add_12" [label="(1, 197, 768)", style=solid]; -"517 add_12" -> "518 layer_norm_12" [label="(1, 197, 768)", style=solid]; -"517 add_12" -> "563 add_13" [label="(1, 197, 768)", style=solid]; -"518 layer_norm_12" -> "519 transpose_36" [label="(1, 197, 768)", style=solid]; -"519 transpose_36" -> "520 transpose_36_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; -"520 transpose_36_0_0_nncf_smooth_quant_0" -> "521 quantize_per_tensor_default_37" [label="(197, 1, 768)", style=solid]; -"521 quantize_per_tensor_default_37" -> "522 dequantize_per_tensor_default_37" [label="(197, 1, 768)", style=solid]; -"522 dequantize_per_tensor_default_37" -> "527 linear_24" [label="(197, 1, 768)", style=solid]; -"523 linear_24_scale_0" -> "526 dequantize_per_channel_default_25" [label="(2304,)", style=solid]; -"524 linear_24_zero_point_0" -> "526 dequantize_per_channel_default_25" [label="(2304,)", style=solid]; -"525 _frozen_param25" -> "526 dequantize_per_channel_default_25" [label="(2304, 768)", style=solid]; -"526 dequantize_per_channel_default_25" -> "527 linear_24" [label="(2304, 768)", style=solid]; -"527 linear_24" -> "528 unflatten_6" [label="(197, 1, 2304)", style=solid]; -"528 unflatten_6" -> "529 unsqueeze_6" [label="(197, 1, 3, 768)", style=solid]; -"529 unsqueeze_6" -> "530 transpose_37" [label="(1, 197, 1, 3, 768)", style=solid]; -"530 transpose_37" -> "531 squeeze_6" [label="(3, 197, 1, 1, 768)", style=solid]; -"531 squeeze_6" -> "532 contiguous_6" [label="(3, 197, 1, 768)", style=solid]; -"532 contiguous_6" -> "533 quantize_per_tensor_default_38" [label="(3, 197, 1, 768)", style=solid]; -"532 contiguous_6" -> "536 quantize_per_tensor_default_39" [label="(3, 197, 1, 768)", style=solid]; -"532 contiguous_6" -> "539 select_20" [label="(3, 197, 1, 768)", style=solid]; -"533 quantize_per_tensor_default_38" -> "534 dequantize_per_tensor_default_38" [label="(3, 197, 1, 768)", style=solid]; -"534 dequantize_per_tensor_default_38" -> "535 select_18" [label="(3, 197, 1, 768)", style=solid]; -"535 select_18" -> "540 view_48" [label="(197, 1, 768)", style=solid]; -"536 quantize_per_tensor_default_39" -> "537 dequantize_per_tensor_default_39" [label="(3, 197, 1, 768)", style=solid]; -"537 dequantize_per_tensor_default_39" -> "538 select_19" [label="(3, 197, 1, 768)", style=solid]; -"538 select_19" -> "542 view_49" [label="(197, 1, 768)", style=solid]; -"539 select_20" -> "544 view_50" [label="(197, 1, 768)", style=solid]; -"540 view_48" -> "541 transpose_38" [label="(197, 12, 64)", style=solid]; -"541 transpose_38" -> "546 view_51" [label="(12, 197, 64)", style=solid]; -"542 view_49" -> "543 transpose_39" [label="(197, 12, 64)", style=solid]; -"543 transpose_39" -> "547 view_52" [label="(12, 197, 64)", style=solid]; -"544 view_50" -> "545 transpose_40" [label="(197, 12, 64)", style=solid]; -"545 transpose_40" -> "548 view_53" [label="(12, 197, 64)", style=solid]; -"546 view_51" -> "549 scaled_dot_product_attention_6" [label="(1, 12, 197, 64)", style=solid]; -"547 view_52" -> "549 scaled_dot_product_attention_6" [label="(1, 12, 197, 64)", style=solid]; -"548 view_53" -> "549 scaled_dot_product_attention_6" [label="(1, 12, 197, 64)", style=solid]; -"549 scaled_dot_product_attention_6" -> "550 permute_7" [label="(1, 12, 197, 64)", style=solid]; -"550 permute_7" -> "551 view_54" [label="(197, 1, 12, 64)", style=solid]; -"551 view_54" -> "552 view_54_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; -"552 view_54_0_0_nncf_smooth_quant_0" -> "553 quantize_per_tensor_default_40" [label="(197, 768)", style=solid]; -"553 quantize_per_tensor_default_40" -> "554 dequantize_per_tensor_default_40" [label="(197, 768)", style=solid]; -"554 dequantize_per_tensor_default_40" -> "559 linear_25" [label="(197, 768)", style=solid]; -"555 linear_25_scale_0" -> "558 dequantize_per_channel_default_26" [label="(768,)", style=solid]; -"556 linear_25_zero_point_0" -> "558 dequantize_per_channel_default_26" [label="(768,)", style=solid]; -"557 _frozen_param26" -> "558 dequantize_per_channel_default_26" [label="(768, 768)", style=solid]; -"558 dequantize_per_channel_default_26" -> "559 linear_25" [label="(768, 768)", style=solid]; -"559 linear_25" -> "560 view_55" [label="(197, 768)", style=solid]; -"560 view_55" -> "561 transpose_41" [label="(197, 1, 768)", style=solid]; -"561 transpose_41" -> "562 dropout_19" [label="(1, 197, 768)", style=solid]; -"562 dropout_19" -> "563 add_13" [label="(1, 197, 768)", style=solid]; -"563 add_13" -> "564 layer_norm_13" [label="(1, 197, 768)", style=solid]; -"563 add_13" -> "584 add_14" [label="(1, 197, 768)", style=solid]; -"564 layer_norm_13" -> "565 layer_norm_13_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; -"565 layer_norm_13_0_0_nncf_smooth_quant_0" -> "566 quantize_per_tensor_default_41" [label="(1, 197, 768)", style=solid]; -"566 quantize_per_tensor_default_41" -> "567 dequantize_per_tensor_default_41" [label="(1, 197, 768)", style=solid]; -"567 dequantize_per_tensor_default_41" -> "572 linear_26" [label="(1, 197, 768)", style=solid]; -"568 linear_26_scale_0" -> "571 dequantize_per_channel_default_27" [label="(3072,)", style=solid]; -"569 linear_26_zero_point_0" -> "571 dequantize_per_channel_default_27" [label="(3072,)", style=solid]; -"570 _frozen_param27" -> "571 dequantize_per_channel_default_27" [label="(3072, 768)", style=solid]; -"571 dequantize_per_channel_default_27" -> "572 linear_26" [label="(3072, 768)", style=solid]; -"572 linear_26" -> "573 gelu_6" [label="(1, 197, 3072)", style=solid]; -"573 gelu_6" -> "574 dropout_20" [label="(1, 197, 3072)", style=solid]; -"574 dropout_20" -> "575 dropout_20_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; -"575 dropout_20_0_0_nncf_smooth_quant_0" -> "576 quantize_per_tensor_default_42" [label="(1, 197, 3072)", style=solid]; -"576 quantize_per_tensor_default_42" -> "577 dequantize_per_tensor_default_42" [label="(1, 197, 3072)", style=solid]; -"577 dequantize_per_tensor_default_42" -> "582 linear_27" [label="(1, 197, 3072)", style=solid]; -"578 linear_27_scale_0" -> "581 dequantize_per_channel_default_28" [label="(768,)", style=solid]; -"579 linear_27_zero_point_0" -> "581 dequantize_per_channel_default_28" [label="(768,)", style=solid]; -"580 _frozen_param28" -> "581 dequantize_per_channel_default_28" [label="(768, 3072)", style=solid]; -"581 dequantize_per_channel_default_28" -> "582 linear_27" [label="(768, 3072)", style=solid]; -"582 linear_27" -> "583 dropout_21" [label="(1, 197, 768)", style=solid]; -"583 dropout_21" -> "584 add_14" [label="(1, 197, 768)", style=solid]; -"584 add_14" -> "585 layer_norm_14" [label="(1, 197, 768)", style=solid]; -"584 add_14" -> "630 add_15" [label="(1, 197, 768)", style=solid]; -"585 layer_norm_14" -> "586 transpose_42" [label="(1, 197, 768)", style=solid]; -"586 transpose_42" -> "587 transpose_42_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; -"587 transpose_42_0_0_nncf_smooth_quant_0" -> "588 quantize_per_tensor_default_43" [label="(197, 1, 768)", style=solid]; -"588 quantize_per_tensor_default_43" -> "589 dequantize_per_tensor_default_43" [label="(197, 1, 768)", style=solid]; -"589 dequantize_per_tensor_default_43" -> "594 linear_28" [label="(197, 1, 768)", style=solid]; -"590 linear_28_scale_0" -> "593 dequantize_per_channel_default_29" [label="(2304,)", style=solid]; -"591 linear_28_zero_point_0" -> "593 dequantize_per_channel_default_29" [label="(2304,)", style=solid]; -"592 _frozen_param29" -> "593 dequantize_per_channel_default_29" [label="(2304, 768)", style=solid]; -"593 dequantize_per_channel_default_29" -> "594 linear_28" [label="(2304, 768)", style=solid]; -"594 linear_28" -> "595 unflatten_7" [label="(197, 1, 2304)", style=solid]; -"595 unflatten_7" -> "596 unsqueeze_7" [label="(197, 1, 3, 768)", style=solid]; -"596 unsqueeze_7" -> "597 transpose_43" [label="(1, 197, 1, 3, 768)", style=solid]; -"597 transpose_43" -> "598 squeeze_7" [label="(3, 197, 1, 1, 768)", style=solid]; -"598 squeeze_7" -> "599 contiguous_7" [label="(3, 197, 1, 768)", style=solid]; -"599 contiguous_7" -> "600 quantize_per_tensor_default_44" [label="(3, 197, 1, 768)", style=solid]; -"599 contiguous_7" -> "603 quantize_per_tensor_default_45" [label="(3, 197, 1, 768)", style=solid]; -"599 contiguous_7" -> "606 select_23" [label="(3, 197, 1, 768)", style=solid]; -"600 quantize_per_tensor_default_44" -> "601 dequantize_per_tensor_default_44" [label="(3, 197, 1, 768)", style=solid]; -"601 dequantize_per_tensor_default_44" -> "602 select_21" [label="(3, 197, 1, 768)", style=solid]; -"602 select_21" -> "607 view_56" [label="(197, 1, 768)", style=solid]; -"603 quantize_per_tensor_default_45" -> "604 dequantize_per_tensor_default_45" [label="(3, 197, 1, 768)", style=solid]; -"604 dequantize_per_tensor_default_45" -> "605 select_22" [label="(3, 197, 1, 768)", style=solid]; -"605 select_22" -> "609 view_57" [label="(197, 1, 768)", style=solid]; -"606 select_23" -> "611 view_58" [label="(197, 1, 768)", style=solid]; -"607 view_56" -> "608 transpose_44" [label="(197, 12, 64)", style=solid]; -"608 transpose_44" -> "613 view_59" [label="(12, 197, 64)", style=solid]; -"609 view_57" -> "610 transpose_45" [label="(197, 12, 64)", style=solid]; -"610 transpose_45" -> "614 view_60" [label="(12, 197, 64)", style=solid]; -"611 view_58" -> "612 transpose_46" [label="(197, 12, 64)", style=solid]; -"612 transpose_46" -> "615 view_61" [label="(12, 197, 64)", style=solid]; -"613 view_59" -> "616 scaled_dot_product_attention_7" [label="(1, 12, 197, 64)", style=solid]; -"614 view_60" -> "616 scaled_dot_product_attention_7" [label="(1, 12, 197, 64)", style=solid]; -"615 view_61" -> "616 scaled_dot_product_attention_7" [label="(1, 12, 197, 64)", style=solid]; -"616 scaled_dot_product_attention_7" -> "617 permute_8" [label="(1, 12, 197, 64)", style=solid]; -"617 permute_8" -> "618 view_62" [label="(197, 1, 12, 64)", style=solid]; -"618 view_62" -> "619 view_62_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; -"619 view_62_0_0_nncf_smooth_quant_0" -> "620 quantize_per_tensor_default_46" [label="(197, 768)", style=solid]; -"620 quantize_per_tensor_default_46" -> "621 dequantize_per_tensor_default_46" [label="(197, 768)", style=solid]; -"621 dequantize_per_tensor_default_46" -> "626 linear_29" [label="(197, 768)", style=solid]; -"622 linear_29_scale_0" -> "625 dequantize_per_channel_default_30" [label="(768,)", style=solid]; -"623 linear_29_zero_point_0" -> "625 dequantize_per_channel_default_30" [label="(768,)", style=solid]; -"624 _frozen_param30" -> "625 dequantize_per_channel_default_30" [label="(768, 768)", style=solid]; -"625 dequantize_per_channel_default_30" -> "626 linear_29" [label="(768, 768)", style=solid]; -"626 linear_29" -> "627 view_63" [label="(197, 768)", style=solid]; -"627 view_63" -> "628 transpose_47" [label="(197, 1, 768)", style=solid]; -"628 transpose_47" -> "629 dropout_22" [label="(1, 197, 768)", style=solid]; -"629 dropout_22" -> "630 add_15" [label="(1, 197, 768)", style=solid]; -"630 add_15" -> "631 layer_norm_15" [label="(1, 197, 768)", style=solid]; -"630 add_15" -> "651 add_16" [label="(1, 197, 768)", style=solid]; -"631 layer_norm_15" -> "632 layer_norm_15_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; -"632 layer_norm_15_0_0_nncf_smooth_quant_0" -> "633 quantize_per_tensor_default_47" [label="(1, 197, 768)", style=solid]; -"633 quantize_per_tensor_default_47" -> "634 dequantize_per_tensor_default_47" [label="(1, 197, 768)", style=solid]; -"634 dequantize_per_tensor_default_47" -> "639 linear_30" [label="(1, 197, 768)", style=solid]; -"635 linear_30_scale_0" -> "638 dequantize_per_channel_default_31" [label="(3072,)", style=solid]; -"636 linear_30_zero_point_0" -> "638 dequantize_per_channel_default_31" [label="(3072,)", style=solid]; -"637 _frozen_param31" -> "638 dequantize_per_channel_default_31" [label="(3072, 768)", style=solid]; -"638 dequantize_per_channel_default_31" -> "639 linear_30" [label="(3072, 768)", style=solid]; -"639 linear_30" -> "640 gelu_7" [label="(1, 197, 3072)", style=solid]; -"640 gelu_7" -> "641 dropout_23" [label="(1, 197, 3072)", style=solid]; -"641 dropout_23" -> "642 dropout_23_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; -"642 dropout_23_0_0_nncf_smooth_quant_0" -> "643 quantize_per_tensor_default_48" [label="(1, 197, 3072)", style=solid]; -"643 quantize_per_tensor_default_48" -> "644 dequantize_per_tensor_default_48" [label="(1, 197, 3072)", style=solid]; -"644 dequantize_per_tensor_default_48" -> "649 linear_31" [label="(1, 197, 3072)", style=solid]; -"645 linear_31_scale_0" -> "648 dequantize_per_channel_default_32" [label="(768,)", style=solid]; -"646 linear_31_zero_point_0" -> "648 dequantize_per_channel_default_32" [label="(768,)", style=solid]; -"647 _frozen_param32" -> "648 dequantize_per_channel_default_32" [label="(768, 3072)", style=solid]; -"648 dequantize_per_channel_default_32" -> "649 linear_31" [label="(768, 3072)", style=solid]; -"649 linear_31" -> "650 dropout_24" [label="(1, 197, 768)", style=solid]; -"650 dropout_24" -> "651 add_16" [label="(1, 197, 768)", style=solid]; -"651 add_16" -> "652 layer_norm_16" [label="(1, 197, 768)", style=solid]; -"651 add_16" -> "697 add_17" [label="(1, 197, 768)", style=solid]; -"652 layer_norm_16" -> "653 transpose_48" [label="(1, 197, 768)", style=solid]; -"653 transpose_48" -> "654 transpose_48_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; -"654 transpose_48_0_0_nncf_smooth_quant_0" -> "655 quantize_per_tensor_default_49" [label="(197, 1, 768)", style=solid]; -"655 quantize_per_tensor_default_49" -> "656 dequantize_per_tensor_default_49" [label="(197, 1, 768)", style=solid]; -"656 dequantize_per_tensor_default_49" -> "661 linear_32" [label="(197, 1, 768)", style=solid]; -"657 linear_32_scale_0" -> "660 dequantize_per_channel_default_33" [label="(2304,)", style=solid]; -"658 linear_32_zero_point_0" -> "660 dequantize_per_channel_default_33" [label="(2304,)", style=solid]; -"659 _frozen_param33" -> "660 dequantize_per_channel_default_33" [label="(2304, 768)", style=solid]; -"660 dequantize_per_channel_default_33" -> "661 linear_32" [label="(2304, 768)", style=solid]; -"661 linear_32" -> "662 unflatten_8" [label="(197, 1, 2304)", style=solid]; -"662 unflatten_8" -> "663 unsqueeze_8" [label="(197, 1, 3, 768)", style=solid]; -"663 unsqueeze_8" -> "664 transpose_49" [label="(1, 197, 1, 3, 768)", style=solid]; -"664 transpose_49" -> "665 squeeze_8" [label="(3, 197, 1, 1, 768)", style=solid]; -"665 squeeze_8" -> "666 contiguous_8" [label="(3, 197, 1, 768)", style=solid]; -"666 contiguous_8" -> "667 quantize_per_tensor_default_50" [label="(3, 197, 1, 768)", style=solid]; -"666 contiguous_8" -> "670 quantize_per_tensor_default_51" [label="(3, 197, 1, 768)", style=solid]; -"666 contiguous_8" -> "673 select_26" [label="(3, 197, 1, 768)", style=solid]; -"667 quantize_per_tensor_default_50" -> "668 dequantize_per_tensor_default_50" [label="(3, 197, 1, 768)", style=solid]; -"668 dequantize_per_tensor_default_50" -> "669 select_24" [label="(3, 197, 1, 768)", style=solid]; -"669 select_24" -> "674 view_64" [label="(197, 1, 768)", style=solid]; -"670 quantize_per_tensor_default_51" -> "671 dequantize_per_tensor_default_51" [label="(3, 197, 1, 768)", style=solid]; -"671 dequantize_per_tensor_default_51" -> "672 select_25" [label="(3, 197, 1, 768)", style=solid]; -"672 select_25" -> "676 view_65" [label="(197, 1, 768)", style=solid]; -"673 select_26" -> "678 view_66" [label="(197, 1, 768)", style=solid]; -"674 view_64" -> "675 transpose_50" [label="(197, 12, 64)", style=solid]; -"675 transpose_50" -> "680 view_67" [label="(12, 197, 64)", style=solid]; -"676 view_65" -> "677 transpose_51" [label="(197, 12, 64)", style=solid]; -"677 transpose_51" -> "681 view_68" [label="(12, 197, 64)", style=solid]; -"678 view_66" -> "679 transpose_52" [label="(197, 12, 64)", style=solid]; -"679 transpose_52" -> "682 view_69" [label="(12, 197, 64)", style=solid]; -"680 view_67" -> "683 scaled_dot_product_attention_8" [label="(1, 12, 197, 64)", style=solid]; -"681 view_68" -> "683 scaled_dot_product_attention_8" [label="(1, 12, 197, 64)", style=solid]; -"682 view_69" -> "683 scaled_dot_product_attention_8" [label="(1, 12, 197, 64)", style=solid]; -"683 scaled_dot_product_attention_8" -> "684 permute_9" [label="(1, 12, 197, 64)", style=solid]; -"684 permute_9" -> "685 view_70" [label="(197, 1, 12, 64)", style=solid]; -"685 view_70" -> "686 view_70_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; -"686 view_70_0_0_nncf_smooth_quant_0" -> "687 quantize_per_tensor_default_52" [label="(197, 768)", style=solid]; -"687 quantize_per_tensor_default_52" -> "688 dequantize_per_tensor_default_52" [label="(197, 768)", style=solid]; -"688 dequantize_per_tensor_default_52" -> "693 linear_33" [label="(197, 768)", style=solid]; -"689 linear_33_scale_0" -> "692 dequantize_per_channel_default_34" [label="(768,)", style=solid]; -"690 linear_33_zero_point_0" -> "692 dequantize_per_channel_default_34" [label="(768,)", style=solid]; -"691 _frozen_param34" -> "692 dequantize_per_channel_default_34" [label="(768, 768)", style=solid]; -"692 dequantize_per_channel_default_34" -> "693 linear_33" [label="(768, 768)", style=solid]; -"693 linear_33" -> "694 view_71" [label="(197, 768)", style=solid]; -"694 view_71" -> "695 transpose_53" [label="(197, 1, 768)", style=solid]; -"695 transpose_53" -> "696 dropout_25" [label="(1, 197, 768)", style=solid]; -"696 dropout_25" -> "697 add_17" [label="(1, 197, 768)", style=solid]; -"697 add_17" -> "698 layer_norm_17" [label="(1, 197, 768)", style=solid]; -"697 add_17" -> "718 add_18" [label="(1, 197, 768)", style=solid]; -"698 layer_norm_17" -> "699 layer_norm_17_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; -"699 layer_norm_17_0_0_nncf_smooth_quant_0" -> "700 quantize_per_tensor_default_53" [label="(1, 197, 768)", style=solid]; -"700 quantize_per_tensor_default_53" -> "701 dequantize_per_tensor_default_53" [label="(1, 197, 768)", style=solid]; -"701 dequantize_per_tensor_default_53" -> "706 linear_34" [label="(1, 197, 768)", style=solid]; -"702 linear_34_scale_0" -> "705 dequantize_per_channel_default_35" [label="(3072,)", style=solid]; -"703 linear_34_zero_point_0" -> "705 dequantize_per_channel_default_35" [label="(3072,)", style=solid]; -"704 _frozen_param35" -> "705 dequantize_per_channel_default_35" [label="(3072, 768)", style=solid]; -"705 dequantize_per_channel_default_35" -> "706 linear_34" [label="(3072, 768)", style=solid]; -"706 linear_34" -> "707 gelu_8" [label="(1, 197, 3072)", style=solid]; -"707 gelu_8" -> "708 dropout_26" [label="(1, 197, 3072)", style=solid]; -"708 dropout_26" -> "709 dropout_26_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; -"709 dropout_26_0_0_nncf_smooth_quant_0" -> "710 quantize_per_tensor_default_54" [label="(1, 197, 3072)", style=solid]; -"710 quantize_per_tensor_default_54" -> "711 dequantize_per_tensor_default_54" [label="(1, 197, 3072)", style=solid]; -"711 dequantize_per_tensor_default_54" -> "716 linear_35" [label="(1, 197, 3072)", style=solid]; -"712 linear_35_scale_0" -> "715 dequantize_per_channel_default_36" [label="(768,)", style=solid]; -"713 linear_35_zero_point_0" -> "715 dequantize_per_channel_default_36" [label="(768,)", style=solid]; -"714 _frozen_param36" -> "715 dequantize_per_channel_default_36" [label="(768, 3072)", style=solid]; -"715 dequantize_per_channel_default_36" -> "716 linear_35" [label="(768, 3072)", style=solid]; -"716 linear_35" -> "717 dropout_27" [label="(1, 197, 768)", style=solid]; -"717 dropout_27" -> "718 add_18" [label="(1, 197, 768)", style=solid]; -"718 add_18" -> "719 layer_norm_18" [label="(1, 197, 768)", style=solid]; -"718 add_18" -> "764 add_19" [label="(1, 197, 768)", style=solid]; -"719 layer_norm_18" -> "720 transpose_54" [label="(1, 197, 768)", style=solid]; -"720 transpose_54" -> "721 transpose_54_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; -"721 transpose_54_0_0_nncf_smooth_quant_0" -> "722 quantize_per_tensor_default_55" [label="(197, 1, 768)", style=solid]; -"722 quantize_per_tensor_default_55" -> "723 dequantize_per_tensor_default_55" [label="(197, 1, 768)", style=solid]; -"723 dequantize_per_tensor_default_55" -> "728 linear_36" [label="(197, 1, 768)", style=solid]; -"724 linear_36_scale_0" -> "727 dequantize_per_channel_default_37" [label="(2304,)", style=solid]; -"725 linear_36_zero_point_0" -> "727 dequantize_per_channel_default_37" [label="(2304,)", style=solid]; -"726 _frozen_param37" -> "727 dequantize_per_channel_default_37" [label="(2304, 768)", style=solid]; -"727 dequantize_per_channel_default_37" -> "728 linear_36" [label="(2304, 768)", style=solid]; -"728 linear_36" -> "729 unflatten_9" [label="(197, 1, 2304)", style=solid]; -"729 unflatten_9" -> "730 unsqueeze_9" [label="(197, 1, 3, 768)", style=solid]; -"730 unsqueeze_9" -> "731 transpose_55" [label="(1, 197, 1, 3, 768)", style=solid]; -"731 transpose_55" -> "732 squeeze_9" [label="(3, 197, 1, 1, 768)", style=solid]; -"732 squeeze_9" -> "733 contiguous_9" [label="(3, 197, 1, 768)", style=solid]; -"733 contiguous_9" -> "734 quantize_per_tensor_default_56" [label="(3, 197, 1, 768)", style=solid]; -"733 contiguous_9" -> "737 quantize_per_tensor_default_57" [label="(3, 197, 1, 768)", style=solid]; -"733 contiguous_9" -> "740 select_29" [label="(3, 197, 1, 768)", style=solid]; -"734 quantize_per_tensor_default_56" -> "735 dequantize_per_tensor_default_56" [label="(3, 197, 1, 768)", style=solid]; -"735 dequantize_per_tensor_default_56" -> "736 select_27" [label="(3, 197, 1, 768)", style=solid]; -"736 select_27" -> "741 view_72" [label="(197, 1, 768)", style=solid]; -"737 quantize_per_tensor_default_57" -> "738 dequantize_per_tensor_default_57" [label="(3, 197, 1, 768)", style=solid]; -"738 dequantize_per_tensor_default_57" -> "739 select_28" [label="(3, 197, 1, 768)", style=solid]; -"739 select_28" -> "743 view_73" [label="(197, 1, 768)", style=solid]; -"740 select_29" -> "745 view_74" [label="(197, 1, 768)", style=solid]; -"741 view_72" -> "742 transpose_56" [label="(197, 12, 64)", style=solid]; -"742 transpose_56" -> "747 view_75" [label="(12, 197, 64)", style=solid]; -"743 view_73" -> "744 transpose_57" [label="(197, 12, 64)", style=solid]; -"744 transpose_57" -> "748 view_76" [label="(12, 197, 64)", style=solid]; -"745 view_74" -> "746 transpose_58" [label="(197, 12, 64)", style=solid]; -"746 transpose_58" -> "749 view_77" [label="(12, 197, 64)", style=solid]; -"747 view_75" -> "750 scaled_dot_product_attention_9" [label="(1, 12, 197, 64)", style=solid]; -"748 view_76" -> "750 scaled_dot_product_attention_9" [label="(1, 12, 197, 64)", style=solid]; -"749 view_77" -> "750 scaled_dot_product_attention_9" [label="(1, 12, 197, 64)", style=solid]; -"750 scaled_dot_product_attention_9" -> "751 permute_10" [label="(1, 12, 197, 64)", style=solid]; -"751 permute_10" -> "752 view_78" [label="(197, 1, 12, 64)", style=solid]; -"752 view_78" -> "753 view_78_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; -"753 view_78_0_0_nncf_smooth_quant_0" -> "754 quantize_per_tensor_default_58" [label="(197, 768)", style=solid]; -"754 quantize_per_tensor_default_58" -> "755 dequantize_per_tensor_default_58" [label="(197, 768)", style=solid]; -"755 dequantize_per_tensor_default_58" -> "760 linear_37" [label="(197, 768)", style=solid]; -"756 linear_37_scale_0" -> "759 dequantize_per_channel_default_38" [label="(768,)", style=solid]; -"757 linear_37_zero_point_0" -> "759 dequantize_per_channel_default_38" [label="(768,)", style=solid]; -"758 _frozen_param38" -> "759 dequantize_per_channel_default_38" [label="(768, 768)", style=solid]; -"759 dequantize_per_channel_default_38" -> "760 linear_37" [label="(768, 768)", style=solid]; -"760 linear_37" -> "761 view_79" [label="(197, 768)", style=solid]; -"761 view_79" -> "762 transpose_59" [label="(197, 1, 768)", style=solid]; -"762 transpose_59" -> "763 dropout_28" [label="(1, 197, 768)", style=solid]; -"763 dropout_28" -> "764 add_19" [label="(1, 197, 768)", style=solid]; -"764 add_19" -> "765 layer_norm_19" [label="(1, 197, 768)", style=solid]; -"764 add_19" -> "785 add_20" [label="(1, 197, 768)", style=solid]; -"765 layer_norm_19" -> "766 layer_norm_19_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; -"766 layer_norm_19_0_0_nncf_smooth_quant_0" -> "767 quantize_per_tensor_default_59" [label="(1, 197, 768)", style=solid]; -"767 quantize_per_tensor_default_59" -> "768 dequantize_per_tensor_default_59" [label="(1, 197, 768)", style=solid]; -"768 dequantize_per_tensor_default_59" -> "773 linear_38" [label="(1, 197, 768)", style=solid]; -"769 linear_38_scale_0" -> "772 dequantize_per_channel_default_39" [label="(3072,)", style=solid]; -"770 linear_38_zero_point_0" -> "772 dequantize_per_channel_default_39" [label="(3072,)", style=solid]; -"771 _frozen_param39" -> "772 dequantize_per_channel_default_39" [label="(3072, 768)", style=solid]; -"772 dequantize_per_channel_default_39" -> "773 linear_38" [label="(3072, 768)", style=solid]; -"773 linear_38" -> "774 gelu_9" [label="(1, 197, 3072)", style=solid]; -"774 gelu_9" -> "775 dropout_29" [label="(1, 197, 3072)", style=solid]; -"775 dropout_29" -> "776 dropout_29_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; -"776 dropout_29_0_0_nncf_smooth_quant_0" -> "777 quantize_per_tensor_default_60" [label="(1, 197, 3072)", style=solid]; -"777 quantize_per_tensor_default_60" -> "778 dequantize_per_tensor_default_60" [label="(1, 197, 3072)", style=solid]; -"778 dequantize_per_tensor_default_60" -> "783 linear_39" [label="(1, 197, 3072)", style=solid]; -"779 linear_39_scale_0" -> "782 dequantize_per_channel_default_40" [label="(768,)", style=solid]; -"780 linear_39_zero_point_0" -> "782 dequantize_per_channel_default_40" [label="(768,)", style=solid]; -"781 _frozen_param40" -> "782 dequantize_per_channel_default_40" [label="(768, 3072)", style=solid]; -"782 dequantize_per_channel_default_40" -> "783 linear_39" [label="(768, 3072)", style=solid]; -"783 linear_39" -> "784 dropout_30" [label="(1, 197, 768)", style=solid]; -"784 dropout_30" -> "785 add_20" [label="(1, 197, 768)", style=solid]; -"785 add_20" -> "786 layer_norm_20" [label="(1, 197, 768)", style=solid]; -"785 add_20" -> "831 add_21" [label="(1, 197, 768)", style=solid]; -"786 layer_norm_20" -> "787 transpose_60" [label="(1, 197, 768)", style=solid]; -"787 transpose_60" -> "788 transpose_60_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; -"788 transpose_60_0_0_nncf_smooth_quant_0" -> "789 quantize_per_tensor_default_61" [label="(197, 1, 768)", style=solid]; -"789 quantize_per_tensor_default_61" -> "790 dequantize_per_tensor_default_61" [label="(197, 1, 768)", style=solid]; -"790 dequantize_per_tensor_default_61" -> "795 linear_40" [label="(197, 1, 768)", style=solid]; -"791 linear_40_scale_0" -> "794 dequantize_per_channel_default_41" [label="(2304,)", style=solid]; -"792 linear_40_zero_point_0" -> "794 dequantize_per_channel_default_41" [label="(2304,)", style=solid]; -"793 _frozen_param41" -> "794 dequantize_per_channel_default_41" [label="(2304, 768)", style=solid]; -"794 dequantize_per_channel_default_41" -> "795 linear_40" [label="(2304, 768)", style=solid]; -"795 linear_40" -> "796 unflatten_10" [label="(197, 1, 2304)", style=solid]; -"796 unflatten_10" -> "797 unsqueeze_10" [label="(197, 1, 3, 768)", style=solid]; -"797 unsqueeze_10" -> "798 transpose_61" [label="(1, 197, 1, 3, 768)", style=solid]; -"798 transpose_61" -> "799 squeeze_10" [label="(3, 197, 1, 1, 768)", style=solid]; -"799 squeeze_10" -> "800 contiguous_10" [label="(3, 197, 1, 768)", style=solid]; -"800 contiguous_10" -> "801 quantize_per_tensor_default_62" [label="(3, 197, 1, 768)", style=solid]; -"800 contiguous_10" -> "804 quantize_per_tensor_default_63" [label="(3, 197, 1, 768)", style=solid]; -"800 contiguous_10" -> "807 select_32" [label="(3, 197, 1, 768)", style=solid]; -"801 quantize_per_tensor_default_62" -> "802 dequantize_per_tensor_default_62" [label="(3, 197, 1, 768)", style=solid]; -"802 dequantize_per_tensor_default_62" -> "803 select_30" [label="(3, 197, 1, 768)", style=solid]; -"803 select_30" -> "808 view_80" [label="(197, 1, 768)", style=solid]; -"804 quantize_per_tensor_default_63" -> "805 dequantize_per_tensor_default_63" [label="(3, 197, 1, 768)", style=solid]; -"805 dequantize_per_tensor_default_63" -> "806 select_31" [label="(3, 197, 1, 768)", style=solid]; -"806 select_31" -> "810 view_81" [label="(197, 1, 768)", style=solid]; -"807 select_32" -> "812 view_82" [label="(197, 1, 768)", style=solid]; -"808 view_80" -> "809 transpose_62" [label="(197, 12, 64)", style=solid]; -"809 transpose_62" -> "814 view_83" [label="(12, 197, 64)", style=solid]; -"810 view_81" -> "811 transpose_63" [label="(197, 12, 64)", style=solid]; -"811 transpose_63" -> "815 view_84" [label="(12, 197, 64)", style=solid]; -"812 view_82" -> "813 transpose_64" [label="(197, 12, 64)", style=solid]; -"813 transpose_64" -> "816 view_85" [label="(12, 197, 64)", style=solid]; -"814 view_83" -> "817 scaled_dot_product_attention_10" [label="(1, 12, 197, 64)", style=solid]; -"815 view_84" -> "817 scaled_dot_product_attention_10" [label="(1, 12, 197, 64)", style=solid]; -"816 view_85" -> "817 scaled_dot_product_attention_10" [label="(1, 12, 197, 64)", style=solid]; -"817 scaled_dot_product_attention_10" -> "818 permute_11" [label="(1, 12, 197, 64)", style=solid]; -"818 permute_11" -> "819 view_86" [label="(197, 1, 12, 64)", style=solid]; -"819 view_86" -> "820 view_86_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; -"820 view_86_0_0_nncf_smooth_quant_0" -> "821 quantize_per_tensor_default_64" [label="(197, 768)", style=solid]; -"821 quantize_per_tensor_default_64" -> "822 dequantize_per_tensor_default_64" [label="(197, 768)", style=solid]; -"822 dequantize_per_tensor_default_64" -> "827 linear_41" [label="(197, 768)", style=solid]; -"823 linear_41_scale_0" -> "826 dequantize_per_channel_default_42" [label="(768,)", style=solid]; -"824 linear_41_zero_point_0" -> "826 dequantize_per_channel_default_42" [label="(768,)", style=solid]; -"825 _frozen_param42" -> "826 dequantize_per_channel_default_42" [label="(768, 768)", style=solid]; -"826 dequantize_per_channel_default_42" -> "827 linear_41" [label="(768, 768)", style=solid]; -"827 linear_41" -> "828 view_87" [label="(197, 768)", style=solid]; -"828 view_87" -> "829 transpose_65" [label="(197, 1, 768)", style=solid]; -"829 transpose_65" -> "830 dropout_31" [label="(1, 197, 768)", style=solid]; -"830 dropout_31" -> "831 add_21" [label="(1, 197, 768)", style=solid]; -"831 add_21" -> "832 layer_norm_21" [label="(1, 197, 768)", style=solid]; -"831 add_21" -> "852 add_22" [label="(1, 197, 768)", style=solid]; -"832 layer_norm_21" -> "833 layer_norm_21_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; -"833 layer_norm_21_0_0_nncf_smooth_quant_0" -> "834 quantize_per_tensor_default_65" [label="(1, 197, 768)", style=solid]; -"834 quantize_per_tensor_default_65" -> "835 dequantize_per_tensor_default_65" [label="(1, 197, 768)", style=solid]; -"835 dequantize_per_tensor_default_65" -> "840 linear_42" [label="(1, 197, 768)", style=solid]; -"836 linear_42_scale_0" -> "839 dequantize_per_channel_default_43" [label="(3072,)", style=solid]; -"837 linear_42_zero_point_0" -> "839 dequantize_per_channel_default_43" [label="(3072,)", style=solid]; -"838 _frozen_param43" -> "839 dequantize_per_channel_default_43" [label="(3072, 768)", style=solid]; -"839 dequantize_per_channel_default_43" -> "840 linear_42" [label="(3072, 768)", style=solid]; -"840 linear_42" -> "841 gelu_10" [label="(1, 197, 3072)", style=solid]; -"841 gelu_10" -> "842 dropout_32" [label="(1, 197, 3072)", style=solid]; -"842 dropout_32" -> "843 dropout_32_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; -"843 dropout_32_0_0_nncf_smooth_quant_0" -> "844 quantize_per_tensor_default_66" [label="(1, 197, 3072)", style=solid]; -"844 quantize_per_tensor_default_66" -> "845 dequantize_per_tensor_default_66" [label="(1, 197, 3072)", style=solid]; -"845 dequantize_per_tensor_default_66" -> "850 linear_43" [label="(1, 197, 3072)", style=solid]; -"846 linear_43_scale_0" -> "849 dequantize_per_channel_default_44" [label="(768,)", style=solid]; -"847 linear_43_zero_point_0" -> "849 dequantize_per_channel_default_44" [label="(768,)", style=solid]; -"848 _frozen_param44" -> "849 dequantize_per_channel_default_44" [label="(768, 3072)", style=solid]; -"849 dequantize_per_channel_default_44" -> "850 linear_43" [label="(768, 3072)", style=solid]; -"850 linear_43" -> "851 dropout_33" [label="(1, 197, 768)", style=solid]; -"851 dropout_33" -> "852 add_22" [label="(1, 197, 768)", style=solid]; -"852 add_22" -> "853 layer_norm_22" [label="(1, 197, 768)", style=solid]; -"852 add_22" -> "898 add_23" [label="(1, 197, 768)", style=solid]; -"853 layer_norm_22" -> "854 transpose_66" [label="(1, 197, 768)", style=solid]; -"854 transpose_66" -> "855 transpose_66_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; -"855 transpose_66_0_0_nncf_smooth_quant_0" -> "856 quantize_per_tensor_default_67" [label="(197, 1, 768)", style=solid]; -"856 quantize_per_tensor_default_67" -> "857 dequantize_per_tensor_default_67" [label="(197, 1, 768)", style=solid]; -"857 dequantize_per_tensor_default_67" -> "862 linear_44" [label="(197, 1, 768)", style=solid]; -"858 linear_44_scale_0" -> "861 dequantize_per_channel_default_45" [label="(2304,)", style=solid]; -"859 linear_44_zero_point_0" -> "861 dequantize_per_channel_default_45" [label="(2304,)", style=solid]; -"860 _frozen_param45" -> "861 dequantize_per_channel_default_45" [label="(2304, 768)", style=solid]; -"861 dequantize_per_channel_default_45" -> "862 linear_44" [label="(2304, 768)", style=solid]; -"862 linear_44" -> "863 unflatten_11" [label="(197, 1, 2304)", style=solid]; -"863 unflatten_11" -> "864 unsqueeze_11" [label="(197, 1, 3, 768)", style=solid]; -"864 unsqueeze_11" -> "865 transpose_67" [label="(1, 197, 1, 3, 768)", style=solid]; -"865 transpose_67" -> "866 squeeze_11" [label="(3, 197, 1, 1, 768)", style=solid]; -"866 squeeze_11" -> "867 contiguous_11" [label="(3, 197, 1, 768)", style=solid]; -"867 contiguous_11" -> "868 quantize_per_tensor_default_68" [label="(3, 197, 1, 768)", style=solid]; -"867 contiguous_11" -> "871 quantize_per_tensor_default_69" [label="(3, 197, 1, 768)", style=solid]; -"867 contiguous_11" -> "874 select_35" [label="(3, 197, 1, 768)", style=solid]; -"868 quantize_per_tensor_default_68" -> "869 dequantize_per_tensor_default_68" [label="(3, 197, 1, 768)", style=solid]; -"869 dequantize_per_tensor_default_68" -> "870 select_33" [label="(3, 197, 1, 768)", style=solid]; -"870 select_33" -> "875 view_88" [label="(197, 1, 768)", style=solid]; -"871 quantize_per_tensor_default_69" -> "872 dequantize_per_tensor_default_69" [label="(3, 197, 1, 768)", style=solid]; -"872 dequantize_per_tensor_default_69" -> "873 select_34" [label="(3, 197, 1, 768)", style=solid]; -"873 select_34" -> "877 view_89" [label="(197, 1, 768)", style=solid]; -"874 select_35" -> "879 view_90" [label="(197, 1, 768)", style=solid]; -"875 view_88" -> "876 transpose_68" [label="(197, 12, 64)", style=solid]; -"876 transpose_68" -> "881 view_91" [label="(12, 197, 64)", style=solid]; -"877 view_89" -> "878 transpose_69" [label="(197, 12, 64)", style=solid]; -"878 transpose_69" -> "882 view_92" [label="(12, 197, 64)", style=solid]; -"879 view_90" -> "880 transpose_70" [label="(197, 12, 64)", style=solid]; -"880 transpose_70" -> "883 view_93" [label="(12, 197, 64)", style=solid]; -"881 view_91" -> "884 scaled_dot_product_attention_11" [label="(1, 12, 197, 64)", style=solid]; -"882 view_92" -> "884 scaled_dot_product_attention_11" [label="(1, 12, 197, 64)", style=solid]; -"883 view_93" -> "884 scaled_dot_product_attention_11" [label="(1, 12, 197, 64)", style=solid]; -"884 scaled_dot_product_attention_11" -> "885 permute_12" [label="(1, 12, 197, 64)", style=solid]; -"885 permute_12" -> "886 view_94" [label="(197, 1, 12, 64)", style=solid]; -"886 view_94" -> "887 view_94_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; -"887 view_94_0_0_nncf_smooth_quant_0" -> "888 quantize_per_tensor_default_70" [label="(197, 768)", style=solid]; -"888 quantize_per_tensor_default_70" -> "889 dequantize_per_tensor_default_70" [label="(197, 768)", style=solid]; -"889 dequantize_per_tensor_default_70" -> "894 linear_45" [label="(197, 768)", style=solid]; -"890 linear_45_scale_0" -> "893 dequantize_per_channel_default_46" [label="(768,)", style=solid]; -"891 linear_45_zero_point_0" -> "893 dequantize_per_channel_default_46" [label="(768,)", style=solid]; -"892 _frozen_param46" -> "893 dequantize_per_channel_default_46" [label="(768, 768)", style=solid]; -"893 dequantize_per_channel_default_46" -> "894 linear_45" [label="(768, 768)", style=solid]; -"894 linear_45" -> "895 view_95" [label="(197, 768)", style=solid]; -"895 view_95" -> "896 transpose_71" [label="(197, 1, 768)", style=solid]; -"896 transpose_71" -> "897 dropout_34" [label="(1, 197, 768)", style=solid]; -"897 dropout_34" -> "898 add_23" [label="(1, 197, 768)", style=solid]; -"898 add_23" -> "899 layer_norm_23" [label="(1, 197, 768)", style=solid]; -"898 add_23" -> "919 add_24" [label="(1, 197, 768)", style=solid]; -"899 layer_norm_23" -> "900 layer_norm_23_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; -"900 layer_norm_23_0_0_nncf_smooth_quant_0" -> "901 quantize_per_tensor_default_71" [label="(1, 197, 768)", style=solid]; -"901 quantize_per_tensor_default_71" -> "902 dequantize_per_tensor_default_71" [label="(1, 197, 768)", style=solid]; -"902 dequantize_per_tensor_default_71" -> "907 linear_46" [label="(1, 197, 768)", style=solid]; -"903 linear_46_scale_0" -> "906 dequantize_per_channel_default_47" [label="(3072,)", style=solid]; -"904 linear_46_zero_point_0" -> "906 dequantize_per_channel_default_47" [label="(3072,)", style=solid]; -"905 _frozen_param47" -> "906 dequantize_per_channel_default_47" [label="(3072, 768)", style=solid]; -"906 dequantize_per_channel_default_47" -> "907 linear_46" [label="(3072, 768)", style=solid]; -"907 linear_46" -> "908 gelu_11" [label="(1, 197, 3072)", style=solid]; -"908 gelu_11" -> "909 dropout_35" [label="(1, 197, 3072)", style=solid]; -"909 dropout_35" -> "910 dropout_35_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; -"910 dropout_35_0_0_nncf_smooth_quant_0" -> "911 quantize_per_tensor_default_72" [label="(1, 197, 3072)", style=solid]; -"911 quantize_per_tensor_default_72" -> "912 dequantize_per_tensor_default_72" [label="(1, 197, 3072)", style=solid]; -"912 dequantize_per_tensor_default_72" -> "917 linear_47" [label="(1, 197, 3072)", style=solid]; -"913 linear_47_scale_0" -> "916 dequantize_per_channel_default_48" [label="(768,)", style=solid]; -"914 linear_47_zero_point_0" -> "916 dequantize_per_channel_default_48" [label="(768,)", style=solid]; -"915 _frozen_param48" -> "916 dequantize_per_channel_default_48" [label="(768, 3072)", style=solid]; -"916 dequantize_per_channel_default_48" -> "917 linear_47" [label="(768, 3072)", style=solid]; -"917 linear_47" -> "918 dropout_36" [label="(1, 197, 768)", style=solid]; -"918 dropout_36" -> "919 add_24" [label="(1, 197, 768)", style=solid]; -"919 add_24" -> "920 layer_norm_24" [label="(1, 197, 768)", style=solid]; -"920 layer_norm_24" -> "921 slice_1" [label="(1, 197, 768)", style=solid]; -"921 slice_1" -> "922 select_36" [label="(1, 197, 768)", style=solid]; -"922 select_36" -> "923 select_36_0_0_nncf_smooth_quant_0" [label="(1, 768)", style=solid]; -"923 select_36_0_0_nncf_smooth_quant_0" -> "924 quantize_per_tensor_default_73" [label="(1, 768)", style=solid]; -"924 quantize_per_tensor_default_73" -> "925 dequantize_per_tensor_default_73" [label="(1, 768)", style=solid]; -"925 dequantize_per_tensor_default_73" -> "930 linear_48" [label="(1, 768)", style=solid]; -"926 linear_48_scale_0" -> "929 dequantize_per_channel_default_49" [label="(1000,)", style=solid]; -"927 linear_48_zero_point_0" -> "929 dequantize_per_channel_default_49" [label="(1000,)", style=solid]; -"928 _frozen_param49" -> "929 dequantize_per_channel_default_49" [label="(1000, 768)", style=solid]; -"929 dequantize_per_channel_default_49" -> "930 linear_48" [label="(1000, 768)", style=solid]; -"930 linear_48" -> "931 output" [label="(1, 1000)", style=solid]; +"0 encoder_pos_embedding" [id=0, type=get_attr]; +"1 conv_proj_bias" [id=1, type=get_attr]; +"2 encoder_layers_encoder_layer_0_ln_1_weight" [id=2, type=get_attr]; +"3 encoder_layers_encoder_layer_0_ln_1_bias" [id=3, type=get_attr]; +"4 encoder_layers_encoder_layer_0_self_attention_in_proj_bias" [id=4, type=get_attr]; +"5 encoder_layers_encoder_layer_0_self_attention_out_proj_bias" [id=5, type=get_attr]; +"6 encoder_layers_encoder_layer_0_ln_2_weight" [id=6, type=get_attr]; +"7 encoder_layers_encoder_layer_0_ln_2_bias" [id=7, type=get_attr]; +"8 encoder_layers_encoder_layer_0_mlp_0_bias" [id=8, type=get_attr]; +"9 encoder_layers_encoder_layer_0_mlp_3_bias" [id=9, type=get_attr]; +"10 encoder_layers_encoder_layer_1_ln_1_weight" [id=10, type=get_attr]; +"11 encoder_layers_encoder_layer_1_ln_1_bias" [id=11, type=get_attr]; +"12 encoder_layers_encoder_layer_1_self_attention_in_proj_bias" [id=12, type=get_attr]; +"13 encoder_layers_encoder_layer_1_self_attention_out_proj_bias" [id=13, type=get_attr]; +"14 encoder_layers_encoder_layer_1_ln_2_weight" [id=14, type=get_attr]; +"15 encoder_layers_encoder_layer_1_ln_2_bias" [id=15, type=get_attr]; +"16 encoder_layers_encoder_layer_1_mlp_0_bias" [id=16, type=get_attr]; +"17 encoder_layers_encoder_layer_1_mlp_3_bias" [id=17, type=get_attr]; +"18 encoder_layers_encoder_layer_2_ln_1_weight" [id=18, type=get_attr]; +"19 encoder_layers_encoder_layer_2_ln_1_bias" [id=19, type=get_attr]; +"20 encoder_layers_encoder_layer_2_self_attention_in_proj_bias" [id=20, type=get_attr]; +"21 encoder_layers_encoder_layer_2_self_attention_out_proj_bias" [id=21, type=get_attr]; +"22 encoder_layers_encoder_layer_2_ln_2_weight" [id=22, type=get_attr]; +"23 encoder_layers_encoder_layer_2_ln_2_bias" [id=23, type=get_attr]; +"24 encoder_layers_encoder_layer_2_mlp_0_bias" [id=24, type=get_attr]; +"25 encoder_layers_encoder_layer_2_mlp_3_bias" [id=25, type=get_attr]; +"26 encoder_layers_encoder_layer_3_ln_1_weight" [id=26, type=get_attr]; +"27 encoder_layers_encoder_layer_3_ln_1_bias" [id=27, type=get_attr]; +"28 encoder_layers_encoder_layer_3_self_attention_in_proj_bias" [id=28, type=get_attr]; +"29 encoder_layers_encoder_layer_3_self_attention_out_proj_bias" [id=29, type=get_attr]; +"30 encoder_layers_encoder_layer_3_ln_2_weight" [id=30, type=get_attr]; +"31 encoder_layers_encoder_layer_3_ln_2_bias" [id=31, type=get_attr]; +"32 encoder_layers_encoder_layer_3_mlp_0_bias" [id=32, type=get_attr]; +"33 encoder_layers_encoder_layer_3_mlp_3_bias" [id=33, type=get_attr]; +"34 encoder_layers_encoder_layer_4_ln_1_weight" [id=34, type=get_attr]; +"35 encoder_layers_encoder_layer_4_ln_1_bias" [id=35, type=get_attr]; +"36 encoder_layers_encoder_layer_4_self_attention_in_proj_bias" [id=36, type=get_attr]; +"37 encoder_layers_encoder_layer_4_self_attention_out_proj_bias" [id=37, type=get_attr]; +"38 encoder_layers_encoder_layer_4_ln_2_weight" [id=38, type=get_attr]; +"39 encoder_layers_encoder_layer_4_ln_2_bias" [id=39, type=get_attr]; +"40 encoder_layers_encoder_layer_4_mlp_0_bias" [id=40, type=get_attr]; +"41 encoder_layers_encoder_layer_4_mlp_3_bias" [id=41, type=get_attr]; +"42 encoder_layers_encoder_layer_5_ln_1_weight" [id=42, type=get_attr]; +"43 encoder_layers_encoder_layer_5_ln_1_bias" [id=43, type=get_attr]; +"44 encoder_layers_encoder_layer_5_self_attention_in_proj_bias" [id=44, type=get_attr]; +"45 encoder_layers_encoder_layer_5_self_attention_out_proj_bias" [id=45, type=get_attr]; +"46 encoder_layers_encoder_layer_5_ln_2_weight" [id=46, type=get_attr]; +"47 encoder_layers_encoder_layer_5_ln_2_bias" [id=47, type=get_attr]; +"48 encoder_layers_encoder_layer_5_mlp_0_bias" [id=48, type=get_attr]; +"49 encoder_layers_encoder_layer_5_mlp_3_bias" [id=49, type=get_attr]; +"50 encoder_layers_encoder_layer_6_ln_1_weight" [id=50, type=get_attr]; +"51 encoder_layers_encoder_layer_6_ln_1_bias" [id=51, type=get_attr]; +"52 encoder_layers_encoder_layer_6_self_attention_in_proj_bias" [id=52, type=get_attr]; +"53 encoder_layers_encoder_layer_6_self_attention_out_proj_bias" [id=53, type=get_attr]; +"54 encoder_layers_encoder_layer_6_ln_2_weight" [id=54, type=get_attr]; +"55 encoder_layers_encoder_layer_6_ln_2_bias" [id=55, type=get_attr]; +"56 encoder_layers_encoder_layer_6_mlp_0_bias" [id=56, type=get_attr]; +"57 encoder_layers_encoder_layer_6_mlp_3_bias" [id=57, type=get_attr]; +"58 encoder_layers_encoder_layer_7_ln_1_weight" [id=58, type=get_attr]; +"59 encoder_layers_encoder_layer_7_ln_1_bias" [id=59, type=get_attr]; +"60 encoder_layers_encoder_layer_7_self_attention_in_proj_bias" [id=60, type=get_attr]; +"61 encoder_layers_encoder_layer_7_self_attention_out_proj_bias" [id=61, type=get_attr]; +"62 encoder_layers_encoder_layer_7_ln_2_weight" [id=62, type=get_attr]; +"63 encoder_layers_encoder_layer_7_ln_2_bias" [id=63, type=get_attr]; +"64 encoder_layers_encoder_layer_7_mlp_0_bias" [id=64, type=get_attr]; +"65 encoder_layers_encoder_layer_7_mlp_3_bias" [id=65, type=get_attr]; +"66 encoder_layers_encoder_layer_8_ln_1_weight" [id=66, type=get_attr]; +"67 encoder_layers_encoder_layer_8_ln_1_bias" [id=67, type=get_attr]; +"68 encoder_layers_encoder_layer_8_self_attention_in_proj_bias" [id=68, type=get_attr]; +"69 encoder_layers_encoder_layer_8_self_attention_out_proj_bias" [id=69, type=get_attr]; +"70 encoder_layers_encoder_layer_8_ln_2_weight" [id=70, type=get_attr]; +"71 encoder_layers_encoder_layer_8_ln_2_bias" [id=71, type=get_attr]; +"72 encoder_layers_encoder_layer_8_mlp_0_bias" [id=72, type=get_attr]; +"73 encoder_layers_encoder_layer_8_mlp_3_bias" [id=73, type=get_attr]; +"74 encoder_layers_encoder_layer_9_ln_1_weight" [id=74, type=get_attr]; +"75 encoder_layers_encoder_layer_9_ln_1_bias" [id=75, type=get_attr]; +"76 encoder_layers_encoder_layer_9_self_attention_in_proj_bias" [id=76, type=get_attr]; +"77 encoder_layers_encoder_layer_9_self_attention_out_proj_bias" [id=77, type=get_attr]; +"78 encoder_layers_encoder_layer_9_ln_2_weight" [id=78, type=get_attr]; +"79 encoder_layers_encoder_layer_9_ln_2_bias" [id=79, type=get_attr]; +"80 encoder_layers_encoder_layer_9_mlp_0_bias" [id=80, type=get_attr]; +"81 encoder_layers_encoder_layer_9_mlp_3_bias" [id=81, type=get_attr]; +"82 encoder_layers_encoder_layer_10_ln_1_weight" [id=82, type=get_attr]; +"83 encoder_layers_encoder_layer_10_ln_1_bias" [id=83, type=get_attr]; +"84 encoder_layers_encoder_layer_10_self_attention_in_proj_bias" [id=84, type=get_attr]; +"85 encoder_layers_encoder_layer_10_self_attention_out_proj_bias" [id=85, type=get_attr]; +"86 encoder_layers_encoder_layer_10_ln_2_weight" [id=86, type=get_attr]; +"87 encoder_layers_encoder_layer_10_ln_2_bias" [id=87, type=get_attr]; +"88 encoder_layers_encoder_layer_10_mlp_0_bias" [id=88, type=get_attr]; +"89 encoder_layers_encoder_layer_10_mlp_3_bias" [id=89, type=get_attr]; +"90 encoder_layers_encoder_layer_11_ln_1_weight" [id=90, type=get_attr]; +"91 encoder_layers_encoder_layer_11_ln_1_bias" [id=91, type=get_attr]; +"92 encoder_layers_encoder_layer_11_self_attention_in_proj_bias" [id=92, type=get_attr]; +"93 encoder_layers_encoder_layer_11_self_attention_out_proj_bias" [id=93, type=get_attr]; +"94 encoder_layers_encoder_layer_11_ln_2_weight" [id=94, type=get_attr]; +"95 encoder_layers_encoder_layer_11_ln_2_bias" [id=95, type=get_attr]; +"96 encoder_layers_encoder_layer_11_mlp_0_bias" [id=96, type=get_attr]; +"97 encoder_layers_encoder_layer_11_mlp_3_bias" [id=97, type=get_attr]; +"98 encoder_ln_weight" [id=98, type=get_attr]; +"99 encoder_ln_bias" [id=99, type=get_attr]; +"100 heads_head_bias" [id=100, type=get_attr]; +"101 x" [id=101, type=input]; +"102 quantize_per_tensor_default" [id=102, type=quantize_per_tensor]; +"103 dequantize_per_tensor_default" [id=103, type=dequantize_per_tensor]; +"104 conv2d_scale_0" [id=104, type=get_attr]; +"105 conv2d_zero_point_0" [id=105, type=get_attr]; +"106 _frozen_param1" [id=106, type=get_attr]; +"107 dequantize_per_channel_default" [id=107, type=dequantize_per_channel]; +"108 conv2d" [id=108, type=conv2d]; +"109 reshape" [id=109, type=reshape]; +"110 permute" [id=110, type=permute]; +"111 _frozen_param0" [id=111, type=get_attr]; +"112 cat" [id=112, type=cat]; +"113 add" [id=113, type=add]; +"114 dropout" [id=114, type=dropout]; +"115 layer_norm" [id=115, type=layer_norm]; +"116 transpose" [id=116, type=transpose]; +"117 transpose_0_0_nncf_smooth_quant_0" [id=117, type=call_module]; +"118 quantize_per_tensor_default_1" [id=118, type=quantize_per_tensor]; +"119 dequantize_per_tensor_default_1" [id=119, type=dequantize_per_tensor]; +"120 linear_scale_0" [id=120, type=get_attr]; +"121 linear_zero_point_0" [id=121, type=get_attr]; +"122 _frozen_param2" [id=122, type=get_attr]; +"123 dequantize_per_channel_default_1" [id=123, type=dequantize_per_channel]; +"124 linear" [id=124, type=linear]; +"125 unflatten" [id=125, type=unflatten]; +"126 unsqueeze" [id=126, type=unsqueeze]; +"127 transpose_1" [id=127, type=transpose]; +"128 squeeze" [id=128, type=squeeze]; +"129 contiguous" [id=129, type=contiguous]; +"130 quantize_per_tensor_default_2" [id=130, type=quantize_per_tensor]; +"131 dequantize_per_tensor_default_2" [id=131, type=dequantize_per_tensor]; +"132 select" [id=132, type=select]; +"133 quantize_per_tensor_default_3" [id=133, type=quantize_per_tensor]; +"134 dequantize_per_tensor_default_3" [id=134, type=dequantize_per_tensor]; +"135 select_1" [id=135, type=select]; +"136 select_2" [id=136, type=select]; +"137 view" [id=137, type=view]; +"138 transpose_2" [id=138, type=transpose]; +"139 view_1" [id=139, type=view]; +"140 transpose_3" [id=140, type=transpose]; +"141 view_2" [id=141, type=view]; +"142 transpose_4" [id=142, type=transpose]; +"143 view_3" [id=143, type=view]; +"144 view_4" [id=144, type=view]; +"145 view_5" [id=145, type=view]; +"146 scaled_dot_product_attention" [id=146, type=scaled_dot_product_attention]; +"147 permute_1" [id=147, type=permute]; +"148 view_6" [id=148, type=view]; +"149 view_6_0_0_nncf_smooth_quant_0" [id=149, type=call_module]; +"150 quantize_per_tensor_default_4" [id=150, type=quantize_per_tensor]; +"151 dequantize_per_tensor_default_4" [id=151, type=dequantize_per_tensor]; +"152 linear_1_scale_0" [id=152, type=get_attr]; +"153 linear_1_zero_point_0" [id=153, type=get_attr]; +"154 _frozen_param3" [id=154, type=get_attr]; +"155 dequantize_per_channel_default_2" [id=155, type=dequantize_per_channel]; +"156 linear_1" [id=156, type=linear]; +"157 view_7" [id=157, type=view]; +"158 transpose_5" [id=158, type=transpose]; +"159 dropout_1" [id=159, type=dropout]; +"160 add_1" [id=160, type=add]; +"161 layer_norm_1" [id=161, type=layer_norm]; +"162 layer_norm_1_0_0_nncf_smooth_quant_0" [id=162, type=call_module]; +"163 quantize_per_tensor_default_5" [id=163, type=quantize_per_tensor]; +"164 dequantize_per_tensor_default_5" [id=164, type=dequantize_per_tensor]; +"165 linear_2_scale_0" [id=165, type=get_attr]; +"166 linear_2_zero_point_0" [id=166, type=get_attr]; +"167 _frozen_param4" [id=167, type=get_attr]; +"168 dequantize_per_channel_default_3" [id=168, type=dequantize_per_channel]; +"169 linear_2" [id=169, type=linear]; +"170 gelu" [id=170, type=gelu]; +"171 dropout_2" [id=171, type=dropout]; +"172 dropout_2_0_0_nncf_smooth_quant_0" [id=172, type=call_module]; +"173 quantize_per_tensor_default_6" [id=173, type=quantize_per_tensor]; +"174 dequantize_per_tensor_default_6" [id=174, type=dequantize_per_tensor]; +"175 linear_3_scale_0" [id=175, type=get_attr]; +"176 linear_3_zero_point_0" [id=176, type=get_attr]; +"177 _frozen_param5" [id=177, type=get_attr]; +"178 dequantize_per_channel_default_4" [id=178, type=dequantize_per_channel]; +"179 linear_3" [id=179, type=linear]; +"180 dropout_3" [id=180, type=dropout]; +"181 add_2" [id=181, type=add]; +"182 layer_norm_2" [id=182, type=layer_norm]; +"183 transpose_6" [id=183, type=transpose]; +"184 transpose_6_0_0_nncf_smooth_quant_0" [id=184, type=call_module]; +"185 quantize_per_tensor_default_7" [id=185, type=quantize_per_tensor]; +"186 dequantize_per_tensor_default_7" [id=186, type=dequantize_per_tensor]; +"187 linear_4_scale_0" [id=187, type=get_attr]; +"188 linear_4_zero_point_0" [id=188, type=get_attr]; +"189 _frozen_param6" [id=189, type=get_attr]; +"190 dequantize_per_channel_default_5" [id=190, type=dequantize_per_channel]; +"191 linear_4" [id=191, type=linear]; +"192 unflatten_1" [id=192, type=unflatten]; +"193 unsqueeze_1" [id=193, type=unsqueeze]; +"194 transpose_7" [id=194, type=transpose]; +"195 squeeze_1" [id=195, type=squeeze]; +"196 contiguous_1" [id=196, type=contiguous]; +"197 quantize_per_tensor_default_8" [id=197, type=quantize_per_tensor]; +"198 dequantize_per_tensor_default_8" [id=198, type=dequantize_per_tensor]; +"199 select_3" [id=199, type=select]; +"200 quantize_per_tensor_default_9" [id=200, type=quantize_per_tensor]; +"201 dequantize_per_tensor_default_9" [id=201, type=dequantize_per_tensor]; +"202 select_4" [id=202, type=select]; +"203 select_5" [id=203, type=select]; +"204 view_8" [id=204, type=view]; +"205 transpose_8" [id=205, type=transpose]; +"206 view_9" [id=206, type=view]; +"207 transpose_9" [id=207, type=transpose]; +"208 view_10" [id=208, type=view]; +"209 transpose_10" [id=209, type=transpose]; +"210 view_11" [id=210, type=view]; +"211 view_12" [id=211, type=view]; +"212 view_13" [id=212, type=view]; +"213 scaled_dot_product_attention_1" [id=213, type=scaled_dot_product_attention]; +"214 permute_2" [id=214, type=permute]; +"215 view_14" [id=215, type=view]; +"216 view_14_0_0_nncf_smooth_quant_0" [id=216, type=call_module]; +"217 quantize_per_tensor_default_10" [id=217, type=quantize_per_tensor]; +"218 dequantize_per_tensor_default_10" [id=218, type=dequantize_per_tensor]; +"219 linear_5_scale_0" [id=219, type=get_attr]; +"220 linear_5_zero_point_0" [id=220, type=get_attr]; +"221 _frozen_param7" [id=221, type=get_attr]; +"222 dequantize_per_channel_default_6" [id=222, type=dequantize_per_channel]; +"223 linear_5" [id=223, type=linear]; +"224 view_15" [id=224, type=view]; +"225 transpose_11" [id=225, type=transpose]; +"226 dropout_4" [id=226, type=dropout]; +"227 add_3" [id=227, type=add]; +"228 layer_norm_3" [id=228, type=layer_norm]; +"229 layer_norm_3_0_0_nncf_smooth_quant_0" [id=229, type=call_module]; +"230 quantize_per_tensor_default_11" [id=230, type=quantize_per_tensor]; +"231 dequantize_per_tensor_default_11" [id=231, type=dequantize_per_tensor]; +"232 linear_6_scale_0" [id=232, type=get_attr]; +"233 linear_6_zero_point_0" [id=233, type=get_attr]; +"234 _frozen_param8" [id=234, type=get_attr]; +"235 dequantize_per_channel_default_7" [id=235, type=dequantize_per_channel]; +"236 linear_6" [id=236, type=linear]; +"237 gelu_1" [id=237, type=gelu]; +"238 dropout_5" [id=238, type=dropout]; +"239 dropout_5_0_0_nncf_smooth_quant_0" [id=239, type=call_module]; +"240 quantize_per_tensor_default_12" [id=240, type=quantize_per_tensor]; +"241 dequantize_per_tensor_default_12" [id=241, type=dequantize_per_tensor]; +"242 linear_7_scale_0" [id=242, type=get_attr]; +"243 linear_7_zero_point_0" [id=243, type=get_attr]; +"244 _frozen_param9" [id=244, type=get_attr]; +"245 dequantize_per_channel_default_8" [id=245, type=dequantize_per_channel]; +"246 linear_7" [id=246, type=linear]; +"247 dropout_6" [id=247, type=dropout]; +"248 add_4" [id=248, type=add]; +"249 layer_norm_4" [id=249, type=layer_norm]; +"250 transpose_12" [id=250, type=transpose]; +"251 transpose_12_0_0_nncf_smooth_quant_0" [id=251, type=call_module]; +"252 quantize_per_tensor_default_13" [id=252, type=quantize_per_tensor]; +"253 dequantize_per_tensor_default_13" [id=253, type=dequantize_per_tensor]; +"254 linear_8_scale_0" [id=254, type=get_attr]; +"255 linear_8_zero_point_0" [id=255, type=get_attr]; +"256 _frozen_param10" [id=256, type=get_attr]; +"257 dequantize_per_channel_default_9" [id=257, type=dequantize_per_channel]; +"258 linear_8" [id=258, type=linear]; +"259 unflatten_2" [id=259, type=unflatten]; +"260 unsqueeze_2" [id=260, type=unsqueeze]; +"261 transpose_13" [id=261, type=transpose]; +"262 squeeze_2" [id=262, type=squeeze]; +"263 contiguous_2" [id=263, type=contiguous]; +"264 quantize_per_tensor_default_14" [id=264, type=quantize_per_tensor]; +"265 dequantize_per_tensor_default_14" [id=265, type=dequantize_per_tensor]; +"266 select_6" [id=266, type=select]; +"267 quantize_per_tensor_default_15" [id=267, type=quantize_per_tensor]; +"268 dequantize_per_tensor_default_15" [id=268, type=dequantize_per_tensor]; +"269 select_7" [id=269, type=select]; +"270 select_8" [id=270, type=select]; +"271 view_16" [id=271, type=view]; +"272 transpose_14" [id=272, type=transpose]; +"273 view_17" [id=273, type=view]; +"274 transpose_15" [id=274, type=transpose]; +"275 view_18" [id=275, type=view]; +"276 transpose_16" [id=276, type=transpose]; +"277 view_19" [id=277, type=view]; +"278 view_20" [id=278, type=view]; +"279 view_21" [id=279, type=view]; +"280 scaled_dot_product_attention_2" [id=280, type=scaled_dot_product_attention]; +"281 permute_3" [id=281, type=permute]; +"282 view_22" [id=282, type=view]; +"283 view_22_0_0_nncf_smooth_quant_0" [id=283, type=call_module]; +"284 quantize_per_tensor_default_16" [id=284, type=quantize_per_tensor]; +"285 dequantize_per_tensor_default_16" [id=285, type=dequantize_per_tensor]; +"286 linear_9_scale_0" [id=286, type=get_attr]; +"287 linear_9_zero_point_0" [id=287, type=get_attr]; +"288 _frozen_param11" [id=288, type=get_attr]; +"289 dequantize_per_channel_default_10" [id=289, type=dequantize_per_channel]; +"290 linear_9" [id=290, type=linear]; +"291 view_23" [id=291, type=view]; +"292 transpose_17" [id=292, type=transpose]; +"293 dropout_7" [id=293, type=dropout]; +"294 add_5" [id=294, type=add]; +"295 layer_norm_5" [id=295, type=layer_norm]; +"296 layer_norm_5_0_0_nncf_smooth_quant_0" [id=296, type=call_module]; +"297 quantize_per_tensor_default_17" [id=297, type=quantize_per_tensor]; +"298 dequantize_per_tensor_default_17" [id=298, type=dequantize_per_tensor]; +"299 linear_10_scale_0" [id=299, type=get_attr]; +"300 linear_10_zero_point_0" [id=300, type=get_attr]; +"301 _frozen_param12" [id=301, type=get_attr]; +"302 dequantize_per_channel_default_11" [id=302, type=dequantize_per_channel]; +"303 linear_10" [id=303, type=linear]; +"304 gelu_2" [id=304, type=gelu]; +"305 dropout_8" [id=305, type=dropout]; +"306 dropout_8_0_0_nncf_smooth_quant_0" [id=306, type=call_module]; +"307 quantize_per_tensor_default_18" [id=307, type=quantize_per_tensor]; +"308 dequantize_per_tensor_default_18" [id=308, type=dequantize_per_tensor]; +"309 linear_11_scale_0" [id=309, type=get_attr]; +"310 linear_11_zero_point_0" [id=310, type=get_attr]; +"311 _frozen_param13" [id=311, type=get_attr]; +"312 dequantize_per_channel_default_12" [id=312, type=dequantize_per_channel]; +"313 linear_11" [id=313, type=linear]; +"314 dropout_9" [id=314, type=dropout]; +"315 add_6" [id=315, type=add]; +"316 layer_norm_6" [id=316, type=layer_norm]; +"317 transpose_18" [id=317, type=transpose]; +"318 transpose_18_0_0_nncf_smooth_quant_0" [id=318, type=call_module]; +"319 quantize_per_tensor_default_19" [id=319, type=quantize_per_tensor]; +"320 dequantize_per_tensor_default_19" [id=320, type=dequantize_per_tensor]; +"321 linear_12_scale_0" [id=321, type=get_attr]; +"322 linear_12_zero_point_0" [id=322, type=get_attr]; +"323 _frozen_param14" [id=323, type=get_attr]; +"324 dequantize_per_channel_default_13" [id=324, type=dequantize_per_channel]; +"325 linear_12" [id=325, type=linear]; +"326 unflatten_3" [id=326, type=unflatten]; +"327 unsqueeze_3" [id=327, type=unsqueeze]; +"328 transpose_19" [id=328, type=transpose]; +"329 squeeze_3" [id=329, type=squeeze]; +"330 contiguous_3" [id=330, type=contiguous]; +"331 quantize_per_tensor_default_20" [id=331, type=quantize_per_tensor]; +"332 dequantize_per_tensor_default_20" [id=332, type=dequantize_per_tensor]; +"333 select_9" [id=333, type=select]; +"334 quantize_per_tensor_default_21" [id=334, type=quantize_per_tensor]; +"335 dequantize_per_tensor_default_21" [id=335, type=dequantize_per_tensor]; +"336 select_10" [id=336, type=select]; +"337 select_11" [id=337, type=select]; +"338 view_24" [id=338, type=view]; +"339 transpose_20" [id=339, type=transpose]; +"340 view_25" [id=340, type=view]; +"341 transpose_21" [id=341, type=transpose]; +"342 view_26" [id=342, type=view]; +"343 transpose_22" [id=343, type=transpose]; +"344 view_27" [id=344, type=view]; +"345 view_28" [id=345, type=view]; +"346 view_29" [id=346, type=view]; +"347 scaled_dot_product_attention_3" [id=347, type=scaled_dot_product_attention]; +"348 permute_4" [id=348, type=permute]; +"349 view_30" [id=349, type=view]; +"350 view_30_0_0_nncf_smooth_quant_0" [id=350, type=call_module]; +"351 quantize_per_tensor_default_22" [id=351, type=quantize_per_tensor]; +"352 dequantize_per_tensor_default_22" [id=352, type=dequantize_per_tensor]; +"353 linear_13_scale_0" [id=353, type=get_attr]; +"354 linear_13_zero_point_0" [id=354, type=get_attr]; +"355 _frozen_param15" [id=355, type=get_attr]; +"356 dequantize_per_channel_default_14" [id=356, type=dequantize_per_channel]; +"357 linear_13" [id=357, type=linear]; +"358 view_31" [id=358, type=view]; +"359 transpose_23" [id=359, type=transpose]; +"360 dropout_10" [id=360, type=dropout]; +"361 add_7" [id=361, type=add]; +"362 layer_norm_7" [id=362, type=layer_norm]; +"363 layer_norm_7_0_0_nncf_smooth_quant_0" [id=363, type=call_module]; +"364 quantize_per_tensor_default_23" [id=364, type=quantize_per_tensor]; +"365 dequantize_per_tensor_default_23" [id=365, type=dequantize_per_tensor]; +"366 linear_14_scale_0" [id=366, type=get_attr]; +"367 linear_14_zero_point_0" [id=367, type=get_attr]; +"368 _frozen_param16" [id=368, type=get_attr]; +"369 dequantize_per_channel_default_15" [id=369, type=dequantize_per_channel]; +"370 linear_14" [id=370, type=linear]; +"371 gelu_3" [id=371, type=gelu]; +"372 dropout_11" [id=372, type=dropout]; +"373 dropout_11_0_0_nncf_smooth_quant_0" [id=373, type=call_module]; +"374 quantize_per_tensor_default_24" [id=374, type=quantize_per_tensor]; +"375 dequantize_per_tensor_default_24" [id=375, type=dequantize_per_tensor]; +"376 linear_15_scale_0" [id=376, type=get_attr]; +"377 linear_15_zero_point_0" [id=377, type=get_attr]; +"378 _frozen_param17" [id=378, type=get_attr]; +"379 dequantize_per_channel_default_16" [id=379, type=dequantize_per_channel]; +"380 linear_15" [id=380, type=linear]; +"381 dropout_12" [id=381, type=dropout]; +"382 add_8" [id=382, type=add]; +"383 layer_norm_8" [id=383, type=layer_norm]; +"384 transpose_24" [id=384, type=transpose]; +"385 transpose_24_0_0_nncf_smooth_quant_0" [id=385, type=call_module]; +"386 quantize_per_tensor_default_25" [id=386, type=quantize_per_tensor]; +"387 dequantize_per_tensor_default_25" [id=387, type=dequantize_per_tensor]; +"388 linear_16_scale_0" [id=388, type=get_attr]; +"389 linear_16_zero_point_0" [id=389, type=get_attr]; +"390 _frozen_param18" [id=390, type=get_attr]; +"391 dequantize_per_channel_default_17" [id=391, type=dequantize_per_channel]; +"392 linear_16" [id=392, type=linear]; +"393 unflatten_4" [id=393, type=unflatten]; +"394 unsqueeze_4" [id=394, type=unsqueeze]; +"395 transpose_25" [id=395, type=transpose]; +"396 squeeze_4" [id=396, type=squeeze]; +"397 contiguous_4" [id=397, type=contiguous]; +"398 quantize_per_tensor_default_26" [id=398, type=quantize_per_tensor]; +"399 dequantize_per_tensor_default_26" [id=399, type=dequantize_per_tensor]; +"400 select_12" [id=400, type=select]; +"401 quantize_per_tensor_default_27" [id=401, type=quantize_per_tensor]; +"402 dequantize_per_tensor_default_27" [id=402, type=dequantize_per_tensor]; +"403 select_13" [id=403, type=select]; +"404 select_14" [id=404, type=select]; +"405 view_32" [id=405, type=view]; +"406 transpose_26" [id=406, type=transpose]; +"407 view_33" [id=407, type=view]; +"408 transpose_27" [id=408, type=transpose]; +"409 view_34" [id=409, type=view]; +"410 transpose_28" [id=410, type=transpose]; +"411 view_35" [id=411, type=view]; +"412 view_36" [id=412, type=view]; +"413 view_37" [id=413, type=view]; +"414 scaled_dot_product_attention_4" [id=414, type=scaled_dot_product_attention]; +"415 permute_5" [id=415, type=permute]; +"416 view_38" [id=416, type=view]; +"417 view_38_0_0_nncf_smooth_quant_0" [id=417, type=call_module]; +"418 quantize_per_tensor_default_28" [id=418, type=quantize_per_tensor]; +"419 dequantize_per_tensor_default_28" [id=419, type=dequantize_per_tensor]; +"420 linear_17_scale_0" [id=420, type=get_attr]; +"421 linear_17_zero_point_0" [id=421, type=get_attr]; +"422 _frozen_param19" [id=422, type=get_attr]; +"423 dequantize_per_channel_default_18" [id=423, type=dequantize_per_channel]; +"424 linear_17" [id=424, type=linear]; +"425 view_39" [id=425, type=view]; +"426 transpose_29" [id=426, type=transpose]; +"427 dropout_13" [id=427, type=dropout]; +"428 add_9" [id=428, type=add]; +"429 layer_norm_9" [id=429, type=layer_norm]; +"430 layer_norm_9_0_0_nncf_smooth_quant_0" [id=430, type=call_module]; +"431 quantize_per_tensor_default_29" [id=431, type=quantize_per_tensor]; +"432 dequantize_per_tensor_default_29" [id=432, type=dequantize_per_tensor]; +"433 linear_18_scale_0" [id=433, type=get_attr]; +"434 linear_18_zero_point_0" [id=434, type=get_attr]; +"435 _frozen_param20" [id=435, type=get_attr]; +"436 dequantize_per_channel_default_19" [id=436, type=dequantize_per_channel]; +"437 linear_18" [id=437, type=linear]; +"438 gelu_4" [id=438, type=gelu]; +"439 dropout_14" [id=439, type=dropout]; +"440 dropout_14_0_0_nncf_smooth_quant_0" [id=440, type=call_module]; +"441 quantize_per_tensor_default_30" [id=441, type=quantize_per_tensor]; +"442 dequantize_per_tensor_default_30" [id=442, type=dequantize_per_tensor]; +"443 linear_19_scale_0" [id=443, type=get_attr]; +"444 linear_19_zero_point_0" [id=444, type=get_attr]; +"445 _frozen_param21" [id=445, type=get_attr]; +"446 dequantize_per_channel_default_20" [id=446, type=dequantize_per_channel]; +"447 linear_19" [id=447, type=linear]; +"448 dropout_15" [id=448, type=dropout]; +"449 add_10" [id=449, type=add]; +"450 layer_norm_10" [id=450, type=layer_norm]; +"451 transpose_30" [id=451, type=transpose]; +"452 transpose_30_0_0_nncf_smooth_quant_0" [id=452, type=call_module]; +"453 quantize_per_tensor_default_31" [id=453, type=quantize_per_tensor]; +"454 dequantize_per_tensor_default_31" [id=454, type=dequantize_per_tensor]; +"455 linear_20_scale_0" [id=455, type=get_attr]; +"456 linear_20_zero_point_0" [id=456, type=get_attr]; +"457 _frozen_param22" [id=457, type=get_attr]; +"458 dequantize_per_channel_default_21" [id=458, type=dequantize_per_channel]; +"459 linear_20" [id=459, type=linear]; +"460 unflatten_5" [id=460, type=unflatten]; +"461 unsqueeze_5" [id=461, type=unsqueeze]; +"462 transpose_31" [id=462, type=transpose]; +"463 squeeze_5" [id=463, type=squeeze]; +"464 contiguous_5" [id=464, type=contiguous]; +"465 quantize_per_tensor_default_32" [id=465, type=quantize_per_tensor]; +"466 dequantize_per_tensor_default_32" [id=466, type=dequantize_per_tensor]; +"467 select_15" [id=467, type=select]; +"468 quantize_per_tensor_default_33" [id=468, type=quantize_per_tensor]; +"469 dequantize_per_tensor_default_33" [id=469, type=dequantize_per_tensor]; +"470 select_16" [id=470, type=select]; +"471 select_17" [id=471, type=select]; +"472 view_40" [id=472, type=view]; +"473 transpose_32" [id=473, type=transpose]; +"474 view_41" [id=474, type=view]; +"475 transpose_33" [id=475, type=transpose]; +"476 view_42" [id=476, type=view]; +"477 transpose_34" [id=477, type=transpose]; +"478 view_43" [id=478, type=view]; +"479 view_44" [id=479, type=view]; +"480 view_45" [id=480, type=view]; +"481 scaled_dot_product_attention_5" [id=481, type=scaled_dot_product_attention]; +"482 permute_6" [id=482, type=permute]; +"483 view_46" [id=483, type=view]; +"484 view_46_0_0_nncf_smooth_quant_0" [id=484, type=call_module]; +"485 quantize_per_tensor_default_34" [id=485, type=quantize_per_tensor]; +"486 dequantize_per_tensor_default_34" [id=486, type=dequantize_per_tensor]; +"487 linear_21_scale_0" [id=487, type=get_attr]; +"488 linear_21_zero_point_0" [id=488, type=get_attr]; +"489 _frozen_param23" [id=489, type=get_attr]; +"490 dequantize_per_channel_default_22" [id=490, type=dequantize_per_channel]; +"491 linear_21" [id=491, type=linear]; +"492 view_47" [id=492, type=view]; +"493 transpose_35" [id=493, type=transpose]; +"494 dropout_16" [id=494, type=dropout]; +"495 add_11" [id=495, type=add]; +"496 layer_norm_11" [id=496, type=layer_norm]; +"497 layer_norm_11_0_0_nncf_smooth_quant_0" [id=497, type=call_module]; +"498 quantize_per_tensor_default_35" [id=498, type=quantize_per_tensor]; +"499 dequantize_per_tensor_default_35" [id=499, type=dequantize_per_tensor]; +"500 linear_22_scale_0" [id=500, type=get_attr]; +"501 linear_22_zero_point_0" [id=501, type=get_attr]; +"502 _frozen_param24" [id=502, type=get_attr]; +"503 dequantize_per_channel_default_23" [id=503, type=dequantize_per_channel]; +"504 linear_22" [id=504, type=linear]; +"505 gelu_5" [id=505, type=gelu]; +"506 dropout_17" [id=506, type=dropout]; +"507 dropout_17_0_0_nncf_smooth_quant_0" [id=507, type=call_module]; +"508 quantize_per_tensor_default_36" [id=508, type=quantize_per_tensor]; +"509 dequantize_per_tensor_default_36" [id=509, type=dequantize_per_tensor]; +"510 linear_23_scale_0" [id=510, type=get_attr]; +"511 linear_23_zero_point_0" [id=511, type=get_attr]; +"512 _frozen_param25" [id=512, type=get_attr]; +"513 dequantize_per_channel_default_24" [id=513, type=dequantize_per_channel]; +"514 linear_23" [id=514, type=linear]; +"515 dropout_18" [id=515, type=dropout]; +"516 add_12" [id=516, type=add]; +"517 layer_norm_12" [id=517, type=layer_norm]; +"518 transpose_36" [id=518, type=transpose]; +"519 transpose_36_0_0_nncf_smooth_quant_0" [id=519, type=call_module]; +"520 quantize_per_tensor_default_37" [id=520, type=quantize_per_tensor]; +"521 dequantize_per_tensor_default_37" [id=521, type=dequantize_per_tensor]; +"522 linear_24_scale_0" [id=522, type=get_attr]; +"523 linear_24_zero_point_0" [id=523, type=get_attr]; +"524 _frozen_param26" [id=524, type=get_attr]; +"525 dequantize_per_channel_default_25" [id=525, type=dequantize_per_channel]; +"526 linear_24" [id=526, type=linear]; +"527 unflatten_6" [id=527, type=unflatten]; +"528 unsqueeze_6" [id=528, type=unsqueeze]; +"529 transpose_37" [id=529, type=transpose]; +"530 squeeze_6" [id=530, type=squeeze]; +"531 contiguous_6" [id=531, type=contiguous]; +"532 quantize_per_tensor_default_38" [id=532, type=quantize_per_tensor]; +"533 dequantize_per_tensor_default_38" [id=533, type=dequantize_per_tensor]; +"534 select_18" [id=534, type=select]; +"535 quantize_per_tensor_default_39" [id=535, type=quantize_per_tensor]; +"536 dequantize_per_tensor_default_39" [id=536, type=dequantize_per_tensor]; +"537 select_19" [id=537, type=select]; +"538 select_20" [id=538, type=select]; +"539 view_48" [id=539, type=view]; +"540 transpose_38" [id=540, type=transpose]; +"541 view_49" [id=541, type=view]; +"542 transpose_39" [id=542, type=transpose]; +"543 view_50" [id=543, type=view]; +"544 transpose_40" [id=544, type=transpose]; +"545 view_51" [id=545, type=view]; +"546 view_52" [id=546, type=view]; +"547 view_53" [id=547, type=view]; +"548 scaled_dot_product_attention_6" [id=548, type=scaled_dot_product_attention]; +"549 permute_7" [id=549, type=permute]; +"550 view_54" [id=550, type=view]; +"551 view_54_0_0_nncf_smooth_quant_0" [id=551, type=call_module]; +"552 quantize_per_tensor_default_40" [id=552, type=quantize_per_tensor]; +"553 dequantize_per_tensor_default_40" [id=553, type=dequantize_per_tensor]; +"554 linear_25_scale_0" [id=554, type=get_attr]; +"555 linear_25_zero_point_0" [id=555, type=get_attr]; +"556 _frozen_param27" [id=556, type=get_attr]; +"557 dequantize_per_channel_default_26" [id=557, type=dequantize_per_channel]; +"558 linear_25" [id=558, type=linear]; +"559 view_55" [id=559, type=view]; +"560 transpose_41" [id=560, type=transpose]; +"561 dropout_19" [id=561, type=dropout]; +"562 add_13" [id=562, type=add]; +"563 layer_norm_13" [id=563, type=layer_norm]; +"564 layer_norm_13_0_0_nncf_smooth_quant_0" [id=564, type=call_module]; +"565 quantize_per_tensor_default_41" [id=565, type=quantize_per_tensor]; +"566 dequantize_per_tensor_default_41" [id=566, type=dequantize_per_tensor]; +"567 linear_26_scale_0" [id=567, type=get_attr]; +"568 linear_26_zero_point_0" [id=568, type=get_attr]; +"569 _frozen_param28" [id=569, type=get_attr]; +"570 dequantize_per_channel_default_27" [id=570, type=dequantize_per_channel]; +"571 linear_26" [id=571, type=linear]; +"572 gelu_6" [id=572, type=gelu]; +"573 dropout_20" [id=573, type=dropout]; +"574 dropout_20_0_0_nncf_smooth_quant_0" [id=574, type=call_module]; +"575 quantize_per_tensor_default_42" [id=575, type=quantize_per_tensor]; +"576 dequantize_per_tensor_default_42" [id=576, type=dequantize_per_tensor]; +"577 linear_27_scale_0" [id=577, type=get_attr]; +"578 linear_27_zero_point_0" [id=578, type=get_attr]; +"579 _frozen_param29" [id=579, type=get_attr]; +"580 dequantize_per_channel_default_28" [id=580, type=dequantize_per_channel]; +"581 linear_27" [id=581, type=linear]; +"582 dropout_21" [id=582, type=dropout]; +"583 add_14" [id=583, type=add]; +"584 layer_norm_14" [id=584, type=layer_norm]; +"585 transpose_42" [id=585, type=transpose]; +"586 transpose_42_0_0_nncf_smooth_quant_0" [id=586, type=call_module]; +"587 quantize_per_tensor_default_43" [id=587, type=quantize_per_tensor]; +"588 dequantize_per_tensor_default_43" [id=588, type=dequantize_per_tensor]; +"589 linear_28_scale_0" [id=589, type=get_attr]; +"590 linear_28_zero_point_0" [id=590, type=get_attr]; +"591 _frozen_param30" [id=591, type=get_attr]; +"592 dequantize_per_channel_default_29" [id=592, type=dequantize_per_channel]; +"593 linear_28" [id=593, type=linear]; +"594 unflatten_7" [id=594, type=unflatten]; +"595 unsqueeze_7" [id=595, type=unsqueeze]; +"596 transpose_43" [id=596, type=transpose]; +"597 squeeze_7" [id=597, type=squeeze]; +"598 contiguous_7" [id=598, type=contiguous]; +"599 quantize_per_tensor_default_44" [id=599, type=quantize_per_tensor]; +"600 dequantize_per_tensor_default_44" [id=600, type=dequantize_per_tensor]; +"601 select_21" [id=601, type=select]; +"602 quantize_per_tensor_default_45" [id=602, type=quantize_per_tensor]; +"603 dequantize_per_tensor_default_45" [id=603, type=dequantize_per_tensor]; +"604 select_22" [id=604, type=select]; +"605 select_23" [id=605, type=select]; +"606 view_56" [id=606, type=view]; +"607 transpose_44" [id=607, type=transpose]; +"608 view_57" [id=608, type=view]; +"609 transpose_45" [id=609, type=transpose]; +"610 view_58" [id=610, type=view]; +"611 transpose_46" [id=611, type=transpose]; +"612 view_59" [id=612, type=view]; +"613 view_60" [id=613, type=view]; +"614 view_61" [id=614, type=view]; +"615 scaled_dot_product_attention_7" [id=615, type=scaled_dot_product_attention]; +"616 permute_8" [id=616, type=permute]; +"617 view_62" [id=617, type=view]; +"618 view_62_0_0_nncf_smooth_quant_0" [id=618, type=call_module]; +"619 quantize_per_tensor_default_46" [id=619, type=quantize_per_tensor]; +"620 dequantize_per_tensor_default_46" [id=620, type=dequantize_per_tensor]; +"621 linear_29_scale_0" [id=621, type=get_attr]; +"622 linear_29_zero_point_0" [id=622, type=get_attr]; +"623 _frozen_param31" [id=623, type=get_attr]; +"624 dequantize_per_channel_default_30" [id=624, type=dequantize_per_channel]; +"625 linear_29" [id=625, type=linear]; +"626 view_63" [id=626, type=view]; +"627 transpose_47" [id=627, type=transpose]; +"628 dropout_22" [id=628, type=dropout]; +"629 add_15" [id=629, type=add]; +"630 layer_norm_15" [id=630, type=layer_norm]; +"631 layer_norm_15_0_0_nncf_smooth_quant_0" [id=631, type=call_module]; +"632 quantize_per_tensor_default_47" [id=632, type=quantize_per_tensor]; +"633 dequantize_per_tensor_default_47" [id=633, type=dequantize_per_tensor]; +"634 linear_30_scale_0" [id=634, type=get_attr]; +"635 linear_30_zero_point_0" [id=635, type=get_attr]; +"636 _frozen_param32" [id=636, type=get_attr]; +"637 dequantize_per_channel_default_31" [id=637, type=dequantize_per_channel]; +"638 linear_30" [id=638, type=linear]; +"639 gelu_7" [id=639, type=gelu]; +"640 dropout_23" [id=640, type=dropout]; +"641 dropout_23_0_0_nncf_smooth_quant_0" [id=641, type=call_module]; +"642 quantize_per_tensor_default_48" [id=642, type=quantize_per_tensor]; +"643 dequantize_per_tensor_default_48" [id=643, type=dequantize_per_tensor]; +"644 linear_31_scale_0" [id=644, type=get_attr]; +"645 linear_31_zero_point_0" [id=645, type=get_attr]; +"646 _frozen_param33" [id=646, type=get_attr]; +"647 dequantize_per_channel_default_32" [id=647, type=dequantize_per_channel]; +"648 linear_31" [id=648, type=linear]; +"649 dropout_24" [id=649, type=dropout]; +"650 add_16" [id=650, type=add]; +"651 layer_norm_16" [id=651, type=layer_norm]; +"652 transpose_48" [id=652, type=transpose]; +"653 transpose_48_0_0_nncf_smooth_quant_0" [id=653, type=call_module]; +"654 quantize_per_tensor_default_49" [id=654, type=quantize_per_tensor]; +"655 dequantize_per_tensor_default_49" [id=655, type=dequantize_per_tensor]; +"656 linear_32_scale_0" [id=656, type=get_attr]; +"657 linear_32_zero_point_0" [id=657, type=get_attr]; +"658 _frozen_param34" [id=658, type=get_attr]; +"659 dequantize_per_channel_default_33" [id=659, type=dequantize_per_channel]; +"660 linear_32" [id=660, type=linear]; +"661 unflatten_8" [id=661, type=unflatten]; +"662 unsqueeze_8" [id=662, type=unsqueeze]; +"663 transpose_49" [id=663, type=transpose]; +"664 squeeze_8" [id=664, type=squeeze]; +"665 contiguous_8" [id=665, type=contiguous]; +"666 quantize_per_tensor_default_50" [id=666, type=quantize_per_tensor]; +"667 dequantize_per_tensor_default_50" [id=667, type=dequantize_per_tensor]; +"668 select_24" [id=668, type=select]; +"669 quantize_per_tensor_default_51" [id=669, type=quantize_per_tensor]; +"670 dequantize_per_tensor_default_51" [id=670, type=dequantize_per_tensor]; +"671 select_25" [id=671, type=select]; +"672 select_26" [id=672, type=select]; +"673 view_64" [id=673, type=view]; +"674 transpose_50" [id=674, type=transpose]; +"675 view_65" [id=675, type=view]; +"676 transpose_51" [id=676, type=transpose]; +"677 view_66" [id=677, type=view]; +"678 transpose_52" [id=678, type=transpose]; +"679 view_67" [id=679, type=view]; +"680 view_68" [id=680, type=view]; +"681 view_69" [id=681, type=view]; +"682 scaled_dot_product_attention_8" [id=682, type=scaled_dot_product_attention]; +"683 permute_9" [id=683, type=permute]; +"684 view_70" [id=684, type=view]; +"685 view_70_0_0_nncf_smooth_quant_0" [id=685, type=call_module]; +"686 quantize_per_tensor_default_52" [id=686, type=quantize_per_tensor]; +"687 dequantize_per_tensor_default_52" [id=687, type=dequantize_per_tensor]; +"688 linear_33_scale_0" [id=688, type=get_attr]; +"689 linear_33_zero_point_0" [id=689, type=get_attr]; +"690 _frozen_param35" [id=690, type=get_attr]; +"691 dequantize_per_channel_default_34" [id=691, type=dequantize_per_channel]; +"692 linear_33" [id=692, type=linear]; +"693 view_71" [id=693, type=view]; +"694 transpose_53" [id=694, type=transpose]; +"695 dropout_25" [id=695, type=dropout]; +"696 add_17" [id=696, type=add]; +"697 layer_norm_17" [id=697, type=layer_norm]; +"698 layer_norm_17_0_0_nncf_smooth_quant_0" [id=698, type=call_module]; +"699 quantize_per_tensor_default_53" [id=699, type=quantize_per_tensor]; +"700 dequantize_per_tensor_default_53" [id=700, type=dequantize_per_tensor]; +"701 linear_34_scale_0" [id=701, type=get_attr]; +"702 linear_34_zero_point_0" [id=702, type=get_attr]; +"703 _frozen_param36" [id=703, type=get_attr]; +"704 dequantize_per_channel_default_35" [id=704, type=dequantize_per_channel]; +"705 linear_34" [id=705, type=linear]; +"706 gelu_8" [id=706, type=gelu]; +"707 dropout_26" [id=707, type=dropout]; +"708 dropout_26_0_0_nncf_smooth_quant_0" [id=708, type=call_module]; +"709 quantize_per_tensor_default_54" [id=709, type=quantize_per_tensor]; +"710 dequantize_per_tensor_default_54" [id=710, type=dequantize_per_tensor]; +"711 linear_35_scale_0" [id=711, type=get_attr]; +"712 linear_35_zero_point_0" [id=712, type=get_attr]; +"713 _frozen_param37" [id=713, type=get_attr]; +"714 dequantize_per_channel_default_36" [id=714, type=dequantize_per_channel]; +"715 linear_35" [id=715, type=linear]; +"716 dropout_27" [id=716, type=dropout]; +"717 add_18" [id=717, type=add]; +"718 layer_norm_18" [id=718, type=layer_norm]; +"719 transpose_54" [id=719, type=transpose]; +"720 transpose_54_0_0_nncf_smooth_quant_0" [id=720, type=call_module]; +"721 quantize_per_tensor_default_55" [id=721, type=quantize_per_tensor]; +"722 dequantize_per_tensor_default_55" [id=722, type=dequantize_per_tensor]; +"723 linear_36_scale_0" [id=723, type=get_attr]; +"724 linear_36_zero_point_0" [id=724, type=get_attr]; +"725 _frozen_param38" [id=725, type=get_attr]; +"726 dequantize_per_channel_default_37" [id=726, type=dequantize_per_channel]; +"727 linear_36" [id=727, type=linear]; +"728 unflatten_9" [id=728, type=unflatten]; +"729 unsqueeze_9" [id=729, type=unsqueeze]; +"730 transpose_55" [id=730, type=transpose]; +"731 squeeze_9" [id=731, type=squeeze]; +"732 contiguous_9" [id=732, type=contiguous]; +"733 quantize_per_tensor_default_56" [id=733, type=quantize_per_tensor]; +"734 dequantize_per_tensor_default_56" [id=734, type=dequantize_per_tensor]; +"735 select_27" [id=735, type=select]; +"736 quantize_per_tensor_default_57" [id=736, type=quantize_per_tensor]; +"737 dequantize_per_tensor_default_57" [id=737, type=dequantize_per_tensor]; +"738 select_28" [id=738, type=select]; +"739 select_29" [id=739, type=select]; +"740 view_72" [id=740, type=view]; +"741 transpose_56" [id=741, type=transpose]; +"742 view_73" [id=742, type=view]; +"743 transpose_57" [id=743, type=transpose]; +"744 view_74" [id=744, type=view]; +"745 transpose_58" [id=745, type=transpose]; +"746 view_75" [id=746, type=view]; +"747 view_76" [id=747, type=view]; +"748 view_77" [id=748, type=view]; +"749 scaled_dot_product_attention_9" [id=749, type=scaled_dot_product_attention]; +"750 permute_10" [id=750, type=permute]; +"751 view_78" [id=751, type=view]; +"752 view_78_0_0_nncf_smooth_quant_0" [id=752, type=call_module]; +"753 quantize_per_tensor_default_58" [id=753, type=quantize_per_tensor]; +"754 dequantize_per_tensor_default_58" [id=754, type=dequantize_per_tensor]; +"755 linear_37_scale_0" [id=755, type=get_attr]; +"756 linear_37_zero_point_0" [id=756, type=get_attr]; +"757 _frozen_param39" [id=757, type=get_attr]; +"758 dequantize_per_channel_default_38" [id=758, type=dequantize_per_channel]; +"759 linear_37" [id=759, type=linear]; +"760 view_79" [id=760, type=view]; +"761 transpose_59" [id=761, type=transpose]; +"762 dropout_28" [id=762, type=dropout]; +"763 add_19" [id=763, type=add]; +"764 layer_norm_19" [id=764, type=layer_norm]; +"765 layer_norm_19_0_0_nncf_smooth_quant_0" [id=765, type=call_module]; +"766 quantize_per_tensor_default_59" [id=766, type=quantize_per_tensor]; +"767 dequantize_per_tensor_default_59" [id=767, type=dequantize_per_tensor]; +"768 linear_38_scale_0" [id=768, type=get_attr]; +"769 linear_38_zero_point_0" [id=769, type=get_attr]; +"770 _frozen_param40" [id=770, type=get_attr]; +"771 dequantize_per_channel_default_39" [id=771, type=dequantize_per_channel]; +"772 linear_38" [id=772, type=linear]; +"773 gelu_9" [id=773, type=gelu]; +"774 dropout_29" [id=774, type=dropout]; +"775 dropout_29_0_0_nncf_smooth_quant_0" [id=775, type=call_module]; +"776 quantize_per_tensor_default_60" [id=776, type=quantize_per_tensor]; +"777 dequantize_per_tensor_default_60" [id=777, type=dequantize_per_tensor]; +"778 linear_39_scale_0" [id=778, type=get_attr]; +"779 linear_39_zero_point_0" [id=779, type=get_attr]; +"780 _frozen_param41" [id=780, type=get_attr]; +"781 dequantize_per_channel_default_40" [id=781, type=dequantize_per_channel]; +"782 linear_39" [id=782, type=linear]; +"783 dropout_30" [id=783, type=dropout]; +"784 add_20" [id=784, type=add]; +"785 layer_norm_20" [id=785, type=layer_norm]; +"786 transpose_60" [id=786, type=transpose]; +"787 transpose_60_0_0_nncf_smooth_quant_0" [id=787, type=call_module]; +"788 quantize_per_tensor_default_61" [id=788, type=quantize_per_tensor]; +"789 dequantize_per_tensor_default_61" [id=789, type=dequantize_per_tensor]; +"790 linear_40_scale_0" [id=790, type=get_attr]; +"791 linear_40_zero_point_0" [id=791, type=get_attr]; +"792 _frozen_param42" [id=792, type=get_attr]; +"793 dequantize_per_channel_default_41" [id=793, type=dequantize_per_channel]; +"794 linear_40" [id=794, type=linear]; +"795 unflatten_10" [id=795, type=unflatten]; +"796 unsqueeze_10" [id=796, type=unsqueeze]; +"797 transpose_61" [id=797, type=transpose]; +"798 squeeze_10" [id=798, type=squeeze]; +"799 contiguous_10" [id=799, type=contiguous]; +"800 quantize_per_tensor_default_62" [id=800, type=quantize_per_tensor]; +"801 dequantize_per_tensor_default_62" [id=801, type=dequantize_per_tensor]; +"802 select_30" [id=802, type=select]; +"803 quantize_per_tensor_default_63" [id=803, type=quantize_per_tensor]; +"804 dequantize_per_tensor_default_63" [id=804, type=dequantize_per_tensor]; +"805 select_31" [id=805, type=select]; +"806 select_32" [id=806, type=select]; +"807 view_80" [id=807, type=view]; +"808 transpose_62" [id=808, type=transpose]; +"809 view_81" [id=809, type=view]; +"810 transpose_63" [id=810, type=transpose]; +"811 view_82" [id=811, type=view]; +"812 transpose_64" [id=812, type=transpose]; +"813 view_83" [id=813, type=view]; +"814 view_84" [id=814, type=view]; +"815 view_85" [id=815, type=view]; +"816 scaled_dot_product_attention_10" [id=816, type=scaled_dot_product_attention]; +"817 permute_11" [id=817, type=permute]; +"818 view_86" [id=818, type=view]; +"819 view_86_0_0_nncf_smooth_quant_0" [id=819, type=call_module]; +"820 quantize_per_tensor_default_64" [id=820, type=quantize_per_tensor]; +"821 dequantize_per_tensor_default_64" [id=821, type=dequantize_per_tensor]; +"822 linear_41_scale_0" [id=822, type=get_attr]; +"823 linear_41_zero_point_0" [id=823, type=get_attr]; +"824 _frozen_param43" [id=824, type=get_attr]; +"825 dequantize_per_channel_default_42" [id=825, type=dequantize_per_channel]; +"826 linear_41" [id=826, type=linear]; +"827 view_87" [id=827, type=view]; +"828 transpose_65" [id=828, type=transpose]; +"829 dropout_31" [id=829, type=dropout]; +"830 add_21" [id=830, type=add]; +"831 layer_norm_21" [id=831, type=layer_norm]; +"832 layer_norm_21_0_0_nncf_smooth_quant_0" [id=832, type=call_module]; +"833 quantize_per_tensor_default_65" [id=833, type=quantize_per_tensor]; +"834 dequantize_per_tensor_default_65" [id=834, type=dequantize_per_tensor]; +"835 linear_42_scale_0" [id=835, type=get_attr]; +"836 linear_42_zero_point_0" [id=836, type=get_attr]; +"837 _frozen_param44" [id=837, type=get_attr]; +"838 dequantize_per_channel_default_43" [id=838, type=dequantize_per_channel]; +"839 linear_42" [id=839, type=linear]; +"840 gelu_10" [id=840, type=gelu]; +"841 dropout_32" [id=841, type=dropout]; +"842 dropout_32_0_0_nncf_smooth_quant_0" [id=842, type=call_module]; +"843 quantize_per_tensor_default_66" [id=843, type=quantize_per_tensor]; +"844 dequantize_per_tensor_default_66" [id=844, type=dequantize_per_tensor]; +"845 linear_43_scale_0" [id=845, type=get_attr]; +"846 linear_43_zero_point_0" [id=846, type=get_attr]; +"847 _frozen_param45" [id=847, type=get_attr]; +"848 dequantize_per_channel_default_44" [id=848, type=dequantize_per_channel]; +"849 linear_43" [id=849, type=linear]; +"850 dropout_33" [id=850, type=dropout]; +"851 add_22" [id=851, type=add]; +"852 layer_norm_22" [id=852, type=layer_norm]; +"853 transpose_66" [id=853, type=transpose]; +"854 transpose_66_0_0_nncf_smooth_quant_0" [id=854, type=call_module]; +"855 quantize_per_tensor_default_67" [id=855, type=quantize_per_tensor]; +"856 dequantize_per_tensor_default_67" [id=856, type=dequantize_per_tensor]; +"857 linear_44_scale_0" [id=857, type=get_attr]; +"858 linear_44_zero_point_0" [id=858, type=get_attr]; +"859 _frozen_param46" [id=859, type=get_attr]; +"860 dequantize_per_channel_default_45" [id=860, type=dequantize_per_channel]; +"861 linear_44" [id=861, type=linear]; +"862 unflatten_11" [id=862, type=unflatten]; +"863 unsqueeze_11" [id=863, type=unsqueeze]; +"864 transpose_67" [id=864, type=transpose]; +"865 squeeze_11" [id=865, type=squeeze]; +"866 contiguous_11" [id=866, type=contiguous]; +"867 quantize_per_tensor_default_68" [id=867, type=quantize_per_tensor]; +"868 dequantize_per_tensor_default_68" [id=868, type=dequantize_per_tensor]; +"869 select_33" [id=869, type=select]; +"870 quantize_per_tensor_default_69" [id=870, type=quantize_per_tensor]; +"871 dequantize_per_tensor_default_69" [id=871, type=dequantize_per_tensor]; +"872 select_34" [id=872, type=select]; +"873 select_35" [id=873, type=select]; +"874 view_88" [id=874, type=view]; +"875 transpose_68" [id=875, type=transpose]; +"876 view_89" [id=876, type=view]; +"877 transpose_69" [id=877, type=transpose]; +"878 view_90" [id=878, type=view]; +"879 transpose_70" [id=879, type=transpose]; +"880 view_91" [id=880, type=view]; +"881 view_92" [id=881, type=view]; +"882 view_93" [id=882, type=view]; +"883 scaled_dot_product_attention_11" [id=883, type=scaled_dot_product_attention]; +"884 permute_12" [id=884, type=permute]; +"885 view_94" [id=885, type=view]; +"886 view_94_0_0_nncf_smooth_quant_0" [id=886, type=call_module]; +"887 quantize_per_tensor_default_70" [id=887, type=quantize_per_tensor]; +"888 dequantize_per_tensor_default_70" [id=888, type=dequantize_per_tensor]; +"889 linear_45_scale_0" [id=889, type=get_attr]; +"890 linear_45_zero_point_0" [id=890, type=get_attr]; +"891 _frozen_param47" [id=891, type=get_attr]; +"892 dequantize_per_channel_default_46" [id=892, type=dequantize_per_channel]; +"893 linear_45" [id=893, type=linear]; +"894 view_95" [id=894, type=view]; +"895 transpose_71" [id=895, type=transpose]; +"896 dropout_34" [id=896, type=dropout]; +"897 add_23" [id=897, type=add]; +"898 layer_norm_23" [id=898, type=layer_norm]; +"899 layer_norm_23_0_0_nncf_smooth_quant_0" [id=899, type=call_module]; +"900 quantize_per_tensor_default_71" [id=900, type=quantize_per_tensor]; +"901 dequantize_per_tensor_default_71" [id=901, type=dequantize_per_tensor]; +"902 linear_46_scale_0" [id=902, type=get_attr]; +"903 linear_46_zero_point_0" [id=903, type=get_attr]; +"904 _frozen_param48" [id=904, type=get_attr]; +"905 dequantize_per_channel_default_47" [id=905, type=dequantize_per_channel]; +"906 linear_46" [id=906, type=linear]; +"907 gelu_11" [id=907, type=gelu]; +"908 dropout_35" [id=908, type=dropout]; +"909 dropout_35_0_0_nncf_smooth_quant_0" [id=909, type=call_module]; +"910 quantize_per_tensor_default_72" [id=910, type=quantize_per_tensor]; +"911 dequantize_per_tensor_default_72" [id=911, type=dequantize_per_tensor]; +"912 linear_47_scale_0" [id=912, type=get_attr]; +"913 linear_47_zero_point_0" [id=913, type=get_attr]; +"914 _frozen_param49" [id=914, type=get_attr]; +"915 dequantize_per_channel_default_48" [id=915, type=dequantize_per_channel]; +"916 linear_47" [id=916, type=linear]; +"917 dropout_36" [id=917, type=dropout]; +"918 add_24" [id=918, type=add]; +"919 layer_norm_24" [id=919, type=layer_norm]; +"920 slice_1" [id=920, type=slice]; +"921 select_36" [id=921, type=select]; +"922 select_36_0_0_nncf_smooth_quant_0" [id=922, type=call_module]; +"923 quantize_per_tensor_default_73" [id=923, type=quantize_per_tensor]; +"924 dequantize_per_tensor_default_73" [id=924, type=dequantize_per_tensor]; +"925 linear_48_scale_0" [id=925, type=get_attr]; +"926 linear_48_zero_point_0" [id=926, type=get_attr]; +"927 _frozen_param50" [id=927, type=get_attr]; +"928 dequantize_per_channel_default_49" [id=928, type=dequantize_per_channel]; +"929 linear_48" [id=929, type=linear]; +"930 output" [id=930, type=output]; +"0 encoder_pos_embedding" -> "113 add" [label="(1, 197, 768)", style=solid]; +"1 conv_proj_bias" -> "108 conv2d" [label="(768,)", style=solid]; +"2 encoder_layers_encoder_layer_0_ln_1_weight" -> "115 layer_norm" [label="(768,)", style=solid]; +"3 encoder_layers_encoder_layer_0_ln_1_bias" -> "115 layer_norm" [label="(768,)", style=solid]; +"4 encoder_layers_encoder_layer_0_self_attention_in_proj_bias" -> "124 linear" [label="(2304,)", style=solid]; +"5 encoder_layers_encoder_layer_0_self_attention_out_proj_bias" -> "156 linear_1" [label="(768,)", style=solid]; +"6 encoder_layers_encoder_layer_0_ln_2_weight" -> "161 layer_norm_1" [label="(768,)", style=solid]; +"7 encoder_layers_encoder_layer_0_ln_2_bias" -> "161 layer_norm_1" [label="(768,)", style=solid]; +"8 encoder_layers_encoder_layer_0_mlp_0_bias" -> "169 linear_2" [label="(3072,)", style=solid]; +"9 encoder_layers_encoder_layer_0_mlp_3_bias" -> "179 linear_3" [label="(768,)", style=solid]; +"10 encoder_layers_encoder_layer_1_ln_1_weight" -> "182 layer_norm_2" [label="(768,)", style=solid]; +"11 encoder_layers_encoder_layer_1_ln_1_bias" -> "182 layer_norm_2" [label="(768,)", style=solid]; +"12 encoder_layers_encoder_layer_1_self_attention_in_proj_bias" -> "191 linear_4" [label="(2304,)", style=solid]; +"13 encoder_layers_encoder_layer_1_self_attention_out_proj_bias" -> "223 linear_5" [label="(768,)", style=solid]; +"14 encoder_layers_encoder_layer_1_ln_2_weight" -> "228 layer_norm_3" [label="(768,)", style=solid]; +"15 encoder_layers_encoder_layer_1_ln_2_bias" -> "228 layer_norm_3" [label="(768,)", style=solid]; +"16 encoder_layers_encoder_layer_1_mlp_0_bias" -> "236 linear_6" [label="(3072,)", style=solid]; +"17 encoder_layers_encoder_layer_1_mlp_3_bias" -> "246 linear_7" [label="(768,)", style=solid]; +"18 encoder_layers_encoder_layer_2_ln_1_weight" -> "249 layer_norm_4" [label="(768,)", style=solid]; +"19 encoder_layers_encoder_layer_2_ln_1_bias" -> "249 layer_norm_4" [label="(768,)", style=solid]; +"20 encoder_layers_encoder_layer_2_self_attention_in_proj_bias" -> "258 linear_8" [label="(2304,)", style=solid]; +"21 encoder_layers_encoder_layer_2_self_attention_out_proj_bias" -> "290 linear_9" [label="(768,)", style=solid]; +"22 encoder_layers_encoder_layer_2_ln_2_weight" -> "295 layer_norm_5" [label="(768,)", style=solid]; +"23 encoder_layers_encoder_layer_2_ln_2_bias" -> "295 layer_norm_5" [label="(768,)", style=solid]; +"24 encoder_layers_encoder_layer_2_mlp_0_bias" -> "303 linear_10" [label="(3072,)", style=solid]; +"25 encoder_layers_encoder_layer_2_mlp_3_bias" -> "313 linear_11" [label="(768,)", style=solid]; +"26 encoder_layers_encoder_layer_3_ln_1_weight" -> "316 layer_norm_6" [label="(768,)", style=solid]; +"27 encoder_layers_encoder_layer_3_ln_1_bias" -> "316 layer_norm_6" [label="(768,)", style=solid]; +"28 encoder_layers_encoder_layer_3_self_attention_in_proj_bias" -> "325 linear_12" [label="(2304,)", style=solid]; +"29 encoder_layers_encoder_layer_3_self_attention_out_proj_bias" -> "357 linear_13" [label="(768,)", style=solid]; +"30 encoder_layers_encoder_layer_3_ln_2_weight" -> "362 layer_norm_7" [label="(768,)", style=solid]; +"31 encoder_layers_encoder_layer_3_ln_2_bias" -> "362 layer_norm_7" [label="(768,)", style=solid]; +"32 encoder_layers_encoder_layer_3_mlp_0_bias" -> "370 linear_14" [label="(3072,)", style=solid]; +"33 encoder_layers_encoder_layer_3_mlp_3_bias" -> "380 linear_15" [label="(768,)", style=solid]; +"34 encoder_layers_encoder_layer_4_ln_1_weight" -> "383 layer_norm_8" [label="(768,)", style=solid]; +"35 encoder_layers_encoder_layer_4_ln_1_bias" -> "383 layer_norm_8" [label="(768,)", style=solid]; +"36 encoder_layers_encoder_layer_4_self_attention_in_proj_bias" -> "392 linear_16" [label="(2304,)", style=solid]; +"37 encoder_layers_encoder_layer_4_self_attention_out_proj_bias" -> "424 linear_17" [label="(768,)", style=solid]; +"38 encoder_layers_encoder_layer_4_ln_2_weight" -> "429 layer_norm_9" [label="(768,)", style=solid]; +"39 encoder_layers_encoder_layer_4_ln_2_bias" -> "429 layer_norm_9" [label="(768,)", style=solid]; +"40 encoder_layers_encoder_layer_4_mlp_0_bias" -> "437 linear_18" [label="(3072,)", style=solid]; +"41 encoder_layers_encoder_layer_4_mlp_3_bias" -> "447 linear_19" [label="(768,)", style=solid]; +"42 encoder_layers_encoder_layer_5_ln_1_weight" -> "450 layer_norm_10" [label="(768,)", style=solid]; +"43 encoder_layers_encoder_layer_5_ln_1_bias" -> "450 layer_norm_10" [label="(768,)", style=solid]; +"44 encoder_layers_encoder_layer_5_self_attention_in_proj_bias" -> "459 linear_20" [label="(2304,)", style=solid]; +"45 encoder_layers_encoder_layer_5_self_attention_out_proj_bias" -> "491 linear_21" [label="(768,)", style=solid]; +"46 encoder_layers_encoder_layer_5_ln_2_weight" -> "496 layer_norm_11" [label="(768,)", style=solid]; +"47 encoder_layers_encoder_layer_5_ln_2_bias" -> "496 layer_norm_11" [label="(768,)", style=solid]; +"48 encoder_layers_encoder_layer_5_mlp_0_bias" -> "504 linear_22" [label="(3072,)", style=solid]; +"49 encoder_layers_encoder_layer_5_mlp_3_bias" -> "514 linear_23" [label="(768,)", style=solid]; +"50 encoder_layers_encoder_layer_6_ln_1_weight" -> "517 layer_norm_12" [label="(768,)", style=solid]; +"51 encoder_layers_encoder_layer_6_ln_1_bias" -> "517 layer_norm_12" [label="(768,)", style=solid]; +"52 encoder_layers_encoder_layer_6_self_attention_in_proj_bias" -> "526 linear_24" [label="(2304,)", style=solid]; +"53 encoder_layers_encoder_layer_6_self_attention_out_proj_bias" -> "558 linear_25" [label="(768,)", style=solid]; +"54 encoder_layers_encoder_layer_6_ln_2_weight" -> "563 layer_norm_13" [label="(768,)", style=solid]; +"55 encoder_layers_encoder_layer_6_ln_2_bias" -> "563 layer_norm_13" [label="(768,)", style=solid]; +"56 encoder_layers_encoder_layer_6_mlp_0_bias" -> "571 linear_26" [label="(3072,)", style=solid]; +"57 encoder_layers_encoder_layer_6_mlp_3_bias" -> "581 linear_27" [label="(768,)", style=solid]; +"58 encoder_layers_encoder_layer_7_ln_1_weight" -> "584 layer_norm_14" [label="(768,)", style=solid]; +"59 encoder_layers_encoder_layer_7_ln_1_bias" -> "584 layer_norm_14" [label="(768,)", style=solid]; +"60 encoder_layers_encoder_layer_7_self_attention_in_proj_bias" -> "593 linear_28" [label="(2304,)", style=solid]; +"61 encoder_layers_encoder_layer_7_self_attention_out_proj_bias" -> "625 linear_29" [label="(768,)", style=solid]; +"62 encoder_layers_encoder_layer_7_ln_2_weight" -> "630 layer_norm_15" [label="(768,)", style=solid]; +"63 encoder_layers_encoder_layer_7_ln_2_bias" -> "630 layer_norm_15" [label="(768,)", style=solid]; +"64 encoder_layers_encoder_layer_7_mlp_0_bias" -> "638 linear_30" [label="(3072,)", style=solid]; +"65 encoder_layers_encoder_layer_7_mlp_3_bias" -> "648 linear_31" [label="(768,)", style=solid]; +"66 encoder_layers_encoder_layer_8_ln_1_weight" -> "651 layer_norm_16" [label="(768,)", style=solid]; +"67 encoder_layers_encoder_layer_8_ln_1_bias" -> "651 layer_norm_16" [label="(768,)", style=solid]; +"68 encoder_layers_encoder_layer_8_self_attention_in_proj_bias" -> "660 linear_32" [label="(2304,)", style=solid]; +"69 encoder_layers_encoder_layer_8_self_attention_out_proj_bias" -> "692 linear_33" [label="(768,)", style=solid]; +"70 encoder_layers_encoder_layer_8_ln_2_weight" -> "697 layer_norm_17" [label="(768,)", style=solid]; +"71 encoder_layers_encoder_layer_8_ln_2_bias" -> "697 layer_norm_17" [label="(768,)", style=solid]; +"72 encoder_layers_encoder_layer_8_mlp_0_bias" -> "705 linear_34" [label="(3072,)", style=solid]; +"73 encoder_layers_encoder_layer_8_mlp_3_bias" -> "715 linear_35" [label="(768,)", style=solid]; +"74 encoder_layers_encoder_layer_9_ln_1_weight" -> "718 layer_norm_18" [label="(768,)", style=solid]; +"75 encoder_layers_encoder_layer_9_ln_1_bias" -> "718 layer_norm_18" [label="(768,)", style=solid]; +"76 encoder_layers_encoder_layer_9_self_attention_in_proj_bias" -> "727 linear_36" [label="(2304,)", style=solid]; +"77 encoder_layers_encoder_layer_9_self_attention_out_proj_bias" -> "759 linear_37" [label="(768,)", style=solid]; +"78 encoder_layers_encoder_layer_9_ln_2_weight" -> "764 layer_norm_19" [label="(768,)", style=solid]; +"79 encoder_layers_encoder_layer_9_ln_2_bias" -> "764 layer_norm_19" [label="(768,)", style=solid]; +"80 encoder_layers_encoder_layer_9_mlp_0_bias" -> "772 linear_38" [label="(3072,)", style=solid]; +"81 encoder_layers_encoder_layer_9_mlp_3_bias" -> "782 linear_39" [label="(768,)", style=solid]; +"82 encoder_layers_encoder_layer_10_ln_1_weight" -> "785 layer_norm_20" [label="(768,)", style=solid]; +"83 encoder_layers_encoder_layer_10_ln_1_bias" -> "785 layer_norm_20" [label="(768,)", style=solid]; +"84 encoder_layers_encoder_layer_10_self_attention_in_proj_bias" -> "794 linear_40" [label="(2304,)", style=solid]; +"85 encoder_layers_encoder_layer_10_self_attention_out_proj_bias" -> "826 linear_41" [label="(768,)", style=solid]; +"86 encoder_layers_encoder_layer_10_ln_2_weight" -> "831 layer_norm_21" [label="(768,)", style=solid]; +"87 encoder_layers_encoder_layer_10_ln_2_bias" -> "831 layer_norm_21" [label="(768,)", style=solid]; +"88 encoder_layers_encoder_layer_10_mlp_0_bias" -> "839 linear_42" [label="(3072,)", style=solid]; +"89 encoder_layers_encoder_layer_10_mlp_3_bias" -> "849 linear_43" [label="(768,)", style=solid]; +"90 encoder_layers_encoder_layer_11_ln_1_weight" -> "852 layer_norm_22" [label="(768,)", style=solid]; +"91 encoder_layers_encoder_layer_11_ln_1_bias" -> "852 layer_norm_22" [label="(768,)", style=solid]; +"92 encoder_layers_encoder_layer_11_self_attention_in_proj_bias" -> "861 linear_44" [label="(2304,)", style=solid]; +"93 encoder_layers_encoder_layer_11_self_attention_out_proj_bias" -> "893 linear_45" [label="(768,)", style=solid]; +"94 encoder_layers_encoder_layer_11_ln_2_weight" -> "898 layer_norm_23" [label="(768,)", style=solid]; +"95 encoder_layers_encoder_layer_11_ln_2_bias" -> "898 layer_norm_23" [label="(768,)", style=solid]; +"96 encoder_layers_encoder_layer_11_mlp_0_bias" -> "906 linear_46" [label="(3072,)", style=solid]; +"97 encoder_layers_encoder_layer_11_mlp_3_bias" -> "916 linear_47" [label="(768,)", style=solid]; +"98 encoder_ln_weight" -> "919 layer_norm_24" [label="(768,)", style=solid]; +"99 encoder_ln_bias" -> "919 layer_norm_24" [label="(768,)", style=solid]; +"100 heads_head_bias" -> "929 linear_48" [label="(1000,)", style=solid]; +"101 x" -> "102 quantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; +"102 quantize_per_tensor_default" -> "103 dequantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; +"103 dequantize_per_tensor_default" -> "108 conv2d" [label="(1, 3, 224, 224)", style=solid]; +"104 conv2d_scale_0" -> "107 dequantize_per_channel_default" [label="(768,)", style=solid]; +"105 conv2d_zero_point_0" -> "107 dequantize_per_channel_default" [label="(768,)", style=solid]; +"106 _frozen_param1" -> "107 dequantize_per_channel_default" [label="(768, 3, 16, 16)", style=solid]; +"107 dequantize_per_channel_default" -> "108 conv2d" [label="(768, 3, 16, 16)", style=solid]; +"108 conv2d" -> "109 reshape" [label="(1, 768, 14, 14)", style=solid]; +"109 reshape" -> "110 permute" [label="(1, 768, 196)", style=solid]; +"110 permute" -> "112 cat" [label="(1, 196, 768)", style=solid]; +"111 _frozen_param0" -> "112 cat" [label="(1, 1, 768)", style=solid]; +"112 cat" -> "113 add" [label="(1, 197, 768)", style=solid]; +"113 add" -> "114 dropout" [label="(1, 197, 768)", style=solid]; +"114 dropout" -> "115 layer_norm" [label="(1, 197, 768)", style=solid]; +"114 dropout" -> "160 add_1" [label="(1, 197, 768)", style=solid]; +"115 layer_norm" -> "116 transpose" [label="(1, 197, 768)", style=solid]; +"116 transpose" -> "117 transpose_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"117 transpose_0_0_nncf_smooth_quant_0" -> "118 quantize_per_tensor_default_1" [label="(197, 1, 768)", style=solid]; +"118 quantize_per_tensor_default_1" -> "119 dequantize_per_tensor_default_1" [label="(197, 1, 768)", style=solid]; +"119 dequantize_per_tensor_default_1" -> "124 linear" [label="(197, 1, 768)", style=solid]; +"120 linear_scale_0" -> "123 dequantize_per_channel_default_1" [label="(2304,)", style=solid]; +"121 linear_zero_point_0" -> "123 dequantize_per_channel_default_1" [label="(2304,)", style=solid]; +"122 _frozen_param2" -> "123 dequantize_per_channel_default_1" [label="(2304, 768)", style=solid]; +"123 dequantize_per_channel_default_1" -> "124 linear" [label="(2304, 768)", style=solid]; +"124 linear" -> "125 unflatten" [label="(197, 1, 2304)", style=solid]; +"125 unflatten" -> "126 unsqueeze" [label="(197, 1, 3, 768)", style=solid]; +"126 unsqueeze" -> "127 transpose_1" [label="(1, 197, 1, 3, 768)", style=solid]; +"127 transpose_1" -> "128 squeeze" [label="(3, 197, 1, 1, 768)", style=solid]; +"128 squeeze" -> "129 contiguous" [label="(3, 197, 1, 768)", style=solid]; +"129 contiguous" -> "130 quantize_per_tensor_default_2" [label="(3, 197, 1, 768)", style=solid]; +"129 contiguous" -> "133 quantize_per_tensor_default_3" [label="(3, 197, 1, 768)", style=solid]; +"129 contiguous" -> "136 select_2" [label="(3, 197, 1, 768)", style=solid]; +"130 quantize_per_tensor_default_2" -> "131 dequantize_per_tensor_default_2" [label="(3, 197, 1, 768)", style=solid]; +"131 dequantize_per_tensor_default_2" -> "132 select" [label="(3, 197, 1, 768)", style=solid]; +"132 select" -> "137 view" [label="(197, 1, 768)", style=solid]; +"133 quantize_per_tensor_default_3" -> "134 dequantize_per_tensor_default_3" [label="(3, 197, 1, 768)", style=solid]; +"134 dequantize_per_tensor_default_3" -> "135 select_1" [label="(3, 197, 1, 768)", style=solid]; +"135 select_1" -> "139 view_1" [label="(197, 1, 768)", style=solid]; +"136 select_2" -> "141 view_2" [label="(197, 1, 768)", style=solid]; +"137 view" -> "138 transpose_2" [label="(197, 12, 64)", style=solid]; +"138 transpose_2" -> "143 view_3" [label="(12, 197, 64)", style=solid]; +"139 view_1" -> "140 transpose_3" [label="(197, 12, 64)", style=solid]; +"140 transpose_3" -> "144 view_4" [label="(12, 197, 64)", style=solid]; +"141 view_2" -> "142 transpose_4" [label="(197, 12, 64)", style=solid]; +"142 transpose_4" -> "145 view_5" [label="(12, 197, 64)", style=solid]; +"143 view_3" -> "146 scaled_dot_product_attention" [label="(1, 12, 197, 64)", style=solid]; +"144 view_4" -> "146 scaled_dot_product_attention" [label="(1, 12, 197, 64)", style=solid]; +"145 view_5" -> "146 scaled_dot_product_attention" [label="(1, 12, 197, 64)", style=solid]; +"146 scaled_dot_product_attention" -> "147 permute_1" [label="(1, 12, 197, 64)", style=solid]; +"147 permute_1" -> "148 view_6" [label="(197, 1, 12, 64)", style=solid]; +"148 view_6" -> "149 view_6_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"149 view_6_0_0_nncf_smooth_quant_0" -> "150 quantize_per_tensor_default_4" [label="(197, 768)", style=solid]; +"150 quantize_per_tensor_default_4" -> "151 dequantize_per_tensor_default_4" [label="(197, 768)", style=solid]; +"151 dequantize_per_tensor_default_4" -> "156 linear_1" [label="(197, 768)", style=solid]; +"152 linear_1_scale_0" -> "155 dequantize_per_channel_default_2" [label="(768,)", style=solid]; +"153 linear_1_zero_point_0" -> "155 dequantize_per_channel_default_2" [label="(768,)", style=solid]; +"154 _frozen_param3" -> "155 dequantize_per_channel_default_2" [label="(768, 768)", style=solid]; +"155 dequantize_per_channel_default_2" -> "156 linear_1" [label="(768, 768)", style=solid]; +"156 linear_1" -> "157 view_7" [label="(197, 768)", style=solid]; +"157 view_7" -> "158 transpose_5" [label="(197, 1, 768)", style=solid]; +"158 transpose_5" -> "159 dropout_1" [label="(1, 197, 768)", style=solid]; +"159 dropout_1" -> "160 add_1" [label="(1, 197, 768)", style=solid]; +"160 add_1" -> "161 layer_norm_1" [label="(1, 197, 768)", style=solid]; +"160 add_1" -> "181 add_2" [label="(1, 197, 768)", style=solid]; +"161 layer_norm_1" -> "162 layer_norm_1_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"162 layer_norm_1_0_0_nncf_smooth_quant_0" -> "163 quantize_per_tensor_default_5" [label="(1, 197, 768)", style=solid]; +"163 quantize_per_tensor_default_5" -> "164 dequantize_per_tensor_default_5" [label="(1, 197, 768)", style=solid]; +"164 dequantize_per_tensor_default_5" -> "169 linear_2" [label="(1, 197, 768)", style=solid]; +"165 linear_2_scale_0" -> "168 dequantize_per_channel_default_3" [label="(3072,)", style=solid]; +"166 linear_2_zero_point_0" -> "168 dequantize_per_channel_default_3" [label="(3072,)", style=solid]; +"167 _frozen_param4" -> "168 dequantize_per_channel_default_3" [label="(3072, 768)", style=solid]; +"168 dequantize_per_channel_default_3" -> "169 linear_2" [label="(3072, 768)", style=solid]; +"169 linear_2" -> "170 gelu" [label="(1, 197, 3072)", style=solid]; +"170 gelu" -> "171 dropout_2" [label="(1, 197, 3072)", style=solid]; +"171 dropout_2" -> "172 dropout_2_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"172 dropout_2_0_0_nncf_smooth_quant_0" -> "173 quantize_per_tensor_default_6" [label="(1, 197, 3072)", style=solid]; +"173 quantize_per_tensor_default_6" -> "174 dequantize_per_tensor_default_6" [label="(1, 197, 3072)", style=solid]; +"174 dequantize_per_tensor_default_6" -> "179 linear_3" [label="(1, 197, 3072)", style=solid]; +"175 linear_3_scale_0" -> "178 dequantize_per_channel_default_4" [label="(768,)", style=solid]; +"176 linear_3_zero_point_0" -> "178 dequantize_per_channel_default_4" [label="(768,)", style=solid]; +"177 _frozen_param5" -> "178 dequantize_per_channel_default_4" [label="(768, 3072)", style=solid]; +"178 dequantize_per_channel_default_4" -> "179 linear_3" [label="(768, 3072)", style=solid]; +"179 linear_3" -> "180 dropout_3" [label="(1, 197, 768)", style=solid]; +"180 dropout_3" -> "181 add_2" [label="(1, 197, 768)", style=solid]; +"181 add_2" -> "182 layer_norm_2" [label="(1, 197, 768)", style=solid]; +"181 add_2" -> "227 add_3" [label="(1, 197, 768)", style=solid]; +"182 layer_norm_2" -> "183 transpose_6" [label="(1, 197, 768)", style=solid]; +"183 transpose_6" -> "184 transpose_6_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"184 transpose_6_0_0_nncf_smooth_quant_0" -> "185 quantize_per_tensor_default_7" [label="(197, 1, 768)", style=solid]; +"185 quantize_per_tensor_default_7" -> "186 dequantize_per_tensor_default_7" [label="(197, 1, 768)", style=solid]; +"186 dequantize_per_tensor_default_7" -> "191 linear_4" [label="(197, 1, 768)", style=solid]; +"187 linear_4_scale_0" -> "190 dequantize_per_channel_default_5" [label="(2304,)", style=solid]; +"188 linear_4_zero_point_0" -> "190 dequantize_per_channel_default_5" [label="(2304,)", style=solid]; +"189 _frozen_param6" -> "190 dequantize_per_channel_default_5" [label="(2304, 768)", style=solid]; +"190 dequantize_per_channel_default_5" -> "191 linear_4" [label="(2304, 768)", style=solid]; +"191 linear_4" -> "192 unflatten_1" [label="(197, 1, 2304)", style=solid]; +"192 unflatten_1" -> "193 unsqueeze_1" [label="(197, 1, 3, 768)", style=solid]; +"193 unsqueeze_1" -> "194 transpose_7" [label="(1, 197, 1, 3, 768)", style=solid]; +"194 transpose_7" -> "195 squeeze_1" [label="(3, 197, 1, 1, 768)", style=solid]; +"195 squeeze_1" -> "196 contiguous_1" [label="(3, 197, 1, 768)", style=solid]; +"196 contiguous_1" -> "197 quantize_per_tensor_default_8" [label="(3, 197, 1, 768)", style=solid]; +"196 contiguous_1" -> "200 quantize_per_tensor_default_9" [label="(3, 197, 1, 768)", style=solid]; +"196 contiguous_1" -> "203 select_5" [label="(3, 197, 1, 768)", style=solid]; +"197 quantize_per_tensor_default_8" -> "198 dequantize_per_tensor_default_8" [label="(3, 197, 1, 768)", style=solid]; +"198 dequantize_per_tensor_default_8" -> "199 select_3" [label="(3, 197, 1, 768)", style=solid]; +"199 select_3" -> "204 view_8" [label="(197, 1, 768)", style=solid]; +"200 quantize_per_tensor_default_9" -> "201 dequantize_per_tensor_default_9" [label="(3, 197, 1, 768)", style=solid]; +"201 dequantize_per_tensor_default_9" -> "202 select_4" [label="(3, 197, 1, 768)", style=solid]; +"202 select_4" -> "206 view_9" [label="(197, 1, 768)", style=solid]; +"203 select_5" -> "208 view_10" [label="(197, 1, 768)", style=solid]; +"204 view_8" -> "205 transpose_8" [label="(197, 12, 64)", style=solid]; +"205 transpose_8" -> "210 view_11" [label="(12, 197, 64)", style=solid]; +"206 view_9" -> "207 transpose_9" [label="(197, 12, 64)", style=solid]; +"207 transpose_9" -> "211 view_12" [label="(12, 197, 64)", style=solid]; +"208 view_10" -> "209 transpose_10" [label="(197, 12, 64)", style=solid]; +"209 transpose_10" -> "212 view_13" [label="(12, 197, 64)", style=solid]; +"210 view_11" -> "213 scaled_dot_product_attention_1" [label="(1, 12, 197, 64)", style=solid]; +"211 view_12" -> "213 scaled_dot_product_attention_1" [label="(1, 12, 197, 64)", style=solid]; +"212 view_13" -> "213 scaled_dot_product_attention_1" [label="(1, 12, 197, 64)", style=solid]; +"213 scaled_dot_product_attention_1" -> "214 permute_2" [label="(1, 12, 197, 64)", style=solid]; +"214 permute_2" -> "215 view_14" [label="(197, 1, 12, 64)", style=solid]; +"215 view_14" -> "216 view_14_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"216 view_14_0_0_nncf_smooth_quant_0" -> "217 quantize_per_tensor_default_10" [label="(197, 768)", style=solid]; +"217 quantize_per_tensor_default_10" -> "218 dequantize_per_tensor_default_10" [label="(197, 768)", style=solid]; +"218 dequantize_per_tensor_default_10" -> "223 linear_5" [label="(197, 768)", style=solid]; +"219 linear_5_scale_0" -> "222 dequantize_per_channel_default_6" [label="(768,)", style=solid]; +"220 linear_5_zero_point_0" -> "222 dequantize_per_channel_default_6" [label="(768,)", style=solid]; +"221 _frozen_param7" -> "222 dequantize_per_channel_default_6" [label="(768, 768)", style=solid]; +"222 dequantize_per_channel_default_6" -> "223 linear_5" [label="(768, 768)", style=solid]; +"223 linear_5" -> "224 view_15" [label="(197, 768)", style=solid]; +"224 view_15" -> "225 transpose_11" [label="(197, 1, 768)", style=solid]; +"225 transpose_11" -> "226 dropout_4" [label="(1, 197, 768)", style=solid]; +"226 dropout_4" -> "227 add_3" [label="(1, 197, 768)", style=solid]; +"227 add_3" -> "228 layer_norm_3" [label="(1, 197, 768)", style=solid]; +"227 add_3" -> "248 add_4" [label="(1, 197, 768)", style=solid]; +"228 layer_norm_3" -> "229 layer_norm_3_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"229 layer_norm_3_0_0_nncf_smooth_quant_0" -> "230 quantize_per_tensor_default_11" [label="(1, 197, 768)", style=solid]; +"230 quantize_per_tensor_default_11" -> "231 dequantize_per_tensor_default_11" [label="(1, 197, 768)", style=solid]; +"231 dequantize_per_tensor_default_11" -> "236 linear_6" [label="(1, 197, 768)", style=solid]; +"232 linear_6_scale_0" -> "235 dequantize_per_channel_default_7" [label="(3072,)", style=solid]; +"233 linear_6_zero_point_0" -> "235 dequantize_per_channel_default_7" [label="(3072,)", style=solid]; +"234 _frozen_param8" -> "235 dequantize_per_channel_default_7" [label="(3072, 768)", style=solid]; +"235 dequantize_per_channel_default_7" -> "236 linear_6" [label="(3072, 768)", style=solid]; +"236 linear_6" -> "237 gelu_1" [label="(1, 197, 3072)", style=solid]; +"237 gelu_1" -> "238 dropout_5" [label="(1, 197, 3072)", style=solid]; +"238 dropout_5" -> "239 dropout_5_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"239 dropout_5_0_0_nncf_smooth_quant_0" -> "240 quantize_per_tensor_default_12" [label="(1, 197, 3072)", style=solid]; +"240 quantize_per_tensor_default_12" -> "241 dequantize_per_tensor_default_12" [label="(1, 197, 3072)", style=solid]; +"241 dequantize_per_tensor_default_12" -> "246 linear_7" [label="(1, 197, 3072)", style=solid]; +"242 linear_7_scale_0" -> "245 dequantize_per_channel_default_8" [label="(768,)", style=solid]; +"243 linear_7_zero_point_0" -> "245 dequantize_per_channel_default_8" [label="(768,)", style=solid]; +"244 _frozen_param9" -> "245 dequantize_per_channel_default_8" [label="(768, 3072)", style=solid]; +"245 dequantize_per_channel_default_8" -> "246 linear_7" [label="(768, 3072)", style=solid]; +"246 linear_7" -> "247 dropout_6" [label="(1, 197, 768)", style=solid]; +"247 dropout_6" -> "248 add_4" [label="(1, 197, 768)", style=solid]; +"248 add_4" -> "249 layer_norm_4" [label="(1, 197, 768)", style=solid]; +"248 add_4" -> "294 add_5" [label="(1, 197, 768)", style=solid]; +"249 layer_norm_4" -> "250 transpose_12" [label="(1, 197, 768)", style=solid]; +"250 transpose_12" -> "251 transpose_12_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"251 transpose_12_0_0_nncf_smooth_quant_0" -> "252 quantize_per_tensor_default_13" [label="(197, 1, 768)", style=solid]; +"252 quantize_per_tensor_default_13" -> "253 dequantize_per_tensor_default_13" [label="(197, 1, 768)", style=solid]; +"253 dequantize_per_tensor_default_13" -> "258 linear_8" [label="(197, 1, 768)", style=solid]; +"254 linear_8_scale_0" -> "257 dequantize_per_channel_default_9" [label="(2304,)", style=solid]; +"255 linear_8_zero_point_0" -> "257 dequantize_per_channel_default_9" [label="(2304,)", style=solid]; +"256 _frozen_param10" -> "257 dequantize_per_channel_default_9" [label="(2304, 768)", style=solid]; +"257 dequantize_per_channel_default_9" -> "258 linear_8" [label="(2304, 768)", style=solid]; +"258 linear_8" -> "259 unflatten_2" [label="(197, 1, 2304)", style=solid]; +"259 unflatten_2" -> "260 unsqueeze_2" [label="(197, 1, 3, 768)", style=solid]; +"260 unsqueeze_2" -> "261 transpose_13" [label="(1, 197, 1, 3, 768)", style=solid]; +"261 transpose_13" -> "262 squeeze_2" [label="(3, 197, 1, 1, 768)", style=solid]; +"262 squeeze_2" -> "263 contiguous_2" [label="(3, 197, 1, 768)", style=solid]; +"263 contiguous_2" -> "264 quantize_per_tensor_default_14" [label="(3, 197, 1, 768)", style=solid]; +"263 contiguous_2" -> "267 quantize_per_tensor_default_15" [label="(3, 197, 1, 768)", style=solid]; +"263 contiguous_2" -> "270 select_8" [label="(3, 197, 1, 768)", style=solid]; +"264 quantize_per_tensor_default_14" -> "265 dequantize_per_tensor_default_14" [label="(3, 197, 1, 768)", style=solid]; +"265 dequantize_per_tensor_default_14" -> "266 select_6" [label="(3, 197, 1, 768)", style=solid]; +"266 select_6" -> "271 view_16" [label="(197, 1, 768)", style=solid]; +"267 quantize_per_tensor_default_15" -> "268 dequantize_per_tensor_default_15" [label="(3, 197, 1, 768)", style=solid]; +"268 dequantize_per_tensor_default_15" -> "269 select_7" [label="(3, 197, 1, 768)", style=solid]; +"269 select_7" -> "273 view_17" [label="(197, 1, 768)", style=solid]; +"270 select_8" -> "275 view_18" [label="(197, 1, 768)", style=solid]; +"271 view_16" -> "272 transpose_14" [label="(197, 12, 64)", style=solid]; +"272 transpose_14" -> "277 view_19" [label="(12, 197, 64)", style=solid]; +"273 view_17" -> "274 transpose_15" [label="(197, 12, 64)", style=solid]; +"274 transpose_15" -> "278 view_20" [label="(12, 197, 64)", style=solid]; +"275 view_18" -> "276 transpose_16" [label="(197, 12, 64)", style=solid]; +"276 transpose_16" -> "279 view_21" [label="(12, 197, 64)", style=solid]; +"277 view_19" -> "280 scaled_dot_product_attention_2" [label="(1, 12, 197, 64)", style=solid]; +"278 view_20" -> "280 scaled_dot_product_attention_2" [label="(1, 12, 197, 64)", style=solid]; +"279 view_21" -> "280 scaled_dot_product_attention_2" [label="(1, 12, 197, 64)", style=solid]; +"280 scaled_dot_product_attention_2" -> "281 permute_3" [label="(1, 12, 197, 64)", style=solid]; +"281 permute_3" -> "282 view_22" [label="(197, 1, 12, 64)", style=solid]; +"282 view_22" -> "283 view_22_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"283 view_22_0_0_nncf_smooth_quant_0" -> "284 quantize_per_tensor_default_16" [label="(197, 768)", style=solid]; +"284 quantize_per_tensor_default_16" -> "285 dequantize_per_tensor_default_16" [label="(197, 768)", style=solid]; +"285 dequantize_per_tensor_default_16" -> "290 linear_9" [label="(197, 768)", style=solid]; +"286 linear_9_scale_0" -> "289 dequantize_per_channel_default_10" [label="(768,)", style=solid]; +"287 linear_9_zero_point_0" -> "289 dequantize_per_channel_default_10" [label="(768,)", style=solid]; +"288 _frozen_param11" -> "289 dequantize_per_channel_default_10" [label="(768, 768)", style=solid]; +"289 dequantize_per_channel_default_10" -> "290 linear_9" [label="(768, 768)", style=solid]; +"290 linear_9" -> "291 view_23" [label="(197, 768)", style=solid]; +"291 view_23" -> "292 transpose_17" [label="(197, 1, 768)", style=solid]; +"292 transpose_17" -> "293 dropout_7" [label="(1, 197, 768)", style=solid]; +"293 dropout_7" -> "294 add_5" [label="(1, 197, 768)", style=solid]; +"294 add_5" -> "295 layer_norm_5" [label="(1, 197, 768)", style=solid]; +"294 add_5" -> "315 add_6" [label="(1, 197, 768)", style=solid]; +"295 layer_norm_5" -> "296 layer_norm_5_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"296 layer_norm_5_0_0_nncf_smooth_quant_0" -> "297 quantize_per_tensor_default_17" [label="(1, 197, 768)", style=solid]; +"297 quantize_per_tensor_default_17" -> "298 dequantize_per_tensor_default_17" [label="(1, 197, 768)", style=solid]; +"298 dequantize_per_tensor_default_17" -> "303 linear_10" [label="(1, 197, 768)", style=solid]; +"299 linear_10_scale_0" -> "302 dequantize_per_channel_default_11" [label="(3072,)", style=solid]; +"300 linear_10_zero_point_0" -> "302 dequantize_per_channel_default_11" [label="(3072,)", style=solid]; +"301 _frozen_param12" -> "302 dequantize_per_channel_default_11" [label="(3072, 768)", style=solid]; +"302 dequantize_per_channel_default_11" -> "303 linear_10" [label="(3072, 768)", style=solid]; +"303 linear_10" -> "304 gelu_2" [label="(1, 197, 3072)", style=solid]; +"304 gelu_2" -> "305 dropout_8" [label="(1, 197, 3072)", style=solid]; +"305 dropout_8" -> "306 dropout_8_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"306 dropout_8_0_0_nncf_smooth_quant_0" -> "307 quantize_per_tensor_default_18" [label="(1, 197, 3072)", style=solid]; +"307 quantize_per_tensor_default_18" -> "308 dequantize_per_tensor_default_18" [label="(1, 197, 3072)", style=solid]; +"308 dequantize_per_tensor_default_18" -> "313 linear_11" [label="(1, 197, 3072)", style=solid]; +"309 linear_11_scale_0" -> "312 dequantize_per_channel_default_12" [label="(768,)", style=solid]; +"310 linear_11_zero_point_0" -> "312 dequantize_per_channel_default_12" [label="(768,)", style=solid]; +"311 _frozen_param13" -> "312 dequantize_per_channel_default_12" [label="(768, 3072)", style=solid]; +"312 dequantize_per_channel_default_12" -> "313 linear_11" [label="(768, 3072)", style=solid]; +"313 linear_11" -> "314 dropout_9" [label="(1, 197, 768)", style=solid]; +"314 dropout_9" -> "315 add_6" [label="(1, 197, 768)", style=solid]; +"315 add_6" -> "316 layer_norm_6" [label="(1, 197, 768)", style=solid]; +"315 add_6" -> "361 add_7" [label="(1, 197, 768)", style=solid]; +"316 layer_norm_6" -> "317 transpose_18" [label="(1, 197, 768)", style=solid]; +"317 transpose_18" -> "318 transpose_18_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"318 transpose_18_0_0_nncf_smooth_quant_0" -> "319 quantize_per_tensor_default_19" [label="(197, 1, 768)", style=solid]; +"319 quantize_per_tensor_default_19" -> "320 dequantize_per_tensor_default_19" [label="(197, 1, 768)", style=solid]; +"320 dequantize_per_tensor_default_19" -> "325 linear_12" [label="(197, 1, 768)", style=solid]; +"321 linear_12_scale_0" -> "324 dequantize_per_channel_default_13" [label="(2304,)", style=solid]; +"322 linear_12_zero_point_0" -> "324 dequantize_per_channel_default_13" [label="(2304,)", style=solid]; +"323 _frozen_param14" -> "324 dequantize_per_channel_default_13" [label="(2304, 768)", style=solid]; +"324 dequantize_per_channel_default_13" -> "325 linear_12" [label="(2304, 768)", style=solid]; +"325 linear_12" -> "326 unflatten_3" [label="(197, 1, 2304)", style=solid]; +"326 unflatten_3" -> "327 unsqueeze_3" [label="(197, 1, 3, 768)", style=solid]; +"327 unsqueeze_3" -> "328 transpose_19" [label="(1, 197, 1, 3, 768)", style=solid]; +"328 transpose_19" -> "329 squeeze_3" [label="(3, 197, 1, 1, 768)", style=solid]; +"329 squeeze_3" -> "330 contiguous_3" [label="(3, 197, 1, 768)", style=solid]; +"330 contiguous_3" -> "331 quantize_per_tensor_default_20" [label="(3, 197, 1, 768)", style=solid]; +"330 contiguous_3" -> "334 quantize_per_tensor_default_21" [label="(3, 197, 1, 768)", style=solid]; +"330 contiguous_3" -> "337 select_11" [label="(3, 197, 1, 768)", style=solid]; +"331 quantize_per_tensor_default_20" -> "332 dequantize_per_tensor_default_20" [label="(3, 197, 1, 768)", style=solid]; +"332 dequantize_per_tensor_default_20" -> "333 select_9" [label="(3, 197, 1, 768)", style=solid]; +"333 select_9" -> "338 view_24" [label="(197, 1, 768)", style=solid]; +"334 quantize_per_tensor_default_21" -> "335 dequantize_per_tensor_default_21" [label="(3, 197, 1, 768)", style=solid]; +"335 dequantize_per_tensor_default_21" -> "336 select_10" [label="(3, 197, 1, 768)", style=solid]; +"336 select_10" -> "340 view_25" [label="(197, 1, 768)", style=solid]; +"337 select_11" -> "342 view_26" [label="(197, 1, 768)", style=solid]; +"338 view_24" -> "339 transpose_20" [label="(197, 12, 64)", style=solid]; +"339 transpose_20" -> "344 view_27" [label="(12, 197, 64)", style=solid]; +"340 view_25" -> "341 transpose_21" [label="(197, 12, 64)", style=solid]; +"341 transpose_21" -> "345 view_28" [label="(12, 197, 64)", style=solid]; +"342 view_26" -> "343 transpose_22" [label="(197, 12, 64)", style=solid]; +"343 transpose_22" -> "346 view_29" [label="(12, 197, 64)", style=solid]; +"344 view_27" -> "347 scaled_dot_product_attention_3" [label="(1, 12, 197, 64)", style=solid]; +"345 view_28" -> "347 scaled_dot_product_attention_3" [label="(1, 12, 197, 64)", style=solid]; +"346 view_29" -> "347 scaled_dot_product_attention_3" [label="(1, 12, 197, 64)", style=solid]; +"347 scaled_dot_product_attention_3" -> "348 permute_4" [label="(1, 12, 197, 64)", style=solid]; +"348 permute_4" -> "349 view_30" [label="(197, 1, 12, 64)", style=solid]; +"349 view_30" -> "350 view_30_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"350 view_30_0_0_nncf_smooth_quant_0" -> "351 quantize_per_tensor_default_22" [label="(197, 768)", style=solid]; +"351 quantize_per_tensor_default_22" -> "352 dequantize_per_tensor_default_22" [label="(197, 768)", style=solid]; +"352 dequantize_per_tensor_default_22" -> "357 linear_13" [label="(197, 768)", style=solid]; +"353 linear_13_scale_0" -> "356 dequantize_per_channel_default_14" [label="(768,)", style=solid]; +"354 linear_13_zero_point_0" -> "356 dequantize_per_channel_default_14" [label="(768,)", style=solid]; +"355 _frozen_param15" -> "356 dequantize_per_channel_default_14" [label="(768, 768)", style=solid]; +"356 dequantize_per_channel_default_14" -> "357 linear_13" [label="(768, 768)", style=solid]; +"357 linear_13" -> "358 view_31" [label="(197, 768)", style=solid]; +"358 view_31" -> "359 transpose_23" [label="(197, 1, 768)", style=solid]; +"359 transpose_23" -> "360 dropout_10" [label="(1, 197, 768)", style=solid]; +"360 dropout_10" -> "361 add_7" [label="(1, 197, 768)", style=solid]; +"361 add_7" -> "362 layer_norm_7" [label="(1, 197, 768)", style=solid]; +"361 add_7" -> "382 add_8" [label="(1, 197, 768)", style=solid]; +"362 layer_norm_7" -> "363 layer_norm_7_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"363 layer_norm_7_0_0_nncf_smooth_quant_0" -> "364 quantize_per_tensor_default_23" [label="(1, 197, 768)", style=solid]; +"364 quantize_per_tensor_default_23" -> "365 dequantize_per_tensor_default_23" [label="(1, 197, 768)", style=solid]; +"365 dequantize_per_tensor_default_23" -> "370 linear_14" [label="(1, 197, 768)", style=solid]; +"366 linear_14_scale_0" -> "369 dequantize_per_channel_default_15" [label="(3072,)", style=solid]; +"367 linear_14_zero_point_0" -> "369 dequantize_per_channel_default_15" [label="(3072,)", style=solid]; +"368 _frozen_param16" -> "369 dequantize_per_channel_default_15" [label="(3072, 768)", style=solid]; +"369 dequantize_per_channel_default_15" -> "370 linear_14" [label="(3072, 768)", style=solid]; +"370 linear_14" -> "371 gelu_3" [label="(1, 197, 3072)", style=solid]; +"371 gelu_3" -> "372 dropout_11" [label="(1, 197, 3072)", style=solid]; +"372 dropout_11" -> "373 dropout_11_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"373 dropout_11_0_0_nncf_smooth_quant_0" -> "374 quantize_per_tensor_default_24" [label="(1, 197, 3072)", style=solid]; +"374 quantize_per_tensor_default_24" -> "375 dequantize_per_tensor_default_24" [label="(1, 197, 3072)", style=solid]; +"375 dequantize_per_tensor_default_24" -> "380 linear_15" [label="(1, 197, 3072)", style=solid]; +"376 linear_15_scale_0" -> "379 dequantize_per_channel_default_16" [label="(768,)", style=solid]; +"377 linear_15_zero_point_0" -> "379 dequantize_per_channel_default_16" [label="(768,)", style=solid]; +"378 _frozen_param17" -> "379 dequantize_per_channel_default_16" [label="(768, 3072)", style=solid]; +"379 dequantize_per_channel_default_16" -> "380 linear_15" [label="(768, 3072)", style=solid]; +"380 linear_15" -> "381 dropout_12" [label="(1, 197, 768)", style=solid]; +"381 dropout_12" -> "382 add_8" [label="(1, 197, 768)", style=solid]; +"382 add_8" -> "383 layer_norm_8" [label="(1, 197, 768)", style=solid]; +"382 add_8" -> "428 add_9" [label="(1, 197, 768)", style=solid]; +"383 layer_norm_8" -> "384 transpose_24" [label="(1, 197, 768)", style=solid]; +"384 transpose_24" -> "385 transpose_24_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"385 transpose_24_0_0_nncf_smooth_quant_0" -> "386 quantize_per_tensor_default_25" [label="(197, 1, 768)", style=solid]; +"386 quantize_per_tensor_default_25" -> "387 dequantize_per_tensor_default_25" [label="(197, 1, 768)", style=solid]; +"387 dequantize_per_tensor_default_25" -> "392 linear_16" [label="(197, 1, 768)", style=solid]; +"388 linear_16_scale_0" -> "391 dequantize_per_channel_default_17" [label="(2304,)", style=solid]; +"389 linear_16_zero_point_0" -> "391 dequantize_per_channel_default_17" [label="(2304,)", style=solid]; +"390 _frozen_param18" -> "391 dequantize_per_channel_default_17" [label="(2304, 768)", style=solid]; +"391 dequantize_per_channel_default_17" -> "392 linear_16" [label="(2304, 768)", style=solid]; +"392 linear_16" -> "393 unflatten_4" [label="(197, 1, 2304)", style=solid]; +"393 unflatten_4" -> "394 unsqueeze_4" [label="(197, 1, 3, 768)", style=solid]; +"394 unsqueeze_4" -> "395 transpose_25" [label="(1, 197, 1, 3, 768)", style=solid]; +"395 transpose_25" -> "396 squeeze_4" [label="(3, 197, 1, 1, 768)", style=solid]; +"396 squeeze_4" -> "397 contiguous_4" [label="(3, 197, 1, 768)", style=solid]; +"397 contiguous_4" -> "398 quantize_per_tensor_default_26" [label="(3, 197, 1, 768)", style=solid]; +"397 contiguous_4" -> "401 quantize_per_tensor_default_27" [label="(3, 197, 1, 768)", style=solid]; +"397 contiguous_4" -> "404 select_14" [label="(3, 197, 1, 768)", style=solid]; +"398 quantize_per_tensor_default_26" -> "399 dequantize_per_tensor_default_26" [label="(3, 197, 1, 768)", style=solid]; +"399 dequantize_per_tensor_default_26" -> "400 select_12" [label="(3, 197, 1, 768)", style=solid]; +"400 select_12" -> "405 view_32" [label="(197, 1, 768)", style=solid]; +"401 quantize_per_tensor_default_27" -> "402 dequantize_per_tensor_default_27" [label="(3, 197, 1, 768)", style=solid]; +"402 dequantize_per_tensor_default_27" -> "403 select_13" [label="(3, 197, 1, 768)", style=solid]; +"403 select_13" -> "407 view_33" [label="(197, 1, 768)", style=solid]; +"404 select_14" -> "409 view_34" [label="(197, 1, 768)", style=solid]; +"405 view_32" -> "406 transpose_26" [label="(197, 12, 64)", style=solid]; +"406 transpose_26" -> "411 view_35" [label="(12, 197, 64)", style=solid]; +"407 view_33" -> "408 transpose_27" [label="(197, 12, 64)", style=solid]; +"408 transpose_27" -> "412 view_36" [label="(12, 197, 64)", style=solid]; +"409 view_34" -> "410 transpose_28" [label="(197, 12, 64)", style=solid]; +"410 transpose_28" -> "413 view_37" [label="(12, 197, 64)", style=solid]; +"411 view_35" -> "414 scaled_dot_product_attention_4" [label="(1, 12, 197, 64)", style=solid]; +"412 view_36" -> "414 scaled_dot_product_attention_4" [label="(1, 12, 197, 64)", style=solid]; +"413 view_37" -> "414 scaled_dot_product_attention_4" [label="(1, 12, 197, 64)", style=solid]; +"414 scaled_dot_product_attention_4" -> "415 permute_5" [label="(1, 12, 197, 64)", style=solid]; +"415 permute_5" -> "416 view_38" [label="(197, 1, 12, 64)", style=solid]; +"416 view_38" -> "417 view_38_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"417 view_38_0_0_nncf_smooth_quant_0" -> "418 quantize_per_tensor_default_28" [label="(197, 768)", style=solid]; +"418 quantize_per_tensor_default_28" -> "419 dequantize_per_tensor_default_28" [label="(197, 768)", style=solid]; +"419 dequantize_per_tensor_default_28" -> "424 linear_17" [label="(197, 768)", style=solid]; +"420 linear_17_scale_0" -> "423 dequantize_per_channel_default_18" [label="(768,)", style=solid]; +"421 linear_17_zero_point_0" -> "423 dequantize_per_channel_default_18" [label="(768,)", style=solid]; +"422 _frozen_param19" -> "423 dequantize_per_channel_default_18" [label="(768, 768)", style=solid]; +"423 dequantize_per_channel_default_18" -> "424 linear_17" [label="(768, 768)", style=solid]; +"424 linear_17" -> "425 view_39" [label="(197, 768)", style=solid]; +"425 view_39" -> "426 transpose_29" [label="(197, 1, 768)", style=solid]; +"426 transpose_29" -> "427 dropout_13" [label="(1, 197, 768)", style=solid]; +"427 dropout_13" -> "428 add_9" [label="(1, 197, 768)", style=solid]; +"428 add_9" -> "429 layer_norm_9" [label="(1, 197, 768)", style=solid]; +"428 add_9" -> "449 add_10" [label="(1, 197, 768)", style=solid]; +"429 layer_norm_9" -> "430 layer_norm_9_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"430 layer_norm_9_0_0_nncf_smooth_quant_0" -> "431 quantize_per_tensor_default_29" [label="(1, 197, 768)", style=solid]; +"431 quantize_per_tensor_default_29" -> "432 dequantize_per_tensor_default_29" [label="(1, 197, 768)", style=solid]; +"432 dequantize_per_tensor_default_29" -> "437 linear_18" [label="(1, 197, 768)", style=solid]; +"433 linear_18_scale_0" -> "436 dequantize_per_channel_default_19" [label="(3072,)", style=solid]; +"434 linear_18_zero_point_0" -> "436 dequantize_per_channel_default_19" [label="(3072,)", style=solid]; +"435 _frozen_param20" -> "436 dequantize_per_channel_default_19" [label="(3072, 768)", style=solid]; +"436 dequantize_per_channel_default_19" -> "437 linear_18" [label="(3072, 768)", style=solid]; +"437 linear_18" -> "438 gelu_4" [label="(1, 197, 3072)", style=solid]; +"438 gelu_4" -> "439 dropout_14" [label="(1, 197, 3072)", style=solid]; +"439 dropout_14" -> "440 dropout_14_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"440 dropout_14_0_0_nncf_smooth_quant_0" -> "441 quantize_per_tensor_default_30" [label="(1, 197, 3072)", style=solid]; +"441 quantize_per_tensor_default_30" -> "442 dequantize_per_tensor_default_30" [label="(1, 197, 3072)", style=solid]; +"442 dequantize_per_tensor_default_30" -> "447 linear_19" [label="(1, 197, 3072)", style=solid]; +"443 linear_19_scale_0" -> "446 dequantize_per_channel_default_20" [label="(768,)", style=solid]; +"444 linear_19_zero_point_0" -> "446 dequantize_per_channel_default_20" [label="(768,)", style=solid]; +"445 _frozen_param21" -> "446 dequantize_per_channel_default_20" [label="(768, 3072)", style=solid]; +"446 dequantize_per_channel_default_20" -> "447 linear_19" [label="(768, 3072)", style=solid]; +"447 linear_19" -> "448 dropout_15" [label="(1, 197, 768)", style=solid]; +"448 dropout_15" -> "449 add_10" [label="(1, 197, 768)", style=solid]; +"449 add_10" -> "450 layer_norm_10" [label="(1, 197, 768)", style=solid]; +"449 add_10" -> "495 add_11" [label="(1, 197, 768)", style=solid]; +"450 layer_norm_10" -> "451 transpose_30" [label="(1, 197, 768)", style=solid]; +"451 transpose_30" -> "452 transpose_30_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"452 transpose_30_0_0_nncf_smooth_quant_0" -> "453 quantize_per_tensor_default_31" [label="(197, 1, 768)", style=solid]; +"453 quantize_per_tensor_default_31" -> "454 dequantize_per_tensor_default_31" [label="(197, 1, 768)", style=solid]; +"454 dequantize_per_tensor_default_31" -> "459 linear_20" [label="(197, 1, 768)", style=solid]; +"455 linear_20_scale_0" -> "458 dequantize_per_channel_default_21" [label="(2304,)", style=solid]; +"456 linear_20_zero_point_0" -> "458 dequantize_per_channel_default_21" [label="(2304,)", style=solid]; +"457 _frozen_param22" -> "458 dequantize_per_channel_default_21" [label="(2304, 768)", style=solid]; +"458 dequantize_per_channel_default_21" -> "459 linear_20" [label="(2304, 768)", style=solid]; +"459 linear_20" -> "460 unflatten_5" [label="(197, 1, 2304)", style=solid]; +"460 unflatten_5" -> "461 unsqueeze_5" [label="(197, 1, 3, 768)", style=solid]; +"461 unsqueeze_5" -> "462 transpose_31" [label="(1, 197, 1, 3, 768)", style=solid]; +"462 transpose_31" -> "463 squeeze_5" [label="(3, 197, 1, 1, 768)", style=solid]; +"463 squeeze_5" -> "464 contiguous_5" [label="(3, 197, 1, 768)", style=solid]; +"464 contiguous_5" -> "465 quantize_per_tensor_default_32" [label="(3, 197, 1, 768)", style=solid]; +"464 contiguous_5" -> "468 quantize_per_tensor_default_33" [label="(3, 197, 1, 768)", style=solid]; +"464 contiguous_5" -> "471 select_17" [label="(3, 197, 1, 768)", style=solid]; +"465 quantize_per_tensor_default_32" -> "466 dequantize_per_tensor_default_32" [label="(3, 197, 1, 768)", style=solid]; +"466 dequantize_per_tensor_default_32" -> "467 select_15" [label="(3, 197, 1, 768)", style=solid]; +"467 select_15" -> "472 view_40" [label="(197, 1, 768)", style=solid]; +"468 quantize_per_tensor_default_33" -> "469 dequantize_per_tensor_default_33" [label="(3, 197, 1, 768)", style=solid]; +"469 dequantize_per_tensor_default_33" -> "470 select_16" [label="(3, 197, 1, 768)", style=solid]; +"470 select_16" -> "474 view_41" [label="(197, 1, 768)", style=solid]; +"471 select_17" -> "476 view_42" [label="(197, 1, 768)", style=solid]; +"472 view_40" -> "473 transpose_32" [label="(197, 12, 64)", style=solid]; +"473 transpose_32" -> "478 view_43" [label="(12, 197, 64)", style=solid]; +"474 view_41" -> "475 transpose_33" [label="(197, 12, 64)", style=solid]; +"475 transpose_33" -> "479 view_44" [label="(12, 197, 64)", style=solid]; +"476 view_42" -> "477 transpose_34" [label="(197, 12, 64)", style=solid]; +"477 transpose_34" -> "480 view_45" [label="(12, 197, 64)", style=solid]; +"478 view_43" -> "481 scaled_dot_product_attention_5" [label="(1, 12, 197, 64)", style=solid]; +"479 view_44" -> "481 scaled_dot_product_attention_5" [label="(1, 12, 197, 64)", style=solid]; +"480 view_45" -> "481 scaled_dot_product_attention_5" [label="(1, 12, 197, 64)", style=solid]; +"481 scaled_dot_product_attention_5" -> "482 permute_6" [label="(1, 12, 197, 64)", style=solid]; +"482 permute_6" -> "483 view_46" [label="(197, 1, 12, 64)", style=solid]; +"483 view_46" -> "484 view_46_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"484 view_46_0_0_nncf_smooth_quant_0" -> "485 quantize_per_tensor_default_34" [label="(197, 768)", style=solid]; +"485 quantize_per_tensor_default_34" -> "486 dequantize_per_tensor_default_34" [label="(197, 768)", style=solid]; +"486 dequantize_per_tensor_default_34" -> "491 linear_21" [label="(197, 768)", style=solid]; +"487 linear_21_scale_0" -> "490 dequantize_per_channel_default_22" [label="(768,)", style=solid]; +"488 linear_21_zero_point_0" -> "490 dequantize_per_channel_default_22" [label="(768,)", style=solid]; +"489 _frozen_param23" -> "490 dequantize_per_channel_default_22" [label="(768, 768)", style=solid]; +"490 dequantize_per_channel_default_22" -> "491 linear_21" [label="(768, 768)", style=solid]; +"491 linear_21" -> "492 view_47" [label="(197, 768)", style=solid]; +"492 view_47" -> "493 transpose_35" [label="(197, 1, 768)", style=solid]; +"493 transpose_35" -> "494 dropout_16" [label="(1, 197, 768)", style=solid]; +"494 dropout_16" -> "495 add_11" [label="(1, 197, 768)", style=solid]; +"495 add_11" -> "496 layer_norm_11" [label="(1, 197, 768)", style=solid]; +"495 add_11" -> "516 add_12" [label="(1, 197, 768)", style=solid]; +"496 layer_norm_11" -> "497 layer_norm_11_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"497 layer_norm_11_0_0_nncf_smooth_quant_0" -> "498 quantize_per_tensor_default_35" [label="(1, 197, 768)", style=solid]; +"498 quantize_per_tensor_default_35" -> "499 dequantize_per_tensor_default_35" [label="(1, 197, 768)", style=solid]; +"499 dequantize_per_tensor_default_35" -> "504 linear_22" [label="(1, 197, 768)", style=solid]; +"500 linear_22_scale_0" -> "503 dequantize_per_channel_default_23" [label="(3072,)", style=solid]; +"501 linear_22_zero_point_0" -> "503 dequantize_per_channel_default_23" [label="(3072,)", style=solid]; +"502 _frozen_param24" -> "503 dequantize_per_channel_default_23" [label="(3072, 768)", style=solid]; +"503 dequantize_per_channel_default_23" -> "504 linear_22" [label="(3072, 768)", style=solid]; +"504 linear_22" -> "505 gelu_5" [label="(1, 197, 3072)", style=solid]; +"505 gelu_5" -> "506 dropout_17" [label="(1, 197, 3072)", style=solid]; +"506 dropout_17" -> "507 dropout_17_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"507 dropout_17_0_0_nncf_smooth_quant_0" -> "508 quantize_per_tensor_default_36" [label="(1, 197, 3072)", style=solid]; +"508 quantize_per_tensor_default_36" -> "509 dequantize_per_tensor_default_36" [label="(1, 197, 3072)", style=solid]; +"509 dequantize_per_tensor_default_36" -> "514 linear_23" [label="(1, 197, 3072)", style=solid]; +"510 linear_23_scale_0" -> "513 dequantize_per_channel_default_24" [label="(768,)", style=solid]; +"511 linear_23_zero_point_0" -> "513 dequantize_per_channel_default_24" [label="(768,)", style=solid]; +"512 _frozen_param25" -> "513 dequantize_per_channel_default_24" [label="(768, 3072)", style=solid]; +"513 dequantize_per_channel_default_24" -> "514 linear_23" [label="(768, 3072)", style=solid]; +"514 linear_23" -> "515 dropout_18" [label="(1, 197, 768)", style=solid]; +"515 dropout_18" -> "516 add_12" [label="(1, 197, 768)", style=solid]; +"516 add_12" -> "517 layer_norm_12" [label="(1, 197, 768)", style=solid]; +"516 add_12" -> "562 add_13" [label="(1, 197, 768)", style=solid]; +"517 layer_norm_12" -> "518 transpose_36" [label="(1, 197, 768)", style=solid]; +"518 transpose_36" -> "519 transpose_36_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"519 transpose_36_0_0_nncf_smooth_quant_0" -> "520 quantize_per_tensor_default_37" [label="(197, 1, 768)", style=solid]; +"520 quantize_per_tensor_default_37" -> "521 dequantize_per_tensor_default_37" [label="(197, 1, 768)", style=solid]; +"521 dequantize_per_tensor_default_37" -> "526 linear_24" [label="(197, 1, 768)", style=solid]; +"522 linear_24_scale_0" -> "525 dequantize_per_channel_default_25" [label="(2304,)", style=solid]; +"523 linear_24_zero_point_0" -> "525 dequantize_per_channel_default_25" [label="(2304,)", style=solid]; +"524 _frozen_param26" -> "525 dequantize_per_channel_default_25" [label="(2304, 768)", style=solid]; +"525 dequantize_per_channel_default_25" -> "526 linear_24" [label="(2304, 768)", style=solid]; +"526 linear_24" -> "527 unflatten_6" [label="(197, 1, 2304)", style=solid]; +"527 unflatten_6" -> "528 unsqueeze_6" [label="(197, 1, 3, 768)", style=solid]; +"528 unsqueeze_6" -> "529 transpose_37" [label="(1, 197, 1, 3, 768)", style=solid]; +"529 transpose_37" -> "530 squeeze_6" [label="(3, 197, 1, 1, 768)", style=solid]; +"530 squeeze_6" -> "531 contiguous_6" [label="(3, 197, 1, 768)", style=solid]; +"531 contiguous_6" -> "532 quantize_per_tensor_default_38" [label="(3, 197, 1, 768)", style=solid]; +"531 contiguous_6" -> "535 quantize_per_tensor_default_39" [label="(3, 197, 1, 768)", style=solid]; +"531 contiguous_6" -> "538 select_20" [label="(3, 197, 1, 768)", style=solid]; +"532 quantize_per_tensor_default_38" -> "533 dequantize_per_tensor_default_38" [label="(3, 197, 1, 768)", style=solid]; +"533 dequantize_per_tensor_default_38" -> "534 select_18" [label="(3, 197, 1, 768)", style=solid]; +"534 select_18" -> "539 view_48" [label="(197, 1, 768)", style=solid]; +"535 quantize_per_tensor_default_39" -> "536 dequantize_per_tensor_default_39" [label="(3, 197, 1, 768)", style=solid]; +"536 dequantize_per_tensor_default_39" -> "537 select_19" [label="(3, 197, 1, 768)", style=solid]; +"537 select_19" -> "541 view_49" [label="(197, 1, 768)", style=solid]; +"538 select_20" -> "543 view_50" [label="(197, 1, 768)", style=solid]; +"539 view_48" -> "540 transpose_38" [label="(197, 12, 64)", style=solid]; +"540 transpose_38" -> "545 view_51" [label="(12, 197, 64)", style=solid]; +"541 view_49" -> "542 transpose_39" [label="(197, 12, 64)", style=solid]; +"542 transpose_39" -> "546 view_52" [label="(12, 197, 64)", style=solid]; +"543 view_50" -> "544 transpose_40" [label="(197, 12, 64)", style=solid]; +"544 transpose_40" -> "547 view_53" [label="(12, 197, 64)", style=solid]; +"545 view_51" -> "548 scaled_dot_product_attention_6" [label="(1, 12, 197, 64)", style=solid]; +"546 view_52" -> "548 scaled_dot_product_attention_6" [label="(1, 12, 197, 64)", style=solid]; +"547 view_53" -> "548 scaled_dot_product_attention_6" [label="(1, 12, 197, 64)", style=solid]; +"548 scaled_dot_product_attention_6" -> "549 permute_7" [label="(1, 12, 197, 64)", style=solid]; +"549 permute_7" -> "550 view_54" [label="(197, 1, 12, 64)", style=solid]; +"550 view_54" -> "551 view_54_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"551 view_54_0_0_nncf_smooth_quant_0" -> "552 quantize_per_tensor_default_40" [label="(197, 768)", style=solid]; +"552 quantize_per_tensor_default_40" -> "553 dequantize_per_tensor_default_40" [label="(197, 768)", style=solid]; +"553 dequantize_per_tensor_default_40" -> "558 linear_25" [label="(197, 768)", style=solid]; +"554 linear_25_scale_0" -> "557 dequantize_per_channel_default_26" [label="(768,)", style=solid]; +"555 linear_25_zero_point_0" -> "557 dequantize_per_channel_default_26" [label="(768,)", style=solid]; +"556 _frozen_param27" -> "557 dequantize_per_channel_default_26" [label="(768, 768)", style=solid]; +"557 dequantize_per_channel_default_26" -> "558 linear_25" [label="(768, 768)", style=solid]; +"558 linear_25" -> "559 view_55" [label="(197, 768)", style=solid]; +"559 view_55" -> "560 transpose_41" [label="(197, 1, 768)", style=solid]; +"560 transpose_41" -> "561 dropout_19" [label="(1, 197, 768)", style=solid]; +"561 dropout_19" -> "562 add_13" [label="(1, 197, 768)", style=solid]; +"562 add_13" -> "563 layer_norm_13" [label="(1, 197, 768)", style=solid]; +"562 add_13" -> "583 add_14" [label="(1, 197, 768)", style=solid]; +"563 layer_norm_13" -> "564 layer_norm_13_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"564 layer_norm_13_0_0_nncf_smooth_quant_0" -> "565 quantize_per_tensor_default_41" [label="(1, 197, 768)", style=solid]; +"565 quantize_per_tensor_default_41" -> "566 dequantize_per_tensor_default_41" [label="(1, 197, 768)", style=solid]; +"566 dequantize_per_tensor_default_41" -> "571 linear_26" [label="(1, 197, 768)", style=solid]; +"567 linear_26_scale_0" -> "570 dequantize_per_channel_default_27" [label="(3072,)", style=solid]; +"568 linear_26_zero_point_0" -> "570 dequantize_per_channel_default_27" [label="(3072,)", style=solid]; +"569 _frozen_param28" -> "570 dequantize_per_channel_default_27" [label="(3072, 768)", style=solid]; +"570 dequantize_per_channel_default_27" -> "571 linear_26" [label="(3072, 768)", style=solid]; +"571 linear_26" -> "572 gelu_6" [label="(1, 197, 3072)", style=solid]; +"572 gelu_6" -> "573 dropout_20" [label="(1, 197, 3072)", style=solid]; +"573 dropout_20" -> "574 dropout_20_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"574 dropout_20_0_0_nncf_smooth_quant_0" -> "575 quantize_per_tensor_default_42" [label="(1, 197, 3072)", style=solid]; +"575 quantize_per_tensor_default_42" -> "576 dequantize_per_tensor_default_42" [label="(1, 197, 3072)", style=solid]; +"576 dequantize_per_tensor_default_42" -> "581 linear_27" [label="(1, 197, 3072)", style=solid]; +"577 linear_27_scale_0" -> "580 dequantize_per_channel_default_28" [label="(768,)", style=solid]; +"578 linear_27_zero_point_0" -> "580 dequantize_per_channel_default_28" [label="(768,)", style=solid]; +"579 _frozen_param29" -> "580 dequantize_per_channel_default_28" [label="(768, 3072)", style=solid]; +"580 dequantize_per_channel_default_28" -> "581 linear_27" [label="(768, 3072)", style=solid]; +"581 linear_27" -> "582 dropout_21" [label="(1, 197, 768)", style=solid]; +"582 dropout_21" -> "583 add_14" [label="(1, 197, 768)", style=solid]; +"583 add_14" -> "584 layer_norm_14" [label="(1, 197, 768)", style=solid]; +"583 add_14" -> "629 add_15" [label="(1, 197, 768)", style=solid]; +"584 layer_norm_14" -> "585 transpose_42" [label="(1, 197, 768)", style=solid]; +"585 transpose_42" -> "586 transpose_42_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"586 transpose_42_0_0_nncf_smooth_quant_0" -> "587 quantize_per_tensor_default_43" [label="(197, 1, 768)", style=solid]; +"587 quantize_per_tensor_default_43" -> "588 dequantize_per_tensor_default_43" [label="(197, 1, 768)", style=solid]; +"588 dequantize_per_tensor_default_43" -> "593 linear_28" [label="(197, 1, 768)", style=solid]; +"589 linear_28_scale_0" -> "592 dequantize_per_channel_default_29" [label="(2304,)", style=solid]; +"590 linear_28_zero_point_0" -> "592 dequantize_per_channel_default_29" [label="(2304,)", style=solid]; +"591 _frozen_param30" -> "592 dequantize_per_channel_default_29" [label="(2304, 768)", style=solid]; +"592 dequantize_per_channel_default_29" -> "593 linear_28" [label="(2304, 768)", style=solid]; +"593 linear_28" -> "594 unflatten_7" [label="(197, 1, 2304)", style=solid]; +"594 unflatten_7" -> "595 unsqueeze_7" [label="(197, 1, 3, 768)", style=solid]; +"595 unsqueeze_7" -> "596 transpose_43" [label="(1, 197, 1, 3, 768)", style=solid]; +"596 transpose_43" -> "597 squeeze_7" [label="(3, 197, 1, 1, 768)", style=solid]; +"597 squeeze_7" -> "598 contiguous_7" [label="(3, 197, 1, 768)", style=solid]; +"598 contiguous_7" -> "599 quantize_per_tensor_default_44" [label="(3, 197, 1, 768)", style=solid]; +"598 contiguous_7" -> "602 quantize_per_tensor_default_45" [label="(3, 197, 1, 768)", style=solid]; +"598 contiguous_7" -> "605 select_23" [label="(3, 197, 1, 768)", style=solid]; +"599 quantize_per_tensor_default_44" -> "600 dequantize_per_tensor_default_44" [label="(3, 197, 1, 768)", style=solid]; +"600 dequantize_per_tensor_default_44" -> "601 select_21" [label="(3, 197, 1, 768)", style=solid]; +"601 select_21" -> "606 view_56" [label="(197, 1, 768)", style=solid]; +"602 quantize_per_tensor_default_45" -> "603 dequantize_per_tensor_default_45" [label="(3, 197, 1, 768)", style=solid]; +"603 dequantize_per_tensor_default_45" -> "604 select_22" [label="(3, 197, 1, 768)", style=solid]; +"604 select_22" -> "608 view_57" [label="(197, 1, 768)", style=solid]; +"605 select_23" -> "610 view_58" [label="(197, 1, 768)", style=solid]; +"606 view_56" -> "607 transpose_44" [label="(197, 12, 64)", style=solid]; +"607 transpose_44" -> "612 view_59" [label="(12, 197, 64)", style=solid]; +"608 view_57" -> "609 transpose_45" [label="(197, 12, 64)", style=solid]; +"609 transpose_45" -> "613 view_60" [label="(12, 197, 64)", style=solid]; +"610 view_58" -> "611 transpose_46" [label="(197, 12, 64)", style=solid]; +"611 transpose_46" -> "614 view_61" [label="(12, 197, 64)", style=solid]; +"612 view_59" -> "615 scaled_dot_product_attention_7" [label="(1, 12, 197, 64)", style=solid]; +"613 view_60" -> "615 scaled_dot_product_attention_7" [label="(1, 12, 197, 64)", style=solid]; +"614 view_61" -> "615 scaled_dot_product_attention_7" [label="(1, 12, 197, 64)", style=solid]; +"615 scaled_dot_product_attention_7" -> "616 permute_8" [label="(1, 12, 197, 64)", style=solid]; +"616 permute_8" -> "617 view_62" [label="(197, 1, 12, 64)", style=solid]; +"617 view_62" -> "618 view_62_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"618 view_62_0_0_nncf_smooth_quant_0" -> "619 quantize_per_tensor_default_46" [label="(197, 768)", style=solid]; +"619 quantize_per_tensor_default_46" -> "620 dequantize_per_tensor_default_46" [label="(197, 768)", style=solid]; +"620 dequantize_per_tensor_default_46" -> "625 linear_29" [label="(197, 768)", style=solid]; +"621 linear_29_scale_0" -> "624 dequantize_per_channel_default_30" [label="(768,)", style=solid]; +"622 linear_29_zero_point_0" -> "624 dequantize_per_channel_default_30" [label="(768,)", style=solid]; +"623 _frozen_param31" -> "624 dequantize_per_channel_default_30" [label="(768, 768)", style=solid]; +"624 dequantize_per_channel_default_30" -> "625 linear_29" [label="(768, 768)", style=solid]; +"625 linear_29" -> "626 view_63" [label="(197, 768)", style=solid]; +"626 view_63" -> "627 transpose_47" [label="(197, 1, 768)", style=solid]; +"627 transpose_47" -> "628 dropout_22" [label="(1, 197, 768)", style=solid]; +"628 dropout_22" -> "629 add_15" [label="(1, 197, 768)", style=solid]; +"629 add_15" -> "630 layer_norm_15" [label="(1, 197, 768)", style=solid]; +"629 add_15" -> "650 add_16" [label="(1, 197, 768)", style=solid]; +"630 layer_norm_15" -> "631 layer_norm_15_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"631 layer_norm_15_0_0_nncf_smooth_quant_0" -> "632 quantize_per_tensor_default_47" [label="(1, 197, 768)", style=solid]; +"632 quantize_per_tensor_default_47" -> "633 dequantize_per_tensor_default_47" [label="(1, 197, 768)", style=solid]; +"633 dequantize_per_tensor_default_47" -> "638 linear_30" [label="(1, 197, 768)", style=solid]; +"634 linear_30_scale_0" -> "637 dequantize_per_channel_default_31" [label="(3072,)", style=solid]; +"635 linear_30_zero_point_0" -> "637 dequantize_per_channel_default_31" [label="(3072,)", style=solid]; +"636 _frozen_param32" -> "637 dequantize_per_channel_default_31" [label="(3072, 768)", style=solid]; +"637 dequantize_per_channel_default_31" -> "638 linear_30" [label="(3072, 768)", style=solid]; +"638 linear_30" -> "639 gelu_7" [label="(1, 197, 3072)", style=solid]; +"639 gelu_7" -> "640 dropout_23" [label="(1, 197, 3072)", style=solid]; +"640 dropout_23" -> "641 dropout_23_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"641 dropout_23_0_0_nncf_smooth_quant_0" -> "642 quantize_per_tensor_default_48" [label="(1, 197, 3072)", style=solid]; +"642 quantize_per_tensor_default_48" -> "643 dequantize_per_tensor_default_48" [label="(1, 197, 3072)", style=solid]; +"643 dequantize_per_tensor_default_48" -> "648 linear_31" [label="(1, 197, 3072)", style=solid]; +"644 linear_31_scale_0" -> "647 dequantize_per_channel_default_32" [label="(768,)", style=solid]; +"645 linear_31_zero_point_0" -> "647 dequantize_per_channel_default_32" [label="(768,)", style=solid]; +"646 _frozen_param33" -> "647 dequantize_per_channel_default_32" [label="(768, 3072)", style=solid]; +"647 dequantize_per_channel_default_32" -> "648 linear_31" [label="(768, 3072)", style=solid]; +"648 linear_31" -> "649 dropout_24" [label="(1, 197, 768)", style=solid]; +"649 dropout_24" -> "650 add_16" [label="(1, 197, 768)", style=solid]; +"650 add_16" -> "651 layer_norm_16" [label="(1, 197, 768)", style=solid]; +"650 add_16" -> "696 add_17" [label="(1, 197, 768)", style=solid]; +"651 layer_norm_16" -> "652 transpose_48" [label="(1, 197, 768)", style=solid]; +"652 transpose_48" -> "653 transpose_48_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"653 transpose_48_0_0_nncf_smooth_quant_0" -> "654 quantize_per_tensor_default_49" [label="(197, 1, 768)", style=solid]; +"654 quantize_per_tensor_default_49" -> "655 dequantize_per_tensor_default_49" [label="(197, 1, 768)", style=solid]; +"655 dequantize_per_tensor_default_49" -> "660 linear_32" [label="(197, 1, 768)", style=solid]; +"656 linear_32_scale_0" -> "659 dequantize_per_channel_default_33" [label="(2304,)", style=solid]; +"657 linear_32_zero_point_0" -> "659 dequantize_per_channel_default_33" [label="(2304,)", style=solid]; +"658 _frozen_param34" -> "659 dequantize_per_channel_default_33" [label="(2304, 768)", style=solid]; +"659 dequantize_per_channel_default_33" -> "660 linear_32" [label="(2304, 768)", style=solid]; +"660 linear_32" -> "661 unflatten_8" [label="(197, 1, 2304)", style=solid]; +"661 unflatten_8" -> "662 unsqueeze_8" [label="(197, 1, 3, 768)", style=solid]; +"662 unsqueeze_8" -> "663 transpose_49" [label="(1, 197, 1, 3, 768)", style=solid]; +"663 transpose_49" -> "664 squeeze_8" [label="(3, 197, 1, 1, 768)", style=solid]; +"664 squeeze_8" -> "665 contiguous_8" [label="(3, 197, 1, 768)", style=solid]; +"665 contiguous_8" -> "666 quantize_per_tensor_default_50" [label="(3, 197, 1, 768)", style=solid]; +"665 contiguous_8" -> "669 quantize_per_tensor_default_51" [label="(3, 197, 1, 768)", style=solid]; +"665 contiguous_8" -> "672 select_26" [label="(3, 197, 1, 768)", style=solid]; +"666 quantize_per_tensor_default_50" -> "667 dequantize_per_tensor_default_50" [label="(3, 197, 1, 768)", style=solid]; +"667 dequantize_per_tensor_default_50" -> "668 select_24" [label="(3, 197, 1, 768)", style=solid]; +"668 select_24" -> "673 view_64" [label="(197, 1, 768)", style=solid]; +"669 quantize_per_tensor_default_51" -> "670 dequantize_per_tensor_default_51" [label="(3, 197, 1, 768)", style=solid]; +"670 dequantize_per_tensor_default_51" -> "671 select_25" [label="(3, 197, 1, 768)", style=solid]; +"671 select_25" -> "675 view_65" [label="(197, 1, 768)", style=solid]; +"672 select_26" -> "677 view_66" [label="(197, 1, 768)", style=solid]; +"673 view_64" -> "674 transpose_50" [label="(197, 12, 64)", style=solid]; +"674 transpose_50" -> "679 view_67" [label="(12, 197, 64)", style=solid]; +"675 view_65" -> "676 transpose_51" [label="(197, 12, 64)", style=solid]; +"676 transpose_51" -> "680 view_68" [label="(12, 197, 64)", style=solid]; +"677 view_66" -> "678 transpose_52" [label="(197, 12, 64)", style=solid]; +"678 transpose_52" -> "681 view_69" [label="(12, 197, 64)", style=solid]; +"679 view_67" -> "682 scaled_dot_product_attention_8" [label="(1, 12, 197, 64)", style=solid]; +"680 view_68" -> "682 scaled_dot_product_attention_8" [label="(1, 12, 197, 64)", style=solid]; +"681 view_69" -> "682 scaled_dot_product_attention_8" [label="(1, 12, 197, 64)", style=solid]; +"682 scaled_dot_product_attention_8" -> "683 permute_9" [label="(1, 12, 197, 64)", style=solid]; +"683 permute_9" -> "684 view_70" [label="(197, 1, 12, 64)", style=solid]; +"684 view_70" -> "685 view_70_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"685 view_70_0_0_nncf_smooth_quant_0" -> "686 quantize_per_tensor_default_52" [label="(197, 768)", style=solid]; +"686 quantize_per_tensor_default_52" -> "687 dequantize_per_tensor_default_52" [label="(197, 768)", style=solid]; +"687 dequantize_per_tensor_default_52" -> "692 linear_33" [label="(197, 768)", style=solid]; +"688 linear_33_scale_0" -> "691 dequantize_per_channel_default_34" [label="(768,)", style=solid]; +"689 linear_33_zero_point_0" -> "691 dequantize_per_channel_default_34" [label="(768,)", style=solid]; +"690 _frozen_param35" -> "691 dequantize_per_channel_default_34" [label="(768, 768)", style=solid]; +"691 dequantize_per_channel_default_34" -> "692 linear_33" [label="(768, 768)", style=solid]; +"692 linear_33" -> "693 view_71" [label="(197, 768)", style=solid]; +"693 view_71" -> "694 transpose_53" [label="(197, 1, 768)", style=solid]; +"694 transpose_53" -> "695 dropout_25" [label="(1, 197, 768)", style=solid]; +"695 dropout_25" -> "696 add_17" [label="(1, 197, 768)", style=solid]; +"696 add_17" -> "697 layer_norm_17" [label="(1, 197, 768)", style=solid]; +"696 add_17" -> "717 add_18" [label="(1, 197, 768)", style=solid]; +"697 layer_norm_17" -> "698 layer_norm_17_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"698 layer_norm_17_0_0_nncf_smooth_quant_0" -> "699 quantize_per_tensor_default_53" [label="(1, 197, 768)", style=solid]; +"699 quantize_per_tensor_default_53" -> "700 dequantize_per_tensor_default_53" [label="(1, 197, 768)", style=solid]; +"700 dequantize_per_tensor_default_53" -> "705 linear_34" [label="(1, 197, 768)", style=solid]; +"701 linear_34_scale_0" -> "704 dequantize_per_channel_default_35" [label="(3072,)", style=solid]; +"702 linear_34_zero_point_0" -> "704 dequantize_per_channel_default_35" [label="(3072,)", style=solid]; +"703 _frozen_param36" -> "704 dequantize_per_channel_default_35" [label="(3072, 768)", style=solid]; +"704 dequantize_per_channel_default_35" -> "705 linear_34" [label="(3072, 768)", style=solid]; +"705 linear_34" -> "706 gelu_8" [label="(1, 197, 3072)", style=solid]; +"706 gelu_8" -> "707 dropout_26" [label="(1, 197, 3072)", style=solid]; +"707 dropout_26" -> "708 dropout_26_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"708 dropout_26_0_0_nncf_smooth_quant_0" -> "709 quantize_per_tensor_default_54" [label="(1, 197, 3072)", style=solid]; +"709 quantize_per_tensor_default_54" -> "710 dequantize_per_tensor_default_54" [label="(1, 197, 3072)", style=solid]; +"710 dequantize_per_tensor_default_54" -> "715 linear_35" [label="(1, 197, 3072)", style=solid]; +"711 linear_35_scale_0" -> "714 dequantize_per_channel_default_36" [label="(768,)", style=solid]; +"712 linear_35_zero_point_0" -> "714 dequantize_per_channel_default_36" [label="(768,)", style=solid]; +"713 _frozen_param37" -> "714 dequantize_per_channel_default_36" [label="(768, 3072)", style=solid]; +"714 dequantize_per_channel_default_36" -> "715 linear_35" [label="(768, 3072)", style=solid]; +"715 linear_35" -> "716 dropout_27" [label="(1, 197, 768)", style=solid]; +"716 dropout_27" -> "717 add_18" [label="(1, 197, 768)", style=solid]; +"717 add_18" -> "718 layer_norm_18" [label="(1, 197, 768)", style=solid]; +"717 add_18" -> "763 add_19" [label="(1, 197, 768)", style=solid]; +"718 layer_norm_18" -> "719 transpose_54" [label="(1, 197, 768)", style=solid]; +"719 transpose_54" -> "720 transpose_54_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"720 transpose_54_0_0_nncf_smooth_quant_0" -> "721 quantize_per_tensor_default_55" [label="(197, 1, 768)", style=solid]; +"721 quantize_per_tensor_default_55" -> "722 dequantize_per_tensor_default_55" [label="(197, 1, 768)", style=solid]; +"722 dequantize_per_tensor_default_55" -> "727 linear_36" [label="(197, 1, 768)", style=solid]; +"723 linear_36_scale_0" -> "726 dequantize_per_channel_default_37" [label="(2304,)", style=solid]; +"724 linear_36_zero_point_0" -> "726 dequantize_per_channel_default_37" [label="(2304,)", style=solid]; +"725 _frozen_param38" -> "726 dequantize_per_channel_default_37" [label="(2304, 768)", style=solid]; +"726 dequantize_per_channel_default_37" -> "727 linear_36" [label="(2304, 768)", style=solid]; +"727 linear_36" -> "728 unflatten_9" [label="(197, 1, 2304)", style=solid]; +"728 unflatten_9" -> "729 unsqueeze_9" [label="(197, 1, 3, 768)", style=solid]; +"729 unsqueeze_9" -> "730 transpose_55" [label="(1, 197, 1, 3, 768)", style=solid]; +"730 transpose_55" -> "731 squeeze_9" [label="(3, 197, 1, 1, 768)", style=solid]; +"731 squeeze_9" -> "732 contiguous_9" [label="(3, 197, 1, 768)", style=solid]; +"732 contiguous_9" -> "733 quantize_per_tensor_default_56" [label="(3, 197, 1, 768)", style=solid]; +"732 contiguous_9" -> "736 quantize_per_tensor_default_57" [label="(3, 197, 1, 768)", style=solid]; +"732 contiguous_9" -> "739 select_29" [label="(3, 197, 1, 768)", style=solid]; +"733 quantize_per_tensor_default_56" -> "734 dequantize_per_tensor_default_56" [label="(3, 197, 1, 768)", style=solid]; +"734 dequantize_per_tensor_default_56" -> "735 select_27" [label="(3, 197, 1, 768)", style=solid]; +"735 select_27" -> "740 view_72" [label="(197, 1, 768)", style=solid]; +"736 quantize_per_tensor_default_57" -> "737 dequantize_per_tensor_default_57" [label="(3, 197, 1, 768)", style=solid]; +"737 dequantize_per_tensor_default_57" -> "738 select_28" [label="(3, 197, 1, 768)", style=solid]; +"738 select_28" -> "742 view_73" [label="(197, 1, 768)", style=solid]; +"739 select_29" -> "744 view_74" [label="(197, 1, 768)", style=solid]; +"740 view_72" -> "741 transpose_56" [label="(197, 12, 64)", style=solid]; +"741 transpose_56" -> "746 view_75" [label="(12, 197, 64)", style=solid]; +"742 view_73" -> "743 transpose_57" [label="(197, 12, 64)", style=solid]; +"743 transpose_57" -> "747 view_76" [label="(12, 197, 64)", style=solid]; +"744 view_74" -> "745 transpose_58" [label="(197, 12, 64)", style=solid]; +"745 transpose_58" -> "748 view_77" [label="(12, 197, 64)", style=solid]; +"746 view_75" -> "749 scaled_dot_product_attention_9" [label="(1, 12, 197, 64)", style=solid]; +"747 view_76" -> "749 scaled_dot_product_attention_9" [label="(1, 12, 197, 64)", style=solid]; +"748 view_77" -> "749 scaled_dot_product_attention_9" [label="(1, 12, 197, 64)", style=solid]; +"749 scaled_dot_product_attention_9" -> "750 permute_10" [label="(1, 12, 197, 64)", style=solid]; +"750 permute_10" -> "751 view_78" [label="(197, 1, 12, 64)", style=solid]; +"751 view_78" -> "752 view_78_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"752 view_78_0_0_nncf_smooth_quant_0" -> "753 quantize_per_tensor_default_58" [label="(197, 768)", style=solid]; +"753 quantize_per_tensor_default_58" -> "754 dequantize_per_tensor_default_58" [label="(197, 768)", style=solid]; +"754 dequantize_per_tensor_default_58" -> "759 linear_37" [label="(197, 768)", style=solid]; +"755 linear_37_scale_0" -> "758 dequantize_per_channel_default_38" [label="(768,)", style=solid]; +"756 linear_37_zero_point_0" -> "758 dequantize_per_channel_default_38" [label="(768,)", style=solid]; +"757 _frozen_param39" -> "758 dequantize_per_channel_default_38" [label="(768, 768)", style=solid]; +"758 dequantize_per_channel_default_38" -> "759 linear_37" [label="(768, 768)", style=solid]; +"759 linear_37" -> "760 view_79" [label="(197, 768)", style=solid]; +"760 view_79" -> "761 transpose_59" [label="(197, 1, 768)", style=solid]; +"761 transpose_59" -> "762 dropout_28" [label="(1, 197, 768)", style=solid]; +"762 dropout_28" -> "763 add_19" [label="(1, 197, 768)", style=solid]; +"763 add_19" -> "764 layer_norm_19" [label="(1, 197, 768)", style=solid]; +"763 add_19" -> "784 add_20" [label="(1, 197, 768)", style=solid]; +"764 layer_norm_19" -> "765 layer_norm_19_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"765 layer_norm_19_0_0_nncf_smooth_quant_0" -> "766 quantize_per_tensor_default_59" [label="(1, 197, 768)", style=solid]; +"766 quantize_per_tensor_default_59" -> "767 dequantize_per_tensor_default_59" [label="(1, 197, 768)", style=solid]; +"767 dequantize_per_tensor_default_59" -> "772 linear_38" [label="(1, 197, 768)", style=solid]; +"768 linear_38_scale_0" -> "771 dequantize_per_channel_default_39" [label="(3072,)", style=solid]; +"769 linear_38_zero_point_0" -> "771 dequantize_per_channel_default_39" [label="(3072,)", style=solid]; +"770 _frozen_param40" -> "771 dequantize_per_channel_default_39" [label="(3072, 768)", style=solid]; +"771 dequantize_per_channel_default_39" -> "772 linear_38" [label="(3072, 768)", style=solid]; +"772 linear_38" -> "773 gelu_9" [label="(1, 197, 3072)", style=solid]; +"773 gelu_9" -> "774 dropout_29" [label="(1, 197, 3072)", style=solid]; +"774 dropout_29" -> "775 dropout_29_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"775 dropout_29_0_0_nncf_smooth_quant_0" -> "776 quantize_per_tensor_default_60" [label="(1, 197, 3072)", style=solid]; +"776 quantize_per_tensor_default_60" -> "777 dequantize_per_tensor_default_60" [label="(1, 197, 3072)", style=solid]; +"777 dequantize_per_tensor_default_60" -> "782 linear_39" [label="(1, 197, 3072)", style=solid]; +"778 linear_39_scale_0" -> "781 dequantize_per_channel_default_40" [label="(768,)", style=solid]; +"779 linear_39_zero_point_0" -> "781 dequantize_per_channel_default_40" [label="(768,)", style=solid]; +"780 _frozen_param41" -> "781 dequantize_per_channel_default_40" [label="(768, 3072)", style=solid]; +"781 dequantize_per_channel_default_40" -> "782 linear_39" [label="(768, 3072)", style=solid]; +"782 linear_39" -> "783 dropout_30" [label="(1, 197, 768)", style=solid]; +"783 dropout_30" -> "784 add_20" [label="(1, 197, 768)", style=solid]; +"784 add_20" -> "785 layer_norm_20" [label="(1, 197, 768)", style=solid]; +"784 add_20" -> "830 add_21" [label="(1, 197, 768)", style=solid]; +"785 layer_norm_20" -> "786 transpose_60" [label="(1, 197, 768)", style=solid]; +"786 transpose_60" -> "787 transpose_60_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"787 transpose_60_0_0_nncf_smooth_quant_0" -> "788 quantize_per_tensor_default_61" [label="(197, 1, 768)", style=solid]; +"788 quantize_per_tensor_default_61" -> "789 dequantize_per_tensor_default_61" [label="(197, 1, 768)", style=solid]; +"789 dequantize_per_tensor_default_61" -> "794 linear_40" [label="(197, 1, 768)", style=solid]; +"790 linear_40_scale_0" -> "793 dequantize_per_channel_default_41" [label="(2304,)", style=solid]; +"791 linear_40_zero_point_0" -> "793 dequantize_per_channel_default_41" [label="(2304,)", style=solid]; +"792 _frozen_param42" -> "793 dequantize_per_channel_default_41" [label="(2304, 768)", style=solid]; +"793 dequantize_per_channel_default_41" -> "794 linear_40" [label="(2304, 768)", style=solid]; +"794 linear_40" -> "795 unflatten_10" [label="(197, 1, 2304)", style=solid]; +"795 unflatten_10" -> "796 unsqueeze_10" [label="(197, 1, 3, 768)", style=solid]; +"796 unsqueeze_10" -> "797 transpose_61" [label="(1, 197, 1, 3, 768)", style=solid]; +"797 transpose_61" -> "798 squeeze_10" [label="(3, 197, 1, 1, 768)", style=solid]; +"798 squeeze_10" -> "799 contiguous_10" [label="(3, 197, 1, 768)", style=solid]; +"799 contiguous_10" -> "800 quantize_per_tensor_default_62" [label="(3, 197, 1, 768)", style=solid]; +"799 contiguous_10" -> "803 quantize_per_tensor_default_63" [label="(3, 197, 1, 768)", style=solid]; +"799 contiguous_10" -> "806 select_32" [label="(3, 197, 1, 768)", style=solid]; +"800 quantize_per_tensor_default_62" -> "801 dequantize_per_tensor_default_62" [label="(3, 197, 1, 768)", style=solid]; +"801 dequantize_per_tensor_default_62" -> "802 select_30" [label="(3, 197, 1, 768)", style=solid]; +"802 select_30" -> "807 view_80" [label="(197, 1, 768)", style=solid]; +"803 quantize_per_tensor_default_63" -> "804 dequantize_per_tensor_default_63" [label="(3, 197, 1, 768)", style=solid]; +"804 dequantize_per_tensor_default_63" -> "805 select_31" [label="(3, 197, 1, 768)", style=solid]; +"805 select_31" -> "809 view_81" [label="(197, 1, 768)", style=solid]; +"806 select_32" -> "811 view_82" [label="(197, 1, 768)", style=solid]; +"807 view_80" -> "808 transpose_62" [label="(197, 12, 64)", style=solid]; +"808 transpose_62" -> "813 view_83" [label="(12, 197, 64)", style=solid]; +"809 view_81" -> "810 transpose_63" [label="(197, 12, 64)", style=solid]; +"810 transpose_63" -> "814 view_84" [label="(12, 197, 64)", style=solid]; +"811 view_82" -> "812 transpose_64" [label="(197, 12, 64)", style=solid]; +"812 transpose_64" -> "815 view_85" [label="(12, 197, 64)", style=solid]; +"813 view_83" -> "816 scaled_dot_product_attention_10" [label="(1, 12, 197, 64)", style=solid]; +"814 view_84" -> "816 scaled_dot_product_attention_10" [label="(1, 12, 197, 64)", style=solid]; +"815 view_85" -> "816 scaled_dot_product_attention_10" [label="(1, 12, 197, 64)", style=solid]; +"816 scaled_dot_product_attention_10" -> "817 permute_11" [label="(1, 12, 197, 64)", style=solid]; +"817 permute_11" -> "818 view_86" [label="(197, 1, 12, 64)", style=solid]; +"818 view_86" -> "819 view_86_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"819 view_86_0_0_nncf_smooth_quant_0" -> "820 quantize_per_tensor_default_64" [label="(197, 768)", style=solid]; +"820 quantize_per_tensor_default_64" -> "821 dequantize_per_tensor_default_64" [label="(197, 768)", style=solid]; +"821 dequantize_per_tensor_default_64" -> "826 linear_41" [label="(197, 768)", style=solid]; +"822 linear_41_scale_0" -> "825 dequantize_per_channel_default_42" [label="(768,)", style=solid]; +"823 linear_41_zero_point_0" -> "825 dequantize_per_channel_default_42" [label="(768,)", style=solid]; +"824 _frozen_param43" -> "825 dequantize_per_channel_default_42" [label="(768, 768)", style=solid]; +"825 dequantize_per_channel_default_42" -> "826 linear_41" [label="(768, 768)", style=solid]; +"826 linear_41" -> "827 view_87" [label="(197, 768)", style=solid]; +"827 view_87" -> "828 transpose_65" [label="(197, 1, 768)", style=solid]; +"828 transpose_65" -> "829 dropout_31" [label="(1, 197, 768)", style=solid]; +"829 dropout_31" -> "830 add_21" [label="(1, 197, 768)", style=solid]; +"830 add_21" -> "831 layer_norm_21" [label="(1, 197, 768)", style=solid]; +"830 add_21" -> "851 add_22" [label="(1, 197, 768)", style=solid]; +"831 layer_norm_21" -> "832 layer_norm_21_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"832 layer_norm_21_0_0_nncf_smooth_quant_0" -> "833 quantize_per_tensor_default_65" [label="(1, 197, 768)", style=solid]; +"833 quantize_per_tensor_default_65" -> "834 dequantize_per_tensor_default_65" [label="(1, 197, 768)", style=solid]; +"834 dequantize_per_tensor_default_65" -> "839 linear_42" [label="(1, 197, 768)", style=solid]; +"835 linear_42_scale_0" -> "838 dequantize_per_channel_default_43" [label="(3072,)", style=solid]; +"836 linear_42_zero_point_0" -> "838 dequantize_per_channel_default_43" [label="(3072,)", style=solid]; +"837 _frozen_param44" -> "838 dequantize_per_channel_default_43" [label="(3072, 768)", style=solid]; +"838 dequantize_per_channel_default_43" -> "839 linear_42" [label="(3072, 768)", style=solid]; +"839 linear_42" -> "840 gelu_10" [label="(1, 197, 3072)", style=solid]; +"840 gelu_10" -> "841 dropout_32" [label="(1, 197, 3072)", style=solid]; +"841 dropout_32" -> "842 dropout_32_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"842 dropout_32_0_0_nncf_smooth_quant_0" -> "843 quantize_per_tensor_default_66" [label="(1, 197, 3072)", style=solid]; +"843 quantize_per_tensor_default_66" -> "844 dequantize_per_tensor_default_66" [label="(1, 197, 3072)", style=solid]; +"844 dequantize_per_tensor_default_66" -> "849 linear_43" [label="(1, 197, 3072)", style=solid]; +"845 linear_43_scale_0" -> "848 dequantize_per_channel_default_44" [label="(768,)", style=solid]; +"846 linear_43_zero_point_0" -> "848 dequantize_per_channel_default_44" [label="(768,)", style=solid]; +"847 _frozen_param45" -> "848 dequantize_per_channel_default_44" [label="(768, 3072)", style=solid]; +"848 dequantize_per_channel_default_44" -> "849 linear_43" [label="(768, 3072)", style=solid]; +"849 linear_43" -> "850 dropout_33" [label="(1, 197, 768)", style=solid]; +"850 dropout_33" -> "851 add_22" [label="(1, 197, 768)", style=solid]; +"851 add_22" -> "852 layer_norm_22" [label="(1, 197, 768)", style=solid]; +"851 add_22" -> "897 add_23" [label="(1, 197, 768)", style=solid]; +"852 layer_norm_22" -> "853 transpose_66" [label="(1, 197, 768)", style=solid]; +"853 transpose_66" -> "854 transpose_66_0_0_nncf_smooth_quant_0" [label="(197, 1, 768)", style=solid]; +"854 transpose_66_0_0_nncf_smooth_quant_0" -> "855 quantize_per_tensor_default_67" [label="(197, 1, 768)", style=solid]; +"855 quantize_per_tensor_default_67" -> "856 dequantize_per_tensor_default_67" [label="(197, 1, 768)", style=solid]; +"856 dequantize_per_tensor_default_67" -> "861 linear_44" [label="(197, 1, 768)", style=solid]; +"857 linear_44_scale_0" -> "860 dequantize_per_channel_default_45" [label="(2304,)", style=solid]; +"858 linear_44_zero_point_0" -> "860 dequantize_per_channel_default_45" [label="(2304,)", style=solid]; +"859 _frozen_param46" -> "860 dequantize_per_channel_default_45" [label="(2304, 768)", style=solid]; +"860 dequantize_per_channel_default_45" -> "861 linear_44" [label="(2304, 768)", style=solid]; +"861 linear_44" -> "862 unflatten_11" [label="(197, 1, 2304)", style=solid]; +"862 unflatten_11" -> "863 unsqueeze_11" [label="(197, 1, 3, 768)", style=solid]; +"863 unsqueeze_11" -> "864 transpose_67" [label="(1, 197, 1, 3, 768)", style=solid]; +"864 transpose_67" -> "865 squeeze_11" [label="(3, 197, 1, 1, 768)", style=solid]; +"865 squeeze_11" -> "866 contiguous_11" [label="(3, 197, 1, 768)", style=solid]; +"866 contiguous_11" -> "867 quantize_per_tensor_default_68" [label="(3, 197, 1, 768)", style=solid]; +"866 contiguous_11" -> "870 quantize_per_tensor_default_69" [label="(3, 197, 1, 768)", style=solid]; +"866 contiguous_11" -> "873 select_35" [label="(3, 197, 1, 768)", style=solid]; +"867 quantize_per_tensor_default_68" -> "868 dequantize_per_tensor_default_68" [label="(3, 197, 1, 768)", style=solid]; +"868 dequantize_per_tensor_default_68" -> "869 select_33" [label="(3, 197, 1, 768)", style=solid]; +"869 select_33" -> "874 view_88" [label="(197, 1, 768)", style=solid]; +"870 quantize_per_tensor_default_69" -> "871 dequantize_per_tensor_default_69" [label="(3, 197, 1, 768)", style=solid]; +"871 dequantize_per_tensor_default_69" -> "872 select_34" [label="(3, 197, 1, 768)", style=solid]; +"872 select_34" -> "876 view_89" [label="(197, 1, 768)", style=solid]; +"873 select_35" -> "878 view_90" [label="(197, 1, 768)", style=solid]; +"874 view_88" -> "875 transpose_68" [label="(197, 12, 64)", style=solid]; +"875 transpose_68" -> "880 view_91" [label="(12, 197, 64)", style=solid]; +"876 view_89" -> "877 transpose_69" [label="(197, 12, 64)", style=solid]; +"877 transpose_69" -> "881 view_92" [label="(12, 197, 64)", style=solid]; +"878 view_90" -> "879 transpose_70" [label="(197, 12, 64)", style=solid]; +"879 transpose_70" -> "882 view_93" [label="(12, 197, 64)", style=solid]; +"880 view_91" -> "883 scaled_dot_product_attention_11" [label="(1, 12, 197, 64)", style=solid]; +"881 view_92" -> "883 scaled_dot_product_attention_11" [label="(1, 12, 197, 64)", style=solid]; +"882 view_93" -> "883 scaled_dot_product_attention_11" [label="(1, 12, 197, 64)", style=solid]; +"883 scaled_dot_product_attention_11" -> "884 permute_12" [label="(1, 12, 197, 64)", style=solid]; +"884 permute_12" -> "885 view_94" [label="(197, 1, 12, 64)", style=solid]; +"885 view_94" -> "886 view_94_0_0_nncf_smooth_quant_0" [label="(197, 768)", style=solid]; +"886 view_94_0_0_nncf_smooth_quant_0" -> "887 quantize_per_tensor_default_70" [label="(197, 768)", style=solid]; +"887 quantize_per_tensor_default_70" -> "888 dequantize_per_tensor_default_70" [label="(197, 768)", style=solid]; +"888 dequantize_per_tensor_default_70" -> "893 linear_45" [label="(197, 768)", style=solid]; +"889 linear_45_scale_0" -> "892 dequantize_per_channel_default_46" [label="(768,)", style=solid]; +"890 linear_45_zero_point_0" -> "892 dequantize_per_channel_default_46" [label="(768,)", style=solid]; +"891 _frozen_param47" -> "892 dequantize_per_channel_default_46" [label="(768, 768)", style=solid]; +"892 dequantize_per_channel_default_46" -> "893 linear_45" [label="(768, 768)", style=solid]; +"893 linear_45" -> "894 view_95" [label="(197, 768)", style=solid]; +"894 view_95" -> "895 transpose_71" [label="(197, 1, 768)", style=solid]; +"895 transpose_71" -> "896 dropout_34" [label="(1, 197, 768)", style=solid]; +"896 dropout_34" -> "897 add_23" [label="(1, 197, 768)", style=solid]; +"897 add_23" -> "898 layer_norm_23" [label="(1, 197, 768)", style=solid]; +"897 add_23" -> "918 add_24" [label="(1, 197, 768)", style=solid]; +"898 layer_norm_23" -> "899 layer_norm_23_0_0_nncf_smooth_quant_0" [label="(1, 197, 768)", style=solid]; +"899 layer_norm_23_0_0_nncf_smooth_quant_0" -> "900 quantize_per_tensor_default_71" [label="(1, 197, 768)", style=solid]; +"900 quantize_per_tensor_default_71" -> "901 dequantize_per_tensor_default_71" [label="(1, 197, 768)", style=solid]; +"901 dequantize_per_tensor_default_71" -> "906 linear_46" [label="(1, 197, 768)", style=solid]; +"902 linear_46_scale_0" -> "905 dequantize_per_channel_default_47" [label="(3072,)", style=solid]; +"903 linear_46_zero_point_0" -> "905 dequantize_per_channel_default_47" [label="(3072,)", style=solid]; +"904 _frozen_param48" -> "905 dequantize_per_channel_default_47" [label="(3072, 768)", style=solid]; +"905 dequantize_per_channel_default_47" -> "906 linear_46" [label="(3072, 768)", style=solid]; +"906 linear_46" -> "907 gelu_11" [label="(1, 197, 3072)", style=solid]; +"907 gelu_11" -> "908 dropout_35" [label="(1, 197, 3072)", style=solid]; +"908 dropout_35" -> "909 dropout_35_0_0_nncf_smooth_quant_0" [label="(1, 197, 3072)", style=solid]; +"909 dropout_35_0_0_nncf_smooth_quant_0" -> "910 quantize_per_tensor_default_72" [label="(1, 197, 3072)", style=solid]; +"910 quantize_per_tensor_default_72" -> "911 dequantize_per_tensor_default_72" [label="(1, 197, 3072)", style=solid]; +"911 dequantize_per_tensor_default_72" -> "916 linear_47" [label="(1, 197, 3072)", style=solid]; +"912 linear_47_scale_0" -> "915 dequantize_per_channel_default_48" [label="(768,)", style=solid]; +"913 linear_47_zero_point_0" -> "915 dequantize_per_channel_default_48" [label="(768,)", style=solid]; +"914 _frozen_param49" -> "915 dequantize_per_channel_default_48" [label="(768, 3072)", style=solid]; +"915 dequantize_per_channel_default_48" -> "916 linear_47" [label="(768, 3072)", style=solid]; +"916 linear_47" -> "917 dropout_36" [label="(1, 197, 768)", style=solid]; +"917 dropout_36" -> "918 add_24" [label="(1, 197, 768)", style=solid]; +"918 add_24" -> "919 layer_norm_24" [label="(1, 197, 768)", style=solid]; +"919 layer_norm_24" -> "920 slice_1" [label="(1, 197, 768)", style=solid]; +"920 slice_1" -> "921 select_36" [label="(1, 197, 768)", style=solid]; +"921 select_36" -> "922 select_36_0_0_nncf_smooth_quant_0" [label="(1, 768)", style=solid]; +"922 select_36_0_0_nncf_smooth_quant_0" -> "923 quantize_per_tensor_default_73" [label="(1, 768)", style=solid]; +"923 quantize_per_tensor_default_73" -> "924 dequantize_per_tensor_default_73" [label="(1, 768)", style=solid]; +"924 dequantize_per_tensor_default_73" -> "929 linear_48" [label="(1, 768)", style=solid]; +"925 linear_48_scale_0" -> "928 dequantize_per_channel_default_49" [label="(1000,)", style=solid]; +"926 linear_48_zero_point_0" -> "928 dequantize_per_channel_default_49" [label="(1000,)", style=solid]; +"927 _frozen_param50" -> "928 dequantize_per_channel_default_49" [label="(1000, 768)", style=solid]; +"928 dequantize_per_channel_default_49" -> "929 linear_48" [label="(1000, 768)", style=solid]; +"929 linear_48" -> "930 output" [label="(1, 1000)", style=solid]; } diff --git a/tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/swin_v2_s.dot b/tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/swin_v2_s.dot index a3e4f7384fa..ed9adcf0eaf 100644 --- a/tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/swin_v2_s.dot +++ b/tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/swin_v2_s.dot @@ -1,7100 +1,4862 @@ strict digraph { "0 _scale_0" [id=0, type=get_attr]; "1 _zero_point_0" [id=1, type=get_attr]; -"2 _frozen_param0" [id=2, type=get_attr]; +"2 _frozen_param2_1" [id=2, type=get_attr]; "3 dequantize_per_channel_default" [id=3, type=dequantize_per_channel]; "4 _scale_1" [id=4, type=get_attr]; "5 _zero_point_1" [id=5, type=get_attr]; -"6 _frozen_param1" [id=6, type=get_attr]; +"6 _frozen_param6_1" [id=6, type=get_attr]; "7 dequantize_per_channel_default_1" [id=7, type=dequantize_per_channel]; -"8 features_1_0_attn_qkv_bias" [id=8, type=get_attr]; -"9 features_1_0_attn_proj_bias" [id=9, type=get_attr]; -"10 features_1_0_attn_logit_scale" [id=10, type=get_attr]; -"11 _scale_2" [id=11, type=get_attr]; -"12 _zero_point_2" [id=12, type=get_attr]; -"13 _frozen_param2" [id=13, type=get_attr]; -"14 dequantize_per_channel_default_2" [id=14, type=dequantize_per_channel]; -"15 _scale_3" [id=15, type=get_attr]; -"16 _zero_point_3" [id=16, type=get_attr]; -"17 _frozen_param3" [id=17, type=get_attr]; -"18 dequantize_per_channel_default_3" [id=18, type=dequantize_per_channel]; -"19 features_1_1_attn_qkv_bias" [id=19, type=get_attr]; -"20 features_1_1_attn_proj_bias" [id=20, type=get_attr]; -"21 features_1_1_attn_logit_scale" [id=21, type=get_attr]; -"22 _scale_4" [id=22, type=get_attr]; -"23 _zero_point_4" [id=23, type=get_attr]; -"24 _frozen_param4" [id=24, type=get_attr]; -"25 dequantize_per_channel_default_4" [id=25, type=dequantize_per_channel]; -"26 _scale_5" [id=26, type=get_attr]; -"27 _zero_point_5" [id=27, type=get_attr]; -"28 _frozen_param5" [id=28, type=get_attr]; -"29 dequantize_per_channel_default_5" [id=29, type=dequantize_per_channel]; -"30 features_3_0_attn_qkv_bias" [id=30, type=get_attr]; -"31 features_3_0_attn_proj_bias" [id=31, type=get_attr]; -"32 features_3_0_attn_logit_scale" [id=32, type=get_attr]; -"33 _scale_6" [id=33, type=get_attr]; -"34 _zero_point_6" [id=34, type=get_attr]; -"35 _frozen_param6" [id=35, type=get_attr]; -"36 dequantize_per_channel_default_6" [id=36, type=dequantize_per_channel]; -"37 _scale_7" [id=37, type=get_attr]; -"38 _zero_point_7" [id=38, type=get_attr]; -"39 _frozen_param7" [id=39, type=get_attr]; -"40 dequantize_per_channel_default_7" [id=40, type=dequantize_per_channel]; -"41 features_3_1_attn_qkv_bias" [id=41, type=get_attr]; -"42 features_3_1_attn_proj_bias" [id=42, type=get_attr]; -"43 features_3_1_attn_logit_scale" [id=43, type=get_attr]; -"44 _scale_8" [id=44, type=get_attr]; -"45 _zero_point_8" [id=45, type=get_attr]; -"46 _frozen_param8" [id=46, type=get_attr]; -"47 dequantize_per_channel_default_8" [id=47, type=dequantize_per_channel]; -"48 _scale_9" [id=48, type=get_attr]; -"49 _zero_point_9" [id=49, type=get_attr]; -"50 _frozen_param9" [id=50, type=get_attr]; -"51 dequantize_per_channel_default_9" [id=51, type=dequantize_per_channel]; -"52 features_5_0_attn_qkv_bias" [id=52, type=get_attr]; -"53 features_5_0_attn_proj_bias" [id=53, type=get_attr]; -"54 features_5_0_attn_logit_scale" [id=54, type=get_attr]; -"55 _scale_10" [id=55, type=get_attr]; -"56 _zero_point_10" [id=56, type=get_attr]; -"57 _frozen_param10" [id=57, type=get_attr]; -"58 dequantize_per_channel_default_10" [id=58, type=dequantize_per_channel]; -"59 _scale_11" [id=59, type=get_attr]; -"60 _zero_point_11" [id=60, type=get_attr]; -"61 _frozen_param11" [id=61, type=get_attr]; -"62 dequantize_per_channel_default_11" [id=62, type=dequantize_per_channel]; -"63 features_5_1_attn_qkv_bias" [id=63, type=get_attr]; -"64 features_5_1_attn_proj_bias" [id=64, type=get_attr]; -"65 features_5_1_attn_logit_scale" [id=65, type=get_attr]; -"66 _scale_12" [id=66, type=get_attr]; -"67 _zero_point_12" [id=67, type=get_attr]; -"68 _frozen_param12" [id=68, type=get_attr]; -"69 dequantize_per_channel_default_12" [id=69, type=dequantize_per_channel]; -"70 _scale_13" [id=70, type=get_attr]; -"71 _zero_point_13" [id=71, type=get_attr]; -"72 _frozen_param13" [id=72, type=get_attr]; -"73 dequantize_per_channel_default_13" [id=73, type=dequantize_per_channel]; -"74 features_5_2_attn_qkv_bias" [id=74, type=get_attr]; -"75 features_5_2_attn_proj_bias" [id=75, type=get_attr]; -"76 features_5_2_attn_logit_scale" [id=76, type=get_attr]; -"77 _scale_14" [id=77, type=get_attr]; -"78 _zero_point_14" [id=78, type=get_attr]; -"79 _frozen_param14" [id=79, type=get_attr]; -"80 dequantize_per_channel_default_14" [id=80, type=dequantize_per_channel]; -"81 _scale_15" [id=81, type=get_attr]; -"82 _zero_point_15" [id=82, type=get_attr]; -"83 _frozen_param15" [id=83, type=get_attr]; -"84 dequantize_per_channel_default_15" [id=84, type=dequantize_per_channel]; -"85 features_5_3_attn_qkv_bias" [id=85, type=get_attr]; -"86 features_5_3_attn_proj_bias" [id=86, type=get_attr]; -"87 features_5_3_attn_logit_scale" [id=87, type=get_attr]; -"88 _scale_16" [id=88, type=get_attr]; -"89 _zero_point_16" [id=89, type=get_attr]; -"90 _frozen_param16" [id=90, type=get_attr]; -"91 dequantize_per_channel_default_16" [id=91, type=dequantize_per_channel]; -"92 _scale_17" [id=92, type=get_attr]; -"93 _zero_point_17" [id=93, type=get_attr]; -"94 _frozen_param17" [id=94, type=get_attr]; -"95 dequantize_per_channel_default_17" [id=95, type=dequantize_per_channel]; -"96 features_5_4_attn_qkv_bias" [id=96, type=get_attr]; -"97 features_5_4_attn_proj_bias" [id=97, type=get_attr]; -"98 features_5_4_attn_logit_scale" [id=98, type=get_attr]; -"99 _scale_18" [id=99, type=get_attr]; -"100 _zero_point_18" [id=100, type=get_attr]; -"101 _frozen_param18" [id=101, type=get_attr]; -"102 dequantize_per_channel_default_18" [id=102, type=dequantize_per_channel]; -"103 _scale_19" [id=103, type=get_attr]; -"104 _zero_point_19" [id=104, type=get_attr]; -"105 _frozen_param19" [id=105, type=get_attr]; -"106 dequantize_per_channel_default_19" [id=106, type=dequantize_per_channel]; -"107 features_5_5_attn_qkv_bias" [id=107, type=get_attr]; -"108 features_5_5_attn_proj_bias" [id=108, type=get_attr]; -"109 features_5_5_attn_logit_scale" [id=109, type=get_attr]; -"110 _scale_20" [id=110, type=get_attr]; -"111 _zero_point_20" [id=111, type=get_attr]; -"112 _frozen_param20" [id=112, type=get_attr]; -"113 dequantize_per_channel_default_20" [id=113, type=dequantize_per_channel]; -"114 _scale_21" [id=114, type=get_attr]; -"115 _zero_point_21" [id=115, type=get_attr]; -"116 _frozen_param21" [id=116, type=get_attr]; -"117 dequantize_per_channel_default_21" [id=117, type=dequantize_per_channel]; -"118 features_5_6_attn_qkv_bias" [id=118, type=get_attr]; -"119 features_5_6_attn_proj_bias" [id=119, type=get_attr]; -"120 features_5_6_attn_logit_scale" [id=120, type=get_attr]; -"121 _scale_22" [id=121, type=get_attr]; -"122 _zero_point_22" [id=122, type=get_attr]; -"123 _frozen_param22" [id=123, type=get_attr]; -"124 dequantize_per_channel_default_22" [id=124, type=dequantize_per_channel]; -"125 _scale_23" [id=125, type=get_attr]; -"126 _zero_point_23" [id=126, type=get_attr]; -"127 _frozen_param23" [id=127, type=get_attr]; -"128 dequantize_per_channel_default_23" [id=128, type=dequantize_per_channel]; -"129 features_5_7_attn_qkv_bias" [id=129, type=get_attr]; -"130 features_5_7_attn_proj_bias" [id=130, type=get_attr]; -"131 features_5_7_attn_logit_scale" [id=131, type=get_attr]; -"132 _scale_24" [id=132, type=get_attr]; -"133 _zero_point_24" [id=133, type=get_attr]; -"134 _frozen_param24" [id=134, type=get_attr]; -"135 dequantize_per_channel_default_24" [id=135, type=dequantize_per_channel]; -"136 _scale_25" [id=136, type=get_attr]; -"137 _zero_point_25" [id=137, type=get_attr]; -"138 _frozen_param25" [id=138, type=get_attr]; -"139 dequantize_per_channel_default_25" [id=139, type=dequantize_per_channel]; -"140 features_5_8_attn_qkv_bias" [id=140, type=get_attr]; -"141 features_5_8_attn_proj_bias" [id=141, type=get_attr]; -"142 features_5_8_attn_logit_scale" [id=142, type=get_attr]; -"143 _scale_26" [id=143, type=get_attr]; -"144 _zero_point_26" [id=144, type=get_attr]; -"145 _frozen_param26" [id=145, type=get_attr]; -"146 dequantize_per_channel_default_26" [id=146, type=dequantize_per_channel]; -"147 _scale_27" [id=147, type=get_attr]; -"148 _zero_point_27" [id=148, type=get_attr]; -"149 _frozen_param27" [id=149, type=get_attr]; -"150 dequantize_per_channel_default_27" [id=150, type=dequantize_per_channel]; -"151 features_5_9_attn_qkv_bias" [id=151, type=get_attr]; -"152 features_5_9_attn_proj_bias" [id=152, type=get_attr]; -"153 features_5_9_attn_logit_scale" [id=153, type=get_attr]; -"154 _scale_28" [id=154, type=get_attr]; -"155 _zero_point_28" [id=155, type=get_attr]; -"156 _frozen_param28" [id=156, type=get_attr]; -"157 dequantize_per_channel_default_28" [id=157, type=dequantize_per_channel]; -"158 _scale_29" [id=158, type=get_attr]; -"159 _zero_point_29" [id=159, type=get_attr]; -"160 _frozen_param29" [id=160, type=get_attr]; -"161 dequantize_per_channel_default_29" [id=161, type=dequantize_per_channel]; -"162 features_5_10_attn_qkv_bias" [id=162, type=get_attr]; -"163 features_5_10_attn_proj_bias" [id=163, type=get_attr]; -"164 features_5_10_attn_logit_scale" [id=164, type=get_attr]; -"165 _scale_30" [id=165, type=get_attr]; -"166 _zero_point_30" [id=166, type=get_attr]; -"167 _frozen_param30" [id=167, type=get_attr]; -"168 dequantize_per_channel_default_30" [id=168, type=dequantize_per_channel]; -"169 _scale_31" [id=169, type=get_attr]; -"170 _zero_point_31" [id=170, type=get_attr]; -"171 _frozen_param31" [id=171, type=get_attr]; -"172 dequantize_per_channel_default_31" [id=172, type=dequantize_per_channel]; -"173 features_5_11_attn_qkv_bias" [id=173, type=get_attr]; -"174 features_5_11_attn_proj_bias" [id=174, type=get_attr]; -"175 features_5_11_attn_logit_scale" [id=175, type=get_attr]; -"176 _scale_32" [id=176, type=get_attr]; -"177 _zero_point_32" [id=177, type=get_attr]; -"178 _frozen_param32" [id=178, type=get_attr]; -"179 dequantize_per_channel_default_32" [id=179, type=dequantize_per_channel]; -"180 _scale_33" [id=180, type=get_attr]; -"181 _zero_point_33" [id=181, type=get_attr]; -"182 _frozen_param33" [id=182, type=get_attr]; -"183 dequantize_per_channel_default_33" [id=183, type=dequantize_per_channel]; -"184 features_5_12_attn_qkv_bias" [id=184, type=get_attr]; -"185 features_5_12_attn_proj_bias" [id=185, type=get_attr]; -"186 features_5_12_attn_logit_scale" [id=186, type=get_attr]; -"187 _scale_34" [id=187, type=get_attr]; -"188 _zero_point_34" [id=188, type=get_attr]; -"189 _frozen_param34" [id=189, type=get_attr]; -"190 dequantize_per_channel_default_34" [id=190, type=dequantize_per_channel]; -"191 _scale_35" [id=191, type=get_attr]; -"192 _zero_point_35" [id=192, type=get_attr]; -"193 _frozen_param35" [id=193, type=get_attr]; -"194 dequantize_per_channel_default_35" [id=194, type=dequantize_per_channel]; -"195 features_5_13_attn_qkv_bias" [id=195, type=get_attr]; -"196 features_5_13_attn_proj_bias" [id=196, type=get_attr]; -"197 features_5_13_attn_logit_scale" [id=197, type=get_attr]; -"198 _scale_36" [id=198, type=get_attr]; -"199 _zero_point_36" [id=199, type=get_attr]; -"200 _frozen_param36" [id=200, type=get_attr]; -"201 dequantize_per_channel_default_36" [id=201, type=dequantize_per_channel]; -"202 _scale_37" [id=202, type=get_attr]; -"203 _zero_point_37" [id=203, type=get_attr]; -"204 _frozen_param37" [id=204, type=get_attr]; -"205 dequantize_per_channel_default_37" [id=205, type=dequantize_per_channel]; -"206 features_5_14_attn_qkv_bias" [id=206, type=get_attr]; -"207 features_5_14_attn_proj_bias" [id=207, type=get_attr]; -"208 features_5_14_attn_logit_scale" [id=208, type=get_attr]; -"209 _scale_38" [id=209, type=get_attr]; -"210 _zero_point_38" [id=210, type=get_attr]; -"211 _frozen_param38" [id=211, type=get_attr]; -"212 dequantize_per_channel_default_38" [id=212, type=dequantize_per_channel]; -"213 _scale_39" [id=213, type=get_attr]; -"214 _zero_point_39" [id=214, type=get_attr]; -"215 _frozen_param39" [id=215, type=get_attr]; -"216 dequantize_per_channel_default_39" [id=216, type=dequantize_per_channel]; -"217 features_5_15_attn_qkv_bias" [id=217, type=get_attr]; -"218 features_5_15_attn_proj_bias" [id=218, type=get_attr]; -"219 features_5_15_attn_logit_scale" [id=219, type=get_attr]; -"220 _scale_40" [id=220, type=get_attr]; -"221 _zero_point_40" [id=221, type=get_attr]; -"222 _frozen_param40" [id=222, type=get_attr]; -"223 dequantize_per_channel_default_40" [id=223, type=dequantize_per_channel]; -"224 _scale_41" [id=224, type=get_attr]; -"225 _zero_point_41" [id=225, type=get_attr]; -"226 _frozen_param41" [id=226, type=get_attr]; -"227 dequantize_per_channel_default_41" [id=227, type=dequantize_per_channel]; -"228 features_5_16_attn_qkv_bias" [id=228, type=get_attr]; -"229 features_5_16_attn_proj_bias" [id=229, type=get_attr]; -"230 features_5_16_attn_logit_scale" [id=230, type=get_attr]; -"231 _scale_42" [id=231, type=get_attr]; -"232 _zero_point_42" [id=232, type=get_attr]; -"233 _frozen_param42" [id=233, type=get_attr]; -"234 dequantize_per_channel_default_42" [id=234, type=dequantize_per_channel]; -"235 _scale_43" [id=235, type=get_attr]; -"236 _zero_point_43" [id=236, type=get_attr]; -"237 _frozen_param43" [id=237, type=get_attr]; -"238 dequantize_per_channel_default_43" [id=238, type=dequantize_per_channel]; -"239 features_5_17_attn_qkv_bias" [id=239, type=get_attr]; -"240 features_5_17_attn_proj_bias" [id=240, type=get_attr]; -"241 features_5_17_attn_logit_scale" [id=241, type=get_attr]; -"242 _scale_44" [id=242, type=get_attr]; -"243 _zero_point_44" [id=243, type=get_attr]; -"244 _frozen_param44" [id=244, type=get_attr]; -"245 dequantize_per_channel_default_44" [id=245, type=dequantize_per_channel]; -"246 _scale_45" [id=246, type=get_attr]; -"247 _zero_point_45" [id=247, type=get_attr]; -"248 _frozen_param45" [id=248, type=get_attr]; -"249 dequantize_per_channel_default_45" [id=249, type=dequantize_per_channel]; -"250 features_7_0_attn_qkv_bias" [id=250, type=get_attr]; -"251 features_7_0_attn_proj_bias" [id=251, type=get_attr]; -"252 features_7_0_attn_logit_scale" [id=252, type=get_attr]; -"253 _scale_46" [id=253, type=get_attr]; -"254 _zero_point_46" [id=254, type=get_attr]; -"255 _frozen_param46" [id=255, type=get_attr]; -"256 dequantize_per_channel_default_46" [id=256, type=dequantize_per_channel]; -"257 _scale_47" [id=257, type=get_attr]; -"258 _zero_point_47" [id=258, type=get_attr]; -"259 _frozen_param47" [id=259, type=get_attr]; -"260 dequantize_per_channel_default_47" [id=260, type=dequantize_per_channel]; -"261 features_7_1_attn_qkv_bias" [id=261, type=get_attr]; -"262 features_7_1_attn_proj_bias" [id=262, type=get_attr]; -"263 features_7_1_attn_logit_scale" [id=263, type=get_attr]; -"264 _scale_48" [id=264, type=get_attr]; -"265 _zero_point_48" [id=265, type=get_attr]; -"266 _frozen_param48" [id=266, type=get_attr]; -"267 dequantize_per_channel_default_48" [id=267, type=dequantize_per_channel]; -"268 features_0_0_bias" [id=268, type=get_attr]; -"269 features_0_2_weight" [id=269, type=get_attr]; -"270 features_0_2_bias" [id=270, type=get_attr]; -"271 _scale_49" [id=271, type=get_attr]; -"272 _zero_point_49" [id=272, type=get_attr]; -"273 _frozen_param49" [id=273, type=get_attr]; -"274 dequantize_per_channel_default_49" [id=274, type=dequantize_per_channel]; -"275 features_1_0_attn_cpb_mlp_0_bias" [id=275, type=get_attr]; -"276 _scale_50" [id=276, type=get_attr]; -"277 _zero_point_50" [id=277, type=get_attr]; -"278 _frozen_param50" [id=278, type=get_attr]; -"279 dequantize_per_channel_default_50" [id=279, type=dequantize_per_channel]; -"280 features_1_0_norm1_weight" [id=280, type=get_attr]; -"281 features_1_0_norm1_bias" [id=281, type=get_attr]; -"282 _scale_51" [id=282, type=get_attr]; -"283 _zero_point_51" [id=283, type=get_attr]; -"284 _frozen_param51" [id=284, type=get_attr]; -"285 dequantize_per_channel_default_51" [id=285, type=dequantize_per_channel]; -"286 features_1_0_mlp_0_bias" [id=286, type=get_attr]; -"287 _scale_52" [id=287, type=get_attr]; -"288 _zero_point_52" [id=288, type=get_attr]; -"289 _frozen_param52" [id=289, type=get_attr]; -"290 dequantize_per_channel_default_52" [id=290, type=dequantize_per_channel]; -"291 features_1_0_mlp_3_bias" [id=291, type=get_attr]; -"292 features_1_0_norm2_weight" [id=292, type=get_attr]; -"293 features_1_0_norm2_bias" [id=293, type=get_attr]; -"294 _scale_53" [id=294, type=get_attr]; -"295 _zero_point_53" [id=295, type=get_attr]; -"296 _frozen_param53" [id=296, type=get_attr]; -"297 dequantize_per_channel_default_53" [id=297, type=dequantize_per_channel]; -"298 features_1_1_attn_cpb_mlp_0_bias" [id=298, type=get_attr]; -"299 _scale_54" [id=299, type=get_attr]; -"300 _zero_point_54" [id=300, type=get_attr]; -"301 _frozen_param54" [id=301, type=get_attr]; -"302 dequantize_per_channel_default_54" [id=302, type=dequantize_per_channel]; -"303 features_1_1_norm1_weight" [id=303, type=get_attr]; -"304 features_1_1_norm1_bias" [id=304, type=get_attr]; -"305 _scale_55" [id=305, type=get_attr]; -"306 _zero_point_55" [id=306, type=get_attr]; -"307 _frozen_param55" [id=307, type=get_attr]; -"308 dequantize_per_channel_default_55" [id=308, type=dequantize_per_channel]; -"309 features_1_1_mlp_0_bias" [id=309, type=get_attr]; -"310 _scale_56" [id=310, type=get_attr]; -"311 _zero_point_56" [id=311, type=get_attr]; -"312 _frozen_param56" [id=312, type=get_attr]; -"313 dequantize_per_channel_default_56" [id=313, type=dequantize_per_channel]; -"314 features_1_1_mlp_3_bias" [id=314, type=get_attr]; -"315 features_1_1_norm2_weight" [id=315, type=get_attr]; -"316 features_1_1_norm2_bias" [id=316, type=get_attr]; -"317 _scale_57" [id=317, type=get_attr]; -"318 _zero_point_57" [id=318, type=get_attr]; -"319 _frozen_param57" [id=319, type=get_attr]; -"320 dequantize_per_channel_default_57" [id=320, type=dequantize_per_channel]; -"321 features_2_norm_weight" [id=321, type=get_attr]; -"322 features_2_norm_bias" [id=322, type=get_attr]; -"323 _scale_58" [id=323, type=get_attr]; -"324 _zero_point_58" [id=324, type=get_attr]; -"325 _frozen_param58" [id=325, type=get_attr]; -"326 dequantize_per_channel_default_58" [id=326, type=dequantize_per_channel]; -"327 features_3_0_attn_cpb_mlp_0_bias" [id=327, type=get_attr]; -"328 _scale_59" [id=328, type=get_attr]; -"329 _zero_point_59" [id=329, type=get_attr]; -"330 _frozen_param59" [id=330, type=get_attr]; -"331 dequantize_per_channel_default_59" [id=331, type=dequantize_per_channel]; -"332 features_3_0_norm1_weight" [id=332, type=get_attr]; -"333 features_3_0_norm1_bias" [id=333, type=get_attr]; -"334 _scale_60" [id=334, type=get_attr]; -"335 _zero_point_60" [id=335, type=get_attr]; -"336 _frozen_param60" [id=336, type=get_attr]; -"337 dequantize_per_channel_default_60" [id=337, type=dequantize_per_channel]; -"338 features_3_0_mlp_0_bias" [id=338, type=get_attr]; -"339 _scale_61" [id=339, type=get_attr]; -"340 _zero_point_61" [id=340, type=get_attr]; -"341 _frozen_param61" [id=341, type=get_attr]; -"342 dequantize_per_channel_default_61" [id=342, type=dequantize_per_channel]; -"343 features_3_0_mlp_3_bias" [id=343, type=get_attr]; -"344 features_3_0_norm2_weight" [id=344, type=get_attr]; -"345 features_3_0_norm2_bias" [id=345, type=get_attr]; -"346 _scale_62" [id=346, type=get_attr]; -"347 _zero_point_62" [id=347, type=get_attr]; -"348 _frozen_param62" [id=348, type=get_attr]; -"349 dequantize_per_channel_default_62" [id=349, type=dequantize_per_channel]; -"350 features_3_1_attn_cpb_mlp_0_bias" [id=350, type=get_attr]; -"351 _scale_63" [id=351, type=get_attr]; -"352 _zero_point_63" [id=352, type=get_attr]; -"353 _frozen_param63" [id=353, type=get_attr]; -"354 dequantize_per_channel_default_63" [id=354, type=dequantize_per_channel]; -"355 features_3_1_norm1_weight" [id=355, type=get_attr]; -"356 features_3_1_norm1_bias" [id=356, type=get_attr]; -"357 _scale_64" [id=357, type=get_attr]; -"358 _zero_point_64" [id=358, type=get_attr]; -"359 _frozen_param64" [id=359, type=get_attr]; -"360 dequantize_per_channel_default_64" [id=360, type=dequantize_per_channel]; -"361 features_3_1_mlp_0_bias" [id=361, type=get_attr]; -"362 _scale_65" [id=362, type=get_attr]; -"363 _zero_point_65" [id=363, type=get_attr]; -"364 _frozen_param65" [id=364, type=get_attr]; -"365 dequantize_per_channel_default_65" [id=365, type=dequantize_per_channel]; -"366 features_3_1_mlp_3_bias" [id=366, type=get_attr]; -"367 features_3_1_norm2_weight" [id=367, type=get_attr]; -"368 features_3_1_norm2_bias" [id=368, type=get_attr]; -"369 _scale_66" [id=369, type=get_attr]; -"370 _zero_point_66" [id=370, type=get_attr]; -"371 _frozen_param66" [id=371, type=get_attr]; -"372 dequantize_per_channel_default_66" [id=372, type=dequantize_per_channel]; -"373 features_4_norm_weight" [id=373, type=get_attr]; -"374 features_4_norm_bias" [id=374, type=get_attr]; -"375 _scale_67" [id=375, type=get_attr]; -"376 _zero_point_67" [id=376, type=get_attr]; -"377 _frozen_param67" [id=377, type=get_attr]; -"378 dequantize_per_channel_default_67" [id=378, type=dequantize_per_channel]; -"379 features_5_0_attn_cpb_mlp_0_bias" [id=379, type=get_attr]; -"380 _scale_68" [id=380, type=get_attr]; -"381 _zero_point_68" [id=381, type=get_attr]; -"382 _frozen_param68" [id=382, type=get_attr]; -"383 dequantize_per_channel_default_68" [id=383, type=dequantize_per_channel]; -"384 features_5_0_norm1_weight" [id=384, type=get_attr]; -"385 features_5_0_norm1_bias" [id=385, type=get_attr]; -"386 _scale_69" [id=386, type=get_attr]; -"387 _zero_point_69" [id=387, type=get_attr]; -"388 _frozen_param69" [id=388, type=get_attr]; -"389 dequantize_per_channel_default_69" [id=389, type=dequantize_per_channel]; -"390 features_5_0_mlp_0_bias" [id=390, type=get_attr]; -"391 _scale_70" [id=391, type=get_attr]; -"392 _zero_point_70" [id=392, type=get_attr]; -"393 _frozen_param70" [id=393, type=get_attr]; -"394 dequantize_per_channel_default_70" [id=394, type=dequantize_per_channel]; -"395 features_5_0_mlp_3_bias" [id=395, type=get_attr]; -"396 features_5_0_norm2_weight" [id=396, type=get_attr]; -"397 features_5_0_norm2_bias" [id=397, type=get_attr]; -"398 _scale_71" [id=398, type=get_attr]; -"399 _zero_point_71" [id=399, type=get_attr]; -"400 _frozen_param71" [id=400, type=get_attr]; -"401 dequantize_per_channel_default_71" [id=401, type=dequantize_per_channel]; -"402 features_5_1_attn_cpb_mlp_0_bias" [id=402, type=get_attr]; -"403 _scale_72" [id=403, type=get_attr]; -"404 _zero_point_72" [id=404, type=get_attr]; -"405 _frozen_param72" [id=405, type=get_attr]; -"406 dequantize_per_channel_default_72" [id=406, type=dequantize_per_channel]; -"407 features_5_1_norm1_weight" [id=407, type=get_attr]; -"408 features_5_1_norm1_bias" [id=408, type=get_attr]; -"409 _scale_73" [id=409, type=get_attr]; -"410 _zero_point_73" [id=410, type=get_attr]; -"411 _frozen_param73" [id=411, type=get_attr]; -"412 dequantize_per_channel_default_73" [id=412, type=dequantize_per_channel]; -"413 features_5_1_mlp_0_bias" [id=413, type=get_attr]; -"414 _scale_74" [id=414, type=get_attr]; -"415 _zero_point_74" [id=415, type=get_attr]; -"416 _frozen_param74" [id=416, type=get_attr]; -"417 dequantize_per_channel_default_74" [id=417, type=dequantize_per_channel]; -"418 features_5_1_mlp_3_bias" [id=418, type=get_attr]; -"419 features_5_1_norm2_weight" [id=419, type=get_attr]; -"420 features_5_1_norm2_bias" [id=420, type=get_attr]; -"421 _scale_75" [id=421, type=get_attr]; -"422 _zero_point_75" [id=422, type=get_attr]; -"423 _frozen_param75" [id=423, type=get_attr]; -"424 dequantize_per_channel_default_75" [id=424, type=dequantize_per_channel]; -"425 features_5_2_attn_cpb_mlp_0_bias" [id=425, type=get_attr]; -"426 _scale_76" [id=426, type=get_attr]; -"427 _zero_point_76" [id=427, type=get_attr]; -"428 _frozen_param76" [id=428, type=get_attr]; -"429 dequantize_per_channel_default_76" [id=429, type=dequantize_per_channel]; -"430 features_5_2_norm1_weight" [id=430, type=get_attr]; -"431 features_5_2_norm1_bias" [id=431, type=get_attr]; -"432 _scale_77" [id=432, type=get_attr]; -"433 _zero_point_77" [id=433, type=get_attr]; -"434 _frozen_param77" [id=434, type=get_attr]; -"435 dequantize_per_channel_default_77" [id=435, type=dequantize_per_channel]; -"436 features_5_2_mlp_0_bias" [id=436, type=get_attr]; -"437 _scale_78" [id=437, type=get_attr]; -"438 _zero_point_78" [id=438, type=get_attr]; -"439 _frozen_param78" [id=439, type=get_attr]; -"440 dequantize_per_channel_default_78" [id=440, type=dequantize_per_channel]; -"441 features_5_2_mlp_3_bias" [id=441, type=get_attr]; -"442 features_5_2_norm2_weight" [id=442, type=get_attr]; -"443 features_5_2_norm2_bias" [id=443, type=get_attr]; -"444 _scale_79" [id=444, type=get_attr]; -"445 _zero_point_79" [id=445, type=get_attr]; -"446 _frozen_param79" [id=446, type=get_attr]; -"447 dequantize_per_channel_default_79" [id=447, type=dequantize_per_channel]; -"448 features_5_3_attn_cpb_mlp_0_bias" [id=448, type=get_attr]; -"449 _scale_80" [id=449, type=get_attr]; -"450 _zero_point_80" [id=450, type=get_attr]; -"451 _frozen_param80" [id=451, type=get_attr]; -"452 dequantize_per_channel_default_80" [id=452, type=dequantize_per_channel]; -"453 features_5_3_norm1_weight" [id=453, type=get_attr]; -"454 features_5_3_norm1_bias" [id=454, type=get_attr]; -"455 _scale_81" [id=455, type=get_attr]; -"456 _zero_point_81" [id=456, type=get_attr]; -"457 _frozen_param81" [id=457, type=get_attr]; -"458 dequantize_per_channel_default_81" [id=458, type=dequantize_per_channel]; -"459 features_5_3_mlp_0_bias" [id=459, type=get_attr]; -"460 _scale_82" [id=460, type=get_attr]; -"461 _zero_point_82" [id=461, type=get_attr]; -"462 _frozen_param82" [id=462, type=get_attr]; -"463 dequantize_per_channel_default_82" [id=463, type=dequantize_per_channel]; -"464 features_5_3_mlp_3_bias" [id=464, type=get_attr]; -"465 features_5_3_norm2_weight" [id=465, type=get_attr]; -"466 features_5_3_norm2_bias" [id=466, type=get_attr]; -"467 _scale_83" [id=467, type=get_attr]; -"468 _zero_point_83" [id=468, type=get_attr]; -"469 _frozen_param83" [id=469, type=get_attr]; -"470 dequantize_per_channel_default_83" [id=470, type=dequantize_per_channel]; -"471 features_5_4_attn_cpb_mlp_0_bias" [id=471, type=get_attr]; -"472 _scale_84" [id=472, type=get_attr]; -"473 _zero_point_84" [id=473, type=get_attr]; -"474 _frozen_param84" [id=474, type=get_attr]; -"475 dequantize_per_channel_default_84" [id=475, type=dequantize_per_channel]; -"476 features_5_4_norm1_weight" [id=476, type=get_attr]; -"477 features_5_4_norm1_bias" [id=477, type=get_attr]; -"478 _scale_85" [id=478, type=get_attr]; -"479 _zero_point_85" [id=479, type=get_attr]; -"480 _frozen_param85" [id=480, type=get_attr]; -"481 dequantize_per_channel_default_85" [id=481, type=dequantize_per_channel]; -"482 features_5_4_mlp_0_bias" [id=482, type=get_attr]; -"483 _scale_86" [id=483, type=get_attr]; -"484 _zero_point_86" [id=484, type=get_attr]; -"485 _frozen_param86" [id=485, type=get_attr]; -"486 dequantize_per_channel_default_86" [id=486, type=dequantize_per_channel]; -"487 features_5_4_mlp_3_bias" [id=487, type=get_attr]; -"488 features_5_4_norm2_weight" [id=488, type=get_attr]; -"489 features_5_4_norm2_bias" [id=489, type=get_attr]; -"490 _scale_87" [id=490, type=get_attr]; -"491 _zero_point_87" [id=491, type=get_attr]; -"492 _frozen_param87" [id=492, type=get_attr]; -"493 dequantize_per_channel_default_87" [id=493, type=dequantize_per_channel]; -"494 features_5_5_attn_cpb_mlp_0_bias" [id=494, type=get_attr]; -"495 _scale_88" [id=495, type=get_attr]; -"496 _zero_point_88" [id=496, type=get_attr]; -"497 _frozen_param88" [id=497, type=get_attr]; -"498 dequantize_per_channel_default_88" [id=498, type=dequantize_per_channel]; -"499 features_5_5_norm1_weight" [id=499, type=get_attr]; -"500 features_5_5_norm1_bias" [id=500, type=get_attr]; -"501 _scale_89" [id=501, type=get_attr]; -"502 _zero_point_89" [id=502, type=get_attr]; -"503 _frozen_param89" [id=503, type=get_attr]; -"504 dequantize_per_channel_default_89" [id=504, type=dequantize_per_channel]; -"505 features_5_5_mlp_0_bias" [id=505, type=get_attr]; -"506 _scale_90" [id=506, type=get_attr]; -"507 _zero_point_90" [id=507, type=get_attr]; -"508 _frozen_param90" [id=508, type=get_attr]; -"509 dequantize_per_channel_default_90" [id=509, type=dequantize_per_channel]; -"510 features_5_5_mlp_3_bias" [id=510, type=get_attr]; -"511 features_5_5_norm2_weight" [id=511, type=get_attr]; -"512 features_5_5_norm2_bias" [id=512, type=get_attr]; -"513 _scale_91" [id=513, type=get_attr]; -"514 _zero_point_91" [id=514, type=get_attr]; -"515 _frozen_param91" [id=515, type=get_attr]; -"516 dequantize_per_channel_default_91" [id=516, type=dequantize_per_channel]; -"517 features_5_6_attn_cpb_mlp_0_bias" [id=517, type=get_attr]; -"518 _scale_92" [id=518, type=get_attr]; -"519 _zero_point_92" [id=519, type=get_attr]; -"520 _frozen_param92" [id=520, type=get_attr]; -"521 dequantize_per_channel_default_92" [id=521, type=dequantize_per_channel]; -"522 features_5_6_norm1_weight" [id=522, type=get_attr]; -"523 features_5_6_norm1_bias" [id=523, type=get_attr]; -"524 _scale_93" [id=524, type=get_attr]; -"525 _zero_point_93" [id=525, type=get_attr]; -"526 _frozen_param93" [id=526, type=get_attr]; -"527 dequantize_per_channel_default_93" [id=527, type=dequantize_per_channel]; -"528 features_5_6_mlp_0_bias" [id=528, type=get_attr]; -"529 _scale_94" [id=529, type=get_attr]; -"530 _zero_point_94" [id=530, type=get_attr]; -"531 _frozen_param94" [id=531, type=get_attr]; -"532 dequantize_per_channel_default_94" [id=532, type=dequantize_per_channel]; -"533 features_5_6_mlp_3_bias" [id=533, type=get_attr]; -"534 features_5_6_norm2_weight" [id=534, type=get_attr]; -"535 features_5_6_norm2_bias" [id=535, type=get_attr]; -"536 _scale_95" [id=536, type=get_attr]; -"537 _zero_point_95" [id=537, type=get_attr]; -"538 _frozen_param95" [id=538, type=get_attr]; -"539 dequantize_per_channel_default_95" [id=539, type=dequantize_per_channel]; -"540 features_5_7_attn_cpb_mlp_0_bias" [id=540, type=get_attr]; -"541 _scale_96" [id=541, type=get_attr]; -"542 _zero_point_96" [id=542, type=get_attr]; -"543 _frozen_param96" [id=543, type=get_attr]; -"544 dequantize_per_channel_default_96" [id=544, type=dequantize_per_channel]; -"545 features_5_7_norm1_weight" [id=545, type=get_attr]; -"546 features_5_7_norm1_bias" [id=546, type=get_attr]; -"547 _scale_97" [id=547, type=get_attr]; -"548 _zero_point_97" [id=548, type=get_attr]; -"549 _frozen_param97" [id=549, type=get_attr]; -"550 dequantize_per_channel_default_97" [id=550, type=dequantize_per_channel]; -"551 features_5_7_mlp_0_bias" [id=551, type=get_attr]; -"552 _scale_98" [id=552, type=get_attr]; -"553 _zero_point_98" [id=553, type=get_attr]; -"554 _frozen_param98" [id=554, type=get_attr]; -"555 dequantize_per_channel_default_98" [id=555, type=dequantize_per_channel]; -"556 features_5_7_mlp_3_bias" [id=556, type=get_attr]; -"557 features_5_7_norm2_weight" [id=557, type=get_attr]; -"558 features_5_7_norm2_bias" [id=558, type=get_attr]; -"559 _scale_99" [id=559, type=get_attr]; -"560 _zero_point_99" [id=560, type=get_attr]; -"561 _frozen_param99" [id=561, type=get_attr]; -"562 dequantize_per_channel_default_99" [id=562, type=dequantize_per_channel]; -"563 features_5_8_attn_cpb_mlp_0_bias" [id=563, type=get_attr]; -"564 _scale_100" [id=564, type=get_attr]; -"565 _zero_point_100" [id=565, type=get_attr]; -"566 _frozen_param100" [id=566, type=get_attr]; -"567 dequantize_per_channel_default_100" [id=567, type=dequantize_per_channel]; -"568 features_5_8_norm1_weight" [id=568, type=get_attr]; -"569 features_5_8_norm1_bias" [id=569, type=get_attr]; -"570 _scale_101" [id=570, type=get_attr]; -"571 _zero_point_101" [id=571, type=get_attr]; -"572 _frozen_param101" [id=572, type=get_attr]; -"573 dequantize_per_channel_default_101" [id=573, type=dequantize_per_channel]; -"574 features_5_8_mlp_0_bias" [id=574, type=get_attr]; -"575 _scale_102" [id=575, type=get_attr]; -"576 _zero_point_102" [id=576, type=get_attr]; -"577 _frozen_param102" [id=577, type=get_attr]; -"578 dequantize_per_channel_default_102" [id=578, type=dequantize_per_channel]; -"579 features_5_8_mlp_3_bias" [id=579, type=get_attr]; -"580 features_5_8_norm2_weight" [id=580, type=get_attr]; -"581 features_5_8_norm2_bias" [id=581, type=get_attr]; -"582 _scale_103" [id=582, type=get_attr]; -"583 _zero_point_103" [id=583, type=get_attr]; -"584 _frozen_param103" [id=584, type=get_attr]; -"585 dequantize_per_channel_default_103" [id=585, type=dequantize_per_channel]; -"586 features_5_9_attn_cpb_mlp_0_bias" [id=586, type=get_attr]; -"587 _scale_104" [id=587, type=get_attr]; -"588 _zero_point_104" [id=588, type=get_attr]; -"589 _frozen_param104" [id=589, type=get_attr]; -"590 dequantize_per_channel_default_104" [id=590, type=dequantize_per_channel]; -"591 features_5_9_norm1_weight" [id=591, type=get_attr]; -"592 features_5_9_norm1_bias" [id=592, type=get_attr]; -"593 _scale_105" [id=593, type=get_attr]; -"594 _zero_point_105" [id=594, type=get_attr]; -"595 _frozen_param105" [id=595, type=get_attr]; -"596 dequantize_per_channel_default_105" [id=596, type=dequantize_per_channel]; -"597 features_5_9_mlp_0_bias" [id=597, type=get_attr]; -"598 _scale_106" [id=598, type=get_attr]; -"599 _zero_point_106" [id=599, type=get_attr]; -"600 _frozen_param106" [id=600, type=get_attr]; -"601 dequantize_per_channel_default_106" [id=601, type=dequantize_per_channel]; -"602 features_5_9_mlp_3_bias" [id=602, type=get_attr]; -"603 features_5_9_norm2_weight" [id=603, type=get_attr]; -"604 features_5_9_norm2_bias" [id=604, type=get_attr]; -"605 _scale_107" [id=605, type=get_attr]; -"606 _zero_point_107" [id=606, type=get_attr]; -"607 _frozen_param107" [id=607, type=get_attr]; -"608 dequantize_per_channel_default_107" [id=608, type=dequantize_per_channel]; -"609 features_5_10_attn_cpb_mlp_0_bias" [id=609, type=get_attr]; -"610 _scale_108" [id=610, type=get_attr]; -"611 _zero_point_108" [id=611, type=get_attr]; -"612 _frozen_param108" [id=612, type=get_attr]; -"613 dequantize_per_channel_default_108" [id=613, type=dequantize_per_channel]; -"614 features_5_10_norm1_weight" [id=614, type=get_attr]; -"615 features_5_10_norm1_bias" [id=615, type=get_attr]; -"616 _scale_109" [id=616, type=get_attr]; -"617 _zero_point_109" [id=617, type=get_attr]; -"618 _frozen_param109" [id=618, type=get_attr]; -"619 dequantize_per_channel_default_109" [id=619, type=dequantize_per_channel]; -"620 features_5_10_mlp_0_bias" [id=620, type=get_attr]; -"621 _scale_110" [id=621, type=get_attr]; -"622 _zero_point_110" [id=622, type=get_attr]; -"623 _frozen_param110" [id=623, type=get_attr]; -"624 dequantize_per_channel_default_110" [id=624, type=dequantize_per_channel]; -"625 features_5_10_mlp_3_bias" [id=625, type=get_attr]; -"626 features_5_10_norm2_weight" [id=626, type=get_attr]; -"627 features_5_10_norm2_bias" [id=627, type=get_attr]; -"628 _scale_111" [id=628, type=get_attr]; -"629 _zero_point_111" [id=629, type=get_attr]; -"630 _frozen_param111" [id=630, type=get_attr]; -"631 dequantize_per_channel_default_111" [id=631, type=dequantize_per_channel]; -"632 features_5_11_attn_cpb_mlp_0_bias" [id=632, type=get_attr]; -"633 _scale_112" [id=633, type=get_attr]; -"634 _zero_point_112" [id=634, type=get_attr]; -"635 _frozen_param112" [id=635, type=get_attr]; -"636 dequantize_per_channel_default_112" [id=636, type=dequantize_per_channel]; -"637 features_5_11_norm1_weight" [id=637, type=get_attr]; -"638 features_5_11_norm1_bias" [id=638, type=get_attr]; -"639 _scale_113" [id=639, type=get_attr]; -"640 _zero_point_113" [id=640, type=get_attr]; -"641 _frozen_param113" [id=641, type=get_attr]; -"642 dequantize_per_channel_default_113" [id=642, type=dequantize_per_channel]; -"643 features_5_11_mlp_0_bias" [id=643, type=get_attr]; -"644 _scale_114" [id=644, type=get_attr]; -"645 _zero_point_114" [id=645, type=get_attr]; -"646 _frozen_param114" [id=646, type=get_attr]; -"647 dequantize_per_channel_default_114" [id=647, type=dequantize_per_channel]; -"648 features_5_11_mlp_3_bias" [id=648, type=get_attr]; -"649 features_5_11_norm2_weight" [id=649, type=get_attr]; -"650 features_5_11_norm2_bias" [id=650, type=get_attr]; -"651 _scale_115" [id=651, type=get_attr]; -"652 _zero_point_115" [id=652, type=get_attr]; -"653 _frozen_param115" [id=653, type=get_attr]; -"654 dequantize_per_channel_default_115" [id=654, type=dequantize_per_channel]; -"655 features_5_12_attn_cpb_mlp_0_bias" [id=655, type=get_attr]; -"656 _scale_116" [id=656, type=get_attr]; -"657 _zero_point_116" [id=657, type=get_attr]; -"658 _frozen_param116" [id=658, type=get_attr]; -"659 dequantize_per_channel_default_116" [id=659, type=dequantize_per_channel]; -"660 features_5_12_norm1_weight" [id=660, type=get_attr]; -"661 features_5_12_norm1_bias" [id=661, type=get_attr]; -"662 _scale_117" [id=662, type=get_attr]; -"663 _zero_point_117" [id=663, type=get_attr]; -"664 _frozen_param117" [id=664, type=get_attr]; -"665 dequantize_per_channel_default_117" [id=665, type=dequantize_per_channel]; -"666 features_5_12_mlp_0_bias" [id=666, type=get_attr]; -"667 _scale_118" [id=667, type=get_attr]; -"668 _zero_point_118" [id=668, type=get_attr]; -"669 _frozen_param118" [id=669, type=get_attr]; -"670 dequantize_per_channel_default_118" [id=670, type=dequantize_per_channel]; -"671 features_5_12_mlp_3_bias" [id=671, type=get_attr]; -"672 features_5_12_norm2_weight" [id=672, type=get_attr]; -"673 features_5_12_norm2_bias" [id=673, type=get_attr]; -"674 _scale_119" [id=674, type=get_attr]; -"675 _zero_point_119" [id=675, type=get_attr]; -"676 _frozen_param119" [id=676, type=get_attr]; -"677 dequantize_per_channel_default_119" [id=677, type=dequantize_per_channel]; -"678 features_5_13_attn_cpb_mlp_0_bias" [id=678, type=get_attr]; -"679 _scale_120" [id=679, type=get_attr]; -"680 _zero_point_120" [id=680, type=get_attr]; -"681 _frozen_param120" [id=681, type=get_attr]; -"682 dequantize_per_channel_default_120" [id=682, type=dequantize_per_channel]; -"683 features_5_13_norm1_weight" [id=683, type=get_attr]; -"684 features_5_13_norm1_bias" [id=684, type=get_attr]; -"685 _scale_121" [id=685, type=get_attr]; -"686 _zero_point_121" [id=686, type=get_attr]; -"687 _frozen_param121" [id=687, type=get_attr]; -"688 dequantize_per_channel_default_121" [id=688, type=dequantize_per_channel]; -"689 features_5_13_mlp_0_bias" [id=689, type=get_attr]; -"690 _scale_122" [id=690, type=get_attr]; -"691 _zero_point_122" [id=691, type=get_attr]; -"692 _frozen_param122" [id=692, type=get_attr]; -"693 dequantize_per_channel_default_122" [id=693, type=dequantize_per_channel]; -"694 features_5_13_mlp_3_bias" [id=694, type=get_attr]; -"695 features_5_13_norm2_weight" [id=695, type=get_attr]; -"696 features_5_13_norm2_bias" [id=696, type=get_attr]; -"697 _scale_123" [id=697, type=get_attr]; -"698 _zero_point_123" [id=698, type=get_attr]; -"699 _frozen_param123" [id=699, type=get_attr]; -"700 dequantize_per_channel_default_123" [id=700, type=dequantize_per_channel]; -"701 features_5_14_attn_cpb_mlp_0_bias" [id=701, type=get_attr]; -"702 _scale_124" [id=702, type=get_attr]; -"703 _zero_point_124" [id=703, type=get_attr]; -"704 _frozen_param124" [id=704, type=get_attr]; -"705 dequantize_per_channel_default_124" [id=705, type=dequantize_per_channel]; -"706 features_5_14_norm1_weight" [id=706, type=get_attr]; -"707 features_5_14_norm1_bias" [id=707, type=get_attr]; -"708 _scale_125" [id=708, type=get_attr]; -"709 _zero_point_125" [id=709, type=get_attr]; -"710 _frozen_param125" [id=710, type=get_attr]; -"711 dequantize_per_channel_default_125" [id=711, type=dequantize_per_channel]; -"712 features_5_14_mlp_0_bias" [id=712, type=get_attr]; -"713 _scale_126" [id=713, type=get_attr]; -"714 _zero_point_126" [id=714, type=get_attr]; -"715 _frozen_param126" [id=715, type=get_attr]; -"716 dequantize_per_channel_default_126" [id=716, type=dequantize_per_channel]; -"717 features_5_14_mlp_3_bias" [id=717, type=get_attr]; -"718 features_5_14_norm2_weight" [id=718, type=get_attr]; -"719 features_5_14_norm2_bias" [id=719, type=get_attr]; -"720 _scale_127" [id=720, type=get_attr]; -"721 _zero_point_127" [id=721, type=get_attr]; -"722 _frozen_param127" [id=722, type=get_attr]; -"723 dequantize_per_channel_default_127" [id=723, type=dequantize_per_channel]; -"724 features_5_15_attn_cpb_mlp_0_bias" [id=724, type=get_attr]; -"725 _scale_128" [id=725, type=get_attr]; -"726 _zero_point_128" [id=726, type=get_attr]; -"727 _frozen_param128" [id=727, type=get_attr]; -"728 dequantize_per_channel_default_128" [id=728, type=dequantize_per_channel]; -"729 features_5_15_norm1_weight" [id=729, type=get_attr]; -"730 features_5_15_norm1_bias" [id=730, type=get_attr]; -"731 _scale_129" [id=731, type=get_attr]; -"732 _zero_point_129" [id=732, type=get_attr]; -"733 _frozen_param129" [id=733, type=get_attr]; -"734 dequantize_per_channel_default_129" [id=734, type=dequantize_per_channel]; -"735 features_5_15_mlp_0_bias" [id=735, type=get_attr]; -"736 _scale_130" [id=736, type=get_attr]; -"737 _zero_point_130" [id=737, type=get_attr]; -"738 _frozen_param130" [id=738, type=get_attr]; -"739 dequantize_per_channel_default_130" [id=739, type=dequantize_per_channel]; -"740 features_5_15_mlp_3_bias" [id=740, type=get_attr]; -"741 features_5_15_norm2_weight" [id=741, type=get_attr]; -"742 features_5_15_norm2_bias" [id=742, type=get_attr]; -"743 _scale_131" [id=743, type=get_attr]; -"744 _zero_point_131" [id=744, type=get_attr]; -"745 _frozen_param131" [id=745, type=get_attr]; -"746 dequantize_per_channel_default_131" [id=746, type=dequantize_per_channel]; -"747 features_5_16_attn_cpb_mlp_0_bias" [id=747, type=get_attr]; -"748 _scale_132" [id=748, type=get_attr]; -"749 _zero_point_132" [id=749, type=get_attr]; -"750 _frozen_param132" [id=750, type=get_attr]; -"751 dequantize_per_channel_default_132" [id=751, type=dequantize_per_channel]; -"752 features_5_16_norm1_weight" [id=752, type=get_attr]; -"753 features_5_16_norm1_bias" [id=753, type=get_attr]; -"754 _scale_133" [id=754, type=get_attr]; -"755 _zero_point_133" [id=755, type=get_attr]; -"756 _frozen_param133" [id=756, type=get_attr]; -"757 dequantize_per_channel_default_133" [id=757, type=dequantize_per_channel]; -"758 features_5_16_mlp_0_bias" [id=758, type=get_attr]; -"759 _scale_134" [id=759, type=get_attr]; -"760 _zero_point_134" [id=760, type=get_attr]; -"761 _frozen_param134" [id=761, type=get_attr]; -"762 dequantize_per_channel_default_134" [id=762, type=dequantize_per_channel]; -"763 features_5_16_mlp_3_bias" [id=763, type=get_attr]; -"764 features_5_16_norm2_weight" [id=764, type=get_attr]; -"765 features_5_16_norm2_bias" [id=765, type=get_attr]; -"766 _scale_135" [id=766, type=get_attr]; -"767 _zero_point_135" [id=767, type=get_attr]; -"768 _frozen_param135" [id=768, type=get_attr]; -"769 dequantize_per_channel_default_135" [id=769, type=dequantize_per_channel]; -"770 features_5_17_attn_cpb_mlp_0_bias" [id=770, type=get_attr]; -"771 _scale_136" [id=771, type=get_attr]; -"772 _zero_point_136" [id=772, type=get_attr]; -"773 _frozen_param136" [id=773, type=get_attr]; -"774 dequantize_per_channel_default_136" [id=774, type=dequantize_per_channel]; -"775 features_5_17_norm1_weight" [id=775, type=get_attr]; -"776 features_5_17_norm1_bias" [id=776, type=get_attr]; -"777 _scale_137" [id=777, type=get_attr]; -"778 _zero_point_137" [id=778, type=get_attr]; -"779 _frozen_param137" [id=779, type=get_attr]; -"780 dequantize_per_channel_default_137" [id=780, type=dequantize_per_channel]; -"781 features_5_17_mlp_0_bias" [id=781, type=get_attr]; -"782 _scale_138" [id=782, type=get_attr]; -"783 _zero_point_138" [id=783, type=get_attr]; -"784 _frozen_param138" [id=784, type=get_attr]; -"785 dequantize_per_channel_default_138" [id=785, type=dequantize_per_channel]; -"786 features_5_17_mlp_3_bias" [id=786, type=get_attr]; -"787 features_5_17_norm2_weight" [id=787, type=get_attr]; -"788 features_5_17_norm2_bias" [id=788, type=get_attr]; -"789 _scale_139" [id=789, type=get_attr]; -"790 _zero_point_139" [id=790, type=get_attr]; -"791 _frozen_param139" [id=791, type=get_attr]; -"792 dequantize_per_channel_default_139" [id=792, type=dequantize_per_channel]; -"793 features_6_norm_weight" [id=793, type=get_attr]; -"794 features_6_norm_bias" [id=794, type=get_attr]; -"795 _scale_140" [id=795, type=get_attr]; -"796 _zero_point_140" [id=796, type=get_attr]; -"797 _frozen_param140" [id=797, type=get_attr]; -"798 dequantize_per_channel_default_140" [id=798, type=dequantize_per_channel]; -"799 features_7_0_attn_cpb_mlp_0_bias" [id=799, type=get_attr]; -"800 _scale_141" [id=800, type=get_attr]; -"801 _zero_point_141" [id=801, type=get_attr]; -"802 _frozen_param141" [id=802, type=get_attr]; -"803 dequantize_per_channel_default_141" [id=803, type=dequantize_per_channel]; -"804 features_7_0_norm1_weight" [id=804, type=get_attr]; -"805 features_7_0_norm1_bias" [id=805, type=get_attr]; -"806 _scale_142" [id=806, type=get_attr]; -"807 _zero_point_142" [id=807, type=get_attr]; -"808 _frozen_param142" [id=808, type=get_attr]; -"809 dequantize_per_channel_default_142" [id=809, type=dequantize_per_channel]; -"810 features_7_0_mlp_0_bias" [id=810, type=get_attr]; -"811 _scale_143" [id=811, type=get_attr]; -"812 _zero_point_143" [id=812, type=get_attr]; -"813 _frozen_param143" [id=813, type=get_attr]; -"814 dequantize_per_channel_default_143" [id=814, type=dequantize_per_channel]; -"815 features_7_0_mlp_3_bias" [id=815, type=get_attr]; -"816 features_7_0_norm2_weight" [id=816, type=get_attr]; -"817 features_7_0_norm2_bias" [id=817, type=get_attr]; -"818 _scale_144" [id=818, type=get_attr]; -"819 _zero_point_144" [id=819, type=get_attr]; -"820 _frozen_param144" [id=820, type=get_attr]; -"821 dequantize_per_channel_default_144" [id=821, type=dequantize_per_channel]; -"822 features_7_1_attn_cpb_mlp_0_bias" [id=822, type=get_attr]; -"823 _scale_145" [id=823, type=get_attr]; -"824 _zero_point_145" [id=824, type=get_attr]; -"825 _frozen_param145" [id=825, type=get_attr]; -"826 dequantize_per_channel_default_145" [id=826, type=dequantize_per_channel]; -"827 features_7_1_norm1_weight" [id=827, type=get_attr]; -"828 features_7_1_norm1_bias" [id=828, type=get_attr]; -"829 _scale_146" [id=829, type=get_attr]; -"830 _zero_point_146" [id=830, type=get_attr]; -"831 _frozen_param146" [id=831, type=get_attr]; -"832 dequantize_per_channel_default_146" [id=832, type=dequantize_per_channel]; -"833 features_7_1_mlp_0_bias" [id=833, type=get_attr]; -"834 _scale_147" [id=834, type=get_attr]; -"835 _zero_point_147" [id=835, type=get_attr]; -"836 _frozen_param147" [id=836, type=get_attr]; -"837 dequantize_per_channel_default_147" [id=837, type=dequantize_per_channel]; -"838 features_7_1_mlp_3_bias" [id=838, type=get_attr]; -"839 features_7_1_norm2_weight" [id=839, type=get_attr]; -"840 features_7_1_norm2_bias" [id=840, type=get_attr]; -"841 norm_weight" [id=841, type=get_attr]; -"842 norm_bias" [id=842, type=get_attr]; -"843 _scale_148" [id=843, type=get_attr]; -"844 _zero_point_148" [id=844, type=get_attr]; -"845 _frozen_param148" [id=845, type=get_attr]; -"846 dequantize_per_channel_default_148" [id=846, type=dequantize_per_channel]; -"847 head_bias" [id=847, type=get_attr]; -"848 features_1_0_attn_relative_coords_table" [id=848, type=get_attr]; -"849 features_1_0_attn_relative_position_index" [id=849, type=get_attr]; -"850 features_1_1_attn_relative_coords_table" [id=850, type=get_attr]; -"851 features_1_1_attn_relative_position_index" [id=851, type=get_attr]; -"852 features_3_0_attn_relative_coords_table" [id=852, type=get_attr]; -"853 features_3_0_attn_relative_position_index" [id=853, type=get_attr]; -"854 features_3_1_attn_relative_coords_table" [id=854, type=get_attr]; -"855 features_3_1_attn_relative_position_index" [id=855, type=get_attr]; -"856 features_5_0_attn_relative_coords_table" [id=856, type=get_attr]; -"857 features_5_0_attn_relative_position_index" [id=857, type=get_attr]; -"858 features_5_1_attn_relative_coords_table" [id=858, type=get_attr]; -"859 features_5_1_attn_relative_position_index" [id=859, type=get_attr]; -"860 features_5_2_attn_relative_coords_table" [id=860, type=get_attr]; -"861 features_5_2_attn_relative_position_index" [id=861, type=get_attr]; -"862 features_5_3_attn_relative_coords_table" [id=862, type=get_attr]; -"863 features_5_3_attn_relative_position_index" [id=863, type=get_attr]; -"864 features_5_4_attn_relative_coords_table" [id=864, type=get_attr]; -"865 features_5_4_attn_relative_position_index" [id=865, type=get_attr]; -"866 features_5_5_attn_relative_coords_table" [id=866, type=get_attr]; -"867 features_5_5_attn_relative_position_index" [id=867, type=get_attr]; -"868 features_5_6_attn_relative_coords_table" [id=868, type=get_attr]; -"869 features_5_6_attn_relative_position_index" [id=869, type=get_attr]; -"870 features_5_7_attn_relative_coords_table" [id=870, type=get_attr]; -"871 features_5_7_attn_relative_position_index" [id=871, type=get_attr]; -"872 features_5_8_attn_relative_coords_table" [id=872, type=get_attr]; -"873 features_5_8_attn_relative_position_index" [id=873, type=get_attr]; -"874 features_5_9_attn_relative_coords_table" [id=874, type=get_attr]; -"875 features_5_9_attn_relative_position_index" [id=875, type=get_attr]; -"876 features_5_10_attn_relative_coords_table" [id=876, type=get_attr]; -"877 features_5_10_attn_relative_position_index" [id=877, type=get_attr]; -"878 features_5_11_attn_relative_coords_table" [id=878, type=get_attr]; -"879 features_5_11_attn_relative_position_index" [id=879, type=get_attr]; -"880 features_5_12_attn_relative_coords_table" [id=880, type=get_attr]; -"881 features_5_12_attn_relative_position_index" [id=881, type=get_attr]; -"882 features_5_13_attn_relative_coords_table" [id=882, type=get_attr]; -"883 features_5_13_attn_relative_position_index" [id=883, type=get_attr]; -"884 features_5_14_attn_relative_coords_table" [id=884, type=get_attr]; -"885 features_5_14_attn_relative_position_index" [id=885, type=get_attr]; -"886 features_5_15_attn_relative_coords_table" [id=886, type=get_attr]; -"887 features_5_15_attn_relative_position_index" [id=887, type=get_attr]; -"888 features_5_16_attn_relative_coords_table" [id=888, type=get_attr]; -"889 features_5_16_attn_relative_position_index" [id=889, type=get_attr]; -"890 features_5_17_attn_relative_coords_table" [id=890, type=get_attr]; -"891 features_5_17_attn_relative_position_index" [id=891, type=get_attr]; -"892 features_7_0_attn_relative_coords_table" [id=892, type=get_attr]; -"893 features_7_0_attn_relative_position_index" [id=893, type=get_attr]; -"894 features_7_1_attn_relative_coords_table" [id=894, type=get_attr]; -"895 features_7_1_attn_relative_position_index" [id=895, type=get_attr]; -"896 features_1_1_attn_lifted_tensor_0" [id=896, type=get_attr]; -"897 features_1_1_attn_lifted_tensor_1" [id=897, type=get_attr]; -"898 features_1_1_attn_lifted_tensor_2" [id=898, type=get_attr]; -"899 features_1_1_attn_lifted_tensor_3" [id=899, type=get_attr]; -"900 features_1_1_attn_lifted_tensor_4" [id=900, type=get_attr]; -"901 features_1_1_attn_lifted_tensor_5" [id=901, type=get_attr]; -"902 features_1_1_attn_lifted_tensor_6" [id=902, type=get_attr]; -"903 features_1_1_attn_lifted_tensor_7" [id=903, type=get_attr]; -"904 features_1_1_attn_lifted_tensor_8" [id=904, type=get_attr]; -"905 features_3_1_attn_lifted_tensor_9" [id=905, type=get_attr]; -"906 features_3_1_attn_lifted_tensor_10" [id=906, type=get_attr]; -"907 features_3_1_attn_lifted_tensor_11" [id=907, type=get_attr]; -"908 features_3_1_attn_lifted_tensor_12" [id=908, type=get_attr]; -"909 features_3_1_attn_lifted_tensor_13" [id=909, type=get_attr]; -"910 features_3_1_attn_lifted_tensor_14" [id=910, type=get_attr]; -"911 features_3_1_attn_lifted_tensor_15" [id=911, type=get_attr]; -"912 features_3_1_attn_lifted_tensor_16" [id=912, type=get_attr]; -"913 features_3_1_attn_lifted_tensor_17" [id=913, type=get_attr]; -"914 features_5_1_attn_lifted_tensor_18" [id=914, type=get_attr]; -"915 features_5_1_attn_lifted_tensor_19" [id=915, type=get_attr]; -"916 features_5_1_attn_lifted_tensor_20" [id=916, type=get_attr]; -"917 features_5_1_attn_lifted_tensor_21" [id=917, type=get_attr]; -"918 features_5_1_attn_lifted_tensor_22" [id=918, type=get_attr]; -"919 features_5_1_attn_lifted_tensor_23" [id=919, type=get_attr]; -"920 features_5_1_attn_lifted_tensor_24" [id=920, type=get_attr]; -"921 features_5_1_attn_lifted_tensor_25" [id=921, type=get_attr]; -"922 features_5_1_attn_lifted_tensor_26" [id=922, type=get_attr]; -"923 features_5_3_attn_lifted_tensor_27" [id=923, type=get_attr]; -"924 features_5_3_attn_lifted_tensor_28" [id=924, type=get_attr]; -"925 features_5_3_attn_lifted_tensor_29" [id=925, type=get_attr]; -"926 features_5_3_attn_lifted_tensor_30" [id=926, type=get_attr]; -"927 features_5_3_attn_lifted_tensor_31" [id=927, type=get_attr]; -"928 features_5_3_attn_lifted_tensor_32" [id=928, type=get_attr]; -"929 features_5_3_attn_lifted_tensor_33" [id=929, type=get_attr]; -"930 features_5_3_attn_lifted_tensor_34" [id=930, type=get_attr]; -"931 features_5_3_attn_lifted_tensor_35" [id=931, type=get_attr]; -"932 features_5_5_attn_lifted_tensor_36" [id=932, type=get_attr]; -"933 features_5_5_attn_lifted_tensor_37" [id=933, type=get_attr]; -"934 features_5_5_attn_lifted_tensor_38" [id=934, type=get_attr]; -"935 features_5_5_attn_lifted_tensor_39" [id=935, type=get_attr]; -"936 features_5_5_attn_lifted_tensor_40" [id=936, type=get_attr]; -"937 features_5_5_attn_lifted_tensor_41" [id=937, type=get_attr]; -"938 features_5_5_attn_lifted_tensor_42" [id=938, type=get_attr]; -"939 features_5_5_attn_lifted_tensor_43" [id=939, type=get_attr]; -"940 features_5_5_attn_lifted_tensor_44" [id=940, type=get_attr]; -"941 features_5_7_attn_lifted_tensor_45" [id=941, type=get_attr]; -"942 features_5_7_attn_lifted_tensor_46" [id=942, type=get_attr]; -"943 features_5_7_attn_lifted_tensor_47" [id=943, type=get_attr]; -"944 features_5_7_attn_lifted_tensor_48" [id=944, type=get_attr]; -"945 features_5_7_attn_lifted_tensor_49" [id=945, type=get_attr]; -"946 features_5_7_attn_lifted_tensor_50" [id=946, type=get_attr]; -"947 features_5_7_attn_lifted_tensor_51" [id=947, type=get_attr]; -"948 features_5_7_attn_lifted_tensor_52" [id=948, type=get_attr]; -"949 features_5_7_attn_lifted_tensor_53" [id=949, type=get_attr]; -"950 features_5_9_attn_lifted_tensor_54" [id=950, type=get_attr]; -"951 features_5_9_attn_lifted_tensor_55" [id=951, type=get_attr]; -"952 features_5_9_attn_lifted_tensor_56" [id=952, type=get_attr]; -"953 features_5_9_attn_lifted_tensor_57" [id=953, type=get_attr]; -"954 features_5_9_attn_lifted_tensor_58" [id=954, type=get_attr]; -"955 features_5_9_attn_lifted_tensor_59" [id=955, type=get_attr]; -"956 features_5_9_attn_lifted_tensor_60" [id=956, type=get_attr]; -"957 features_5_9_attn_lifted_tensor_61" [id=957, type=get_attr]; -"958 features_5_9_attn_lifted_tensor_62" [id=958, type=get_attr]; -"959 features_5_11_attn_lifted_tensor_63" [id=959, type=get_attr]; -"960 features_5_11_attn_lifted_tensor_64" [id=960, type=get_attr]; -"961 features_5_11_attn_lifted_tensor_65" [id=961, type=get_attr]; -"962 features_5_11_attn_lifted_tensor_66" [id=962, type=get_attr]; -"963 features_5_11_attn_lifted_tensor_67" [id=963, type=get_attr]; -"964 features_5_11_attn_lifted_tensor_68" [id=964, type=get_attr]; -"965 features_5_11_attn_lifted_tensor_69" [id=965, type=get_attr]; -"966 features_5_11_attn_lifted_tensor_70" [id=966, type=get_attr]; -"967 features_5_11_attn_lifted_tensor_71" [id=967, type=get_attr]; -"968 features_5_13_attn_lifted_tensor_72" [id=968, type=get_attr]; -"969 features_5_13_attn_lifted_tensor_73" [id=969, type=get_attr]; -"970 features_5_13_attn_lifted_tensor_74" [id=970, type=get_attr]; -"971 features_5_13_attn_lifted_tensor_75" [id=971, type=get_attr]; -"972 features_5_13_attn_lifted_tensor_76" [id=972, type=get_attr]; -"973 features_5_13_attn_lifted_tensor_77" [id=973, type=get_attr]; -"974 features_5_13_attn_lifted_tensor_78" [id=974, type=get_attr]; -"975 features_5_13_attn_lifted_tensor_79" [id=975, type=get_attr]; -"976 features_5_13_attn_lifted_tensor_80" [id=976, type=get_attr]; -"977 features_5_15_attn_lifted_tensor_81" [id=977, type=get_attr]; -"978 features_5_15_attn_lifted_tensor_82" [id=978, type=get_attr]; -"979 features_5_15_attn_lifted_tensor_83" [id=979, type=get_attr]; -"980 features_5_15_attn_lifted_tensor_84" [id=980, type=get_attr]; -"981 features_5_15_attn_lifted_tensor_85" [id=981, type=get_attr]; -"982 features_5_15_attn_lifted_tensor_86" [id=982, type=get_attr]; -"983 features_5_15_attn_lifted_tensor_87" [id=983, type=get_attr]; -"984 features_5_15_attn_lifted_tensor_88" [id=984, type=get_attr]; -"985 features_5_15_attn_lifted_tensor_89" [id=985, type=get_attr]; -"986 features_5_17_attn_lifted_tensor_90" [id=986, type=get_attr]; -"987 features_5_17_attn_lifted_tensor_91" [id=987, type=get_attr]; -"988 features_5_17_attn_lifted_tensor_92" [id=988, type=get_attr]; -"989 features_5_17_attn_lifted_tensor_93" [id=989, type=get_attr]; -"990 features_5_17_attn_lifted_tensor_94" [id=990, type=get_attr]; -"991 features_5_17_attn_lifted_tensor_95" [id=991, type=get_attr]; -"992 features_5_17_attn_lifted_tensor_96" [id=992, type=get_attr]; -"993 features_5_17_attn_lifted_tensor_97" [id=993, type=get_attr]; -"994 features_5_17_attn_lifted_tensor_98" [id=994, type=get_attr]; -"995 x" [id=995, type=input]; -"996 quantize_per_tensor_default" [id=996, type=quantize_per_tensor]; -"997 dequantize_per_tensor_default" [id=997, type=dequantize_per_tensor]; -"998 conv2d" [id=998, type=conv2d]; -"999 permute" [id=999, type=permute]; -"1000 layer_norm" [id=1000, type=layer_norm]; -"1001 quantize_per_tensor_default_1" [id=1001, type=quantize_per_tensor]; -"1002 dequantize_per_tensor_default_1" [id=1002, type=dequantize_per_tensor]; -"1003 linear" [id=1003, type=linear]; -"1004 relu_" [id=1004, type=relu_]; -"1005 linear_1" [id=1005, type=linear]; -"1006 view" [id=1006, type=view]; -"1007 index" [id=1007, type=index]; -"1008 view_1" [id=1008, type=view]; -"1009 permute_1" [id=1009, type=permute]; -"1010 contiguous" [id=1010, type=contiguous]; -"1011 unsqueeze" [id=1011, type=unsqueeze]; -"1012 sigmoid" [id=1012, type=sigmoid]; -"1013 mul" [id=1013, type=mul]; -"1014 pad" [id=1014, type=pad]; -"1015 view_2" [id=1015, type=view]; -"1016 permute_2" [id=1016, type=permute]; -"1017 reshape" [id=1017, type=reshape]; -"1018 clone" [id=1018, type=clone]; -"1019 slice_1" [id=1019, type=slice]; -"1020 zero_" [id=1020, type=zero_]; -"1021 linear_2" [id=1021, type=linear]; -"1022 reshape_1" [id=1022, type=reshape]; -"1023 permute_3" [id=1023, type=permute]; -"1024 select" [id=1024, type=select]; -"1025 select_1" [id=1025, type=select]; -"1026 select_2" [id=1026, type=select]; -"1027 linalg_vector_norm" [id=1027, type=linalg_vector_norm]; -"1028 clamp_min" [id=1028, type=clamp_min]; -"1029 expand_as" [id=1029, type=expand_as]; -"1030 div" [id=1030, type=div]; -"1031 quantize_per_tensor_default_2" [id=1031, type=quantize_per_tensor]; -"1032 dequantize_per_tensor_default_2" [id=1032, type=dequantize_per_tensor]; -"1033 linalg_vector_norm_1" [id=1033, type=linalg_vector_norm]; -"1034 clamp_min_1" [id=1034, type=clamp_min]; -"1035 expand_as_1" [id=1035, type=expand_as]; -"1036 div_1" [id=1036, type=div]; -"1037 quantize_per_tensor_default_3" [id=1037, type=quantize_per_tensor]; -"1038 dequantize_per_tensor_default_3" [id=1038, type=dequantize_per_tensor]; -"1039 transpose" [id=1039, type=transpose]; -"1040 matmul" [id=1040, type=matmul]; -"1041 clamp" [id=1041, type=clamp]; -"1042 exp" [id=1042, type=exp]; -"1043 mul_1" [id=1043, type=mul]; -"1044 add" [id=1044, type=add]; -"1045 softmax" [id=1045, type=softmax]; -"1046 dropout" [id=1046, type=dropout]; -"1047 matmul_1" [id=1047, type=matmul]; -"1048 quantize_per_tensor_default_4" [id=1048, type=quantize_per_tensor]; -"1049 dequantize_per_tensor_default_4" [id=1049, type=dequantize_per_tensor]; -"1050 transpose_1" [id=1050, type=transpose]; -"1051 reshape_2" [id=1051, type=reshape]; -"1052 linear_3" [id=1052, type=linear]; -"1053 dropout_1" [id=1053, type=dropout]; -"1054 view_3" [id=1054, type=view]; -"1055 permute_4" [id=1055, type=permute]; -"1056 reshape_3" [id=1056, type=reshape]; -"1057 slice_2" [id=1057, type=slice]; -"1058 slice_3" [id=1058, type=slice]; -"1059 layer_norm_1" [id=1059, type=layer_norm]; -"1060 add_1" [id=1060, type=add]; -"1061 quantize_per_tensor_default_5" [id=1061, type=quantize_per_tensor]; -"1062 dequantize_per_tensor_default_5" [id=1062, type=dequantize_per_tensor]; -"1063 linear_4" [id=1063, type=linear]; -"1064 gelu" [id=1064, type=gelu]; -"1065 quantize_per_tensor_default_6" [id=1065, type=quantize_per_tensor]; -"1066 dequantize_per_tensor_default_6" [id=1066, type=dequantize_per_tensor]; -"1067 dropout_2" [id=1067, type=dropout]; -"1068 linear_5" [id=1068, type=linear]; -"1069 dropout_3" [id=1069, type=dropout]; -"1070 layer_norm_2" [id=1070, type=layer_norm]; -"1071 add_2" [id=1071, type=add]; -"1072 linear_6" [id=1072, type=linear]; -"1073 relu__1" [id=1073, type=relu_]; -"1074 linear_7" [id=1074, type=linear]; -"1075 view_4" [id=1075, type=view]; -"1076 index_1" [id=1076, type=index]; -"1077 view_5" [id=1077, type=view]; -"1078 permute_5" [id=1078, type=permute]; -"1079 contiguous_1" [id=1079, type=contiguous]; -"1080 unsqueeze_1" [id=1080, type=unsqueeze]; -"1081 sigmoid_1" [id=1081, type=sigmoid]; -"1082 mul_2" [id=1082, type=mul]; -"1083 pad_1" [id=1083, type=pad]; -"1084 roll" [id=1084, type=roll]; -"1085 view_6" [id=1085, type=view]; -"1086 permute_6" [id=1086, type=permute]; -"1087 reshape_4" [id=1087, type=reshape]; -"1088 quantize_per_tensor_default_7" [id=1088, type=quantize_per_tensor]; -"1089 dequantize_per_tensor_default_7" [id=1089, type=dequantize_per_tensor]; -"1090 clone_1" [id=1090, type=clone]; -"1091 slice_4" [id=1091, type=slice]; -"1092 zero__1" [id=1092, type=zero_]; -"1093 linear_8" [id=1093, type=linear]; -"1094 reshape_5" [id=1094, type=reshape]; -"1095 permute_7" [id=1095, type=permute]; -"1096 select_3" [id=1096, type=select]; -"1097 select_4" [id=1097, type=select]; -"1098 select_5" [id=1098, type=select]; -"1099 linalg_vector_norm_2" [id=1099, type=linalg_vector_norm]; -"1100 clamp_min_2" [id=1100, type=clamp_min]; -"1101 expand_as_2" [id=1101, type=expand_as]; -"1102 div_2" [id=1102, type=div]; -"1103 quantize_per_tensor_default_8" [id=1103, type=quantize_per_tensor]; -"1104 dequantize_per_tensor_default_8" [id=1104, type=dequantize_per_tensor]; -"1105 linalg_vector_norm_3" [id=1105, type=linalg_vector_norm]; -"1106 clamp_min_3" [id=1106, type=clamp_min]; -"1107 expand_as_3" [id=1107, type=expand_as]; -"1108 div_3" [id=1108, type=div]; -"1109 quantize_per_tensor_default_9" [id=1109, type=quantize_per_tensor]; -"1110 dequantize_per_tensor_default_9" [id=1110, type=dequantize_per_tensor]; -"1111 transpose_2" [id=1111, type=transpose]; -"1112 matmul_2" [id=1112, type=matmul]; -"1113 clamp_1" [id=1113, type=clamp]; -"1114 exp_1" [id=1114, type=exp]; -"1115 mul_3" [id=1115, type=mul]; -"1116 add_3" [id=1116, type=add]; -"1117 new_zeros" [id=1117, type=new_zeros]; -"1118 lift_fresh_copy" [id=1118, type=lift_fresh_copy]; -"1119 slice_5" [id=1119, type=slice]; -"1120 slice_6" [id=1120, type=slice]; -"1121 fill_" [id=1121, type=fill_]; -"1122 lift_fresh_copy_1" [id=1122, type=lift_fresh_copy]; -"1123 slice_7" [id=1123, type=slice]; -"1124 slice_8" [id=1124, type=slice]; -"1125 fill__1" [id=1125, type=fill_]; -"1126 lift_fresh_copy_2" [id=1126, type=lift_fresh_copy]; -"1127 slice_9" [id=1127, type=slice]; -"1128 slice_10" [id=1128, type=slice]; -"1129 fill__2" [id=1129, type=fill_]; -"1130 lift_fresh_copy_3" [id=1130, type=lift_fresh_copy]; -"1131 slice_11" [id=1131, type=slice]; -"1132 slice_12" [id=1132, type=slice]; -"1133 fill__3" [id=1133, type=fill_]; -"1134 lift_fresh_copy_4" [id=1134, type=lift_fresh_copy]; -"1135 slice_13" [id=1135, type=slice]; -"1136 slice_14" [id=1136, type=slice]; -"1137 fill__4" [id=1137, type=fill_]; -"1138 lift_fresh_copy_5" [id=1138, type=lift_fresh_copy]; -"1139 slice_15" [id=1139, type=slice]; -"1140 slice_16" [id=1140, type=slice]; -"1141 fill__5" [id=1141, type=fill_]; -"1142 lift_fresh_copy_6" [id=1142, type=lift_fresh_copy]; -"1143 slice_17" [id=1143, type=slice]; -"1144 slice_18" [id=1144, type=slice]; -"1145 fill__6" [id=1145, type=fill_]; -"1146 lift_fresh_copy_7" [id=1146, type=lift_fresh_copy]; -"1147 slice_19" [id=1147, type=slice]; -"1148 slice_20" [id=1148, type=slice]; -"1149 fill__7" [id=1149, type=fill_]; -"1150 lift_fresh_copy_8" [id=1150, type=lift_fresh_copy]; -"1151 slice_21" [id=1151, type=slice]; -"1152 slice_22" [id=1152, type=slice]; -"1153 fill__8" [id=1153, type=fill_]; -"1154 view_7" [id=1154, type=view]; -"1155 permute_8" [id=1155, type=permute]; -"1156 reshape_6" [id=1156, type=reshape]; -"1157 unsqueeze_2" [id=1157, type=unsqueeze]; -"1158 unsqueeze_3" [id=1158, type=unsqueeze]; -"1159 sub" [id=1159, type=sub]; -"1160 ne" [id=1160, type=ne]; -"1161 masked_fill" [id=1161, type=masked_fill]; -"1162 eq" [id=1162, type=eq]; -"1163 masked_fill_1" [id=1163, type=masked_fill]; -"1164 view_8" [id=1164, type=view]; -"1165 unsqueeze_4" [id=1165, type=unsqueeze]; -"1166 unsqueeze_5" [id=1166, type=unsqueeze]; -"1167 add_4" [id=1167, type=add]; -"1168 view_9" [id=1168, type=view]; -"1169 softmax_1" [id=1169, type=softmax]; -"1170 dropout_4" [id=1170, type=dropout]; -"1171 matmul_3" [id=1171, type=matmul]; -"1172 quantize_per_tensor_default_10" [id=1172, type=quantize_per_tensor]; -"1173 dequantize_per_tensor_default_10" [id=1173, type=dequantize_per_tensor]; -"1174 transpose_3" [id=1174, type=transpose]; -"1175 reshape_7" [id=1175, type=reshape]; -"1176 linear_9" [id=1176, type=linear]; -"1177 dropout_5" [id=1177, type=dropout]; -"1178 view_10" [id=1178, type=view]; -"1179 permute_9" [id=1179, type=permute]; -"1180 reshape_8" [id=1180, type=reshape]; -"1181 roll_1" [id=1181, type=roll]; -"1182 slice_23" [id=1182, type=slice]; -"1183 slice_24" [id=1183, type=slice]; -"1184 layer_norm_3" [id=1184, type=layer_norm]; -"1185 add_5" [id=1185, type=add]; -"1186 quantize_per_tensor_default_11" [id=1186, type=quantize_per_tensor]; -"1187 dequantize_per_tensor_default_11" [id=1187, type=dequantize_per_tensor]; -"1188 linear_10" [id=1188, type=linear]; -"1189 gelu_1" [id=1189, type=gelu]; -"1190 quantize_per_tensor_default_12" [id=1190, type=quantize_per_tensor]; -"1191 dequantize_per_tensor_default_12" [id=1191, type=dequantize_per_tensor]; -"1192 dropout_6" [id=1192, type=dropout]; -"1193 linear_11" [id=1193, type=linear]; -"1194 dropout_7" [id=1194, type=dropout]; -"1195 layer_norm_4" [id=1195, type=layer_norm]; -"1196 add_6" [id=1196, type=add]; -"1197 quantize_per_tensor_default_13" [id=1197, type=quantize_per_tensor]; -"1198 dequantize_per_tensor_default_13" [id=1198, type=dequantize_per_tensor]; -"1199 pad_2" [id=1199, type=pad]; -"1200 slice_25" [id=1200, type=slice]; -"1201 slice_26" [id=1201, type=slice]; -"1202 slice_27" [id=1202, type=slice]; -"1203 slice_28" [id=1203, type=slice]; -"1204 slice_29" [id=1204, type=slice]; -"1205 slice_30" [id=1205, type=slice]; -"1206 slice_31" [id=1206, type=slice]; -"1207 slice_32" [id=1207, type=slice]; -"1208 slice_33" [id=1208, type=slice]; -"1209 slice_34" [id=1209, type=slice]; -"1210 slice_35" [id=1210, type=slice]; -"1211 slice_36" [id=1211, type=slice]; -"1212 cat" [id=1212, type=cat]; -"1213 linear_12" [id=1213, type=linear]; -"1214 layer_norm_5" [id=1214, type=layer_norm]; -"1215 quantize_per_tensor_default_14" [id=1215, type=quantize_per_tensor]; -"1216 dequantize_per_tensor_default_14" [id=1216, type=dequantize_per_tensor]; -"1217 linear_13" [id=1217, type=linear]; -"1218 relu__2" [id=1218, type=relu_]; -"1219 linear_14" [id=1219, type=linear]; -"1220 view_11" [id=1220, type=view]; -"1221 index_2" [id=1221, type=index]; -"1222 view_12" [id=1222, type=view]; -"1223 permute_10" [id=1223, type=permute]; -"1224 contiguous_2" [id=1224, type=contiguous]; -"1225 unsqueeze_6" [id=1225, type=unsqueeze]; -"1226 sigmoid_2" [id=1226, type=sigmoid]; -"1227 mul_4" [id=1227, type=mul]; -"1228 pad_3" [id=1228, type=pad]; -"1229 view_13" [id=1229, type=view]; -"1230 permute_11" [id=1230, type=permute]; -"1231 reshape_9" [id=1231, type=reshape]; -"1232 clone_2" [id=1232, type=clone]; -"1233 slice_37" [id=1233, type=slice]; -"1234 zero__2" [id=1234, type=zero_]; -"1235 linear_15" [id=1235, type=linear]; -"1236 reshape_10" [id=1236, type=reshape]; -"1237 permute_12" [id=1237, type=permute]; -"1238 select_6" [id=1238, type=select]; -"1239 select_7" [id=1239, type=select]; -"1240 select_8" [id=1240, type=select]; -"1241 linalg_vector_norm_4" [id=1241, type=linalg_vector_norm]; -"1242 clamp_min_4" [id=1242, type=clamp_min]; -"1243 expand_as_4" [id=1243, type=expand_as]; -"1244 div_4" [id=1244, type=div]; -"1245 quantize_per_tensor_default_15" [id=1245, type=quantize_per_tensor]; -"1246 dequantize_per_tensor_default_15" [id=1246, type=dequantize_per_tensor]; -"1247 linalg_vector_norm_5" [id=1247, type=linalg_vector_norm]; -"1248 clamp_min_5" [id=1248, type=clamp_min]; -"1249 expand_as_5" [id=1249, type=expand_as]; -"1250 div_5" [id=1250, type=div]; -"1251 quantize_per_tensor_default_16" [id=1251, type=quantize_per_tensor]; -"1252 dequantize_per_tensor_default_16" [id=1252, type=dequantize_per_tensor]; -"1253 transpose_4" [id=1253, type=transpose]; -"1254 matmul_4" [id=1254, type=matmul]; -"1255 clamp_2" [id=1255, type=clamp]; -"1256 exp_2" [id=1256, type=exp]; -"1257 mul_5" [id=1257, type=mul]; -"1258 add_7" [id=1258, type=add]; -"1259 softmax_2" [id=1259, type=softmax]; -"1260 dropout_8" [id=1260, type=dropout]; -"1261 matmul_5" [id=1261, type=matmul]; -"1262 quantize_per_tensor_default_17" [id=1262, type=quantize_per_tensor]; -"1263 dequantize_per_tensor_default_17" [id=1263, type=dequantize_per_tensor]; -"1264 transpose_5" [id=1264, type=transpose]; -"1265 reshape_11" [id=1265, type=reshape]; -"1266 linear_16" [id=1266, type=linear]; -"1267 dropout_9" [id=1267, type=dropout]; -"1268 view_14" [id=1268, type=view]; -"1269 permute_13" [id=1269, type=permute]; -"1270 reshape_12" [id=1270, type=reshape]; -"1271 slice_38" [id=1271, type=slice]; -"1272 slice_39" [id=1272, type=slice]; -"1273 slice_40" [id=1273, type=slice]; -"1274 slice_41" [id=1274, type=slice]; -"1275 contiguous_3" [id=1275, type=contiguous]; -"1276 layer_norm_6" [id=1276, type=layer_norm]; -"1277 add_8" [id=1277, type=add]; -"1278 quantize_per_tensor_default_18" [id=1278, type=quantize_per_tensor]; -"1279 dequantize_per_tensor_default_18" [id=1279, type=dequantize_per_tensor]; -"1280 linear_17" [id=1280, type=linear]; -"1281 gelu_2" [id=1281, type=gelu]; -"1282 quantize_per_tensor_default_19" [id=1282, type=quantize_per_tensor]; -"1283 dequantize_per_tensor_default_19" [id=1283, type=dequantize_per_tensor]; -"1284 dropout_10" [id=1284, type=dropout]; -"1285 linear_18" [id=1285, type=linear]; -"1286 dropout_11" [id=1286, type=dropout]; -"1287 layer_norm_7" [id=1287, type=layer_norm]; -"1288 add_9" [id=1288, type=add]; -"1289 linear_19" [id=1289, type=linear]; -"1290 relu__3" [id=1290, type=relu_]; -"1291 linear_20" [id=1291, type=linear]; -"1292 view_15" [id=1292, type=view]; -"1293 index_3" [id=1293, type=index]; -"1294 view_16" [id=1294, type=view]; -"1295 permute_14" [id=1295, type=permute]; -"1296 contiguous_4" [id=1296, type=contiguous]; -"1297 unsqueeze_7" [id=1297, type=unsqueeze]; -"1298 sigmoid_3" [id=1298, type=sigmoid]; -"1299 mul_6" [id=1299, type=mul]; -"1300 pad_4" [id=1300, type=pad]; -"1301 roll_2" [id=1301, type=roll]; -"1302 view_17" [id=1302, type=view]; -"1303 permute_15" [id=1303, type=permute]; -"1304 reshape_13" [id=1304, type=reshape]; -"1305 quantize_per_tensor_default_20" [id=1305, type=quantize_per_tensor]; -"1306 dequantize_per_tensor_default_20" [id=1306, type=dequantize_per_tensor]; -"1307 clone_3" [id=1307, type=clone]; -"1308 slice_42" [id=1308, type=slice]; -"1309 zero__3" [id=1309, type=zero_]; -"1310 linear_21" [id=1310, type=linear]; -"1311 reshape_14" [id=1311, type=reshape]; -"1312 permute_16" [id=1312, type=permute]; -"1313 select_9" [id=1313, type=select]; -"1314 select_10" [id=1314, type=select]; -"1315 select_11" [id=1315, type=select]; -"1316 linalg_vector_norm_6" [id=1316, type=linalg_vector_norm]; -"1317 clamp_min_6" [id=1317, type=clamp_min]; -"1318 expand_as_6" [id=1318, type=expand_as]; -"1319 div_6" [id=1319, type=div]; -"1320 quantize_per_tensor_default_21" [id=1320, type=quantize_per_tensor]; -"1321 dequantize_per_tensor_default_21" [id=1321, type=dequantize_per_tensor]; -"1322 linalg_vector_norm_7" [id=1322, type=linalg_vector_norm]; -"1323 clamp_min_7" [id=1323, type=clamp_min]; -"1324 expand_as_7" [id=1324, type=expand_as]; -"1325 div_7" [id=1325, type=div]; -"1326 quantize_per_tensor_default_22" [id=1326, type=quantize_per_tensor]; -"1327 dequantize_per_tensor_default_22" [id=1327, type=dequantize_per_tensor]; -"1328 transpose_6" [id=1328, type=transpose]; -"1329 matmul_6" [id=1329, type=matmul]; -"1330 clamp_3" [id=1330, type=clamp]; -"1331 exp_3" [id=1331, type=exp]; -"1332 mul_7" [id=1332, type=mul]; -"1333 add_10" [id=1333, type=add]; -"1334 new_zeros_1" [id=1334, type=new_zeros]; -"1335 lift_fresh_copy_9" [id=1335, type=lift_fresh_copy]; -"1336 slice_43" [id=1336, type=slice]; -"1337 slice_44" [id=1337, type=slice]; -"1338 fill__9" [id=1338, type=fill_]; -"1339 lift_fresh_copy_10" [id=1339, type=lift_fresh_copy]; -"1340 slice_45" [id=1340, type=slice]; -"1341 slice_46" [id=1341, type=slice]; -"1342 fill__10" [id=1342, type=fill_]; -"1343 lift_fresh_copy_11" [id=1343, type=lift_fresh_copy]; -"1344 slice_47" [id=1344, type=slice]; -"1345 slice_48" [id=1345, type=slice]; -"1346 fill__11" [id=1346, type=fill_]; -"1347 lift_fresh_copy_12" [id=1347, type=lift_fresh_copy]; -"1348 slice_49" [id=1348, type=slice]; -"1349 slice_50" [id=1349, type=slice]; -"1350 fill__12" [id=1350, type=fill_]; -"1351 lift_fresh_copy_13" [id=1351, type=lift_fresh_copy]; -"1352 slice_51" [id=1352, type=slice]; -"1353 slice_52" [id=1353, type=slice]; -"1354 fill__13" [id=1354, type=fill_]; -"1355 lift_fresh_copy_14" [id=1355, type=lift_fresh_copy]; -"1356 slice_53" [id=1356, type=slice]; -"1357 slice_54" [id=1357, type=slice]; -"1358 fill__14" [id=1358, type=fill_]; -"1359 lift_fresh_copy_15" [id=1359, type=lift_fresh_copy]; -"1360 slice_55" [id=1360, type=slice]; -"1361 slice_56" [id=1361, type=slice]; -"1362 fill__15" [id=1362, type=fill_]; -"1363 lift_fresh_copy_16" [id=1363, type=lift_fresh_copy]; -"1364 slice_57" [id=1364, type=slice]; -"1365 slice_58" [id=1365, type=slice]; -"1366 fill__16" [id=1366, type=fill_]; -"1367 lift_fresh_copy_17" [id=1367, type=lift_fresh_copy]; -"1368 slice_59" [id=1368, type=slice]; -"1369 slice_60" [id=1369, type=slice]; -"1370 fill__17" [id=1370, type=fill_]; -"1371 view_18" [id=1371, type=view]; -"1372 permute_17" [id=1372, type=permute]; -"1373 reshape_15" [id=1373, type=reshape]; -"1374 unsqueeze_8" [id=1374, type=unsqueeze]; -"1375 unsqueeze_9" [id=1375, type=unsqueeze]; -"1376 sub_1" [id=1376, type=sub]; -"1377 ne_1" [id=1377, type=ne]; -"1378 masked_fill_2" [id=1378, type=masked_fill]; -"1379 eq_1" [id=1379, type=eq]; -"1380 masked_fill_3" [id=1380, type=masked_fill]; -"1381 view_19" [id=1381, type=view]; -"1382 unsqueeze_10" [id=1382, type=unsqueeze]; -"1383 unsqueeze_11" [id=1383, type=unsqueeze]; -"1384 add_11" [id=1384, type=add]; -"1385 view_20" [id=1385, type=view]; -"1386 softmax_3" [id=1386, type=softmax]; -"1387 dropout_12" [id=1387, type=dropout]; -"1388 matmul_7" [id=1388, type=matmul]; -"1389 quantize_per_tensor_default_23" [id=1389, type=quantize_per_tensor]; -"1390 dequantize_per_tensor_default_23" [id=1390, type=dequantize_per_tensor]; -"1391 transpose_7" [id=1391, type=transpose]; -"1392 reshape_16" [id=1392, type=reshape]; -"1393 linear_22" [id=1393, type=linear]; -"1394 dropout_13" [id=1394, type=dropout]; -"1395 view_21" [id=1395, type=view]; -"1396 permute_18" [id=1396, type=permute]; -"1397 reshape_17" [id=1397, type=reshape]; -"1398 roll_3" [id=1398, type=roll]; -"1399 slice_61" [id=1399, type=slice]; -"1400 slice_62" [id=1400, type=slice]; -"1401 slice_63" [id=1401, type=slice]; -"1402 slice_64" [id=1402, type=slice]; -"1403 contiguous_5" [id=1403, type=contiguous]; -"1404 layer_norm_8" [id=1404, type=layer_norm]; -"1405 add_12" [id=1405, type=add]; -"1406 quantize_per_tensor_default_24" [id=1406, type=quantize_per_tensor]; -"1407 dequantize_per_tensor_default_24" [id=1407, type=dequantize_per_tensor]; -"1408 linear_23" [id=1408, type=linear]; -"1409 gelu_3" [id=1409, type=gelu]; -"1410 quantize_per_tensor_default_25" [id=1410, type=quantize_per_tensor]; -"1411 dequantize_per_tensor_default_25" [id=1411, type=dequantize_per_tensor]; -"1412 dropout_14" [id=1412, type=dropout]; -"1413 linear_24" [id=1413, type=linear]; -"1414 dropout_15" [id=1414, type=dropout]; -"1415 layer_norm_9" [id=1415, type=layer_norm]; -"1416 add_13" [id=1416, type=add]; -"1417 quantize_per_tensor_default_26" [id=1417, type=quantize_per_tensor]; -"1418 dequantize_per_tensor_default_26" [id=1418, type=dequantize_per_tensor]; -"1419 pad_5" [id=1419, type=pad]; -"1420 slice_65" [id=1420, type=slice]; -"1421 slice_66" [id=1421, type=slice]; -"1422 slice_67" [id=1422, type=slice]; -"1423 slice_68" [id=1423, type=slice]; -"1424 slice_69" [id=1424, type=slice]; -"1425 slice_70" [id=1425, type=slice]; -"1426 slice_71" [id=1426, type=slice]; -"1427 slice_72" [id=1427, type=slice]; -"1428 slice_73" [id=1428, type=slice]; -"1429 slice_74" [id=1429, type=slice]; -"1430 slice_75" [id=1430, type=slice]; -"1431 slice_76" [id=1431, type=slice]; -"1432 cat_1" [id=1432, type=cat]; -"1433 linear_25" [id=1433, type=linear]; -"1434 layer_norm_10" [id=1434, type=layer_norm]; -"1435 quantize_per_tensor_default_27" [id=1435, type=quantize_per_tensor]; -"1436 dequantize_per_tensor_default_27" [id=1436, type=dequantize_per_tensor]; -"1437 linear_26" [id=1437, type=linear]; -"1438 relu__4" [id=1438, type=relu_]; -"1439 linear_27" [id=1439, type=linear]; -"1440 view_22" [id=1440, type=view]; -"1441 index_4" [id=1441, type=index]; -"1442 view_23" [id=1442, type=view]; -"1443 permute_19" [id=1443, type=permute]; -"1444 contiguous_6" [id=1444, type=contiguous]; -"1445 unsqueeze_12" [id=1445, type=unsqueeze]; -"1446 sigmoid_4" [id=1446, type=sigmoid]; -"1447 mul_8" [id=1447, type=mul]; -"1448 pad_6" [id=1448, type=pad]; -"1449 view_24" [id=1449, type=view]; -"1450 permute_20" [id=1450, type=permute]; -"1451 reshape_18" [id=1451, type=reshape]; -"1452 clone_4" [id=1452, type=clone]; -"1453 slice_77" [id=1453, type=slice]; -"1454 zero__4" [id=1454, type=zero_]; -"1455 linear_28" [id=1455, type=linear]; -"1456 reshape_19" [id=1456, type=reshape]; -"1457 permute_21" [id=1457, type=permute]; -"1458 select_12" [id=1458, type=select]; -"1459 select_13" [id=1459, type=select]; -"1460 select_14" [id=1460, type=select]; -"1461 linalg_vector_norm_8" [id=1461, type=linalg_vector_norm]; -"1462 clamp_min_8" [id=1462, type=clamp_min]; -"1463 expand_as_8" [id=1463, type=expand_as]; -"1464 div_8" [id=1464, type=div]; -"1465 quantize_per_tensor_default_28" [id=1465, type=quantize_per_tensor]; -"1466 dequantize_per_tensor_default_28" [id=1466, type=dequantize_per_tensor]; -"1467 linalg_vector_norm_9" [id=1467, type=linalg_vector_norm]; -"1468 clamp_min_9" [id=1468, type=clamp_min]; -"1469 expand_as_9" [id=1469, type=expand_as]; -"1470 div_9" [id=1470, type=div]; -"1471 quantize_per_tensor_default_29" [id=1471, type=quantize_per_tensor]; -"1472 dequantize_per_tensor_default_29" [id=1472, type=dequantize_per_tensor]; -"1473 transpose_8" [id=1473, type=transpose]; -"1474 matmul_8" [id=1474, type=matmul]; -"1475 clamp_4" [id=1475, type=clamp]; -"1476 exp_4" [id=1476, type=exp]; -"1477 mul_9" [id=1477, type=mul]; -"1478 add_14" [id=1478, type=add]; -"1479 softmax_4" [id=1479, type=softmax]; -"1480 dropout_16" [id=1480, type=dropout]; -"1481 matmul_9" [id=1481, type=matmul]; -"1482 quantize_per_tensor_default_30" [id=1482, type=quantize_per_tensor]; -"1483 dequantize_per_tensor_default_30" [id=1483, type=dequantize_per_tensor]; -"1484 transpose_9" [id=1484, type=transpose]; -"1485 reshape_20" [id=1485, type=reshape]; -"1486 linear_29" [id=1486, type=linear]; -"1487 dropout_17" [id=1487, type=dropout]; -"1488 view_25" [id=1488, type=view]; -"1489 permute_22" [id=1489, type=permute]; -"1490 reshape_21" [id=1490, type=reshape]; -"1491 slice_78" [id=1491, type=slice]; -"1492 slice_79" [id=1492, type=slice]; -"1493 slice_80" [id=1493, type=slice]; -"1494 slice_81" [id=1494, type=slice]; -"1495 contiguous_7" [id=1495, type=contiguous]; -"1496 layer_norm_11" [id=1496, type=layer_norm]; -"1497 add_15" [id=1497, type=add]; -"1498 quantize_per_tensor_default_31" [id=1498, type=quantize_per_tensor]; -"1499 dequantize_per_tensor_default_31" [id=1499, type=dequantize_per_tensor]; -"1500 linear_30" [id=1500, type=linear]; -"1501 gelu_4" [id=1501, type=gelu]; -"1502 quantize_per_tensor_default_32" [id=1502, type=quantize_per_tensor]; -"1503 dequantize_per_tensor_default_32" [id=1503, type=dequantize_per_tensor]; -"1504 dropout_18" [id=1504, type=dropout]; -"1505 linear_31" [id=1505, type=linear]; -"1506 dropout_19" [id=1506, type=dropout]; -"1507 layer_norm_12" [id=1507, type=layer_norm]; -"1508 add_16" [id=1508, type=add]; -"1509 linear_32" [id=1509, type=linear]; -"1510 relu__5" [id=1510, type=relu_]; -"1511 linear_33" [id=1511, type=linear]; -"1512 view_26" [id=1512, type=view]; -"1513 index_5" [id=1513, type=index]; -"1514 view_27" [id=1514, type=view]; -"1515 permute_23" [id=1515, type=permute]; -"1516 contiguous_8" [id=1516, type=contiguous]; -"1517 unsqueeze_13" [id=1517, type=unsqueeze]; -"1518 sigmoid_5" [id=1518, type=sigmoid]; -"1519 mul_10" [id=1519, type=mul]; -"1520 pad_7" [id=1520, type=pad]; -"1521 roll_4" [id=1521, type=roll]; -"1522 view_28" [id=1522, type=view]; -"1523 permute_24" [id=1523, type=permute]; -"1524 reshape_22" [id=1524, type=reshape]; -"1525 quantize_per_tensor_default_33" [id=1525, type=quantize_per_tensor]; -"1526 dequantize_per_tensor_default_33" [id=1526, type=dequantize_per_tensor]; -"1527 clone_5" [id=1527, type=clone]; -"1528 slice_82" [id=1528, type=slice]; -"1529 zero__5" [id=1529, type=zero_]; -"1530 linear_34" [id=1530, type=linear]; -"1531 reshape_23" [id=1531, type=reshape]; -"1532 permute_25" [id=1532, type=permute]; -"1533 select_15" [id=1533, type=select]; -"1534 select_16" [id=1534, type=select]; -"1535 select_17" [id=1535, type=select]; -"1536 linalg_vector_norm_10" [id=1536, type=linalg_vector_norm]; -"1537 clamp_min_10" [id=1537, type=clamp_min]; -"1538 expand_as_10" [id=1538, type=expand_as]; -"1539 div_10" [id=1539, type=div]; -"1540 quantize_per_tensor_default_34" [id=1540, type=quantize_per_tensor]; -"1541 dequantize_per_tensor_default_34" [id=1541, type=dequantize_per_tensor]; -"1542 linalg_vector_norm_11" [id=1542, type=linalg_vector_norm]; -"1543 clamp_min_11" [id=1543, type=clamp_min]; -"1544 expand_as_11" [id=1544, type=expand_as]; -"1545 div_11" [id=1545, type=div]; -"1546 quantize_per_tensor_default_35" [id=1546, type=quantize_per_tensor]; -"1547 dequantize_per_tensor_default_35" [id=1547, type=dequantize_per_tensor]; -"1548 transpose_10" [id=1548, type=transpose]; -"1549 matmul_10" [id=1549, type=matmul]; -"1550 clamp_5" [id=1550, type=clamp]; -"1551 exp_5" [id=1551, type=exp]; -"1552 mul_11" [id=1552, type=mul]; -"1553 add_17" [id=1553, type=add]; -"1554 new_zeros_2" [id=1554, type=new_zeros]; -"1555 lift_fresh_copy_18" [id=1555, type=lift_fresh_copy]; -"1556 slice_83" [id=1556, type=slice]; -"1557 slice_84" [id=1557, type=slice]; -"1558 fill__18" [id=1558, type=fill_]; -"1559 lift_fresh_copy_19" [id=1559, type=lift_fresh_copy]; -"1560 slice_85" [id=1560, type=slice]; -"1561 slice_86" [id=1561, type=slice]; -"1562 fill__19" [id=1562, type=fill_]; -"1563 lift_fresh_copy_20" [id=1563, type=lift_fresh_copy]; -"1564 slice_87" [id=1564, type=slice]; -"1565 slice_88" [id=1565, type=slice]; -"1566 fill__20" [id=1566, type=fill_]; -"1567 lift_fresh_copy_21" [id=1567, type=lift_fresh_copy]; -"1568 slice_89" [id=1568, type=slice]; -"1569 slice_90" [id=1569, type=slice]; -"1570 fill__21" [id=1570, type=fill_]; -"1571 lift_fresh_copy_22" [id=1571, type=lift_fresh_copy]; -"1572 slice_91" [id=1572, type=slice]; -"1573 slice_92" [id=1573, type=slice]; -"1574 fill__22" [id=1574, type=fill_]; -"1575 lift_fresh_copy_23" [id=1575, type=lift_fresh_copy]; -"1576 slice_93" [id=1576, type=slice]; -"1577 slice_94" [id=1577, type=slice]; -"1578 fill__23" [id=1578, type=fill_]; -"1579 lift_fresh_copy_24" [id=1579, type=lift_fresh_copy]; -"1580 slice_95" [id=1580, type=slice]; -"1581 slice_96" [id=1581, type=slice]; -"1582 fill__24" [id=1582, type=fill_]; -"1583 lift_fresh_copy_25" [id=1583, type=lift_fresh_copy]; -"1584 slice_97" [id=1584, type=slice]; -"1585 slice_98" [id=1585, type=slice]; -"1586 fill__25" [id=1586, type=fill_]; -"1587 lift_fresh_copy_26" [id=1587, type=lift_fresh_copy]; -"1588 slice_99" [id=1588, type=slice]; -"1589 slice_100" [id=1589, type=slice]; -"1590 fill__26" [id=1590, type=fill_]; -"1591 view_29" [id=1591, type=view]; -"1592 permute_26" [id=1592, type=permute]; -"1593 reshape_24" [id=1593, type=reshape]; -"1594 unsqueeze_14" [id=1594, type=unsqueeze]; -"1595 unsqueeze_15" [id=1595, type=unsqueeze]; -"1596 sub_2" [id=1596, type=sub]; -"1597 ne_2" [id=1597, type=ne]; -"1598 masked_fill_4" [id=1598, type=masked_fill]; -"1599 eq_2" [id=1599, type=eq]; -"1600 masked_fill_5" [id=1600, type=masked_fill]; -"1601 view_30" [id=1601, type=view]; -"1602 unsqueeze_16" [id=1602, type=unsqueeze]; -"1603 unsqueeze_17" [id=1603, type=unsqueeze]; -"1604 add_18" [id=1604, type=add]; -"1605 view_31" [id=1605, type=view]; -"1606 softmax_5" [id=1606, type=softmax]; -"1607 dropout_20" [id=1607, type=dropout]; -"1608 matmul_11" [id=1608, type=matmul]; -"1609 quantize_per_tensor_default_36" [id=1609, type=quantize_per_tensor]; -"1610 dequantize_per_tensor_default_36" [id=1610, type=dequantize_per_tensor]; -"1611 transpose_11" [id=1611, type=transpose]; -"1612 reshape_25" [id=1612, type=reshape]; -"1613 linear_35" [id=1613, type=linear]; -"1614 dropout_21" [id=1614, type=dropout]; -"1615 view_32" [id=1615, type=view]; -"1616 permute_27" [id=1616, type=permute]; -"1617 reshape_26" [id=1617, type=reshape]; -"1618 roll_5" [id=1618, type=roll]; -"1619 slice_101" [id=1619, type=slice]; -"1620 slice_102" [id=1620, type=slice]; -"1621 slice_103" [id=1621, type=slice]; -"1622 slice_104" [id=1622, type=slice]; -"1623 contiguous_9" [id=1623, type=contiguous]; -"1624 layer_norm_13" [id=1624, type=layer_norm]; -"1625 add_19" [id=1625, type=add]; -"1626 quantize_per_tensor_default_37" [id=1626, type=quantize_per_tensor]; -"1627 dequantize_per_tensor_default_37" [id=1627, type=dequantize_per_tensor]; -"1628 linear_36" [id=1628, type=linear]; -"1629 gelu_5" [id=1629, type=gelu]; -"1630 quantize_per_tensor_default_38" [id=1630, type=quantize_per_tensor]; -"1631 dequantize_per_tensor_default_38" [id=1631, type=dequantize_per_tensor]; -"1632 dropout_22" [id=1632, type=dropout]; -"1633 linear_37" [id=1633, type=linear]; -"1634 dropout_23" [id=1634, type=dropout]; -"1635 layer_norm_14" [id=1635, type=layer_norm]; -"1636 add_20" [id=1636, type=add]; -"1637 quantize_per_tensor_default_39" [id=1637, type=quantize_per_tensor]; -"1638 dequantize_per_tensor_default_39" [id=1638, type=dequantize_per_tensor]; -"1639 linear_38" [id=1639, type=linear]; -"1640 relu__6" [id=1640, type=relu_]; -"1641 linear_39" [id=1641, type=linear]; -"1642 view_33" [id=1642, type=view]; -"1643 index_6" [id=1643, type=index]; -"1644 view_34" [id=1644, type=view]; -"1645 permute_28" [id=1645, type=permute]; -"1646 contiguous_10" [id=1646, type=contiguous]; -"1647 unsqueeze_18" [id=1647, type=unsqueeze]; -"1648 sigmoid_6" [id=1648, type=sigmoid]; -"1649 mul_12" [id=1649, type=mul]; -"1650 pad_8" [id=1650, type=pad]; -"1651 view_35" [id=1651, type=view]; -"1652 permute_29" [id=1652, type=permute]; -"1653 reshape_27" [id=1653, type=reshape]; -"1654 clone_6" [id=1654, type=clone]; -"1655 slice_105" [id=1655, type=slice]; -"1656 zero__6" [id=1656, type=zero_]; -"1657 linear_40" [id=1657, type=linear]; -"1658 reshape_28" [id=1658, type=reshape]; -"1659 permute_30" [id=1659, type=permute]; -"1660 select_18" [id=1660, type=select]; -"1661 select_19" [id=1661, type=select]; -"1662 select_20" [id=1662, type=select]; -"1663 linalg_vector_norm_12" [id=1663, type=linalg_vector_norm]; -"1664 clamp_min_12" [id=1664, type=clamp_min]; -"1665 expand_as_12" [id=1665, type=expand_as]; -"1666 div_12" [id=1666, type=div]; -"1667 quantize_per_tensor_default_40" [id=1667, type=quantize_per_tensor]; -"1668 dequantize_per_tensor_default_40" [id=1668, type=dequantize_per_tensor]; -"1669 linalg_vector_norm_13" [id=1669, type=linalg_vector_norm]; -"1670 clamp_min_13" [id=1670, type=clamp_min]; -"1671 expand_as_13" [id=1671, type=expand_as]; -"1672 div_13" [id=1672, type=div]; -"1673 quantize_per_tensor_default_41" [id=1673, type=quantize_per_tensor]; -"1674 dequantize_per_tensor_default_41" [id=1674, type=dequantize_per_tensor]; -"1675 transpose_12" [id=1675, type=transpose]; -"1676 matmul_12" [id=1676, type=matmul]; -"1677 clamp_6" [id=1677, type=clamp]; -"1678 exp_6" [id=1678, type=exp]; -"1679 mul_13" [id=1679, type=mul]; -"1680 add_21" [id=1680, type=add]; -"1681 softmax_6" [id=1681, type=softmax]; -"1682 dropout_24" [id=1682, type=dropout]; -"1683 matmul_13" [id=1683, type=matmul]; -"1684 quantize_per_tensor_default_42" [id=1684, type=quantize_per_tensor]; -"1685 dequantize_per_tensor_default_42" [id=1685, type=dequantize_per_tensor]; -"1686 transpose_13" [id=1686, type=transpose]; -"1687 reshape_29" [id=1687, type=reshape]; -"1688 linear_41" [id=1688, type=linear]; -"1689 dropout_25" [id=1689, type=dropout]; -"1690 view_36" [id=1690, type=view]; -"1691 permute_31" [id=1691, type=permute]; -"1692 reshape_30" [id=1692, type=reshape]; -"1693 slice_106" [id=1693, type=slice]; -"1694 slice_107" [id=1694, type=slice]; -"1695 slice_108" [id=1695, type=slice]; -"1696 slice_109" [id=1696, type=slice]; -"1697 contiguous_11" [id=1697, type=contiguous]; -"1698 layer_norm_15" [id=1698, type=layer_norm]; -"1699 add_22" [id=1699, type=add]; -"1700 quantize_per_tensor_default_43" [id=1700, type=quantize_per_tensor]; -"1701 dequantize_per_tensor_default_43" [id=1701, type=dequantize_per_tensor]; -"1702 linear_42" [id=1702, type=linear]; -"1703 gelu_6" [id=1703, type=gelu]; -"1704 quantize_per_tensor_default_44" [id=1704, type=quantize_per_tensor]; -"1705 dequantize_per_tensor_default_44" [id=1705, type=dequantize_per_tensor]; -"1706 dropout_26" [id=1706, type=dropout]; -"1707 linear_43" [id=1707, type=linear]; -"1708 dropout_27" [id=1708, type=dropout]; -"1709 layer_norm_16" [id=1709, type=layer_norm]; -"1710 add_23" [id=1710, type=add]; -"1711 linear_44" [id=1711, type=linear]; -"1712 relu__7" [id=1712, type=relu_]; -"1713 linear_45" [id=1713, type=linear]; -"1714 view_37" [id=1714, type=view]; -"1715 index_7" [id=1715, type=index]; -"1716 view_38" [id=1716, type=view]; -"1717 permute_32" [id=1717, type=permute]; -"1718 contiguous_12" [id=1718, type=contiguous]; -"1719 unsqueeze_19" [id=1719, type=unsqueeze]; -"1720 sigmoid_7" [id=1720, type=sigmoid]; -"1721 mul_14" [id=1721, type=mul]; -"1722 pad_9" [id=1722, type=pad]; -"1723 roll_6" [id=1723, type=roll]; -"1724 view_39" [id=1724, type=view]; -"1725 permute_33" [id=1725, type=permute]; -"1726 reshape_31" [id=1726, type=reshape]; -"1727 quantize_per_tensor_default_45" [id=1727, type=quantize_per_tensor]; -"1728 dequantize_per_tensor_default_45" [id=1728, type=dequantize_per_tensor]; -"1729 clone_7" [id=1729, type=clone]; -"1730 slice_110" [id=1730, type=slice]; -"1731 zero__7" [id=1731, type=zero_]; -"1732 linear_46" [id=1732, type=linear]; -"1733 reshape_32" [id=1733, type=reshape]; -"1734 permute_34" [id=1734, type=permute]; -"1735 select_21" [id=1735, type=select]; -"1736 select_22" [id=1736, type=select]; -"1737 select_23" [id=1737, type=select]; -"1738 linalg_vector_norm_14" [id=1738, type=linalg_vector_norm]; -"1739 clamp_min_14" [id=1739, type=clamp_min]; -"1740 expand_as_14" [id=1740, type=expand_as]; -"1741 div_14" [id=1741, type=div]; -"1742 quantize_per_tensor_default_46" [id=1742, type=quantize_per_tensor]; -"1743 dequantize_per_tensor_default_46" [id=1743, type=dequantize_per_tensor]; -"1744 linalg_vector_norm_15" [id=1744, type=linalg_vector_norm]; -"1745 clamp_min_15" [id=1745, type=clamp_min]; -"1746 expand_as_15" [id=1746, type=expand_as]; -"1747 div_15" [id=1747, type=div]; -"1748 quantize_per_tensor_default_47" [id=1748, type=quantize_per_tensor]; -"1749 dequantize_per_tensor_default_47" [id=1749, type=dequantize_per_tensor]; -"1750 transpose_14" [id=1750, type=transpose]; -"1751 matmul_14" [id=1751, type=matmul]; -"1752 clamp_7" [id=1752, type=clamp]; -"1753 exp_7" [id=1753, type=exp]; -"1754 mul_15" [id=1754, type=mul]; -"1755 add_24" [id=1755, type=add]; -"1756 new_zeros_3" [id=1756, type=new_zeros]; -"1757 lift_fresh_copy_27" [id=1757, type=lift_fresh_copy]; -"1758 slice_111" [id=1758, type=slice]; -"1759 slice_112" [id=1759, type=slice]; -"1760 fill__27" [id=1760, type=fill_]; -"1761 lift_fresh_copy_28" [id=1761, type=lift_fresh_copy]; -"1762 slice_113" [id=1762, type=slice]; -"1763 slice_114" [id=1763, type=slice]; -"1764 fill__28" [id=1764, type=fill_]; -"1765 lift_fresh_copy_29" [id=1765, type=lift_fresh_copy]; -"1766 slice_115" [id=1766, type=slice]; -"1767 slice_116" [id=1767, type=slice]; -"1768 fill__29" [id=1768, type=fill_]; -"1769 lift_fresh_copy_30" [id=1769, type=lift_fresh_copy]; -"1770 slice_117" [id=1770, type=slice]; -"1771 slice_118" [id=1771, type=slice]; -"1772 fill__30" [id=1772, type=fill_]; -"1773 lift_fresh_copy_31" [id=1773, type=lift_fresh_copy]; -"1774 slice_119" [id=1774, type=slice]; -"1775 slice_120" [id=1775, type=slice]; -"1776 fill__31" [id=1776, type=fill_]; -"1777 lift_fresh_copy_32" [id=1777, type=lift_fresh_copy]; -"1778 slice_121" [id=1778, type=slice]; -"1779 slice_122" [id=1779, type=slice]; -"1780 fill__32" [id=1780, type=fill_]; -"1781 lift_fresh_copy_33" [id=1781, type=lift_fresh_copy]; -"1782 slice_123" [id=1782, type=slice]; -"1783 slice_124" [id=1783, type=slice]; -"1784 fill__33" [id=1784, type=fill_]; -"1785 lift_fresh_copy_34" [id=1785, type=lift_fresh_copy]; -"1786 slice_125" [id=1786, type=slice]; -"1787 slice_126" [id=1787, type=slice]; -"1788 fill__34" [id=1788, type=fill_]; -"1789 lift_fresh_copy_35" [id=1789, type=lift_fresh_copy]; -"1790 slice_127" [id=1790, type=slice]; -"1791 slice_128" [id=1791, type=slice]; -"1792 fill__35" [id=1792, type=fill_]; -"1793 view_40" [id=1793, type=view]; -"1794 permute_35" [id=1794, type=permute]; -"1795 reshape_33" [id=1795, type=reshape]; -"1796 unsqueeze_20" [id=1796, type=unsqueeze]; -"1797 unsqueeze_21" [id=1797, type=unsqueeze]; -"1798 sub_3" [id=1798, type=sub]; -"1799 ne_3" [id=1799, type=ne]; -"1800 masked_fill_6" [id=1800, type=masked_fill]; -"1801 eq_3" [id=1801, type=eq]; -"1802 masked_fill_7" [id=1802, type=masked_fill]; -"1803 view_41" [id=1803, type=view]; -"1804 unsqueeze_22" [id=1804, type=unsqueeze]; -"1805 unsqueeze_23" [id=1805, type=unsqueeze]; -"1806 add_25" [id=1806, type=add]; -"1807 view_42" [id=1807, type=view]; -"1808 softmax_7" [id=1808, type=softmax]; -"1809 dropout_28" [id=1809, type=dropout]; -"1810 matmul_15" [id=1810, type=matmul]; -"1811 quantize_per_tensor_default_48" [id=1811, type=quantize_per_tensor]; -"1812 dequantize_per_tensor_default_48" [id=1812, type=dequantize_per_tensor]; -"1813 transpose_15" [id=1813, type=transpose]; -"1814 reshape_34" [id=1814, type=reshape]; -"1815 linear_47" [id=1815, type=linear]; -"1816 dropout_29" [id=1816, type=dropout]; -"1817 view_43" [id=1817, type=view]; -"1818 permute_36" [id=1818, type=permute]; -"1819 reshape_35" [id=1819, type=reshape]; -"1820 roll_7" [id=1820, type=roll]; -"1821 slice_129" [id=1821, type=slice]; -"1822 slice_130" [id=1822, type=slice]; -"1823 slice_131" [id=1823, type=slice]; -"1824 slice_132" [id=1824, type=slice]; -"1825 contiguous_13" [id=1825, type=contiguous]; -"1826 layer_norm_17" [id=1826, type=layer_norm]; -"1827 add_26" [id=1827, type=add]; -"1828 quantize_per_tensor_default_49" [id=1828, type=quantize_per_tensor]; -"1829 dequantize_per_tensor_default_49" [id=1829, type=dequantize_per_tensor]; -"1830 linear_48" [id=1830, type=linear]; -"1831 gelu_7" [id=1831, type=gelu]; -"1832 quantize_per_tensor_default_50" [id=1832, type=quantize_per_tensor]; -"1833 dequantize_per_tensor_default_50" [id=1833, type=dequantize_per_tensor]; -"1834 dropout_30" [id=1834, type=dropout]; -"1835 linear_49" [id=1835, type=linear]; -"1836 dropout_31" [id=1836, type=dropout]; -"1837 layer_norm_18" [id=1837, type=layer_norm]; -"1838 add_27" [id=1838, type=add]; -"1839 quantize_per_tensor_default_51" [id=1839, type=quantize_per_tensor]; -"1840 dequantize_per_tensor_default_51" [id=1840, type=dequantize_per_tensor]; -"1841 linear_50" [id=1841, type=linear]; -"1842 relu__8" [id=1842, type=relu_]; -"1843 linear_51" [id=1843, type=linear]; -"1844 view_44" [id=1844, type=view]; -"1845 index_8" [id=1845, type=index]; -"1846 view_45" [id=1846, type=view]; -"1847 permute_37" [id=1847, type=permute]; -"1848 contiguous_14" [id=1848, type=contiguous]; -"1849 unsqueeze_24" [id=1849, type=unsqueeze]; -"1850 sigmoid_8" [id=1850, type=sigmoid]; -"1851 mul_16" [id=1851, type=mul]; -"1852 pad_10" [id=1852, type=pad]; -"1853 view_46" [id=1853, type=view]; -"1854 permute_38" [id=1854, type=permute]; -"1855 reshape_36" [id=1855, type=reshape]; -"1856 clone_8" [id=1856, type=clone]; -"1857 slice_133" [id=1857, type=slice]; -"1858 zero__8" [id=1858, type=zero_]; -"1859 linear_52" [id=1859, type=linear]; -"1860 reshape_37" [id=1860, type=reshape]; -"1861 permute_39" [id=1861, type=permute]; -"1862 select_24" [id=1862, type=select]; -"1863 select_25" [id=1863, type=select]; -"1864 select_26" [id=1864, type=select]; -"1865 linalg_vector_norm_16" [id=1865, type=linalg_vector_norm]; -"1866 clamp_min_16" [id=1866, type=clamp_min]; -"1867 expand_as_16" [id=1867, type=expand_as]; -"1868 div_16" [id=1868, type=div]; -"1869 quantize_per_tensor_default_52" [id=1869, type=quantize_per_tensor]; -"1870 dequantize_per_tensor_default_52" [id=1870, type=dequantize_per_tensor]; -"1871 linalg_vector_norm_17" [id=1871, type=linalg_vector_norm]; -"1872 clamp_min_17" [id=1872, type=clamp_min]; -"1873 expand_as_17" [id=1873, type=expand_as]; -"1874 div_17" [id=1874, type=div]; -"1875 quantize_per_tensor_default_53" [id=1875, type=quantize_per_tensor]; -"1876 dequantize_per_tensor_default_53" [id=1876, type=dequantize_per_tensor]; -"1877 transpose_16" [id=1877, type=transpose]; -"1878 matmul_16" [id=1878, type=matmul]; -"1879 clamp_8" [id=1879, type=clamp]; -"1880 exp_8" [id=1880, type=exp]; -"1881 mul_17" [id=1881, type=mul]; -"1882 add_28" [id=1882, type=add]; -"1883 softmax_8" [id=1883, type=softmax]; -"1884 dropout_32" [id=1884, type=dropout]; -"1885 matmul_17" [id=1885, type=matmul]; -"1886 quantize_per_tensor_default_54" [id=1886, type=quantize_per_tensor]; -"1887 dequantize_per_tensor_default_54" [id=1887, type=dequantize_per_tensor]; -"1888 transpose_17" [id=1888, type=transpose]; -"1889 reshape_38" [id=1889, type=reshape]; -"1890 linear_53" [id=1890, type=linear]; -"1891 dropout_33" [id=1891, type=dropout]; -"1892 view_47" [id=1892, type=view]; -"1893 permute_40" [id=1893, type=permute]; -"1894 reshape_39" [id=1894, type=reshape]; -"1895 slice_134" [id=1895, type=slice]; -"1896 slice_135" [id=1896, type=slice]; -"1897 slice_136" [id=1897, type=slice]; -"1898 slice_137" [id=1898, type=slice]; -"1899 contiguous_15" [id=1899, type=contiguous]; -"1900 layer_norm_19" [id=1900, type=layer_norm]; -"1901 add_29" [id=1901, type=add]; -"1902 quantize_per_tensor_default_55" [id=1902, type=quantize_per_tensor]; -"1903 dequantize_per_tensor_default_55" [id=1903, type=dequantize_per_tensor]; -"1904 linear_54" [id=1904, type=linear]; -"1905 gelu_8" [id=1905, type=gelu]; -"1906 quantize_per_tensor_default_56" [id=1906, type=quantize_per_tensor]; -"1907 dequantize_per_tensor_default_56" [id=1907, type=dequantize_per_tensor]; -"1908 dropout_34" [id=1908, type=dropout]; -"1909 linear_55" [id=1909, type=linear]; -"1910 dropout_35" [id=1910, type=dropout]; -"1911 layer_norm_20" [id=1911, type=layer_norm]; -"1912 add_30" [id=1912, type=add]; -"1913 linear_56" [id=1913, type=linear]; -"1914 relu__9" [id=1914, type=relu_]; -"1915 linear_57" [id=1915, type=linear]; -"1916 view_48" [id=1916, type=view]; -"1917 index_9" [id=1917, type=index]; -"1918 view_49" [id=1918, type=view]; -"1919 permute_41" [id=1919, type=permute]; -"1920 contiguous_16" [id=1920, type=contiguous]; -"1921 unsqueeze_25" [id=1921, type=unsqueeze]; -"1922 sigmoid_9" [id=1922, type=sigmoid]; -"1923 mul_18" [id=1923, type=mul]; -"1924 pad_11" [id=1924, type=pad]; -"1925 roll_8" [id=1925, type=roll]; -"1926 view_50" [id=1926, type=view]; -"1927 permute_42" [id=1927, type=permute]; -"1928 reshape_40" [id=1928, type=reshape]; -"1929 quantize_per_tensor_default_57" [id=1929, type=quantize_per_tensor]; -"1930 dequantize_per_tensor_default_57" [id=1930, type=dequantize_per_tensor]; -"1931 clone_9" [id=1931, type=clone]; -"1932 slice_138" [id=1932, type=slice]; -"1933 zero__9" [id=1933, type=zero_]; -"1934 linear_58" [id=1934, type=linear]; -"1935 reshape_41" [id=1935, type=reshape]; -"1936 permute_43" [id=1936, type=permute]; -"1937 select_27" [id=1937, type=select]; -"1938 select_28" [id=1938, type=select]; -"1939 select_29" [id=1939, type=select]; -"1940 linalg_vector_norm_18" [id=1940, type=linalg_vector_norm]; -"1941 clamp_min_18" [id=1941, type=clamp_min]; -"1942 expand_as_18" [id=1942, type=expand_as]; -"1943 div_18" [id=1943, type=div]; -"1944 quantize_per_tensor_default_58" [id=1944, type=quantize_per_tensor]; -"1945 dequantize_per_tensor_default_58" [id=1945, type=dequantize_per_tensor]; -"1946 linalg_vector_norm_19" [id=1946, type=linalg_vector_norm]; -"1947 clamp_min_19" [id=1947, type=clamp_min]; -"1948 expand_as_19" [id=1948, type=expand_as]; -"1949 div_19" [id=1949, type=div]; -"1950 quantize_per_tensor_default_59" [id=1950, type=quantize_per_tensor]; -"1951 dequantize_per_tensor_default_59" [id=1951, type=dequantize_per_tensor]; -"1952 transpose_18" [id=1952, type=transpose]; -"1953 matmul_18" [id=1953, type=matmul]; -"1954 clamp_9" [id=1954, type=clamp]; -"1955 exp_9" [id=1955, type=exp]; -"1956 mul_19" [id=1956, type=mul]; -"1957 add_31" [id=1957, type=add]; -"1958 new_zeros_4" [id=1958, type=new_zeros]; -"1959 lift_fresh_copy_36" [id=1959, type=lift_fresh_copy]; -"1960 slice_139" [id=1960, type=slice]; -"1961 slice_140" [id=1961, type=slice]; -"1962 fill__36" [id=1962, type=fill_]; -"1963 lift_fresh_copy_37" [id=1963, type=lift_fresh_copy]; -"1964 slice_141" [id=1964, type=slice]; -"1965 slice_142" [id=1965, type=slice]; -"1966 fill__37" [id=1966, type=fill_]; -"1967 lift_fresh_copy_38" [id=1967, type=lift_fresh_copy]; -"1968 slice_143" [id=1968, type=slice]; -"1969 slice_144" [id=1969, type=slice]; -"1970 fill__38" [id=1970, type=fill_]; -"1971 lift_fresh_copy_39" [id=1971, type=lift_fresh_copy]; -"1972 slice_145" [id=1972, type=slice]; -"1973 slice_146" [id=1973, type=slice]; -"1974 fill__39" [id=1974, type=fill_]; -"1975 lift_fresh_copy_40" [id=1975, type=lift_fresh_copy]; -"1976 slice_147" [id=1976, type=slice]; -"1977 slice_148" [id=1977, type=slice]; -"1978 fill__40" [id=1978, type=fill_]; -"1979 lift_fresh_copy_41" [id=1979, type=lift_fresh_copy]; -"1980 slice_149" [id=1980, type=slice]; -"1981 slice_150" [id=1981, type=slice]; -"1982 fill__41" [id=1982, type=fill_]; -"1983 lift_fresh_copy_42" [id=1983, type=lift_fresh_copy]; -"1984 slice_151" [id=1984, type=slice]; -"1985 slice_152" [id=1985, type=slice]; -"1986 fill__42" [id=1986, type=fill_]; -"1987 lift_fresh_copy_43" [id=1987, type=lift_fresh_copy]; -"1988 slice_153" [id=1988, type=slice]; -"1989 slice_154" [id=1989, type=slice]; -"1990 fill__43" [id=1990, type=fill_]; -"1991 lift_fresh_copy_44" [id=1991, type=lift_fresh_copy]; -"1992 slice_155" [id=1992, type=slice]; -"1993 slice_156" [id=1993, type=slice]; -"1994 fill__44" [id=1994, type=fill_]; -"1995 view_51" [id=1995, type=view]; -"1996 permute_44" [id=1996, type=permute]; -"1997 reshape_42" [id=1997, type=reshape]; -"1998 unsqueeze_26" [id=1998, type=unsqueeze]; -"1999 unsqueeze_27" [id=1999, type=unsqueeze]; -"2000 sub_4" [id=2000, type=sub]; -"2001 ne_4" [id=2001, type=ne]; -"2002 masked_fill_8" [id=2002, type=masked_fill]; -"2003 eq_4" [id=2003, type=eq]; -"2004 masked_fill_9" [id=2004, type=masked_fill]; -"2005 view_52" [id=2005, type=view]; -"2006 unsqueeze_28" [id=2006, type=unsqueeze]; -"2007 unsqueeze_29" [id=2007, type=unsqueeze]; -"2008 add_32" [id=2008, type=add]; -"2009 view_53" [id=2009, type=view]; -"2010 softmax_9" [id=2010, type=softmax]; -"2011 dropout_36" [id=2011, type=dropout]; -"2012 matmul_19" [id=2012, type=matmul]; -"2013 quantize_per_tensor_default_60" [id=2013, type=quantize_per_tensor]; -"2014 dequantize_per_tensor_default_60" [id=2014, type=dequantize_per_tensor]; -"2015 transpose_19" [id=2015, type=transpose]; -"2016 reshape_43" [id=2016, type=reshape]; -"2017 linear_59" [id=2017, type=linear]; -"2018 dropout_37" [id=2018, type=dropout]; -"2019 view_54" [id=2019, type=view]; -"2020 permute_45" [id=2020, type=permute]; -"2021 reshape_44" [id=2021, type=reshape]; -"2022 roll_9" [id=2022, type=roll]; -"2023 slice_157" [id=2023, type=slice]; -"2024 slice_158" [id=2024, type=slice]; -"2025 slice_159" [id=2025, type=slice]; -"2026 slice_160" [id=2026, type=slice]; -"2027 contiguous_17" [id=2027, type=contiguous]; -"2028 layer_norm_21" [id=2028, type=layer_norm]; -"2029 add_33" [id=2029, type=add]; -"2030 quantize_per_tensor_default_61" [id=2030, type=quantize_per_tensor]; -"2031 dequantize_per_tensor_default_61" [id=2031, type=dequantize_per_tensor]; -"2032 linear_60" [id=2032, type=linear]; -"2033 gelu_9" [id=2033, type=gelu]; -"2034 quantize_per_tensor_default_62" [id=2034, type=quantize_per_tensor]; -"2035 dequantize_per_tensor_default_62" [id=2035, type=dequantize_per_tensor]; -"2036 dropout_38" [id=2036, type=dropout]; -"2037 linear_61" [id=2037, type=linear]; -"2038 dropout_39" [id=2038, type=dropout]; -"2039 layer_norm_22" [id=2039, type=layer_norm]; -"2040 add_34" [id=2040, type=add]; -"2041 quantize_per_tensor_default_63" [id=2041, type=quantize_per_tensor]; -"2042 dequantize_per_tensor_default_63" [id=2042, type=dequantize_per_tensor]; -"2043 linear_62" [id=2043, type=linear]; -"2044 relu__10" [id=2044, type=relu_]; -"2045 linear_63" [id=2045, type=linear]; -"2046 view_55" [id=2046, type=view]; -"2047 index_10" [id=2047, type=index]; -"2048 view_56" [id=2048, type=view]; -"2049 permute_46" [id=2049, type=permute]; -"2050 contiguous_18" [id=2050, type=contiguous]; -"2051 unsqueeze_30" [id=2051, type=unsqueeze]; -"2052 sigmoid_10" [id=2052, type=sigmoid]; -"2053 mul_20" [id=2053, type=mul]; -"2054 pad_12" [id=2054, type=pad]; -"2055 view_57" [id=2055, type=view]; -"2056 permute_47" [id=2056, type=permute]; -"2057 reshape_45" [id=2057, type=reshape]; -"2058 clone_10" [id=2058, type=clone]; -"2059 slice_161" [id=2059, type=slice]; -"2060 zero__10" [id=2060, type=zero_]; -"2061 linear_64" [id=2061, type=linear]; -"2062 reshape_46" [id=2062, type=reshape]; -"2063 permute_48" [id=2063, type=permute]; -"2064 select_30" [id=2064, type=select]; -"2065 select_31" [id=2065, type=select]; -"2066 select_32" [id=2066, type=select]; -"2067 linalg_vector_norm_20" [id=2067, type=linalg_vector_norm]; -"2068 clamp_min_20" [id=2068, type=clamp_min]; -"2069 expand_as_20" [id=2069, type=expand_as]; -"2070 div_20" [id=2070, type=div]; -"2071 quantize_per_tensor_default_64" [id=2071, type=quantize_per_tensor]; -"2072 dequantize_per_tensor_default_64" [id=2072, type=dequantize_per_tensor]; -"2073 linalg_vector_norm_21" [id=2073, type=linalg_vector_norm]; -"2074 clamp_min_21" [id=2074, type=clamp_min]; -"2075 expand_as_21" [id=2075, type=expand_as]; -"2076 div_21" [id=2076, type=div]; -"2077 quantize_per_tensor_default_65" [id=2077, type=quantize_per_tensor]; -"2078 dequantize_per_tensor_default_65" [id=2078, type=dequantize_per_tensor]; -"2079 transpose_20" [id=2079, type=transpose]; -"2080 matmul_20" [id=2080, type=matmul]; -"2081 clamp_10" [id=2081, type=clamp]; -"2082 exp_10" [id=2082, type=exp]; -"2083 mul_21" [id=2083, type=mul]; -"2084 add_35" [id=2084, type=add]; -"2085 softmax_10" [id=2085, type=softmax]; -"2086 dropout_40" [id=2086, type=dropout]; -"2087 matmul_21" [id=2087, type=matmul]; -"2088 quantize_per_tensor_default_66" [id=2088, type=quantize_per_tensor]; -"2089 dequantize_per_tensor_default_66" [id=2089, type=dequantize_per_tensor]; -"2090 transpose_21" [id=2090, type=transpose]; -"2091 reshape_47" [id=2091, type=reshape]; -"2092 linear_65" [id=2092, type=linear]; -"2093 dropout_41" [id=2093, type=dropout]; -"2094 view_58" [id=2094, type=view]; -"2095 permute_49" [id=2095, type=permute]; -"2096 reshape_48" [id=2096, type=reshape]; -"2097 slice_162" [id=2097, type=slice]; -"2098 slice_163" [id=2098, type=slice]; -"2099 slice_164" [id=2099, type=slice]; -"2100 slice_165" [id=2100, type=slice]; -"2101 contiguous_19" [id=2101, type=contiguous]; -"2102 layer_norm_23" [id=2102, type=layer_norm]; -"2103 add_36" [id=2103, type=add]; -"2104 quantize_per_tensor_default_67" [id=2104, type=quantize_per_tensor]; -"2105 dequantize_per_tensor_default_67" [id=2105, type=dequantize_per_tensor]; -"2106 linear_66" [id=2106, type=linear]; -"2107 gelu_10" [id=2107, type=gelu]; -"2108 quantize_per_tensor_default_68" [id=2108, type=quantize_per_tensor]; -"2109 dequantize_per_tensor_default_68" [id=2109, type=dequantize_per_tensor]; -"2110 dropout_42" [id=2110, type=dropout]; -"2111 linear_67" [id=2111, type=linear]; -"2112 dropout_43" [id=2112, type=dropout]; -"2113 layer_norm_24" [id=2113, type=layer_norm]; -"2114 add_37" [id=2114, type=add]; -"2115 linear_68" [id=2115, type=linear]; -"2116 relu__11" [id=2116, type=relu_]; -"2117 linear_69" [id=2117, type=linear]; -"2118 view_59" [id=2118, type=view]; -"2119 index_11" [id=2119, type=index]; -"2120 view_60" [id=2120, type=view]; -"2121 permute_50" [id=2121, type=permute]; -"2122 contiguous_20" [id=2122, type=contiguous]; -"2123 unsqueeze_31" [id=2123, type=unsqueeze]; -"2124 sigmoid_11" [id=2124, type=sigmoid]; -"2125 mul_22" [id=2125, type=mul]; -"2126 pad_13" [id=2126, type=pad]; -"2127 roll_10" [id=2127, type=roll]; -"2128 view_61" [id=2128, type=view]; -"2129 permute_51" [id=2129, type=permute]; -"2130 reshape_49" [id=2130, type=reshape]; -"2131 quantize_per_tensor_default_69" [id=2131, type=quantize_per_tensor]; -"2132 dequantize_per_tensor_default_69" [id=2132, type=dequantize_per_tensor]; -"2133 clone_11" [id=2133, type=clone]; -"2134 slice_166" [id=2134, type=slice]; -"2135 zero__11" [id=2135, type=zero_]; -"2136 linear_70" [id=2136, type=linear]; -"2137 reshape_50" [id=2137, type=reshape]; -"2138 permute_52" [id=2138, type=permute]; -"2139 select_33" [id=2139, type=select]; -"2140 select_34" [id=2140, type=select]; -"2141 select_35" [id=2141, type=select]; -"2142 linalg_vector_norm_22" [id=2142, type=linalg_vector_norm]; -"2143 clamp_min_22" [id=2143, type=clamp_min]; -"2144 expand_as_22" [id=2144, type=expand_as]; -"2145 div_22" [id=2145, type=div]; -"2146 quantize_per_tensor_default_70" [id=2146, type=quantize_per_tensor]; -"2147 dequantize_per_tensor_default_70" [id=2147, type=dequantize_per_tensor]; -"2148 linalg_vector_norm_23" [id=2148, type=linalg_vector_norm]; -"2149 clamp_min_23" [id=2149, type=clamp_min]; -"2150 expand_as_23" [id=2150, type=expand_as]; -"2151 div_23" [id=2151, type=div]; -"2152 quantize_per_tensor_default_71" [id=2152, type=quantize_per_tensor]; -"2153 dequantize_per_tensor_default_71" [id=2153, type=dequantize_per_tensor]; -"2154 transpose_22" [id=2154, type=transpose]; -"2155 matmul_22" [id=2155, type=matmul]; -"2156 clamp_11" [id=2156, type=clamp]; -"2157 exp_11" [id=2157, type=exp]; -"2158 mul_23" [id=2158, type=mul]; -"2159 add_38" [id=2159, type=add]; -"2160 new_zeros_5" [id=2160, type=new_zeros]; -"2161 lift_fresh_copy_45" [id=2161, type=lift_fresh_copy]; -"2162 slice_167" [id=2162, type=slice]; -"2163 slice_168" [id=2163, type=slice]; -"2164 fill__45" [id=2164, type=fill_]; -"2165 lift_fresh_copy_46" [id=2165, type=lift_fresh_copy]; -"2166 slice_169" [id=2166, type=slice]; -"2167 slice_170" [id=2167, type=slice]; -"2168 fill__46" [id=2168, type=fill_]; -"2169 lift_fresh_copy_47" [id=2169, type=lift_fresh_copy]; -"2170 slice_171" [id=2170, type=slice]; -"2171 slice_172" [id=2171, type=slice]; -"2172 fill__47" [id=2172, type=fill_]; -"2173 lift_fresh_copy_48" [id=2173, type=lift_fresh_copy]; -"2174 slice_173" [id=2174, type=slice]; -"2175 slice_174" [id=2175, type=slice]; -"2176 fill__48" [id=2176, type=fill_]; -"2177 lift_fresh_copy_49" [id=2177, type=lift_fresh_copy]; -"2178 slice_175" [id=2178, type=slice]; -"2179 slice_176" [id=2179, type=slice]; -"2180 fill__49" [id=2180, type=fill_]; -"2181 lift_fresh_copy_50" [id=2181, type=lift_fresh_copy]; -"2182 slice_177" [id=2182, type=slice]; -"2183 slice_178" [id=2183, type=slice]; -"2184 fill__50" [id=2184, type=fill_]; -"2185 lift_fresh_copy_51" [id=2185, type=lift_fresh_copy]; -"2186 slice_179" [id=2186, type=slice]; -"2187 slice_180" [id=2187, type=slice]; -"2188 fill__51" [id=2188, type=fill_]; -"2189 lift_fresh_copy_52" [id=2189, type=lift_fresh_copy]; -"2190 slice_181" [id=2190, type=slice]; -"2191 slice_182" [id=2191, type=slice]; -"2192 fill__52" [id=2192, type=fill_]; -"2193 lift_fresh_copy_53" [id=2193, type=lift_fresh_copy]; -"2194 slice_183" [id=2194, type=slice]; -"2195 slice_184" [id=2195, type=slice]; -"2196 fill__53" [id=2196, type=fill_]; -"2197 view_62" [id=2197, type=view]; -"2198 permute_53" [id=2198, type=permute]; -"2199 reshape_51" [id=2199, type=reshape]; -"2200 unsqueeze_32" [id=2200, type=unsqueeze]; -"2201 unsqueeze_33" [id=2201, type=unsqueeze]; -"2202 sub_5" [id=2202, type=sub]; -"2203 ne_5" [id=2203, type=ne]; -"2204 masked_fill_10" [id=2204, type=masked_fill]; -"2205 eq_5" [id=2205, type=eq]; -"2206 masked_fill_11" [id=2206, type=masked_fill]; -"2207 view_63" [id=2207, type=view]; -"2208 unsqueeze_34" [id=2208, type=unsqueeze]; -"2209 unsqueeze_35" [id=2209, type=unsqueeze]; -"2210 add_39" [id=2210, type=add]; -"2211 view_64" [id=2211, type=view]; -"2212 softmax_11" [id=2212, type=softmax]; -"2213 dropout_44" [id=2213, type=dropout]; -"2214 matmul_23" [id=2214, type=matmul]; -"2215 quantize_per_tensor_default_72" [id=2215, type=quantize_per_tensor]; -"2216 dequantize_per_tensor_default_72" [id=2216, type=dequantize_per_tensor]; -"2217 transpose_23" [id=2217, type=transpose]; -"2218 reshape_52" [id=2218, type=reshape]; -"2219 linear_71" [id=2219, type=linear]; -"2220 dropout_45" [id=2220, type=dropout]; -"2221 view_65" [id=2221, type=view]; -"2222 permute_54" [id=2222, type=permute]; -"2223 reshape_53" [id=2223, type=reshape]; -"2224 roll_11" [id=2224, type=roll]; -"2225 slice_185" [id=2225, type=slice]; -"2226 slice_186" [id=2226, type=slice]; -"2227 slice_187" [id=2227, type=slice]; -"2228 slice_188" [id=2228, type=slice]; -"2229 contiguous_21" [id=2229, type=contiguous]; -"2230 layer_norm_25" [id=2230, type=layer_norm]; -"2231 add_40" [id=2231, type=add]; -"2232 quantize_per_tensor_default_73" [id=2232, type=quantize_per_tensor]; -"2233 dequantize_per_tensor_default_73" [id=2233, type=dequantize_per_tensor]; -"2234 linear_72" [id=2234, type=linear]; -"2235 gelu_11" [id=2235, type=gelu]; -"2236 quantize_per_tensor_default_74" [id=2236, type=quantize_per_tensor]; -"2237 dequantize_per_tensor_default_74" [id=2237, type=dequantize_per_tensor]; -"2238 dropout_46" [id=2238, type=dropout]; -"2239 linear_73" [id=2239, type=linear]; -"2240 dropout_47" [id=2240, type=dropout]; -"2241 layer_norm_26" [id=2241, type=layer_norm]; -"2242 add_41" [id=2242, type=add]; -"2243 quantize_per_tensor_default_75" [id=2243, type=quantize_per_tensor]; -"2244 dequantize_per_tensor_default_75" [id=2244, type=dequantize_per_tensor]; -"2245 linear_74" [id=2245, type=linear]; -"2246 relu__12" [id=2246, type=relu_]; -"2247 linear_75" [id=2247, type=linear]; -"2248 view_66" [id=2248, type=view]; -"2249 index_12" [id=2249, type=index]; -"2250 view_67" [id=2250, type=view]; -"2251 permute_55" [id=2251, type=permute]; -"2252 contiguous_22" [id=2252, type=contiguous]; -"2253 unsqueeze_36" [id=2253, type=unsqueeze]; -"2254 sigmoid_12" [id=2254, type=sigmoid]; -"2255 mul_24" [id=2255, type=mul]; -"2256 pad_14" [id=2256, type=pad]; -"2257 view_68" [id=2257, type=view]; -"2258 permute_56" [id=2258, type=permute]; -"2259 reshape_54" [id=2259, type=reshape]; -"2260 clone_12" [id=2260, type=clone]; -"2261 slice_189" [id=2261, type=slice]; -"2262 zero__12" [id=2262, type=zero_]; -"2263 linear_76" [id=2263, type=linear]; -"2264 reshape_55" [id=2264, type=reshape]; -"2265 permute_57" [id=2265, type=permute]; -"2266 select_36" [id=2266, type=select]; -"2267 select_37" [id=2267, type=select]; -"2268 select_38" [id=2268, type=select]; -"2269 linalg_vector_norm_24" [id=2269, type=linalg_vector_norm]; -"2270 clamp_min_24" [id=2270, type=clamp_min]; -"2271 expand_as_24" [id=2271, type=expand_as]; -"2272 div_24" [id=2272, type=div]; -"2273 quantize_per_tensor_default_76" [id=2273, type=quantize_per_tensor]; -"2274 dequantize_per_tensor_default_76" [id=2274, type=dequantize_per_tensor]; -"2275 linalg_vector_norm_25" [id=2275, type=linalg_vector_norm]; -"2276 clamp_min_25" [id=2276, type=clamp_min]; -"2277 expand_as_25" [id=2277, type=expand_as]; -"2278 div_25" [id=2278, type=div]; -"2279 quantize_per_tensor_default_77" [id=2279, type=quantize_per_tensor]; -"2280 dequantize_per_tensor_default_77" [id=2280, type=dequantize_per_tensor]; -"2281 transpose_24" [id=2281, type=transpose]; -"2282 matmul_24" [id=2282, type=matmul]; -"2283 clamp_12" [id=2283, type=clamp]; -"2284 exp_12" [id=2284, type=exp]; -"2285 mul_25" [id=2285, type=mul]; -"2286 add_42" [id=2286, type=add]; -"2287 softmax_12" [id=2287, type=softmax]; -"2288 dropout_48" [id=2288, type=dropout]; -"2289 matmul_25" [id=2289, type=matmul]; -"2290 quantize_per_tensor_default_78" [id=2290, type=quantize_per_tensor]; -"2291 dequantize_per_tensor_default_78" [id=2291, type=dequantize_per_tensor]; -"2292 transpose_25" [id=2292, type=transpose]; -"2293 reshape_56" [id=2293, type=reshape]; -"2294 linear_77" [id=2294, type=linear]; -"2295 dropout_49" [id=2295, type=dropout]; -"2296 view_69" [id=2296, type=view]; -"2297 permute_58" [id=2297, type=permute]; -"2298 reshape_57" [id=2298, type=reshape]; -"2299 slice_190" [id=2299, type=slice]; -"2300 slice_191" [id=2300, type=slice]; -"2301 slice_192" [id=2301, type=slice]; -"2302 slice_193" [id=2302, type=slice]; -"2303 contiguous_23" [id=2303, type=contiguous]; -"2304 layer_norm_27" [id=2304, type=layer_norm]; -"2305 add_43" [id=2305, type=add]; -"2306 quantize_per_tensor_default_79" [id=2306, type=quantize_per_tensor]; -"2307 dequantize_per_tensor_default_79" [id=2307, type=dequantize_per_tensor]; -"2308 linear_78" [id=2308, type=linear]; -"2309 gelu_12" [id=2309, type=gelu]; -"2310 quantize_per_tensor_default_80" [id=2310, type=quantize_per_tensor]; -"2311 dequantize_per_tensor_default_80" [id=2311, type=dequantize_per_tensor]; -"2312 dropout_50" [id=2312, type=dropout]; -"2313 linear_79" [id=2313, type=linear]; -"2314 dropout_51" [id=2314, type=dropout]; -"2315 layer_norm_28" [id=2315, type=layer_norm]; -"2316 add_44" [id=2316, type=add]; -"2317 linear_80" [id=2317, type=linear]; -"2318 relu__13" [id=2318, type=relu_]; -"2319 linear_81" [id=2319, type=linear]; -"2320 view_70" [id=2320, type=view]; -"2321 index_13" [id=2321, type=index]; -"2322 view_71" [id=2322, type=view]; -"2323 permute_59" [id=2323, type=permute]; -"2324 contiguous_24" [id=2324, type=contiguous]; -"2325 unsqueeze_37" [id=2325, type=unsqueeze]; -"2326 sigmoid_13" [id=2326, type=sigmoid]; -"2327 mul_26" [id=2327, type=mul]; -"2328 pad_15" [id=2328, type=pad]; -"2329 roll_12" [id=2329, type=roll]; -"2330 view_72" [id=2330, type=view]; -"2331 permute_60" [id=2331, type=permute]; -"2332 reshape_58" [id=2332, type=reshape]; -"2333 quantize_per_tensor_default_81" [id=2333, type=quantize_per_tensor]; -"2334 dequantize_per_tensor_default_81" [id=2334, type=dequantize_per_tensor]; -"2335 clone_13" [id=2335, type=clone]; -"2336 slice_194" [id=2336, type=slice]; -"2337 zero__13" [id=2337, type=zero_]; -"2338 linear_82" [id=2338, type=linear]; -"2339 reshape_59" [id=2339, type=reshape]; -"2340 permute_61" [id=2340, type=permute]; -"2341 select_39" [id=2341, type=select]; -"2342 select_40" [id=2342, type=select]; -"2343 select_41" [id=2343, type=select]; -"2344 linalg_vector_norm_26" [id=2344, type=linalg_vector_norm]; -"2345 clamp_min_26" [id=2345, type=clamp_min]; -"2346 expand_as_26" [id=2346, type=expand_as]; -"2347 div_26" [id=2347, type=div]; -"2348 quantize_per_tensor_default_82" [id=2348, type=quantize_per_tensor]; -"2349 dequantize_per_tensor_default_82" [id=2349, type=dequantize_per_tensor]; -"2350 linalg_vector_norm_27" [id=2350, type=linalg_vector_norm]; -"2351 clamp_min_27" [id=2351, type=clamp_min]; -"2352 expand_as_27" [id=2352, type=expand_as]; -"2353 div_27" [id=2353, type=div]; -"2354 quantize_per_tensor_default_83" [id=2354, type=quantize_per_tensor]; -"2355 dequantize_per_tensor_default_83" [id=2355, type=dequantize_per_tensor]; -"2356 transpose_26" [id=2356, type=transpose]; -"2357 matmul_26" [id=2357, type=matmul]; -"2358 clamp_13" [id=2358, type=clamp]; -"2359 exp_13" [id=2359, type=exp]; -"2360 mul_27" [id=2360, type=mul]; -"2361 add_45" [id=2361, type=add]; -"2362 new_zeros_6" [id=2362, type=new_zeros]; -"2363 lift_fresh_copy_54" [id=2363, type=lift_fresh_copy]; -"2364 slice_195" [id=2364, type=slice]; -"2365 slice_196" [id=2365, type=slice]; -"2366 fill__54" [id=2366, type=fill_]; -"2367 lift_fresh_copy_55" [id=2367, type=lift_fresh_copy]; -"2368 slice_197" [id=2368, type=slice]; -"2369 slice_198" [id=2369, type=slice]; -"2370 fill__55" [id=2370, type=fill_]; -"2371 lift_fresh_copy_56" [id=2371, type=lift_fresh_copy]; -"2372 slice_199" [id=2372, type=slice]; -"2373 slice_200" [id=2373, type=slice]; -"2374 fill__56" [id=2374, type=fill_]; -"2375 lift_fresh_copy_57" [id=2375, type=lift_fresh_copy]; -"2376 slice_201" [id=2376, type=slice]; -"2377 slice_202" [id=2377, type=slice]; -"2378 fill__57" [id=2378, type=fill_]; -"2379 lift_fresh_copy_58" [id=2379, type=lift_fresh_copy]; -"2380 slice_203" [id=2380, type=slice]; -"2381 slice_204" [id=2381, type=slice]; -"2382 fill__58" [id=2382, type=fill_]; -"2383 lift_fresh_copy_59" [id=2383, type=lift_fresh_copy]; -"2384 slice_205" [id=2384, type=slice]; -"2385 slice_206" [id=2385, type=slice]; -"2386 fill__59" [id=2386, type=fill_]; -"2387 lift_fresh_copy_60" [id=2387, type=lift_fresh_copy]; -"2388 slice_207" [id=2388, type=slice]; -"2389 slice_208" [id=2389, type=slice]; -"2390 fill__60" [id=2390, type=fill_]; -"2391 lift_fresh_copy_61" [id=2391, type=lift_fresh_copy]; -"2392 slice_209" [id=2392, type=slice]; -"2393 slice_210" [id=2393, type=slice]; -"2394 fill__61" [id=2394, type=fill_]; -"2395 lift_fresh_copy_62" [id=2395, type=lift_fresh_copy]; -"2396 slice_211" [id=2396, type=slice]; -"2397 slice_212" [id=2397, type=slice]; -"2398 fill__62" [id=2398, type=fill_]; -"2399 view_73" [id=2399, type=view]; -"2400 permute_62" [id=2400, type=permute]; -"2401 reshape_60" [id=2401, type=reshape]; -"2402 unsqueeze_38" [id=2402, type=unsqueeze]; -"2403 unsqueeze_39" [id=2403, type=unsqueeze]; -"2404 sub_6" [id=2404, type=sub]; -"2405 ne_6" [id=2405, type=ne]; -"2406 masked_fill_12" [id=2406, type=masked_fill]; -"2407 eq_6" [id=2407, type=eq]; -"2408 masked_fill_13" [id=2408, type=masked_fill]; -"2409 view_74" [id=2409, type=view]; -"2410 unsqueeze_40" [id=2410, type=unsqueeze]; -"2411 unsqueeze_41" [id=2411, type=unsqueeze]; -"2412 add_46" [id=2412, type=add]; -"2413 view_75" [id=2413, type=view]; -"2414 softmax_13" [id=2414, type=softmax]; -"2415 dropout_52" [id=2415, type=dropout]; -"2416 matmul_27" [id=2416, type=matmul]; -"2417 quantize_per_tensor_default_84" [id=2417, type=quantize_per_tensor]; -"2418 dequantize_per_tensor_default_84" [id=2418, type=dequantize_per_tensor]; -"2419 transpose_27" [id=2419, type=transpose]; -"2420 reshape_61" [id=2420, type=reshape]; -"2421 linear_83" [id=2421, type=linear]; -"2422 dropout_53" [id=2422, type=dropout]; -"2423 view_76" [id=2423, type=view]; -"2424 permute_63" [id=2424, type=permute]; -"2425 reshape_62" [id=2425, type=reshape]; -"2426 roll_13" [id=2426, type=roll]; -"2427 slice_213" [id=2427, type=slice]; -"2428 slice_214" [id=2428, type=slice]; -"2429 slice_215" [id=2429, type=slice]; -"2430 slice_216" [id=2430, type=slice]; -"2431 contiguous_25" [id=2431, type=contiguous]; -"2432 layer_norm_29" [id=2432, type=layer_norm]; -"2433 add_47" [id=2433, type=add]; -"2434 quantize_per_tensor_default_85" [id=2434, type=quantize_per_tensor]; -"2435 dequantize_per_tensor_default_85" [id=2435, type=dequantize_per_tensor]; -"2436 linear_84" [id=2436, type=linear]; -"2437 gelu_13" [id=2437, type=gelu]; -"2438 quantize_per_tensor_default_86" [id=2438, type=quantize_per_tensor]; -"2439 dequantize_per_tensor_default_86" [id=2439, type=dequantize_per_tensor]; -"2440 dropout_54" [id=2440, type=dropout]; -"2441 linear_85" [id=2441, type=linear]; -"2442 dropout_55" [id=2442, type=dropout]; -"2443 layer_norm_30" [id=2443, type=layer_norm]; -"2444 add_48" [id=2444, type=add]; -"2445 quantize_per_tensor_default_87" [id=2445, type=quantize_per_tensor]; -"2446 dequantize_per_tensor_default_87" [id=2446, type=dequantize_per_tensor]; -"2447 linear_86" [id=2447, type=linear]; -"2448 relu__14" [id=2448, type=relu_]; -"2449 linear_87" [id=2449, type=linear]; -"2450 view_77" [id=2450, type=view]; -"2451 index_14" [id=2451, type=index]; -"2452 view_78" [id=2452, type=view]; -"2453 permute_64" [id=2453, type=permute]; -"2454 contiguous_26" [id=2454, type=contiguous]; -"2455 unsqueeze_42" [id=2455, type=unsqueeze]; -"2456 sigmoid_14" [id=2456, type=sigmoid]; -"2457 mul_28" [id=2457, type=mul]; -"2458 pad_16" [id=2458, type=pad]; -"2459 view_79" [id=2459, type=view]; -"2460 permute_65" [id=2460, type=permute]; -"2461 reshape_63" [id=2461, type=reshape]; -"2462 clone_14" [id=2462, type=clone]; -"2463 slice_217" [id=2463, type=slice]; -"2464 zero__14" [id=2464, type=zero_]; -"2465 linear_88" [id=2465, type=linear]; -"2466 reshape_64" [id=2466, type=reshape]; -"2467 permute_66" [id=2467, type=permute]; -"2468 select_42" [id=2468, type=select]; -"2469 select_43" [id=2469, type=select]; -"2470 select_44" [id=2470, type=select]; -"2471 linalg_vector_norm_28" [id=2471, type=linalg_vector_norm]; -"2472 clamp_min_28" [id=2472, type=clamp_min]; -"2473 expand_as_28" [id=2473, type=expand_as]; -"2474 div_28" [id=2474, type=div]; -"2475 quantize_per_tensor_default_88" [id=2475, type=quantize_per_tensor]; -"2476 dequantize_per_tensor_default_88" [id=2476, type=dequantize_per_tensor]; -"2477 linalg_vector_norm_29" [id=2477, type=linalg_vector_norm]; -"2478 clamp_min_29" [id=2478, type=clamp_min]; -"2479 expand_as_29" [id=2479, type=expand_as]; -"2480 div_29" [id=2480, type=div]; -"2481 quantize_per_tensor_default_89" [id=2481, type=quantize_per_tensor]; -"2482 dequantize_per_tensor_default_89" [id=2482, type=dequantize_per_tensor]; -"2483 transpose_28" [id=2483, type=transpose]; -"2484 matmul_28" [id=2484, type=matmul]; -"2485 clamp_14" [id=2485, type=clamp]; -"2486 exp_14" [id=2486, type=exp]; -"2487 mul_29" [id=2487, type=mul]; -"2488 add_49" [id=2488, type=add]; -"2489 softmax_14" [id=2489, type=softmax]; -"2490 dropout_56" [id=2490, type=dropout]; -"2491 matmul_29" [id=2491, type=matmul]; -"2492 quantize_per_tensor_default_90" [id=2492, type=quantize_per_tensor]; -"2493 dequantize_per_tensor_default_90" [id=2493, type=dequantize_per_tensor]; -"2494 transpose_29" [id=2494, type=transpose]; -"2495 reshape_65" [id=2495, type=reshape]; -"2496 linear_89" [id=2496, type=linear]; -"2497 dropout_57" [id=2497, type=dropout]; -"2498 view_80" [id=2498, type=view]; -"2499 permute_67" [id=2499, type=permute]; -"2500 reshape_66" [id=2500, type=reshape]; -"2501 slice_218" [id=2501, type=slice]; -"2502 slice_219" [id=2502, type=slice]; -"2503 slice_220" [id=2503, type=slice]; -"2504 slice_221" [id=2504, type=slice]; -"2505 contiguous_27" [id=2505, type=contiguous]; -"2506 layer_norm_31" [id=2506, type=layer_norm]; -"2507 add_50" [id=2507, type=add]; -"2508 quantize_per_tensor_default_91" [id=2508, type=quantize_per_tensor]; -"2509 dequantize_per_tensor_default_91" [id=2509, type=dequantize_per_tensor]; -"2510 linear_90" [id=2510, type=linear]; -"2511 gelu_14" [id=2511, type=gelu]; -"2512 quantize_per_tensor_default_92" [id=2512, type=quantize_per_tensor]; -"2513 dequantize_per_tensor_default_92" [id=2513, type=dequantize_per_tensor]; -"2514 dropout_58" [id=2514, type=dropout]; -"2515 linear_91" [id=2515, type=linear]; -"2516 dropout_59" [id=2516, type=dropout]; -"2517 layer_norm_32" [id=2517, type=layer_norm]; -"2518 add_51" [id=2518, type=add]; -"2519 linear_92" [id=2519, type=linear]; -"2520 relu__15" [id=2520, type=relu_]; -"2521 linear_93" [id=2521, type=linear]; -"2522 view_81" [id=2522, type=view]; -"2523 index_15" [id=2523, type=index]; -"2524 view_82" [id=2524, type=view]; -"2525 permute_68" [id=2525, type=permute]; -"2526 contiguous_28" [id=2526, type=contiguous]; -"2527 unsqueeze_43" [id=2527, type=unsqueeze]; -"2528 sigmoid_15" [id=2528, type=sigmoid]; -"2529 mul_30" [id=2529, type=mul]; -"2530 pad_17" [id=2530, type=pad]; -"2531 roll_14" [id=2531, type=roll]; -"2532 view_83" [id=2532, type=view]; -"2533 permute_69" [id=2533, type=permute]; -"2534 reshape_67" [id=2534, type=reshape]; -"2535 quantize_per_tensor_default_93" [id=2535, type=quantize_per_tensor]; -"2536 dequantize_per_tensor_default_93" [id=2536, type=dequantize_per_tensor]; -"2537 clone_15" [id=2537, type=clone]; -"2538 slice_222" [id=2538, type=slice]; -"2539 zero__15" [id=2539, type=zero_]; -"2540 linear_94" [id=2540, type=linear]; -"2541 reshape_68" [id=2541, type=reshape]; -"2542 permute_70" [id=2542, type=permute]; -"2543 select_45" [id=2543, type=select]; -"2544 select_46" [id=2544, type=select]; -"2545 select_47" [id=2545, type=select]; -"2546 linalg_vector_norm_30" [id=2546, type=linalg_vector_norm]; -"2547 clamp_min_30" [id=2547, type=clamp_min]; -"2548 expand_as_30" [id=2548, type=expand_as]; -"2549 div_30" [id=2549, type=div]; -"2550 quantize_per_tensor_default_94" [id=2550, type=quantize_per_tensor]; -"2551 dequantize_per_tensor_default_94" [id=2551, type=dequantize_per_tensor]; -"2552 linalg_vector_norm_31" [id=2552, type=linalg_vector_norm]; -"2553 clamp_min_31" [id=2553, type=clamp_min]; -"2554 expand_as_31" [id=2554, type=expand_as]; -"2555 div_31" [id=2555, type=div]; -"2556 quantize_per_tensor_default_95" [id=2556, type=quantize_per_tensor]; -"2557 dequantize_per_tensor_default_95" [id=2557, type=dequantize_per_tensor]; -"2558 transpose_30" [id=2558, type=transpose]; -"2559 matmul_30" [id=2559, type=matmul]; -"2560 clamp_15" [id=2560, type=clamp]; -"2561 exp_15" [id=2561, type=exp]; -"2562 mul_31" [id=2562, type=mul]; -"2563 add_52" [id=2563, type=add]; -"2564 new_zeros_7" [id=2564, type=new_zeros]; -"2565 lift_fresh_copy_63" [id=2565, type=lift_fresh_copy]; -"2566 slice_223" [id=2566, type=slice]; -"2567 slice_224" [id=2567, type=slice]; -"2568 fill__63" [id=2568, type=fill_]; -"2569 lift_fresh_copy_64" [id=2569, type=lift_fresh_copy]; -"2570 slice_225" [id=2570, type=slice]; -"2571 slice_226" [id=2571, type=slice]; -"2572 fill__64" [id=2572, type=fill_]; -"2573 lift_fresh_copy_65" [id=2573, type=lift_fresh_copy]; -"2574 slice_227" [id=2574, type=slice]; -"2575 slice_228" [id=2575, type=slice]; -"2576 fill__65" [id=2576, type=fill_]; -"2577 lift_fresh_copy_66" [id=2577, type=lift_fresh_copy]; -"2578 slice_229" [id=2578, type=slice]; -"2579 slice_230" [id=2579, type=slice]; -"2580 fill__66" [id=2580, type=fill_]; -"2581 lift_fresh_copy_67" [id=2581, type=lift_fresh_copy]; -"2582 slice_231" [id=2582, type=slice]; -"2583 slice_232" [id=2583, type=slice]; -"2584 fill__67" [id=2584, type=fill_]; -"2585 lift_fresh_copy_68" [id=2585, type=lift_fresh_copy]; -"2586 slice_233" [id=2586, type=slice]; -"2587 slice_234" [id=2587, type=slice]; -"2588 fill__68" [id=2588, type=fill_]; -"2589 lift_fresh_copy_69" [id=2589, type=lift_fresh_copy]; -"2590 slice_235" [id=2590, type=slice]; -"2591 slice_236" [id=2591, type=slice]; -"2592 fill__69" [id=2592, type=fill_]; -"2593 lift_fresh_copy_70" [id=2593, type=lift_fresh_copy]; -"2594 slice_237" [id=2594, type=slice]; -"2595 slice_238" [id=2595, type=slice]; -"2596 fill__70" [id=2596, type=fill_]; -"2597 lift_fresh_copy_71" [id=2597, type=lift_fresh_copy]; -"2598 slice_239" [id=2598, type=slice]; -"2599 slice_240" [id=2599, type=slice]; -"2600 fill__71" [id=2600, type=fill_]; -"2601 view_84" [id=2601, type=view]; -"2602 permute_71" [id=2602, type=permute]; -"2603 reshape_69" [id=2603, type=reshape]; -"2604 unsqueeze_44" [id=2604, type=unsqueeze]; -"2605 unsqueeze_45" [id=2605, type=unsqueeze]; -"2606 sub_7" [id=2606, type=sub]; -"2607 ne_7" [id=2607, type=ne]; -"2608 masked_fill_14" [id=2608, type=masked_fill]; -"2609 eq_7" [id=2609, type=eq]; -"2610 masked_fill_15" [id=2610, type=masked_fill]; -"2611 view_85" [id=2611, type=view]; -"2612 unsqueeze_46" [id=2612, type=unsqueeze]; -"2613 unsqueeze_47" [id=2613, type=unsqueeze]; -"2614 add_53" [id=2614, type=add]; -"2615 view_86" [id=2615, type=view]; -"2616 softmax_15" [id=2616, type=softmax]; -"2617 dropout_60" [id=2617, type=dropout]; -"2618 matmul_31" [id=2618, type=matmul]; -"2619 quantize_per_tensor_default_96" [id=2619, type=quantize_per_tensor]; -"2620 dequantize_per_tensor_default_96" [id=2620, type=dequantize_per_tensor]; -"2621 transpose_31" [id=2621, type=transpose]; -"2622 reshape_70" [id=2622, type=reshape]; -"2623 linear_95" [id=2623, type=linear]; -"2624 dropout_61" [id=2624, type=dropout]; -"2625 view_87" [id=2625, type=view]; -"2626 permute_72" [id=2626, type=permute]; -"2627 reshape_71" [id=2627, type=reshape]; -"2628 roll_15" [id=2628, type=roll]; -"2629 slice_241" [id=2629, type=slice]; -"2630 slice_242" [id=2630, type=slice]; -"2631 slice_243" [id=2631, type=slice]; -"2632 slice_244" [id=2632, type=slice]; -"2633 contiguous_29" [id=2633, type=contiguous]; -"2634 layer_norm_33" [id=2634, type=layer_norm]; -"2635 add_54" [id=2635, type=add]; -"2636 quantize_per_tensor_default_97" [id=2636, type=quantize_per_tensor]; -"2637 dequantize_per_tensor_default_97" [id=2637, type=dequantize_per_tensor]; -"2638 linear_96" [id=2638, type=linear]; -"2639 gelu_15" [id=2639, type=gelu]; -"2640 quantize_per_tensor_default_98" [id=2640, type=quantize_per_tensor]; -"2641 dequantize_per_tensor_default_98" [id=2641, type=dequantize_per_tensor]; -"2642 dropout_62" [id=2642, type=dropout]; -"2643 linear_97" [id=2643, type=linear]; -"2644 dropout_63" [id=2644, type=dropout]; -"2645 layer_norm_34" [id=2645, type=layer_norm]; -"2646 add_55" [id=2646, type=add]; -"2647 quantize_per_tensor_default_99" [id=2647, type=quantize_per_tensor]; -"2648 dequantize_per_tensor_default_99" [id=2648, type=dequantize_per_tensor]; -"2649 linear_98" [id=2649, type=linear]; -"2650 relu__16" [id=2650, type=relu_]; -"2651 linear_99" [id=2651, type=linear]; -"2652 view_88" [id=2652, type=view]; -"2653 index_16" [id=2653, type=index]; -"2654 view_89" [id=2654, type=view]; -"2655 permute_73" [id=2655, type=permute]; -"2656 contiguous_30" [id=2656, type=contiguous]; -"2657 unsqueeze_48" [id=2657, type=unsqueeze]; -"2658 sigmoid_16" [id=2658, type=sigmoid]; -"2659 mul_32" [id=2659, type=mul]; -"2660 pad_18" [id=2660, type=pad]; -"2661 view_90" [id=2661, type=view]; -"2662 permute_74" [id=2662, type=permute]; -"2663 reshape_72" [id=2663, type=reshape]; -"2664 clone_16" [id=2664, type=clone]; -"2665 slice_245" [id=2665, type=slice]; -"2666 zero__16" [id=2666, type=zero_]; -"2667 linear_100" [id=2667, type=linear]; -"2668 reshape_73" [id=2668, type=reshape]; -"2669 permute_75" [id=2669, type=permute]; -"2670 select_48" [id=2670, type=select]; -"2671 select_49" [id=2671, type=select]; -"2672 select_50" [id=2672, type=select]; -"2673 linalg_vector_norm_32" [id=2673, type=linalg_vector_norm]; -"2674 clamp_min_32" [id=2674, type=clamp_min]; -"2675 expand_as_32" [id=2675, type=expand_as]; -"2676 div_32" [id=2676, type=div]; -"2677 quantize_per_tensor_default_100" [id=2677, type=quantize_per_tensor]; -"2678 dequantize_per_tensor_default_100" [id=2678, type=dequantize_per_tensor]; -"2679 linalg_vector_norm_33" [id=2679, type=linalg_vector_norm]; -"2680 clamp_min_33" [id=2680, type=clamp_min]; -"2681 expand_as_33" [id=2681, type=expand_as]; -"2682 div_33" [id=2682, type=div]; -"2683 quantize_per_tensor_default_101" [id=2683, type=quantize_per_tensor]; -"2684 dequantize_per_tensor_default_101" [id=2684, type=dequantize_per_tensor]; -"2685 transpose_32" [id=2685, type=transpose]; -"2686 matmul_32" [id=2686, type=matmul]; -"2687 clamp_16" [id=2687, type=clamp]; -"2688 exp_16" [id=2688, type=exp]; -"2689 mul_33" [id=2689, type=mul]; -"2690 add_56" [id=2690, type=add]; -"2691 softmax_16" [id=2691, type=softmax]; -"2692 dropout_64" [id=2692, type=dropout]; -"2693 matmul_33" [id=2693, type=matmul]; -"2694 quantize_per_tensor_default_102" [id=2694, type=quantize_per_tensor]; -"2695 dequantize_per_tensor_default_102" [id=2695, type=dequantize_per_tensor]; -"2696 transpose_33" [id=2696, type=transpose]; -"2697 reshape_74" [id=2697, type=reshape]; -"2698 linear_101" [id=2698, type=linear]; -"2699 dropout_65" [id=2699, type=dropout]; -"2700 view_91" [id=2700, type=view]; -"2701 permute_76" [id=2701, type=permute]; -"2702 reshape_75" [id=2702, type=reshape]; -"2703 slice_246" [id=2703, type=slice]; -"2704 slice_247" [id=2704, type=slice]; -"2705 slice_248" [id=2705, type=slice]; -"2706 slice_249" [id=2706, type=slice]; -"2707 contiguous_31" [id=2707, type=contiguous]; -"2708 layer_norm_35" [id=2708, type=layer_norm]; -"2709 add_57" [id=2709, type=add]; -"2710 quantize_per_tensor_default_103" [id=2710, type=quantize_per_tensor]; -"2711 dequantize_per_tensor_default_103" [id=2711, type=dequantize_per_tensor]; -"2712 linear_102" [id=2712, type=linear]; -"2713 gelu_16" [id=2713, type=gelu]; -"2714 quantize_per_tensor_default_104" [id=2714, type=quantize_per_tensor]; -"2715 dequantize_per_tensor_default_104" [id=2715, type=dequantize_per_tensor]; -"2716 dropout_66" [id=2716, type=dropout]; -"2717 linear_103" [id=2717, type=linear]; -"2718 dropout_67" [id=2718, type=dropout]; -"2719 layer_norm_36" [id=2719, type=layer_norm]; -"2720 add_58" [id=2720, type=add]; -"2721 linear_104" [id=2721, type=linear]; -"2722 relu__17" [id=2722, type=relu_]; -"2723 linear_105" [id=2723, type=linear]; -"2724 view_92" [id=2724, type=view]; -"2725 index_17" [id=2725, type=index]; -"2726 view_93" [id=2726, type=view]; -"2727 permute_77" [id=2727, type=permute]; -"2728 contiguous_32" [id=2728, type=contiguous]; -"2729 unsqueeze_49" [id=2729, type=unsqueeze]; -"2730 sigmoid_17" [id=2730, type=sigmoid]; -"2731 mul_34" [id=2731, type=mul]; -"2732 pad_19" [id=2732, type=pad]; -"2733 roll_16" [id=2733, type=roll]; -"2734 view_94" [id=2734, type=view]; -"2735 permute_78" [id=2735, type=permute]; -"2736 reshape_76" [id=2736, type=reshape]; -"2737 quantize_per_tensor_default_105" [id=2737, type=quantize_per_tensor]; -"2738 dequantize_per_tensor_default_105" [id=2738, type=dequantize_per_tensor]; -"2739 clone_17" [id=2739, type=clone]; -"2740 slice_250" [id=2740, type=slice]; -"2741 zero__17" [id=2741, type=zero_]; -"2742 linear_106" [id=2742, type=linear]; -"2743 reshape_77" [id=2743, type=reshape]; -"2744 permute_79" [id=2744, type=permute]; -"2745 select_51" [id=2745, type=select]; -"2746 select_52" [id=2746, type=select]; -"2747 select_53" [id=2747, type=select]; -"2748 linalg_vector_norm_34" [id=2748, type=linalg_vector_norm]; -"2749 clamp_min_34" [id=2749, type=clamp_min]; -"2750 expand_as_34" [id=2750, type=expand_as]; -"2751 div_34" [id=2751, type=div]; -"2752 quantize_per_tensor_default_106" [id=2752, type=quantize_per_tensor]; -"2753 dequantize_per_tensor_default_106" [id=2753, type=dequantize_per_tensor]; -"2754 linalg_vector_norm_35" [id=2754, type=linalg_vector_norm]; -"2755 clamp_min_35" [id=2755, type=clamp_min]; -"2756 expand_as_35" [id=2756, type=expand_as]; -"2757 div_35" [id=2757, type=div]; -"2758 quantize_per_tensor_default_107" [id=2758, type=quantize_per_tensor]; -"2759 dequantize_per_tensor_default_107" [id=2759, type=dequantize_per_tensor]; -"2760 transpose_34" [id=2760, type=transpose]; -"2761 matmul_34" [id=2761, type=matmul]; -"2762 clamp_17" [id=2762, type=clamp]; -"2763 exp_17" [id=2763, type=exp]; -"2764 mul_35" [id=2764, type=mul]; -"2765 add_59" [id=2765, type=add]; -"2766 new_zeros_8" [id=2766, type=new_zeros]; -"2767 lift_fresh_copy_72" [id=2767, type=lift_fresh_copy]; -"2768 slice_251" [id=2768, type=slice]; -"2769 slice_252" [id=2769, type=slice]; -"2770 fill__72" [id=2770, type=fill_]; -"2771 lift_fresh_copy_73" [id=2771, type=lift_fresh_copy]; -"2772 slice_253" [id=2772, type=slice]; -"2773 slice_254" [id=2773, type=slice]; -"2774 fill__73" [id=2774, type=fill_]; -"2775 lift_fresh_copy_74" [id=2775, type=lift_fresh_copy]; -"2776 slice_255" [id=2776, type=slice]; -"2777 slice_256" [id=2777, type=slice]; -"2778 fill__74" [id=2778, type=fill_]; -"2779 lift_fresh_copy_75" [id=2779, type=lift_fresh_copy]; -"2780 slice_257" [id=2780, type=slice]; -"2781 slice_258" [id=2781, type=slice]; -"2782 fill__75" [id=2782, type=fill_]; -"2783 lift_fresh_copy_76" [id=2783, type=lift_fresh_copy]; -"2784 slice_259" [id=2784, type=slice]; -"2785 slice_260" [id=2785, type=slice]; -"2786 fill__76" [id=2786, type=fill_]; -"2787 lift_fresh_copy_77" [id=2787, type=lift_fresh_copy]; -"2788 slice_261" [id=2788, type=slice]; -"2789 slice_262" [id=2789, type=slice]; -"2790 fill__77" [id=2790, type=fill_]; -"2791 lift_fresh_copy_78" [id=2791, type=lift_fresh_copy]; -"2792 slice_263" [id=2792, type=slice]; -"2793 slice_264" [id=2793, type=slice]; -"2794 fill__78" [id=2794, type=fill_]; -"2795 lift_fresh_copy_79" [id=2795, type=lift_fresh_copy]; -"2796 slice_265" [id=2796, type=slice]; -"2797 slice_266" [id=2797, type=slice]; -"2798 fill__79" [id=2798, type=fill_]; -"2799 lift_fresh_copy_80" [id=2799, type=lift_fresh_copy]; -"2800 slice_267" [id=2800, type=slice]; -"2801 slice_268" [id=2801, type=slice]; -"2802 fill__80" [id=2802, type=fill_]; -"2803 view_95" [id=2803, type=view]; -"2804 permute_80" [id=2804, type=permute]; -"2805 reshape_78" [id=2805, type=reshape]; -"2806 unsqueeze_50" [id=2806, type=unsqueeze]; -"2807 unsqueeze_51" [id=2807, type=unsqueeze]; -"2808 sub_8" [id=2808, type=sub]; -"2809 ne_8" [id=2809, type=ne]; -"2810 masked_fill_16" [id=2810, type=masked_fill]; -"2811 eq_8" [id=2811, type=eq]; -"2812 masked_fill_17" [id=2812, type=masked_fill]; -"2813 view_96" [id=2813, type=view]; -"2814 unsqueeze_52" [id=2814, type=unsqueeze]; -"2815 unsqueeze_53" [id=2815, type=unsqueeze]; -"2816 add_60" [id=2816, type=add]; -"2817 view_97" [id=2817, type=view]; -"2818 softmax_17" [id=2818, type=softmax]; -"2819 dropout_68" [id=2819, type=dropout]; -"2820 matmul_35" [id=2820, type=matmul]; -"2821 quantize_per_tensor_default_108" [id=2821, type=quantize_per_tensor]; -"2822 dequantize_per_tensor_default_108" [id=2822, type=dequantize_per_tensor]; -"2823 transpose_35" [id=2823, type=transpose]; -"2824 reshape_79" [id=2824, type=reshape]; -"2825 linear_107" [id=2825, type=linear]; -"2826 dropout_69" [id=2826, type=dropout]; -"2827 view_98" [id=2827, type=view]; -"2828 permute_81" [id=2828, type=permute]; -"2829 reshape_80" [id=2829, type=reshape]; -"2830 roll_17" [id=2830, type=roll]; -"2831 slice_269" [id=2831, type=slice]; -"2832 slice_270" [id=2832, type=slice]; -"2833 slice_271" [id=2833, type=slice]; -"2834 slice_272" [id=2834, type=slice]; -"2835 contiguous_33" [id=2835, type=contiguous]; -"2836 layer_norm_37" [id=2836, type=layer_norm]; -"2837 add_61" [id=2837, type=add]; -"2838 quantize_per_tensor_default_109" [id=2838, type=quantize_per_tensor]; -"2839 dequantize_per_tensor_default_109" [id=2839, type=dequantize_per_tensor]; -"2840 linear_108" [id=2840, type=linear]; -"2841 gelu_17" [id=2841, type=gelu]; -"2842 quantize_per_tensor_default_110" [id=2842, type=quantize_per_tensor]; -"2843 dequantize_per_tensor_default_110" [id=2843, type=dequantize_per_tensor]; -"2844 dropout_70" [id=2844, type=dropout]; -"2845 linear_109" [id=2845, type=linear]; -"2846 dropout_71" [id=2846, type=dropout]; -"2847 layer_norm_38" [id=2847, type=layer_norm]; -"2848 add_62" [id=2848, type=add]; -"2849 quantize_per_tensor_default_111" [id=2849, type=quantize_per_tensor]; -"2850 dequantize_per_tensor_default_111" [id=2850, type=dequantize_per_tensor]; -"2851 linear_110" [id=2851, type=linear]; -"2852 relu__18" [id=2852, type=relu_]; -"2853 linear_111" [id=2853, type=linear]; -"2854 view_99" [id=2854, type=view]; -"2855 index_18" [id=2855, type=index]; -"2856 view_100" [id=2856, type=view]; -"2857 permute_82" [id=2857, type=permute]; -"2858 contiguous_34" [id=2858, type=contiguous]; -"2859 unsqueeze_54" [id=2859, type=unsqueeze]; -"2860 sigmoid_18" [id=2860, type=sigmoid]; -"2861 mul_36" [id=2861, type=mul]; -"2862 pad_20" [id=2862, type=pad]; -"2863 view_101" [id=2863, type=view]; -"2864 permute_83" [id=2864, type=permute]; -"2865 reshape_81" [id=2865, type=reshape]; -"2866 clone_18" [id=2866, type=clone]; -"2867 slice_273" [id=2867, type=slice]; -"2868 zero__18" [id=2868, type=zero_]; -"2869 linear_112" [id=2869, type=linear]; -"2870 reshape_82" [id=2870, type=reshape]; -"2871 permute_84" [id=2871, type=permute]; -"2872 select_54" [id=2872, type=select]; -"2873 select_55" [id=2873, type=select]; -"2874 select_56" [id=2874, type=select]; -"2875 linalg_vector_norm_36" [id=2875, type=linalg_vector_norm]; -"2876 clamp_min_36" [id=2876, type=clamp_min]; -"2877 expand_as_36" [id=2877, type=expand_as]; -"2878 div_36" [id=2878, type=div]; -"2879 quantize_per_tensor_default_112" [id=2879, type=quantize_per_tensor]; -"2880 dequantize_per_tensor_default_112" [id=2880, type=dequantize_per_tensor]; -"2881 linalg_vector_norm_37" [id=2881, type=linalg_vector_norm]; -"2882 clamp_min_37" [id=2882, type=clamp_min]; -"2883 expand_as_37" [id=2883, type=expand_as]; -"2884 div_37" [id=2884, type=div]; -"2885 quantize_per_tensor_default_113" [id=2885, type=quantize_per_tensor]; -"2886 dequantize_per_tensor_default_113" [id=2886, type=dequantize_per_tensor]; -"2887 transpose_36" [id=2887, type=transpose]; -"2888 matmul_36" [id=2888, type=matmul]; -"2889 clamp_18" [id=2889, type=clamp]; -"2890 exp_18" [id=2890, type=exp]; -"2891 mul_37" [id=2891, type=mul]; -"2892 add_63" [id=2892, type=add]; -"2893 softmax_18" [id=2893, type=softmax]; -"2894 dropout_72" [id=2894, type=dropout]; -"2895 matmul_37" [id=2895, type=matmul]; -"2896 quantize_per_tensor_default_114" [id=2896, type=quantize_per_tensor]; -"2897 dequantize_per_tensor_default_114" [id=2897, type=dequantize_per_tensor]; -"2898 transpose_37" [id=2898, type=transpose]; -"2899 reshape_83" [id=2899, type=reshape]; -"2900 linear_113" [id=2900, type=linear]; -"2901 dropout_73" [id=2901, type=dropout]; -"2902 view_102" [id=2902, type=view]; -"2903 permute_85" [id=2903, type=permute]; -"2904 reshape_84" [id=2904, type=reshape]; -"2905 slice_274" [id=2905, type=slice]; -"2906 slice_275" [id=2906, type=slice]; -"2907 slice_276" [id=2907, type=slice]; -"2908 slice_277" [id=2908, type=slice]; -"2909 contiguous_35" [id=2909, type=contiguous]; -"2910 layer_norm_39" [id=2910, type=layer_norm]; -"2911 add_64" [id=2911, type=add]; -"2912 quantize_per_tensor_default_115" [id=2912, type=quantize_per_tensor]; -"2913 dequantize_per_tensor_default_115" [id=2913, type=dequantize_per_tensor]; -"2914 linear_114" [id=2914, type=linear]; -"2915 gelu_18" [id=2915, type=gelu]; -"2916 quantize_per_tensor_default_116" [id=2916, type=quantize_per_tensor]; -"2917 dequantize_per_tensor_default_116" [id=2917, type=dequantize_per_tensor]; -"2918 dropout_74" [id=2918, type=dropout]; -"2919 linear_115" [id=2919, type=linear]; -"2920 dropout_75" [id=2920, type=dropout]; -"2921 layer_norm_40" [id=2921, type=layer_norm]; -"2922 add_65" [id=2922, type=add]; -"2923 linear_116" [id=2923, type=linear]; -"2924 relu__19" [id=2924, type=relu_]; -"2925 linear_117" [id=2925, type=linear]; -"2926 view_103" [id=2926, type=view]; -"2927 index_19" [id=2927, type=index]; -"2928 view_104" [id=2928, type=view]; -"2929 permute_86" [id=2929, type=permute]; -"2930 contiguous_36" [id=2930, type=contiguous]; -"2931 unsqueeze_55" [id=2931, type=unsqueeze]; -"2932 sigmoid_19" [id=2932, type=sigmoid]; -"2933 mul_38" [id=2933, type=mul]; -"2934 pad_21" [id=2934, type=pad]; -"2935 roll_18" [id=2935, type=roll]; -"2936 view_105" [id=2936, type=view]; -"2937 permute_87" [id=2937, type=permute]; -"2938 reshape_85" [id=2938, type=reshape]; -"2939 quantize_per_tensor_default_117" [id=2939, type=quantize_per_tensor]; -"2940 dequantize_per_tensor_default_117" [id=2940, type=dequantize_per_tensor]; -"2941 clone_19" [id=2941, type=clone]; -"2942 slice_278" [id=2942, type=slice]; -"2943 zero__19" [id=2943, type=zero_]; -"2944 linear_118" [id=2944, type=linear]; -"2945 reshape_86" [id=2945, type=reshape]; -"2946 permute_88" [id=2946, type=permute]; -"2947 select_57" [id=2947, type=select]; -"2948 select_58" [id=2948, type=select]; -"2949 select_59" [id=2949, type=select]; -"2950 linalg_vector_norm_38" [id=2950, type=linalg_vector_norm]; -"2951 clamp_min_38" [id=2951, type=clamp_min]; -"2952 expand_as_38" [id=2952, type=expand_as]; -"2953 div_38" [id=2953, type=div]; -"2954 quantize_per_tensor_default_118" [id=2954, type=quantize_per_tensor]; -"2955 dequantize_per_tensor_default_118" [id=2955, type=dequantize_per_tensor]; -"2956 linalg_vector_norm_39" [id=2956, type=linalg_vector_norm]; -"2957 clamp_min_39" [id=2957, type=clamp_min]; -"2958 expand_as_39" [id=2958, type=expand_as]; -"2959 div_39" [id=2959, type=div]; -"2960 quantize_per_tensor_default_119" [id=2960, type=quantize_per_tensor]; -"2961 dequantize_per_tensor_default_119" [id=2961, type=dequantize_per_tensor]; -"2962 transpose_38" [id=2962, type=transpose]; -"2963 matmul_38" [id=2963, type=matmul]; -"2964 clamp_19" [id=2964, type=clamp]; -"2965 exp_19" [id=2965, type=exp]; -"2966 mul_39" [id=2966, type=mul]; -"2967 add_66" [id=2967, type=add]; -"2968 new_zeros_9" [id=2968, type=new_zeros]; -"2969 lift_fresh_copy_81" [id=2969, type=lift_fresh_copy]; -"2970 slice_279" [id=2970, type=slice]; -"2971 slice_280" [id=2971, type=slice]; -"2972 fill__81" [id=2972, type=fill_]; -"2973 lift_fresh_copy_82" [id=2973, type=lift_fresh_copy]; -"2974 slice_281" [id=2974, type=slice]; -"2975 slice_282" [id=2975, type=slice]; -"2976 fill__82" [id=2976, type=fill_]; -"2977 lift_fresh_copy_83" [id=2977, type=lift_fresh_copy]; -"2978 slice_283" [id=2978, type=slice]; -"2979 slice_284" [id=2979, type=slice]; -"2980 fill__83" [id=2980, type=fill_]; -"2981 lift_fresh_copy_84" [id=2981, type=lift_fresh_copy]; -"2982 slice_285" [id=2982, type=slice]; -"2983 slice_286" [id=2983, type=slice]; -"2984 fill__84" [id=2984, type=fill_]; -"2985 lift_fresh_copy_85" [id=2985, type=lift_fresh_copy]; -"2986 slice_287" [id=2986, type=slice]; -"2987 slice_288" [id=2987, type=slice]; -"2988 fill__85" [id=2988, type=fill_]; -"2989 lift_fresh_copy_86" [id=2989, type=lift_fresh_copy]; -"2990 slice_289" [id=2990, type=slice]; -"2991 slice_290" [id=2991, type=slice]; -"2992 fill__86" [id=2992, type=fill_]; -"2993 lift_fresh_copy_87" [id=2993, type=lift_fresh_copy]; -"2994 slice_291" [id=2994, type=slice]; -"2995 slice_292" [id=2995, type=slice]; -"2996 fill__87" [id=2996, type=fill_]; -"2997 lift_fresh_copy_88" [id=2997, type=lift_fresh_copy]; -"2998 slice_293" [id=2998, type=slice]; -"2999 slice_294" [id=2999, type=slice]; -"3000 fill__88" [id=3000, type=fill_]; -"3001 lift_fresh_copy_89" [id=3001, type=lift_fresh_copy]; -"3002 slice_295" [id=3002, type=slice]; -"3003 slice_296" [id=3003, type=slice]; -"3004 fill__89" [id=3004, type=fill_]; -"3005 view_106" [id=3005, type=view]; -"3006 permute_89" [id=3006, type=permute]; -"3007 reshape_87" [id=3007, type=reshape]; -"3008 unsqueeze_56" [id=3008, type=unsqueeze]; -"3009 unsqueeze_57" [id=3009, type=unsqueeze]; -"3010 sub_9" [id=3010, type=sub]; -"3011 ne_9" [id=3011, type=ne]; -"3012 masked_fill_18" [id=3012, type=masked_fill]; -"3013 eq_9" [id=3013, type=eq]; -"3014 masked_fill_19" [id=3014, type=masked_fill]; -"3015 view_107" [id=3015, type=view]; -"3016 unsqueeze_58" [id=3016, type=unsqueeze]; -"3017 unsqueeze_59" [id=3017, type=unsqueeze]; -"3018 add_67" [id=3018, type=add]; -"3019 view_108" [id=3019, type=view]; -"3020 softmax_19" [id=3020, type=softmax]; -"3021 dropout_76" [id=3021, type=dropout]; -"3022 matmul_39" [id=3022, type=matmul]; -"3023 quantize_per_tensor_default_120" [id=3023, type=quantize_per_tensor]; -"3024 dequantize_per_tensor_default_120" [id=3024, type=dequantize_per_tensor]; -"3025 transpose_39" [id=3025, type=transpose]; -"3026 reshape_88" [id=3026, type=reshape]; -"3027 linear_119" [id=3027, type=linear]; -"3028 dropout_77" [id=3028, type=dropout]; -"3029 view_109" [id=3029, type=view]; -"3030 permute_90" [id=3030, type=permute]; -"3031 reshape_89" [id=3031, type=reshape]; -"3032 roll_19" [id=3032, type=roll]; -"3033 slice_297" [id=3033, type=slice]; -"3034 slice_298" [id=3034, type=slice]; -"3035 slice_299" [id=3035, type=slice]; -"3036 slice_300" [id=3036, type=slice]; -"3037 contiguous_37" [id=3037, type=contiguous]; -"3038 layer_norm_41" [id=3038, type=layer_norm]; -"3039 add_68" [id=3039, type=add]; -"3040 quantize_per_tensor_default_121" [id=3040, type=quantize_per_tensor]; -"3041 dequantize_per_tensor_default_121" [id=3041, type=dequantize_per_tensor]; -"3042 linear_120" [id=3042, type=linear]; -"3043 gelu_19" [id=3043, type=gelu]; -"3044 quantize_per_tensor_default_122" [id=3044, type=quantize_per_tensor]; -"3045 dequantize_per_tensor_default_122" [id=3045, type=dequantize_per_tensor]; -"3046 dropout_78" [id=3046, type=dropout]; -"3047 linear_121" [id=3047, type=linear]; -"3048 dropout_79" [id=3048, type=dropout]; -"3049 layer_norm_42" [id=3049, type=layer_norm]; -"3050 add_69" [id=3050, type=add]; -"3051 quantize_per_tensor_default_123" [id=3051, type=quantize_per_tensor]; -"3052 dequantize_per_tensor_default_123" [id=3052, type=dequantize_per_tensor]; -"3053 linear_122" [id=3053, type=linear]; -"3054 relu__20" [id=3054, type=relu_]; -"3055 linear_123" [id=3055, type=linear]; -"3056 view_110" [id=3056, type=view]; -"3057 index_20" [id=3057, type=index]; -"3058 view_111" [id=3058, type=view]; -"3059 permute_91" [id=3059, type=permute]; -"3060 contiguous_38" [id=3060, type=contiguous]; -"3061 unsqueeze_60" [id=3061, type=unsqueeze]; -"3062 sigmoid_20" [id=3062, type=sigmoid]; -"3063 mul_40" [id=3063, type=mul]; -"3064 pad_22" [id=3064, type=pad]; -"3065 view_112" [id=3065, type=view]; -"3066 permute_92" [id=3066, type=permute]; -"3067 reshape_90" [id=3067, type=reshape]; -"3068 clone_20" [id=3068, type=clone]; -"3069 slice_301" [id=3069, type=slice]; -"3070 zero__20" [id=3070, type=zero_]; -"3071 linear_124" [id=3071, type=linear]; -"3072 reshape_91" [id=3072, type=reshape]; -"3073 permute_93" [id=3073, type=permute]; -"3074 select_60" [id=3074, type=select]; -"3075 select_61" [id=3075, type=select]; -"3076 select_62" [id=3076, type=select]; -"3077 linalg_vector_norm_40" [id=3077, type=linalg_vector_norm]; -"3078 clamp_min_40" [id=3078, type=clamp_min]; -"3079 expand_as_40" [id=3079, type=expand_as]; -"3080 div_40" [id=3080, type=div]; -"3081 quantize_per_tensor_default_124" [id=3081, type=quantize_per_tensor]; -"3082 dequantize_per_tensor_default_124" [id=3082, type=dequantize_per_tensor]; -"3083 linalg_vector_norm_41" [id=3083, type=linalg_vector_norm]; -"3084 clamp_min_41" [id=3084, type=clamp_min]; -"3085 expand_as_41" [id=3085, type=expand_as]; -"3086 div_41" [id=3086, type=div]; -"3087 quantize_per_tensor_default_125" [id=3087, type=quantize_per_tensor]; -"3088 dequantize_per_tensor_default_125" [id=3088, type=dequantize_per_tensor]; -"3089 transpose_40" [id=3089, type=transpose]; -"3090 matmul_40" [id=3090, type=matmul]; -"3091 clamp_20" [id=3091, type=clamp]; -"3092 exp_20" [id=3092, type=exp]; -"3093 mul_41" [id=3093, type=mul]; -"3094 add_70" [id=3094, type=add]; -"3095 softmax_20" [id=3095, type=softmax]; -"3096 dropout_80" [id=3096, type=dropout]; -"3097 matmul_41" [id=3097, type=matmul]; -"3098 quantize_per_tensor_default_126" [id=3098, type=quantize_per_tensor]; -"3099 dequantize_per_tensor_default_126" [id=3099, type=dequantize_per_tensor]; -"3100 transpose_41" [id=3100, type=transpose]; -"3101 reshape_92" [id=3101, type=reshape]; -"3102 linear_125" [id=3102, type=linear]; -"3103 dropout_81" [id=3103, type=dropout]; -"3104 view_113" [id=3104, type=view]; -"3105 permute_94" [id=3105, type=permute]; -"3106 reshape_93" [id=3106, type=reshape]; -"3107 slice_302" [id=3107, type=slice]; -"3108 slice_303" [id=3108, type=slice]; -"3109 slice_304" [id=3109, type=slice]; -"3110 slice_305" [id=3110, type=slice]; -"3111 contiguous_39" [id=3111, type=contiguous]; -"3112 layer_norm_43" [id=3112, type=layer_norm]; -"3113 add_71" [id=3113, type=add]; -"3114 quantize_per_tensor_default_127" [id=3114, type=quantize_per_tensor]; -"3115 dequantize_per_tensor_default_127" [id=3115, type=dequantize_per_tensor]; -"3116 linear_126" [id=3116, type=linear]; -"3117 gelu_20" [id=3117, type=gelu]; -"3118 quantize_per_tensor_default_128" [id=3118, type=quantize_per_tensor]; -"3119 dequantize_per_tensor_default_128" [id=3119, type=dequantize_per_tensor]; -"3120 dropout_82" [id=3120, type=dropout]; -"3121 linear_127" [id=3121, type=linear]; -"3122 dropout_83" [id=3122, type=dropout]; -"3123 layer_norm_44" [id=3123, type=layer_norm]; -"3124 add_72" [id=3124, type=add]; -"3125 linear_128" [id=3125, type=linear]; -"3126 relu__21" [id=3126, type=relu_]; -"3127 linear_129" [id=3127, type=linear]; -"3128 view_114" [id=3128, type=view]; -"3129 index_21" [id=3129, type=index]; -"3130 view_115" [id=3130, type=view]; -"3131 permute_95" [id=3131, type=permute]; -"3132 contiguous_40" [id=3132, type=contiguous]; -"3133 unsqueeze_61" [id=3133, type=unsqueeze]; -"3134 sigmoid_21" [id=3134, type=sigmoid]; -"3135 mul_42" [id=3135, type=mul]; -"3136 pad_23" [id=3136, type=pad]; -"3137 roll_20" [id=3137, type=roll]; -"3138 view_116" [id=3138, type=view]; -"3139 permute_96" [id=3139, type=permute]; -"3140 reshape_94" [id=3140, type=reshape]; -"3141 quantize_per_tensor_default_129" [id=3141, type=quantize_per_tensor]; -"3142 dequantize_per_tensor_default_129" [id=3142, type=dequantize_per_tensor]; -"3143 clone_21" [id=3143, type=clone]; -"3144 slice_306" [id=3144, type=slice]; -"3145 zero__21" [id=3145, type=zero_]; -"3146 linear_130" [id=3146, type=linear]; -"3147 reshape_95" [id=3147, type=reshape]; -"3148 permute_97" [id=3148, type=permute]; -"3149 select_63" [id=3149, type=select]; -"3150 select_64" [id=3150, type=select]; -"3151 select_65" [id=3151, type=select]; -"3152 linalg_vector_norm_42" [id=3152, type=linalg_vector_norm]; -"3153 clamp_min_42" [id=3153, type=clamp_min]; -"3154 expand_as_42" [id=3154, type=expand_as]; -"3155 div_42" [id=3155, type=div]; -"3156 quantize_per_tensor_default_130" [id=3156, type=quantize_per_tensor]; -"3157 dequantize_per_tensor_default_130" [id=3157, type=dequantize_per_tensor]; -"3158 linalg_vector_norm_43" [id=3158, type=linalg_vector_norm]; -"3159 clamp_min_43" [id=3159, type=clamp_min]; -"3160 expand_as_43" [id=3160, type=expand_as]; -"3161 div_43" [id=3161, type=div]; -"3162 quantize_per_tensor_default_131" [id=3162, type=quantize_per_tensor]; -"3163 dequantize_per_tensor_default_131" [id=3163, type=dequantize_per_tensor]; -"3164 transpose_42" [id=3164, type=transpose]; -"3165 matmul_42" [id=3165, type=matmul]; -"3166 clamp_21" [id=3166, type=clamp]; -"3167 exp_21" [id=3167, type=exp]; -"3168 mul_43" [id=3168, type=mul]; -"3169 add_73" [id=3169, type=add]; -"3170 new_zeros_10" [id=3170, type=new_zeros]; -"3171 lift_fresh_copy_90" [id=3171, type=lift_fresh_copy]; -"3172 slice_307" [id=3172, type=slice]; -"3173 slice_308" [id=3173, type=slice]; -"3174 fill__90" [id=3174, type=fill_]; -"3175 lift_fresh_copy_91" [id=3175, type=lift_fresh_copy]; -"3176 slice_309" [id=3176, type=slice]; -"3177 slice_310" [id=3177, type=slice]; -"3178 fill__91" [id=3178, type=fill_]; -"3179 lift_fresh_copy_92" [id=3179, type=lift_fresh_copy]; -"3180 slice_311" [id=3180, type=slice]; -"3181 slice_312" [id=3181, type=slice]; -"3182 fill__92" [id=3182, type=fill_]; -"3183 lift_fresh_copy_93" [id=3183, type=lift_fresh_copy]; -"3184 slice_313" [id=3184, type=slice]; -"3185 slice_314" [id=3185, type=slice]; -"3186 fill__93" [id=3186, type=fill_]; -"3187 lift_fresh_copy_94" [id=3187, type=lift_fresh_copy]; -"3188 slice_315" [id=3188, type=slice]; -"3189 slice_316" [id=3189, type=slice]; -"3190 fill__94" [id=3190, type=fill_]; -"3191 lift_fresh_copy_95" [id=3191, type=lift_fresh_copy]; -"3192 slice_317" [id=3192, type=slice]; -"3193 slice_318" [id=3193, type=slice]; -"3194 fill__95" [id=3194, type=fill_]; -"3195 lift_fresh_copy_96" [id=3195, type=lift_fresh_copy]; -"3196 slice_319" [id=3196, type=slice]; -"3197 slice_320" [id=3197, type=slice]; -"3198 fill__96" [id=3198, type=fill_]; -"3199 lift_fresh_copy_97" [id=3199, type=lift_fresh_copy]; -"3200 slice_321" [id=3200, type=slice]; -"3201 slice_322" [id=3201, type=slice]; -"3202 fill__97" [id=3202, type=fill_]; -"3203 lift_fresh_copy_98" [id=3203, type=lift_fresh_copy]; -"3204 slice_323" [id=3204, type=slice]; -"3205 slice_324" [id=3205, type=slice]; -"3206 fill__98" [id=3206, type=fill_]; -"3207 view_117" [id=3207, type=view]; -"3208 permute_98" [id=3208, type=permute]; -"3209 reshape_96" [id=3209, type=reshape]; -"3210 unsqueeze_62" [id=3210, type=unsqueeze]; -"3211 unsqueeze_63" [id=3211, type=unsqueeze]; -"3212 sub_10" [id=3212, type=sub]; -"3213 ne_10" [id=3213, type=ne]; -"3214 masked_fill_20" [id=3214, type=masked_fill]; -"3215 eq_10" [id=3215, type=eq]; -"3216 masked_fill_21" [id=3216, type=masked_fill]; -"3217 view_118" [id=3217, type=view]; -"3218 unsqueeze_64" [id=3218, type=unsqueeze]; -"3219 unsqueeze_65" [id=3219, type=unsqueeze]; -"3220 add_74" [id=3220, type=add]; -"3221 view_119" [id=3221, type=view]; -"3222 softmax_21" [id=3222, type=softmax]; -"3223 dropout_84" [id=3223, type=dropout]; -"3224 matmul_43" [id=3224, type=matmul]; -"3225 quantize_per_tensor_default_132" [id=3225, type=quantize_per_tensor]; -"3226 dequantize_per_tensor_default_132" [id=3226, type=dequantize_per_tensor]; -"3227 transpose_43" [id=3227, type=transpose]; -"3228 reshape_97" [id=3228, type=reshape]; -"3229 linear_131" [id=3229, type=linear]; -"3230 dropout_85" [id=3230, type=dropout]; -"3231 view_120" [id=3231, type=view]; -"3232 permute_99" [id=3232, type=permute]; -"3233 reshape_98" [id=3233, type=reshape]; -"3234 roll_21" [id=3234, type=roll]; -"3235 slice_325" [id=3235, type=slice]; -"3236 slice_326" [id=3236, type=slice]; -"3237 slice_327" [id=3237, type=slice]; -"3238 slice_328" [id=3238, type=slice]; -"3239 contiguous_41" [id=3239, type=contiguous]; -"3240 layer_norm_45" [id=3240, type=layer_norm]; -"3241 add_75" [id=3241, type=add]; -"3242 quantize_per_tensor_default_133" [id=3242, type=quantize_per_tensor]; -"3243 dequantize_per_tensor_default_133" [id=3243, type=dequantize_per_tensor]; -"3244 linear_132" [id=3244, type=linear]; -"3245 gelu_21" [id=3245, type=gelu]; -"3246 quantize_per_tensor_default_134" [id=3246, type=quantize_per_tensor]; -"3247 dequantize_per_tensor_default_134" [id=3247, type=dequantize_per_tensor]; -"3248 dropout_86" [id=3248, type=dropout]; -"3249 linear_133" [id=3249, type=linear]; -"3250 dropout_87" [id=3250, type=dropout]; -"3251 layer_norm_46" [id=3251, type=layer_norm]; -"3252 add_76" [id=3252, type=add]; -"3253 quantize_per_tensor_default_135" [id=3253, type=quantize_per_tensor]; -"3254 dequantize_per_tensor_default_135" [id=3254, type=dequantize_per_tensor]; -"3255 pad_24" [id=3255, type=pad]; -"3256 slice_329" [id=3256, type=slice]; -"3257 slice_330" [id=3257, type=slice]; -"3258 slice_331" [id=3258, type=slice]; -"3259 slice_332" [id=3259, type=slice]; -"3260 slice_333" [id=3260, type=slice]; -"3261 slice_334" [id=3261, type=slice]; -"3262 slice_335" [id=3262, type=slice]; -"3263 slice_336" [id=3263, type=slice]; -"3264 slice_337" [id=3264, type=slice]; -"3265 slice_338" [id=3265, type=slice]; -"3266 slice_339" [id=3266, type=slice]; -"3267 slice_340" [id=3267, type=slice]; -"3268 cat_2" [id=3268, type=cat]; -"3269 linear_134" [id=3269, type=linear]; -"3270 layer_norm_47" [id=3270, type=layer_norm]; -"3271 quantize_per_tensor_default_136" [id=3271, type=quantize_per_tensor]; -"3272 dequantize_per_tensor_default_136" [id=3272, type=dequantize_per_tensor]; -"3273 linear_135" [id=3273, type=linear]; -"3274 relu__22" [id=3274, type=relu_]; -"3275 linear_136" [id=3275, type=linear]; -"3276 view_121" [id=3276, type=view]; -"3277 index_22" [id=3277, type=index]; -"3278 view_122" [id=3278, type=view]; -"3279 permute_100" [id=3279, type=permute]; -"3280 contiguous_42" [id=3280, type=contiguous]; -"3281 unsqueeze_66" [id=3281, type=unsqueeze]; -"3282 sigmoid_22" [id=3282, type=sigmoid]; -"3283 mul_44" [id=3283, type=mul]; -"3284 pad_25" [id=3284, type=pad]; -"3285 view_123" [id=3285, type=view]; -"3286 permute_101" [id=3286, type=permute]; -"3287 reshape_99" [id=3287, type=reshape]; -"3288 clone_22" [id=3288, type=clone]; -"3289 slice_341" [id=3289, type=slice]; -"3290 zero__22" [id=3290, type=zero_]; -"3291 linear_137" [id=3291, type=linear]; -"3292 reshape_100" [id=3292, type=reshape]; -"3293 permute_102" [id=3293, type=permute]; -"3294 select_66" [id=3294, type=select]; -"3295 select_67" [id=3295, type=select]; -"3296 select_68" [id=3296, type=select]; -"3297 linalg_vector_norm_44" [id=3297, type=linalg_vector_norm]; -"3298 clamp_min_44" [id=3298, type=clamp_min]; -"3299 expand_as_44" [id=3299, type=expand_as]; -"3300 div_44" [id=3300, type=div]; -"3301 quantize_per_tensor_default_137" [id=3301, type=quantize_per_tensor]; -"3302 dequantize_per_tensor_default_137" [id=3302, type=dequantize_per_tensor]; -"3303 linalg_vector_norm_45" [id=3303, type=linalg_vector_norm]; -"3304 clamp_min_45" [id=3304, type=clamp_min]; -"3305 expand_as_45" [id=3305, type=expand_as]; -"3306 div_45" [id=3306, type=div]; -"3307 quantize_per_tensor_default_138" [id=3307, type=quantize_per_tensor]; -"3308 dequantize_per_tensor_default_138" [id=3308, type=dequantize_per_tensor]; -"3309 transpose_44" [id=3309, type=transpose]; -"3310 matmul_44" [id=3310, type=matmul]; -"3311 clamp_22" [id=3311, type=clamp]; -"3312 exp_22" [id=3312, type=exp]; -"3313 mul_45" [id=3313, type=mul]; -"3314 add_77" [id=3314, type=add]; -"3315 softmax_22" [id=3315, type=softmax]; -"3316 dropout_88" [id=3316, type=dropout]; -"3317 matmul_45" [id=3317, type=matmul]; -"3318 quantize_per_tensor_default_139" [id=3318, type=quantize_per_tensor]; -"3319 dequantize_per_tensor_default_139" [id=3319, type=dequantize_per_tensor]; -"3320 transpose_45" [id=3320, type=transpose]; -"3321 reshape_101" [id=3321, type=reshape]; -"3322 linear_138" [id=3322, type=linear]; -"3323 dropout_89" [id=3323, type=dropout]; -"3324 view_124" [id=3324, type=view]; -"3325 permute_103" [id=3325, type=permute]; -"3326 reshape_102" [id=3326, type=reshape]; -"3327 slice_342" [id=3327, type=slice]; -"3328 slice_343" [id=3328, type=slice]; -"3329 slice_344" [id=3329, type=slice]; -"3330 slice_345" [id=3330, type=slice]; -"3331 contiguous_43" [id=3331, type=contiguous]; -"3332 layer_norm_48" [id=3332, type=layer_norm]; -"3333 add_78" [id=3333, type=add]; -"3334 quantize_per_tensor_default_140" [id=3334, type=quantize_per_tensor]; -"3335 dequantize_per_tensor_default_140" [id=3335, type=dequantize_per_tensor]; -"3336 linear_139" [id=3336, type=linear]; -"3337 gelu_22" [id=3337, type=gelu]; -"3338 quantize_per_tensor_default_141" [id=3338, type=quantize_per_tensor]; -"3339 dequantize_per_tensor_default_141" [id=3339, type=dequantize_per_tensor]; -"3340 dropout_90" [id=3340, type=dropout]; -"3341 linear_140" [id=3341, type=linear]; -"3342 dropout_91" [id=3342, type=dropout]; -"3343 layer_norm_49" [id=3343, type=layer_norm]; -"3344 add_79" [id=3344, type=add]; -"3345 quantize_per_tensor_default_142" [id=3345, type=quantize_per_tensor]; -"3346 dequantize_per_tensor_default_142" [id=3346, type=dequantize_per_tensor]; -"3347 linear_141" [id=3347, type=linear]; -"3348 relu__23" [id=3348, type=relu_]; -"3349 linear_142" [id=3349, type=linear]; -"3350 view_125" [id=3350, type=view]; -"3351 index_23" [id=3351, type=index]; -"3352 view_126" [id=3352, type=view]; -"3353 permute_104" [id=3353, type=permute]; -"3354 contiguous_44" [id=3354, type=contiguous]; -"3355 unsqueeze_67" [id=3355, type=unsqueeze]; -"3356 sigmoid_23" [id=3356, type=sigmoid]; -"3357 mul_46" [id=3357, type=mul]; -"3358 pad_26" [id=3358, type=pad]; -"3359 view_127" [id=3359, type=view]; -"3360 permute_105" [id=3360, type=permute]; -"3361 reshape_103" [id=3361, type=reshape]; -"3362 clone_23" [id=3362, type=clone]; -"3363 slice_346" [id=3363, type=slice]; -"3364 zero__23" [id=3364, type=zero_]; -"3365 linear_143" [id=3365, type=linear]; -"3366 reshape_104" [id=3366, type=reshape]; -"3367 permute_106" [id=3367, type=permute]; -"3368 select_69" [id=3368, type=select]; -"3369 select_70" [id=3369, type=select]; -"3370 select_71" [id=3370, type=select]; -"3371 linalg_vector_norm_46" [id=3371, type=linalg_vector_norm]; -"3372 clamp_min_46" [id=3372, type=clamp_min]; -"3373 expand_as_46" [id=3373, type=expand_as]; -"3374 div_46" [id=3374, type=div]; -"3375 quantize_per_tensor_default_143" [id=3375, type=quantize_per_tensor]; -"3376 dequantize_per_tensor_default_143" [id=3376, type=dequantize_per_tensor]; -"3377 linalg_vector_norm_47" [id=3377, type=linalg_vector_norm]; -"3378 clamp_min_47" [id=3378, type=clamp_min]; -"3379 expand_as_47" [id=3379, type=expand_as]; -"3380 div_47" [id=3380, type=div]; -"3381 quantize_per_tensor_default_144" [id=3381, type=quantize_per_tensor]; -"3382 dequantize_per_tensor_default_144" [id=3382, type=dequantize_per_tensor]; -"3383 transpose_46" [id=3383, type=transpose]; -"3384 matmul_46" [id=3384, type=matmul]; -"3385 clamp_23" [id=3385, type=clamp]; -"3386 exp_23" [id=3386, type=exp]; -"3387 mul_47" [id=3387, type=mul]; -"3388 add_80" [id=3388, type=add]; -"3389 softmax_23" [id=3389, type=softmax]; -"3390 dropout_92" [id=3390, type=dropout]; -"3391 matmul_47" [id=3391, type=matmul]; -"3392 quantize_per_tensor_default_145" [id=3392, type=quantize_per_tensor]; -"3393 dequantize_per_tensor_default_145" [id=3393, type=dequantize_per_tensor]; -"3394 transpose_47" [id=3394, type=transpose]; -"3395 reshape_105" [id=3395, type=reshape]; -"3396 linear_144" [id=3396, type=linear]; -"3397 dropout_93" [id=3397, type=dropout]; -"3398 view_128" [id=3398, type=view]; -"3399 permute_107" [id=3399, type=permute]; -"3400 reshape_106" [id=3400, type=reshape]; -"3401 slice_347" [id=3401, type=slice]; -"3402 slice_348" [id=3402, type=slice]; -"3403 slice_349" [id=3403, type=slice]; -"3404 slice_350" [id=3404, type=slice]; -"3405 contiguous_45" [id=3405, type=contiguous]; -"3406 layer_norm_50" [id=3406, type=layer_norm]; -"3407 add_81" [id=3407, type=add]; -"3408 quantize_per_tensor_default_146" [id=3408, type=quantize_per_tensor]; -"3409 dequantize_per_tensor_default_146" [id=3409, type=dequantize_per_tensor]; -"3410 linear_145" [id=3410, type=linear]; -"3411 gelu_23" [id=3411, type=gelu]; -"3412 quantize_per_tensor_default_147" [id=3412, type=quantize_per_tensor]; -"3413 dequantize_per_tensor_default_147" [id=3413, type=dequantize_per_tensor]; -"3414 dropout_94" [id=3414, type=dropout]; -"3415 linear_146" [id=3415, type=linear]; -"3416 dropout_95" [id=3416, type=dropout]; -"3417 layer_norm_51" [id=3417, type=layer_norm]; -"3418 add_82" [id=3418, type=add]; -"3419 layer_norm_52" [id=3419, type=layer_norm]; -"3420 permute_108" [id=3420, type=permute]; -"3421 adaptive_avg_pool2d" [id=3421, type=adaptive_avg_pool2d]; -"3422 quantize_per_tensor_default_148" [id=3422, type=quantize_per_tensor]; -"3423 dequantize_per_tensor_default_148" [id=3423, type=dequantize_per_tensor]; -"3424 flatten" [id=3424, type=flatten]; -"3425 linear_147" [id=3425, type=linear]; -"3426 output_1" [id=3426, type=output]; +"8 features_1_0_attn_proj_bias" [id=8, type=get_attr]; +"9 _scale_2" [id=9, type=get_attr]; +"10 _zero_point_2" [id=10, type=get_attr]; +"11 _frozen_param8_1" [id=11, type=get_attr]; +"12 dequantize_per_channel_default_2" [id=12, type=dequantize_per_channel]; +"13 _scale_3" [id=13, type=get_attr]; +"14 _zero_point_3" [id=14, type=get_attr]; +"15 _frozen_param9_1" [id=15, type=get_attr]; +"16 dequantize_per_channel_default_3" [id=16, type=dequantize_per_channel]; +"17 features_1_1_attn_proj_bias" [id=17, type=get_attr]; +"18 _scale_4" [id=18, type=get_attr]; +"19 _zero_point_4" [id=19, type=get_attr]; +"20 _frozen_param10_1" [id=20, type=get_attr]; +"21 dequantize_per_channel_default_4" [id=21, type=dequantize_per_channel]; +"22 _scale_5" [id=22, type=get_attr]; +"23 _zero_point_5" [id=23, type=get_attr]; +"24 _frozen_param11_1" [id=24, type=get_attr]; +"25 dequantize_per_channel_default_5" [id=25, type=dequantize_per_channel]; +"26 features_3_0_attn_proj_bias" [id=26, type=get_attr]; +"27 _scale_6" [id=27, type=get_attr]; +"28 _zero_point_6" [id=28, type=get_attr]; +"29 _frozen_param12_1" [id=29, type=get_attr]; +"30 dequantize_per_channel_default_6" [id=30, type=dequantize_per_channel]; +"31 _scale_7" [id=31, type=get_attr]; +"32 _zero_point_7" [id=32, type=get_attr]; +"33 _frozen_param13_1" [id=33, type=get_attr]; +"34 dequantize_per_channel_default_7" [id=34, type=dequantize_per_channel]; +"35 features_3_1_attn_proj_bias" [id=35, type=get_attr]; +"36 _scale_8" [id=36, type=get_attr]; +"37 _zero_point_8" [id=37, type=get_attr]; +"38 _frozen_param14_1" [id=38, type=get_attr]; +"39 dequantize_per_channel_default_8" [id=39, type=dequantize_per_channel]; +"40 _scale_9" [id=40, type=get_attr]; +"41 _zero_point_9" [id=41, type=get_attr]; +"42 _frozen_param15_1" [id=42, type=get_attr]; +"43 dequantize_per_channel_default_9" [id=43, type=dequantize_per_channel]; +"44 features_5_0_attn_proj_bias" [id=44, type=get_attr]; +"45 _scale_10" [id=45, type=get_attr]; +"46 _zero_point_10" [id=46, type=get_attr]; +"47 _frozen_param16_1" [id=47, type=get_attr]; +"48 dequantize_per_channel_default_10" [id=48, type=dequantize_per_channel]; +"49 _scale_11" [id=49, type=get_attr]; +"50 _zero_point_11" [id=50, type=get_attr]; +"51 _frozen_param19_1" [id=51, type=get_attr]; +"52 dequantize_per_channel_default_11" [id=52, type=dequantize_per_channel]; +"53 features_5_1_attn_proj_bias" [id=53, type=get_attr]; +"54 _scale_12" [id=54, type=get_attr]; +"55 _zero_point_12" [id=55, type=get_attr]; +"56 _frozen_param23_1" [id=56, type=get_attr]; +"57 dequantize_per_channel_default_12" [id=57, type=dequantize_per_channel]; +"58 _scale_13" [id=58, type=get_attr]; +"59 _zero_point_13" [id=59, type=get_attr]; +"60 _frozen_param25_1" [id=60, type=get_attr]; +"61 dequantize_per_channel_default_13" [id=61, type=dequantize_per_channel]; +"62 features_5_2_attn_proj_bias" [id=62, type=get_attr]; +"63 _scale_14" [id=63, type=get_attr]; +"64 _zero_point_14" [id=64, type=get_attr]; +"65 _frozen_param26_1" [id=65, type=get_attr]; +"66 dequantize_per_channel_default_14" [id=66, type=dequantize_per_channel]; +"67 _scale_15" [id=67, type=get_attr]; +"68 _zero_point_15" [id=68, type=get_attr]; +"69 _frozen_param27_1" [id=69, type=get_attr]; +"70 dequantize_per_channel_default_15" [id=70, type=dequantize_per_channel]; +"71 features_5_3_attn_proj_bias" [id=71, type=get_attr]; +"72 _scale_16" [id=72, type=get_attr]; +"73 _zero_point_16" [id=73, type=get_attr]; +"74 _frozen_param28_1" [id=74, type=get_attr]; +"75 dequantize_per_channel_default_16" [id=75, type=dequantize_per_channel]; +"76 _scale_17" [id=76, type=get_attr]; +"77 _zero_point_17" [id=77, type=get_attr]; +"78 _frozen_param29_1" [id=78, type=get_attr]; +"79 dequantize_per_channel_default_17" [id=79, type=dequantize_per_channel]; +"80 features_5_4_attn_proj_bias" [id=80, type=get_attr]; +"81 _scale_18" [id=81, type=get_attr]; +"82 _zero_point_18" [id=82, type=get_attr]; +"83 _frozen_param30_1" [id=83, type=get_attr]; +"84 dequantize_per_channel_default_18" [id=84, type=dequantize_per_channel]; +"85 _scale_19" [id=85, type=get_attr]; +"86 _zero_point_19" [id=86, type=get_attr]; +"87 _frozen_param31_1" [id=87, type=get_attr]; +"88 dequantize_per_channel_default_19" [id=88, type=dequantize_per_channel]; +"89 features_5_5_attn_proj_bias" [id=89, type=get_attr]; +"90 _scale_20" [id=90, type=get_attr]; +"91 _zero_point_20" [id=91, type=get_attr]; +"92 _frozen_param32_1" [id=92, type=get_attr]; +"93 dequantize_per_channel_default_20" [id=93, type=dequantize_per_channel]; +"94 _scale_21" [id=94, type=get_attr]; +"95 _zero_point_21" [id=95, type=get_attr]; +"96 _frozen_param33_1" [id=96, type=get_attr]; +"97 dequantize_per_channel_default_21" [id=97, type=dequantize_per_channel]; +"98 features_5_6_attn_proj_bias" [id=98, type=get_attr]; +"99 _scale_22" [id=99, type=get_attr]; +"100 _zero_point_22" [id=100, type=get_attr]; +"101 _frozen_param36_1" [id=101, type=get_attr]; +"102 dequantize_per_channel_default_22" [id=102, type=dequantize_per_channel]; +"103 _scale_23" [id=103, type=get_attr]; +"104 _zero_point_23" [id=104, type=get_attr]; +"105 _frozen_param40_1" [id=105, type=get_attr]; +"106 dequantize_per_channel_default_23" [id=106, type=dequantize_per_channel]; +"107 features_5_7_attn_proj_bias" [id=107, type=get_attr]; +"108 _scale_24" [id=108, type=get_attr]; +"109 _zero_point_24" [id=109, type=get_attr]; +"110 _frozen_param42_1" [id=110, type=get_attr]; +"111 dequantize_per_channel_default_24" [id=111, type=dequantize_per_channel]; +"112 _scale_25" [id=112, type=get_attr]; +"113 _zero_point_25" [id=113, type=get_attr]; +"114 _frozen_param43_1" [id=114, type=get_attr]; +"115 dequantize_per_channel_default_25" [id=115, type=dequantize_per_channel]; +"116 features_5_8_attn_proj_bias" [id=116, type=get_attr]; +"117 _scale_26" [id=117, type=get_attr]; +"118 _zero_point_26" [id=118, type=get_attr]; +"119 _frozen_param44_1" [id=119, type=get_attr]; +"120 dequantize_per_channel_default_26" [id=120, type=dequantize_per_channel]; +"121 _scale_27" [id=121, type=get_attr]; +"122 _zero_point_27" [id=122, type=get_attr]; +"123 _frozen_param45_1" [id=123, type=get_attr]; +"124 dequantize_per_channel_default_27" [id=124, type=dequantize_per_channel]; +"125 features_5_9_attn_proj_bias" [id=125, type=get_attr]; +"126 _scale_28" [id=126, type=get_attr]; +"127 _zero_point_28" [id=127, type=get_attr]; +"128 _frozen_param46_1" [id=128, type=get_attr]; +"129 dequantize_per_channel_default_28" [id=129, type=dequantize_per_channel]; +"130 _scale_29" [id=130, type=get_attr]; +"131 _zero_point_29" [id=131, type=get_attr]; +"132 _frozen_param47_1" [id=132, type=get_attr]; +"133 dequantize_per_channel_default_29" [id=133, type=dequantize_per_channel]; +"134 features_5_10_attn_proj_bias" [id=134, type=get_attr]; +"135 _scale_30" [id=135, type=get_attr]; +"136 _zero_point_30" [id=136, type=get_attr]; +"137 _frozen_param48_1" [id=137, type=get_attr]; +"138 dequantize_per_channel_default_30" [id=138, type=dequantize_per_channel]; +"139 _scale_31" [id=139, type=get_attr]; +"140 _zero_point_31" [id=140, type=get_attr]; +"141 _frozen_param49_1" [id=141, type=get_attr]; +"142 dequantize_per_channel_default_31" [id=142, type=dequantize_per_channel]; +"143 features_5_11_attn_proj_bias" [id=143, type=get_attr]; +"144 _scale_32" [id=144, type=get_attr]; +"145 _zero_point_32" [id=145, type=get_attr]; +"146 _frozen_param50_1" [id=146, type=get_attr]; +"147 dequantize_per_channel_default_32" [id=147, type=dequantize_per_channel]; +"148 _scale_33" [id=148, type=get_attr]; +"149 _zero_point_33" [id=149, type=get_attr]; +"150 _frozen_param53_1" [id=150, type=get_attr]; +"151 dequantize_per_channel_default_33" [id=151, type=dequantize_per_channel]; +"152 features_5_12_attn_proj_bias" [id=152, type=get_attr]; +"153 _scale_34" [id=153, type=get_attr]; +"154 _zero_point_34" [id=154, type=get_attr]; +"155 _frozen_param57_1" [id=155, type=get_attr]; +"156 dequantize_per_channel_default_34" [id=156, type=dequantize_per_channel]; +"157 _scale_35" [id=157, type=get_attr]; +"158 _zero_point_35" [id=158, type=get_attr]; +"159 _frozen_param59_1" [id=159, type=get_attr]; +"160 dequantize_per_channel_default_35" [id=160, type=dequantize_per_channel]; +"161 features_5_13_attn_proj_bias" [id=161, type=get_attr]; +"162 _scale_36" [id=162, type=get_attr]; +"163 _zero_point_36" [id=163, type=get_attr]; +"164 _frozen_param60_1" [id=164, type=get_attr]; +"165 dequantize_per_channel_default_36" [id=165, type=dequantize_per_channel]; +"166 _scale_37" [id=166, type=get_attr]; +"167 _zero_point_37" [id=167, type=get_attr]; +"168 _frozen_param61_1" [id=168, type=get_attr]; +"169 dequantize_per_channel_default_37" [id=169, type=dequantize_per_channel]; +"170 features_5_14_attn_proj_bias" [id=170, type=get_attr]; +"171 _scale_38" [id=171, type=get_attr]; +"172 _zero_point_38" [id=172, type=get_attr]; +"173 _frozen_param62_1" [id=173, type=get_attr]; +"174 dequantize_per_channel_default_38" [id=174, type=dequantize_per_channel]; +"175 _scale_39" [id=175, type=get_attr]; +"176 _zero_point_39" [id=176, type=get_attr]; +"177 _frozen_param63_1" [id=177, type=get_attr]; +"178 dequantize_per_channel_default_39" [id=178, type=dequantize_per_channel]; +"179 features_5_15_attn_proj_bias" [id=179, type=get_attr]; +"180 _scale_40" [id=180, type=get_attr]; +"181 _zero_point_40" [id=181, type=get_attr]; +"182 _frozen_param64_1" [id=182, type=get_attr]; +"183 dequantize_per_channel_default_40" [id=183, type=dequantize_per_channel]; +"184 _scale_41" [id=184, type=get_attr]; +"185 _zero_point_41" [id=185, type=get_attr]; +"186 _frozen_param65_1" [id=186, type=get_attr]; +"187 dequantize_per_channel_default_41" [id=187, type=dequantize_per_channel]; +"188 features_5_16_attn_proj_bias" [id=188, type=get_attr]; +"189 _scale_42" [id=189, type=get_attr]; +"190 _zero_point_42" [id=190, type=get_attr]; +"191 _frozen_param66_1" [id=191, type=get_attr]; +"192 dequantize_per_channel_default_42" [id=192, type=dequantize_per_channel]; +"193 _scale_43" [id=193, type=get_attr]; +"194 _zero_point_43" [id=194, type=get_attr]; +"195 _frozen_param67_1" [id=195, type=get_attr]; +"196 dequantize_per_channel_default_43" [id=196, type=dequantize_per_channel]; +"197 features_5_17_attn_proj_bias" [id=197, type=get_attr]; +"198 _scale_44" [id=198, type=get_attr]; +"199 _zero_point_44" [id=199, type=get_attr]; +"200 _frozen_param70_1" [id=200, type=get_attr]; +"201 dequantize_per_channel_default_44" [id=201, type=dequantize_per_channel]; +"202 _scale_45" [id=202, type=get_attr]; +"203 _zero_point_45" [id=203, type=get_attr]; +"204 _frozen_param74_1" [id=204, type=get_attr]; +"205 dequantize_per_channel_default_45" [id=205, type=dequantize_per_channel]; +"206 features_7_0_attn_proj_bias" [id=206, type=get_attr]; +"207 _scale_46" [id=207, type=get_attr]; +"208 _zero_point_46" [id=208, type=get_attr]; +"209 _frozen_param76_1" [id=209, type=get_attr]; +"210 dequantize_per_channel_default_46" [id=210, type=dequantize_per_channel]; +"211 _scale_47" [id=211, type=get_attr]; +"212 _zero_point_47" [id=212, type=get_attr]; +"213 _frozen_param77_1" [id=213, type=get_attr]; +"214 dequantize_per_channel_default_47" [id=214, type=dequantize_per_channel]; +"215 features_7_1_attn_proj_bias" [id=215, type=get_attr]; +"216 _scale_48" [id=216, type=get_attr]; +"217 _zero_point_48" [id=217, type=get_attr]; +"218 _frozen_param78_1" [id=218, type=get_attr]; +"219 dequantize_per_channel_default_48" [id=219, type=dequantize_per_channel]; +"220 features_0_0_bias" [id=220, type=get_attr]; +"221 features_0_2_weight" [id=221, type=get_attr]; +"222 features_0_2_bias" [id=222, type=get_attr]; +"223 features_1_0_norm1_weight" [id=223, type=get_attr]; +"224 features_1_0_norm1_bias" [id=224, type=get_attr]; +"225 _scale_49" [id=225, type=get_attr]; +"226 _zero_point_49" [id=226, type=get_attr]; +"227 _frozen_param79_1" [id=227, type=get_attr]; +"228 dequantize_per_channel_default_49" [id=228, type=dequantize_per_channel]; +"229 features_1_0_mlp_0_bias" [id=229, type=get_attr]; +"230 _scale_50" [id=230, type=get_attr]; +"231 _zero_point_50" [id=231, type=get_attr]; +"232 _frozen_param80_1" [id=232, type=get_attr]; +"233 dequantize_per_channel_default_50" [id=233, type=dequantize_per_channel]; +"234 features_1_0_mlp_3_bias" [id=234, type=get_attr]; +"235 features_1_0_norm2_weight" [id=235, type=get_attr]; +"236 features_1_0_norm2_bias" [id=236, type=get_attr]; +"237 features_1_1_norm1_weight" [id=237, type=get_attr]; +"238 features_1_1_norm1_bias" [id=238, type=get_attr]; +"239 _scale_51" [id=239, type=get_attr]; +"240 _zero_point_51" [id=240, type=get_attr]; +"241 _frozen_param81_1" [id=241, type=get_attr]; +"242 dequantize_per_channel_default_51" [id=242, type=dequantize_per_channel]; +"243 features_1_1_mlp_0_bias" [id=243, type=get_attr]; +"244 _scale_52" [id=244, type=get_attr]; +"245 _zero_point_52" [id=245, type=get_attr]; +"246 _frozen_param82_1" [id=246, type=get_attr]; +"247 dequantize_per_channel_default_52" [id=247, type=dequantize_per_channel]; +"248 features_1_1_mlp_3_bias" [id=248, type=get_attr]; +"249 features_1_1_norm2_weight" [id=249, type=get_attr]; +"250 features_1_1_norm2_bias" [id=250, type=get_attr]; +"251 _scale_53" [id=251, type=get_attr]; +"252 _zero_point_53" [id=252, type=get_attr]; +"253 _frozen_param83_1" [id=253, type=get_attr]; +"254 dequantize_per_channel_default_53" [id=254, type=dequantize_per_channel]; +"255 features_2_norm_weight" [id=255, type=get_attr]; +"256 features_2_norm_bias" [id=256, type=get_attr]; +"257 features_3_0_norm1_weight" [id=257, type=get_attr]; +"258 features_3_0_norm1_bias" [id=258, type=get_attr]; +"259 _scale_54" [id=259, type=get_attr]; +"260 _zero_point_54" [id=260, type=get_attr]; +"261 _frozen_param84_1" [id=261, type=get_attr]; +"262 dequantize_per_channel_default_54" [id=262, type=dequantize_per_channel]; +"263 features_3_0_mlp_0_bias" [id=263, type=get_attr]; +"264 _scale_55" [id=264, type=get_attr]; +"265 _zero_point_55" [id=265, type=get_attr]; +"266 _frozen_param87_1" [id=266, type=get_attr]; +"267 dequantize_per_channel_default_55" [id=267, type=dequantize_per_channel]; +"268 features_3_0_mlp_3_bias" [id=268, type=get_attr]; +"269 features_3_0_norm2_weight" [id=269, type=get_attr]; +"270 features_3_0_norm2_bias" [id=270, type=get_attr]; +"271 features_3_1_norm1_weight" [id=271, type=get_attr]; +"272 features_3_1_norm1_bias" [id=272, type=get_attr]; +"273 _scale_56" [id=273, type=get_attr]; +"274 _zero_point_56" [id=274, type=get_attr]; +"275 _frozen_param91_1" [id=275, type=get_attr]; +"276 dequantize_per_channel_default_56" [id=276, type=dequantize_per_channel]; +"277 features_3_1_mlp_0_bias" [id=277, type=get_attr]; +"278 _scale_57" [id=278, type=get_attr]; +"279 _zero_point_57" [id=279, type=get_attr]; +"280 _frozen_param93_1" [id=280, type=get_attr]; +"281 dequantize_per_channel_default_57" [id=281, type=dequantize_per_channel]; +"282 features_3_1_mlp_3_bias" [id=282, type=get_attr]; +"283 features_3_1_norm2_weight" [id=283, type=get_attr]; +"284 features_3_1_norm2_bias" [id=284, type=get_attr]; +"285 _scale_58" [id=285, type=get_attr]; +"286 _zero_point_58" [id=286, type=get_attr]; +"287 _frozen_param94_1" [id=287, type=get_attr]; +"288 dequantize_per_channel_default_58" [id=288, type=dequantize_per_channel]; +"289 features_4_norm_weight" [id=289, type=get_attr]; +"290 features_4_norm_bias" [id=290, type=get_attr]; +"291 features_5_0_norm1_weight" [id=291, type=get_attr]; +"292 features_5_0_norm1_bias" [id=292, type=get_attr]; +"293 _scale_59" [id=293, type=get_attr]; +"294 _zero_point_59" [id=294, type=get_attr]; +"295 _frozen_param95_1" [id=295, type=get_attr]; +"296 dequantize_per_channel_default_59" [id=296, type=dequantize_per_channel]; +"297 features_5_0_mlp_0_bias" [id=297, type=get_attr]; +"298 _scale_60" [id=298, type=get_attr]; +"299 _zero_point_60" [id=299, type=get_attr]; +"300 _frozen_param96_1" [id=300, type=get_attr]; +"301 dequantize_per_channel_default_60" [id=301, type=dequantize_per_channel]; +"302 features_5_0_mlp_3_bias" [id=302, type=get_attr]; +"303 features_5_0_norm2_weight" [id=303, type=get_attr]; +"304 features_5_0_norm2_bias" [id=304, type=get_attr]; +"305 features_5_1_norm1_weight" [id=305, type=get_attr]; +"306 features_5_1_norm1_bias" [id=306, type=get_attr]; +"307 _scale_61" [id=307, type=get_attr]; +"308 _zero_point_61" [id=308, type=get_attr]; +"309 _frozen_param97_1" [id=309, type=get_attr]; +"310 dequantize_per_channel_default_61" [id=310, type=dequantize_per_channel]; +"311 features_5_1_mlp_0_bias" [id=311, type=get_attr]; +"312 _scale_62" [id=312, type=get_attr]; +"313 _zero_point_62" [id=313, type=get_attr]; +"314 _frozen_param98_1" [id=314, type=get_attr]; +"315 dequantize_per_channel_default_62" [id=315, type=dequantize_per_channel]; +"316 features_5_1_mlp_3_bias" [id=316, type=get_attr]; +"317 features_5_1_norm2_weight" [id=317, type=get_attr]; +"318 features_5_1_norm2_bias" [id=318, type=get_attr]; +"319 features_5_2_norm1_weight" [id=319, type=get_attr]; +"320 features_5_2_norm1_bias" [id=320, type=get_attr]; +"321 _scale_63" [id=321, type=get_attr]; +"322 _zero_point_63" [id=322, type=get_attr]; +"323 _frozen_param99_1" [id=323, type=get_attr]; +"324 dequantize_per_channel_default_63" [id=324, type=dequantize_per_channel]; +"325 features_5_2_mlp_0_bias" [id=325, type=get_attr]; +"326 _scale_64" [id=326, type=get_attr]; +"327 _zero_point_64" [id=327, type=get_attr]; +"328 _frozen_param100_1" [id=328, type=get_attr]; +"329 dequantize_per_channel_default_64" [id=329, type=dequantize_per_channel]; +"330 features_5_2_mlp_3_bias" [id=330, type=get_attr]; +"331 features_5_2_norm2_weight" [id=331, type=get_attr]; +"332 features_5_2_norm2_bias" [id=332, type=get_attr]; +"333 features_5_3_norm1_weight" [id=333, type=get_attr]; +"334 features_5_3_norm1_bias" [id=334, type=get_attr]; +"335 _scale_65" [id=335, type=get_attr]; +"336 _zero_point_65" [id=336, type=get_attr]; +"337 _frozen_param101_1" [id=337, type=get_attr]; +"338 dequantize_per_channel_default_65" [id=338, type=dequantize_per_channel]; +"339 features_5_3_mlp_0_bias" [id=339, type=get_attr]; +"340 _scale_66" [id=340, type=get_attr]; +"341 _zero_point_66" [id=341, type=get_attr]; +"342 _frozen_param104_1" [id=342, type=get_attr]; +"343 dequantize_per_channel_default_66" [id=343, type=dequantize_per_channel]; +"344 features_5_3_mlp_3_bias" [id=344, type=get_attr]; +"345 features_5_3_norm2_weight" [id=345, type=get_attr]; +"346 features_5_3_norm2_bias" [id=346, type=get_attr]; +"347 features_5_4_norm1_weight" [id=347, type=get_attr]; +"348 features_5_4_norm1_bias" [id=348, type=get_attr]; +"349 _scale_67" [id=349, type=get_attr]; +"350 _zero_point_67" [id=350, type=get_attr]; +"351 _frozen_param108_1" [id=351, type=get_attr]; +"352 dequantize_per_channel_default_67" [id=352, type=dequantize_per_channel]; +"353 features_5_4_mlp_0_bias" [id=353, type=get_attr]; +"354 _scale_68" [id=354, type=get_attr]; +"355 _zero_point_68" [id=355, type=get_attr]; +"356 _frozen_param110_1" [id=356, type=get_attr]; +"357 dequantize_per_channel_default_68" [id=357, type=dequantize_per_channel]; +"358 features_5_4_mlp_3_bias" [id=358, type=get_attr]; +"359 features_5_4_norm2_weight" [id=359, type=get_attr]; +"360 features_5_4_norm2_bias" [id=360, type=get_attr]; +"361 features_5_5_norm1_weight" [id=361, type=get_attr]; +"362 features_5_5_norm1_bias" [id=362, type=get_attr]; +"363 _scale_69" [id=363, type=get_attr]; +"364 _zero_point_69" [id=364, type=get_attr]; +"365 _frozen_param111_1" [id=365, type=get_attr]; +"366 dequantize_per_channel_default_69" [id=366, type=dequantize_per_channel]; +"367 features_5_5_mlp_0_bias" [id=367, type=get_attr]; +"368 _scale_70" [id=368, type=get_attr]; +"369 _zero_point_70" [id=369, type=get_attr]; +"370 _frozen_param112_1" [id=370, type=get_attr]; +"371 dequantize_per_channel_default_70" [id=371, type=dequantize_per_channel]; +"372 features_5_5_mlp_3_bias" [id=372, type=get_attr]; +"373 features_5_5_norm2_weight" [id=373, type=get_attr]; +"374 features_5_5_norm2_bias" [id=374, type=get_attr]; +"375 features_5_6_norm1_weight" [id=375, type=get_attr]; +"376 features_5_6_norm1_bias" [id=376, type=get_attr]; +"377 _scale_71" [id=377, type=get_attr]; +"378 _zero_point_71" [id=378, type=get_attr]; +"379 _frozen_param113_1" [id=379, type=get_attr]; +"380 dequantize_per_channel_default_71" [id=380, type=dequantize_per_channel]; +"381 features_5_6_mlp_0_bias" [id=381, type=get_attr]; +"382 _scale_72" [id=382, type=get_attr]; +"383 _zero_point_72" [id=383, type=get_attr]; +"384 _frozen_param114_1" [id=384, type=get_attr]; +"385 dequantize_per_channel_default_72" [id=385, type=dequantize_per_channel]; +"386 features_5_6_mlp_3_bias" [id=386, type=get_attr]; +"387 features_5_6_norm2_weight" [id=387, type=get_attr]; +"388 features_5_6_norm2_bias" [id=388, type=get_attr]; +"389 features_5_7_norm1_weight" [id=389, type=get_attr]; +"390 features_5_7_norm1_bias" [id=390, type=get_attr]; +"391 _scale_73" [id=391, type=get_attr]; +"392 _zero_point_73" [id=392, type=get_attr]; +"393 _frozen_param115_1" [id=393, type=get_attr]; +"394 dequantize_per_channel_default_73" [id=394, type=dequantize_per_channel]; +"395 features_5_7_mlp_0_bias" [id=395, type=get_attr]; +"396 _scale_74" [id=396, type=get_attr]; +"397 _zero_point_74" [id=397, type=get_attr]; +"398 _frozen_param116_1" [id=398, type=get_attr]; +"399 dequantize_per_channel_default_74" [id=399, type=dequantize_per_channel]; +"400 features_5_7_mlp_3_bias" [id=400, type=get_attr]; +"401 features_5_7_norm2_weight" [id=401, type=get_attr]; +"402 features_5_7_norm2_bias" [id=402, type=get_attr]; +"403 features_5_8_norm1_weight" [id=403, type=get_attr]; +"404 features_5_8_norm1_bias" [id=404, type=get_attr]; +"405 _scale_75" [id=405, type=get_attr]; +"406 _zero_point_75" [id=406, type=get_attr]; +"407 _frozen_param117_1" [id=407, type=get_attr]; +"408 dequantize_per_channel_default_75" [id=408, type=dequantize_per_channel]; +"409 features_5_8_mlp_0_bias" [id=409, type=get_attr]; +"410 _scale_76" [id=410, type=get_attr]; +"411 _zero_point_76" [id=411, type=get_attr]; +"412 _frozen_param118_1" [id=412, type=get_attr]; +"413 dequantize_per_channel_default_76" [id=413, type=dequantize_per_channel]; +"414 features_5_8_mlp_3_bias" [id=414, type=get_attr]; +"415 features_5_8_norm2_weight" [id=415, type=get_attr]; +"416 features_5_8_norm2_bias" [id=416, type=get_attr]; +"417 features_5_9_norm1_weight" [id=417, type=get_attr]; +"418 features_5_9_norm1_bias" [id=418, type=get_attr]; +"419 _scale_77" [id=419, type=get_attr]; +"420 _zero_point_77" [id=420, type=get_attr]; +"421 _frozen_param121_1" [id=421, type=get_attr]; +"422 dequantize_per_channel_default_77" [id=422, type=dequantize_per_channel]; +"423 features_5_9_mlp_0_bias" [id=423, type=get_attr]; +"424 _scale_78" [id=424, type=get_attr]; +"425 _zero_point_78" [id=425, type=get_attr]; +"426 _frozen_param125_1" [id=426, type=get_attr]; +"427 dequantize_per_channel_default_78" [id=427, type=dequantize_per_channel]; +"428 features_5_9_mlp_3_bias" [id=428, type=get_attr]; +"429 features_5_9_norm2_weight" [id=429, type=get_attr]; +"430 features_5_9_norm2_bias" [id=430, type=get_attr]; +"431 features_5_10_norm1_weight" [id=431, type=get_attr]; +"432 features_5_10_norm1_bias" [id=432, type=get_attr]; +"433 _scale_79" [id=433, type=get_attr]; +"434 _zero_point_79" [id=434, type=get_attr]; +"435 _frozen_param127_1" [id=435, type=get_attr]; +"436 dequantize_per_channel_default_79" [id=436, type=dequantize_per_channel]; +"437 features_5_10_mlp_0_bias" [id=437, type=get_attr]; +"438 _scale_80" [id=438, type=get_attr]; +"439 _zero_point_80" [id=439, type=get_attr]; +"440 _frozen_param128_1" [id=440, type=get_attr]; +"441 dequantize_per_channel_default_80" [id=441, type=dequantize_per_channel]; +"442 features_5_10_mlp_3_bias" [id=442, type=get_attr]; +"443 features_5_10_norm2_weight" [id=443, type=get_attr]; +"444 features_5_10_norm2_bias" [id=444, type=get_attr]; +"445 features_5_11_norm1_weight" [id=445, type=get_attr]; +"446 features_5_11_norm1_bias" [id=446, type=get_attr]; +"447 _scale_81" [id=447, type=get_attr]; +"448 _zero_point_81" [id=448, type=get_attr]; +"449 _frozen_param129_1" [id=449, type=get_attr]; +"450 dequantize_per_channel_default_81" [id=450, type=dequantize_per_channel]; +"451 features_5_11_mlp_0_bias" [id=451, type=get_attr]; +"452 _scale_82" [id=452, type=get_attr]; +"453 _zero_point_82" [id=453, type=get_attr]; +"454 _frozen_param130_1" [id=454, type=get_attr]; +"455 dequantize_per_channel_default_82" [id=455, type=dequantize_per_channel]; +"456 features_5_11_mlp_3_bias" [id=456, type=get_attr]; +"457 features_5_11_norm2_weight" [id=457, type=get_attr]; +"458 features_5_11_norm2_bias" [id=458, type=get_attr]; +"459 features_5_12_norm1_weight" [id=459, type=get_attr]; +"460 features_5_12_norm1_bias" [id=460, type=get_attr]; +"461 _scale_83" [id=461, type=get_attr]; +"462 _zero_point_83" [id=462, type=get_attr]; +"463 _frozen_param131_1" [id=463, type=get_attr]; +"464 dequantize_per_channel_default_83" [id=464, type=dequantize_per_channel]; +"465 features_5_12_mlp_0_bias" [id=465, type=get_attr]; +"466 _scale_84" [id=466, type=get_attr]; +"467 _zero_point_84" [id=467, type=get_attr]; +"468 _frozen_param132_1" [id=468, type=get_attr]; +"469 dequantize_per_channel_default_84" [id=469, type=dequantize_per_channel]; +"470 features_5_12_mlp_3_bias" [id=470, type=get_attr]; +"471 features_5_12_norm2_weight" [id=471, type=get_attr]; +"472 features_5_12_norm2_bias" [id=472, type=get_attr]; +"473 features_5_13_norm1_weight" [id=473, type=get_attr]; +"474 features_5_13_norm1_bias" [id=474, type=get_attr]; +"475 _scale_85" [id=475, type=get_attr]; +"476 _zero_point_85" [id=476, type=get_attr]; +"477 _frozen_param133_1" [id=477, type=get_attr]; +"478 dequantize_per_channel_default_85" [id=478, type=dequantize_per_channel]; +"479 features_5_13_mlp_0_bias" [id=479, type=get_attr]; +"480 _scale_86" [id=480, type=get_attr]; +"481 _zero_point_86" [id=481, type=get_attr]; +"482 _frozen_param134_1" [id=482, type=get_attr]; +"483 dequantize_per_channel_default_86" [id=483, type=dequantize_per_channel]; +"484 features_5_13_mlp_3_bias" [id=484, type=get_attr]; +"485 features_5_13_norm2_weight" [id=485, type=get_attr]; +"486 features_5_13_norm2_bias" [id=486, type=get_attr]; +"487 features_5_14_norm1_weight" [id=487, type=get_attr]; +"488 features_5_14_norm1_bias" [id=488, type=get_attr]; +"489 _scale_87" [id=489, type=get_attr]; +"490 _zero_point_87" [id=490, type=get_attr]; +"491 _frozen_param135_1" [id=491, type=get_attr]; +"492 dequantize_per_channel_default_87" [id=492, type=dequantize_per_channel]; +"493 features_5_14_mlp_0_bias" [id=493, type=get_attr]; +"494 _scale_88" [id=494, type=get_attr]; +"495 _zero_point_88" [id=495, type=get_attr]; +"496 _frozen_param138_1" [id=496, type=get_attr]; +"497 dequantize_per_channel_default_88" [id=497, type=dequantize_per_channel]; +"498 features_5_14_mlp_3_bias" [id=498, type=get_attr]; +"499 features_5_14_norm2_weight" [id=499, type=get_attr]; +"500 features_5_14_norm2_bias" [id=500, type=get_attr]; +"501 features_5_15_norm1_weight" [id=501, type=get_attr]; +"502 features_5_15_norm1_bias" [id=502, type=get_attr]; +"503 _scale_89" [id=503, type=get_attr]; +"504 _zero_point_89" [id=504, type=get_attr]; +"505 _frozen_param142_1" [id=505, type=get_attr]; +"506 dequantize_per_channel_default_89" [id=506, type=dequantize_per_channel]; +"507 features_5_15_mlp_0_bias" [id=507, type=get_attr]; +"508 _scale_90" [id=508, type=get_attr]; +"509 _zero_point_90" [id=509, type=get_attr]; +"510 _frozen_param144_1" [id=510, type=get_attr]; +"511 dequantize_per_channel_default_90" [id=511, type=dequantize_per_channel]; +"512 features_5_15_mlp_3_bias" [id=512, type=get_attr]; +"513 features_5_15_norm2_weight" [id=513, type=get_attr]; +"514 features_5_15_norm2_bias" [id=514, type=get_attr]; +"515 features_5_16_norm1_weight" [id=515, type=get_attr]; +"516 features_5_16_norm1_bias" [id=516, type=get_attr]; +"517 _scale_91" [id=517, type=get_attr]; +"518 _zero_point_91" [id=518, type=get_attr]; +"519 _frozen_param145_1" [id=519, type=get_attr]; +"520 dequantize_per_channel_default_91" [id=520, type=dequantize_per_channel]; +"521 features_5_16_mlp_0_bias" [id=521, type=get_attr]; +"522 _scale_92" [id=522, type=get_attr]; +"523 _zero_point_92" [id=523, type=get_attr]; +"524 _frozen_param146_1" [id=524, type=get_attr]; +"525 dequantize_per_channel_default_92" [id=525, type=dequantize_per_channel]; +"526 features_5_16_mlp_3_bias" [id=526, type=get_attr]; +"527 features_5_16_norm2_weight" [id=527, type=get_attr]; +"528 features_5_16_norm2_bias" [id=528, type=get_attr]; +"529 features_5_17_norm1_weight" [id=529, type=get_attr]; +"530 features_5_17_norm1_bias" [id=530, type=get_attr]; +"531 _scale_93" [id=531, type=get_attr]; +"532 _zero_point_93" [id=532, type=get_attr]; +"533 _frozen_param147_1" [id=533, type=get_attr]; +"534 dequantize_per_channel_default_93" [id=534, type=dequantize_per_channel]; +"535 features_5_17_mlp_0_bias" [id=535, type=get_attr]; +"536 _scale_94" [id=536, type=get_attr]; +"537 _zero_point_94" [id=537, type=get_attr]; +"538 _frozen_param148_1" [id=538, type=get_attr]; +"539 dequantize_per_channel_default_94" [id=539, type=dequantize_per_channel]; +"540 features_5_17_mlp_3_bias" [id=540, type=get_attr]; +"541 features_5_17_norm2_weight" [id=541, type=get_attr]; +"542 features_5_17_norm2_bias" [id=542, type=get_attr]; +"543 _scale_95" [id=543, type=get_attr]; +"544 _zero_point_95" [id=544, type=get_attr]; +"545 _frozen_param149_1" [id=545, type=get_attr]; +"546 dequantize_per_channel_default_95" [id=546, type=dequantize_per_channel]; +"547 features_6_norm_weight" [id=547, type=get_attr]; +"548 features_6_norm_bias" [id=548, type=get_attr]; +"549 features_7_0_norm1_weight" [id=549, type=get_attr]; +"550 features_7_0_norm1_bias" [id=550, type=get_attr]; +"551 _scale_96" [id=551, type=get_attr]; +"552 _zero_point_96" [id=552, type=get_attr]; +"553 _frozen_param150_1" [id=553, type=get_attr]; +"554 dequantize_per_channel_default_96" [id=554, type=dequantize_per_channel]; +"555 features_7_0_mlp_0_bias" [id=555, type=get_attr]; +"556 _scale_97" [id=556, type=get_attr]; +"557 _zero_point_97" [id=557, type=get_attr]; +"558 _frozen_param151_1" [id=558, type=get_attr]; +"559 dequantize_per_channel_default_97" [id=559, type=dequantize_per_channel]; +"560 features_7_0_mlp_3_bias" [id=560, type=get_attr]; +"561 features_7_0_norm2_weight" [id=561, type=get_attr]; +"562 features_7_0_norm2_bias" [id=562, type=get_attr]; +"563 features_7_1_norm1_weight" [id=563, type=get_attr]; +"564 features_7_1_norm1_bias" [id=564, type=get_attr]; +"565 _scale_98" [id=565, type=get_attr]; +"566 _zero_point_98" [id=566, type=get_attr]; +"567 _frozen_param152_1" [id=567, type=get_attr]; +"568 dequantize_per_channel_default_98" [id=568, type=dequantize_per_channel]; +"569 features_7_1_mlp_0_bias" [id=569, type=get_attr]; +"570 _scale_99" [id=570, type=get_attr]; +"571 _zero_point_99" [id=571, type=get_attr]; +"572 _frozen_param155_1" [id=572, type=get_attr]; +"573 dequantize_per_channel_default_99" [id=573, type=dequantize_per_channel]; +"574 features_7_1_mlp_3_bias" [id=574, type=get_attr]; +"575 features_7_1_norm2_weight" [id=575, type=get_attr]; +"576 features_7_1_norm2_bias" [id=576, type=get_attr]; +"577 norm_weight" [id=577, type=get_attr]; +"578 norm_bias" [id=578, type=get_attr]; +"579 _scale_100" [id=579, type=get_attr]; +"580 _zero_point_100" [id=580, type=get_attr]; +"581 _frozen_param159_1" [id=581, type=get_attr]; +"582 dequantize_per_channel_default_100" [id=582, type=dequantize_per_channel]; +"583 head_bias" [id=583, type=get_attr]; +"584 x" [id=584, type=input]; +"585 quantize_per_tensor_default" [id=585, type=quantize_per_tensor]; +"586 dequantize_per_tensor_default" [id=586, type=dequantize_per_tensor]; +"587 conv2d" [id=587, type=conv2d]; +"588 permute" [id=588, type=permute]; +"589 layer_norm" [id=589, type=layer_norm]; +"590 quantize_per_tensor_default_1" [id=590, type=quantize_per_tensor]; +"591 dequantize_per_tensor_default_1" [id=591, type=dequantize_per_tensor]; +"592 _frozen_param0" [id=592, type=get_attr]; +"593 pad" [id=593, type=pad]; +"594 view_2" [id=594, type=view]; +"595 permute_2" [id=595, type=permute]; +"596 reshape" [id=596, type=reshape]; +"597 _frozen_param1" [id=597, type=get_attr]; +"598 linear_2" [id=598, type=linear]; +"599 reshape_1" [id=599, type=reshape]; +"600 permute_3" [id=600, type=permute]; +"601 select" [id=601, type=select]; +"602 select_1" [id=602, type=select]; +"603 select_2" [id=603, type=select]; +"604 linalg_vector_norm" [id=604, type=linalg_vector_norm]; +"605 clamp_min" [id=605, type=clamp_min]; +"606 expand_as" [id=606, type=expand_as]; +"607 div" [id=607, type=div]; +"608 quantize_per_tensor_default_2" [id=608, type=quantize_per_tensor]; +"609 dequantize_per_tensor_default_2" [id=609, type=dequantize_per_tensor]; +"610 linalg_vector_norm_1" [id=610, type=linalg_vector_norm]; +"611 clamp_min_1" [id=611, type=clamp_min]; +"612 expand_as_1" [id=612, type=expand_as]; +"613 div_1" [id=613, type=div]; +"614 quantize_per_tensor_default_3" [id=614, type=quantize_per_tensor]; +"615 dequantize_per_tensor_default_3" [id=615, type=dequantize_per_tensor]; +"616 transpose" [id=616, type=transpose]; +"617 matmul" [id=617, type=matmul]; +"618 _frozen_param3" [id=618, type=get_attr]; +"619 mul_1" [id=619, type=mul]; +"620 add" [id=620, type=add]; +"621 softmax" [id=621, type=softmax]; +"622 dropout" [id=622, type=dropout]; +"623 matmul_1" [id=623, type=matmul]; +"624 quantize_per_tensor_default_4" [id=624, type=quantize_per_tensor]; +"625 dequantize_per_tensor_default_4" [id=625, type=dequantize_per_tensor]; +"626 transpose_1" [id=626, type=transpose]; +"627 reshape_2" [id=627, type=reshape]; +"628 linear_3" [id=628, type=linear]; +"629 dropout_1" [id=629, type=dropout]; +"630 view_3" [id=630, type=view]; +"631 permute_4" [id=631, type=permute]; +"632 reshape_3" [id=632, type=reshape]; +"633 slice_2" [id=633, type=slice]; +"634 slice_3" [id=634, type=slice]; +"635 layer_norm_1" [id=635, type=layer_norm]; +"636 add_1" [id=636, type=add]; +"637 quantize_per_tensor_default_5" [id=637, type=quantize_per_tensor]; +"638 dequantize_per_tensor_default_5" [id=638, type=dequantize_per_tensor]; +"639 linear_4" [id=639, type=linear]; +"640 gelu" [id=640, type=gelu]; +"641 quantize_per_tensor_default_6" [id=641, type=quantize_per_tensor]; +"642 dequantize_per_tensor_default_6" [id=642, type=dequantize_per_tensor]; +"643 dropout_2" [id=643, type=dropout]; +"644 linear_5" [id=644, type=linear]; +"645 dropout_3" [id=645, type=dropout]; +"646 layer_norm_2" [id=646, type=layer_norm]; +"647 add_2" [id=647, type=add]; +"648 _frozen_param4" [id=648, type=get_attr]; +"649 pad_1" [id=649, type=pad]; +"650 roll" [id=650, type=roll]; +"651 view_6" [id=651, type=view]; +"652 permute_6" [id=652, type=permute]; +"653 reshape_4" [id=653, type=reshape]; +"654 quantize_per_tensor_default_7" [id=654, type=quantize_per_tensor]; +"655 dequantize_per_tensor_default_7" [id=655, type=dequantize_per_tensor]; +"656 _frozen_param5" [id=656, type=get_attr]; +"657 linear_8" [id=657, type=linear]; +"658 reshape_5" [id=658, type=reshape]; +"659 permute_7" [id=659, type=permute]; +"660 select_3" [id=660, type=select]; +"661 select_4" [id=661, type=select]; +"662 select_5" [id=662, type=select]; +"663 linalg_vector_norm_2" [id=663, type=linalg_vector_norm]; +"664 clamp_min_2" [id=664, type=clamp_min]; +"665 expand_as_2" [id=665, type=expand_as]; +"666 div_2" [id=666, type=div]; +"667 quantize_per_tensor_default_8" [id=667, type=quantize_per_tensor]; +"668 dequantize_per_tensor_default_8" [id=668, type=dequantize_per_tensor]; +"669 linalg_vector_norm_3" [id=669, type=linalg_vector_norm]; +"670 clamp_min_3" [id=670, type=clamp_min]; +"671 expand_as_3" [id=671, type=expand_as]; +"672 div_3" [id=672, type=div]; +"673 quantize_per_tensor_default_9" [id=673, type=quantize_per_tensor]; +"674 dequantize_per_tensor_default_9" [id=674, type=dequantize_per_tensor]; +"675 transpose_2" [id=675, type=transpose]; +"676 matmul_2" [id=676, type=matmul]; +"677 _frozen_param7" [id=677, type=get_attr]; +"678 mul_3" [id=678, type=mul]; +"679 add_3" [id=679, type=add]; +"680 new_zeros" [id=680, type=new_zeros]; +"681 view_7" [id=681, type=view]; +"682 permute_8" [id=682, type=permute]; +"683 reshape_6" [id=683, type=reshape]; +"684 unsqueeze_2" [id=684, type=unsqueeze]; +"685 unsqueeze_3" [id=685, type=unsqueeze]; +"686 sub" [id=686, type=sub]; +"687 ne" [id=687, type=ne]; +"688 masked_fill" [id=688, type=masked_fill]; +"689 eq" [id=689, type=eq]; +"690 masked_fill_1" [id=690, type=masked_fill]; +"691 view_8" [id=691, type=view]; +"692 unsqueeze_4" [id=692, type=unsqueeze]; +"693 unsqueeze_5" [id=693, type=unsqueeze]; +"694 add_4" [id=694, type=add]; +"695 view_9" [id=695, type=view]; +"696 softmax_1" [id=696, type=softmax]; +"697 dropout_4" [id=697, type=dropout]; +"698 matmul_3" [id=698, type=matmul]; +"699 quantize_per_tensor_default_10" [id=699, type=quantize_per_tensor]; +"700 dequantize_per_tensor_default_10" [id=700, type=dequantize_per_tensor]; +"701 transpose_3" [id=701, type=transpose]; +"702 reshape_7" [id=702, type=reshape]; +"703 linear_9" [id=703, type=linear]; +"704 dropout_5" [id=704, type=dropout]; +"705 view_10" [id=705, type=view]; +"706 permute_9" [id=706, type=permute]; +"707 reshape_8" [id=707, type=reshape]; +"708 roll_1" [id=708, type=roll]; +"709 slice_23" [id=709, type=slice]; +"710 slice_24" [id=710, type=slice]; +"711 layer_norm_3" [id=711, type=layer_norm]; +"712 add_5" [id=712, type=add]; +"713 quantize_per_tensor_default_11" [id=713, type=quantize_per_tensor]; +"714 dequantize_per_tensor_default_11" [id=714, type=dequantize_per_tensor]; +"715 linear_10" [id=715, type=linear]; +"716 gelu_1" [id=716, type=gelu]; +"717 quantize_per_tensor_default_12" [id=717, type=quantize_per_tensor]; +"718 dequantize_per_tensor_default_12" [id=718, type=dequantize_per_tensor]; +"719 dropout_6" [id=719, type=dropout]; +"720 linear_11" [id=720, type=linear]; +"721 dropout_7" [id=721, type=dropout]; +"722 layer_norm_4" [id=722, type=layer_norm]; +"723 add_6" [id=723, type=add]; +"724 quantize_per_tensor_default_13" [id=724, type=quantize_per_tensor]; +"725 dequantize_per_tensor_default_13" [id=725, type=dequantize_per_tensor]; +"726 pad_2" [id=726, type=pad]; +"727 slice_25" [id=727, type=slice]; +"728 slice_26" [id=728, type=slice]; +"729 slice_27" [id=729, type=slice]; +"730 slice_28" [id=730, type=slice]; +"731 slice_29" [id=731, type=slice]; +"732 slice_30" [id=732, type=slice]; +"733 slice_31" [id=733, type=slice]; +"734 slice_32" [id=734, type=slice]; +"735 slice_33" [id=735, type=slice]; +"736 slice_34" [id=736, type=slice]; +"737 slice_35" [id=737, type=slice]; +"738 slice_36" [id=738, type=slice]; +"739 cat" [id=739, type=cat]; +"740 linear_12" [id=740, type=linear]; +"741 layer_norm_5" [id=741, type=layer_norm]; +"742 quantize_per_tensor_default_14" [id=742, type=quantize_per_tensor]; +"743 dequantize_per_tensor_default_14" [id=743, type=dequantize_per_tensor]; +"744 _frozen_param17" [id=744, type=get_attr]; +"745 pad_3" [id=745, type=pad]; +"746 view_13" [id=746, type=view]; +"747 permute_11" [id=747, type=permute]; +"748 reshape_9" [id=748, type=reshape]; +"749 _frozen_param18" [id=749, type=get_attr]; +"750 linear_15" [id=750, type=linear]; +"751 reshape_10" [id=751, type=reshape]; +"752 permute_12" [id=752, type=permute]; +"753 select_6" [id=753, type=select]; +"754 select_7" [id=754, type=select]; +"755 select_8" [id=755, type=select]; +"756 linalg_vector_norm_4" [id=756, type=linalg_vector_norm]; +"757 clamp_min_4" [id=757, type=clamp_min]; +"758 expand_as_4" [id=758, type=expand_as]; +"759 div_4" [id=759, type=div]; +"760 quantize_per_tensor_default_15" [id=760, type=quantize_per_tensor]; +"761 dequantize_per_tensor_default_15" [id=761, type=dequantize_per_tensor]; +"762 linalg_vector_norm_5" [id=762, type=linalg_vector_norm]; +"763 clamp_min_5" [id=763, type=clamp_min]; +"764 expand_as_5" [id=764, type=expand_as]; +"765 div_5" [id=765, type=div]; +"766 quantize_per_tensor_default_16" [id=766, type=quantize_per_tensor]; +"767 dequantize_per_tensor_default_16" [id=767, type=dequantize_per_tensor]; +"768 transpose_4" [id=768, type=transpose]; +"769 matmul_4" [id=769, type=matmul]; +"770 _frozen_param20" [id=770, type=get_attr]; +"771 mul_5" [id=771, type=mul]; +"772 add_7" [id=772, type=add]; +"773 softmax_2" [id=773, type=softmax]; +"774 dropout_8" [id=774, type=dropout]; +"775 matmul_5" [id=775, type=matmul]; +"776 quantize_per_tensor_default_17" [id=776, type=quantize_per_tensor]; +"777 dequantize_per_tensor_default_17" [id=777, type=dequantize_per_tensor]; +"778 transpose_5" [id=778, type=transpose]; +"779 reshape_11" [id=779, type=reshape]; +"780 linear_16" [id=780, type=linear]; +"781 dropout_9" [id=781, type=dropout]; +"782 view_14" [id=782, type=view]; +"783 permute_13" [id=783, type=permute]; +"784 reshape_12" [id=784, type=reshape]; +"785 slice_38" [id=785, type=slice]; +"786 slice_39" [id=786, type=slice]; +"787 slice_40" [id=787, type=slice]; +"788 slice_41" [id=788, type=slice]; +"789 contiguous_3" [id=789, type=contiguous]; +"790 layer_norm_6" [id=790, type=layer_norm]; +"791 add_8" [id=791, type=add]; +"792 quantize_per_tensor_default_18" [id=792, type=quantize_per_tensor]; +"793 dequantize_per_tensor_default_18" [id=793, type=dequantize_per_tensor]; +"794 linear_17" [id=794, type=linear]; +"795 gelu_2" [id=795, type=gelu]; +"796 quantize_per_tensor_default_19" [id=796, type=quantize_per_tensor]; +"797 dequantize_per_tensor_default_19" [id=797, type=dequantize_per_tensor]; +"798 dropout_10" [id=798, type=dropout]; +"799 linear_18" [id=799, type=linear]; +"800 dropout_11" [id=800, type=dropout]; +"801 layer_norm_7" [id=801, type=layer_norm]; +"802 add_9" [id=802, type=add]; +"803 _frozen_param21" [id=803, type=get_attr]; +"804 pad_4" [id=804, type=pad]; +"805 roll_2" [id=805, type=roll]; +"806 view_17" [id=806, type=view]; +"807 permute_15" [id=807, type=permute]; +"808 reshape_13" [id=808, type=reshape]; +"809 quantize_per_tensor_default_20" [id=809, type=quantize_per_tensor]; +"810 dequantize_per_tensor_default_20" [id=810, type=dequantize_per_tensor]; +"811 _frozen_param22" [id=811, type=get_attr]; +"812 linear_21" [id=812, type=linear]; +"813 reshape_14" [id=813, type=reshape]; +"814 permute_16" [id=814, type=permute]; +"815 select_9" [id=815, type=select]; +"816 select_10" [id=816, type=select]; +"817 select_11" [id=817, type=select]; +"818 linalg_vector_norm_6" [id=818, type=linalg_vector_norm]; +"819 clamp_min_6" [id=819, type=clamp_min]; +"820 expand_as_6" [id=820, type=expand_as]; +"821 div_6" [id=821, type=div]; +"822 quantize_per_tensor_default_21" [id=822, type=quantize_per_tensor]; +"823 dequantize_per_tensor_default_21" [id=823, type=dequantize_per_tensor]; +"824 linalg_vector_norm_7" [id=824, type=linalg_vector_norm]; +"825 clamp_min_7" [id=825, type=clamp_min]; +"826 expand_as_7" [id=826, type=expand_as]; +"827 div_7" [id=827, type=div]; +"828 quantize_per_tensor_default_22" [id=828, type=quantize_per_tensor]; +"829 dequantize_per_tensor_default_22" [id=829, type=dequantize_per_tensor]; +"830 transpose_6" [id=830, type=transpose]; +"831 matmul_6" [id=831, type=matmul]; +"832 _frozen_param24" [id=832, type=get_attr]; +"833 mul_7" [id=833, type=mul]; +"834 add_10" [id=834, type=add]; +"835 new_zeros_1" [id=835, type=new_zeros]; +"836 view_18" [id=836, type=view]; +"837 permute_17" [id=837, type=permute]; +"838 reshape_15" [id=838, type=reshape]; +"839 unsqueeze_8" [id=839, type=unsqueeze]; +"840 unsqueeze_9" [id=840, type=unsqueeze]; +"841 sub_1" [id=841, type=sub]; +"842 ne_1" [id=842, type=ne]; +"843 masked_fill_2" [id=843, type=masked_fill]; +"844 eq_1" [id=844, type=eq]; +"845 masked_fill_3" [id=845, type=masked_fill]; +"846 view_19" [id=846, type=view]; +"847 unsqueeze_10" [id=847, type=unsqueeze]; +"848 unsqueeze_11" [id=848, type=unsqueeze]; +"849 add_11" [id=849, type=add]; +"850 view_20" [id=850, type=view]; +"851 softmax_3" [id=851, type=softmax]; +"852 dropout_12" [id=852, type=dropout]; +"853 matmul_7" [id=853, type=matmul]; +"854 quantize_per_tensor_default_23" [id=854, type=quantize_per_tensor]; +"855 dequantize_per_tensor_default_23" [id=855, type=dequantize_per_tensor]; +"856 transpose_7" [id=856, type=transpose]; +"857 reshape_16" [id=857, type=reshape]; +"858 linear_22" [id=858, type=linear]; +"859 dropout_13" [id=859, type=dropout]; +"860 view_21" [id=860, type=view]; +"861 permute_18" [id=861, type=permute]; +"862 reshape_17" [id=862, type=reshape]; +"863 roll_3" [id=863, type=roll]; +"864 slice_61" [id=864, type=slice]; +"865 slice_62" [id=865, type=slice]; +"866 slice_63" [id=866, type=slice]; +"867 slice_64" [id=867, type=slice]; +"868 contiguous_5" [id=868, type=contiguous]; +"869 layer_norm_8" [id=869, type=layer_norm]; +"870 add_12" [id=870, type=add]; +"871 quantize_per_tensor_default_24" [id=871, type=quantize_per_tensor]; +"872 dequantize_per_tensor_default_24" [id=872, type=dequantize_per_tensor]; +"873 linear_23" [id=873, type=linear]; +"874 gelu_3" [id=874, type=gelu]; +"875 quantize_per_tensor_default_25" [id=875, type=quantize_per_tensor]; +"876 dequantize_per_tensor_default_25" [id=876, type=dequantize_per_tensor]; +"877 dropout_14" [id=877, type=dropout]; +"878 linear_24" [id=878, type=linear]; +"879 dropout_15" [id=879, type=dropout]; +"880 layer_norm_9" [id=880, type=layer_norm]; +"881 add_13" [id=881, type=add]; +"882 quantize_per_tensor_default_26" [id=882, type=quantize_per_tensor]; +"883 dequantize_per_tensor_default_26" [id=883, type=dequantize_per_tensor]; +"884 pad_5" [id=884, type=pad]; +"885 slice_65" [id=885, type=slice]; +"886 slice_66" [id=886, type=slice]; +"887 slice_67" [id=887, type=slice]; +"888 slice_68" [id=888, type=slice]; +"889 slice_69" [id=889, type=slice]; +"890 slice_70" [id=890, type=slice]; +"891 slice_71" [id=891, type=slice]; +"892 slice_72" [id=892, type=slice]; +"893 slice_73" [id=893, type=slice]; +"894 slice_74" [id=894, type=slice]; +"895 slice_75" [id=895, type=slice]; +"896 slice_76" [id=896, type=slice]; +"897 cat_1" [id=897, type=cat]; +"898 linear_25" [id=898, type=linear]; +"899 layer_norm_10" [id=899, type=layer_norm]; +"900 quantize_per_tensor_default_27" [id=900, type=quantize_per_tensor]; +"901 dequantize_per_tensor_default_27" [id=901, type=dequantize_per_tensor]; +"902 _frozen_param34" [id=902, type=get_attr]; +"903 pad_6" [id=903, type=pad]; +"904 view_24" [id=904, type=view]; +"905 permute_20" [id=905, type=permute]; +"906 reshape_18" [id=906, type=reshape]; +"907 _frozen_param35" [id=907, type=get_attr]; +"908 linear_28" [id=908, type=linear]; +"909 reshape_19" [id=909, type=reshape]; +"910 permute_21" [id=910, type=permute]; +"911 select_12" [id=911, type=select]; +"912 select_13" [id=912, type=select]; +"913 select_14" [id=913, type=select]; +"914 linalg_vector_norm_8" [id=914, type=linalg_vector_norm]; +"915 clamp_min_8" [id=915, type=clamp_min]; +"916 expand_as_8" [id=916, type=expand_as]; +"917 div_8" [id=917, type=div]; +"918 quantize_per_tensor_default_28" [id=918, type=quantize_per_tensor]; +"919 dequantize_per_tensor_default_28" [id=919, type=dequantize_per_tensor]; +"920 linalg_vector_norm_9" [id=920, type=linalg_vector_norm]; +"921 clamp_min_9" [id=921, type=clamp_min]; +"922 expand_as_9" [id=922, type=expand_as]; +"923 div_9" [id=923, type=div]; +"924 quantize_per_tensor_default_29" [id=924, type=quantize_per_tensor]; +"925 dequantize_per_tensor_default_29" [id=925, type=dequantize_per_tensor]; +"926 transpose_8" [id=926, type=transpose]; +"927 matmul_8" [id=927, type=matmul]; +"928 _frozen_param37" [id=928, type=get_attr]; +"929 mul_9" [id=929, type=mul]; +"930 add_14" [id=930, type=add]; +"931 softmax_4" [id=931, type=softmax]; +"932 dropout_16" [id=932, type=dropout]; +"933 matmul_9" [id=933, type=matmul]; +"934 quantize_per_tensor_default_30" [id=934, type=quantize_per_tensor]; +"935 dequantize_per_tensor_default_30" [id=935, type=dequantize_per_tensor]; +"936 transpose_9" [id=936, type=transpose]; +"937 reshape_20" [id=937, type=reshape]; +"938 linear_29" [id=938, type=linear]; +"939 dropout_17" [id=939, type=dropout]; +"940 view_25" [id=940, type=view]; +"941 permute_22" [id=941, type=permute]; +"942 reshape_21" [id=942, type=reshape]; +"943 slice_78" [id=943, type=slice]; +"944 slice_79" [id=944, type=slice]; +"945 slice_80" [id=945, type=slice]; +"946 slice_81" [id=946, type=slice]; +"947 contiguous_7" [id=947, type=contiguous]; +"948 layer_norm_11" [id=948, type=layer_norm]; +"949 add_15" [id=949, type=add]; +"950 quantize_per_tensor_default_31" [id=950, type=quantize_per_tensor]; +"951 dequantize_per_tensor_default_31" [id=951, type=dequantize_per_tensor]; +"952 linear_30" [id=952, type=linear]; +"953 gelu_4" [id=953, type=gelu]; +"954 quantize_per_tensor_default_32" [id=954, type=quantize_per_tensor]; +"955 dequantize_per_tensor_default_32" [id=955, type=dequantize_per_tensor]; +"956 dropout_18" [id=956, type=dropout]; +"957 linear_31" [id=957, type=linear]; +"958 dropout_19" [id=958, type=dropout]; +"959 layer_norm_12" [id=959, type=layer_norm]; +"960 add_16" [id=960, type=add]; +"961 _frozen_param38" [id=961, type=get_attr]; +"962 pad_7" [id=962, type=pad]; +"963 roll_4" [id=963, type=roll]; +"964 view_28" [id=964, type=view]; +"965 permute_24" [id=965, type=permute]; +"966 reshape_22" [id=966, type=reshape]; +"967 quantize_per_tensor_default_33" [id=967, type=quantize_per_tensor]; +"968 dequantize_per_tensor_default_33" [id=968, type=dequantize_per_tensor]; +"969 _frozen_param39" [id=969, type=get_attr]; +"970 linear_34" [id=970, type=linear]; +"971 reshape_23" [id=971, type=reshape]; +"972 permute_25" [id=972, type=permute]; +"973 select_15" [id=973, type=select]; +"974 select_16" [id=974, type=select]; +"975 select_17" [id=975, type=select]; +"976 linalg_vector_norm_10" [id=976, type=linalg_vector_norm]; +"977 clamp_min_10" [id=977, type=clamp_min]; +"978 expand_as_10" [id=978, type=expand_as]; +"979 div_10" [id=979, type=div]; +"980 quantize_per_tensor_default_34" [id=980, type=quantize_per_tensor]; +"981 dequantize_per_tensor_default_34" [id=981, type=dequantize_per_tensor]; +"982 linalg_vector_norm_11" [id=982, type=linalg_vector_norm]; +"983 clamp_min_11" [id=983, type=clamp_min]; +"984 expand_as_11" [id=984, type=expand_as]; +"985 div_11" [id=985, type=div]; +"986 quantize_per_tensor_default_35" [id=986, type=quantize_per_tensor]; +"987 dequantize_per_tensor_default_35" [id=987, type=dequantize_per_tensor]; +"988 transpose_10" [id=988, type=transpose]; +"989 matmul_10" [id=989, type=matmul]; +"990 _frozen_param41" [id=990, type=get_attr]; +"991 mul_11" [id=991, type=mul]; +"992 add_17" [id=992, type=add]; +"993 new_zeros_2" [id=993, type=new_zeros]; +"994 view_29" [id=994, type=view]; +"995 permute_26" [id=995, type=permute]; +"996 reshape_24" [id=996, type=reshape]; +"997 unsqueeze_14" [id=997, type=unsqueeze]; +"998 unsqueeze_15" [id=998, type=unsqueeze]; +"999 sub_2" [id=999, type=sub]; +"1000 ne_2" [id=1000, type=ne]; +"1001 masked_fill_4" [id=1001, type=masked_fill]; +"1002 eq_2" [id=1002, type=eq]; +"1003 masked_fill_5" [id=1003, type=masked_fill]; +"1004 view_30" [id=1004, type=view]; +"1005 unsqueeze_16" [id=1005, type=unsqueeze]; +"1006 unsqueeze_17" [id=1006, type=unsqueeze]; +"1007 add_18" [id=1007, type=add]; +"1008 view_31" [id=1008, type=view]; +"1009 softmax_5" [id=1009, type=softmax]; +"1010 dropout_20" [id=1010, type=dropout]; +"1011 matmul_11" [id=1011, type=matmul]; +"1012 quantize_per_tensor_default_36" [id=1012, type=quantize_per_tensor]; +"1013 dequantize_per_tensor_default_36" [id=1013, type=dequantize_per_tensor]; +"1014 transpose_11" [id=1014, type=transpose]; +"1015 reshape_25" [id=1015, type=reshape]; +"1016 linear_35" [id=1016, type=linear]; +"1017 dropout_21" [id=1017, type=dropout]; +"1018 view_32" [id=1018, type=view]; +"1019 permute_27" [id=1019, type=permute]; +"1020 reshape_26" [id=1020, type=reshape]; +"1021 roll_5" [id=1021, type=roll]; +"1022 slice_101" [id=1022, type=slice]; +"1023 slice_102" [id=1023, type=slice]; +"1024 slice_103" [id=1024, type=slice]; +"1025 slice_104" [id=1025, type=slice]; +"1026 contiguous_9" [id=1026, type=contiguous]; +"1027 layer_norm_13" [id=1027, type=layer_norm]; +"1028 add_19" [id=1028, type=add]; +"1029 quantize_per_tensor_default_37" [id=1029, type=quantize_per_tensor]; +"1030 dequantize_per_tensor_default_37" [id=1030, type=dequantize_per_tensor]; +"1031 linear_36" [id=1031, type=linear]; +"1032 gelu_5" [id=1032, type=gelu]; +"1033 quantize_per_tensor_default_38" [id=1033, type=quantize_per_tensor]; +"1034 dequantize_per_tensor_default_38" [id=1034, type=dequantize_per_tensor]; +"1035 dropout_22" [id=1035, type=dropout]; +"1036 linear_37" [id=1036, type=linear]; +"1037 dropout_23" [id=1037, type=dropout]; +"1038 layer_norm_14" [id=1038, type=layer_norm]; +"1039 add_20" [id=1039, type=add]; +"1040 quantize_per_tensor_default_39" [id=1040, type=quantize_per_tensor]; +"1041 dequantize_per_tensor_default_39" [id=1041, type=dequantize_per_tensor]; +"1042 _frozen_param51" [id=1042, type=get_attr]; +"1043 pad_8" [id=1043, type=pad]; +"1044 view_35" [id=1044, type=view]; +"1045 permute_29" [id=1045, type=permute]; +"1046 reshape_27" [id=1046, type=reshape]; +"1047 _frozen_param52" [id=1047, type=get_attr]; +"1048 linear_40" [id=1048, type=linear]; +"1049 reshape_28" [id=1049, type=reshape]; +"1050 permute_30" [id=1050, type=permute]; +"1051 select_18" [id=1051, type=select]; +"1052 select_19" [id=1052, type=select]; +"1053 select_20" [id=1053, type=select]; +"1054 linalg_vector_norm_12" [id=1054, type=linalg_vector_norm]; +"1055 clamp_min_12" [id=1055, type=clamp_min]; +"1056 expand_as_12" [id=1056, type=expand_as]; +"1057 div_12" [id=1057, type=div]; +"1058 quantize_per_tensor_default_40" [id=1058, type=quantize_per_tensor]; +"1059 dequantize_per_tensor_default_40" [id=1059, type=dequantize_per_tensor]; +"1060 linalg_vector_norm_13" [id=1060, type=linalg_vector_norm]; +"1061 clamp_min_13" [id=1061, type=clamp_min]; +"1062 expand_as_13" [id=1062, type=expand_as]; +"1063 div_13" [id=1063, type=div]; +"1064 quantize_per_tensor_default_41" [id=1064, type=quantize_per_tensor]; +"1065 dequantize_per_tensor_default_41" [id=1065, type=dequantize_per_tensor]; +"1066 transpose_12" [id=1066, type=transpose]; +"1067 matmul_12" [id=1067, type=matmul]; +"1068 _frozen_param54" [id=1068, type=get_attr]; +"1069 mul_13" [id=1069, type=mul]; +"1070 add_21" [id=1070, type=add]; +"1071 softmax_6" [id=1071, type=softmax]; +"1072 dropout_24" [id=1072, type=dropout]; +"1073 matmul_13" [id=1073, type=matmul]; +"1074 quantize_per_tensor_default_42" [id=1074, type=quantize_per_tensor]; +"1075 dequantize_per_tensor_default_42" [id=1075, type=dequantize_per_tensor]; +"1076 transpose_13" [id=1076, type=transpose]; +"1077 reshape_29" [id=1077, type=reshape]; +"1078 linear_41" [id=1078, type=linear]; +"1079 dropout_25" [id=1079, type=dropout]; +"1080 view_36" [id=1080, type=view]; +"1081 permute_31" [id=1081, type=permute]; +"1082 reshape_30" [id=1082, type=reshape]; +"1083 slice_106" [id=1083, type=slice]; +"1084 slice_107" [id=1084, type=slice]; +"1085 slice_108" [id=1085, type=slice]; +"1086 slice_109" [id=1086, type=slice]; +"1087 contiguous_11" [id=1087, type=contiguous]; +"1088 layer_norm_15" [id=1088, type=layer_norm]; +"1089 add_22" [id=1089, type=add]; +"1090 quantize_per_tensor_default_43" [id=1090, type=quantize_per_tensor]; +"1091 dequantize_per_tensor_default_43" [id=1091, type=dequantize_per_tensor]; +"1092 linear_42" [id=1092, type=linear]; +"1093 gelu_6" [id=1093, type=gelu]; +"1094 quantize_per_tensor_default_44" [id=1094, type=quantize_per_tensor]; +"1095 dequantize_per_tensor_default_44" [id=1095, type=dequantize_per_tensor]; +"1096 dropout_26" [id=1096, type=dropout]; +"1097 linear_43" [id=1097, type=linear]; +"1098 dropout_27" [id=1098, type=dropout]; +"1099 layer_norm_16" [id=1099, type=layer_norm]; +"1100 add_23" [id=1100, type=add]; +"1101 _frozen_param55" [id=1101, type=get_attr]; +"1102 pad_9" [id=1102, type=pad]; +"1103 roll_6" [id=1103, type=roll]; +"1104 view_39" [id=1104, type=view]; +"1105 permute_33" [id=1105, type=permute]; +"1106 reshape_31" [id=1106, type=reshape]; +"1107 quantize_per_tensor_default_45" [id=1107, type=quantize_per_tensor]; +"1108 dequantize_per_tensor_default_45" [id=1108, type=dequantize_per_tensor]; +"1109 _frozen_param56" [id=1109, type=get_attr]; +"1110 linear_46" [id=1110, type=linear]; +"1111 reshape_32" [id=1111, type=reshape]; +"1112 permute_34" [id=1112, type=permute]; +"1113 select_21" [id=1113, type=select]; +"1114 select_22" [id=1114, type=select]; +"1115 select_23" [id=1115, type=select]; +"1116 linalg_vector_norm_14" [id=1116, type=linalg_vector_norm]; +"1117 clamp_min_14" [id=1117, type=clamp_min]; +"1118 expand_as_14" [id=1118, type=expand_as]; +"1119 div_14" [id=1119, type=div]; +"1120 quantize_per_tensor_default_46" [id=1120, type=quantize_per_tensor]; +"1121 dequantize_per_tensor_default_46" [id=1121, type=dequantize_per_tensor]; +"1122 linalg_vector_norm_15" [id=1122, type=linalg_vector_norm]; +"1123 clamp_min_15" [id=1123, type=clamp_min]; +"1124 expand_as_15" [id=1124, type=expand_as]; +"1125 div_15" [id=1125, type=div]; +"1126 quantize_per_tensor_default_47" [id=1126, type=quantize_per_tensor]; +"1127 dequantize_per_tensor_default_47" [id=1127, type=dequantize_per_tensor]; +"1128 transpose_14" [id=1128, type=transpose]; +"1129 matmul_14" [id=1129, type=matmul]; +"1130 _frozen_param58" [id=1130, type=get_attr]; +"1131 mul_15" [id=1131, type=mul]; +"1132 add_24" [id=1132, type=add]; +"1133 new_zeros_3" [id=1133, type=new_zeros]; +"1134 view_40" [id=1134, type=view]; +"1135 permute_35" [id=1135, type=permute]; +"1136 reshape_33" [id=1136, type=reshape]; +"1137 unsqueeze_20" [id=1137, type=unsqueeze]; +"1138 unsqueeze_21" [id=1138, type=unsqueeze]; +"1139 sub_3" [id=1139, type=sub]; +"1140 ne_3" [id=1140, type=ne]; +"1141 masked_fill_6" [id=1141, type=masked_fill]; +"1142 eq_3" [id=1142, type=eq]; +"1143 masked_fill_7" [id=1143, type=masked_fill]; +"1144 view_41" [id=1144, type=view]; +"1145 unsqueeze_22" [id=1145, type=unsqueeze]; +"1146 unsqueeze_23" [id=1146, type=unsqueeze]; +"1147 add_25" [id=1147, type=add]; +"1148 view_42" [id=1148, type=view]; +"1149 softmax_7" [id=1149, type=softmax]; +"1150 dropout_28" [id=1150, type=dropout]; +"1151 matmul_15" [id=1151, type=matmul]; +"1152 quantize_per_tensor_default_48" [id=1152, type=quantize_per_tensor]; +"1153 dequantize_per_tensor_default_48" [id=1153, type=dequantize_per_tensor]; +"1154 transpose_15" [id=1154, type=transpose]; +"1155 reshape_34" [id=1155, type=reshape]; +"1156 linear_47" [id=1156, type=linear]; +"1157 dropout_29" [id=1157, type=dropout]; +"1158 view_43" [id=1158, type=view]; +"1159 permute_36" [id=1159, type=permute]; +"1160 reshape_35" [id=1160, type=reshape]; +"1161 roll_7" [id=1161, type=roll]; +"1162 slice_129" [id=1162, type=slice]; +"1163 slice_130" [id=1163, type=slice]; +"1164 slice_131" [id=1164, type=slice]; +"1165 slice_132" [id=1165, type=slice]; +"1166 contiguous_13" [id=1166, type=contiguous]; +"1167 layer_norm_17" [id=1167, type=layer_norm]; +"1168 add_26" [id=1168, type=add]; +"1169 quantize_per_tensor_default_49" [id=1169, type=quantize_per_tensor]; +"1170 dequantize_per_tensor_default_49" [id=1170, type=dequantize_per_tensor]; +"1171 linear_48" [id=1171, type=linear]; +"1172 gelu_7" [id=1172, type=gelu]; +"1173 quantize_per_tensor_default_50" [id=1173, type=quantize_per_tensor]; +"1174 dequantize_per_tensor_default_50" [id=1174, type=dequantize_per_tensor]; +"1175 dropout_30" [id=1175, type=dropout]; +"1176 linear_49" [id=1176, type=linear]; +"1177 dropout_31" [id=1177, type=dropout]; +"1178 layer_norm_18" [id=1178, type=layer_norm]; +"1179 add_27" [id=1179, type=add]; +"1180 quantize_per_tensor_default_51" [id=1180, type=quantize_per_tensor]; +"1181 dequantize_per_tensor_default_51" [id=1181, type=dequantize_per_tensor]; +"1182 _frozen_param68" [id=1182, type=get_attr]; +"1183 pad_10" [id=1183, type=pad]; +"1184 view_46" [id=1184, type=view]; +"1185 permute_38" [id=1185, type=permute]; +"1186 reshape_36" [id=1186, type=reshape]; +"1187 _frozen_param69" [id=1187, type=get_attr]; +"1188 linear_52" [id=1188, type=linear]; +"1189 reshape_37" [id=1189, type=reshape]; +"1190 permute_39" [id=1190, type=permute]; +"1191 select_24" [id=1191, type=select]; +"1192 select_25" [id=1192, type=select]; +"1193 select_26" [id=1193, type=select]; +"1194 linalg_vector_norm_16" [id=1194, type=linalg_vector_norm]; +"1195 clamp_min_16" [id=1195, type=clamp_min]; +"1196 expand_as_16" [id=1196, type=expand_as]; +"1197 div_16" [id=1197, type=div]; +"1198 quantize_per_tensor_default_52" [id=1198, type=quantize_per_tensor]; +"1199 dequantize_per_tensor_default_52" [id=1199, type=dequantize_per_tensor]; +"1200 linalg_vector_norm_17" [id=1200, type=linalg_vector_norm]; +"1201 clamp_min_17" [id=1201, type=clamp_min]; +"1202 expand_as_17" [id=1202, type=expand_as]; +"1203 div_17" [id=1203, type=div]; +"1204 quantize_per_tensor_default_53" [id=1204, type=quantize_per_tensor]; +"1205 dequantize_per_tensor_default_53" [id=1205, type=dequantize_per_tensor]; +"1206 transpose_16" [id=1206, type=transpose]; +"1207 matmul_16" [id=1207, type=matmul]; +"1208 _frozen_param71" [id=1208, type=get_attr]; +"1209 mul_17" [id=1209, type=mul]; +"1210 add_28" [id=1210, type=add]; +"1211 softmax_8" [id=1211, type=softmax]; +"1212 dropout_32" [id=1212, type=dropout]; +"1213 matmul_17" [id=1213, type=matmul]; +"1214 quantize_per_tensor_default_54" [id=1214, type=quantize_per_tensor]; +"1215 dequantize_per_tensor_default_54" [id=1215, type=dequantize_per_tensor]; +"1216 transpose_17" [id=1216, type=transpose]; +"1217 reshape_38" [id=1217, type=reshape]; +"1218 linear_53" [id=1218, type=linear]; +"1219 dropout_33" [id=1219, type=dropout]; +"1220 view_47" [id=1220, type=view]; +"1221 permute_40" [id=1221, type=permute]; +"1222 reshape_39" [id=1222, type=reshape]; +"1223 slice_134" [id=1223, type=slice]; +"1224 slice_135" [id=1224, type=slice]; +"1225 slice_136" [id=1225, type=slice]; +"1226 slice_137" [id=1226, type=slice]; +"1227 contiguous_15" [id=1227, type=contiguous]; +"1228 layer_norm_19" [id=1228, type=layer_norm]; +"1229 add_29" [id=1229, type=add]; +"1230 quantize_per_tensor_default_55" [id=1230, type=quantize_per_tensor]; +"1231 dequantize_per_tensor_default_55" [id=1231, type=dequantize_per_tensor]; +"1232 linear_54" [id=1232, type=linear]; +"1233 gelu_8" [id=1233, type=gelu]; +"1234 quantize_per_tensor_default_56" [id=1234, type=quantize_per_tensor]; +"1235 dequantize_per_tensor_default_56" [id=1235, type=dequantize_per_tensor]; +"1236 dropout_34" [id=1236, type=dropout]; +"1237 linear_55" [id=1237, type=linear]; +"1238 dropout_35" [id=1238, type=dropout]; +"1239 layer_norm_20" [id=1239, type=layer_norm]; +"1240 add_30" [id=1240, type=add]; +"1241 _frozen_param72" [id=1241, type=get_attr]; +"1242 pad_11" [id=1242, type=pad]; +"1243 roll_8" [id=1243, type=roll]; +"1244 view_50" [id=1244, type=view]; +"1245 permute_42" [id=1245, type=permute]; +"1246 reshape_40" [id=1246, type=reshape]; +"1247 quantize_per_tensor_default_57" [id=1247, type=quantize_per_tensor]; +"1248 dequantize_per_tensor_default_57" [id=1248, type=dequantize_per_tensor]; +"1249 _frozen_param73" [id=1249, type=get_attr]; +"1250 linear_58" [id=1250, type=linear]; +"1251 reshape_41" [id=1251, type=reshape]; +"1252 permute_43" [id=1252, type=permute]; +"1253 select_27" [id=1253, type=select]; +"1254 select_28" [id=1254, type=select]; +"1255 select_29" [id=1255, type=select]; +"1256 linalg_vector_norm_18" [id=1256, type=linalg_vector_norm]; +"1257 clamp_min_18" [id=1257, type=clamp_min]; +"1258 expand_as_18" [id=1258, type=expand_as]; +"1259 div_18" [id=1259, type=div]; +"1260 quantize_per_tensor_default_58" [id=1260, type=quantize_per_tensor]; +"1261 dequantize_per_tensor_default_58" [id=1261, type=dequantize_per_tensor]; +"1262 linalg_vector_norm_19" [id=1262, type=linalg_vector_norm]; +"1263 clamp_min_19" [id=1263, type=clamp_min]; +"1264 expand_as_19" [id=1264, type=expand_as]; +"1265 div_19" [id=1265, type=div]; +"1266 quantize_per_tensor_default_59" [id=1266, type=quantize_per_tensor]; +"1267 dequantize_per_tensor_default_59" [id=1267, type=dequantize_per_tensor]; +"1268 transpose_18" [id=1268, type=transpose]; +"1269 matmul_18" [id=1269, type=matmul]; +"1270 _frozen_param75" [id=1270, type=get_attr]; +"1271 mul_19" [id=1271, type=mul]; +"1272 add_31" [id=1272, type=add]; +"1273 new_zeros_4" [id=1273, type=new_zeros]; +"1274 view_51" [id=1274, type=view]; +"1275 permute_44" [id=1275, type=permute]; +"1276 reshape_42" [id=1276, type=reshape]; +"1277 unsqueeze_26" [id=1277, type=unsqueeze]; +"1278 unsqueeze_27" [id=1278, type=unsqueeze]; +"1279 sub_4" [id=1279, type=sub]; +"1280 ne_4" [id=1280, type=ne]; +"1281 masked_fill_8" [id=1281, type=masked_fill]; +"1282 eq_4" [id=1282, type=eq]; +"1283 masked_fill_9" [id=1283, type=masked_fill]; +"1284 view_52" [id=1284, type=view]; +"1285 unsqueeze_28" [id=1285, type=unsqueeze]; +"1286 unsqueeze_29" [id=1286, type=unsqueeze]; +"1287 add_32" [id=1287, type=add]; +"1288 view_53" [id=1288, type=view]; +"1289 softmax_9" [id=1289, type=softmax]; +"1290 dropout_36" [id=1290, type=dropout]; +"1291 matmul_19" [id=1291, type=matmul]; +"1292 quantize_per_tensor_default_60" [id=1292, type=quantize_per_tensor]; +"1293 dequantize_per_tensor_default_60" [id=1293, type=dequantize_per_tensor]; +"1294 transpose_19" [id=1294, type=transpose]; +"1295 reshape_43" [id=1295, type=reshape]; +"1296 linear_59" [id=1296, type=linear]; +"1297 dropout_37" [id=1297, type=dropout]; +"1298 view_54" [id=1298, type=view]; +"1299 permute_45" [id=1299, type=permute]; +"1300 reshape_44" [id=1300, type=reshape]; +"1301 roll_9" [id=1301, type=roll]; +"1302 slice_157" [id=1302, type=slice]; +"1303 slice_158" [id=1303, type=slice]; +"1304 slice_159" [id=1304, type=slice]; +"1305 slice_160" [id=1305, type=slice]; +"1306 contiguous_17" [id=1306, type=contiguous]; +"1307 layer_norm_21" [id=1307, type=layer_norm]; +"1308 add_33" [id=1308, type=add]; +"1309 quantize_per_tensor_default_61" [id=1309, type=quantize_per_tensor]; +"1310 dequantize_per_tensor_default_61" [id=1310, type=dequantize_per_tensor]; +"1311 linear_60" [id=1311, type=linear]; +"1312 gelu_9" [id=1312, type=gelu]; +"1313 quantize_per_tensor_default_62" [id=1313, type=quantize_per_tensor]; +"1314 dequantize_per_tensor_default_62" [id=1314, type=dequantize_per_tensor]; +"1315 dropout_38" [id=1315, type=dropout]; +"1316 linear_61" [id=1316, type=linear]; +"1317 dropout_39" [id=1317, type=dropout]; +"1318 layer_norm_22" [id=1318, type=layer_norm]; +"1319 add_34" [id=1319, type=add]; +"1320 quantize_per_tensor_default_63" [id=1320, type=quantize_per_tensor]; +"1321 dequantize_per_tensor_default_63" [id=1321, type=dequantize_per_tensor]; +"1322 _frozen_param85" [id=1322, type=get_attr]; +"1323 pad_12" [id=1323, type=pad]; +"1324 view_57" [id=1324, type=view]; +"1325 permute_47" [id=1325, type=permute]; +"1326 reshape_45" [id=1326, type=reshape]; +"1327 _frozen_param86" [id=1327, type=get_attr]; +"1328 linear_64" [id=1328, type=linear]; +"1329 reshape_46" [id=1329, type=reshape]; +"1330 permute_48" [id=1330, type=permute]; +"1331 select_30" [id=1331, type=select]; +"1332 select_31" [id=1332, type=select]; +"1333 select_32" [id=1333, type=select]; +"1334 linalg_vector_norm_20" [id=1334, type=linalg_vector_norm]; +"1335 clamp_min_20" [id=1335, type=clamp_min]; +"1336 expand_as_20" [id=1336, type=expand_as]; +"1337 div_20" [id=1337, type=div]; +"1338 quantize_per_tensor_default_64" [id=1338, type=quantize_per_tensor]; +"1339 dequantize_per_tensor_default_64" [id=1339, type=dequantize_per_tensor]; +"1340 linalg_vector_norm_21" [id=1340, type=linalg_vector_norm]; +"1341 clamp_min_21" [id=1341, type=clamp_min]; +"1342 expand_as_21" [id=1342, type=expand_as]; +"1343 div_21" [id=1343, type=div]; +"1344 quantize_per_tensor_default_65" [id=1344, type=quantize_per_tensor]; +"1345 dequantize_per_tensor_default_65" [id=1345, type=dequantize_per_tensor]; +"1346 transpose_20" [id=1346, type=transpose]; +"1347 matmul_20" [id=1347, type=matmul]; +"1348 _frozen_param88" [id=1348, type=get_attr]; +"1349 mul_21" [id=1349, type=mul]; +"1350 add_35" [id=1350, type=add]; +"1351 softmax_10" [id=1351, type=softmax]; +"1352 dropout_40" [id=1352, type=dropout]; +"1353 matmul_21" [id=1353, type=matmul]; +"1354 quantize_per_tensor_default_66" [id=1354, type=quantize_per_tensor]; +"1355 dequantize_per_tensor_default_66" [id=1355, type=dequantize_per_tensor]; +"1356 transpose_21" [id=1356, type=transpose]; +"1357 reshape_47" [id=1357, type=reshape]; +"1358 linear_65" [id=1358, type=linear]; +"1359 dropout_41" [id=1359, type=dropout]; +"1360 view_58" [id=1360, type=view]; +"1361 permute_49" [id=1361, type=permute]; +"1362 reshape_48" [id=1362, type=reshape]; +"1363 slice_162" [id=1363, type=slice]; +"1364 slice_163" [id=1364, type=slice]; +"1365 slice_164" [id=1365, type=slice]; +"1366 slice_165" [id=1366, type=slice]; +"1367 contiguous_19" [id=1367, type=contiguous]; +"1368 layer_norm_23" [id=1368, type=layer_norm]; +"1369 add_36" [id=1369, type=add]; +"1370 quantize_per_tensor_default_67" [id=1370, type=quantize_per_tensor]; +"1371 dequantize_per_tensor_default_67" [id=1371, type=dequantize_per_tensor]; +"1372 linear_66" [id=1372, type=linear]; +"1373 gelu_10" [id=1373, type=gelu]; +"1374 quantize_per_tensor_default_68" [id=1374, type=quantize_per_tensor]; +"1375 dequantize_per_tensor_default_68" [id=1375, type=dequantize_per_tensor]; +"1376 dropout_42" [id=1376, type=dropout]; +"1377 linear_67" [id=1377, type=linear]; +"1378 dropout_43" [id=1378, type=dropout]; +"1379 layer_norm_24" [id=1379, type=layer_norm]; +"1380 add_37" [id=1380, type=add]; +"1381 _frozen_param89" [id=1381, type=get_attr]; +"1382 pad_13" [id=1382, type=pad]; +"1383 roll_10" [id=1383, type=roll]; +"1384 view_61" [id=1384, type=view]; +"1385 permute_51" [id=1385, type=permute]; +"1386 reshape_49" [id=1386, type=reshape]; +"1387 quantize_per_tensor_default_69" [id=1387, type=quantize_per_tensor]; +"1388 dequantize_per_tensor_default_69" [id=1388, type=dequantize_per_tensor]; +"1389 _frozen_param90" [id=1389, type=get_attr]; +"1390 linear_70" [id=1390, type=linear]; +"1391 reshape_50" [id=1391, type=reshape]; +"1392 permute_52" [id=1392, type=permute]; +"1393 select_33" [id=1393, type=select]; +"1394 select_34" [id=1394, type=select]; +"1395 select_35" [id=1395, type=select]; +"1396 linalg_vector_norm_22" [id=1396, type=linalg_vector_norm]; +"1397 clamp_min_22" [id=1397, type=clamp_min]; +"1398 expand_as_22" [id=1398, type=expand_as]; +"1399 div_22" [id=1399, type=div]; +"1400 quantize_per_tensor_default_70" [id=1400, type=quantize_per_tensor]; +"1401 dequantize_per_tensor_default_70" [id=1401, type=dequantize_per_tensor]; +"1402 linalg_vector_norm_23" [id=1402, type=linalg_vector_norm]; +"1403 clamp_min_23" [id=1403, type=clamp_min]; +"1404 expand_as_23" [id=1404, type=expand_as]; +"1405 div_23" [id=1405, type=div]; +"1406 quantize_per_tensor_default_71" [id=1406, type=quantize_per_tensor]; +"1407 dequantize_per_tensor_default_71" [id=1407, type=dequantize_per_tensor]; +"1408 transpose_22" [id=1408, type=transpose]; +"1409 matmul_22" [id=1409, type=matmul]; +"1410 _frozen_param92" [id=1410, type=get_attr]; +"1411 mul_23" [id=1411, type=mul]; +"1412 add_38" [id=1412, type=add]; +"1413 new_zeros_5" [id=1413, type=new_zeros]; +"1414 view_62" [id=1414, type=view]; +"1415 permute_53" [id=1415, type=permute]; +"1416 reshape_51" [id=1416, type=reshape]; +"1417 unsqueeze_32" [id=1417, type=unsqueeze]; +"1418 unsqueeze_33" [id=1418, type=unsqueeze]; +"1419 sub_5" [id=1419, type=sub]; +"1420 ne_5" [id=1420, type=ne]; +"1421 masked_fill_10" [id=1421, type=masked_fill]; +"1422 eq_5" [id=1422, type=eq]; +"1423 masked_fill_11" [id=1423, type=masked_fill]; +"1424 view_63" [id=1424, type=view]; +"1425 unsqueeze_34" [id=1425, type=unsqueeze]; +"1426 unsqueeze_35" [id=1426, type=unsqueeze]; +"1427 add_39" [id=1427, type=add]; +"1428 view_64" [id=1428, type=view]; +"1429 softmax_11" [id=1429, type=softmax]; +"1430 dropout_44" [id=1430, type=dropout]; +"1431 matmul_23" [id=1431, type=matmul]; +"1432 quantize_per_tensor_default_72" [id=1432, type=quantize_per_tensor]; +"1433 dequantize_per_tensor_default_72" [id=1433, type=dequantize_per_tensor]; +"1434 transpose_23" [id=1434, type=transpose]; +"1435 reshape_52" [id=1435, type=reshape]; +"1436 linear_71" [id=1436, type=linear]; +"1437 dropout_45" [id=1437, type=dropout]; +"1438 view_65" [id=1438, type=view]; +"1439 permute_54" [id=1439, type=permute]; +"1440 reshape_53" [id=1440, type=reshape]; +"1441 roll_11" [id=1441, type=roll]; +"1442 slice_185" [id=1442, type=slice]; +"1443 slice_186" [id=1443, type=slice]; +"1444 slice_187" [id=1444, type=slice]; +"1445 slice_188" [id=1445, type=slice]; +"1446 contiguous_21" [id=1446, type=contiguous]; +"1447 layer_norm_25" [id=1447, type=layer_norm]; +"1448 add_40" [id=1448, type=add]; +"1449 quantize_per_tensor_default_73" [id=1449, type=quantize_per_tensor]; +"1450 dequantize_per_tensor_default_73" [id=1450, type=dequantize_per_tensor]; +"1451 linear_72" [id=1451, type=linear]; +"1452 gelu_11" [id=1452, type=gelu]; +"1453 quantize_per_tensor_default_74" [id=1453, type=quantize_per_tensor]; +"1454 dequantize_per_tensor_default_74" [id=1454, type=dequantize_per_tensor]; +"1455 dropout_46" [id=1455, type=dropout]; +"1456 linear_73" [id=1456, type=linear]; +"1457 dropout_47" [id=1457, type=dropout]; +"1458 layer_norm_26" [id=1458, type=layer_norm]; +"1459 add_41" [id=1459, type=add]; +"1460 quantize_per_tensor_default_75" [id=1460, type=quantize_per_tensor]; +"1461 dequantize_per_tensor_default_75" [id=1461, type=dequantize_per_tensor]; +"1462 _frozen_param102" [id=1462, type=get_attr]; +"1463 pad_14" [id=1463, type=pad]; +"1464 view_68" [id=1464, type=view]; +"1465 permute_56" [id=1465, type=permute]; +"1466 reshape_54" [id=1466, type=reshape]; +"1467 _frozen_param103" [id=1467, type=get_attr]; +"1468 linear_76" [id=1468, type=linear]; +"1469 reshape_55" [id=1469, type=reshape]; +"1470 permute_57" [id=1470, type=permute]; +"1471 select_36" [id=1471, type=select]; +"1472 select_37" [id=1472, type=select]; +"1473 select_38" [id=1473, type=select]; +"1474 linalg_vector_norm_24" [id=1474, type=linalg_vector_norm]; +"1475 clamp_min_24" [id=1475, type=clamp_min]; +"1476 expand_as_24" [id=1476, type=expand_as]; +"1477 div_24" [id=1477, type=div]; +"1478 quantize_per_tensor_default_76" [id=1478, type=quantize_per_tensor]; +"1479 dequantize_per_tensor_default_76" [id=1479, type=dequantize_per_tensor]; +"1480 linalg_vector_norm_25" [id=1480, type=linalg_vector_norm]; +"1481 clamp_min_25" [id=1481, type=clamp_min]; +"1482 expand_as_25" [id=1482, type=expand_as]; +"1483 div_25" [id=1483, type=div]; +"1484 quantize_per_tensor_default_77" [id=1484, type=quantize_per_tensor]; +"1485 dequantize_per_tensor_default_77" [id=1485, type=dequantize_per_tensor]; +"1486 transpose_24" [id=1486, type=transpose]; +"1487 matmul_24" [id=1487, type=matmul]; +"1488 _frozen_param105" [id=1488, type=get_attr]; +"1489 mul_25" [id=1489, type=mul]; +"1490 add_42" [id=1490, type=add]; +"1491 softmax_12" [id=1491, type=softmax]; +"1492 dropout_48" [id=1492, type=dropout]; +"1493 matmul_25" [id=1493, type=matmul]; +"1494 quantize_per_tensor_default_78" [id=1494, type=quantize_per_tensor]; +"1495 dequantize_per_tensor_default_78" [id=1495, type=dequantize_per_tensor]; +"1496 transpose_25" [id=1496, type=transpose]; +"1497 reshape_56" [id=1497, type=reshape]; +"1498 linear_77" [id=1498, type=linear]; +"1499 dropout_49" [id=1499, type=dropout]; +"1500 view_69" [id=1500, type=view]; +"1501 permute_58" [id=1501, type=permute]; +"1502 reshape_57" [id=1502, type=reshape]; +"1503 slice_190" [id=1503, type=slice]; +"1504 slice_191" [id=1504, type=slice]; +"1505 slice_192" [id=1505, type=slice]; +"1506 slice_193" [id=1506, type=slice]; +"1507 contiguous_23" [id=1507, type=contiguous]; +"1508 layer_norm_27" [id=1508, type=layer_norm]; +"1509 add_43" [id=1509, type=add]; +"1510 quantize_per_tensor_default_79" [id=1510, type=quantize_per_tensor]; +"1511 dequantize_per_tensor_default_79" [id=1511, type=dequantize_per_tensor]; +"1512 linear_78" [id=1512, type=linear]; +"1513 gelu_12" [id=1513, type=gelu]; +"1514 quantize_per_tensor_default_80" [id=1514, type=quantize_per_tensor]; +"1515 dequantize_per_tensor_default_80" [id=1515, type=dequantize_per_tensor]; +"1516 dropout_50" [id=1516, type=dropout]; +"1517 linear_79" [id=1517, type=linear]; +"1518 dropout_51" [id=1518, type=dropout]; +"1519 layer_norm_28" [id=1519, type=layer_norm]; +"1520 add_44" [id=1520, type=add]; +"1521 _frozen_param106" [id=1521, type=get_attr]; +"1522 pad_15" [id=1522, type=pad]; +"1523 roll_12" [id=1523, type=roll]; +"1524 view_72" [id=1524, type=view]; +"1525 permute_60" [id=1525, type=permute]; +"1526 reshape_58" [id=1526, type=reshape]; +"1527 quantize_per_tensor_default_81" [id=1527, type=quantize_per_tensor]; +"1528 dequantize_per_tensor_default_81" [id=1528, type=dequantize_per_tensor]; +"1529 _frozen_param107" [id=1529, type=get_attr]; +"1530 linear_82" [id=1530, type=linear]; +"1531 reshape_59" [id=1531, type=reshape]; +"1532 permute_61" [id=1532, type=permute]; +"1533 select_39" [id=1533, type=select]; +"1534 select_40" [id=1534, type=select]; +"1535 select_41" [id=1535, type=select]; +"1536 linalg_vector_norm_26" [id=1536, type=linalg_vector_norm]; +"1537 clamp_min_26" [id=1537, type=clamp_min]; +"1538 expand_as_26" [id=1538, type=expand_as]; +"1539 div_26" [id=1539, type=div]; +"1540 quantize_per_tensor_default_82" [id=1540, type=quantize_per_tensor]; +"1541 dequantize_per_tensor_default_82" [id=1541, type=dequantize_per_tensor]; +"1542 linalg_vector_norm_27" [id=1542, type=linalg_vector_norm]; +"1543 clamp_min_27" [id=1543, type=clamp_min]; +"1544 expand_as_27" [id=1544, type=expand_as]; +"1545 div_27" [id=1545, type=div]; +"1546 quantize_per_tensor_default_83" [id=1546, type=quantize_per_tensor]; +"1547 dequantize_per_tensor_default_83" [id=1547, type=dequantize_per_tensor]; +"1548 transpose_26" [id=1548, type=transpose]; +"1549 matmul_26" [id=1549, type=matmul]; +"1550 _frozen_param109" [id=1550, type=get_attr]; +"1551 mul_27" [id=1551, type=mul]; +"1552 add_45" [id=1552, type=add]; +"1553 new_zeros_6" [id=1553, type=new_zeros]; +"1554 view_73" [id=1554, type=view]; +"1555 permute_62" [id=1555, type=permute]; +"1556 reshape_60" [id=1556, type=reshape]; +"1557 unsqueeze_38" [id=1557, type=unsqueeze]; +"1558 unsqueeze_39" [id=1558, type=unsqueeze]; +"1559 sub_6" [id=1559, type=sub]; +"1560 ne_6" [id=1560, type=ne]; +"1561 masked_fill_12" [id=1561, type=masked_fill]; +"1562 eq_6" [id=1562, type=eq]; +"1563 masked_fill_13" [id=1563, type=masked_fill]; +"1564 view_74" [id=1564, type=view]; +"1565 unsqueeze_40" [id=1565, type=unsqueeze]; +"1566 unsqueeze_41" [id=1566, type=unsqueeze]; +"1567 add_46" [id=1567, type=add]; +"1568 view_75" [id=1568, type=view]; +"1569 softmax_13" [id=1569, type=softmax]; +"1570 dropout_52" [id=1570, type=dropout]; +"1571 matmul_27" [id=1571, type=matmul]; +"1572 quantize_per_tensor_default_84" [id=1572, type=quantize_per_tensor]; +"1573 dequantize_per_tensor_default_84" [id=1573, type=dequantize_per_tensor]; +"1574 transpose_27" [id=1574, type=transpose]; +"1575 reshape_61" [id=1575, type=reshape]; +"1576 linear_83" [id=1576, type=linear]; +"1577 dropout_53" [id=1577, type=dropout]; +"1578 view_76" [id=1578, type=view]; +"1579 permute_63" [id=1579, type=permute]; +"1580 reshape_62" [id=1580, type=reshape]; +"1581 roll_13" [id=1581, type=roll]; +"1582 slice_213" [id=1582, type=slice]; +"1583 slice_214" [id=1583, type=slice]; +"1584 slice_215" [id=1584, type=slice]; +"1585 slice_216" [id=1585, type=slice]; +"1586 contiguous_25" [id=1586, type=contiguous]; +"1587 layer_norm_29" [id=1587, type=layer_norm]; +"1588 add_47" [id=1588, type=add]; +"1589 quantize_per_tensor_default_85" [id=1589, type=quantize_per_tensor]; +"1590 dequantize_per_tensor_default_85" [id=1590, type=dequantize_per_tensor]; +"1591 linear_84" [id=1591, type=linear]; +"1592 gelu_13" [id=1592, type=gelu]; +"1593 quantize_per_tensor_default_86" [id=1593, type=quantize_per_tensor]; +"1594 dequantize_per_tensor_default_86" [id=1594, type=dequantize_per_tensor]; +"1595 dropout_54" [id=1595, type=dropout]; +"1596 linear_85" [id=1596, type=linear]; +"1597 dropout_55" [id=1597, type=dropout]; +"1598 layer_norm_30" [id=1598, type=layer_norm]; +"1599 add_48" [id=1599, type=add]; +"1600 quantize_per_tensor_default_87" [id=1600, type=quantize_per_tensor]; +"1601 dequantize_per_tensor_default_87" [id=1601, type=dequantize_per_tensor]; +"1602 _frozen_param119" [id=1602, type=get_attr]; +"1603 pad_16" [id=1603, type=pad]; +"1604 view_79" [id=1604, type=view]; +"1605 permute_65" [id=1605, type=permute]; +"1606 reshape_63" [id=1606, type=reshape]; +"1607 _frozen_param120" [id=1607, type=get_attr]; +"1608 linear_88" [id=1608, type=linear]; +"1609 reshape_64" [id=1609, type=reshape]; +"1610 permute_66" [id=1610, type=permute]; +"1611 select_42" [id=1611, type=select]; +"1612 select_43" [id=1612, type=select]; +"1613 select_44" [id=1613, type=select]; +"1614 linalg_vector_norm_28" [id=1614, type=linalg_vector_norm]; +"1615 clamp_min_28" [id=1615, type=clamp_min]; +"1616 expand_as_28" [id=1616, type=expand_as]; +"1617 div_28" [id=1617, type=div]; +"1618 quantize_per_tensor_default_88" [id=1618, type=quantize_per_tensor]; +"1619 dequantize_per_tensor_default_88" [id=1619, type=dequantize_per_tensor]; +"1620 linalg_vector_norm_29" [id=1620, type=linalg_vector_norm]; +"1621 clamp_min_29" [id=1621, type=clamp_min]; +"1622 expand_as_29" [id=1622, type=expand_as]; +"1623 div_29" [id=1623, type=div]; +"1624 quantize_per_tensor_default_89" [id=1624, type=quantize_per_tensor]; +"1625 dequantize_per_tensor_default_89" [id=1625, type=dequantize_per_tensor]; +"1626 transpose_28" [id=1626, type=transpose]; +"1627 matmul_28" [id=1627, type=matmul]; +"1628 _frozen_param122" [id=1628, type=get_attr]; +"1629 mul_29" [id=1629, type=mul]; +"1630 add_49" [id=1630, type=add]; +"1631 softmax_14" [id=1631, type=softmax]; +"1632 dropout_56" [id=1632, type=dropout]; +"1633 matmul_29" [id=1633, type=matmul]; +"1634 quantize_per_tensor_default_90" [id=1634, type=quantize_per_tensor]; +"1635 dequantize_per_tensor_default_90" [id=1635, type=dequantize_per_tensor]; +"1636 transpose_29" [id=1636, type=transpose]; +"1637 reshape_65" [id=1637, type=reshape]; +"1638 linear_89" [id=1638, type=linear]; +"1639 dropout_57" [id=1639, type=dropout]; +"1640 view_80" [id=1640, type=view]; +"1641 permute_67" [id=1641, type=permute]; +"1642 reshape_66" [id=1642, type=reshape]; +"1643 slice_218" [id=1643, type=slice]; +"1644 slice_219" [id=1644, type=slice]; +"1645 slice_220" [id=1645, type=slice]; +"1646 slice_221" [id=1646, type=slice]; +"1647 contiguous_27" [id=1647, type=contiguous]; +"1648 layer_norm_31" [id=1648, type=layer_norm]; +"1649 add_50" [id=1649, type=add]; +"1650 quantize_per_tensor_default_91" [id=1650, type=quantize_per_tensor]; +"1651 dequantize_per_tensor_default_91" [id=1651, type=dequantize_per_tensor]; +"1652 linear_90" [id=1652, type=linear]; +"1653 gelu_14" [id=1653, type=gelu]; +"1654 quantize_per_tensor_default_92" [id=1654, type=quantize_per_tensor]; +"1655 dequantize_per_tensor_default_92" [id=1655, type=dequantize_per_tensor]; +"1656 dropout_58" [id=1656, type=dropout]; +"1657 linear_91" [id=1657, type=linear]; +"1658 dropout_59" [id=1658, type=dropout]; +"1659 layer_norm_32" [id=1659, type=layer_norm]; +"1660 add_51" [id=1660, type=add]; +"1661 _frozen_param123" [id=1661, type=get_attr]; +"1662 pad_17" [id=1662, type=pad]; +"1663 roll_14" [id=1663, type=roll]; +"1664 view_83" [id=1664, type=view]; +"1665 permute_69" [id=1665, type=permute]; +"1666 reshape_67" [id=1666, type=reshape]; +"1667 quantize_per_tensor_default_93" [id=1667, type=quantize_per_tensor]; +"1668 dequantize_per_tensor_default_93" [id=1668, type=dequantize_per_tensor]; +"1669 _frozen_param124" [id=1669, type=get_attr]; +"1670 linear_94" [id=1670, type=linear]; +"1671 reshape_68" [id=1671, type=reshape]; +"1672 permute_70" [id=1672, type=permute]; +"1673 select_45" [id=1673, type=select]; +"1674 select_46" [id=1674, type=select]; +"1675 select_47" [id=1675, type=select]; +"1676 linalg_vector_norm_30" [id=1676, type=linalg_vector_norm]; +"1677 clamp_min_30" [id=1677, type=clamp_min]; +"1678 expand_as_30" [id=1678, type=expand_as]; +"1679 div_30" [id=1679, type=div]; +"1680 quantize_per_tensor_default_94" [id=1680, type=quantize_per_tensor]; +"1681 dequantize_per_tensor_default_94" [id=1681, type=dequantize_per_tensor]; +"1682 linalg_vector_norm_31" [id=1682, type=linalg_vector_norm]; +"1683 clamp_min_31" [id=1683, type=clamp_min]; +"1684 expand_as_31" [id=1684, type=expand_as]; +"1685 div_31" [id=1685, type=div]; +"1686 quantize_per_tensor_default_95" [id=1686, type=quantize_per_tensor]; +"1687 dequantize_per_tensor_default_95" [id=1687, type=dequantize_per_tensor]; +"1688 transpose_30" [id=1688, type=transpose]; +"1689 matmul_30" [id=1689, type=matmul]; +"1690 _frozen_param126" [id=1690, type=get_attr]; +"1691 mul_31" [id=1691, type=mul]; +"1692 add_52" [id=1692, type=add]; +"1693 new_zeros_7" [id=1693, type=new_zeros]; +"1694 view_84" [id=1694, type=view]; +"1695 permute_71" [id=1695, type=permute]; +"1696 reshape_69" [id=1696, type=reshape]; +"1697 unsqueeze_44" [id=1697, type=unsqueeze]; +"1698 unsqueeze_45" [id=1698, type=unsqueeze]; +"1699 sub_7" [id=1699, type=sub]; +"1700 ne_7" [id=1700, type=ne]; +"1701 masked_fill_14" [id=1701, type=masked_fill]; +"1702 eq_7" [id=1702, type=eq]; +"1703 masked_fill_15" [id=1703, type=masked_fill]; +"1704 view_85" [id=1704, type=view]; +"1705 unsqueeze_46" [id=1705, type=unsqueeze]; +"1706 unsqueeze_47" [id=1706, type=unsqueeze]; +"1707 add_53" [id=1707, type=add]; +"1708 view_86" [id=1708, type=view]; +"1709 softmax_15" [id=1709, type=softmax]; +"1710 dropout_60" [id=1710, type=dropout]; +"1711 matmul_31" [id=1711, type=matmul]; +"1712 quantize_per_tensor_default_96" [id=1712, type=quantize_per_tensor]; +"1713 dequantize_per_tensor_default_96" [id=1713, type=dequantize_per_tensor]; +"1714 transpose_31" [id=1714, type=transpose]; +"1715 reshape_70" [id=1715, type=reshape]; +"1716 linear_95" [id=1716, type=linear]; +"1717 dropout_61" [id=1717, type=dropout]; +"1718 view_87" [id=1718, type=view]; +"1719 permute_72" [id=1719, type=permute]; +"1720 reshape_71" [id=1720, type=reshape]; +"1721 roll_15" [id=1721, type=roll]; +"1722 slice_241" [id=1722, type=slice]; +"1723 slice_242" [id=1723, type=slice]; +"1724 slice_243" [id=1724, type=slice]; +"1725 slice_244" [id=1725, type=slice]; +"1726 contiguous_29" [id=1726, type=contiguous]; +"1727 layer_norm_33" [id=1727, type=layer_norm]; +"1728 add_54" [id=1728, type=add]; +"1729 quantize_per_tensor_default_97" [id=1729, type=quantize_per_tensor]; +"1730 dequantize_per_tensor_default_97" [id=1730, type=dequantize_per_tensor]; +"1731 linear_96" [id=1731, type=linear]; +"1732 gelu_15" [id=1732, type=gelu]; +"1733 quantize_per_tensor_default_98" [id=1733, type=quantize_per_tensor]; +"1734 dequantize_per_tensor_default_98" [id=1734, type=dequantize_per_tensor]; +"1735 dropout_62" [id=1735, type=dropout]; +"1736 linear_97" [id=1736, type=linear]; +"1737 dropout_63" [id=1737, type=dropout]; +"1738 layer_norm_34" [id=1738, type=layer_norm]; +"1739 add_55" [id=1739, type=add]; +"1740 quantize_per_tensor_default_99" [id=1740, type=quantize_per_tensor]; +"1741 dequantize_per_tensor_default_99" [id=1741, type=dequantize_per_tensor]; +"1742 _frozen_param136" [id=1742, type=get_attr]; +"1743 pad_18" [id=1743, type=pad]; +"1744 view_90" [id=1744, type=view]; +"1745 permute_74" [id=1745, type=permute]; +"1746 reshape_72" [id=1746, type=reshape]; +"1747 _frozen_param137" [id=1747, type=get_attr]; +"1748 linear_100" [id=1748, type=linear]; +"1749 reshape_73" [id=1749, type=reshape]; +"1750 permute_75" [id=1750, type=permute]; +"1751 select_48" [id=1751, type=select]; +"1752 select_49" [id=1752, type=select]; +"1753 select_50" [id=1753, type=select]; +"1754 linalg_vector_norm_32" [id=1754, type=linalg_vector_norm]; +"1755 clamp_min_32" [id=1755, type=clamp_min]; +"1756 expand_as_32" [id=1756, type=expand_as]; +"1757 div_32" [id=1757, type=div]; +"1758 quantize_per_tensor_default_100" [id=1758, type=quantize_per_tensor]; +"1759 dequantize_per_tensor_default_100" [id=1759, type=dequantize_per_tensor]; +"1760 linalg_vector_norm_33" [id=1760, type=linalg_vector_norm]; +"1761 clamp_min_33" [id=1761, type=clamp_min]; +"1762 expand_as_33" [id=1762, type=expand_as]; +"1763 div_33" [id=1763, type=div]; +"1764 quantize_per_tensor_default_101" [id=1764, type=quantize_per_tensor]; +"1765 dequantize_per_tensor_default_101" [id=1765, type=dequantize_per_tensor]; +"1766 transpose_32" [id=1766, type=transpose]; +"1767 matmul_32" [id=1767, type=matmul]; +"1768 _frozen_param139" [id=1768, type=get_attr]; +"1769 mul_33" [id=1769, type=mul]; +"1770 add_56" [id=1770, type=add]; +"1771 softmax_16" [id=1771, type=softmax]; +"1772 dropout_64" [id=1772, type=dropout]; +"1773 matmul_33" [id=1773, type=matmul]; +"1774 quantize_per_tensor_default_102" [id=1774, type=quantize_per_tensor]; +"1775 dequantize_per_tensor_default_102" [id=1775, type=dequantize_per_tensor]; +"1776 transpose_33" [id=1776, type=transpose]; +"1777 reshape_74" [id=1777, type=reshape]; +"1778 linear_101" [id=1778, type=linear]; +"1779 dropout_65" [id=1779, type=dropout]; +"1780 view_91" [id=1780, type=view]; +"1781 permute_76" [id=1781, type=permute]; +"1782 reshape_75" [id=1782, type=reshape]; +"1783 slice_246" [id=1783, type=slice]; +"1784 slice_247" [id=1784, type=slice]; +"1785 slice_248" [id=1785, type=slice]; +"1786 slice_249" [id=1786, type=slice]; +"1787 contiguous_31" [id=1787, type=contiguous]; +"1788 layer_norm_35" [id=1788, type=layer_norm]; +"1789 add_57" [id=1789, type=add]; +"1790 quantize_per_tensor_default_103" [id=1790, type=quantize_per_tensor]; +"1791 dequantize_per_tensor_default_103" [id=1791, type=dequantize_per_tensor]; +"1792 linear_102" [id=1792, type=linear]; +"1793 gelu_16" [id=1793, type=gelu]; +"1794 quantize_per_tensor_default_104" [id=1794, type=quantize_per_tensor]; +"1795 dequantize_per_tensor_default_104" [id=1795, type=dequantize_per_tensor]; +"1796 dropout_66" [id=1796, type=dropout]; +"1797 linear_103" [id=1797, type=linear]; +"1798 dropout_67" [id=1798, type=dropout]; +"1799 layer_norm_36" [id=1799, type=layer_norm]; +"1800 add_58" [id=1800, type=add]; +"1801 _frozen_param140" [id=1801, type=get_attr]; +"1802 pad_19" [id=1802, type=pad]; +"1803 roll_16" [id=1803, type=roll]; +"1804 view_94" [id=1804, type=view]; +"1805 permute_78" [id=1805, type=permute]; +"1806 reshape_76" [id=1806, type=reshape]; +"1807 quantize_per_tensor_default_105" [id=1807, type=quantize_per_tensor]; +"1808 dequantize_per_tensor_default_105" [id=1808, type=dequantize_per_tensor]; +"1809 _frozen_param141" [id=1809, type=get_attr]; +"1810 linear_106" [id=1810, type=linear]; +"1811 reshape_77" [id=1811, type=reshape]; +"1812 permute_79" [id=1812, type=permute]; +"1813 select_51" [id=1813, type=select]; +"1814 select_52" [id=1814, type=select]; +"1815 select_53" [id=1815, type=select]; +"1816 linalg_vector_norm_34" [id=1816, type=linalg_vector_norm]; +"1817 clamp_min_34" [id=1817, type=clamp_min]; +"1818 expand_as_34" [id=1818, type=expand_as]; +"1819 div_34" [id=1819, type=div]; +"1820 quantize_per_tensor_default_106" [id=1820, type=quantize_per_tensor]; +"1821 dequantize_per_tensor_default_106" [id=1821, type=dequantize_per_tensor]; +"1822 linalg_vector_norm_35" [id=1822, type=linalg_vector_norm]; +"1823 clamp_min_35" [id=1823, type=clamp_min]; +"1824 expand_as_35" [id=1824, type=expand_as]; +"1825 div_35" [id=1825, type=div]; +"1826 quantize_per_tensor_default_107" [id=1826, type=quantize_per_tensor]; +"1827 dequantize_per_tensor_default_107" [id=1827, type=dequantize_per_tensor]; +"1828 transpose_34" [id=1828, type=transpose]; +"1829 matmul_34" [id=1829, type=matmul]; +"1830 _frozen_param143" [id=1830, type=get_attr]; +"1831 mul_35" [id=1831, type=mul]; +"1832 add_59" [id=1832, type=add]; +"1833 new_zeros_8" [id=1833, type=new_zeros]; +"1834 view_95" [id=1834, type=view]; +"1835 permute_80" [id=1835, type=permute]; +"1836 reshape_78" [id=1836, type=reshape]; +"1837 unsqueeze_50" [id=1837, type=unsqueeze]; +"1838 unsqueeze_51" [id=1838, type=unsqueeze]; +"1839 sub_8" [id=1839, type=sub]; +"1840 ne_8" [id=1840, type=ne]; +"1841 masked_fill_16" [id=1841, type=masked_fill]; +"1842 eq_8" [id=1842, type=eq]; +"1843 masked_fill_17" [id=1843, type=masked_fill]; +"1844 view_96" [id=1844, type=view]; +"1845 unsqueeze_52" [id=1845, type=unsqueeze]; +"1846 unsqueeze_53" [id=1846, type=unsqueeze]; +"1847 add_60" [id=1847, type=add]; +"1848 view_97" [id=1848, type=view]; +"1849 softmax_17" [id=1849, type=softmax]; +"1850 dropout_68" [id=1850, type=dropout]; +"1851 matmul_35" [id=1851, type=matmul]; +"1852 quantize_per_tensor_default_108" [id=1852, type=quantize_per_tensor]; +"1853 dequantize_per_tensor_default_108" [id=1853, type=dequantize_per_tensor]; +"1854 transpose_35" [id=1854, type=transpose]; +"1855 reshape_79" [id=1855, type=reshape]; +"1856 linear_107" [id=1856, type=linear]; +"1857 dropout_69" [id=1857, type=dropout]; +"1858 view_98" [id=1858, type=view]; +"1859 permute_81" [id=1859, type=permute]; +"1860 reshape_80" [id=1860, type=reshape]; +"1861 roll_17" [id=1861, type=roll]; +"1862 slice_269" [id=1862, type=slice]; +"1863 slice_270" [id=1863, type=slice]; +"1864 slice_271" [id=1864, type=slice]; +"1865 slice_272" [id=1865, type=slice]; +"1866 contiguous_33" [id=1866, type=contiguous]; +"1867 layer_norm_37" [id=1867, type=layer_norm]; +"1868 add_61" [id=1868, type=add]; +"1869 quantize_per_tensor_default_109" [id=1869, type=quantize_per_tensor]; +"1870 dequantize_per_tensor_default_109" [id=1870, type=dequantize_per_tensor]; +"1871 linear_108" [id=1871, type=linear]; +"1872 gelu_17" [id=1872, type=gelu]; +"1873 quantize_per_tensor_default_110" [id=1873, type=quantize_per_tensor]; +"1874 dequantize_per_tensor_default_110" [id=1874, type=dequantize_per_tensor]; +"1875 dropout_70" [id=1875, type=dropout]; +"1876 linear_109" [id=1876, type=linear]; +"1877 dropout_71" [id=1877, type=dropout]; +"1878 layer_norm_38" [id=1878, type=layer_norm]; +"1879 add_62" [id=1879, type=add]; +"1880 quantize_per_tensor_default_111" [id=1880, type=quantize_per_tensor]; +"1881 dequantize_per_tensor_default_111" [id=1881, type=dequantize_per_tensor]; +"1882 _frozen_param153" [id=1882, type=get_attr]; +"1883 pad_20" [id=1883, type=pad]; +"1884 view_101" [id=1884, type=view]; +"1885 permute_83" [id=1885, type=permute]; +"1886 reshape_81" [id=1886, type=reshape]; +"1887 _frozen_param154" [id=1887, type=get_attr]; +"1888 linear_112" [id=1888, type=linear]; +"1889 reshape_82" [id=1889, type=reshape]; +"1890 permute_84" [id=1890, type=permute]; +"1891 select_54" [id=1891, type=select]; +"1892 select_55" [id=1892, type=select]; +"1893 select_56" [id=1893, type=select]; +"1894 linalg_vector_norm_36" [id=1894, type=linalg_vector_norm]; +"1895 clamp_min_36" [id=1895, type=clamp_min]; +"1896 expand_as_36" [id=1896, type=expand_as]; +"1897 div_36" [id=1897, type=div]; +"1898 quantize_per_tensor_default_112" [id=1898, type=quantize_per_tensor]; +"1899 dequantize_per_tensor_default_112" [id=1899, type=dequantize_per_tensor]; +"1900 linalg_vector_norm_37" [id=1900, type=linalg_vector_norm]; +"1901 clamp_min_37" [id=1901, type=clamp_min]; +"1902 expand_as_37" [id=1902, type=expand_as]; +"1903 div_37" [id=1903, type=div]; +"1904 quantize_per_tensor_default_113" [id=1904, type=quantize_per_tensor]; +"1905 dequantize_per_tensor_default_113" [id=1905, type=dequantize_per_tensor]; +"1906 transpose_36" [id=1906, type=transpose]; +"1907 matmul_36" [id=1907, type=matmul]; +"1908 _frozen_param156" [id=1908, type=get_attr]; +"1909 mul_37" [id=1909, type=mul]; +"1910 add_63" [id=1910, type=add]; +"1911 softmax_18" [id=1911, type=softmax]; +"1912 dropout_72" [id=1912, type=dropout]; +"1913 matmul_37" [id=1913, type=matmul]; +"1914 quantize_per_tensor_default_114" [id=1914, type=quantize_per_tensor]; +"1915 dequantize_per_tensor_default_114" [id=1915, type=dequantize_per_tensor]; +"1916 transpose_37" [id=1916, type=transpose]; +"1917 reshape_83" [id=1917, type=reshape]; +"1918 linear_113" [id=1918, type=linear]; +"1919 dropout_73" [id=1919, type=dropout]; +"1920 view_102" [id=1920, type=view]; +"1921 permute_85" [id=1921, type=permute]; +"1922 reshape_84" [id=1922, type=reshape]; +"1923 slice_274" [id=1923, type=slice]; +"1924 slice_275" [id=1924, type=slice]; +"1925 slice_276" [id=1925, type=slice]; +"1926 slice_277" [id=1926, type=slice]; +"1927 contiguous_35" [id=1927, type=contiguous]; +"1928 layer_norm_39" [id=1928, type=layer_norm]; +"1929 add_64" [id=1929, type=add]; +"1930 quantize_per_tensor_default_115" [id=1930, type=quantize_per_tensor]; +"1931 dequantize_per_tensor_default_115" [id=1931, type=dequantize_per_tensor]; +"1932 linear_114" [id=1932, type=linear]; +"1933 gelu_18" [id=1933, type=gelu]; +"1934 quantize_per_tensor_default_116" [id=1934, type=quantize_per_tensor]; +"1935 dequantize_per_tensor_default_116" [id=1935, type=dequantize_per_tensor]; +"1936 dropout_74" [id=1936, type=dropout]; +"1937 linear_115" [id=1937, type=linear]; +"1938 dropout_75" [id=1938, type=dropout]; +"1939 layer_norm_40" [id=1939, type=layer_norm]; +"1940 add_65" [id=1940, type=add]; +"1941 _frozen_param157" [id=1941, type=get_attr]; +"1942 pad_21" [id=1942, type=pad]; +"1943 roll_18" [id=1943, type=roll]; +"1944 view_105" [id=1944, type=view]; +"1945 permute_87" [id=1945, type=permute]; +"1946 reshape_85" [id=1946, type=reshape]; +"1947 quantize_per_tensor_default_117" [id=1947, type=quantize_per_tensor]; +"1948 dequantize_per_tensor_default_117" [id=1948, type=dequantize_per_tensor]; +"1949 _frozen_param158" [id=1949, type=get_attr]; +"1950 linear_118" [id=1950, type=linear]; +"1951 reshape_86" [id=1951, type=reshape]; +"1952 permute_88" [id=1952, type=permute]; +"1953 select_57" [id=1953, type=select]; +"1954 select_58" [id=1954, type=select]; +"1955 select_59" [id=1955, type=select]; +"1956 linalg_vector_norm_38" [id=1956, type=linalg_vector_norm]; +"1957 clamp_min_38" [id=1957, type=clamp_min]; +"1958 expand_as_38" [id=1958, type=expand_as]; +"1959 div_38" [id=1959, type=div]; +"1960 quantize_per_tensor_default_118" [id=1960, type=quantize_per_tensor]; +"1961 dequantize_per_tensor_default_118" [id=1961, type=dequantize_per_tensor]; +"1962 linalg_vector_norm_39" [id=1962, type=linalg_vector_norm]; +"1963 clamp_min_39" [id=1963, type=clamp_min]; +"1964 expand_as_39" [id=1964, type=expand_as]; +"1965 div_39" [id=1965, type=div]; +"1966 quantize_per_tensor_default_119" [id=1966, type=quantize_per_tensor]; +"1967 dequantize_per_tensor_default_119" [id=1967, type=dequantize_per_tensor]; +"1968 transpose_38" [id=1968, type=transpose]; +"1969 matmul_38" [id=1969, type=matmul]; +"1970 _frozen_param160" [id=1970, type=get_attr]; +"1971 mul_39" [id=1971, type=mul]; +"1972 add_66" [id=1972, type=add]; +"1973 new_zeros_9" [id=1973, type=new_zeros]; +"1974 view_106" [id=1974, type=view]; +"1975 permute_89" [id=1975, type=permute]; +"1976 reshape_87" [id=1976, type=reshape]; +"1977 unsqueeze_56" [id=1977, type=unsqueeze]; +"1978 unsqueeze_57" [id=1978, type=unsqueeze]; +"1979 sub_9" [id=1979, type=sub]; +"1980 ne_9" [id=1980, type=ne]; +"1981 masked_fill_18" [id=1981, type=masked_fill]; +"1982 eq_9" [id=1982, type=eq]; +"1983 masked_fill_19" [id=1983, type=masked_fill]; +"1984 view_107" [id=1984, type=view]; +"1985 unsqueeze_58" [id=1985, type=unsqueeze]; +"1986 unsqueeze_59" [id=1986, type=unsqueeze]; +"1987 add_67" [id=1987, type=add]; +"1988 view_108" [id=1988, type=view]; +"1989 softmax_19" [id=1989, type=softmax]; +"1990 dropout_76" [id=1990, type=dropout]; +"1991 matmul_39" [id=1991, type=matmul]; +"1992 quantize_per_tensor_default_120" [id=1992, type=quantize_per_tensor]; +"1993 dequantize_per_tensor_default_120" [id=1993, type=dequantize_per_tensor]; +"1994 transpose_39" [id=1994, type=transpose]; +"1995 reshape_88" [id=1995, type=reshape]; +"1996 linear_119" [id=1996, type=linear]; +"1997 dropout_77" [id=1997, type=dropout]; +"1998 view_109" [id=1998, type=view]; +"1999 permute_90" [id=1999, type=permute]; +"2000 reshape_89" [id=2000, type=reshape]; +"2001 roll_19" [id=2001, type=roll]; +"2002 slice_297" [id=2002, type=slice]; +"2003 slice_298" [id=2003, type=slice]; +"2004 slice_299" [id=2004, type=slice]; +"2005 slice_300" [id=2005, type=slice]; +"2006 contiguous_37" [id=2006, type=contiguous]; +"2007 layer_norm_41" [id=2007, type=layer_norm]; +"2008 add_68" [id=2008, type=add]; +"2009 quantize_per_tensor_default_121" [id=2009, type=quantize_per_tensor]; +"2010 dequantize_per_tensor_default_121" [id=2010, type=dequantize_per_tensor]; +"2011 linear_120" [id=2011, type=linear]; +"2012 gelu_19" [id=2012, type=gelu]; +"2013 quantize_per_tensor_default_122" [id=2013, type=quantize_per_tensor]; +"2014 dequantize_per_tensor_default_122" [id=2014, type=dequantize_per_tensor]; +"2015 dropout_78" [id=2015, type=dropout]; +"2016 linear_121" [id=2016, type=linear]; +"2017 dropout_79" [id=2017, type=dropout]; +"2018 layer_norm_42" [id=2018, type=layer_norm]; +"2019 add_69" [id=2019, type=add]; +"2020 quantize_per_tensor_default_123" [id=2020, type=quantize_per_tensor]; +"2021 dequantize_per_tensor_default_123" [id=2021, type=dequantize_per_tensor]; +"2022 _frozen_param170" [id=2022, type=get_attr]; +"2023 pad_22" [id=2023, type=pad]; +"2024 view_112" [id=2024, type=view]; +"2025 permute_92" [id=2025, type=permute]; +"2026 reshape_90" [id=2026, type=reshape]; +"2027 _frozen_param171" [id=2027, type=get_attr]; +"2028 linear_124" [id=2028, type=linear]; +"2029 reshape_91" [id=2029, type=reshape]; +"2030 permute_93" [id=2030, type=permute]; +"2031 select_60" [id=2031, type=select]; +"2032 select_61" [id=2032, type=select]; +"2033 select_62" [id=2033, type=select]; +"2034 linalg_vector_norm_40" [id=2034, type=linalg_vector_norm]; +"2035 clamp_min_40" [id=2035, type=clamp_min]; +"2036 expand_as_40" [id=2036, type=expand_as]; +"2037 div_40" [id=2037, type=div]; +"2038 quantize_per_tensor_default_124" [id=2038, type=quantize_per_tensor]; +"2039 dequantize_per_tensor_default_124" [id=2039, type=dequantize_per_tensor]; +"2040 linalg_vector_norm_41" [id=2040, type=linalg_vector_norm]; +"2041 clamp_min_41" [id=2041, type=clamp_min]; +"2042 expand_as_41" [id=2042, type=expand_as]; +"2043 div_41" [id=2043, type=div]; +"2044 quantize_per_tensor_default_125" [id=2044, type=quantize_per_tensor]; +"2045 dequantize_per_tensor_default_125" [id=2045, type=dequantize_per_tensor]; +"2046 transpose_40" [id=2046, type=transpose]; +"2047 matmul_40" [id=2047, type=matmul]; +"2048 _frozen_param173" [id=2048, type=get_attr]; +"2049 mul_41" [id=2049, type=mul]; +"2050 add_70" [id=2050, type=add]; +"2051 softmax_20" [id=2051, type=softmax]; +"2052 dropout_80" [id=2052, type=dropout]; +"2053 matmul_41" [id=2053, type=matmul]; +"2054 quantize_per_tensor_default_126" [id=2054, type=quantize_per_tensor]; +"2055 dequantize_per_tensor_default_126" [id=2055, type=dequantize_per_tensor]; +"2056 transpose_41" [id=2056, type=transpose]; +"2057 reshape_92" [id=2057, type=reshape]; +"2058 linear_125" [id=2058, type=linear]; +"2059 dropout_81" [id=2059, type=dropout]; +"2060 view_113" [id=2060, type=view]; +"2061 permute_94" [id=2061, type=permute]; +"2062 reshape_93" [id=2062, type=reshape]; +"2063 slice_302" [id=2063, type=slice]; +"2064 slice_303" [id=2064, type=slice]; +"2065 slice_304" [id=2065, type=slice]; +"2066 slice_305" [id=2066, type=slice]; +"2067 contiguous_39" [id=2067, type=contiguous]; +"2068 layer_norm_43" [id=2068, type=layer_norm]; +"2069 add_71" [id=2069, type=add]; +"2070 quantize_per_tensor_default_127" [id=2070, type=quantize_per_tensor]; +"2071 dequantize_per_tensor_default_127" [id=2071, type=dequantize_per_tensor]; +"2072 linear_126" [id=2072, type=linear]; +"2073 gelu_20" [id=2073, type=gelu]; +"2074 quantize_per_tensor_default_128" [id=2074, type=quantize_per_tensor]; +"2075 dequantize_per_tensor_default_128" [id=2075, type=dequantize_per_tensor]; +"2076 dropout_82" [id=2076, type=dropout]; +"2077 linear_127" [id=2077, type=linear]; +"2078 dropout_83" [id=2078, type=dropout]; +"2079 layer_norm_44" [id=2079, type=layer_norm]; +"2080 add_72" [id=2080, type=add]; +"2081 _frozen_param174" [id=2081, type=get_attr]; +"2082 pad_23" [id=2082, type=pad]; +"2083 roll_20" [id=2083, type=roll]; +"2084 view_116" [id=2084, type=view]; +"2085 permute_96" [id=2085, type=permute]; +"2086 reshape_94" [id=2086, type=reshape]; +"2087 quantize_per_tensor_default_129" [id=2087, type=quantize_per_tensor]; +"2088 dequantize_per_tensor_default_129" [id=2088, type=dequantize_per_tensor]; +"2089 _frozen_param175" [id=2089, type=get_attr]; +"2090 linear_130" [id=2090, type=linear]; +"2091 reshape_95" [id=2091, type=reshape]; +"2092 permute_97" [id=2092, type=permute]; +"2093 select_63" [id=2093, type=select]; +"2094 select_64" [id=2094, type=select]; +"2095 select_65" [id=2095, type=select]; +"2096 linalg_vector_norm_42" [id=2096, type=linalg_vector_norm]; +"2097 clamp_min_42" [id=2097, type=clamp_min]; +"2098 expand_as_42" [id=2098, type=expand_as]; +"2099 div_42" [id=2099, type=div]; +"2100 quantize_per_tensor_default_130" [id=2100, type=quantize_per_tensor]; +"2101 dequantize_per_tensor_default_130" [id=2101, type=dequantize_per_tensor]; +"2102 linalg_vector_norm_43" [id=2102, type=linalg_vector_norm]; +"2103 clamp_min_43" [id=2103, type=clamp_min]; +"2104 expand_as_43" [id=2104, type=expand_as]; +"2105 div_43" [id=2105, type=div]; +"2106 quantize_per_tensor_default_131" [id=2106, type=quantize_per_tensor]; +"2107 dequantize_per_tensor_default_131" [id=2107, type=dequantize_per_tensor]; +"2108 transpose_42" [id=2108, type=transpose]; +"2109 matmul_42" [id=2109, type=matmul]; +"2110 _frozen_param177" [id=2110, type=get_attr]; +"2111 mul_43" [id=2111, type=mul]; +"2112 add_73" [id=2112, type=add]; +"2113 new_zeros_10" [id=2113, type=new_zeros]; +"2114 view_117" [id=2114, type=view]; +"2115 permute_98" [id=2115, type=permute]; +"2116 reshape_96" [id=2116, type=reshape]; +"2117 unsqueeze_62" [id=2117, type=unsqueeze]; +"2118 unsqueeze_63" [id=2118, type=unsqueeze]; +"2119 sub_10" [id=2119, type=sub]; +"2120 ne_10" [id=2120, type=ne]; +"2121 masked_fill_20" [id=2121, type=masked_fill]; +"2122 eq_10" [id=2122, type=eq]; +"2123 masked_fill_21" [id=2123, type=masked_fill]; +"2124 view_118" [id=2124, type=view]; +"2125 unsqueeze_64" [id=2125, type=unsqueeze]; +"2126 unsqueeze_65" [id=2126, type=unsqueeze]; +"2127 add_74" [id=2127, type=add]; +"2128 view_119" [id=2128, type=view]; +"2129 softmax_21" [id=2129, type=softmax]; +"2130 dropout_84" [id=2130, type=dropout]; +"2131 matmul_43" [id=2131, type=matmul]; +"2132 quantize_per_tensor_default_132" [id=2132, type=quantize_per_tensor]; +"2133 dequantize_per_tensor_default_132" [id=2133, type=dequantize_per_tensor]; +"2134 transpose_43" [id=2134, type=transpose]; +"2135 reshape_97" [id=2135, type=reshape]; +"2136 linear_131" [id=2136, type=linear]; +"2137 dropout_85" [id=2137, type=dropout]; +"2138 view_120" [id=2138, type=view]; +"2139 permute_99" [id=2139, type=permute]; +"2140 reshape_98" [id=2140, type=reshape]; +"2141 roll_21" [id=2141, type=roll]; +"2142 slice_325" [id=2142, type=slice]; +"2143 slice_326" [id=2143, type=slice]; +"2144 slice_327" [id=2144, type=slice]; +"2145 slice_328" [id=2145, type=slice]; +"2146 contiguous_41" [id=2146, type=contiguous]; +"2147 layer_norm_45" [id=2147, type=layer_norm]; +"2148 add_75" [id=2148, type=add]; +"2149 quantize_per_tensor_default_133" [id=2149, type=quantize_per_tensor]; +"2150 dequantize_per_tensor_default_133" [id=2150, type=dequantize_per_tensor]; +"2151 linear_132" [id=2151, type=linear]; +"2152 gelu_21" [id=2152, type=gelu]; +"2153 quantize_per_tensor_default_134" [id=2153, type=quantize_per_tensor]; +"2154 dequantize_per_tensor_default_134" [id=2154, type=dequantize_per_tensor]; +"2155 dropout_86" [id=2155, type=dropout]; +"2156 linear_133" [id=2156, type=linear]; +"2157 dropout_87" [id=2157, type=dropout]; +"2158 layer_norm_46" [id=2158, type=layer_norm]; +"2159 add_76" [id=2159, type=add]; +"2160 quantize_per_tensor_default_135" [id=2160, type=quantize_per_tensor]; +"2161 dequantize_per_tensor_default_135" [id=2161, type=dequantize_per_tensor]; +"2162 pad_24" [id=2162, type=pad]; +"2163 slice_329" [id=2163, type=slice]; +"2164 slice_330" [id=2164, type=slice]; +"2165 slice_331" [id=2165, type=slice]; +"2166 slice_332" [id=2166, type=slice]; +"2167 slice_333" [id=2167, type=slice]; +"2168 slice_334" [id=2168, type=slice]; +"2169 slice_335" [id=2169, type=slice]; +"2170 slice_336" [id=2170, type=slice]; +"2171 slice_337" [id=2171, type=slice]; +"2172 slice_338" [id=2172, type=slice]; +"2173 slice_339" [id=2173, type=slice]; +"2174 slice_340" [id=2174, type=slice]; +"2175 cat_2" [id=2175, type=cat]; +"2176 linear_134" [id=2176, type=linear]; +"2177 layer_norm_47" [id=2177, type=layer_norm]; +"2178 quantize_per_tensor_default_136" [id=2178, type=quantize_per_tensor]; +"2179 dequantize_per_tensor_default_136" [id=2179, type=dequantize_per_tensor]; +"2180 _frozen_param187" [id=2180, type=get_attr]; +"2181 pad_25" [id=2181, type=pad]; +"2182 view_123" [id=2182, type=view]; +"2183 permute_101" [id=2183, type=permute]; +"2184 reshape_99" [id=2184, type=reshape]; +"2185 _frozen_param188" [id=2185, type=get_attr]; +"2186 linear_137" [id=2186, type=linear]; +"2187 reshape_100" [id=2187, type=reshape]; +"2188 permute_102" [id=2188, type=permute]; +"2189 select_66" [id=2189, type=select]; +"2190 select_67" [id=2190, type=select]; +"2191 select_68" [id=2191, type=select]; +"2192 linalg_vector_norm_44" [id=2192, type=linalg_vector_norm]; +"2193 clamp_min_44" [id=2193, type=clamp_min]; +"2194 expand_as_44" [id=2194, type=expand_as]; +"2195 div_44" [id=2195, type=div]; +"2196 quantize_per_tensor_default_137" [id=2196, type=quantize_per_tensor]; +"2197 dequantize_per_tensor_default_137" [id=2197, type=dequantize_per_tensor]; +"2198 linalg_vector_norm_45" [id=2198, type=linalg_vector_norm]; +"2199 clamp_min_45" [id=2199, type=clamp_min]; +"2200 expand_as_45" [id=2200, type=expand_as]; +"2201 div_45" [id=2201, type=div]; +"2202 quantize_per_tensor_default_138" [id=2202, type=quantize_per_tensor]; +"2203 dequantize_per_tensor_default_138" [id=2203, type=dequantize_per_tensor]; +"2204 transpose_44" [id=2204, type=transpose]; +"2205 matmul_44" [id=2205, type=matmul]; +"2206 _frozen_param190" [id=2206, type=get_attr]; +"2207 mul_45" [id=2207, type=mul]; +"2208 add_77" [id=2208, type=add]; +"2209 softmax_22" [id=2209, type=softmax]; +"2210 dropout_88" [id=2210, type=dropout]; +"2211 matmul_45" [id=2211, type=matmul]; +"2212 quantize_per_tensor_default_139" [id=2212, type=quantize_per_tensor]; +"2213 dequantize_per_tensor_default_139" [id=2213, type=dequantize_per_tensor]; +"2214 transpose_45" [id=2214, type=transpose]; +"2215 reshape_101" [id=2215, type=reshape]; +"2216 linear_138" [id=2216, type=linear]; +"2217 dropout_89" [id=2217, type=dropout]; +"2218 view_124" [id=2218, type=view]; +"2219 permute_103" [id=2219, type=permute]; +"2220 reshape_102" [id=2220, type=reshape]; +"2221 slice_342" [id=2221, type=slice]; +"2222 slice_343" [id=2222, type=slice]; +"2223 slice_344" [id=2223, type=slice]; +"2224 slice_345" [id=2224, type=slice]; +"2225 contiguous_43" [id=2225, type=contiguous]; +"2226 layer_norm_48" [id=2226, type=layer_norm]; +"2227 add_78" [id=2227, type=add]; +"2228 quantize_per_tensor_default_140" [id=2228, type=quantize_per_tensor]; +"2229 dequantize_per_tensor_default_140" [id=2229, type=dequantize_per_tensor]; +"2230 linear_139" [id=2230, type=linear]; +"2231 gelu_22" [id=2231, type=gelu]; +"2232 quantize_per_tensor_default_141" [id=2232, type=quantize_per_tensor]; +"2233 dequantize_per_tensor_default_141" [id=2233, type=dequantize_per_tensor]; +"2234 dropout_90" [id=2234, type=dropout]; +"2235 linear_140" [id=2235, type=linear]; +"2236 dropout_91" [id=2236, type=dropout]; +"2237 layer_norm_49" [id=2237, type=layer_norm]; +"2238 add_79" [id=2238, type=add]; +"2239 quantize_per_tensor_default_142" [id=2239, type=quantize_per_tensor]; +"2240 dequantize_per_tensor_default_142" [id=2240, type=dequantize_per_tensor]; +"2241 _frozen_param191" [id=2241, type=get_attr]; +"2242 pad_26" [id=2242, type=pad]; +"2243 view_127" [id=2243, type=view]; +"2244 permute_105" [id=2244, type=permute]; +"2245 reshape_103" [id=2245, type=reshape]; +"2246 _frozen_param192" [id=2246, type=get_attr]; +"2247 linear_143" [id=2247, type=linear]; +"2248 reshape_104" [id=2248, type=reshape]; +"2249 permute_106" [id=2249, type=permute]; +"2250 select_69" [id=2250, type=select]; +"2251 select_70" [id=2251, type=select]; +"2252 select_71" [id=2252, type=select]; +"2253 linalg_vector_norm_46" [id=2253, type=linalg_vector_norm]; +"2254 clamp_min_46" [id=2254, type=clamp_min]; +"2255 expand_as_46" [id=2255, type=expand_as]; +"2256 div_46" [id=2256, type=div]; +"2257 quantize_per_tensor_default_143" [id=2257, type=quantize_per_tensor]; +"2258 dequantize_per_tensor_default_143" [id=2258, type=dequantize_per_tensor]; +"2259 linalg_vector_norm_47" [id=2259, type=linalg_vector_norm]; +"2260 clamp_min_47" [id=2260, type=clamp_min]; +"2261 expand_as_47" [id=2261, type=expand_as]; +"2262 div_47" [id=2262, type=div]; +"2263 quantize_per_tensor_default_144" [id=2263, type=quantize_per_tensor]; +"2264 dequantize_per_tensor_default_144" [id=2264, type=dequantize_per_tensor]; +"2265 transpose_46" [id=2265, type=transpose]; +"2266 matmul_46" [id=2266, type=matmul]; +"2267 _frozen_param194" [id=2267, type=get_attr]; +"2268 mul_47" [id=2268, type=mul]; +"2269 add_80" [id=2269, type=add]; +"2270 softmax_23" [id=2270, type=softmax]; +"2271 dropout_92" [id=2271, type=dropout]; +"2272 matmul_47" [id=2272, type=matmul]; +"2273 quantize_per_tensor_default_145" [id=2273, type=quantize_per_tensor]; +"2274 dequantize_per_tensor_default_145" [id=2274, type=dequantize_per_tensor]; +"2275 transpose_47" [id=2275, type=transpose]; +"2276 reshape_105" [id=2276, type=reshape]; +"2277 linear_144" [id=2277, type=linear]; +"2278 dropout_93" [id=2278, type=dropout]; +"2279 view_128" [id=2279, type=view]; +"2280 permute_107" [id=2280, type=permute]; +"2281 reshape_106" [id=2281, type=reshape]; +"2282 slice_347" [id=2282, type=slice]; +"2283 slice_348" [id=2283, type=slice]; +"2284 slice_349" [id=2284, type=slice]; +"2285 slice_350" [id=2285, type=slice]; +"2286 contiguous_45" [id=2286, type=contiguous]; +"2287 layer_norm_50" [id=2287, type=layer_norm]; +"2288 add_81" [id=2288, type=add]; +"2289 quantize_per_tensor_default_146" [id=2289, type=quantize_per_tensor]; +"2290 dequantize_per_tensor_default_146" [id=2290, type=dequantize_per_tensor]; +"2291 linear_145" [id=2291, type=linear]; +"2292 gelu_23" [id=2292, type=gelu]; +"2293 quantize_per_tensor_default_147" [id=2293, type=quantize_per_tensor]; +"2294 dequantize_per_tensor_default_147" [id=2294, type=dequantize_per_tensor]; +"2295 dropout_94" [id=2295, type=dropout]; +"2296 linear_146" [id=2296, type=linear]; +"2297 dropout_95" [id=2297, type=dropout]; +"2298 layer_norm_51" [id=2298, type=layer_norm]; +"2299 add_82" [id=2299, type=add]; +"2300 layer_norm_52" [id=2300, type=layer_norm]; +"2301 permute_108" [id=2301, type=permute]; +"2302 adaptive_avg_pool2d" [id=2302, type=adaptive_avg_pool2d]; +"2303 quantize_per_tensor_default_148" [id=2303, type=quantize_per_tensor]; +"2304 dequantize_per_tensor_default_148" [id=2304, type=dequantize_per_tensor]; +"2305 flatten" [id=2305, type=flatten]; +"2306 linear_147" [id=2306, type=linear]; +"2307 output_1" [id=2307, type=output]; "0 _scale_0" -> "3 dequantize_per_channel_default" [label="(288,)", style=solid]; "1 _zero_point_0" -> "3 dequantize_per_channel_default" [label="(288,)", style=solid]; -"2 _frozen_param0" -> "3 dequantize_per_channel_default" [label="(288, 96)", style=solid]; -"3 dequantize_per_channel_default" -> "1021 linear_2" [label=None, style=solid]; +"2 _frozen_param2_1" -> "3 dequantize_per_channel_default" [label="(288, 96)", style=solid]; +"3 dequantize_per_channel_default" -> "598 linear_2" [label=None, style=solid]; "4 _scale_1" -> "7 dequantize_per_channel_default_1" [label="(96,)", style=solid]; "5 _zero_point_1" -> "7 dequantize_per_channel_default_1" [label="(96,)", style=solid]; -"6 _frozen_param1" -> "7 dequantize_per_channel_default_1" [label="(96, 96)", style=solid]; -"7 dequantize_per_channel_default_1" -> "1052 linear_3" [label=None, style=solid]; -"8 features_1_0_attn_qkv_bias" -> "1018 clone" [label="(288,)", style=solid]; -"9 features_1_0_attn_proj_bias" -> "1052 linear_3" [label="(96,)", style=solid]; -"10 features_1_0_attn_logit_scale" -> "1041 clamp" [label="(3, 1, 1)", style=solid]; -"11 _scale_2" -> "14 dequantize_per_channel_default_2" [label="(288,)", style=solid]; -"12 _zero_point_2" -> "14 dequantize_per_channel_default_2" [label="(288,)", style=solid]; -"13 _frozen_param2" -> "14 dequantize_per_channel_default_2" [label="(288, 96)", style=solid]; -"14 dequantize_per_channel_default_2" -> "1093 linear_8" [label=None, style=solid]; -"15 _scale_3" -> "18 dequantize_per_channel_default_3" [label="(96,)", style=solid]; -"16 _zero_point_3" -> "18 dequantize_per_channel_default_3" [label="(96,)", style=solid]; -"17 _frozen_param3" -> "18 dequantize_per_channel_default_3" [label="(96, 96)", style=solid]; -"18 dequantize_per_channel_default_3" -> "1176 linear_9" [label=None, style=solid]; -"19 features_1_1_attn_qkv_bias" -> "1090 clone_1" [label="(288,)", style=solid]; -"20 features_1_1_attn_proj_bias" -> "1176 linear_9" [label="(96,)", style=solid]; -"21 features_1_1_attn_logit_scale" -> "1113 clamp_1" [label="(3, 1, 1)", style=solid]; -"22 _scale_4" -> "25 dequantize_per_channel_default_4" [label="(576,)", style=solid]; -"23 _zero_point_4" -> "25 dequantize_per_channel_default_4" [label="(576,)", style=solid]; -"24 _frozen_param4" -> "25 dequantize_per_channel_default_4" [label="(576, 192)", style=solid]; -"25 dequantize_per_channel_default_4" -> "1235 linear_15" [label=None, style=solid]; -"26 _scale_5" -> "29 dequantize_per_channel_default_5" [label="(192,)", style=solid]; -"27 _zero_point_5" -> "29 dequantize_per_channel_default_5" [label="(192,)", style=solid]; -"28 _frozen_param5" -> "29 dequantize_per_channel_default_5" [label="(192, 192)", style=solid]; -"29 dequantize_per_channel_default_5" -> "1266 linear_16" [label=None, style=solid]; -"30 features_3_0_attn_qkv_bias" -> "1232 clone_2" [label="(576,)", style=solid]; -"31 features_3_0_attn_proj_bias" -> "1266 linear_16" [label="(192,)", style=solid]; -"32 features_3_0_attn_logit_scale" -> "1255 clamp_2" [label="(6, 1, 1)", style=solid]; -"33 _scale_6" -> "36 dequantize_per_channel_default_6" [label="(576,)", style=solid]; -"34 _zero_point_6" -> "36 dequantize_per_channel_default_6" [label="(576,)", style=solid]; -"35 _frozen_param6" -> "36 dequantize_per_channel_default_6" [label="(576, 192)", style=solid]; -"36 dequantize_per_channel_default_6" -> "1310 linear_21" [label=None, style=solid]; -"37 _scale_7" -> "40 dequantize_per_channel_default_7" [label="(192,)", style=solid]; -"38 _zero_point_7" -> "40 dequantize_per_channel_default_7" [label="(192,)", style=solid]; -"39 _frozen_param7" -> "40 dequantize_per_channel_default_7" [label="(192, 192)", style=solid]; -"40 dequantize_per_channel_default_7" -> "1393 linear_22" [label=None, style=solid]; -"41 features_3_1_attn_qkv_bias" -> "1307 clone_3" [label="(576,)", style=solid]; -"42 features_3_1_attn_proj_bias" -> "1393 linear_22" [label="(192,)", style=solid]; -"43 features_3_1_attn_logit_scale" -> "1330 clamp_3" [label="(6, 1, 1)", style=solid]; -"44 _scale_8" -> "47 dequantize_per_channel_default_8" [label="(1152,)", style=solid]; -"45 _zero_point_8" -> "47 dequantize_per_channel_default_8" [label="(1152,)", style=solid]; -"46 _frozen_param8" -> "47 dequantize_per_channel_default_8" [label="(1152, 384)", style=solid]; -"47 dequantize_per_channel_default_8" -> "1455 linear_28" [label=None, style=solid]; -"48 _scale_9" -> "51 dequantize_per_channel_default_9" [label="(384,)", style=solid]; -"49 _zero_point_9" -> "51 dequantize_per_channel_default_9" [label="(384,)", style=solid]; -"50 _frozen_param9" -> "51 dequantize_per_channel_default_9" [label="(384, 384)", style=solid]; -"51 dequantize_per_channel_default_9" -> "1486 linear_29" [label=None, style=solid]; -"52 features_5_0_attn_qkv_bias" -> "1452 clone_4" [label="(1152,)", style=solid]; -"53 features_5_0_attn_proj_bias" -> "1486 linear_29" [label="(384,)", style=solid]; -"54 features_5_0_attn_logit_scale" -> "1475 clamp_4" [label="(12, 1, 1)", style=solid]; -"55 _scale_10" -> "58 dequantize_per_channel_default_10" [label="(1152,)", style=solid]; -"56 _zero_point_10" -> "58 dequantize_per_channel_default_10" [label="(1152,)", style=solid]; -"57 _frozen_param10" -> "58 dequantize_per_channel_default_10" [label="(1152, 384)", style=solid]; -"58 dequantize_per_channel_default_10" -> "1530 linear_34" [label=None, style=solid]; -"59 _scale_11" -> "62 dequantize_per_channel_default_11" [label="(384,)", style=solid]; -"60 _zero_point_11" -> "62 dequantize_per_channel_default_11" [label="(384,)", style=solid]; -"61 _frozen_param11" -> "62 dequantize_per_channel_default_11" [label="(384, 384)", style=solid]; -"62 dequantize_per_channel_default_11" -> "1613 linear_35" [label=None, style=solid]; -"63 features_5_1_attn_qkv_bias" -> "1527 clone_5" [label="(1152,)", style=solid]; -"64 features_5_1_attn_proj_bias" -> "1613 linear_35" [label="(384,)", style=solid]; -"65 features_5_1_attn_logit_scale" -> "1550 clamp_5" [label="(12, 1, 1)", style=solid]; -"66 _scale_12" -> "69 dequantize_per_channel_default_12" [label="(1152,)", style=solid]; -"67 _zero_point_12" -> "69 dequantize_per_channel_default_12" [label="(1152,)", style=solid]; -"68 _frozen_param12" -> "69 dequantize_per_channel_default_12" [label="(1152, 384)", style=solid]; -"69 dequantize_per_channel_default_12" -> "1657 linear_40" [label=None, style=solid]; -"70 _scale_13" -> "73 dequantize_per_channel_default_13" [label="(384,)", style=solid]; -"71 _zero_point_13" -> "73 dequantize_per_channel_default_13" [label="(384,)", style=solid]; -"72 _frozen_param13" -> "73 dequantize_per_channel_default_13" [label="(384, 384)", style=solid]; -"73 dequantize_per_channel_default_13" -> "1688 linear_41" [label=None, style=solid]; -"74 features_5_2_attn_qkv_bias" -> "1654 clone_6" [label="(1152,)", style=solid]; -"75 features_5_2_attn_proj_bias" -> "1688 linear_41" [label="(384,)", style=solid]; -"76 features_5_2_attn_logit_scale" -> "1677 clamp_6" [label="(12, 1, 1)", style=solid]; -"77 _scale_14" -> "80 dequantize_per_channel_default_14" [label="(1152,)", style=solid]; -"78 _zero_point_14" -> "80 dequantize_per_channel_default_14" [label="(1152,)", style=solid]; -"79 _frozen_param14" -> "80 dequantize_per_channel_default_14" [label="(1152, 384)", style=solid]; -"80 dequantize_per_channel_default_14" -> "1732 linear_46" [label=None, style=solid]; -"81 _scale_15" -> "84 dequantize_per_channel_default_15" [label="(384,)", style=solid]; -"82 _zero_point_15" -> "84 dequantize_per_channel_default_15" [label="(384,)", style=solid]; -"83 _frozen_param15" -> "84 dequantize_per_channel_default_15" [label="(384, 384)", style=solid]; -"84 dequantize_per_channel_default_15" -> "1815 linear_47" [label=None, style=solid]; -"85 features_5_3_attn_qkv_bias" -> "1729 clone_7" [label="(1152,)", style=solid]; -"86 features_5_3_attn_proj_bias" -> "1815 linear_47" [label="(384,)", style=solid]; -"87 features_5_3_attn_logit_scale" -> "1752 clamp_7" [label="(12, 1, 1)", style=solid]; -"88 _scale_16" -> "91 dequantize_per_channel_default_16" [label="(1152,)", style=solid]; -"89 _zero_point_16" -> "91 dequantize_per_channel_default_16" [label="(1152,)", style=solid]; -"90 _frozen_param16" -> "91 dequantize_per_channel_default_16" [label="(1152, 384)", style=solid]; -"91 dequantize_per_channel_default_16" -> "1859 linear_52" [label=None, style=solid]; -"92 _scale_17" -> "95 dequantize_per_channel_default_17" [label="(384,)", style=solid]; -"93 _zero_point_17" -> "95 dequantize_per_channel_default_17" [label="(384,)", style=solid]; -"94 _frozen_param17" -> "95 dequantize_per_channel_default_17" [label="(384, 384)", style=solid]; -"95 dequantize_per_channel_default_17" -> "1890 linear_53" [label=None, style=solid]; -"96 features_5_4_attn_qkv_bias" -> "1856 clone_8" [label="(1152,)", style=solid]; -"97 features_5_4_attn_proj_bias" -> "1890 linear_53" [label="(384,)", style=solid]; -"98 features_5_4_attn_logit_scale" -> "1879 clamp_8" [label="(12, 1, 1)", style=solid]; -"99 _scale_18" -> "102 dequantize_per_channel_default_18" [label="(1152,)", style=solid]; -"100 _zero_point_18" -> "102 dequantize_per_channel_default_18" [label="(1152,)", style=solid]; -"101 _frozen_param18" -> "102 dequantize_per_channel_default_18" [label="(1152, 384)", style=solid]; -"102 dequantize_per_channel_default_18" -> "1934 linear_58" [label=None, style=solid]; -"103 _scale_19" -> "106 dequantize_per_channel_default_19" [label="(384,)", style=solid]; -"104 _zero_point_19" -> "106 dequantize_per_channel_default_19" [label="(384,)", style=solid]; -"105 _frozen_param19" -> "106 dequantize_per_channel_default_19" [label="(384, 384)", style=solid]; -"106 dequantize_per_channel_default_19" -> "2017 linear_59" [label=None, style=solid]; -"107 features_5_5_attn_qkv_bias" -> "1931 clone_9" [label="(1152,)", style=solid]; -"108 features_5_5_attn_proj_bias" -> "2017 linear_59" [label="(384,)", style=solid]; -"109 features_5_5_attn_logit_scale" -> "1954 clamp_9" [label="(12, 1, 1)", style=solid]; -"110 _scale_20" -> "113 dequantize_per_channel_default_20" [label="(1152,)", style=solid]; -"111 _zero_point_20" -> "113 dequantize_per_channel_default_20" [label="(1152,)", style=solid]; -"112 _frozen_param20" -> "113 dequantize_per_channel_default_20" [label="(1152, 384)", style=solid]; -"113 dequantize_per_channel_default_20" -> "2061 linear_64" [label=None, style=solid]; -"114 _scale_21" -> "117 dequantize_per_channel_default_21" [label="(384,)", style=solid]; -"115 _zero_point_21" -> "117 dequantize_per_channel_default_21" [label="(384,)", style=solid]; -"116 _frozen_param21" -> "117 dequantize_per_channel_default_21" [label="(384, 384)", style=solid]; -"117 dequantize_per_channel_default_21" -> "2092 linear_65" [label=None, style=solid]; -"118 features_5_6_attn_qkv_bias" -> "2058 clone_10" [label="(1152,)", style=solid]; -"119 features_5_6_attn_proj_bias" -> "2092 linear_65" [label="(384,)", style=solid]; -"120 features_5_6_attn_logit_scale" -> "2081 clamp_10" [label="(12, 1, 1)", style=solid]; -"121 _scale_22" -> "124 dequantize_per_channel_default_22" [label="(1152,)", style=solid]; -"122 _zero_point_22" -> "124 dequantize_per_channel_default_22" [label="(1152,)", style=solid]; -"123 _frozen_param22" -> "124 dequantize_per_channel_default_22" [label="(1152, 384)", style=solid]; -"124 dequantize_per_channel_default_22" -> "2136 linear_70" [label=None, style=solid]; -"125 _scale_23" -> "128 dequantize_per_channel_default_23" [label="(384,)", style=solid]; -"126 _zero_point_23" -> "128 dequantize_per_channel_default_23" [label="(384,)", style=solid]; -"127 _frozen_param23" -> "128 dequantize_per_channel_default_23" [label="(384, 384)", style=solid]; -"128 dequantize_per_channel_default_23" -> "2219 linear_71" [label=None, style=solid]; -"129 features_5_7_attn_qkv_bias" -> "2133 clone_11" [label="(1152,)", style=solid]; -"130 features_5_7_attn_proj_bias" -> "2219 linear_71" [label="(384,)", style=solid]; -"131 features_5_7_attn_logit_scale" -> "2156 clamp_11" [label="(12, 1, 1)", style=solid]; -"132 _scale_24" -> "135 dequantize_per_channel_default_24" [label="(1152,)", style=solid]; -"133 _zero_point_24" -> "135 dequantize_per_channel_default_24" [label="(1152,)", style=solid]; -"134 _frozen_param24" -> "135 dequantize_per_channel_default_24" [label="(1152, 384)", style=solid]; -"135 dequantize_per_channel_default_24" -> "2263 linear_76" [label=None, style=solid]; -"136 _scale_25" -> "139 dequantize_per_channel_default_25" [label="(384,)", style=solid]; -"137 _zero_point_25" -> "139 dequantize_per_channel_default_25" [label="(384,)", style=solid]; -"138 _frozen_param25" -> "139 dequantize_per_channel_default_25" [label="(384, 384)", style=solid]; -"139 dequantize_per_channel_default_25" -> "2294 linear_77" [label=None, style=solid]; -"140 features_5_8_attn_qkv_bias" -> "2260 clone_12" [label="(1152,)", style=solid]; -"141 features_5_8_attn_proj_bias" -> "2294 linear_77" [label="(384,)", style=solid]; -"142 features_5_8_attn_logit_scale" -> "2283 clamp_12" [label="(12, 1, 1)", style=solid]; -"143 _scale_26" -> "146 dequantize_per_channel_default_26" [label="(1152,)", style=solid]; -"144 _zero_point_26" -> "146 dequantize_per_channel_default_26" [label="(1152,)", style=solid]; -"145 _frozen_param26" -> "146 dequantize_per_channel_default_26" [label="(1152, 384)", style=solid]; -"146 dequantize_per_channel_default_26" -> "2338 linear_82" [label=None, style=solid]; -"147 _scale_27" -> "150 dequantize_per_channel_default_27" [label="(384,)", style=solid]; -"148 _zero_point_27" -> "150 dequantize_per_channel_default_27" [label="(384,)", style=solid]; -"149 _frozen_param27" -> "150 dequantize_per_channel_default_27" [label="(384, 384)", style=solid]; -"150 dequantize_per_channel_default_27" -> "2421 linear_83" [label=None, style=solid]; -"151 features_5_9_attn_qkv_bias" -> "2335 clone_13" [label="(1152,)", style=solid]; -"152 features_5_9_attn_proj_bias" -> "2421 linear_83" [label="(384,)", style=solid]; -"153 features_5_9_attn_logit_scale" -> "2358 clamp_13" [label="(12, 1, 1)", style=solid]; -"154 _scale_28" -> "157 dequantize_per_channel_default_28" [label="(1152,)", style=solid]; -"155 _zero_point_28" -> "157 dequantize_per_channel_default_28" [label="(1152,)", style=solid]; -"156 _frozen_param28" -> "157 dequantize_per_channel_default_28" [label="(1152, 384)", style=solid]; -"157 dequantize_per_channel_default_28" -> "2465 linear_88" [label=None, style=solid]; -"158 _scale_29" -> "161 dequantize_per_channel_default_29" [label="(384,)", style=solid]; -"159 _zero_point_29" -> "161 dequantize_per_channel_default_29" [label="(384,)", style=solid]; -"160 _frozen_param29" -> "161 dequantize_per_channel_default_29" [label="(384, 384)", style=solid]; -"161 dequantize_per_channel_default_29" -> "2496 linear_89" [label=None, style=solid]; -"162 features_5_10_attn_qkv_bias" -> "2462 clone_14" [label="(1152,)", style=solid]; -"163 features_5_10_attn_proj_bias" -> "2496 linear_89" [label="(384,)", style=solid]; -"164 features_5_10_attn_logit_scale" -> "2485 clamp_14" [label="(12, 1, 1)", style=solid]; -"165 _scale_30" -> "168 dequantize_per_channel_default_30" [label="(1152,)", style=solid]; -"166 _zero_point_30" -> "168 dequantize_per_channel_default_30" [label="(1152,)", style=solid]; -"167 _frozen_param30" -> "168 dequantize_per_channel_default_30" [label="(1152, 384)", style=solid]; -"168 dequantize_per_channel_default_30" -> "2540 linear_94" [label=None, style=solid]; -"169 _scale_31" -> "172 dequantize_per_channel_default_31" [label="(384,)", style=solid]; -"170 _zero_point_31" -> "172 dequantize_per_channel_default_31" [label="(384,)", style=solid]; -"171 _frozen_param31" -> "172 dequantize_per_channel_default_31" [label="(384, 384)", style=solid]; -"172 dequantize_per_channel_default_31" -> "2623 linear_95" [label=None, style=solid]; -"173 features_5_11_attn_qkv_bias" -> "2537 clone_15" [label="(1152,)", style=solid]; -"174 features_5_11_attn_proj_bias" -> "2623 linear_95" [label="(384,)", style=solid]; -"175 features_5_11_attn_logit_scale" -> "2560 clamp_15" [label="(12, 1, 1)", style=solid]; -"176 _scale_32" -> "179 dequantize_per_channel_default_32" [label="(1152,)", style=solid]; -"177 _zero_point_32" -> "179 dequantize_per_channel_default_32" [label="(1152,)", style=solid]; -"178 _frozen_param32" -> "179 dequantize_per_channel_default_32" [label="(1152, 384)", style=solid]; -"179 dequantize_per_channel_default_32" -> "2667 linear_100" [label=None, style=solid]; -"180 _scale_33" -> "183 dequantize_per_channel_default_33" [label="(384,)", style=solid]; -"181 _zero_point_33" -> "183 dequantize_per_channel_default_33" [label="(384,)", style=solid]; -"182 _frozen_param33" -> "183 dequantize_per_channel_default_33" [label="(384, 384)", style=solid]; -"183 dequantize_per_channel_default_33" -> "2698 linear_101" [label=None, style=solid]; -"184 features_5_12_attn_qkv_bias" -> "2664 clone_16" [label="(1152,)", style=solid]; -"185 features_5_12_attn_proj_bias" -> "2698 linear_101" [label="(384,)", style=solid]; -"186 features_5_12_attn_logit_scale" -> "2687 clamp_16" [label="(12, 1, 1)", style=solid]; -"187 _scale_34" -> "190 dequantize_per_channel_default_34" [label="(1152,)", style=solid]; -"188 _zero_point_34" -> "190 dequantize_per_channel_default_34" [label="(1152,)", style=solid]; -"189 _frozen_param34" -> "190 dequantize_per_channel_default_34" [label="(1152, 384)", style=solid]; -"190 dequantize_per_channel_default_34" -> "2742 linear_106" [label=None, style=solid]; -"191 _scale_35" -> "194 dequantize_per_channel_default_35" [label="(384,)", style=solid]; -"192 _zero_point_35" -> "194 dequantize_per_channel_default_35" [label="(384,)", style=solid]; -"193 _frozen_param35" -> "194 dequantize_per_channel_default_35" [label="(384, 384)", style=solid]; -"194 dequantize_per_channel_default_35" -> "2825 linear_107" [label=None, style=solid]; -"195 features_5_13_attn_qkv_bias" -> "2739 clone_17" [label="(1152,)", style=solid]; -"196 features_5_13_attn_proj_bias" -> "2825 linear_107" [label="(384,)", style=solid]; -"197 features_5_13_attn_logit_scale" -> "2762 clamp_17" [label="(12, 1, 1)", style=solid]; -"198 _scale_36" -> "201 dequantize_per_channel_default_36" [label="(1152,)", style=solid]; -"199 _zero_point_36" -> "201 dequantize_per_channel_default_36" [label="(1152,)", style=solid]; -"200 _frozen_param36" -> "201 dequantize_per_channel_default_36" [label="(1152, 384)", style=solid]; -"201 dequantize_per_channel_default_36" -> "2869 linear_112" [label=None, style=solid]; -"202 _scale_37" -> "205 dequantize_per_channel_default_37" [label="(384,)", style=solid]; -"203 _zero_point_37" -> "205 dequantize_per_channel_default_37" [label="(384,)", style=solid]; -"204 _frozen_param37" -> "205 dequantize_per_channel_default_37" [label="(384, 384)", style=solid]; -"205 dequantize_per_channel_default_37" -> "2900 linear_113" [label=None, style=solid]; -"206 features_5_14_attn_qkv_bias" -> "2866 clone_18" [label="(1152,)", style=solid]; -"207 features_5_14_attn_proj_bias" -> "2900 linear_113" [label="(384,)", style=solid]; -"208 features_5_14_attn_logit_scale" -> "2889 clamp_18" [label="(12, 1, 1)", style=solid]; -"209 _scale_38" -> "212 dequantize_per_channel_default_38" [label="(1152,)", style=solid]; -"210 _zero_point_38" -> "212 dequantize_per_channel_default_38" [label="(1152,)", style=solid]; -"211 _frozen_param38" -> "212 dequantize_per_channel_default_38" [label="(1152, 384)", style=solid]; -"212 dequantize_per_channel_default_38" -> "2944 linear_118" [label=None, style=solid]; -"213 _scale_39" -> "216 dequantize_per_channel_default_39" [label="(384,)", style=solid]; -"214 _zero_point_39" -> "216 dequantize_per_channel_default_39" [label="(384,)", style=solid]; -"215 _frozen_param39" -> "216 dequantize_per_channel_default_39" [label="(384, 384)", style=solid]; -"216 dequantize_per_channel_default_39" -> "3027 linear_119" [label=None, style=solid]; -"217 features_5_15_attn_qkv_bias" -> "2941 clone_19" [label="(1152,)", style=solid]; -"218 features_5_15_attn_proj_bias" -> "3027 linear_119" [label="(384,)", style=solid]; -"219 features_5_15_attn_logit_scale" -> "2964 clamp_19" [label="(12, 1, 1)", style=solid]; -"220 _scale_40" -> "223 dequantize_per_channel_default_40" [label="(1152,)", style=solid]; -"221 _zero_point_40" -> "223 dequantize_per_channel_default_40" [label="(1152,)", style=solid]; -"222 _frozen_param40" -> "223 dequantize_per_channel_default_40" [label="(1152, 384)", style=solid]; -"223 dequantize_per_channel_default_40" -> "3071 linear_124" [label=None, style=solid]; -"224 _scale_41" -> "227 dequantize_per_channel_default_41" [label="(384,)", style=solid]; -"225 _zero_point_41" -> "227 dequantize_per_channel_default_41" [label="(384,)", style=solid]; -"226 _frozen_param41" -> "227 dequantize_per_channel_default_41" [label="(384, 384)", style=solid]; -"227 dequantize_per_channel_default_41" -> "3102 linear_125" [label=None, style=solid]; -"228 features_5_16_attn_qkv_bias" -> "3068 clone_20" [label="(1152,)", style=solid]; -"229 features_5_16_attn_proj_bias" -> "3102 linear_125" [label="(384,)", style=solid]; -"230 features_5_16_attn_logit_scale" -> "3091 clamp_20" [label="(12, 1, 1)", style=solid]; -"231 _scale_42" -> "234 dequantize_per_channel_default_42" [label="(1152,)", style=solid]; -"232 _zero_point_42" -> "234 dequantize_per_channel_default_42" [label="(1152,)", style=solid]; -"233 _frozen_param42" -> "234 dequantize_per_channel_default_42" [label="(1152, 384)", style=solid]; -"234 dequantize_per_channel_default_42" -> "3146 linear_130" [label=None, style=solid]; -"235 _scale_43" -> "238 dequantize_per_channel_default_43" [label="(384,)", style=solid]; -"236 _zero_point_43" -> "238 dequantize_per_channel_default_43" [label="(384,)", style=solid]; -"237 _frozen_param43" -> "238 dequantize_per_channel_default_43" [label="(384, 384)", style=solid]; -"238 dequantize_per_channel_default_43" -> "3229 linear_131" [label=None, style=solid]; -"239 features_5_17_attn_qkv_bias" -> "3143 clone_21" [label="(1152,)", style=solid]; -"240 features_5_17_attn_proj_bias" -> "3229 linear_131" [label="(384,)", style=solid]; -"241 features_5_17_attn_logit_scale" -> "3166 clamp_21" [label="(12, 1, 1)", style=solid]; -"242 _scale_44" -> "245 dequantize_per_channel_default_44" [label="(2304,)", style=solid]; -"243 _zero_point_44" -> "245 dequantize_per_channel_default_44" [label="(2304,)", style=solid]; -"244 _frozen_param44" -> "245 dequantize_per_channel_default_44" [label="(2304, 768)", style=solid]; -"245 dequantize_per_channel_default_44" -> "3291 linear_137" [label=None, style=solid]; -"246 _scale_45" -> "249 dequantize_per_channel_default_45" [label="(768,)", style=solid]; -"247 _zero_point_45" -> "249 dequantize_per_channel_default_45" [label="(768,)", style=solid]; -"248 _frozen_param45" -> "249 dequantize_per_channel_default_45" [label="(768, 768)", style=solid]; -"249 dequantize_per_channel_default_45" -> "3322 linear_138" [label=None, style=solid]; -"250 features_7_0_attn_qkv_bias" -> "3288 clone_22" [label="(2304,)", style=solid]; -"251 features_7_0_attn_proj_bias" -> "3322 linear_138" [label="(768,)", style=solid]; -"252 features_7_0_attn_logit_scale" -> "3311 clamp_22" [label="(24, 1, 1)", style=solid]; -"253 _scale_46" -> "256 dequantize_per_channel_default_46" [label="(2304,)", style=solid]; -"254 _zero_point_46" -> "256 dequantize_per_channel_default_46" [label="(2304,)", style=solid]; -"255 _frozen_param46" -> "256 dequantize_per_channel_default_46" [label="(2304, 768)", style=solid]; -"256 dequantize_per_channel_default_46" -> "3365 linear_143" [label=None, style=solid]; -"257 _scale_47" -> "260 dequantize_per_channel_default_47" [label="(768,)", style=solid]; -"258 _zero_point_47" -> "260 dequantize_per_channel_default_47" [label="(768,)", style=solid]; -"259 _frozen_param47" -> "260 dequantize_per_channel_default_47" [label="(768, 768)", style=solid]; -"260 dequantize_per_channel_default_47" -> "3396 linear_144" [label=None, style=solid]; -"261 features_7_1_attn_qkv_bias" -> "3362 clone_23" [label="(2304,)", style=solid]; -"262 features_7_1_attn_proj_bias" -> "3396 linear_144" [label="(768,)", style=solid]; -"263 features_7_1_attn_logit_scale" -> "3385 clamp_23" [label="(24, 1, 1)", style=solid]; -"264 _scale_48" -> "267 dequantize_per_channel_default_48" [label="(96,)", style=solid]; -"265 _zero_point_48" -> "267 dequantize_per_channel_default_48" [label="(96,)", style=solid]; -"266 _frozen_param48" -> "267 dequantize_per_channel_default_48" [label="(96, 3, 4, 4)", style=solid]; -"267 dequantize_per_channel_default_48" -> "998 conv2d" [label=None, style=solid]; -"268 features_0_0_bias" -> "998 conv2d" [label="(96,)", style=solid]; -"269 features_0_2_weight" -> "1000 layer_norm" [label="(96,)", style=solid]; -"270 features_0_2_bias" -> "1000 layer_norm" [label="(96,)", style=solid]; -"271 _scale_49" -> "274 dequantize_per_channel_default_49" [label="(512,)", style=solid]; -"272 _zero_point_49" -> "274 dequantize_per_channel_default_49" [label="(512,)", style=solid]; -"273 _frozen_param49" -> "274 dequantize_per_channel_default_49" [label="(512, 2)", style=solid]; -"274 dequantize_per_channel_default_49" -> "1003 linear" [label=None, style=solid]; -"275 features_1_0_attn_cpb_mlp_0_bias" -> "1003 linear" [label="(512,)", style=solid]; -"276 _scale_50" -> "279 dequantize_per_channel_default_50" [label="(3,)", style=solid]; -"277 _zero_point_50" -> "279 dequantize_per_channel_default_50" [label="(3,)", style=solid]; -"278 _frozen_param50" -> "279 dequantize_per_channel_default_50" [label="(3, 512)", style=solid]; -"279 dequantize_per_channel_default_50" -> "1005 linear_1" [label=None, style=solid]; -"280 features_1_0_norm1_weight" -> "1059 layer_norm_1" [label="(96,)", style=solid]; -"281 features_1_0_norm1_bias" -> "1059 layer_norm_1" [label="(96,)", style=solid]; -"282 _scale_51" -> "285 dequantize_per_channel_default_51" [label="(384,)", style=solid]; -"283 _zero_point_51" -> "285 dequantize_per_channel_default_51" [label="(384,)", style=solid]; -"284 _frozen_param51" -> "285 dequantize_per_channel_default_51" [label="(384, 96)", style=solid]; -"285 dequantize_per_channel_default_51" -> "1063 linear_4" [label=None, style=solid]; -"286 features_1_0_mlp_0_bias" -> "1063 linear_4" [label="(384,)", style=solid]; -"287 _scale_52" -> "290 dequantize_per_channel_default_52" [label="(96,)", style=solid]; -"288 _zero_point_52" -> "290 dequantize_per_channel_default_52" [label="(96,)", style=solid]; -"289 _frozen_param52" -> "290 dequantize_per_channel_default_52" [label="(96, 384)", style=solid]; -"290 dequantize_per_channel_default_52" -> "1068 linear_5" [label=None, style=solid]; -"291 features_1_0_mlp_3_bias" -> "1068 linear_5" [label="(96,)", style=solid]; -"292 features_1_0_norm2_weight" -> "1070 layer_norm_2" [label="(96,)", style=solid]; -"293 features_1_0_norm2_bias" -> "1070 layer_norm_2" [label="(96,)", style=solid]; -"294 _scale_53" -> "297 dequantize_per_channel_default_53" [label="(512,)", style=solid]; -"295 _zero_point_53" -> "297 dequantize_per_channel_default_53" [label="(512,)", style=solid]; -"296 _frozen_param53" -> "297 dequantize_per_channel_default_53" [label="(512, 2)", style=solid]; -"297 dequantize_per_channel_default_53" -> "1072 linear_6" [label=None, style=solid]; -"298 features_1_1_attn_cpb_mlp_0_bias" -> "1072 linear_6" [label="(512,)", style=solid]; -"299 _scale_54" -> "302 dequantize_per_channel_default_54" [label="(3,)", style=solid]; -"300 _zero_point_54" -> "302 dequantize_per_channel_default_54" [label="(3,)", style=solid]; -"301 _frozen_param54" -> "302 dequantize_per_channel_default_54" [label="(3, 512)", style=solid]; -"302 dequantize_per_channel_default_54" -> "1074 linear_7" [label=None, style=solid]; -"303 features_1_1_norm1_weight" -> "1184 layer_norm_3" [label="(96,)", style=solid]; -"304 features_1_1_norm1_bias" -> "1184 layer_norm_3" [label="(96,)", style=solid]; -"305 _scale_55" -> "308 dequantize_per_channel_default_55" [label="(384,)", style=solid]; -"306 _zero_point_55" -> "308 dequantize_per_channel_default_55" [label="(384,)", style=solid]; -"307 _frozen_param55" -> "308 dequantize_per_channel_default_55" [label="(384, 96)", style=solid]; -"308 dequantize_per_channel_default_55" -> "1188 linear_10" [label=None, style=solid]; -"309 features_1_1_mlp_0_bias" -> "1188 linear_10" [label="(384,)", style=solid]; -"310 _scale_56" -> "313 dequantize_per_channel_default_56" [label="(96,)", style=solid]; -"311 _zero_point_56" -> "313 dequantize_per_channel_default_56" [label="(96,)", style=solid]; -"312 _frozen_param56" -> "313 dequantize_per_channel_default_56" [label="(96, 384)", style=solid]; -"313 dequantize_per_channel_default_56" -> "1193 linear_11" [label=None, style=solid]; -"314 features_1_1_mlp_3_bias" -> "1193 linear_11" [label="(96,)", style=solid]; -"315 features_1_1_norm2_weight" -> "1195 layer_norm_4" [label="(96,)", style=solid]; -"316 features_1_1_norm2_bias" -> "1195 layer_norm_4" [label="(96,)", style=solid]; -"317 _scale_57" -> "320 dequantize_per_channel_default_57" [label="(192,)", style=solid]; -"318 _zero_point_57" -> "320 dequantize_per_channel_default_57" [label="(192,)", style=solid]; -"319 _frozen_param57" -> "320 dequantize_per_channel_default_57" [label="(192, 384)", style=solid]; -"320 dequantize_per_channel_default_57" -> "1213 linear_12" [label=None, style=solid]; -"321 features_2_norm_weight" -> "1214 layer_norm_5" [label="(192,)", style=solid]; -"322 features_2_norm_bias" -> "1214 layer_norm_5" [label="(192,)", style=solid]; -"323 _scale_58" -> "326 dequantize_per_channel_default_58" [label="(512,)", style=solid]; -"324 _zero_point_58" -> "326 dequantize_per_channel_default_58" [label="(512,)", style=solid]; -"325 _frozen_param58" -> "326 dequantize_per_channel_default_58" [label="(512, 2)", style=solid]; -"326 dequantize_per_channel_default_58" -> "1217 linear_13" [label=None, style=solid]; -"327 features_3_0_attn_cpb_mlp_0_bias" -> "1217 linear_13" [label="(512,)", style=solid]; -"328 _scale_59" -> "331 dequantize_per_channel_default_59" [label="(6,)", style=solid]; -"329 _zero_point_59" -> "331 dequantize_per_channel_default_59" [label="(6,)", style=solid]; -"330 _frozen_param59" -> "331 dequantize_per_channel_default_59" [label="(6, 512)", style=solid]; -"331 dequantize_per_channel_default_59" -> "1219 linear_14" [label=None, style=solid]; -"332 features_3_0_norm1_weight" -> "1276 layer_norm_6" [label="(192,)", style=solid]; -"333 features_3_0_norm1_bias" -> "1276 layer_norm_6" [label="(192,)", style=solid]; -"334 _scale_60" -> "337 dequantize_per_channel_default_60" [label="(768,)", style=solid]; -"335 _zero_point_60" -> "337 dequantize_per_channel_default_60" [label="(768,)", style=solid]; -"336 _frozen_param60" -> "337 dequantize_per_channel_default_60" [label="(768, 192)", style=solid]; -"337 dequantize_per_channel_default_60" -> "1280 linear_17" [label=None, style=solid]; -"338 features_3_0_mlp_0_bias" -> "1280 linear_17" [label="(768,)", style=solid]; -"339 _scale_61" -> "342 dequantize_per_channel_default_61" [label="(192,)", style=solid]; -"340 _zero_point_61" -> "342 dequantize_per_channel_default_61" [label="(192,)", style=solid]; -"341 _frozen_param61" -> "342 dequantize_per_channel_default_61" [label="(192, 768)", style=solid]; -"342 dequantize_per_channel_default_61" -> "1285 linear_18" [label=None, style=solid]; -"343 features_3_0_mlp_3_bias" -> "1285 linear_18" [label="(192,)", style=solid]; -"344 features_3_0_norm2_weight" -> "1287 layer_norm_7" [label="(192,)", style=solid]; -"345 features_3_0_norm2_bias" -> "1287 layer_norm_7" [label="(192,)", style=solid]; -"346 _scale_62" -> "349 dequantize_per_channel_default_62" [label="(512,)", style=solid]; -"347 _zero_point_62" -> "349 dequantize_per_channel_default_62" [label="(512,)", style=solid]; -"348 _frozen_param62" -> "349 dequantize_per_channel_default_62" [label="(512, 2)", style=solid]; -"349 dequantize_per_channel_default_62" -> "1289 linear_19" [label=None, style=solid]; -"350 features_3_1_attn_cpb_mlp_0_bias" -> "1289 linear_19" [label="(512,)", style=solid]; -"351 _scale_63" -> "354 dequantize_per_channel_default_63" [label="(6,)", style=solid]; -"352 _zero_point_63" -> "354 dequantize_per_channel_default_63" [label="(6,)", style=solid]; -"353 _frozen_param63" -> "354 dequantize_per_channel_default_63" [label="(6, 512)", style=solid]; -"354 dequantize_per_channel_default_63" -> "1291 linear_20" [label=None, style=solid]; -"355 features_3_1_norm1_weight" -> "1404 layer_norm_8" [label="(192,)", style=solid]; -"356 features_3_1_norm1_bias" -> "1404 layer_norm_8" [label="(192,)", style=solid]; -"357 _scale_64" -> "360 dequantize_per_channel_default_64" [label="(768,)", style=solid]; -"358 _zero_point_64" -> "360 dequantize_per_channel_default_64" [label="(768,)", style=solid]; -"359 _frozen_param64" -> "360 dequantize_per_channel_default_64" [label="(768, 192)", style=solid]; -"360 dequantize_per_channel_default_64" -> "1408 linear_23" [label=None, style=solid]; -"361 features_3_1_mlp_0_bias" -> "1408 linear_23" [label="(768,)", style=solid]; -"362 _scale_65" -> "365 dequantize_per_channel_default_65" [label="(192,)", style=solid]; -"363 _zero_point_65" -> "365 dequantize_per_channel_default_65" [label="(192,)", style=solid]; -"364 _frozen_param65" -> "365 dequantize_per_channel_default_65" [label="(192, 768)", style=solid]; -"365 dequantize_per_channel_default_65" -> "1413 linear_24" [label=None, style=solid]; -"366 features_3_1_mlp_3_bias" -> "1413 linear_24" [label="(192,)", style=solid]; -"367 features_3_1_norm2_weight" -> "1415 layer_norm_9" [label="(192,)", style=solid]; -"368 features_3_1_norm2_bias" -> "1415 layer_norm_9" [label="(192,)", style=solid]; -"369 _scale_66" -> "372 dequantize_per_channel_default_66" [label="(384,)", style=solid]; -"370 _zero_point_66" -> "372 dequantize_per_channel_default_66" [label="(384,)", style=solid]; -"371 _frozen_param66" -> "372 dequantize_per_channel_default_66" [label="(384, 768)", style=solid]; -"372 dequantize_per_channel_default_66" -> "1433 linear_25" [label=None, style=solid]; -"373 features_4_norm_weight" -> "1434 layer_norm_10" [label="(384,)", style=solid]; -"374 features_4_norm_bias" -> "1434 layer_norm_10" [label="(384,)", style=solid]; -"375 _scale_67" -> "378 dequantize_per_channel_default_67" [label="(512,)", style=solid]; -"376 _zero_point_67" -> "378 dequantize_per_channel_default_67" [label="(512,)", style=solid]; -"377 _frozen_param67" -> "378 dequantize_per_channel_default_67" [label="(512, 2)", style=solid]; -"378 dequantize_per_channel_default_67" -> "1437 linear_26" [label=None, style=solid]; -"379 features_5_0_attn_cpb_mlp_0_bias" -> "1437 linear_26" [label="(512,)", style=solid]; -"380 _scale_68" -> "383 dequantize_per_channel_default_68" [label="(12,)", style=solid]; -"381 _zero_point_68" -> "383 dequantize_per_channel_default_68" [label="(12,)", style=solid]; -"382 _frozen_param68" -> "383 dequantize_per_channel_default_68" [label="(12, 512)", style=solid]; -"383 dequantize_per_channel_default_68" -> "1439 linear_27" [label=None, style=solid]; -"384 features_5_0_norm1_weight" -> "1496 layer_norm_11" [label="(384,)", style=solid]; -"385 features_5_0_norm1_bias" -> "1496 layer_norm_11" [label="(384,)", style=solid]; -"386 _scale_69" -> "389 dequantize_per_channel_default_69" [label="(1536,)", style=solid]; -"387 _zero_point_69" -> "389 dequantize_per_channel_default_69" [label="(1536,)", style=solid]; -"388 _frozen_param69" -> "389 dequantize_per_channel_default_69" [label="(1536, 384)", style=solid]; -"389 dequantize_per_channel_default_69" -> "1500 linear_30" [label=None, style=solid]; -"390 features_5_0_mlp_0_bias" -> "1500 linear_30" [label="(1536,)", style=solid]; -"391 _scale_70" -> "394 dequantize_per_channel_default_70" [label="(384,)", style=solid]; -"392 _zero_point_70" -> "394 dequantize_per_channel_default_70" [label="(384,)", style=solid]; -"393 _frozen_param70" -> "394 dequantize_per_channel_default_70" [label="(384, 1536)", style=solid]; -"394 dequantize_per_channel_default_70" -> "1505 linear_31" [label=None, style=solid]; -"395 features_5_0_mlp_3_bias" -> "1505 linear_31" [label="(384,)", style=solid]; -"396 features_5_0_norm2_weight" -> "1507 layer_norm_12" [label="(384,)", style=solid]; -"397 features_5_0_norm2_bias" -> "1507 layer_norm_12" [label="(384,)", style=solid]; -"398 _scale_71" -> "401 dequantize_per_channel_default_71" [label="(512,)", style=solid]; -"399 _zero_point_71" -> "401 dequantize_per_channel_default_71" [label="(512,)", style=solid]; -"400 _frozen_param71" -> "401 dequantize_per_channel_default_71" [label="(512, 2)", style=solid]; -"401 dequantize_per_channel_default_71" -> "1509 linear_32" [label=None, style=solid]; -"402 features_5_1_attn_cpb_mlp_0_bias" -> "1509 linear_32" [label="(512,)", style=solid]; -"403 _scale_72" -> "406 dequantize_per_channel_default_72" [label="(12,)", style=solid]; -"404 _zero_point_72" -> "406 dequantize_per_channel_default_72" [label="(12,)", style=solid]; -"405 _frozen_param72" -> "406 dequantize_per_channel_default_72" [label="(12, 512)", style=solid]; -"406 dequantize_per_channel_default_72" -> "1511 linear_33" [label=None, style=solid]; -"407 features_5_1_norm1_weight" -> "1624 layer_norm_13" [label="(384,)", style=solid]; -"408 features_5_1_norm1_bias" -> "1624 layer_norm_13" [label="(384,)", style=solid]; -"409 _scale_73" -> "412 dequantize_per_channel_default_73" [label="(1536,)", style=solid]; -"410 _zero_point_73" -> "412 dequantize_per_channel_default_73" [label="(1536,)", style=solid]; -"411 _frozen_param73" -> "412 dequantize_per_channel_default_73" [label="(1536, 384)", style=solid]; -"412 dequantize_per_channel_default_73" -> "1628 linear_36" [label=None, style=solid]; -"413 features_5_1_mlp_0_bias" -> "1628 linear_36" [label="(1536,)", style=solid]; -"414 _scale_74" -> "417 dequantize_per_channel_default_74" [label="(384,)", style=solid]; -"415 _zero_point_74" -> "417 dequantize_per_channel_default_74" [label="(384,)", style=solid]; -"416 _frozen_param74" -> "417 dequantize_per_channel_default_74" [label="(384, 1536)", style=solid]; -"417 dequantize_per_channel_default_74" -> "1633 linear_37" [label=None, style=solid]; -"418 features_5_1_mlp_3_bias" -> "1633 linear_37" [label="(384,)", style=solid]; -"419 features_5_1_norm2_weight" -> "1635 layer_norm_14" [label="(384,)", style=solid]; -"420 features_5_1_norm2_bias" -> "1635 layer_norm_14" [label="(384,)", style=solid]; -"421 _scale_75" -> "424 dequantize_per_channel_default_75" [label="(512,)", style=solid]; -"422 _zero_point_75" -> "424 dequantize_per_channel_default_75" [label="(512,)", style=solid]; -"423 _frozen_param75" -> "424 dequantize_per_channel_default_75" [label="(512, 2)", style=solid]; -"424 dequantize_per_channel_default_75" -> "1639 linear_38" [label=None, style=solid]; -"425 features_5_2_attn_cpb_mlp_0_bias" -> "1639 linear_38" [label="(512,)", style=solid]; -"426 _scale_76" -> "429 dequantize_per_channel_default_76" [label="(12,)", style=solid]; -"427 _zero_point_76" -> "429 dequantize_per_channel_default_76" [label="(12,)", style=solid]; -"428 _frozen_param76" -> "429 dequantize_per_channel_default_76" [label="(12, 512)", style=solid]; -"429 dequantize_per_channel_default_76" -> "1641 linear_39" [label=None, style=solid]; -"430 features_5_2_norm1_weight" -> "1698 layer_norm_15" [label="(384,)", style=solid]; -"431 features_5_2_norm1_bias" -> "1698 layer_norm_15" [label="(384,)", style=solid]; -"432 _scale_77" -> "435 dequantize_per_channel_default_77" [label="(1536,)", style=solid]; -"433 _zero_point_77" -> "435 dequantize_per_channel_default_77" [label="(1536,)", style=solid]; -"434 _frozen_param77" -> "435 dequantize_per_channel_default_77" [label="(1536, 384)", style=solid]; -"435 dequantize_per_channel_default_77" -> "1702 linear_42" [label=None, style=solid]; -"436 features_5_2_mlp_0_bias" -> "1702 linear_42" [label="(1536,)", style=solid]; -"437 _scale_78" -> "440 dequantize_per_channel_default_78" [label="(384,)", style=solid]; -"438 _zero_point_78" -> "440 dequantize_per_channel_default_78" [label="(384,)", style=solid]; -"439 _frozen_param78" -> "440 dequantize_per_channel_default_78" [label="(384, 1536)", style=solid]; -"440 dequantize_per_channel_default_78" -> "1707 linear_43" [label=None, style=solid]; -"441 features_5_2_mlp_3_bias" -> "1707 linear_43" [label="(384,)", style=solid]; -"442 features_5_2_norm2_weight" -> "1709 layer_norm_16" [label="(384,)", style=solid]; -"443 features_5_2_norm2_bias" -> "1709 layer_norm_16" [label="(384,)", style=solid]; -"444 _scale_79" -> "447 dequantize_per_channel_default_79" [label="(512,)", style=solid]; -"445 _zero_point_79" -> "447 dequantize_per_channel_default_79" [label="(512,)", style=solid]; -"446 _frozen_param79" -> "447 dequantize_per_channel_default_79" [label="(512, 2)", style=solid]; -"447 dequantize_per_channel_default_79" -> "1711 linear_44" [label=None, style=solid]; -"448 features_5_3_attn_cpb_mlp_0_bias" -> "1711 linear_44" [label="(512,)", style=solid]; -"449 _scale_80" -> "452 dequantize_per_channel_default_80" [label="(12,)", style=solid]; -"450 _zero_point_80" -> "452 dequantize_per_channel_default_80" [label="(12,)", style=solid]; -"451 _frozen_param80" -> "452 dequantize_per_channel_default_80" [label="(12, 512)", style=solid]; -"452 dequantize_per_channel_default_80" -> "1713 linear_45" [label=None, style=solid]; -"453 features_5_3_norm1_weight" -> "1826 layer_norm_17" [label="(384,)", style=solid]; -"454 features_5_3_norm1_bias" -> "1826 layer_norm_17" [label="(384,)", style=solid]; -"455 _scale_81" -> "458 dequantize_per_channel_default_81" [label="(1536,)", style=solid]; -"456 _zero_point_81" -> "458 dequantize_per_channel_default_81" [label="(1536,)", style=solid]; -"457 _frozen_param81" -> "458 dequantize_per_channel_default_81" [label="(1536, 384)", style=solid]; -"458 dequantize_per_channel_default_81" -> "1830 linear_48" [label=None, style=solid]; -"459 features_5_3_mlp_0_bias" -> "1830 linear_48" [label="(1536,)", style=solid]; -"460 _scale_82" -> "463 dequantize_per_channel_default_82" [label="(384,)", style=solid]; -"461 _zero_point_82" -> "463 dequantize_per_channel_default_82" [label="(384,)", style=solid]; -"462 _frozen_param82" -> "463 dequantize_per_channel_default_82" [label="(384, 1536)", style=solid]; -"463 dequantize_per_channel_default_82" -> "1835 linear_49" [label=None, style=solid]; -"464 features_5_3_mlp_3_bias" -> "1835 linear_49" [label="(384,)", style=solid]; -"465 features_5_3_norm2_weight" -> "1837 layer_norm_18" [label="(384,)", style=solid]; -"466 features_5_3_norm2_bias" -> "1837 layer_norm_18" [label="(384,)", style=solid]; -"467 _scale_83" -> "470 dequantize_per_channel_default_83" [label="(512,)", style=solid]; -"468 _zero_point_83" -> "470 dequantize_per_channel_default_83" [label="(512,)", style=solid]; -"469 _frozen_param83" -> "470 dequantize_per_channel_default_83" [label="(512, 2)", style=solid]; -"470 dequantize_per_channel_default_83" -> "1841 linear_50" [label=None, style=solid]; -"471 features_5_4_attn_cpb_mlp_0_bias" -> "1841 linear_50" [label="(512,)", style=solid]; -"472 _scale_84" -> "475 dequantize_per_channel_default_84" [label="(12,)", style=solid]; -"473 _zero_point_84" -> "475 dequantize_per_channel_default_84" [label="(12,)", style=solid]; -"474 _frozen_param84" -> "475 dequantize_per_channel_default_84" [label="(12, 512)", style=solid]; -"475 dequantize_per_channel_default_84" -> "1843 linear_51" [label=None, style=solid]; -"476 features_5_4_norm1_weight" -> "1900 layer_norm_19" [label="(384,)", style=solid]; -"477 features_5_4_norm1_bias" -> "1900 layer_norm_19" [label="(384,)", style=solid]; -"478 _scale_85" -> "481 dequantize_per_channel_default_85" [label="(1536,)", style=solid]; -"479 _zero_point_85" -> "481 dequantize_per_channel_default_85" [label="(1536,)", style=solid]; -"480 _frozen_param85" -> "481 dequantize_per_channel_default_85" [label="(1536, 384)", style=solid]; -"481 dequantize_per_channel_default_85" -> "1904 linear_54" [label=None, style=solid]; -"482 features_5_4_mlp_0_bias" -> "1904 linear_54" [label="(1536,)", style=solid]; -"483 _scale_86" -> "486 dequantize_per_channel_default_86" [label="(384,)", style=solid]; -"484 _zero_point_86" -> "486 dequantize_per_channel_default_86" [label="(384,)", style=solid]; -"485 _frozen_param86" -> "486 dequantize_per_channel_default_86" [label="(384, 1536)", style=solid]; -"486 dequantize_per_channel_default_86" -> "1909 linear_55" [label=None, style=solid]; -"487 features_5_4_mlp_3_bias" -> "1909 linear_55" [label="(384,)", style=solid]; -"488 features_5_4_norm2_weight" -> "1911 layer_norm_20" [label="(384,)", style=solid]; -"489 features_5_4_norm2_bias" -> "1911 layer_norm_20" [label="(384,)", style=solid]; -"490 _scale_87" -> "493 dequantize_per_channel_default_87" [label="(512,)", style=solid]; -"491 _zero_point_87" -> "493 dequantize_per_channel_default_87" [label="(512,)", style=solid]; -"492 _frozen_param87" -> "493 dequantize_per_channel_default_87" [label="(512, 2)", style=solid]; -"493 dequantize_per_channel_default_87" -> "1913 linear_56" [label=None, style=solid]; -"494 features_5_5_attn_cpb_mlp_0_bias" -> "1913 linear_56" [label="(512,)", style=solid]; -"495 _scale_88" -> "498 dequantize_per_channel_default_88" [label="(12,)", style=solid]; -"496 _zero_point_88" -> "498 dequantize_per_channel_default_88" [label="(12,)", style=solid]; -"497 _frozen_param88" -> "498 dequantize_per_channel_default_88" [label="(12, 512)", style=solid]; -"498 dequantize_per_channel_default_88" -> "1915 linear_57" [label=None, style=solid]; -"499 features_5_5_norm1_weight" -> "2028 layer_norm_21" [label="(384,)", style=solid]; -"500 features_5_5_norm1_bias" -> "2028 layer_norm_21" [label="(384,)", style=solid]; -"501 _scale_89" -> "504 dequantize_per_channel_default_89" [label="(1536,)", style=solid]; -"502 _zero_point_89" -> "504 dequantize_per_channel_default_89" [label="(1536,)", style=solid]; -"503 _frozen_param89" -> "504 dequantize_per_channel_default_89" [label="(1536, 384)", style=solid]; -"504 dequantize_per_channel_default_89" -> "2032 linear_60" [label=None, style=solid]; -"505 features_5_5_mlp_0_bias" -> "2032 linear_60" [label="(1536,)", style=solid]; -"506 _scale_90" -> "509 dequantize_per_channel_default_90" [label="(384,)", style=solid]; -"507 _zero_point_90" -> "509 dequantize_per_channel_default_90" [label="(384,)", style=solid]; -"508 _frozen_param90" -> "509 dequantize_per_channel_default_90" [label="(384, 1536)", style=solid]; -"509 dequantize_per_channel_default_90" -> "2037 linear_61" [label=None, style=solid]; -"510 features_5_5_mlp_3_bias" -> "2037 linear_61" [label="(384,)", style=solid]; -"511 features_5_5_norm2_weight" -> "2039 layer_norm_22" [label="(384,)", style=solid]; -"512 features_5_5_norm2_bias" -> "2039 layer_norm_22" [label="(384,)", style=solid]; -"513 _scale_91" -> "516 dequantize_per_channel_default_91" [label="(512,)", style=solid]; -"514 _zero_point_91" -> "516 dequantize_per_channel_default_91" [label="(512,)", style=solid]; -"515 _frozen_param91" -> "516 dequantize_per_channel_default_91" [label="(512, 2)", style=solid]; -"516 dequantize_per_channel_default_91" -> "2043 linear_62" [label=None, style=solid]; -"517 features_5_6_attn_cpb_mlp_0_bias" -> "2043 linear_62" [label="(512,)", style=solid]; -"518 _scale_92" -> "521 dequantize_per_channel_default_92" [label="(12,)", style=solid]; -"519 _zero_point_92" -> "521 dequantize_per_channel_default_92" [label="(12,)", style=solid]; -"520 _frozen_param92" -> "521 dequantize_per_channel_default_92" [label="(12, 512)", style=solid]; -"521 dequantize_per_channel_default_92" -> "2045 linear_63" [label=None, style=solid]; -"522 features_5_6_norm1_weight" -> "2102 layer_norm_23" [label="(384,)", style=solid]; -"523 features_5_6_norm1_bias" -> "2102 layer_norm_23" [label="(384,)", style=solid]; -"524 _scale_93" -> "527 dequantize_per_channel_default_93" [label="(1536,)", style=solid]; -"525 _zero_point_93" -> "527 dequantize_per_channel_default_93" [label="(1536,)", style=solid]; -"526 _frozen_param93" -> "527 dequantize_per_channel_default_93" [label="(1536, 384)", style=solid]; -"527 dequantize_per_channel_default_93" -> "2106 linear_66" [label=None, style=solid]; -"528 features_5_6_mlp_0_bias" -> "2106 linear_66" [label="(1536,)", style=solid]; -"529 _scale_94" -> "532 dequantize_per_channel_default_94" [label="(384,)", style=solid]; -"530 _zero_point_94" -> "532 dequantize_per_channel_default_94" [label="(384,)", style=solid]; -"531 _frozen_param94" -> "532 dequantize_per_channel_default_94" [label="(384, 1536)", style=solid]; -"532 dequantize_per_channel_default_94" -> "2111 linear_67" [label=None, style=solid]; -"533 features_5_6_mlp_3_bias" -> "2111 linear_67" [label="(384,)", style=solid]; -"534 features_5_6_norm2_weight" -> "2113 layer_norm_24" [label="(384,)", style=solid]; -"535 features_5_6_norm2_bias" -> "2113 layer_norm_24" [label="(384,)", style=solid]; -"536 _scale_95" -> "539 dequantize_per_channel_default_95" [label="(512,)", style=solid]; -"537 _zero_point_95" -> "539 dequantize_per_channel_default_95" [label="(512,)", style=solid]; -"538 _frozen_param95" -> "539 dequantize_per_channel_default_95" [label="(512, 2)", style=solid]; -"539 dequantize_per_channel_default_95" -> "2115 linear_68" [label=None, style=solid]; -"540 features_5_7_attn_cpb_mlp_0_bias" -> "2115 linear_68" [label="(512,)", style=solid]; -"541 _scale_96" -> "544 dequantize_per_channel_default_96" [label="(12,)", style=solid]; -"542 _zero_point_96" -> "544 dequantize_per_channel_default_96" [label="(12,)", style=solid]; -"543 _frozen_param96" -> "544 dequantize_per_channel_default_96" [label="(12, 512)", style=solid]; -"544 dequantize_per_channel_default_96" -> "2117 linear_69" [label=None, style=solid]; -"545 features_5_7_norm1_weight" -> "2230 layer_norm_25" [label="(384,)", style=solid]; -"546 features_5_7_norm1_bias" -> "2230 layer_norm_25" [label="(384,)", style=solid]; -"547 _scale_97" -> "550 dequantize_per_channel_default_97" [label="(1536,)", style=solid]; -"548 _zero_point_97" -> "550 dequantize_per_channel_default_97" [label="(1536,)", style=solid]; -"549 _frozen_param97" -> "550 dequantize_per_channel_default_97" [label="(1536, 384)", style=solid]; -"550 dequantize_per_channel_default_97" -> "2234 linear_72" [label=None, style=solid]; -"551 features_5_7_mlp_0_bias" -> "2234 linear_72" [label="(1536,)", style=solid]; -"552 _scale_98" -> "555 dequantize_per_channel_default_98" [label="(384,)", style=solid]; -"553 _zero_point_98" -> "555 dequantize_per_channel_default_98" [label="(384,)", style=solid]; -"554 _frozen_param98" -> "555 dequantize_per_channel_default_98" [label="(384, 1536)", style=solid]; -"555 dequantize_per_channel_default_98" -> "2239 linear_73" [label=None, style=solid]; -"556 features_5_7_mlp_3_bias" -> "2239 linear_73" [label="(384,)", style=solid]; -"557 features_5_7_norm2_weight" -> "2241 layer_norm_26" [label="(384,)", style=solid]; -"558 features_5_7_norm2_bias" -> "2241 layer_norm_26" [label="(384,)", style=solid]; -"559 _scale_99" -> "562 dequantize_per_channel_default_99" [label="(512,)", style=solid]; -"560 _zero_point_99" -> "562 dequantize_per_channel_default_99" [label="(512,)", style=solid]; -"561 _frozen_param99" -> "562 dequantize_per_channel_default_99" [label="(512, 2)", style=solid]; -"562 dequantize_per_channel_default_99" -> "2245 linear_74" [label=None, style=solid]; -"563 features_5_8_attn_cpb_mlp_0_bias" -> "2245 linear_74" [label="(512,)", style=solid]; -"564 _scale_100" -> "567 dequantize_per_channel_default_100" [label="(12,)", style=solid]; -"565 _zero_point_100" -> "567 dequantize_per_channel_default_100" [label="(12,)", style=solid]; -"566 _frozen_param100" -> "567 dequantize_per_channel_default_100" [label="(12, 512)", style=solid]; -"567 dequantize_per_channel_default_100" -> "2247 linear_75" [label=None, style=solid]; -"568 features_5_8_norm1_weight" -> "2304 layer_norm_27" [label="(384,)", style=solid]; -"569 features_5_8_norm1_bias" -> "2304 layer_norm_27" [label="(384,)", style=solid]; -"570 _scale_101" -> "573 dequantize_per_channel_default_101" [label="(1536,)", style=solid]; -"571 _zero_point_101" -> "573 dequantize_per_channel_default_101" [label="(1536,)", style=solid]; -"572 _frozen_param101" -> "573 dequantize_per_channel_default_101" [label="(1536, 384)", style=solid]; -"573 dequantize_per_channel_default_101" -> "2308 linear_78" [label=None, style=solid]; -"574 features_5_8_mlp_0_bias" -> "2308 linear_78" [label="(1536,)", style=solid]; -"575 _scale_102" -> "578 dequantize_per_channel_default_102" [label="(384,)", style=solid]; -"576 _zero_point_102" -> "578 dequantize_per_channel_default_102" [label="(384,)", style=solid]; -"577 _frozen_param102" -> "578 dequantize_per_channel_default_102" [label="(384, 1536)", style=solid]; -"578 dequantize_per_channel_default_102" -> "2313 linear_79" [label=None, style=solid]; -"579 features_5_8_mlp_3_bias" -> "2313 linear_79" [label="(384,)", style=solid]; -"580 features_5_8_norm2_weight" -> "2315 layer_norm_28" [label="(384,)", style=solid]; -"581 features_5_8_norm2_bias" -> "2315 layer_norm_28" [label="(384,)", style=solid]; -"582 _scale_103" -> "585 dequantize_per_channel_default_103" [label="(512,)", style=solid]; -"583 _zero_point_103" -> "585 dequantize_per_channel_default_103" [label="(512,)", style=solid]; -"584 _frozen_param103" -> "585 dequantize_per_channel_default_103" [label="(512, 2)", style=solid]; -"585 dequantize_per_channel_default_103" -> "2317 linear_80" [label=None, style=solid]; -"586 features_5_9_attn_cpb_mlp_0_bias" -> "2317 linear_80" [label="(512,)", style=solid]; -"587 _scale_104" -> "590 dequantize_per_channel_default_104" [label="(12,)", style=solid]; -"588 _zero_point_104" -> "590 dequantize_per_channel_default_104" [label="(12,)", style=solid]; -"589 _frozen_param104" -> "590 dequantize_per_channel_default_104" [label="(12, 512)", style=solid]; -"590 dequantize_per_channel_default_104" -> "2319 linear_81" [label=None, style=solid]; -"591 features_5_9_norm1_weight" -> "2432 layer_norm_29" [label="(384,)", style=solid]; -"592 features_5_9_norm1_bias" -> "2432 layer_norm_29" [label="(384,)", style=solid]; -"593 _scale_105" -> "596 dequantize_per_channel_default_105" [label="(1536,)", style=solid]; -"594 _zero_point_105" -> "596 dequantize_per_channel_default_105" [label="(1536,)", style=solid]; -"595 _frozen_param105" -> "596 dequantize_per_channel_default_105" [label="(1536, 384)", style=solid]; -"596 dequantize_per_channel_default_105" -> "2436 linear_84" [label=None, style=solid]; -"597 features_5_9_mlp_0_bias" -> "2436 linear_84" [label="(1536,)", style=solid]; -"598 _scale_106" -> "601 dequantize_per_channel_default_106" [label="(384,)", style=solid]; -"599 _zero_point_106" -> "601 dequantize_per_channel_default_106" [label="(384,)", style=solid]; -"600 _frozen_param106" -> "601 dequantize_per_channel_default_106" [label="(384, 1536)", style=solid]; -"601 dequantize_per_channel_default_106" -> "2441 linear_85" [label=None, style=solid]; -"602 features_5_9_mlp_3_bias" -> "2441 linear_85" [label="(384,)", style=solid]; -"603 features_5_9_norm2_weight" -> "2443 layer_norm_30" [label="(384,)", style=solid]; -"604 features_5_9_norm2_bias" -> "2443 layer_norm_30" [label="(384,)", style=solid]; -"605 _scale_107" -> "608 dequantize_per_channel_default_107" [label="(512,)", style=solid]; -"606 _zero_point_107" -> "608 dequantize_per_channel_default_107" [label="(512,)", style=solid]; -"607 _frozen_param107" -> "608 dequantize_per_channel_default_107" [label="(512, 2)", style=solid]; -"608 dequantize_per_channel_default_107" -> "2447 linear_86" [label=None, style=solid]; -"609 features_5_10_attn_cpb_mlp_0_bias" -> "2447 linear_86" [label="(512,)", style=solid]; -"610 _scale_108" -> "613 dequantize_per_channel_default_108" [label="(12,)", style=solid]; -"611 _zero_point_108" -> "613 dequantize_per_channel_default_108" [label="(12,)", style=solid]; -"612 _frozen_param108" -> "613 dequantize_per_channel_default_108" [label="(12, 512)", style=solid]; -"613 dequantize_per_channel_default_108" -> "2449 linear_87" [label=None, style=solid]; -"614 features_5_10_norm1_weight" -> "2506 layer_norm_31" [label="(384,)", style=solid]; -"615 features_5_10_norm1_bias" -> "2506 layer_norm_31" [label="(384,)", style=solid]; -"616 _scale_109" -> "619 dequantize_per_channel_default_109" [label="(1536,)", style=solid]; -"617 _zero_point_109" -> "619 dequantize_per_channel_default_109" [label="(1536,)", style=solid]; -"618 _frozen_param109" -> "619 dequantize_per_channel_default_109" [label="(1536, 384)", style=solid]; -"619 dequantize_per_channel_default_109" -> "2510 linear_90" [label=None, style=solid]; -"620 features_5_10_mlp_0_bias" -> "2510 linear_90" [label="(1536,)", style=solid]; -"621 _scale_110" -> "624 dequantize_per_channel_default_110" [label="(384,)", style=solid]; -"622 _zero_point_110" -> "624 dequantize_per_channel_default_110" [label="(384,)", style=solid]; -"623 _frozen_param110" -> "624 dequantize_per_channel_default_110" [label="(384, 1536)", style=solid]; -"624 dequantize_per_channel_default_110" -> "2515 linear_91" [label=None, style=solid]; -"625 features_5_10_mlp_3_bias" -> "2515 linear_91" [label="(384,)", style=solid]; -"626 features_5_10_norm2_weight" -> "2517 layer_norm_32" [label="(384,)", style=solid]; -"627 features_5_10_norm2_bias" -> "2517 layer_norm_32" [label="(384,)", style=solid]; -"628 _scale_111" -> "631 dequantize_per_channel_default_111" [label="(512,)", style=solid]; -"629 _zero_point_111" -> "631 dequantize_per_channel_default_111" [label="(512,)", style=solid]; -"630 _frozen_param111" -> "631 dequantize_per_channel_default_111" [label="(512, 2)", style=solid]; -"631 dequantize_per_channel_default_111" -> "2519 linear_92" [label=None, style=solid]; -"632 features_5_11_attn_cpb_mlp_0_bias" -> "2519 linear_92" [label="(512,)", style=solid]; -"633 _scale_112" -> "636 dequantize_per_channel_default_112" [label="(12,)", style=solid]; -"634 _zero_point_112" -> "636 dequantize_per_channel_default_112" [label="(12,)", style=solid]; -"635 _frozen_param112" -> "636 dequantize_per_channel_default_112" [label="(12, 512)", style=solid]; -"636 dequantize_per_channel_default_112" -> "2521 linear_93" [label=None, style=solid]; -"637 features_5_11_norm1_weight" -> "2634 layer_norm_33" [label="(384,)", style=solid]; -"638 features_5_11_norm1_bias" -> "2634 layer_norm_33" [label="(384,)", style=solid]; -"639 _scale_113" -> "642 dequantize_per_channel_default_113" [label="(1536,)", style=solid]; -"640 _zero_point_113" -> "642 dequantize_per_channel_default_113" [label="(1536,)", style=solid]; -"641 _frozen_param113" -> "642 dequantize_per_channel_default_113" [label="(1536, 384)", style=solid]; -"642 dequantize_per_channel_default_113" -> "2638 linear_96" [label=None, style=solid]; -"643 features_5_11_mlp_0_bias" -> "2638 linear_96" [label="(1536,)", style=solid]; -"644 _scale_114" -> "647 dequantize_per_channel_default_114" [label="(384,)", style=solid]; -"645 _zero_point_114" -> "647 dequantize_per_channel_default_114" [label="(384,)", style=solid]; -"646 _frozen_param114" -> "647 dequantize_per_channel_default_114" [label="(384, 1536)", style=solid]; -"647 dequantize_per_channel_default_114" -> "2643 linear_97" [label=None, style=solid]; -"648 features_5_11_mlp_3_bias" -> "2643 linear_97" [label="(384,)", style=solid]; -"649 features_5_11_norm2_weight" -> "2645 layer_norm_34" [label="(384,)", style=solid]; -"650 features_5_11_norm2_bias" -> "2645 layer_norm_34" [label="(384,)", style=solid]; -"651 _scale_115" -> "654 dequantize_per_channel_default_115" [label="(512,)", style=solid]; -"652 _zero_point_115" -> "654 dequantize_per_channel_default_115" [label="(512,)", style=solid]; -"653 _frozen_param115" -> "654 dequantize_per_channel_default_115" [label="(512, 2)", style=solid]; -"654 dequantize_per_channel_default_115" -> "2649 linear_98" [label=None, style=solid]; -"655 features_5_12_attn_cpb_mlp_0_bias" -> "2649 linear_98" [label="(512,)", style=solid]; -"656 _scale_116" -> "659 dequantize_per_channel_default_116" [label="(12,)", style=solid]; -"657 _zero_point_116" -> "659 dequantize_per_channel_default_116" [label="(12,)", style=solid]; -"658 _frozen_param116" -> "659 dequantize_per_channel_default_116" [label="(12, 512)", style=solid]; -"659 dequantize_per_channel_default_116" -> "2651 linear_99" [label=None, style=solid]; -"660 features_5_12_norm1_weight" -> "2708 layer_norm_35" [label="(384,)", style=solid]; -"661 features_5_12_norm1_bias" -> "2708 layer_norm_35" [label="(384,)", style=solid]; -"662 _scale_117" -> "665 dequantize_per_channel_default_117" [label="(1536,)", style=solid]; -"663 _zero_point_117" -> "665 dequantize_per_channel_default_117" [label="(1536,)", style=solid]; -"664 _frozen_param117" -> "665 dequantize_per_channel_default_117" [label="(1536, 384)", style=solid]; -"665 dequantize_per_channel_default_117" -> "2712 linear_102" [label=None, style=solid]; -"666 features_5_12_mlp_0_bias" -> "2712 linear_102" [label="(1536,)", style=solid]; -"667 _scale_118" -> "670 dequantize_per_channel_default_118" [label="(384,)", style=solid]; -"668 _zero_point_118" -> "670 dequantize_per_channel_default_118" [label="(384,)", style=solid]; -"669 _frozen_param118" -> "670 dequantize_per_channel_default_118" [label="(384, 1536)", style=solid]; -"670 dequantize_per_channel_default_118" -> "2717 linear_103" [label=None, style=solid]; -"671 features_5_12_mlp_3_bias" -> "2717 linear_103" [label="(384,)", style=solid]; -"672 features_5_12_norm2_weight" -> "2719 layer_norm_36" [label="(384,)", style=solid]; -"673 features_5_12_norm2_bias" -> "2719 layer_norm_36" [label="(384,)", style=solid]; -"674 _scale_119" -> "677 dequantize_per_channel_default_119" [label="(512,)", style=solid]; -"675 _zero_point_119" -> "677 dequantize_per_channel_default_119" [label="(512,)", style=solid]; -"676 _frozen_param119" -> "677 dequantize_per_channel_default_119" [label="(512, 2)", style=solid]; -"677 dequantize_per_channel_default_119" -> "2721 linear_104" [label=None, style=solid]; -"678 features_5_13_attn_cpb_mlp_0_bias" -> "2721 linear_104" [label="(512,)", style=solid]; -"679 _scale_120" -> "682 dequantize_per_channel_default_120" [label="(12,)", style=solid]; -"680 _zero_point_120" -> "682 dequantize_per_channel_default_120" [label="(12,)", style=solid]; -"681 _frozen_param120" -> "682 dequantize_per_channel_default_120" [label="(12, 512)", style=solid]; -"682 dequantize_per_channel_default_120" -> "2723 linear_105" [label=None, style=solid]; -"683 features_5_13_norm1_weight" -> "2836 layer_norm_37" [label="(384,)", style=solid]; -"684 features_5_13_norm1_bias" -> "2836 layer_norm_37" [label="(384,)", style=solid]; -"685 _scale_121" -> "688 dequantize_per_channel_default_121" [label="(1536,)", style=solid]; -"686 _zero_point_121" -> "688 dequantize_per_channel_default_121" [label="(1536,)", style=solid]; -"687 _frozen_param121" -> "688 dequantize_per_channel_default_121" [label="(1536, 384)", style=solid]; -"688 dequantize_per_channel_default_121" -> "2840 linear_108" [label=None, style=solid]; -"689 features_5_13_mlp_0_bias" -> "2840 linear_108" [label="(1536,)", style=solid]; -"690 _scale_122" -> "693 dequantize_per_channel_default_122" [label="(384,)", style=solid]; -"691 _zero_point_122" -> "693 dequantize_per_channel_default_122" [label="(384,)", style=solid]; -"692 _frozen_param122" -> "693 dequantize_per_channel_default_122" [label="(384, 1536)", style=solid]; -"693 dequantize_per_channel_default_122" -> "2845 linear_109" [label=None, style=solid]; -"694 features_5_13_mlp_3_bias" -> "2845 linear_109" [label="(384,)", style=solid]; -"695 features_5_13_norm2_weight" -> "2847 layer_norm_38" [label="(384,)", style=solid]; -"696 features_5_13_norm2_bias" -> "2847 layer_norm_38" [label="(384,)", style=solid]; -"697 _scale_123" -> "700 dequantize_per_channel_default_123" [label="(512,)", style=solid]; -"698 _zero_point_123" -> "700 dequantize_per_channel_default_123" [label="(512,)", style=solid]; -"699 _frozen_param123" -> "700 dequantize_per_channel_default_123" [label="(512, 2)", style=solid]; -"700 dequantize_per_channel_default_123" -> "2851 linear_110" [label=None, style=solid]; -"701 features_5_14_attn_cpb_mlp_0_bias" -> "2851 linear_110" [label="(512,)", style=solid]; -"702 _scale_124" -> "705 dequantize_per_channel_default_124" [label="(12,)", style=solid]; -"703 _zero_point_124" -> "705 dequantize_per_channel_default_124" [label="(12,)", style=solid]; -"704 _frozen_param124" -> "705 dequantize_per_channel_default_124" [label="(12, 512)", style=solid]; -"705 dequantize_per_channel_default_124" -> "2853 linear_111" [label=None, style=solid]; -"706 features_5_14_norm1_weight" -> "2910 layer_norm_39" [label="(384,)", style=solid]; -"707 features_5_14_norm1_bias" -> "2910 layer_norm_39" [label="(384,)", style=solid]; -"708 _scale_125" -> "711 dequantize_per_channel_default_125" [label="(1536,)", style=solid]; -"709 _zero_point_125" -> "711 dequantize_per_channel_default_125" [label="(1536,)", style=solid]; -"710 _frozen_param125" -> "711 dequantize_per_channel_default_125" [label="(1536, 384)", style=solid]; -"711 dequantize_per_channel_default_125" -> "2914 linear_114" [label=None, style=solid]; -"712 features_5_14_mlp_0_bias" -> "2914 linear_114" [label="(1536,)", style=solid]; -"713 _scale_126" -> "716 dequantize_per_channel_default_126" [label="(384,)", style=solid]; -"714 _zero_point_126" -> "716 dequantize_per_channel_default_126" [label="(384,)", style=solid]; -"715 _frozen_param126" -> "716 dequantize_per_channel_default_126" [label="(384, 1536)", style=solid]; -"716 dequantize_per_channel_default_126" -> "2919 linear_115" [label=None, style=solid]; -"717 features_5_14_mlp_3_bias" -> "2919 linear_115" [label="(384,)", style=solid]; -"718 features_5_14_norm2_weight" -> "2921 layer_norm_40" [label="(384,)", style=solid]; -"719 features_5_14_norm2_bias" -> "2921 layer_norm_40" [label="(384,)", style=solid]; -"720 _scale_127" -> "723 dequantize_per_channel_default_127" [label="(512,)", style=solid]; -"721 _zero_point_127" -> "723 dequantize_per_channel_default_127" [label="(512,)", style=solid]; -"722 _frozen_param127" -> "723 dequantize_per_channel_default_127" [label="(512, 2)", style=solid]; -"723 dequantize_per_channel_default_127" -> "2923 linear_116" [label=None, style=solid]; -"724 features_5_15_attn_cpb_mlp_0_bias" -> "2923 linear_116" [label="(512,)", style=solid]; -"725 _scale_128" -> "728 dequantize_per_channel_default_128" [label="(12,)", style=solid]; -"726 _zero_point_128" -> "728 dequantize_per_channel_default_128" [label="(12,)", style=solid]; -"727 _frozen_param128" -> "728 dequantize_per_channel_default_128" [label="(12, 512)", style=solid]; -"728 dequantize_per_channel_default_128" -> "2925 linear_117" [label=None, style=solid]; -"729 features_5_15_norm1_weight" -> "3038 layer_norm_41" [label="(384,)", style=solid]; -"730 features_5_15_norm1_bias" -> "3038 layer_norm_41" [label="(384,)", style=solid]; -"731 _scale_129" -> "734 dequantize_per_channel_default_129" [label="(1536,)", style=solid]; -"732 _zero_point_129" -> "734 dequantize_per_channel_default_129" [label="(1536,)", style=solid]; -"733 _frozen_param129" -> "734 dequantize_per_channel_default_129" [label="(1536, 384)", style=solid]; -"734 dequantize_per_channel_default_129" -> "3042 linear_120" [label=None, style=solid]; -"735 features_5_15_mlp_0_bias" -> "3042 linear_120" [label="(1536,)", style=solid]; -"736 _scale_130" -> "739 dequantize_per_channel_default_130" [label="(384,)", style=solid]; -"737 _zero_point_130" -> "739 dequantize_per_channel_default_130" [label="(384,)", style=solid]; -"738 _frozen_param130" -> "739 dequantize_per_channel_default_130" [label="(384, 1536)", style=solid]; -"739 dequantize_per_channel_default_130" -> "3047 linear_121" [label=None, style=solid]; -"740 features_5_15_mlp_3_bias" -> "3047 linear_121" [label="(384,)", style=solid]; -"741 features_5_15_norm2_weight" -> "3049 layer_norm_42" [label="(384,)", style=solid]; -"742 features_5_15_norm2_bias" -> "3049 layer_norm_42" [label="(384,)", style=solid]; -"743 _scale_131" -> "746 dequantize_per_channel_default_131" [label="(512,)", style=solid]; -"744 _zero_point_131" -> "746 dequantize_per_channel_default_131" [label="(512,)", style=solid]; -"745 _frozen_param131" -> "746 dequantize_per_channel_default_131" [label="(512, 2)", style=solid]; -"746 dequantize_per_channel_default_131" -> "3053 linear_122" [label=None, style=solid]; -"747 features_5_16_attn_cpb_mlp_0_bias" -> "3053 linear_122" [label="(512,)", style=solid]; -"748 _scale_132" -> "751 dequantize_per_channel_default_132" [label="(12,)", style=solid]; -"749 _zero_point_132" -> "751 dequantize_per_channel_default_132" [label="(12,)", style=solid]; -"750 _frozen_param132" -> "751 dequantize_per_channel_default_132" [label="(12, 512)", style=solid]; -"751 dequantize_per_channel_default_132" -> "3055 linear_123" [label=None, style=solid]; -"752 features_5_16_norm1_weight" -> "3112 layer_norm_43" [label="(384,)", style=solid]; -"753 features_5_16_norm1_bias" -> "3112 layer_norm_43" [label="(384,)", style=solid]; -"754 _scale_133" -> "757 dequantize_per_channel_default_133" [label="(1536,)", style=solid]; -"755 _zero_point_133" -> "757 dequantize_per_channel_default_133" [label="(1536,)", style=solid]; -"756 _frozen_param133" -> "757 dequantize_per_channel_default_133" [label="(1536, 384)", style=solid]; -"757 dequantize_per_channel_default_133" -> "3116 linear_126" [label=None, style=solid]; -"758 features_5_16_mlp_0_bias" -> "3116 linear_126" [label="(1536,)", style=solid]; -"759 _scale_134" -> "762 dequantize_per_channel_default_134" [label="(384,)", style=solid]; -"760 _zero_point_134" -> "762 dequantize_per_channel_default_134" [label="(384,)", style=solid]; -"761 _frozen_param134" -> "762 dequantize_per_channel_default_134" [label="(384, 1536)", style=solid]; -"762 dequantize_per_channel_default_134" -> "3121 linear_127" [label=None, style=solid]; -"763 features_5_16_mlp_3_bias" -> "3121 linear_127" [label="(384,)", style=solid]; -"764 features_5_16_norm2_weight" -> "3123 layer_norm_44" [label="(384,)", style=solid]; -"765 features_5_16_norm2_bias" -> "3123 layer_norm_44" [label="(384,)", style=solid]; -"766 _scale_135" -> "769 dequantize_per_channel_default_135" [label="(512,)", style=solid]; -"767 _zero_point_135" -> "769 dequantize_per_channel_default_135" [label="(512,)", style=solid]; -"768 _frozen_param135" -> "769 dequantize_per_channel_default_135" [label="(512, 2)", style=solid]; -"769 dequantize_per_channel_default_135" -> "3125 linear_128" [label=None, style=solid]; -"770 features_5_17_attn_cpb_mlp_0_bias" -> "3125 linear_128" [label="(512,)", style=solid]; -"771 _scale_136" -> "774 dequantize_per_channel_default_136" [label="(12,)", style=solid]; -"772 _zero_point_136" -> "774 dequantize_per_channel_default_136" [label="(12,)", style=solid]; -"773 _frozen_param136" -> "774 dequantize_per_channel_default_136" [label="(12, 512)", style=solid]; -"774 dequantize_per_channel_default_136" -> "3127 linear_129" [label=None, style=solid]; -"775 features_5_17_norm1_weight" -> "3240 layer_norm_45" [label="(384,)", style=solid]; -"776 features_5_17_norm1_bias" -> "3240 layer_norm_45" [label="(384,)", style=solid]; -"777 _scale_137" -> "780 dequantize_per_channel_default_137" [label="(1536,)", style=solid]; -"778 _zero_point_137" -> "780 dequantize_per_channel_default_137" [label="(1536,)", style=solid]; -"779 _frozen_param137" -> "780 dequantize_per_channel_default_137" [label="(1536, 384)", style=solid]; -"780 dequantize_per_channel_default_137" -> "3244 linear_132" [label=None, style=solid]; -"781 features_5_17_mlp_0_bias" -> "3244 linear_132" [label="(1536,)", style=solid]; -"782 _scale_138" -> "785 dequantize_per_channel_default_138" [label="(384,)", style=solid]; -"783 _zero_point_138" -> "785 dequantize_per_channel_default_138" [label="(384,)", style=solid]; -"784 _frozen_param138" -> "785 dequantize_per_channel_default_138" [label="(384, 1536)", style=solid]; -"785 dequantize_per_channel_default_138" -> "3249 linear_133" [label=None, style=solid]; -"786 features_5_17_mlp_3_bias" -> "3249 linear_133" [label="(384,)", style=solid]; -"787 features_5_17_norm2_weight" -> "3251 layer_norm_46" [label="(384,)", style=solid]; -"788 features_5_17_norm2_bias" -> "3251 layer_norm_46" [label="(384,)", style=solid]; -"789 _scale_139" -> "792 dequantize_per_channel_default_139" [label="(768,)", style=solid]; -"790 _zero_point_139" -> "792 dequantize_per_channel_default_139" [label="(768,)", style=solid]; -"791 _frozen_param139" -> "792 dequantize_per_channel_default_139" [label="(768, 1536)", style=solid]; -"792 dequantize_per_channel_default_139" -> "3269 linear_134" [label=None, style=solid]; -"793 features_6_norm_weight" -> "3270 layer_norm_47" [label="(768,)", style=solid]; -"794 features_6_norm_bias" -> "3270 layer_norm_47" [label="(768,)", style=solid]; -"795 _scale_140" -> "798 dequantize_per_channel_default_140" [label="(512,)", style=solid]; -"796 _zero_point_140" -> "798 dequantize_per_channel_default_140" [label="(512,)", style=solid]; -"797 _frozen_param140" -> "798 dequantize_per_channel_default_140" [label="(512, 2)", style=solid]; -"798 dequantize_per_channel_default_140" -> "3273 linear_135" [label=None, style=solid]; -"799 features_7_0_attn_cpb_mlp_0_bias" -> "3273 linear_135" [label="(512,)", style=solid]; -"800 _scale_141" -> "803 dequantize_per_channel_default_141" [label="(24,)", style=solid]; -"801 _zero_point_141" -> "803 dequantize_per_channel_default_141" [label="(24,)", style=solid]; -"802 _frozen_param141" -> "803 dequantize_per_channel_default_141" [label="(24, 512)", style=solid]; -"803 dequantize_per_channel_default_141" -> "3275 linear_136" [label=None, style=solid]; -"804 features_7_0_norm1_weight" -> "3332 layer_norm_48" [label="(768,)", style=solid]; -"805 features_7_0_norm1_bias" -> "3332 layer_norm_48" [label="(768,)", style=solid]; -"806 _scale_142" -> "809 dequantize_per_channel_default_142" [label="(3072,)", style=solid]; -"807 _zero_point_142" -> "809 dequantize_per_channel_default_142" [label="(3072,)", style=solid]; -"808 _frozen_param142" -> "809 dequantize_per_channel_default_142" [label="(3072, 768)", style=solid]; -"809 dequantize_per_channel_default_142" -> "3336 linear_139" [label=None, style=solid]; -"810 features_7_0_mlp_0_bias" -> "3336 linear_139" [label="(3072,)", style=solid]; -"811 _scale_143" -> "814 dequantize_per_channel_default_143" [label="(768,)", style=solid]; -"812 _zero_point_143" -> "814 dequantize_per_channel_default_143" [label="(768,)", style=solid]; -"813 _frozen_param143" -> "814 dequantize_per_channel_default_143" [label="(768, 3072)", style=solid]; -"814 dequantize_per_channel_default_143" -> "3341 linear_140" [label=None, style=solid]; -"815 features_7_0_mlp_3_bias" -> "3341 linear_140" [label="(768,)", style=solid]; -"816 features_7_0_norm2_weight" -> "3343 layer_norm_49" [label="(768,)", style=solid]; -"817 features_7_0_norm2_bias" -> "3343 layer_norm_49" [label="(768,)", style=solid]; -"818 _scale_144" -> "821 dequantize_per_channel_default_144" [label="(512,)", style=solid]; -"819 _zero_point_144" -> "821 dequantize_per_channel_default_144" [label="(512,)", style=solid]; -"820 _frozen_param144" -> "821 dequantize_per_channel_default_144" [label="(512, 2)", style=solid]; -"821 dequantize_per_channel_default_144" -> "3347 linear_141" [label=None, style=solid]; -"822 features_7_1_attn_cpb_mlp_0_bias" -> "3347 linear_141" [label="(512,)", style=solid]; -"823 _scale_145" -> "826 dequantize_per_channel_default_145" [label="(24,)", style=solid]; -"824 _zero_point_145" -> "826 dequantize_per_channel_default_145" [label="(24,)", style=solid]; -"825 _frozen_param145" -> "826 dequantize_per_channel_default_145" [label="(24, 512)", style=solid]; -"826 dequantize_per_channel_default_145" -> "3349 linear_142" [label=None, style=solid]; -"827 features_7_1_norm1_weight" -> "3406 layer_norm_50" [label="(768,)", style=solid]; -"828 features_7_1_norm1_bias" -> "3406 layer_norm_50" [label="(768,)", style=solid]; -"829 _scale_146" -> "832 dequantize_per_channel_default_146" [label="(3072,)", style=solid]; -"830 _zero_point_146" -> "832 dequantize_per_channel_default_146" [label="(3072,)", style=solid]; -"831 _frozen_param146" -> "832 dequantize_per_channel_default_146" [label="(3072, 768)", style=solid]; -"832 dequantize_per_channel_default_146" -> "3410 linear_145" [label=None, style=solid]; -"833 features_7_1_mlp_0_bias" -> "3410 linear_145" [label="(3072,)", style=solid]; -"834 _scale_147" -> "837 dequantize_per_channel_default_147" [label="(768,)", style=solid]; -"835 _zero_point_147" -> "837 dequantize_per_channel_default_147" [label="(768,)", style=solid]; -"836 _frozen_param147" -> "837 dequantize_per_channel_default_147" [label="(768, 3072)", style=solid]; -"837 dequantize_per_channel_default_147" -> "3415 linear_146" [label=None, style=solid]; -"838 features_7_1_mlp_3_bias" -> "3415 linear_146" [label="(768,)", style=solid]; -"839 features_7_1_norm2_weight" -> "3417 layer_norm_51" [label="(768,)", style=solid]; -"840 features_7_1_norm2_bias" -> "3417 layer_norm_51" [label="(768,)", style=solid]; -"841 norm_weight" -> "3419 layer_norm_52" [label="(768,)", style=solid]; -"842 norm_bias" -> "3419 layer_norm_52" [label="(768,)", style=solid]; -"843 _scale_148" -> "846 dequantize_per_channel_default_148" [label="(1000,)", style=solid]; -"844 _zero_point_148" -> "846 dequantize_per_channel_default_148" [label="(1000,)", style=solid]; -"845 _frozen_param148" -> "846 dequantize_per_channel_default_148" [label="(1000, 768)", style=solid]; -"846 dequantize_per_channel_default_148" -> "3425 linear_147" [label=None, style=solid]; -"847 head_bias" -> "3425 linear_147" [label="(1000,)", style=solid]; -"848 features_1_0_attn_relative_coords_table" -> "1003 linear" [label="(1, 15, 15, 2)", style=solid]; -"849 features_1_0_attn_relative_position_index" -> "1007 index" [label="(4096,)", style=solid]; -"850 features_1_1_attn_relative_coords_table" -> "1072 linear_6" [label="(1, 15, 15, 2)", style=solid]; -"851 features_1_1_attn_relative_position_index" -> "1076 index_1" [label="(4096,)", style=solid]; -"852 features_3_0_attn_relative_coords_table" -> "1217 linear_13" [label="(1, 15, 15, 2)", style=solid]; -"853 features_3_0_attn_relative_position_index" -> "1221 index_2" [label="(4096,)", style=solid]; -"854 features_3_1_attn_relative_coords_table" -> "1289 linear_19" [label="(1, 15, 15, 2)", style=solid]; -"855 features_3_1_attn_relative_position_index" -> "1293 index_3" [label="(4096,)", style=solid]; -"856 features_5_0_attn_relative_coords_table" -> "1437 linear_26" [label="(1, 15, 15, 2)", style=solid]; -"857 features_5_0_attn_relative_position_index" -> "1441 index_4" [label="(4096,)", style=solid]; -"858 features_5_1_attn_relative_coords_table" -> "1509 linear_32" [label="(1, 15, 15, 2)", style=solid]; -"859 features_5_1_attn_relative_position_index" -> "1513 index_5" [label="(4096,)", style=solid]; -"860 features_5_2_attn_relative_coords_table" -> "1639 linear_38" [label="(1, 15, 15, 2)", style=solid]; -"861 features_5_2_attn_relative_position_index" -> "1643 index_6" [label="(4096,)", style=solid]; -"862 features_5_3_attn_relative_coords_table" -> "1711 linear_44" [label="(1, 15, 15, 2)", style=solid]; -"863 features_5_3_attn_relative_position_index" -> "1715 index_7" [label="(4096,)", style=solid]; -"864 features_5_4_attn_relative_coords_table" -> "1841 linear_50" [label="(1, 15, 15, 2)", style=solid]; -"865 features_5_4_attn_relative_position_index" -> "1845 index_8" [label="(4096,)", style=solid]; -"866 features_5_5_attn_relative_coords_table" -> "1913 linear_56" [label="(1, 15, 15, 2)", style=solid]; -"867 features_5_5_attn_relative_position_index" -> "1917 index_9" [label="(4096,)", style=solid]; -"868 features_5_6_attn_relative_coords_table" -> "2043 linear_62" [label="(1, 15, 15, 2)", style=solid]; -"869 features_5_6_attn_relative_position_index" -> "2047 index_10" [label="(4096,)", style=solid]; -"870 features_5_7_attn_relative_coords_table" -> "2115 linear_68" [label="(1, 15, 15, 2)", style=solid]; -"871 features_5_7_attn_relative_position_index" -> "2119 index_11" [label="(4096,)", style=solid]; -"872 features_5_8_attn_relative_coords_table" -> "2245 linear_74" [label="(1, 15, 15, 2)", style=solid]; -"873 features_5_8_attn_relative_position_index" -> "2249 index_12" [label="(4096,)", style=solid]; -"874 features_5_9_attn_relative_coords_table" -> "2317 linear_80" [label="(1, 15, 15, 2)", style=solid]; -"875 features_5_9_attn_relative_position_index" -> "2321 index_13" [label="(4096,)", style=solid]; -"876 features_5_10_attn_relative_coords_table" -> "2447 linear_86" [label="(1, 15, 15, 2)", style=solid]; -"877 features_5_10_attn_relative_position_index" -> "2451 index_14" [label="(4096,)", style=solid]; -"878 features_5_11_attn_relative_coords_table" -> "2519 linear_92" [label="(1, 15, 15, 2)", style=solid]; -"879 features_5_11_attn_relative_position_index" -> "2523 index_15" [label="(4096,)", style=solid]; -"880 features_5_12_attn_relative_coords_table" -> "2649 linear_98" [label="(1, 15, 15, 2)", style=solid]; -"881 features_5_12_attn_relative_position_index" -> "2653 index_16" [label="(4096,)", style=solid]; -"882 features_5_13_attn_relative_coords_table" -> "2721 linear_104" [label="(1, 15, 15, 2)", style=solid]; -"883 features_5_13_attn_relative_position_index" -> "2725 index_17" [label="(4096,)", style=solid]; -"884 features_5_14_attn_relative_coords_table" -> "2851 linear_110" [label="(1, 15, 15, 2)", style=solid]; -"885 features_5_14_attn_relative_position_index" -> "2855 index_18" [label="(4096,)", style=solid]; -"886 features_5_15_attn_relative_coords_table" -> "2923 linear_116" [label="(1, 15, 15, 2)", style=solid]; -"887 features_5_15_attn_relative_position_index" -> "2927 index_19" [label="(4096,)", style=solid]; -"888 features_5_16_attn_relative_coords_table" -> "3053 linear_122" [label="(1, 15, 15, 2)", style=solid]; -"889 features_5_16_attn_relative_position_index" -> "3057 index_20" [label="(4096,)", style=solid]; -"890 features_5_17_attn_relative_coords_table" -> "3125 linear_128" [label="(1, 15, 15, 2)", style=solid]; -"891 features_5_17_attn_relative_position_index" -> "3129 index_21" [label="(4096,)", style=solid]; -"892 features_7_0_attn_relative_coords_table" -> "3273 linear_135" [label="(1, 15, 15, 2)", style=solid]; -"893 features_7_0_attn_relative_position_index" -> "3277 index_22" [label="(4096,)", style=solid]; -"894 features_7_1_attn_relative_coords_table" -> "3347 linear_141" [label="(1, 15, 15, 2)", style=solid]; -"895 features_7_1_attn_relative_position_index" -> "3351 index_23" [label="(4096,)", style=solid]; -"896 features_1_1_attn_lifted_tensor_0" -> "1118 lift_fresh_copy" [label="()", style=solid]; -"897 features_1_1_attn_lifted_tensor_1" -> "1122 lift_fresh_copy_1" [label="()", style=solid]; -"898 features_1_1_attn_lifted_tensor_2" -> "1126 lift_fresh_copy_2" [label="()", style=solid]; -"899 features_1_1_attn_lifted_tensor_3" -> "1130 lift_fresh_copy_3" [label="()", style=solid]; -"900 features_1_1_attn_lifted_tensor_4" -> "1134 lift_fresh_copy_4" [label="()", style=solid]; -"901 features_1_1_attn_lifted_tensor_5" -> "1138 lift_fresh_copy_5" [label="()", style=solid]; -"902 features_1_1_attn_lifted_tensor_6" -> "1142 lift_fresh_copy_6" [label="()", style=solid]; -"903 features_1_1_attn_lifted_tensor_7" -> "1146 lift_fresh_copy_7" [label="()", style=solid]; -"904 features_1_1_attn_lifted_tensor_8" -> "1150 lift_fresh_copy_8" [label="()", style=solid]; -"905 features_3_1_attn_lifted_tensor_9" -> "1335 lift_fresh_copy_9" [label="()", style=solid]; -"906 features_3_1_attn_lifted_tensor_10" -> "1339 lift_fresh_copy_10" [label="()", style=solid]; -"907 features_3_1_attn_lifted_tensor_11" -> "1343 lift_fresh_copy_11" [label="()", style=solid]; -"908 features_3_1_attn_lifted_tensor_12" -> "1347 lift_fresh_copy_12" [label="()", style=solid]; -"909 features_3_1_attn_lifted_tensor_13" -> "1351 lift_fresh_copy_13" [label="()", style=solid]; -"910 features_3_1_attn_lifted_tensor_14" -> "1355 lift_fresh_copy_14" [label="()", style=solid]; -"911 features_3_1_attn_lifted_tensor_15" -> "1359 lift_fresh_copy_15" [label="()", style=solid]; -"912 features_3_1_attn_lifted_tensor_16" -> "1363 lift_fresh_copy_16" [label="()", style=solid]; -"913 features_3_1_attn_lifted_tensor_17" -> "1367 lift_fresh_copy_17" [label="()", style=solid]; -"914 features_5_1_attn_lifted_tensor_18" -> "1555 lift_fresh_copy_18" [label="()", style=solid]; -"915 features_5_1_attn_lifted_tensor_19" -> "1559 lift_fresh_copy_19" [label="()", style=solid]; -"916 features_5_1_attn_lifted_tensor_20" -> "1563 lift_fresh_copy_20" [label="()", style=solid]; -"917 features_5_1_attn_lifted_tensor_21" -> "1567 lift_fresh_copy_21" [label="()", style=solid]; -"918 features_5_1_attn_lifted_tensor_22" -> "1571 lift_fresh_copy_22" [label="()", style=solid]; -"919 features_5_1_attn_lifted_tensor_23" -> "1575 lift_fresh_copy_23" [label="()", style=solid]; -"920 features_5_1_attn_lifted_tensor_24" -> "1579 lift_fresh_copy_24" [label="()", style=solid]; -"921 features_5_1_attn_lifted_tensor_25" -> "1583 lift_fresh_copy_25" [label="()", style=solid]; -"922 features_5_1_attn_lifted_tensor_26" -> "1587 lift_fresh_copy_26" [label="()", style=solid]; -"923 features_5_3_attn_lifted_tensor_27" -> "1757 lift_fresh_copy_27" [label="()", style=solid]; -"924 features_5_3_attn_lifted_tensor_28" -> "1761 lift_fresh_copy_28" [label="()", style=solid]; -"925 features_5_3_attn_lifted_tensor_29" -> "1765 lift_fresh_copy_29" [label="()", style=solid]; -"926 features_5_3_attn_lifted_tensor_30" -> "1769 lift_fresh_copy_30" [label="()", style=solid]; -"927 features_5_3_attn_lifted_tensor_31" -> "1773 lift_fresh_copy_31" [label="()", style=solid]; -"928 features_5_3_attn_lifted_tensor_32" -> "1777 lift_fresh_copy_32" [label="()", style=solid]; -"929 features_5_3_attn_lifted_tensor_33" -> "1781 lift_fresh_copy_33" [label="()", style=solid]; -"930 features_5_3_attn_lifted_tensor_34" -> "1785 lift_fresh_copy_34" [label="()", style=solid]; -"931 features_5_3_attn_lifted_tensor_35" -> "1789 lift_fresh_copy_35" [label="()", style=solid]; -"932 features_5_5_attn_lifted_tensor_36" -> "1959 lift_fresh_copy_36" [label="()", style=solid]; -"933 features_5_5_attn_lifted_tensor_37" -> "1963 lift_fresh_copy_37" [label="()", style=solid]; -"934 features_5_5_attn_lifted_tensor_38" -> "1967 lift_fresh_copy_38" [label="()", style=solid]; -"935 features_5_5_attn_lifted_tensor_39" -> "1971 lift_fresh_copy_39" [label="()", style=solid]; -"936 features_5_5_attn_lifted_tensor_40" -> "1975 lift_fresh_copy_40" [label="()", style=solid]; -"937 features_5_5_attn_lifted_tensor_41" -> "1979 lift_fresh_copy_41" [label="()", style=solid]; -"938 features_5_5_attn_lifted_tensor_42" -> "1983 lift_fresh_copy_42" [label="()", style=solid]; -"939 features_5_5_attn_lifted_tensor_43" -> "1987 lift_fresh_copy_43" [label="()", style=solid]; -"940 features_5_5_attn_lifted_tensor_44" -> "1991 lift_fresh_copy_44" [label="()", style=solid]; -"941 features_5_7_attn_lifted_tensor_45" -> "2161 lift_fresh_copy_45" [label="()", style=solid]; -"942 features_5_7_attn_lifted_tensor_46" -> "2165 lift_fresh_copy_46" [label="()", style=solid]; -"943 features_5_7_attn_lifted_tensor_47" -> "2169 lift_fresh_copy_47" [label="()", style=solid]; -"944 features_5_7_attn_lifted_tensor_48" -> "2173 lift_fresh_copy_48" [label="()", style=solid]; -"945 features_5_7_attn_lifted_tensor_49" -> "2177 lift_fresh_copy_49" [label="()", style=solid]; -"946 features_5_7_attn_lifted_tensor_50" -> "2181 lift_fresh_copy_50" [label="()", style=solid]; -"947 features_5_7_attn_lifted_tensor_51" -> "2185 lift_fresh_copy_51" [label="()", style=solid]; -"948 features_5_7_attn_lifted_tensor_52" -> "2189 lift_fresh_copy_52" [label="()", style=solid]; -"949 features_5_7_attn_lifted_tensor_53" -> "2193 lift_fresh_copy_53" [label="()", style=solid]; -"950 features_5_9_attn_lifted_tensor_54" -> "2363 lift_fresh_copy_54" [label="()", style=solid]; -"951 features_5_9_attn_lifted_tensor_55" -> "2367 lift_fresh_copy_55" [label="()", style=solid]; -"952 features_5_9_attn_lifted_tensor_56" -> "2371 lift_fresh_copy_56" [label="()", style=solid]; -"953 features_5_9_attn_lifted_tensor_57" -> "2375 lift_fresh_copy_57" [label="()", style=solid]; -"954 features_5_9_attn_lifted_tensor_58" -> "2379 lift_fresh_copy_58" [label="()", style=solid]; -"955 features_5_9_attn_lifted_tensor_59" -> "2383 lift_fresh_copy_59" [label="()", style=solid]; -"956 features_5_9_attn_lifted_tensor_60" -> "2387 lift_fresh_copy_60" [label="()", style=solid]; -"957 features_5_9_attn_lifted_tensor_61" -> "2391 lift_fresh_copy_61" [label="()", style=solid]; -"958 features_5_9_attn_lifted_tensor_62" -> "2395 lift_fresh_copy_62" [label="()", style=solid]; -"959 features_5_11_attn_lifted_tensor_63" -> "2565 lift_fresh_copy_63" [label="()", style=solid]; -"960 features_5_11_attn_lifted_tensor_64" -> "2569 lift_fresh_copy_64" [label="()", style=solid]; -"961 features_5_11_attn_lifted_tensor_65" -> "2573 lift_fresh_copy_65" [label="()", style=solid]; -"962 features_5_11_attn_lifted_tensor_66" -> "2577 lift_fresh_copy_66" [label="()", style=solid]; -"963 features_5_11_attn_lifted_tensor_67" -> "2581 lift_fresh_copy_67" [label="()", style=solid]; -"964 features_5_11_attn_lifted_tensor_68" -> "2585 lift_fresh_copy_68" [label="()", style=solid]; -"965 features_5_11_attn_lifted_tensor_69" -> "2589 lift_fresh_copy_69" [label="()", style=solid]; -"966 features_5_11_attn_lifted_tensor_70" -> "2593 lift_fresh_copy_70" [label="()", style=solid]; -"967 features_5_11_attn_lifted_tensor_71" -> "2597 lift_fresh_copy_71" [label="()", style=solid]; -"968 features_5_13_attn_lifted_tensor_72" -> "2767 lift_fresh_copy_72" [label="()", style=solid]; -"969 features_5_13_attn_lifted_tensor_73" -> "2771 lift_fresh_copy_73" [label="()", style=solid]; -"970 features_5_13_attn_lifted_tensor_74" -> "2775 lift_fresh_copy_74" [label="()", style=solid]; -"971 features_5_13_attn_lifted_tensor_75" -> "2779 lift_fresh_copy_75" [label="()", style=solid]; -"972 features_5_13_attn_lifted_tensor_76" -> "2783 lift_fresh_copy_76" [label="()", style=solid]; -"973 features_5_13_attn_lifted_tensor_77" -> "2787 lift_fresh_copy_77" [label="()", style=solid]; -"974 features_5_13_attn_lifted_tensor_78" -> "2791 lift_fresh_copy_78" [label="()", style=solid]; -"975 features_5_13_attn_lifted_tensor_79" -> "2795 lift_fresh_copy_79" [label="()", style=solid]; -"976 features_5_13_attn_lifted_tensor_80" -> "2799 lift_fresh_copy_80" [label="()", style=solid]; -"977 features_5_15_attn_lifted_tensor_81" -> "2969 lift_fresh_copy_81" [label="()", style=solid]; -"978 features_5_15_attn_lifted_tensor_82" -> "2973 lift_fresh_copy_82" [label="()", style=solid]; -"979 features_5_15_attn_lifted_tensor_83" -> "2977 lift_fresh_copy_83" [label="()", style=solid]; -"980 features_5_15_attn_lifted_tensor_84" -> "2981 lift_fresh_copy_84" [label="()", style=solid]; -"981 features_5_15_attn_lifted_tensor_85" -> "2985 lift_fresh_copy_85" [label="()", style=solid]; -"982 features_5_15_attn_lifted_tensor_86" -> "2989 lift_fresh_copy_86" [label="()", style=solid]; -"983 features_5_15_attn_lifted_tensor_87" -> "2993 lift_fresh_copy_87" [label="()", style=solid]; -"984 features_5_15_attn_lifted_tensor_88" -> "2997 lift_fresh_copy_88" [label="()", style=solid]; -"985 features_5_15_attn_lifted_tensor_89" -> "3001 lift_fresh_copy_89" [label="()", style=solid]; -"986 features_5_17_attn_lifted_tensor_90" -> "3171 lift_fresh_copy_90" [label="()", style=solid]; -"987 features_5_17_attn_lifted_tensor_91" -> "3175 lift_fresh_copy_91" [label="()", style=solid]; -"988 features_5_17_attn_lifted_tensor_92" -> "3179 lift_fresh_copy_92" [label="()", style=solid]; -"989 features_5_17_attn_lifted_tensor_93" -> "3183 lift_fresh_copy_93" [label="()", style=solid]; -"990 features_5_17_attn_lifted_tensor_94" -> "3187 lift_fresh_copy_94" [label="()", style=solid]; -"991 features_5_17_attn_lifted_tensor_95" -> "3191 lift_fresh_copy_95" [label="()", style=solid]; -"992 features_5_17_attn_lifted_tensor_96" -> "3195 lift_fresh_copy_96" [label="()", style=solid]; -"993 features_5_17_attn_lifted_tensor_97" -> "3199 lift_fresh_copy_97" [label="()", style=solid]; -"994 features_5_17_attn_lifted_tensor_98" -> "3203 lift_fresh_copy_98" [label="()", style=solid]; -"995 x" -> "996 quantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; -"996 quantize_per_tensor_default" -> "997 dequantize_per_tensor_default" [label=None, style=solid]; -"997 dequantize_per_tensor_default" -> "998 conv2d" [label=None, style=solid]; -"998 conv2d" -> "999 permute" [label="(1, 96, 56, 56)", style=solid]; -"999 permute" -> "1000 layer_norm" [label="(1, 56, 56, 96)", style=solid]; -"1000 layer_norm" -> "1001 quantize_per_tensor_default_1" [label="(1, 56, 56, 96)", style=solid]; -"1000 layer_norm" -> "1060 add_1" [label="(1, 56, 56, 96)", style=solid]; -"1001 quantize_per_tensor_default_1" -> "1002 dequantize_per_tensor_default_1" [label=None, style=solid]; -"1002 dequantize_per_tensor_default_1" -> "1014 pad" [label=None, style=solid]; -"1003 linear" -> "1004 relu_" [label="(1, 15, 15, 512)", style=solid]; -"1004 relu_" -> "1005 linear_1" [label="(1, 15, 15, 512)", style=solid]; -"1005 linear_1" -> "1006 view" [label="(1, 15, 15, 3)", style=solid]; -"1006 view" -> "1007 index" [label="(225, 3)", style=solid]; -"1007 index" -> "1008 view_1" [label="(4096, 3)", style=solid]; -"1008 view_1" -> "1009 permute_1" [label="(64, 64, 3)", style=solid]; -"1009 permute_1" -> "1010 contiguous" [label="(3, 64, 64)", style=solid]; -"1010 contiguous" -> "1011 unsqueeze" [label="(3, 64, 64)", style=solid]; -"1011 unsqueeze" -> "1012 sigmoid" [label="(1, 3, 64, 64)", style=solid]; -"1012 sigmoid" -> "1013 mul" [label="(1, 3, 64, 64)", style=solid]; -"1013 mul" -> "1044 add" [label="(1, 3, 64, 64)", style=solid]; -"1014 pad" -> "1015 view_2" [label="(1, 56, 56, 96)", style=solid]; -"1015 view_2" -> "1016 permute_2" [label="(1, 7, 8, 7, 8, 96)", style=solid]; -"1016 permute_2" -> "1017 reshape" [label="(1, 7, 7, 8, 8, 96)", style=solid]; -"1017 reshape" -> "1021 linear_2" [label="(49, 64, 96)", style=solid]; -"1018 clone" -> "1019 slice_1" [label="(288,)", style=solid]; -"1018 clone" -> "1021 linear_2" [label="(288,)", style=solid]; -"1019 slice_1" -> "1020 zero_" [label="(96,)", style=solid]; -"1021 linear_2" -> "1022 reshape_1" [label="(49, 64, 288)", style=solid]; -"1022 reshape_1" -> "1023 permute_3" [label="(49, 64, 3, 3, 32)", style=solid]; -"1023 permute_3" -> "1024 select" [label="(3, 49, 3, 64, 32)", style=solid]; -"1023 permute_3" -> "1025 select_1" [label="(3, 49, 3, 64, 32)", style=solid]; -"1023 permute_3" -> "1026 select_2" [label="(3, 49, 3, 64, 32)", style=solid]; -"1024 select" -> "1027 linalg_vector_norm" [label="(49, 3, 64, 32)", style=solid]; -"1024 select" -> "1029 expand_as" [label="(49, 3, 64, 32)", style=solid]; -"1024 select" -> "1030 div" [label="(49, 3, 64, 32)", style=solid]; -"1025 select_1" -> "1033 linalg_vector_norm_1" [label="(49, 3, 64, 32)", style=solid]; -"1025 select_1" -> "1035 expand_as_1" [label="(49, 3, 64, 32)", style=solid]; -"1025 select_1" -> "1036 div_1" [label="(49, 3, 64, 32)", style=solid]; -"1026 select_2" -> "1047 matmul_1" [label="(49, 3, 64, 32)", style=solid]; -"1027 linalg_vector_norm" -> "1028 clamp_min" [label="(49, 3, 64, 1)", style=solid]; -"1028 clamp_min" -> "1029 expand_as" [label="(49, 3, 64, 1)", style=solid]; -"1029 expand_as" -> "1030 div" [label="(49, 3, 64, 32)", style=solid]; -"1030 div" -> "1031 quantize_per_tensor_default_2" [label="(49, 3, 64, 32)", style=solid]; -"1031 quantize_per_tensor_default_2" -> "1032 dequantize_per_tensor_default_2" [label=None, style=solid]; -"1032 dequantize_per_tensor_default_2" -> "1040 matmul" [label=None, style=solid]; -"1033 linalg_vector_norm_1" -> "1034 clamp_min_1" [label="(49, 3, 64, 1)", style=solid]; -"1034 clamp_min_1" -> "1035 expand_as_1" [label="(49, 3, 64, 1)", style=solid]; -"1035 expand_as_1" -> "1036 div_1" [label="(49, 3, 64, 32)", style=solid]; -"1036 div_1" -> "1037 quantize_per_tensor_default_3" [label="(49, 3, 64, 32)", style=solid]; -"1037 quantize_per_tensor_default_3" -> "1038 dequantize_per_tensor_default_3" [label=None, style=solid]; -"1038 dequantize_per_tensor_default_3" -> "1039 transpose" [label=None, style=solid]; -"1039 transpose" -> "1040 matmul" [label="(49, 3, 32, 64)", style=solid]; -"1040 matmul" -> "1043 mul_1" [label="(49, 3, 64, 64)", style=solid]; -"1041 clamp" -> "1042 exp" [label="(3, 1, 1)", style=solid]; -"1042 exp" -> "1043 mul_1" [label="(3, 1, 1)", style=solid]; -"1043 mul_1" -> "1044 add" [label="(49, 3, 64, 64)", style=solid]; -"1044 add" -> "1045 softmax" [label="(49, 3, 64, 64)", style=solid]; -"1045 softmax" -> "1046 dropout" [label="(49, 3, 64, 64)", style=solid]; -"1046 dropout" -> "1047 matmul_1" [label="(49, 3, 64, 64)", style=solid]; -"1047 matmul_1" -> "1048 quantize_per_tensor_default_4" [label="(49, 3, 64, 32)", style=solid]; -"1048 quantize_per_tensor_default_4" -> "1049 dequantize_per_tensor_default_4" [label=None, style=solid]; -"1049 dequantize_per_tensor_default_4" -> "1050 transpose_1" [label=None, style=solid]; -"1050 transpose_1" -> "1051 reshape_2" [label="(49, 64, 3, 32)", style=solid]; -"1051 reshape_2" -> "1052 linear_3" [label="(49, 64, 96)", style=solid]; -"1052 linear_3" -> "1053 dropout_1" [label="(49, 64, 96)", style=solid]; -"1053 dropout_1" -> "1054 view_3" [label="(49, 64, 96)", style=solid]; -"1054 view_3" -> "1055 permute_4" [label="(1, 7, 7, 8, 8, 96)", style=solid]; -"1055 permute_4" -> "1056 reshape_3" [label="(1, 7, 8, 7, 8, 96)", style=solid]; -"1056 reshape_3" -> "1057 slice_2" [label="(1, 56, 56, 96)", style=solid]; -"1057 slice_2" -> "1058 slice_3" [label="(1, 56, 56, 96)", style=solid]; -"1058 slice_3" -> "1059 layer_norm_1" [label="(1, 56, 56, 96)", style=solid]; -"1059 layer_norm_1" -> "1060 add_1" [label="(1, 56, 56, 96)", style=solid]; -"1060 add_1" -> "1061 quantize_per_tensor_default_5" [label="(1, 56, 56, 96)", style=solid]; -"1060 add_1" -> "1071 add_2" [label="(1, 56, 56, 96)", style=solid]; -"1061 quantize_per_tensor_default_5" -> "1062 dequantize_per_tensor_default_5" [label=None, style=solid]; -"1062 dequantize_per_tensor_default_5" -> "1063 linear_4" [label=None, style=solid]; -"1063 linear_4" -> "1064 gelu" [label="(1, 56, 56, 384)", style=solid]; -"1064 gelu" -> "1065 quantize_per_tensor_default_6" [label="(1, 56, 56, 384)", style=solid]; -"1065 quantize_per_tensor_default_6" -> "1066 dequantize_per_tensor_default_6" [label=None, style=solid]; -"1066 dequantize_per_tensor_default_6" -> "1067 dropout_2" [label=None, style=solid]; -"1067 dropout_2" -> "1068 linear_5" [label="(1, 56, 56, 384)", style=solid]; -"1068 linear_5" -> "1069 dropout_3" [label="(1, 56, 56, 96)", style=solid]; -"1069 dropout_3" -> "1070 layer_norm_2" [label="(1, 56, 56, 96)", style=solid]; -"1070 layer_norm_2" -> "1071 add_2" [label="(1, 56, 56, 96)", style=solid]; -"1071 add_2" -> "1083 pad_1" [label="(1, 56, 56, 96)", style=solid]; -"1071 add_2" -> "1185 add_5" [label="(1, 56, 56, 96)", style=solid]; -"1072 linear_6" -> "1073 relu__1" [label="(1, 15, 15, 512)", style=solid]; -"1073 relu__1" -> "1074 linear_7" [label="(1, 15, 15, 512)", style=solid]; -"1074 linear_7" -> "1075 view_4" [label="(1, 15, 15, 3)", style=solid]; -"1075 view_4" -> "1076 index_1" [label="(225, 3)", style=solid]; -"1076 index_1" -> "1077 view_5" [label="(4096, 3)", style=solid]; -"1077 view_5" -> "1078 permute_5" [label="(64, 64, 3)", style=solid]; -"1078 permute_5" -> "1079 contiguous_1" [label="(3, 64, 64)", style=solid]; -"1079 contiguous_1" -> "1080 unsqueeze_1" [label="(3, 64, 64)", style=solid]; -"1080 unsqueeze_1" -> "1081 sigmoid_1" [label="(1, 3, 64, 64)", style=solid]; -"1081 sigmoid_1" -> "1082 mul_2" [label="(1, 3, 64, 64)", style=solid]; -"1082 mul_2" -> "1116 add_3" [label="(1, 3, 64, 64)", style=solid]; -"1083 pad_1" -> "1084 roll" [label="(1, 56, 56, 96)", style=solid]; -"1084 roll" -> "1085 view_6" [label="(1, 56, 56, 96)", style=solid]; -"1085 view_6" -> "1086 permute_6" [label="(1, 7, 8, 7, 8, 96)", style=solid]; -"1086 permute_6" -> "1087 reshape_4" [label="(1, 7, 7, 8, 8, 96)", style=solid]; -"1087 reshape_4" -> "1088 quantize_per_tensor_default_7" [label="(49, 64, 96)", style=solid]; -"1087 reshape_4" -> "1117 new_zeros" [label="(49, 64, 96)", style=solid]; -"1088 quantize_per_tensor_default_7" -> "1089 dequantize_per_tensor_default_7" [label=None, style=solid]; -"1089 dequantize_per_tensor_default_7" -> "1093 linear_8" [label=None, style=solid]; -"1090 clone_1" -> "1091 slice_4" [label="(288,)", style=solid]; -"1090 clone_1" -> "1093 linear_8" [label="(288,)", style=solid]; -"1091 slice_4" -> "1092 zero__1" [label="(96,)", style=solid]; -"1093 linear_8" -> "1094 reshape_5" [label="(49, 64, 288)", style=solid]; -"1094 reshape_5" -> "1095 permute_7" [label="(49, 64, 3, 3, 32)", style=solid]; -"1095 permute_7" -> "1096 select_3" [label="(3, 49, 3, 64, 32)", style=solid]; -"1095 permute_7" -> "1097 select_4" [label="(3, 49, 3, 64, 32)", style=solid]; -"1095 permute_7" -> "1098 select_5" [label="(3, 49, 3, 64, 32)", style=solid]; -"1096 select_3" -> "1099 linalg_vector_norm_2" [label="(49, 3, 64, 32)", style=solid]; -"1096 select_3" -> "1101 expand_as_2" [label="(49, 3, 64, 32)", style=solid]; -"1096 select_3" -> "1102 div_2" [label="(49, 3, 64, 32)", style=solid]; -"1097 select_4" -> "1105 linalg_vector_norm_3" [label="(49, 3, 64, 32)", style=solid]; -"1097 select_4" -> "1107 expand_as_3" [label="(49, 3, 64, 32)", style=solid]; -"1097 select_4" -> "1108 div_3" [label="(49, 3, 64, 32)", style=solid]; -"1098 select_5" -> "1171 matmul_3" [label="(49, 3, 64, 32)", style=solid]; -"1099 linalg_vector_norm_2" -> "1100 clamp_min_2" [label="(49, 3, 64, 1)", style=solid]; -"1100 clamp_min_2" -> "1101 expand_as_2" [label="(49, 3, 64, 1)", style=solid]; -"1101 expand_as_2" -> "1102 div_2" [label="(49, 3, 64, 32)", style=solid]; -"1102 div_2" -> "1103 quantize_per_tensor_default_8" [label="(49, 3, 64, 32)", style=solid]; -"1103 quantize_per_tensor_default_8" -> "1104 dequantize_per_tensor_default_8" [label=None, style=solid]; -"1104 dequantize_per_tensor_default_8" -> "1112 matmul_2" [label=None, style=solid]; -"1105 linalg_vector_norm_3" -> "1106 clamp_min_3" [label="(49, 3, 64, 1)", style=solid]; -"1106 clamp_min_3" -> "1107 expand_as_3" [label="(49, 3, 64, 1)", style=solid]; -"1107 expand_as_3" -> "1108 div_3" [label="(49, 3, 64, 32)", style=solid]; -"1108 div_3" -> "1109 quantize_per_tensor_default_9" [label="(49, 3, 64, 32)", style=solid]; -"1109 quantize_per_tensor_default_9" -> "1110 dequantize_per_tensor_default_9" [label=None, style=solid]; -"1110 dequantize_per_tensor_default_9" -> "1111 transpose_2" [label=None, style=solid]; -"1111 transpose_2" -> "1112 matmul_2" [label="(49, 3, 32, 64)", style=solid]; -"1112 matmul_2" -> "1115 mul_3" [label="(49, 3, 64, 64)", style=solid]; -"1113 clamp_1" -> "1114 exp_1" [label="(3, 1, 1)", style=solid]; -"1114 exp_1" -> "1115 mul_3" [label="(3, 1, 1)", style=solid]; -"1115 mul_3" -> "1116 add_3" [label="(49, 3, 64, 64)", style=solid]; -"1116 add_3" -> "1164 view_8" [label="(49, 3, 64, 64)", style=solid]; -"1117 new_zeros" -> "1119 slice_5" [label="(56, 56)", style=solid]; -"1117 new_zeros" -> "1123 slice_7" [label="(56, 56)", style=solid]; -"1117 new_zeros" -> "1127 slice_9" [label="(56, 56)", style=solid]; -"1117 new_zeros" -> "1131 slice_11" [label="(56, 56)", style=solid]; -"1117 new_zeros" -> "1135 slice_13" [label="(56, 56)", style=solid]; -"1117 new_zeros" -> "1139 slice_15" [label="(56, 56)", style=solid]; -"1117 new_zeros" -> "1143 slice_17" [label="(56, 56)", style=solid]; -"1117 new_zeros" -> "1147 slice_19" [label="(56, 56)", style=solid]; -"1117 new_zeros" -> "1151 slice_21" [label="(56, 56)", style=solid]; -"1117 new_zeros" -> "1154 view_7" [label="(56, 56)", style=solid]; -"1118 lift_fresh_copy" -> "1121 fill_" [label="()", style=solid]; -"1119 slice_5" -> "1120 slice_6" [label="(48, 56)", style=solid]; -"1120 slice_6" -> "1121 fill_" [label="(48, 48)", style=solid]; -"1122 lift_fresh_copy_1" -> "1125 fill__1" [label="()", style=solid]; -"1123 slice_7" -> "1124 slice_8" [label="(48, 56)", style=solid]; -"1124 slice_8" -> "1125 fill__1" [label="(48, 4)", style=solid]; -"1126 lift_fresh_copy_2" -> "1129 fill__2" [label="()", style=solid]; -"1127 slice_9" -> "1128 slice_10" [label="(48, 56)", style=solid]; -"1128 slice_10" -> "1129 fill__2" [label="(48, 4)", style=solid]; -"1130 lift_fresh_copy_3" -> "1133 fill__3" [label="()", style=solid]; -"1131 slice_11" -> "1132 slice_12" [label="(4, 56)", style=solid]; -"1132 slice_12" -> "1133 fill__3" [label="(4, 48)", style=solid]; -"1134 lift_fresh_copy_4" -> "1137 fill__4" [label="()", style=solid]; -"1135 slice_13" -> "1136 slice_14" [label="(4, 56)", style=solid]; -"1136 slice_14" -> "1137 fill__4" [label="(4, 4)", style=solid]; -"1138 lift_fresh_copy_5" -> "1141 fill__5" [label="()", style=solid]; -"1139 slice_15" -> "1140 slice_16" [label="(4, 56)", style=solid]; -"1140 slice_16" -> "1141 fill__5" [label="(4, 4)", style=solid]; -"1142 lift_fresh_copy_6" -> "1145 fill__6" [label="()", style=solid]; -"1143 slice_17" -> "1144 slice_18" [label="(4, 56)", style=solid]; -"1144 slice_18" -> "1145 fill__6" [label="(4, 48)", style=solid]; -"1146 lift_fresh_copy_7" -> "1149 fill__7" [label="()", style=solid]; -"1147 slice_19" -> "1148 slice_20" [label="(4, 56)", style=solid]; -"1148 slice_20" -> "1149 fill__7" [label="(4, 4)", style=solid]; -"1150 lift_fresh_copy_8" -> "1153 fill__8" [label="()", style=solid]; -"1151 slice_21" -> "1152 slice_22" [label="(4, 56)", style=solid]; -"1152 slice_22" -> "1153 fill__8" [label="(4, 4)", style=solid]; -"1154 view_7" -> "1155 permute_8" [label="(7, 8, 7, 8)", style=solid]; -"1155 permute_8" -> "1156 reshape_6" [label="(7, 7, 8, 8)", style=solid]; -"1156 reshape_6" -> "1157 unsqueeze_2" [label="(49, 64)", style=solid]; -"1156 reshape_6" -> "1158 unsqueeze_3" [label="(49, 64)", style=solid]; -"1157 unsqueeze_2" -> "1159 sub" [label="(49, 1, 64)", style=solid]; -"1158 unsqueeze_3" -> "1159 sub" [label="(49, 64, 1)", style=solid]; -"1159 sub" -> "1160 ne" [label="(49, 64, 64)", style=solid]; -"1159 sub" -> "1161 masked_fill" [label="(49, 64, 64)", style=solid]; -"1159 sub" -> "1162 eq" [label="(49, 64, 64)", style=solid]; -"1160 ne" -> "1161 masked_fill" [label="(49, 64, 64)", style=solid]; -"1161 masked_fill" -> "1163 masked_fill_1" [label="(49, 64, 64)", style=solid]; -"1162 eq" -> "1163 masked_fill_1" [label="(49, 64, 64)", style=solid]; -"1163 masked_fill_1" -> "1165 unsqueeze_4" [label="(49, 64, 64)", style=solid]; -"1164 view_8" -> "1167 add_4" [label="(1, 49, 3, 64, 64)", style=solid]; -"1165 unsqueeze_4" -> "1166 unsqueeze_5" [label="(49, 1, 64, 64)", style=solid]; -"1166 unsqueeze_5" -> "1167 add_4" [label="(1, 49, 1, 64, 64)", style=solid]; -"1167 add_4" -> "1168 view_9" [label="(1, 49, 3, 64, 64)", style=solid]; -"1168 view_9" -> "1169 softmax_1" [label="(49, 3, 64, 64)", style=solid]; -"1169 softmax_1" -> "1170 dropout_4" [label="(49, 3, 64, 64)", style=solid]; -"1170 dropout_4" -> "1171 matmul_3" [label="(49, 3, 64, 64)", style=solid]; -"1171 matmul_3" -> "1172 quantize_per_tensor_default_10" [label="(49, 3, 64, 32)", style=solid]; -"1172 quantize_per_tensor_default_10" -> "1173 dequantize_per_tensor_default_10" [label=None, style=solid]; -"1173 dequantize_per_tensor_default_10" -> "1174 transpose_3" [label=None, style=solid]; -"1174 transpose_3" -> "1175 reshape_7" [label="(49, 64, 3, 32)", style=solid]; -"1175 reshape_7" -> "1176 linear_9" [label="(49, 64, 96)", style=solid]; -"1176 linear_9" -> "1177 dropout_5" [label="(49, 64, 96)", style=solid]; -"1177 dropout_5" -> "1178 view_10" [label="(49, 64, 96)", style=solid]; -"1178 view_10" -> "1179 permute_9" [label="(1, 7, 7, 8, 8, 96)", style=solid]; -"1179 permute_9" -> "1180 reshape_8" [label="(1, 7, 8, 7, 8, 96)", style=solid]; -"1180 reshape_8" -> "1181 roll_1" [label="(1, 56, 56, 96)", style=solid]; -"1181 roll_1" -> "1182 slice_23" [label="(1, 56, 56, 96)", style=solid]; -"1182 slice_23" -> "1183 slice_24" [label="(1, 56, 56, 96)", style=solid]; -"1183 slice_24" -> "1184 layer_norm_3" [label="(1, 56, 56, 96)", style=solid]; -"1184 layer_norm_3" -> "1185 add_5" [label="(1, 56, 56, 96)", style=solid]; -"1185 add_5" -> "1186 quantize_per_tensor_default_11" [label="(1, 56, 56, 96)", style=solid]; -"1185 add_5" -> "1196 add_6" [label="(1, 56, 56, 96)", style=solid]; -"1186 quantize_per_tensor_default_11" -> "1187 dequantize_per_tensor_default_11" [label=None, style=solid]; -"1187 dequantize_per_tensor_default_11" -> "1188 linear_10" [label=None, style=solid]; -"1188 linear_10" -> "1189 gelu_1" [label="(1, 56, 56, 384)", style=solid]; -"1189 gelu_1" -> "1190 quantize_per_tensor_default_12" [label="(1, 56, 56, 384)", style=solid]; -"1190 quantize_per_tensor_default_12" -> "1191 dequantize_per_tensor_default_12" [label=None, style=solid]; -"1191 dequantize_per_tensor_default_12" -> "1192 dropout_6" [label=None, style=solid]; -"1192 dropout_6" -> "1193 linear_11" [label="(1, 56, 56, 384)", style=solid]; -"1193 linear_11" -> "1194 dropout_7" [label="(1, 56, 56, 96)", style=solid]; -"1194 dropout_7" -> "1195 layer_norm_4" [label="(1, 56, 56, 96)", style=solid]; -"1195 layer_norm_4" -> "1196 add_6" [label="(1, 56, 56, 96)", style=solid]; -"1196 add_6" -> "1197 quantize_per_tensor_default_13" [label="(1, 56, 56, 96)", style=solid]; -"1197 quantize_per_tensor_default_13" -> "1198 dequantize_per_tensor_default_13" [label=None, style=solid]; -"1198 dequantize_per_tensor_default_13" -> "1199 pad_2" [label=None, style=solid]; -"1199 pad_2" -> "1200 slice_25" [label="(1, 56, 56, 96)", style=solid]; -"1199 pad_2" -> "1203 slice_28" [label="(1, 56, 56, 96)", style=solid]; -"1199 pad_2" -> "1206 slice_31" [label="(1, 56, 56, 96)", style=solid]; -"1199 pad_2" -> "1209 slice_34" [label="(1, 56, 56, 96)", style=solid]; -"1200 slice_25" -> "1201 slice_26" [label="(1, 28, 56, 96)", style=solid]; -"1201 slice_26" -> "1202 slice_27" [label="(1, 28, 28, 96)", style=solid]; -"1202 slice_27" -> "1212 cat" [label="(1, 28, 28, 96)", style=solid]; -"1203 slice_28" -> "1204 slice_29" [label="(1, 28, 56, 96)", style=solid]; -"1204 slice_29" -> "1205 slice_30" [label="(1, 28, 28, 96)", style=solid]; -"1205 slice_30" -> "1212 cat" [label="(1, 28, 28, 96)", style=solid]; -"1206 slice_31" -> "1207 slice_32" [label="(1, 28, 56, 96)", style=solid]; -"1207 slice_32" -> "1208 slice_33" [label="(1, 28, 28, 96)", style=solid]; -"1208 slice_33" -> "1212 cat" [label="(1, 28, 28, 96)", style=solid]; -"1209 slice_34" -> "1210 slice_35" [label="(1, 28, 56, 96)", style=solid]; -"1210 slice_35" -> "1211 slice_36" [label="(1, 28, 28, 96)", style=solid]; -"1211 slice_36" -> "1212 cat" [label="(1, 28, 28, 96)", style=solid]; -"1212 cat" -> "1213 linear_12" [label="(1, 28, 28, 384)", style=solid]; -"1213 linear_12" -> "1214 layer_norm_5" [label="(1, 28, 28, 192)", style=solid]; -"1214 layer_norm_5" -> "1215 quantize_per_tensor_default_14" [label="(1, 28, 28, 192)", style=solid]; -"1214 layer_norm_5" -> "1277 add_8" [label="(1, 28, 28, 192)", style=solid]; -"1215 quantize_per_tensor_default_14" -> "1216 dequantize_per_tensor_default_14" [label=None, style=solid]; -"1216 dequantize_per_tensor_default_14" -> "1228 pad_3" [label=None, style=solid]; -"1217 linear_13" -> "1218 relu__2" [label="(1, 15, 15, 512)", style=solid]; -"1218 relu__2" -> "1219 linear_14" [label="(1, 15, 15, 512)", style=solid]; -"1219 linear_14" -> "1220 view_11" [label="(1, 15, 15, 6)", style=solid]; -"1220 view_11" -> "1221 index_2" [label="(225, 6)", style=solid]; -"1221 index_2" -> "1222 view_12" [label="(4096, 6)", style=solid]; -"1222 view_12" -> "1223 permute_10" [label="(64, 64, 6)", style=solid]; -"1223 permute_10" -> "1224 contiguous_2" [label="(6, 64, 64)", style=solid]; -"1224 contiguous_2" -> "1225 unsqueeze_6" [label="(6, 64, 64)", style=solid]; -"1225 unsqueeze_6" -> "1226 sigmoid_2" [label="(1, 6, 64, 64)", style=solid]; -"1226 sigmoid_2" -> "1227 mul_4" [label="(1, 6, 64, 64)", style=solid]; -"1227 mul_4" -> "1258 add_7" [label="(1, 6, 64, 64)", style=solid]; -"1228 pad_3" -> "1229 view_13" [label="(1, 32, 32, 192)", style=solid]; -"1229 view_13" -> "1230 permute_11" [label="(1, 4, 8, 4, 8, 192)", style=solid]; -"1230 permute_11" -> "1231 reshape_9" [label="(1, 4, 4, 8, 8, 192)", style=solid]; -"1231 reshape_9" -> "1235 linear_15" [label="(16, 64, 192)", style=solid]; -"1232 clone_2" -> "1233 slice_37" [label="(576,)", style=solid]; -"1232 clone_2" -> "1235 linear_15" [label="(576,)", style=solid]; -"1233 slice_37" -> "1234 zero__2" [label="(192,)", style=solid]; -"1235 linear_15" -> "1236 reshape_10" [label="(16, 64, 576)", style=solid]; -"1236 reshape_10" -> "1237 permute_12" [label="(16, 64, 3, 6, 32)", style=solid]; -"1237 permute_12" -> "1238 select_6" [label="(3, 16, 6, 64, 32)", style=solid]; -"1237 permute_12" -> "1239 select_7" [label="(3, 16, 6, 64, 32)", style=solid]; -"1237 permute_12" -> "1240 select_8" [label="(3, 16, 6, 64, 32)", style=solid]; -"1238 select_6" -> "1241 linalg_vector_norm_4" [label="(16, 6, 64, 32)", style=solid]; -"1238 select_6" -> "1243 expand_as_4" [label="(16, 6, 64, 32)", style=solid]; -"1238 select_6" -> "1244 div_4" [label="(16, 6, 64, 32)", style=solid]; -"1239 select_7" -> "1247 linalg_vector_norm_5" [label="(16, 6, 64, 32)", style=solid]; -"1239 select_7" -> "1249 expand_as_5" [label="(16, 6, 64, 32)", style=solid]; -"1239 select_7" -> "1250 div_5" [label="(16, 6, 64, 32)", style=solid]; -"1240 select_8" -> "1261 matmul_5" [label="(16, 6, 64, 32)", style=solid]; -"1241 linalg_vector_norm_4" -> "1242 clamp_min_4" [label="(16, 6, 64, 1)", style=solid]; -"1242 clamp_min_4" -> "1243 expand_as_4" [label="(16, 6, 64, 1)", style=solid]; -"1243 expand_as_4" -> "1244 div_4" [label="(16, 6, 64, 32)", style=solid]; -"1244 div_4" -> "1245 quantize_per_tensor_default_15" [label="(16, 6, 64, 32)", style=solid]; -"1245 quantize_per_tensor_default_15" -> "1246 dequantize_per_tensor_default_15" [label=None, style=solid]; -"1246 dequantize_per_tensor_default_15" -> "1254 matmul_4" [label=None, style=solid]; -"1247 linalg_vector_norm_5" -> "1248 clamp_min_5" [label="(16, 6, 64, 1)", style=solid]; -"1248 clamp_min_5" -> "1249 expand_as_5" [label="(16, 6, 64, 1)", style=solid]; -"1249 expand_as_5" -> "1250 div_5" [label="(16, 6, 64, 32)", style=solid]; -"1250 div_5" -> "1251 quantize_per_tensor_default_16" [label="(16, 6, 64, 32)", style=solid]; -"1251 quantize_per_tensor_default_16" -> "1252 dequantize_per_tensor_default_16" [label=None, style=solid]; -"1252 dequantize_per_tensor_default_16" -> "1253 transpose_4" [label=None, style=solid]; -"1253 transpose_4" -> "1254 matmul_4" [label="(16, 6, 32, 64)", style=solid]; -"1254 matmul_4" -> "1257 mul_5" [label="(16, 6, 64, 64)", style=solid]; -"1255 clamp_2" -> "1256 exp_2" [label="(6, 1, 1)", style=solid]; -"1256 exp_2" -> "1257 mul_5" [label="(6, 1, 1)", style=solid]; -"1257 mul_5" -> "1258 add_7" [label="(16, 6, 64, 64)", style=solid]; -"1258 add_7" -> "1259 softmax_2" [label="(16, 6, 64, 64)", style=solid]; -"1259 softmax_2" -> "1260 dropout_8" [label="(16, 6, 64, 64)", style=solid]; -"1260 dropout_8" -> "1261 matmul_5" [label="(16, 6, 64, 64)", style=solid]; -"1261 matmul_5" -> "1262 quantize_per_tensor_default_17" [label="(16, 6, 64, 32)", style=solid]; -"1262 quantize_per_tensor_default_17" -> "1263 dequantize_per_tensor_default_17" [label=None, style=solid]; -"1263 dequantize_per_tensor_default_17" -> "1264 transpose_5" [label=None, style=solid]; -"1264 transpose_5" -> "1265 reshape_11" [label="(16, 64, 6, 32)", style=solid]; -"1265 reshape_11" -> "1266 linear_16" [label="(16, 64, 192)", style=solid]; -"1266 linear_16" -> "1267 dropout_9" [label="(16, 64, 192)", style=solid]; -"1267 dropout_9" -> "1268 view_14" [label="(16, 64, 192)", style=solid]; -"1268 view_14" -> "1269 permute_13" [label="(1, 4, 4, 8, 8, 192)", style=solid]; -"1269 permute_13" -> "1270 reshape_12" [label="(1, 4, 8, 4, 8, 192)", style=solid]; -"1270 reshape_12" -> "1271 slice_38" [label="(1, 32, 32, 192)", style=solid]; -"1271 slice_38" -> "1272 slice_39" [label="(1, 32, 32, 192)", style=solid]; -"1272 slice_39" -> "1273 slice_40" [label="(1, 28, 32, 192)", style=solid]; -"1273 slice_40" -> "1274 slice_41" [label="(1, 28, 28, 192)", style=solid]; -"1274 slice_41" -> "1275 contiguous_3" [label="(1, 28, 28, 192)", style=solid]; -"1275 contiguous_3" -> "1276 layer_norm_6" [label="(1, 28, 28, 192)", style=solid]; -"1276 layer_norm_6" -> "1277 add_8" [label="(1, 28, 28, 192)", style=solid]; -"1277 add_8" -> "1278 quantize_per_tensor_default_18" [label="(1, 28, 28, 192)", style=solid]; -"1277 add_8" -> "1288 add_9" [label="(1, 28, 28, 192)", style=solid]; -"1278 quantize_per_tensor_default_18" -> "1279 dequantize_per_tensor_default_18" [label=None, style=solid]; -"1279 dequantize_per_tensor_default_18" -> "1280 linear_17" [label=None, style=solid]; -"1280 linear_17" -> "1281 gelu_2" [label="(1, 28, 28, 768)", style=solid]; -"1281 gelu_2" -> "1282 quantize_per_tensor_default_19" [label="(1, 28, 28, 768)", style=solid]; -"1282 quantize_per_tensor_default_19" -> "1283 dequantize_per_tensor_default_19" [label=None, style=solid]; -"1283 dequantize_per_tensor_default_19" -> "1284 dropout_10" [label=None, style=solid]; -"1284 dropout_10" -> "1285 linear_18" [label="(1, 28, 28, 768)", style=solid]; -"1285 linear_18" -> "1286 dropout_11" [label="(1, 28, 28, 192)", style=solid]; -"1286 dropout_11" -> "1287 layer_norm_7" [label="(1, 28, 28, 192)", style=solid]; -"1287 layer_norm_7" -> "1288 add_9" [label="(1, 28, 28, 192)", style=solid]; -"1288 add_9" -> "1300 pad_4" [label="(1, 28, 28, 192)", style=solid]; -"1288 add_9" -> "1405 add_12" [label="(1, 28, 28, 192)", style=solid]; -"1289 linear_19" -> "1290 relu__3" [label="(1, 15, 15, 512)", style=solid]; -"1290 relu__3" -> "1291 linear_20" [label="(1, 15, 15, 512)", style=solid]; -"1291 linear_20" -> "1292 view_15" [label="(1, 15, 15, 6)", style=solid]; -"1292 view_15" -> "1293 index_3" [label="(225, 6)", style=solid]; -"1293 index_3" -> "1294 view_16" [label="(4096, 6)", style=solid]; -"1294 view_16" -> "1295 permute_14" [label="(64, 64, 6)", style=solid]; -"1295 permute_14" -> "1296 contiguous_4" [label="(6, 64, 64)", style=solid]; -"1296 contiguous_4" -> "1297 unsqueeze_7" [label="(6, 64, 64)", style=solid]; -"1297 unsqueeze_7" -> "1298 sigmoid_3" [label="(1, 6, 64, 64)", style=solid]; -"1298 sigmoid_3" -> "1299 mul_6" [label="(1, 6, 64, 64)", style=solid]; -"1299 mul_6" -> "1333 add_10" [label="(1, 6, 64, 64)", style=solid]; -"1300 pad_4" -> "1301 roll_2" [label="(1, 32, 32, 192)", style=solid]; -"1301 roll_2" -> "1302 view_17" [label="(1, 32, 32, 192)", style=solid]; -"1302 view_17" -> "1303 permute_15" [label="(1, 4, 8, 4, 8, 192)", style=solid]; -"1303 permute_15" -> "1304 reshape_13" [label="(1, 4, 4, 8, 8, 192)", style=solid]; -"1304 reshape_13" -> "1305 quantize_per_tensor_default_20" [label="(16, 64, 192)", style=solid]; -"1304 reshape_13" -> "1334 new_zeros_1" [label="(16, 64, 192)", style=solid]; -"1305 quantize_per_tensor_default_20" -> "1306 dequantize_per_tensor_default_20" [label=None, style=solid]; -"1306 dequantize_per_tensor_default_20" -> "1310 linear_21" [label=None, style=solid]; -"1307 clone_3" -> "1308 slice_42" [label="(576,)", style=solid]; -"1307 clone_3" -> "1310 linear_21" [label="(576,)", style=solid]; -"1308 slice_42" -> "1309 zero__3" [label="(192,)", style=solid]; -"1310 linear_21" -> "1311 reshape_14" [label="(16, 64, 576)", style=solid]; -"1311 reshape_14" -> "1312 permute_16" [label="(16, 64, 3, 6, 32)", style=solid]; -"1312 permute_16" -> "1313 select_9" [label="(3, 16, 6, 64, 32)", style=solid]; -"1312 permute_16" -> "1314 select_10" [label="(3, 16, 6, 64, 32)", style=solid]; -"1312 permute_16" -> "1315 select_11" [label="(3, 16, 6, 64, 32)", style=solid]; -"1313 select_9" -> "1316 linalg_vector_norm_6" [label="(16, 6, 64, 32)", style=solid]; -"1313 select_9" -> "1318 expand_as_6" [label="(16, 6, 64, 32)", style=solid]; -"1313 select_9" -> "1319 div_6" [label="(16, 6, 64, 32)", style=solid]; -"1314 select_10" -> "1322 linalg_vector_norm_7" [label="(16, 6, 64, 32)", style=solid]; -"1314 select_10" -> "1324 expand_as_7" [label="(16, 6, 64, 32)", style=solid]; -"1314 select_10" -> "1325 div_7" [label="(16, 6, 64, 32)", style=solid]; -"1315 select_11" -> "1388 matmul_7" [label="(16, 6, 64, 32)", style=solid]; -"1316 linalg_vector_norm_6" -> "1317 clamp_min_6" [label="(16, 6, 64, 1)", style=solid]; -"1317 clamp_min_6" -> "1318 expand_as_6" [label="(16, 6, 64, 1)", style=solid]; -"1318 expand_as_6" -> "1319 div_6" [label="(16, 6, 64, 32)", style=solid]; -"1319 div_6" -> "1320 quantize_per_tensor_default_21" [label="(16, 6, 64, 32)", style=solid]; -"1320 quantize_per_tensor_default_21" -> "1321 dequantize_per_tensor_default_21" [label=None, style=solid]; -"1321 dequantize_per_tensor_default_21" -> "1329 matmul_6" [label=None, style=solid]; -"1322 linalg_vector_norm_7" -> "1323 clamp_min_7" [label="(16, 6, 64, 1)", style=solid]; -"1323 clamp_min_7" -> "1324 expand_as_7" [label="(16, 6, 64, 1)", style=solid]; -"1324 expand_as_7" -> "1325 div_7" [label="(16, 6, 64, 32)", style=solid]; -"1325 div_7" -> "1326 quantize_per_tensor_default_22" [label="(16, 6, 64, 32)", style=solid]; -"1326 quantize_per_tensor_default_22" -> "1327 dequantize_per_tensor_default_22" [label=None, style=solid]; -"1327 dequantize_per_tensor_default_22" -> "1328 transpose_6" [label=None, style=solid]; -"1328 transpose_6" -> "1329 matmul_6" [label="(16, 6, 32, 64)", style=solid]; -"1329 matmul_6" -> "1332 mul_7" [label="(16, 6, 64, 64)", style=solid]; -"1330 clamp_3" -> "1331 exp_3" [label="(6, 1, 1)", style=solid]; -"1331 exp_3" -> "1332 mul_7" [label="(6, 1, 1)", style=solid]; -"1332 mul_7" -> "1333 add_10" [label="(16, 6, 64, 64)", style=solid]; -"1333 add_10" -> "1381 view_19" [label="(16, 6, 64, 64)", style=solid]; -"1334 new_zeros_1" -> "1336 slice_43" [label="(32, 32)", style=solid]; -"1334 new_zeros_1" -> "1340 slice_45" [label="(32, 32)", style=solid]; -"1334 new_zeros_1" -> "1344 slice_47" [label="(32, 32)", style=solid]; -"1334 new_zeros_1" -> "1348 slice_49" [label="(32, 32)", style=solid]; -"1334 new_zeros_1" -> "1352 slice_51" [label="(32, 32)", style=solid]; -"1334 new_zeros_1" -> "1356 slice_53" [label="(32, 32)", style=solid]; -"1334 new_zeros_1" -> "1360 slice_55" [label="(32, 32)", style=solid]; -"1334 new_zeros_1" -> "1364 slice_57" [label="(32, 32)", style=solid]; -"1334 new_zeros_1" -> "1368 slice_59" [label="(32, 32)", style=solid]; -"1334 new_zeros_1" -> "1371 view_18" [label="(32, 32)", style=solid]; -"1335 lift_fresh_copy_9" -> "1338 fill__9" [label="()", style=solid]; -"1336 slice_43" -> "1337 slice_44" [label="(24, 32)", style=solid]; -"1337 slice_44" -> "1338 fill__9" [label="(24, 24)", style=solid]; -"1339 lift_fresh_copy_10" -> "1342 fill__10" [label="()", style=solid]; -"1340 slice_45" -> "1341 slice_46" [label="(24, 32)", style=solid]; -"1341 slice_46" -> "1342 fill__10" [label="(24, 4)", style=solid]; -"1343 lift_fresh_copy_11" -> "1346 fill__11" [label="()", style=solid]; -"1344 slice_47" -> "1345 slice_48" [label="(24, 32)", style=solid]; -"1345 slice_48" -> "1346 fill__11" [label="(24, 4)", style=solid]; -"1347 lift_fresh_copy_12" -> "1350 fill__12" [label="()", style=solid]; -"1348 slice_49" -> "1349 slice_50" [label="(4, 32)", style=solid]; -"1349 slice_50" -> "1350 fill__12" [label="(4, 24)", style=solid]; -"1351 lift_fresh_copy_13" -> "1354 fill__13" [label="()", style=solid]; -"1352 slice_51" -> "1353 slice_52" [label="(4, 32)", style=solid]; -"1353 slice_52" -> "1354 fill__13" [label="(4, 4)", style=solid]; -"1355 lift_fresh_copy_14" -> "1358 fill__14" [label="()", style=solid]; -"1356 slice_53" -> "1357 slice_54" [label="(4, 32)", style=solid]; -"1357 slice_54" -> "1358 fill__14" [label="(4, 4)", style=solid]; -"1359 lift_fresh_copy_15" -> "1362 fill__15" [label="()", style=solid]; -"1360 slice_55" -> "1361 slice_56" [label="(4, 32)", style=solid]; -"1361 slice_56" -> "1362 fill__15" [label="(4, 24)", style=solid]; -"1363 lift_fresh_copy_16" -> "1366 fill__16" [label="()", style=solid]; -"1364 slice_57" -> "1365 slice_58" [label="(4, 32)", style=solid]; -"1365 slice_58" -> "1366 fill__16" [label="(4, 4)", style=solid]; -"1367 lift_fresh_copy_17" -> "1370 fill__17" [label="()", style=solid]; -"1368 slice_59" -> "1369 slice_60" [label="(4, 32)", style=solid]; -"1369 slice_60" -> "1370 fill__17" [label="(4, 4)", style=solid]; -"1371 view_18" -> "1372 permute_17" [label="(4, 8, 4, 8)", style=solid]; -"1372 permute_17" -> "1373 reshape_15" [label="(4, 4, 8, 8)", style=solid]; -"1373 reshape_15" -> "1374 unsqueeze_8" [label="(16, 64)", style=solid]; -"1373 reshape_15" -> "1375 unsqueeze_9" [label="(16, 64)", style=solid]; -"1374 unsqueeze_8" -> "1376 sub_1" [label="(16, 1, 64)", style=solid]; -"1375 unsqueeze_9" -> "1376 sub_1" [label="(16, 64, 1)", style=solid]; -"1376 sub_1" -> "1377 ne_1" [label="(16, 64, 64)", style=solid]; -"1376 sub_1" -> "1378 masked_fill_2" [label="(16, 64, 64)", style=solid]; -"1376 sub_1" -> "1379 eq_1" [label="(16, 64, 64)", style=solid]; -"1377 ne_1" -> "1378 masked_fill_2" [label="(16, 64, 64)", style=solid]; -"1378 masked_fill_2" -> "1380 masked_fill_3" [label="(16, 64, 64)", style=solid]; -"1379 eq_1" -> "1380 masked_fill_3" [label="(16, 64, 64)", style=solid]; -"1380 masked_fill_3" -> "1382 unsqueeze_10" [label="(16, 64, 64)", style=solid]; -"1381 view_19" -> "1384 add_11" [label="(1, 16, 6, 64, 64)", style=solid]; -"1382 unsqueeze_10" -> "1383 unsqueeze_11" [label="(16, 1, 64, 64)", style=solid]; -"1383 unsqueeze_11" -> "1384 add_11" [label="(1, 16, 1, 64, 64)", style=solid]; -"1384 add_11" -> "1385 view_20" [label="(1, 16, 6, 64, 64)", style=solid]; -"1385 view_20" -> "1386 softmax_3" [label="(16, 6, 64, 64)", style=solid]; -"1386 softmax_3" -> "1387 dropout_12" [label="(16, 6, 64, 64)", style=solid]; -"1387 dropout_12" -> "1388 matmul_7" [label="(16, 6, 64, 64)", style=solid]; -"1388 matmul_7" -> "1389 quantize_per_tensor_default_23" [label="(16, 6, 64, 32)", style=solid]; -"1389 quantize_per_tensor_default_23" -> "1390 dequantize_per_tensor_default_23" [label=None, style=solid]; -"1390 dequantize_per_tensor_default_23" -> "1391 transpose_7" [label=None, style=solid]; -"1391 transpose_7" -> "1392 reshape_16" [label="(16, 64, 6, 32)", style=solid]; -"1392 reshape_16" -> "1393 linear_22" [label="(16, 64, 192)", style=solid]; -"1393 linear_22" -> "1394 dropout_13" [label="(16, 64, 192)", style=solid]; -"1394 dropout_13" -> "1395 view_21" [label="(16, 64, 192)", style=solid]; -"1395 view_21" -> "1396 permute_18" [label="(1, 4, 4, 8, 8, 192)", style=solid]; -"1396 permute_18" -> "1397 reshape_17" [label="(1, 4, 8, 4, 8, 192)", style=solid]; -"1397 reshape_17" -> "1398 roll_3" [label="(1, 32, 32, 192)", style=solid]; -"1398 roll_3" -> "1399 slice_61" [label="(1, 32, 32, 192)", style=solid]; -"1399 slice_61" -> "1400 slice_62" [label="(1, 32, 32, 192)", style=solid]; -"1400 slice_62" -> "1401 slice_63" [label="(1, 28, 32, 192)", style=solid]; -"1401 slice_63" -> "1402 slice_64" [label="(1, 28, 28, 192)", style=solid]; -"1402 slice_64" -> "1403 contiguous_5" [label="(1, 28, 28, 192)", style=solid]; -"1403 contiguous_5" -> "1404 layer_norm_8" [label="(1, 28, 28, 192)", style=solid]; -"1404 layer_norm_8" -> "1405 add_12" [label="(1, 28, 28, 192)", style=solid]; -"1405 add_12" -> "1406 quantize_per_tensor_default_24" [label="(1, 28, 28, 192)", style=solid]; -"1405 add_12" -> "1416 add_13" [label="(1, 28, 28, 192)", style=solid]; -"1406 quantize_per_tensor_default_24" -> "1407 dequantize_per_tensor_default_24" [label=None, style=solid]; -"1407 dequantize_per_tensor_default_24" -> "1408 linear_23" [label=None, style=solid]; -"1408 linear_23" -> "1409 gelu_3" [label="(1, 28, 28, 768)", style=solid]; -"1409 gelu_3" -> "1410 quantize_per_tensor_default_25" [label="(1, 28, 28, 768)", style=solid]; -"1410 quantize_per_tensor_default_25" -> "1411 dequantize_per_tensor_default_25" [label=None, style=solid]; -"1411 dequantize_per_tensor_default_25" -> "1412 dropout_14" [label=None, style=solid]; -"1412 dropout_14" -> "1413 linear_24" [label="(1, 28, 28, 768)", style=solid]; -"1413 linear_24" -> "1414 dropout_15" [label="(1, 28, 28, 192)", style=solid]; -"1414 dropout_15" -> "1415 layer_norm_9" [label="(1, 28, 28, 192)", style=solid]; -"1415 layer_norm_9" -> "1416 add_13" [label="(1, 28, 28, 192)", style=solid]; -"1416 add_13" -> "1417 quantize_per_tensor_default_26" [label="(1, 28, 28, 192)", style=solid]; -"1417 quantize_per_tensor_default_26" -> "1418 dequantize_per_tensor_default_26" [label=None, style=solid]; -"1418 dequantize_per_tensor_default_26" -> "1419 pad_5" [label=None, style=solid]; -"1419 pad_5" -> "1420 slice_65" [label="(1, 28, 28, 192)", style=solid]; -"1419 pad_5" -> "1423 slice_68" [label="(1, 28, 28, 192)", style=solid]; -"1419 pad_5" -> "1426 slice_71" [label="(1, 28, 28, 192)", style=solid]; -"1419 pad_5" -> "1429 slice_74" [label="(1, 28, 28, 192)", style=solid]; -"1420 slice_65" -> "1421 slice_66" [label="(1, 14, 28, 192)", style=solid]; -"1421 slice_66" -> "1422 slice_67" [label="(1, 14, 14, 192)", style=solid]; -"1422 slice_67" -> "1432 cat_1" [label="(1, 14, 14, 192)", style=solid]; -"1423 slice_68" -> "1424 slice_69" [label="(1, 14, 28, 192)", style=solid]; -"1424 slice_69" -> "1425 slice_70" [label="(1, 14, 14, 192)", style=solid]; -"1425 slice_70" -> "1432 cat_1" [label="(1, 14, 14, 192)", style=solid]; -"1426 slice_71" -> "1427 slice_72" [label="(1, 14, 28, 192)", style=solid]; -"1427 slice_72" -> "1428 slice_73" [label="(1, 14, 14, 192)", style=solid]; -"1428 slice_73" -> "1432 cat_1" [label="(1, 14, 14, 192)", style=solid]; -"1429 slice_74" -> "1430 slice_75" [label="(1, 14, 28, 192)", style=solid]; -"1430 slice_75" -> "1431 slice_76" [label="(1, 14, 14, 192)", style=solid]; -"1431 slice_76" -> "1432 cat_1" [label="(1, 14, 14, 192)", style=solid]; -"1432 cat_1" -> "1433 linear_25" [label="(1, 14, 14, 768)", style=solid]; -"1433 linear_25" -> "1434 layer_norm_10" [label="(1, 14, 14, 384)", style=solid]; -"1434 layer_norm_10" -> "1435 quantize_per_tensor_default_27" [label="(1, 14, 14, 384)", style=solid]; -"1434 layer_norm_10" -> "1497 add_15" [label="(1, 14, 14, 384)", style=solid]; -"1435 quantize_per_tensor_default_27" -> "1436 dequantize_per_tensor_default_27" [label=None, style=solid]; -"1436 dequantize_per_tensor_default_27" -> "1448 pad_6" [label=None, style=solid]; -"1437 linear_26" -> "1438 relu__4" [label="(1, 15, 15, 512)", style=solid]; -"1438 relu__4" -> "1439 linear_27" [label="(1, 15, 15, 512)", style=solid]; -"1439 linear_27" -> "1440 view_22" [label="(1, 15, 15, 12)", style=solid]; -"1440 view_22" -> "1441 index_4" [label="(225, 12)", style=solid]; -"1441 index_4" -> "1442 view_23" [label="(4096, 12)", style=solid]; -"1442 view_23" -> "1443 permute_19" [label="(64, 64, 12)", style=solid]; -"1443 permute_19" -> "1444 contiguous_6" [label="(12, 64, 64)", style=solid]; -"1444 contiguous_6" -> "1445 unsqueeze_12" [label="(12, 64, 64)", style=solid]; -"1445 unsqueeze_12" -> "1446 sigmoid_4" [label="(1, 12, 64, 64)", style=solid]; -"1446 sigmoid_4" -> "1447 mul_8" [label="(1, 12, 64, 64)", style=solid]; -"1447 mul_8" -> "1478 add_14" [label="(1, 12, 64, 64)", style=solid]; -"1448 pad_6" -> "1449 view_24" [label="(1, 16, 16, 384)", style=solid]; -"1449 view_24" -> "1450 permute_20" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"1450 permute_20" -> "1451 reshape_18" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"1451 reshape_18" -> "1455 linear_28" [label="(4, 64, 384)", style=solid]; -"1452 clone_4" -> "1453 slice_77" [label="(1152,)", style=solid]; -"1452 clone_4" -> "1455 linear_28" [label="(1152,)", style=solid]; -"1453 slice_77" -> "1454 zero__4" [label="(384,)", style=solid]; -"1455 linear_28" -> "1456 reshape_19" [label="(4, 64, 1152)", style=solid]; -"1456 reshape_19" -> "1457 permute_21" [label="(4, 64, 3, 12, 32)", style=solid]; -"1457 permute_21" -> "1458 select_12" [label="(3, 4, 12, 64, 32)", style=solid]; -"1457 permute_21" -> "1459 select_13" [label="(3, 4, 12, 64, 32)", style=solid]; -"1457 permute_21" -> "1460 select_14" [label="(3, 4, 12, 64, 32)", style=solid]; -"1458 select_12" -> "1461 linalg_vector_norm_8" [label="(4, 12, 64, 32)", style=solid]; -"1458 select_12" -> "1463 expand_as_8" [label="(4, 12, 64, 32)", style=solid]; -"1458 select_12" -> "1464 div_8" [label="(4, 12, 64, 32)", style=solid]; -"1459 select_13" -> "1467 linalg_vector_norm_9" [label="(4, 12, 64, 32)", style=solid]; -"1459 select_13" -> "1469 expand_as_9" [label="(4, 12, 64, 32)", style=solid]; -"1459 select_13" -> "1470 div_9" [label="(4, 12, 64, 32)", style=solid]; -"1460 select_14" -> "1481 matmul_9" [label="(4, 12, 64, 32)", style=solid]; -"1461 linalg_vector_norm_8" -> "1462 clamp_min_8" [label="(4, 12, 64, 1)", style=solid]; -"1462 clamp_min_8" -> "1463 expand_as_8" [label="(4, 12, 64, 1)", style=solid]; -"1463 expand_as_8" -> "1464 div_8" [label="(4, 12, 64, 32)", style=solid]; -"1464 div_8" -> "1465 quantize_per_tensor_default_28" [label="(4, 12, 64, 32)", style=solid]; -"1465 quantize_per_tensor_default_28" -> "1466 dequantize_per_tensor_default_28" [label=None, style=solid]; -"1466 dequantize_per_tensor_default_28" -> "1474 matmul_8" [label=None, style=solid]; -"1467 linalg_vector_norm_9" -> "1468 clamp_min_9" [label="(4, 12, 64, 1)", style=solid]; -"1468 clamp_min_9" -> "1469 expand_as_9" [label="(4, 12, 64, 1)", style=solid]; -"1469 expand_as_9" -> "1470 div_9" [label="(4, 12, 64, 32)", style=solid]; -"1470 div_9" -> "1471 quantize_per_tensor_default_29" [label="(4, 12, 64, 32)", style=solid]; -"1471 quantize_per_tensor_default_29" -> "1472 dequantize_per_tensor_default_29" [label=None, style=solid]; -"1472 dequantize_per_tensor_default_29" -> "1473 transpose_8" [label=None, style=solid]; -"1473 transpose_8" -> "1474 matmul_8" [label="(4, 12, 32, 64)", style=solid]; -"1474 matmul_8" -> "1477 mul_9" [label="(4, 12, 64, 64)", style=solid]; -"1475 clamp_4" -> "1476 exp_4" [label="(12, 1, 1)", style=solid]; -"1476 exp_4" -> "1477 mul_9" [label="(12, 1, 1)", style=solid]; -"1477 mul_9" -> "1478 add_14" [label="(4, 12, 64, 64)", style=solid]; -"1478 add_14" -> "1479 softmax_4" [label="(4, 12, 64, 64)", style=solid]; -"1479 softmax_4" -> "1480 dropout_16" [label="(4, 12, 64, 64)", style=solid]; -"1480 dropout_16" -> "1481 matmul_9" [label="(4, 12, 64, 64)", style=solid]; -"1481 matmul_9" -> "1482 quantize_per_tensor_default_30" [label="(4, 12, 64, 32)", style=solid]; -"1482 quantize_per_tensor_default_30" -> "1483 dequantize_per_tensor_default_30" [label=None, style=solid]; -"1483 dequantize_per_tensor_default_30" -> "1484 transpose_9" [label=None, style=solid]; -"1484 transpose_9" -> "1485 reshape_20" [label="(4, 64, 12, 32)", style=solid]; -"1485 reshape_20" -> "1486 linear_29" [label="(4, 64, 384)", style=solid]; -"1486 linear_29" -> "1487 dropout_17" [label="(4, 64, 384)", style=solid]; -"1487 dropout_17" -> "1488 view_25" [label="(4, 64, 384)", style=solid]; -"1488 view_25" -> "1489 permute_22" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"1489 permute_22" -> "1490 reshape_21" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"1490 reshape_21" -> "1491 slice_78" [label="(1, 16, 16, 384)", style=solid]; -"1491 slice_78" -> "1492 slice_79" [label="(1, 16, 16, 384)", style=solid]; -"1492 slice_79" -> "1493 slice_80" [label="(1, 14, 16, 384)", style=solid]; -"1493 slice_80" -> "1494 slice_81" [label="(1, 14, 14, 384)", style=solid]; -"1494 slice_81" -> "1495 contiguous_7" [label="(1, 14, 14, 384)", style=solid]; -"1495 contiguous_7" -> "1496 layer_norm_11" [label="(1, 14, 14, 384)", style=solid]; -"1496 layer_norm_11" -> "1497 add_15" [label="(1, 14, 14, 384)", style=solid]; -"1497 add_15" -> "1498 quantize_per_tensor_default_31" [label="(1, 14, 14, 384)", style=solid]; -"1497 add_15" -> "1508 add_16" [label="(1, 14, 14, 384)", style=solid]; -"1498 quantize_per_tensor_default_31" -> "1499 dequantize_per_tensor_default_31" [label=None, style=solid]; -"1499 dequantize_per_tensor_default_31" -> "1500 linear_30" [label=None, style=solid]; -"1500 linear_30" -> "1501 gelu_4" [label="(1, 14, 14, 1536)", style=solid]; -"1501 gelu_4" -> "1502 quantize_per_tensor_default_32" [label="(1, 14, 14, 1536)", style=solid]; -"1502 quantize_per_tensor_default_32" -> "1503 dequantize_per_tensor_default_32" [label=None, style=solid]; -"1503 dequantize_per_tensor_default_32" -> "1504 dropout_18" [label=None, style=solid]; -"1504 dropout_18" -> "1505 linear_31" [label="(1, 14, 14, 1536)", style=solid]; -"1505 linear_31" -> "1506 dropout_19" [label="(1, 14, 14, 384)", style=solid]; -"1506 dropout_19" -> "1507 layer_norm_12" [label="(1, 14, 14, 384)", style=solid]; -"1507 layer_norm_12" -> "1508 add_16" [label="(1, 14, 14, 384)", style=solid]; -"1508 add_16" -> "1520 pad_7" [label="(1, 14, 14, 384)", style=solid]; -"1508 add_16" -> "1625 add_19" [label="(1, 14, 14, 384)", style=solid]; -"1509 linear_32" -> "1510 relu__5" [label="(1, 15, 15, 512)", style=solid]; -"1510 relu__5" -> "1511 linear_33" [label="(1, 15, 15, 512)", style=solid]; -"1511 linear_33" -> "1512 view_26" [label="(1, 15, 15, 12)", style=solid]; -"1512 view_26" -> "1513 index_5" [label="(225, 12)", style=solid]; -"1513 index_5" -> "1514 view_27" [label="(4096, 12)", style=solid]; -"1514 view_27" -> "1515 permute_23" [label="(64, 64, 12)", style=solid]; -"1515 permute_23" -> "1516 contiguous_8" [label="(12, 64, 64)", style=solid]; -"1516 contiguous_8" -> "1517 unsqueeze_13" [label="(12, 64, 64)", style=solid]; -"1517 unsqueeze_13" -> "1518 sigmoid_5" [label="(1, 12, 64, 64)", style=solid]; -"1518 sigmoid_5" -> "1519 mul_10" [label="(1, 12, 64, 64)", style=solid]; -"1519 mul_10" -> "1553 add_17" [label="(1, 12, 64, 64)", style=solid]; -"1520 pad_7" -> "1521 roll_4" [label="(1, 16, 16, 384)", style=solid]; -"1521 roll_4" -> "1522 view_28" [label="(1, 16, 16, 384)", style=solid]; -"1522 view_28" -> "1523 permute_24" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"1523 permute_24" -> "1524 reshape_22" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"1524 reshape_22" -> "1525 quantize_per_tensor_default_33" [label="(4, 64, 384)", style=solid]; -"1524 reshape_22" -> "1554 new_zeros_2" [label="(4, 64, 384)", style=solid]; -"1525 quantize_per_tensor_default_33" -> "1526 dequantize_per_tensor_default_33" [label=None, style=solid]; -"1526 dequantize_per_tensor_default_33" -> "1530 linear_34" [label=None, style=solid]; -"1527 clone_5" -> "1528 slice_82" [label="(1152,)", style=solid]; -"1527 clone_5" -> "1530 linear_34" [label="(1152,)", style=solid]; -"1528 slice_82" -> "1529 zero__5" [label="(384,)", style=solid]; -"1530 linear_34" -> "1531 reshape_23" [label="(4, 64, 1152)", style=solid]; -"1531 reshape_23" -> "1532 permute_25" [label="(4, 64, 3, 12, 32)", style=solid]; -"1532 permute_25" -> "1533 select_15" [label="(3, 4, 12, 64, 32)", style=solid]; -"1532 permute_25" -> "1534 select_16" [label="(3, 4, 12, 64, 32)", style=solid]; -"1532 permute_25" -> "1535 select_17" [label="(3, 4, 12, 64, 32)", style=solid]; -"1533 select_15" -> "1536 linalg_vector_norm_10" [label="(4, 12, 64, 32)", style=solid]; -"1533 select_15" -> "1538 expand_as_10" [label="(4, 12, 64, 32)", style=solid]; -"1533 select_15" -> "1539 div_10" [label="(4, 12, 64, 32)", style=solid]; -"1534 select_16" -> "1542 linalg_vector_norm_11" [label="(4, 12, 64, 32)", style=solid]; -"1534 select_16" -> "1544 expand_as_11" [label="(4, 12, 64, 32)", style=solid]; -"1534 select_16" -> "1545 div_11" [label="(4, 12, 64, 32)", style=solid]; -"1535 select_17" -> "1608 matmul_11" [label="(4, 12, 64, 32)", style=solid]; -"1536 linalg_vector_norm_10" -> "1537 clamp_min_10" [label="(4, 12, 64, 1)", style=solid]; -"1537 clamp_min_10" -> "1538 expand_as_10" [label="(4, 12, 64, 1)", style=solid]; -"1538 expand_as_10" -> "1539 div_10" [label="(4, 12, 64, 32)", style=solid]; -"1539 div_10" -> "1540 quantize_per_tensor_default_34" [label="(4, 12, 64, 32)", style=solid]; -"1540 quantize_per_tensor_default_34" -> "1541 dequantize_per_tensor_default_34" [label=None, style=solid]; -"1541 dequantize_per_tensor_default_34" -> "1549 matmul_10" [label=None, style=solid]; -"1542 linalg_vector_norm_11" -> "1543 clamp_min_11" [label="(4, 12, 64, 1)", style=solid]; -"1543 clamp_min_11" -> "1544 expand_as_11" [label="(4, 12, 64, 1)", style=solid]; -"1544 expand_as_11" -> "1545 div_11" [label="(4, 12, 64, 32)", style=solid]; -"1545 div_11" -> "1546 quantize_per_tensor_default_35" [label="(4, 12, 64, 32)", style=solid]; -"1546 quantize_per_tensor_default_35" -> "1547 dequantize_per_tensor_default_35" [label=None, style=solid]; -"1547 dequantize_per_tensor_default_35" -> "1548 transpose_10" [label=None, style=solid]; -"1548 transpose_10" -> "1549 matmul_10" [label="(4, 12, 32, 64)", style=solid]; -"1549 matmul_10" -> "1552 mul_11" [label="(4, 12, 64, 64)", style=solid]; -"1550 clamp_5" -> "1551 exp_5" [label="(12, 1, 1)", style=solid]; -"1551 exp_5" -> "1552 mul_11" [label="(12, 1, 1)", style=solid]; -"1552 mul_11" -> "1553 add_17" [label="(4, 12, 64, 64)", style=solid]; -"1553 add_17" -> "1601 view_30" [label="(4, 12, 64, 64)", style=solid]; -"1554 new_zeros_2" -> "1556 slice_83" [label="(16, 16)", style=solid]; -"1554 new_zeros_2" -> "1560 slice_85" [label="(16, 16)", style=solid]; -"1554 new_zeros_2" -> "1564 slice_87" [label="(16, 16)", style=solid]; -"1554 new_zeros_2" -> "1568 slice_89" [label="(16, 16)", style=solid]; -"1554 new_zeros_2" -> "1572 slice_91" [label="(16, 16)", style=solid]; -"1554 new_zeros_2" -> "1576 slice_93" [label="(16, 16)", style=solid]; -"1554 new_zeros_2" -> "1580 slice_95" [label="(16, 16)", style=solid]; -"1554 new_zeros_2" -> "1584 slice_97" [label="(16, 16)", style=solid]; -"1554 new_zeros_2" -> "1588 slice_99" [label="(16, 16)", style=solid]; -"1554 new_zeros_2" -> "1591 view_29" [label="(16, 16)", style=solid]; -"1555 lift_fresh_copy_18" -> "1558 fill__18" [label="()", style=solid]; -"1556 slice_83" -> "1557 slice_84" [label="(8, 16)", style=solid]; -"1557 slice_84" -> "1558 fill__18" [label="(8, 8)", style=solid]; -"1559 lift_fresh_copy_19" -> "1562 fill__19" [label="()", style=solid]; -"1560 slice_85" -> "1561 slice_86" [label="(8, 16)", style=solid]; -"1561 slice_86" -> "1562 fill__19" [label="(8, 4)", style=solid]; -"1563 lift_fresh_copy_20" -> "1566 fill__20" [label="()", style=solid]; -"1564 slice_87" -> "1565 slice_88" [label="(8, 16)", style=solid]; -"1565 slice_88" -> "1566 fill__20" [label="(8, 4)", style=solid]; -"1567 lift_fresh_copy_21" -> "1570 fill__21" [label="()", style=solid]; -"1568 slice_89" -> "1569 slice_90" [label="(4, 16)", style=solid]; -"1569 slice_90" -> "1570 fill__21" [label="(4, 8)", style=solid]; -"1571 lift_fresh_copy_22" -> "1574 fill__22" [label="()", style=solid]; -"1572 slice_91" -> "1573 slice_92" [label="(4, 16)", style=solid]; -"1573 slice_92" -> "1574 fill__22" [label="(4, 4)", style=solid]; -"1575 lift_fresh_copy_23" -> "1578 fill__23" [label="()", style=solid]; -"1576 slice_93" -> "1577 slice_94" [label="(4, 16)", style=solid]; -"1577 slice_94" -> "1578 fill__23" [label="(4, 4)", style=solid]; -"1579 lift_fresh_copy_24" -> "1582 fill__24" [label="()", style=solid]; -"1580 slice_95" -> "1581 slice_96" [label="(4, 16)", style=solid]; -"1581 slice_96" -> "1582 fill__24" [label="(4, 8)", style=solid]; -"1583 lift_fresh_copy_25" -> "1586 fill__25" [label="()", style=solid]; -"1584 slice_97" -> "1585 slice_98" [label="(4, 16)", style=solid]; -"1585 slice_98" -> "1586 fill__25" [label="(4, 4)", style=solid]; -"1587 lift_fresh_copy_26" -> "1590 fill__26" [label="()", style=solid]; -"1588 slice_99" -> "1589 slice_100" [label="(4, 16)", style=solid]; -"1589 slice_100" -> "1590 fill__26" [label="(4, 4)", style=solid]; -"1591 view_29" -> "1592 permute_26" [label="(2, 8, 2, 8)", style=solid]; -"1592 permute_26" -> "1593 reshape_24" [label="(2, 2, 8, 8)", style=solid]; -"1593 reshape_24" -> "1594 unsqueeze_14" [label="(4, 64)", style=solid]; -"1593 reshape_24" -> "1595 unsqueeze_15" [label="(4, 64)", style=solid]; -"1594 unsqueeze_14" -> "1596 sub_2" [label="(4, 1, 64)", style=solid]; -"1595 unsqueeze_15" -> "1596 sub_2" [label="(4, 64, 1)", style=solid]; -"1596 sub_2" -> "1597 ne_2" [label="(4, 64, 64)", style=solid]; -"1596 sub_2" -> "1598 masked_fill_4" [label="(4, 64, 64)", style=solid]; -"1596 sub_2" -> "1599 eq_2" [label="(4, 64, 64)", style=solid]; -"1597 ne_2" -> "1598 masked_fill_4" [label="(4, 64, 64)", style=solid]; -"1598 masked_fill_4" -> "1600 masked_fill_5" [label="(4, 64, 64)", style=solid]; -"1599 eq_2" -> "1600 masked_fill_5" [label="(4, 64, 64)", style=solid]; -"1600 masked_fill_5" -> "1602 unsqueeze_16" [label="(4, 64, 64)", style=solid]; -"1601 view_30" -> "1604 add_18" [label="(1, 4, 12, 64, 64)", style=solid]; -"1602 unsqueeze_16" -> "1603 unsqueeze_17" [label="(4, 1, 64, 64)", style=solid]; -"1603 unsqueeze_17" -> "1604 add_18" [label="(1, 4, 1, 64, 64)", style=solid]; -"1604 add_18" -> "1605 view_31" [label="(1, 4, 12, 64, 64)", style=solid]; -"1605 view_31" -> "1606 softmax_5" [label="(4, 12, 64, 64)", style=solid]; -"1606 softmax_5" -> "1607 dropout_20" [label="(4, 12, 64, 64)", style=solid]; -"1607 dropout_20" -> "1608 matmul_11" [label="(4, 12, 64, 64)", style=solid]; -"1608 matmul_11" -> "1609 quantize_per_tensor_default_36" [label="(4, 12, 64, 32)", style=solid]; -"1609 quantize_per_tensor_default_36" -> "1610 dequantize_per_tensor_default_36" [label=None, style=solid]; -"1610 dequantize_per_tensor_default_36" -> "1611 transpose_11" [label=None, style=solid]; -"1611 transpose_11" -> "1612 reshape_25" [label="(4, 64, 12, 32)", style=solid]; -"1612 reshape_25" -> "1613 linear_35" [label="(4, 64, 384)", style=solid]; -"1613 linear_35" -> "1614 dropout_21" [label="(4, 64, 384)", style=solid]; -"1614 dropout_21" -> "1615 view_32" [label="(4, 64, 384)", style=solid]; -"1615 view_32" -> "1616 permute_27" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"1616 permute_27" -> "1617 reshape_26" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"1617 reshape_26" -> "1618 roll_5" [label="(1, 16, 16, 384)", style=solid]; -"1618 roll_5" -> "1619 slice_101" [label="(1, 16, 16, 384)", style=solid]; -"1619 slice_101" -> "1620 slice_102" [label="(1, 16, 16, 384)", style=solid]; -"1620 slice_102" -> "1621 slice_103" [label="(1, 14, 16, 384)", style=solid]; -"1621 slice_103" -> "1622 slice_104" [label="(1, 14, 14, 384)", style=solid]; -"1622 slice_104" -> "1623 contiguous_9" [label="(1, 14, 14, 384)", style=solid]; -"1623 contiguous_9" -> "1624 layer_norm_13" [label="(1, 14, 14, 384)", style=solid]; -"1624 layer_norm_13" -> "1625 add_19" [label="(1, 14, 14, 384)", style=solid]; -"1625 add_19" -> "1626 quantize_per_tensor_default_37" [label="(1, 14, 14, 384)", style=solid]; -"1625 add_19" -> "1636 add_20" [label="(1, 14, 14, 384)", style=solid]; -"1626 quantize_per_tensor_default_37" -> "1627 dequantize_per_tensor_default_37" [label=None, style=solid]; -"1627 dequantize_per_tensor_default_37" -> "1628 linear_36" [label=None, style=solid]; -"1628 linear_36" -> "1629 gelu_5" [label="(1, 14, 14, 1536)", style=solid]; -"1629 gelu_5" -> "1630 quantize_per_tensor_default_38" [label="(1, 14, 14, 1536)", style=solid]; -"1630 quantize_per_tensor_default_38" -> "1631 dequantize_per_tensor_default_38" [label=None, style=solid]; -"1631 dequantize_per_tensor_default_38" -> "1632 dropout_22" [label=None, style=solid]; -"1632 dropout_22" -> "1633 linear_37" [label="(1, 14, 14, 1536)", style=solid]; -"1633 linear_37" -> "1634 dropout_23" [label="(1, 14, 14, 384)", style=solid]; -"1634 dropout_23" -> "1635 layer_norm_14" [label="(1, 14, 14, 384)", style=solid]; -"1635 layer_norm_14" -> "1636 add_20" [label="(1, 14, 14, 384)", style=solid]; -"1636 add_20" -> "1637 quantize_per_tensor_default_39" [label="(1, 14, 14, 384)", style=solid]; -"1636 add_20" -> "1699 add_22" [label="(1, 14, 14, 384)", style=solid]; -"1637 quantize_per_tensor_default_39" -> "1638 dequantize_per_tensor_default_39" [label=None, style=solid]; -"1638 dequantize_per_tensor_default_39" -> "1650 pad_8" [label=None, style=solid]; -"1639 linear_38" -> "1640 relu__6" [label="(1, 15, 15, 512)", style=solid]; -"1640 relu__6" -> "1641 linear_39" [label="(1, 15, 15, 512)", style=solid]; -"1641 linear_39" -> "1642 view_33" [label="(1, 15, 15, 12)", style=solid]; -"1642 view_33" -> "1643 index_6" [label="(225, 12)", style=solid]; -"1643 index_6" -> "1644 view_34" [label="(4096, 12)", style=solid]; -"1644 view_34" -> "1645 permute_28" [label="(64, 64, 12)", style=solid]; -"1645 permute_28" -> "1646 contiguous_10" [label="(12, 64, 64)", style=solid]; -"1646 contiguous_10" -> "1647 unsqueeze_18" [label="(12, 64, 64)", style=solid]; -"1647 unsqueeze_18" -> "1648 sigmoid_6" [label="(1, 12, 64, 64)", style=solid]; -"1648 sigmoid_6" -> "1649 mul_12" [label="(1, 12, 64, 64)", style=solid]; -"1649 mul_12" -> "1680 add_21" [label="(1, 12, 64, 64)", style=solid]; -"1650 pad_8" -> "1651 view_35" [label="(1, 16, 16, 384)", style=solid]; -"1651 view_35" -> "1652 permute_29" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"1652 permute_29" -> "1653 reshape_27" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"1653 reshape_27" -> "1657 linear_40" [label="(4, 64, 384)", style=solid]; -"1654 clone_6" -> "1655 slice_105" [label="(1152,)", style=solid]; -"1654 clone_6" -> "1657 linear_40" [label="(1152,)", style=solid]; -"1655 slice_105" -> "1656 zero__6" [label="(384,)", style=solid]; -"1657 linear_40" -> "1658 reshape_28" [label="(4, 64, 1152)", style=solid]; -"1658 reshape_28" -> "1659 permute_30" [label="(4, 64, 3, 12, 32)", style=solid]; -"1659 permute_30" -> "1660 select_18" [label="(3, 4, 12, 64, 32)", style=solid]; -"1659 permute_30" -> "1661 select_19" [label="(3, 4, 12, 64, 32)", style=solid]; -"1659 permute_30" -> "1662 select_20" [label="(3, 4, 12, 64, 32)", style=solid]; -"1660 select_18" -> "1663 linalg_vector_norm_12" [label="(4, 12, 64, 32)", style=solid]; -"1660 select_18" -> "1665 expand_as_12" [label="(4, 12, 64, 32)", style=solid]; -"1660 select_18" -> "1666 div_12" [label="(4, 12, 64, 32)", style=solid]; -"1661 select_19" -> "1669 linalg_vector_norm_13" [label="(4, 12, 64, 32)", style=solid]; -"1661 select_19" -> "1671 expand_as_13" [label="(4, 12, 64, 32)", style=solid]; -"1661 select_19" -> "1672 div_13" [label="(4, 12, 64, 32)", style=solid]; -"1662 select_20" -> "1683 matmul_13" [label="(4, 12, 64, 32)", style=solid]; -"1663 linalg_vector_norm_12" -> "1664 clamp_min_12" [label="(4, 12, 64, 1)", style=solid]; -"1664 clamp_min_12" -> "1665 expand_as_12" [label="(4, 12, 64, 1)", style=solid]; -"1665 expand_as_12" -> "1666 div_12" [label="(4, 12, 64, 32)", style=solid]; -"1666 div_12" -> "1667 quantize_per_tensor_default_40" [label="(4, 12, 64, 32)", style=solid]; -"1667 quantize_per_tensor_default_40" -> "1668 dequantize_per_tensor_default_40" [label=None, style=solid]; -"1668 dequantize_per_tensor_default_40" -> "1676 matmul_12" [label=None, style=solid]; -"1669 linalg_vector_norm_13" -> "1670 clamp_min_13" [label="(4, 12, 64, 1)", style=solid]; -"1670 clamp_min_13" -> "1671 expand_as_13" [label="(4, 12, 64, 1)", style=solid]; -"1671 expand_as_13" -> "1672 div_13" [label="(4, 12, 64, 32)", style=solid]; -"1672 div_13" -> "1673 quantize_per_tensor_default_41" [label="(4, 12, 64, 32)", style=solid]; -"1673 quantize_per_tensor_default_41" -> "1674 dequantize_per_tensor_default_41" [label=None, style=solid]; -"1674 dequantize_per_tensor_default_41" -> "1675 transpose_12" [label=None, style=solid]; -"1675 transpose_12" -> "1676 matmul_12" [label="(4, 12, 32, 64)", style=solid]; -"1676 matmul_12" -> "1679 mul_13" [label="(4, 12, 64, 64)", style=solid]; -"1677 clamp_6" -> "1678 exp_6" [label="(12, 1, 1)", style=solid]; -"1678 exp_6" -> "1679 mul_13" [label="(12, 1, 1)", style=solid]; -"1679 mul_13" -> "1680 add_21" [label="(4, 12, 64, 64)", style=solid]; -"1680 add_21" -> "1681 softmax_6" [label="(4, 12, 64, 64)", style=solid]; -"1681 softmax_6" -> "1682 dropout_24" [label="(4, 12, 64, 64)", style=solid]; -"1682 dropout_24" -> "1683 matmul_13" [label="(4, 12, 64, 64)", style=solid]; -"1683 matmul_13" -> "1684 quantize_per_tensor_default_42" [label="(4, 12, 64, 32)", style=solid]; -"1684 quantize_per_tensor_default_42" -> "1685 dequantize_per_tensor_default_42" [label=None, style=solid]; -"1685 dequantize_per_tensor_default_42" -> "1686 transpose_13" [label=None, style=solid]; -"1686 transpose_13" -> "1687 reshape_29" [label="(4, 64, 12, 32)", style=solid]; -"1687 reshape_29" -> "1688 linear_41" [label="(4, 64, 384)", style=solid]; -"1688 linear_41" -> "1689 dropout_25" [label="(4, 64, 384)", style=solid]; -"1689 dropout_25" -> "1690 view_36" [label="(4, 64, 384)", style=solid]; -"1690 view_36" -> "1691 permute_31" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"1691 permute_31" -> "1692 reshape_30" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"1692 reshape_30" -> "1693 slice_106" [label="(1, 16, 16, 384)", style=solid]; -"1693 slice_106" -> "1694 slice_107" [label="(1, 16, 16, 384)", style=solid]; -"1694 slice_107" -> "1695 slice_108" [label="(1, 14, 16, 384)", style=solid]; -"1695 slice_108" -> "1696 slice_109" [label="(1, 14, 14, 384)", style=solid]; -"1696 slice_109" -> "1697 contiguous_11" [label="(1, 14, 14, 384)", style=solid]; -"1697 contiguous_11" -> "1698 layer_norm_15" [label="(1, 14, 14, 384)", style=solid]; -"1698 layer_norm_15" -> "1699 add_22" [label="(1, 14, 14, 384)", style=solid]; -"1699 add_22" -> "1700 quantize_per_tensor_default_43" [label="(1, 14, 14, 384)", style=solid]; -"1699 add_22" -> "1710 add_23" [label="(1, 14, 14, 384)", style=solid]; -"1700 quantize_per_tensor_default_43" -> "1701 dequantize_per_tensor_default_43" [label=None, style=solid]; -"1701 dequantize_per_tensor_default_43" -> "1702 linear_42" [label=None, style=solid]; -"1702 linear_42" -> "1703 gelu_6" [label="(1, 14, 14, 1536)", style=solid]; -"1703 gelu_6" -> "1704 quantize_per_tensor_default_44" [label="(1, 14, 14, 1536)", style=solid]; -"1704 quantize_per_tensor_default_44" -> "1705 dequantize_per_tensor_default_44" [label=None, style=solid]; -"1705 dequantize_per_tensor_default_44" -> "1706 dropout_26" [label=None, style=solid]; -"1706 dropout_26" -> "1707 linear_43" [label="(1, 14, 14, 1536)", style=solid]; -"1707 linear_43" -> "1708 dropout_27" [label="(1, 14, 14, 384)", style=solid]; -"1708 dropout_27" -> "1709 layer_norm_16" [label="(1, 14, 14, 384)", style=solid]; -"1709 layer_norm_16" -> "1710 add_23" [label="(1, 14, 14, 384)", style=solid]; -"1710 add_23" -> "1722 pad_9" [label="(1, 14, 14, 384)", style=solid]; -"1710 add_23" -> "1827 add_26" [label="(1, 14, 14, 384)", style=solid]; -"1711 linear_44" -> "1712 relu__7" [label="(1, 15, 15, 512)", style=solid]; -"1712 relu__7" -> "1713 linear_45" [label="(1, 15, 15, 512)", style=solid]; -"1713 linear_45" -> "1714 view_37" [label="(1, 15, 15, 12)", style=solid]; -"1714 view_37" -> "1715 index_7" [label="(225, 12)", style=solid]; -"1715 index_7" -> "1716 view_38" [label="(4096, 12)", style=solid]; -"1716 view_38" -> "1717 permute_32" [label="(64, 64, 12)", style=solid]; -"1717 permute_32" -> "1718 contiguous_12" [label="(12, 64, 64)", style=solid]; -"1718 contiguous_12" -> "1719 unsqueeze_19" [label="(12, 64, 64)", style=solid]; -"1719 unsqueeze_19" -> "1720 sigmoid_7" [label="(1, 12, 64, 64)", style=solid]; -"1720 sigmoid_7" -> "1721 mul_14" [label="(1, 12, 64, 64)", style=solid]; -"1721 mul_14" -> "1755 add_24" [label="(1, 12, 64, 64)", style=solid]; -"1722 pad_9" -> "1723 roll_6" [label="(1, 16, 16, 384)", style=solid]; -"1723 roll_6" -> "1724 view_39" [label="(1, 16, 16, 384)", style=solid]; -"1724 view_39" -> "1725 permute_33" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"1725 permute_33" -> "1726 reshape_31" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"1726 reshape_31" -> "1727 quantize_per_tensor_default_45" [label="(4, 64, 384)", style=solid]; -"1726 reshape_31" -> "1756 new_zeros_3" [label="(4, 64, 384)", style=solid]; -"1727 quantize_per_tensor_default_45" -> "1728 dequantize_per_tensor_default_45" [label=None, style=solid]; -"1728 dequantize_per_tensor_default_45" -> "1732 linear_46" [label=None, style=solid]; -"1729 clone_7" -> "1730 slice_110" [label="(1152,)", style=solid]; -"1729 clone_7" -> "1732 linear_46" [label="(1152,)", style=solid]; -"1730 slice_110" -> "1731 zero__7" [label="(384,)", style=solid]; -"1732 linear_46" -> "1733 reshape_32" [label="(4, 64, 1152)", style=solid]; -"1733 reshape_32" -> "1734 permute_34" [label="(4, 64, 3, 12, 32)", style=solid]; -"1734 permute_34" -> "1735 select_21" [label="(3, 4, 12, 64, 32)", style=solid]; -"1734 permute_34" -> "1736 select_22" [label="(3, 4, 12, 64, 32)", style=solid]; -"1734 permute_34" -> "1737 select_23" [label="(3, 4, 12, 64, 32)", style=solid]; -"1735 select_21" -> "1738 linalg_vector_norm_14" [label="(4, 12, 64, 32)", style=solid]; -"1735 select_21" -> "1740 expand_as_14" [label="(4, 12, 64, 32)", style=solid]; -"1735 select_21" -> "1741 div_14" [label="(4, 12, 64, 32)", style=solid]; -"1736 select_22" -> "1744 linalg_vector_norm_15" [label="(4, 12, 64, 32)", style=solid]; -"1736 select_22" -> "1746 expand_as_15" [label="(4, 12, 64, 32)", style=solid]; -"1736 select_22" -> "1747 div_15" [label="(4, 12, 64, 32)", style=solid]; -"1737 select_23" -> "1810 matmul_15" [label="(4, 12, 64, 32)", style=solid]; -"1738 linalg_vector_norm_14" -> "1739 clamp_min_14" [label="(4, 12, 64, 1)", style=solid]; -"1739 clamp_min_14" -> "1740 expand_as_14" [label="(4, 12, 64, 1)", style=solid]; -"1740 expand_as_14" -> "1741 div_14" [label="(4, 12, 64, 32)", style=solid]; -"1741 div_14" -> "1742 quantize_per_tensor_default_46" [label="(4, 12, 64, 32)", style=solid]; -"1742 quantize_per_tensor_default_46" -> "1743 dequantize_per_tensor_default_46" [label=None, style=solid]; -"1743 dequantize_per_tensor_default_46" -> "1751 matmul_14" [label=None, style=solid]; -"1744 linalg_vector_norm_15" -> "1745 clamp_min_15" [label="(4, 12, 64, 1)", style=solid]; -"1745 clamp_min_15" -> "1746 expand_as_15" [label="(4, 12, 64, 1)", style=solid]; -"1746 expand_as_15" -> "1747 div_15" [label="(4, 12, 64, 32)", style=solid]; -"1747 div_15" -> "1748 quantize_per_tensor_default_47" [label="(4, 12, 64, 32)", style=solid]; -"1748 quantize_per_tensor_default_47" -> "1749 dequantize_per_tensor_default_47" [label=None, style=solid]; -"1749 dequantize_per_tensor_default_47" -> "1750 transpose_14" [label=None, style=solid]; -"1750 transpose_14" -> "1751 matmul_14" [label="(4, 12, 32, 64)", style=solid]; -"1751 matmul_14" -> "1754 mul_15" [label="(4, 12, 64, 64)", style=solid]; -"1752 clamp_7" -> "1753 exp_7" [label="(12, 1, 1)", style=solid]; -"1753 exp_7" -> "1754 mul_15" [label="(12, 1, 1)", style=solid]; -"1754 mul_15" -> "1755 add_24" [label="(4, 12, 64, 64)", style=solid]; -"1755 add_24" -> "1803 view_41" [label="(4, 12, 64, 64)", style=solid]; -"1756 new_zeros_3" -> "1758 slice_111" [label="(16, 16)", style=solid]; -"1756 new_zeros_3" -> "1762 slice_113" [label="(16, 16)", style=solid]; -"1756 new_zeros_3" -> "1766 slice_115" [label="(16, 16)", style=solid]; -"1756 new_zeros_3" -> "1770 slice_117" [label="(16, 16)", style=solid]; -"1756 new_zeros_3" -> "1774 slice_119" [label="(16, 16)", style=solid]; -"1756 new_zeros_3" -> "1778 slice_121" [label="(16, 16)", style=solid]; -"1756 new_zeros_3" -> "1782 slice_123" [label="(16, 16)", style=solid]; -"1756 new_zeros_3" -> "1786 slice_125" [label="(16, 16)", style=solid]; -"1756 new_zeros_3" -> "1790 slice_127" [label="(16, 16)", style=solid]; -"1756 new_zeros_3" -> "1793 view_40" [label="(16, 16)", style=solid]; -"1757 lift_fresh_copy_27" -> "1760 fill__27" [label="()", style=solid]; -"1758 slice_111" -> "1759 slice_112" [label="(8, 16)", style=solid]; -"1759 slice_112" -> "1760 fill__27" [label="(8, 8)", style=solid]; -"1761 lift_fresh_copy_28" -> "1764 fill__28" [label="()", style=solid]; -"1762 slice_113" -> "1763 slice_114" [label="(8, 16)", style=solid]; -"1763 slice_114" -> "1764 fill__28" [label="(8, 4)", style=solid]; -"1765 lift_fresh_copy_29" -> "1768 fill__29" [label="()", style=solid]; -"1766 slice_115" -> "1767 slice_116" [label="(8, 16)", style=solid]; -"1767 slice_116" -> "1768 fill__29" [label="(8, 4)", style=solid]; -"1769 lift_fresh_copy_30" -> "1772 fill__30" [label="()", style=solid]; -"1770 slice_117" -> "1771 slice_118" [label="(4, 16)", style=solid]; -"1771 slice_118" -> "1772 fill__30" [label="(4, 8)", style=solid]; -"1773 lift_fresh_copy_31" -> "1776 fill__31" [label="()", style=solid]; -"1774 slice_119" -> "1775 slice_120" [label="(4, 16)", style=solid]; -"1775 slice_120" -> "1776 fill__31" [label="(4, 4)", style=solid]; -"1777 lift_fresh_copy_32" -> "1780 fill__32" [label="()", style=solid]; -"1778 slice_121" -> "1779 slice_122" [label="(4, 16)", style=solid]; -"1779 slice_122" -> "1780 fill__32" [label="(4, 4)", style=solid]; -"1781 lift_fresh_copy_33" -> "1784 fill__33" [label="()", style=solid]; -"1782 slice_123" -> "1783 slice_124" [label="(4, 16)", style=solid]; -"1783 slice_124" -> "1784 fill__33" [label="(4, 8)", style=solid]; -"1785 lift_fresh_copy_34" -> "1788 fill__34" [label="()", style=solid]; -"1786 slice_125" -> "1787 slice_126" [label="(4, 16)", style=solid]; -"1787 slice_126" -> "1788 fill__34" [label="(4, 4)", style=solid]; -"1789 lift_fresh_copy_35" -> "1792 fill__35" [label="()", style=solid]; -"1790 slice_127" -> "1791 slice_128" [label="(4, 16)", style=solid]; -"1791 slice_128" -> "1792 fill__35" [label="(4, 4)", style=solid]; -"1793 view_40" -> "1794 permute_35" [label="(2, 8, 2, 8)", style=solid]; -"1794 permute_35" -> "1795 reshape_33" [label="(2, 2, 8, 8)", style=solid]; -"1795 reshape_33" -> "1796 unsqueeze_20" [label="(4, 64)", style=solid]; -"1795 reshape_33" -> "1797 unsqueeze_21" [label="(4, 64)", style=solid]; -"1796 unsqueeze_20" -> "1798 sub_3" [label="(4, 1, 64)", style=solid]; -"1797 unsqueeze_21" -> "1798 sub_3" [label="(4, 64, 1)", style=solid]; -"1798 sub_3" -> "1799 ne_3" [label="(4, 64, 64)", style=solid]; -"1798 sub_3" -> "1800 masked_fill_6" [label="(4, 64, 64)", style=solid]; -"1798 sub_3" -> "1801 eq_3" [label="(4, 64, 64)", style=solid]; -"1799 ne_3" -> "1800 masked_fill_6" [label="(4, 64, 64)", style=solid]; -"1800 masked_fill_6" -> "1802 masked_fill_7" [label="(4, 64, 64)", style=solid]; -"1801 eq_3" -> "1802 masked_fill_7" [label="(4, 64, 64)", style=solid]; -"1802 masked_fill_7" -> "1804 unsqueeze_22" [label="(4, 64, 64)", style=solid]; -"1803 view_41" -> "1806 add_25" [label="(1, 4, 12, 64, 64)", style=solid]; -"1804 unsqueeze_22" -> "1805 unsqueeze_23" [label="(4, 1, 64, 64)", style=solid]; -"1805 unsqueeze_23" -> "1806 add_25" [label="(1, 4, 1, 64, 64)", style=solid]; -"1806 add_25" -> "1807 view_42" [label="(1, 4, 12, 64, 64)", style=solid]; -"1807 view_42" -> "1808 softmax_7" [label="(4, 12, 64, 64)", style=solid]; -"1808 softmax_7" -> "1809 dropout_28" [label="(4, 12, 64, 64)", style=solid]; -"1809 dropout_28" -> "1810 matmul_15" [label="(4, 12, 64, 64)", style=solid]; -"1810 matmul_15" -> "1811 quantize_per_tensor_default_48" [label="(4, 12, 64, 32)", style=solid]; -"1811 quantize_per_tensor_default_48" -> "1812 dequantize_per_tensor_default_48" [label=None, style=solid]; -"1812 dequantize_per_tensor_default_48" -> "1813 transpose_15" [label=None, style=solid]; -"1813 transpose_15" -> "1814 reshape_34" [label="(4, 64, 12, 32)", style=solid]; -"1814 reshape_34" -> "1815 linear_47" [label="(4, 64, 384)", style=solid]; -"1815 linear_47" -> "1816 dropout_29" [label="(4, 64, 384)", style=solid]; -"1816 dropout_29" -> "1817 view_43" [label="(4, 64, 384)", style=solid]; -"1817 view_43" -> "1818 permute_36" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"1818 permute_36" -> "1819 reshape_35" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"1819 reshape_35" -> "1820 roll_7" [label="(1, 16, 16, 384)", style=solid]; -"1820 roll_7" -> "1821 slice_129" [label="(1, 16, 16, 384)", style=solid]; -"1821 slice_129" -> "1822 slice_130" [label="(1, 16, 16, 384)", style=solid]; -"1822 slice_130" -> "1823 slice_131" [label="(1, 14, 16, 384)", style=solid]; -"1823 slice_131" -> "1824 slice_132" [label="(1, 14, 14, 384)", style=solid]; -"1824 slice_132" -> "1825 contiguous_13" [label="(1, 14, 14, 384)", style=solid]; -"1825 contiguous_13" -> "1826 layer_norm_17" [label="(1, 14, 14, 384)", style=solid]; -"1826 layer_norm_17" -> "1827 add_26" [label="(1, 14, 14, 384)", style=solid]; -"1827 add_26" -> "1828 quantize_per_tensor_default_49" [label="(1, 14, 14, 384)", style=solid]; -"1827 add_26" -> "1838 add_27" [label="(1, 14, 14, 384)", style=solid]; -"1828 quantize_per_tensor_default_49" -> "1829 dequantize_per_tensor_default_49" [label=None, style=solid]; -"1829 dequantize_per_tensor_default_49" -> "1830 linear_48" [label=None, style=solid]; -"1830 linear_48" -> "1831 gelu_7" [label="(1, 14, 14, 1536)", style=solid]; -"1831 gelu_7" -> "1832 quantize_per_tensor_default_50" [label="(1, 14, 14, 1536)", style=solid]; -"1832 quantize_per_tensor_default_50" -> "1833 dequantize_per_tensor_default_50" [label=None, style=solid]; -"1833 dequantize_per_tensor_default_50" -> "1834 dropout_30" [label=None, style=solid]; -"1834 dropout_30" -> "1835 linear_49" [label="(1, 14, 14, 1536)", style=solid]; -"1835 linear_49" -> "1836 dropout_31" [label="(1, 14, 14, 384)", style=solid]; -"1836 dropout_31" -> "1837 layer_norm_18" [label="(1, 14, 14, 384)", style=solid]; -"1837 layer_norm_18" -> "1838 add_27" [label="(1, 14, 14, 384)", style=solid]; -"1838 add_27" -> "1839 quantize_per_tensor_default_51" [label="(1, 14, 14, 384)", style=solid]; -"1838 add_27" -> "1901 add_29" [label="(1, 14, 14, 384)", style=solid]; -"1839 quantize_per_tensor_default_51" -> "1840 dequantize_per_tensor_default_51" [label=None, style=solid]; -"1840 dequantize_per_tensor_default_51" -> "1852 pad_10" [label=None, style=solid]; -"1841 linear_50" -> "1842 relu__8" [label="(1, 15, 15, 512)", style=solid]; -"1842 relu__8" -> "1843 linear_51" [label="(1, 15, 15, 512)", style=solid]; -"1843 linear_51" -> "1844 view_44" [label="(1, 15, 15, 12)", style=solid]; -"1844 view_44" -> "1845 index_8" [label="(225, 12)", style=solid]; -"1845 index_8" -> "1846 view_45" [label="(4096, 12)", style=solid]; -"1846 view_45" -> "1847 permute_37" [label="(64, 64, 12)", style=solid]; -"1847 permute_37" -> "1848 contiguous_14" [label="(12, 64, 64)", style=solid]; -"1848 contiguous_14" -> "1849 unsqueeze_24" [label="(12, 64, 64)", style=solid]; -"1849 unsqueeze_24" -> "1850 sigmoid_8" [label="(1, 12, 64, 64)", style=solid]; -"1850 sigmoid_8" -> "1851 mul_16" [label="(1, 12, 64, 64)", style=solid]; -"1851 mul_16" -> "1882 add_28" [label="(1, 12, 64, 64)", style=solid]; -"1852 pad_10" -> "1853 view_46" [label="(1, 16, 16, 384)", style=solid]; -"1853 view_46" -> "1854 permute_38" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"1854 permute_38" -> "1855 reshape_36" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"1855 reshape_36" -> "1859 linear_52" [label="(4, 64, 384)", style=solid]; -"1856 clone_8" -> "1857 slice_133" [label="(1152,)", style=solid]; -"1856 clone_8" -> "1859 linear_52" [label="(1152,)", style=solid]; -"1857 slice_133" -> "1858 zero__8" [label="(384,)", style=solid]; -"1859 linear_52" -> "1860 reshape_37" [label="(4, 64, 1152)", style=solid]; -"1860 reshape_37" -> "1861 permute_39" [label="(4, 64, 3, 12, 32)", style=solid]; -"1861 permute_39" -> "1862 select_24" [label="(3, 4, 12, 64, 32)", style=solid]; -"1861 permute_39" -> "1863 select_25" [label="(3, 4, 12, 64, 32)", style=solid]; -"1861 permute_39" -> "1864 select_26" [label="(3, 4, 12, 64, 32)", style=solid]; -"1862 select_24" -> "1865 linalg_vector_norm_16" [label="(4, 12, 64, 32)", style=solid]; -"1862 select_24" -> "1867 expand_as_16" [label="(4, 12, 64, 32)", style=solid]; -"1862 select_24" -> "1868 div_16" [label="(4, 12, 64, 32)", style=solid]; -"1863 select_25" -> "1871 linalg_vector_norm_17" [label="(4, 12, 64, 32)", style=solid]; -"1863 select_25" -> "1873 expand_as_17" [label="(4, 12, 64, 32)", style=solid]; -"1863 select_25" -> "1874 div_17" [label="(4, 12, 64, 32)", style=solid]; -"1864 select_26" -> "1885 matmul_17" [label="(4, 12, 64, 32)", style=solid]; -"1865 linalg_vector_norm_16" -> "1866 clamp_min_16" [label="(4, 12, 64, 1)", style=solid]; -"1866 clamp_min_16" -> "1867 expand_as_16" [label="(4, 12, 64, 1)", style=solid]; -"1867 expand_as_16" -> "1868 div_16" [label="(4, 12, 64, 32)", style=solid]; -"1868 div_16" -> "1869 quantize_per_tensor_default_52" [label="(4, 12, 64, 32)", style=solid]; -"1869 quantize_per_tensor_default_52" -> "1870 dequantize_per_tensor_default_52" [label=None, style=solid]; -"1870 dequantize_per_tensor_default_52" -> "1878 matmul_16" [label=None, style=solid]; -"1871 linalg_vector_norm_17" -> "1872 clamp_min_17" [label="(4, 12, 64, 1)", style=solid]; -"1872 clamp_min_17" -> "1873 expand_as_17" [label="(4, 12, 64, 1)", style=solid]; -"1873 expand_as_17" -> "1874 div_17" [label="(4, 12, 64, 32)", style=solid]; -"1874 div_17" -> "1875 quantize_per_tensor_default_53" [label="(4, 12, 64, 32)", style=solid]; -"1875 quantize_per_tensor_default_53" -> "1876 dequantize_per_tensor_default_53" [label=None, style=solid]; -"1876 dequantize_per_tensor_default_53" -> "1877 transpose_16" [label=None, style=solid]; -"1877 transpose_16" -> "1878 matmul_16" [label="(4, 12, 32, 64)", style=solid]; -"1878 matmul_16" -> "1881 mul_17" [label="(4, 12, 64, 64)", style=solid]; -"1879 clamp_8" -> "1880 exp_8" [label="(12, 1, 1)", style=solid]; -"1880 exp_8" -> "1881 mul_17" [label="(12, 1, 1)", style=solid]; -"1881 mul_17" -> "1882 add_28" [label="(4, 12, 64, 64)", style=solid]; -"1882 add_28" -> "1883 softmax_8" [label="(4, 12, 64, 64)", style=solid]; -"1883 softmax_8" -> "1884 dropout_32" [label="(4, 12, 64, 64)", style=solid]; -"1884 dropout_32" -> "1885 matmul_17" [label="(4, 12, 64, 64)", style=solid]; -"1885 matmul_17" -> "1886 quantize_per_tensor_default_54" [label="(4, 12, 64, 32)", style=solid]; -"1886 quantize_per_tensor_default_54" -> "1887 dequantize_per_tensor_default_54" [label=None, style=solid]; -"1887 dequantize_per_tensor_default_54" -> "1888 transpose_17" [label=None, style=solid]; -"1888 transpose_17" -> "1889 reshape_38" [label="(4, 64, 12, 32)", style=solid]; -"1889 reshape_38" -> "1890 linear_53" [label="(4, 64, 384)", style=solid]; -"1890 linear_53" -> "1891 dropout_33" [label="(4, 64, 384)", style=solid]; -"1891 dropout_33" -> "1892 view_47" [label="(4, 64, 384)", style=solid]; -"1892 view_47" -> "1893 permute_40" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"1893 permute_40" -> "1894 reshape_39" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"1894 reshape_39" -> "1895 slice_134" [label="(1, 16, 16, 384)", style=solid]; -"1895 slice_134" -> "1896 slice_135" [label="(1, 16, 16, 384)", style=solid]; -"1896 slice_135" -> "1897 slice_136" [label="(1, 14, 16, 384)", style=solid]; -"1897 slice_136" -> "1898 slice_137" [label="(1, 14, 14, 384)", style=solid]; -"1898 slice_137" -> "1899 contiguous_15" [label="(1, 14, 14, 384)", style=solid]; -"1899 contiguous_15" -> "1900 layer_norm_19" [label="(1, 14, 14, 384)", style=solid]; -"1900 layer_norm_19" -> "1901 add_29" [label="(1, 14, 14, 384)", style=solid]; -"1901 add_29" -> "1902 quantize_per_tensor_default_55" [label="(1, 14, 14, 384)", style=solid]; -"1901 add_29" -> "1912 add_30" [label="(1, 14, 14, 384)", style=solid]; -"1902 quantize_per_tensor_default_55" -> "1903 dequantize_per_tensor_default_55" [label=None, style=solid]; -"1903 dequantize_per_tensor_default_55" -> "1904 linear_54" [label=None, style=solid]; -"1904 linear_54" -> "1905 gelu_8" [label="(1, 14, 14, 1536)", style=solid]; -"1905 gelu_8" -> "1906 quantize_per_tensor_default_56" [label="(1, 14, 14, 1536)", style=solid]; -"1906 quantize_per_tensor_default_56" -> "1907 dequantize_per_tensor_default_56" [label=None, style=solid]; -"1907 dequantize_per_tensor_default_56" -> "1908 dropout_34" [label=None, style=solid]; -"1908 dropout_34" -> "1909 linear_55" [label="(1, 14, 14, 1536)", style=solid]; -"1909 linear_55" -> "1910 dropout_35" [label="(1, 14, 14, 384)", style=solid]; -"1910 dropout_35" -> "1911 layer_norm_20" [label="(1, 14, 14, 384)", style=solid]; -"1911 layer_norm_20" -> "1912 add_30" [label="(1, 14, 14, 384)", style=solid]; -"1912 add_30" -> "1924 pad_11" [label="(1, 14, 14, 384)", style=solid]; -"1912 add_30" -> "2029 add_33" [label="(1, 14, 14, 384)", style=solid]; -"1913 linear_56" -> "1914 relu__9" [label="(1, 15, 15, 512)", style=solid]; -"1914 relu__9" -> "1915 linear_57" [label="(1, 15, 15, 512)", style=solid]; -"1915 linear_57" -> "1916 view_48" [label="(1, 15, 15, 12)", style=solid]; -"1916 view_48" -> "1917 index_9" [label="(225, 12)", style=solid]; -"1917 index_9" -> "1918 view_49" [label="(4096, 12)", style=solid]; -"1918 view_49" -> "1919 permute_41" [label="(64, 64, 12)", style=solid]; -"1919 permute_41" -> "1920 contiguous_16" [label="(12, 64, 64)", style=solid]; -"1920 contiguous_16" -> "1921 unsqueeze_25" [label="(12, 64, 64)", style=solid]; -"1921 unsqueeze_25" -> "1922 sigmoid_9" [label="(1, 12, 64, 64)", style=solid]; -"1922 sigmoid_9" -> "1923 mul_18" [label="(1, 12, 64, 64)", style=solid]; -"1923 mul_18" -> "1957 add_31" [label="(1, 12, 64, 64)", style=solid]; -"1924 pad_11" -> "1925 roll_8" [label="(1, 16, 16, 384)", style=solid]; -"1925 roll_8" -> "1926 view_50" [label="(1, 16, 16, 384)", style=solid]; -"1926 view_50" -> "1927 permute_42" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"1927 permute_42" -> "1928 reshape_40" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"1928 reshape_40" -> "1929 quantize_per_tensor_default_57" [label="(4, 64, 384)", style=solid]; -"1928 reshape_40" -> "1958 new_zeros_4" [label="(4, 64, 384)", style=solid]; -"1929 quantize_per_tensor_default_57" -> "1930 dequantize_per_tensor_default_57" [label=None, style=solid]; -"1930 dequantize_per_tensor_default_57" -> "1934 linear_58" [label=None, style=solid]; -"1931 clone_9" -> "1932 slice_138" [label="(1152,)", style=solid]; -"1931 clone_9" -> "1934 linear_58" [label="(1152,)", style=solid]; -"1932 slice_138" -> "1933 zero__9" [label="(384,)", style=solid]; -"1934 linear_58" -> "1935 reshape_41" [label="(4, 64, 1152)", style=solid]; -"1935 reshape_41" -> "1936 permute_43" [label="(4, 64, 3, 12, 32)", style=solid]; -"1936 permute_43" -> "1937 select_27" [label="(3, 4, 12, 64, 32)", style=solid]; -"1936 permute_43" -> "1938 select_28" [label="(3, 4, 12, 64, 32)", style=solid]; -"1936 permute_43" -> "1939 select_29" [label="(3, 4, 12, 64, 32)", style=solid]; -"1937 select_27" -> "1940 linalg_vector_norm_18" [label="(4, 12, 64, 32)", style=solid]; -"1937 select_27" -> "1942 expand_as_18" [label="(4, 12, 64, 32)", style=solid]; -"1937 select_27" -> "1943 div_18" [label="(4, 12, 64, 32)", style=solid]; -"1938 select_28" -> "1946 linalg_vector_norm_19" [label="(4, 12, 64, 32)", style=solid]; -"1938 select_28" -> "1948 expand_as_19" [label="(4, 12, 64, 32)", style=solid]; -"1938 select_28" -> "1949 div_19" [label="(4, 12, 64, 32)", style=solid]; -"1939 select_29" -> "2012 matmul_19" [label="(4, 12, 64, 32)", style=solid]; -"1940 linalg_vector_norm_18" -> "1941 clamp_min_18" [label="(4, 12, 64, 1)", style=solid]; -"1941 clamp_min_18" -> "1942 expand_as_18" [label="(4, 12, 64, 1)", style=solid]; -"1942 expand_as_18" -> "1943 div_18" [label="(4, 12, 64, 32)", style=solid]; -"1943 div_18" -> "1944 quantize_per_tensor_default_58" [label="(4, 12, 64, 32)", style=solid]; -"1944 quantize_per_tensor_default_58" -> "1945 dequantize_per_tensor_default_58" [label=None, style=solid]; -"1945 dequantize_per_tensor_default_58" -> "1953 matmul_18" [label=None, style=solid]; -"1946 linalg_vector_norm_19" -> "1947 clamp_min_19" [label="(4, 12, 64, 1)", style=solid]; -"1947 clamp_min_19" -> "1948 expand_as_19" [label="(4, 12, 64, 1)", style=solid]; -"1948 expand_as_19" -> "1949 div_19" [label="(4, 12, 64, 32)", style=solid]; -"1949 div_19" -> "1950 quantize_per_tensor_default_59" [label="(4, 12, 64, 32)", style=solid]; -"1950 quantize_per_tensor_default_59" -> "1951 dequantize_per_tensor_default_59" [label=None, style=solid]; -"1951 dequantize_per_tensor_default_59" -> "1952 transpose_18" [label=None, style=solid]; -"1952 transpose_18" -> "1953 matmul_18" [label="(4, 12, 32, 64)", style=solid]; -"1953 matmul_18" -> "1956 mul_19" [label="(4, 12, 64, 64)", style=solid]; -"1954 clamp_9" -> "1955 exp_9" [label="(12, 1, 1)", style=solid]; -"1955 exp_9" -> "1956 mul_19" [label="(12, 1, 1)", style=solid]; -"1956 mul_19" -> "1957 add_31" [label="(4, 12, 64, 64)", style=solid]; -"1957 add_31" -> "2005 view_52" [label="(4, 12, 64, 64)", style=solid]; -"1958 new_zeros_4" -> "1960 slice_139" [label="(16, 16)", style=solid]; -"1958 new_zeros_4" -> "1964 slice_141" [label="(16, 16)", style=solid]; -"1958 new_zeros_4" -> "1968 slice_143" [label="(16, 16)", style=solid]; -"1958 new_zeros_4" -> "1972 slice_145" [label="(16, 16)", style=solid]; -"1958 new_zeros_4" -> "1976 slice_147" [label="(16, 16)", style=solid]; -"1958 new_zeros_4" -> "1980 slice_149" [label="(16, 16)", style=solid]; -"1958 new_zeros_4" -> "1984 slice_151" [label="(16, 16)", style=solid]; -"1958 new_zeros_4" -> "1988 slice_153" [label="(16, 16)", style=solid]; -"1958 new_zeros_4" -> "1992 slice_155" [label="(16, 16)", style=solid]; -"1958 new_zeros_4" -> "1995 view_51" [label="(16, 16)", style=solid]; -"1959 lift_fresh_copy_36" -> "1962 fill__36" [label="()", style=solid]; -"1960 slice_139" -> "1961 slice_140" [label="(8, 16)", style=solid]; -"1961 slice_140" -> "1962 fill__36" [label="(8, 8)", style=solid]; -"1963 lift_fresh_copy_37" -> "1966 fill__37" [label="()", style=solid]; -"1964 slice_141" -> "1965 slice_142" [label="(8, 16)", style=solid]; -"1965 slice_142" -> "1966 fill__37" [label="(8, 4)", style=solid]; -"1967 lift_fresh_copy_38" -> "1970 fill__38" [label="()", style=solid]; -"1968 slice_143" -> "1969 slice_144" [label="(8, 16)", style=solid]; -"1969 slice_144" -> "1970 fill__38" [label="(8, 4)", style=solid]; -"1971 lift_fresh_copy_39" -> "1974 fill__39" [label="()", style=solid]; -"1972 slice_145" -> "1973 slice_146" [label="(4, 16)", style=solid]; -"1973 slice_146" -> "1974 fill__39" [label="(4, 8)", style=solid]; -"1975 lift_fresh_copy_40" -> "1978 fill__40" [label="()", style=solid]; -"1976 slice_147" -> "1977 slice_148" [label="(4, 16)", style=solid]; -"1977 slice_148" -> "1978 fill__40" [label="(4, 4)", style=solid]; -"1979 lift_fresh_copy_41" -> "1982 fill__41" [label="()", style=solid]; -"1980 slice_149" -> "1981 slice_150" [label="(4, 16)", style=solid]; -"1981 slice_150" -> "1982 fill__41" [label="(4, 4)", style=solid]; -"1983 lift_fresh_copy_42" -> "1986 fill__42" [label="()", style=solid]; -"1984 slice_151" -> "1985 slice_152" [label="(4, 16)", style=solid]; -"1985 slice_152" -> "1986 fill__42" [label="(4, 8)", style=solid]; -"1987 lift_fresh_copy_43" -> "1990 fill__43" [label="()", style=solid]; -"1988 slice_153" -> "1989 slice_154" [label="(4, 16)", style=solid]; -"1989 slice_154" -> "1990 fill__43" [label="(4, 4)", style=solid]; -"1991 lift_fresh_copy_44" -> "1994 fill__44" [label="()", style=solid]; -"1992 slice_155" -> "1993 slice_156" [label="(4, 16)", style=solid]; -"1993 slice_156" -> "1994 fill__44" [label="(4, 4)", style=solid]; -"1995 view_51" -> "1996 permute_44" [label="(2, 8, 2, 8)", style=solid]; -"1996 permute_44" -> "1997 reshape_42" [label="(2, 2, 8, 8)", style=solid]; -"1997 reshape_42" -> "1998 unsqueeze_26" [label="(4, 64)", style=solid]; -"1997 reshape_42" -> "1999 unsqueeze_27" [label="(4, 64)", style=solid]; -"1998 unsqueeze_26" -> "2000 sub_4" [label="(4, 1, 64)", style=solid]; -"1999 unsqueeze_27" -> "2000 sub_4" [label="(4, 64, 1)", style=solid]; -"2000 sub_4" -> "2001 ne_4" [label="(4, 64, 64)", style=solid]; -"2000 sub_4" -> "2002 masked_fill_8" [label="(4, 64, 64)", style=solid]; -"2000 sub_4" -> "2003 eq_4" [label="(4, 64, 64)", style=solid]; -"2001 ne_4" -> "2002 masked_fill_8" [label="(4, 64, 64)", style=solid]; -"2002 masked_fill_8" -> "2004 masked_fill_9" [label="(4, 64, 64)", style=solid]; -"2003 eq_4" -> "2004 masked_fill_9" [label="(4, 64, 64)", style=solid]; -"2004 masked_fill_9" -> "2006 unsqueeze_28" [label="(4, 64, 64)", style=solid]; -"2005 view_52" -> "2008 add_32" [label="(1, 4, 12, 64, 64)", style=solid]; -"2006 unsqueeze_28" -> "2007 unsqueeze_29" [label="(4, 1, 64, 64)", style=solid]; -"2007 unsqueeze_29" -> "2008 add_32" [label="(1, 4, 1, 64, 64)", style=solid]; -"2008 add_32" -> "2009 view_53" [label="(1, 4, 12, 64, 64)", style=solid]; -"2009 view_53" -> "2010 softmax_9" [label="(4, 12, 64, 64)", style=solid]; -"2010 softmax_9" -> "2011 dropout_36" [label="(4, 12, 64, 64)", style=solid]; -"2011 dropout_36" -> "2012 matmul_19" [label="(4, 12, 64, 64)", style=solid]; -"2012 matmul_19" -> "2013 quantize_per_tensor_default_60" [label="(4, 12, 64, 32)", style=solid]; -"2013 quantize_per_tensor_default_60" -> "2014 dequantize_per_tensor_default_60" [label=None, style=solid]; -"2014 dequantize_per_tensor_default_60" -> "2015 transpose_19" [label=None, style=solid]; -"2015 transpose_19" -> "2016 reshape_43" [label="(4, 64, 12, 32)", style=solid]; -"2016 reshape_43" -> "2017 linear_59" [label="(4, 64, 384)", style=solid]; -"2017 linear_59" -> "2018 dropout_37" [label="(4, 64, 384)", style=solid]; -"2018 dropout_37" -> "2019 view_54" [label="(4, 64, 384)", style=solid]; -"2019 view_54" -> "2020 permute_45" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2020 permute_45" -> "2021 reshape_44" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2021 reshape_44" -> "2022 roll_9" [label="(1, 16, 16, 384)", style=solid]; -"2022 roll_9" -> "2023 slice_157" [label="(1, 16, 16, 384)", style=solid]; -"2023 slice_157" -> "2024 slice_158" [label="(1, 16, 16, 384)", style=solid]; -"2024 slice_158" -> "2025 slice_159" [label="(1, 14, 16, 384)", style=solid]; -"2025 slice_159" -> "2026 slice_160" [label="(1, 14, 14, 384)", style=solid]; -"2026 slice_160" -> "2027 contiguous_17" [label="(1, 14, 14, 384)", style=solid]; -"2027 contiguous_17" -> "2028 layer_norm_21" [label="(1, 14, 14, 384)", style=solid]; -"2028 layer_norm_21" -> "2029 add_33" [label="(1, 14, 14, 384)", style=solid]; -"2029 add_33" -> "2030 quantize_per_tensor_default_61" [label="(1, 14, 14, 384)", style=solid]; -"2029 add_33" -> "2040 add_34" [label="(1, 14, 14, 384)", style=solid]; -"2030 quantize_per_tensor_default_61" -> "2031 dequantize_per_tensor_default_61" [label=None, style=solid]; -"2031 dequantize_per_tensor_default_61" -> "2032 linear_60" [label=None, style=solid]; -"2032 linear_60" -> "2033 gelu_9" [label="(1, 14, 14, 1536)", style=solid]; -"2033 gelu_9" -> "2034 quantize_per_tensor_default_62" [label="(1, 14, 14, 1536)", style=solid]; -"2034 quantize_per_tensor_default_62" -> "2035 dequantize_per_tensor_default_62" [label=None, style=solid]; -"2035 dequantize_per_tensor_default_62" -> "2036 dropout_38" [label=None, style=solid]; -"2036 dropout_38" -> "2037 linear_61" [label="(1, 14, 14, 1536)", style=solid]; -"2037 linear_61" -> "2038 dropout_39" [label="(1, 14, 14, 384)", style=solid]; -"2038 dropout_39" -> "2039 layer_norm_22" [label="(1, 14, 14, 384)", style=solid]; -"2039 layer_norm_22" -> "2040 add_34" [label="(1, 14, 14, 384)", style=solid]; -"2040 add_34" -> "2041 quantize_per_tensor_default_63" [label="(1, 14, 14, 384)", style=solid]; -"2040 add_34" -> "2103 add_36" [label="(1, 14, 14, 384)", style=solid]; -"2041 quantize_per_tensor_default_63" -> "2042 dequantize_per_tensor_default_63" [label=None, style=solid]; -"2042 dequantize_per_tensor_default_63" -> "2054 pad_12" [label=None, style=solid]; -"2043 linear_62" -> "2044 relu__10" [label="(1, 15, 15, 512)", style=solid]; -"2044 relu__10" -> "2045 linear_63" [label="(1, 15, 15, 512)", style=solid]; -"2045 linear_63" -> "2046 view_55" [label="(1, 15, 15, 12)", style=solid]; -"2046 view_55" -> "2047 index_10" [label="(225, 12)", style=solid]; -"2047 index_10" -> "2048 view_56" [label="(4096, 12)", style=solid]; -"2048 view_56" -> "2049 permute_46" [label="(64, 64, 12)", style=solid]; -"2049 permute_46" -> "2050 contiguous_18" [label="(12, 64, 64)", style=solid]; -"2050 contiguous_18" -> "2051 unsqueeze_30" [label="(12, 64, 64)", style=solid]; -"2051 unsqueeze_30" -> "2052 sigmoid_10" [label="(1, 12, 64, 64)", style=solid]; -"2052 sigmoid_10" -> "2053 mul_20" [label="(1, 12, 64, 64)", style=solid]; -"2053 mul_20" -> "2084 add_35" [label="(1, 12, 64, 64)", style=solid]; -"2054 pad_12" -> "2055 view_57" [label="(1, 16, 16, 384)", style=solid]; -"2055 view_57" -> "2056 permute_47" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2056 permute_47" -> "2057 reshape_45" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2057 reshape_45" -> "2061 linear_64" [label="(4, 64, 384)", style=solid]; -"2058 clone_10" -> "2059 slice_161" [label="(1152,)", style=solid]; -"2058 clone_10" -> "2061 linear_64" [label="(1152,)", style=solid]; -"2059 slice_161" -> "2060 zero__10" [label="(384,)", style=solid]; -"2061 linear_64" -> "2062 reshape_46" [label="(4, 64, 1152)", style=solid]; -"2062 reshape_46" -> "2063 permute_48" [label="(4, 64, 3, 12, 32)", style=solid]; -"2063 permute_48" -> "2064 select_30" [label="(3, 4, 12, 64, 32)", style=solid]; -"2063 permute_48" -> "2065 select_31" [label="(3, 4, 12, 64, 32)", style=solid]; -"2063 permute_48" -> "2066 select_32" [label="(3, 4, 12, 64, 32)", style=solid]; -"2064 select_30" -> "2067 linalg_vector_norm_20" [label="(4, 12, 64, 32)", style=solid]; -"2064 select_30" -> "2069 expand_as_20" [label="(4, 12, 64, 32)", style=solid]; -"2064 select_30" -> "2070 div_20" [label="(4, 12, 64, 32)", style=solid]; -"2065 select_31" -> "2073 linalg_vector_norm_21" [label="(4, 12, 64, 32)", style=solid]; -"2065 select_31" -> "2075 expand_as_21" [label="(4, 12, 64, 32)", style=solid]; -"2065 select_31" -> "2076 div_21" [label="(4, 12, 64, 32)", style=solid]; -"2066 select_32" -> "2087 matmul_21" [label="(4, 12, 64, 32)", style=solid]; -"2067 linalg_vector_norm_20" -> "2068 clamp_min_20" [label="(4, 12, 64, 1)", style=solid]; -"2068 clamp_min_20" -> "2069 expand_as_20" [label="(4, 12, 64, 1)", style=solid]; -"2069 expand_as_20" -> "2070 div_20" [label="(4, 12, 64, 32)", style=solid]; -"2070 div_20" -> "2071 quantize_per_tensor_default_64" [label="(4, 12, 64, 32)", style=solid]; -"2071 quantize_per_tensor_default_64" -> "2072 dequantize_per_tensor_default_64" [label=None, style=solid]; -"2072 dequantize_per_tensor_default_64" -> "2080 matmul_20" [label=None, style=solid]; -"2073 linalg_vector_norm_21" -> "2074 clamp_min_21" [label="(4, 12, 64, 1)", style=solid]; -"2074 clamp_min_21" -> "2075 expand_as_21" [label="(4, 12, 64, 1)", style=solid]; -"2075 expand_as_21" -> "2076 div_21" [label="(4, 12, 64, 32)", style=solid]; -"2076 div_21" -> "2077 quantize_per_tensor_default_65" [label="(4, 12, 64, 32)", style=solid]; -"2077 quantize_per_tensor_default_65" -> "2078 dequantize_per_tensor_default_65" [label=None, style=solid]; -"2078 dequantize_per_tensor_default_65" -> "2079 transpose_20" [label=None, style=solid]; -"2079 transpose_20" -> "2080 matmul_20" [label="(4, 12, 32, 64)", style=solid]; -"2080 matmul_20" -> "2083 mul_21" [label="(4, 12, 64, 64)", style=solid]; -"2081 clamp_10" -> "2082 exp_10" [label="(12, 1, 1)", style=solid]; -"2082 exp_10" -> "2083 mul_21" [label="(12, 1, 1)", style=solid]; -"2083 mul_21" -> "2084 add_35" [label="(4, 12, 64, 64)", style=solid]; -"2084 add_35" -> "2085 softmax_10" [label="(4, 12, 64, 64)", style=solid]; -"2085 softmax_10" -> "2086 dropout_40" [label="(4, 12, 64, 64)", style=solid]; -"2086 dropout_40" -> "2087 matmul_21" [label="(4, 12, 64, 64)", style=solid]; -"2087 matmul_21" -> "2088 quantize_per_tensor_default_66" [label="(4, 12, 64, 32)", style=solid]; -"2088 quantize_per_tensor_default_66" -> "2089 dequantize_per_tensor_default_66" [label=None, style=solid]; -"2089 dequantize_per_tensor_default_66" -> "2090 transpose_21" [label=None, style=solid]; -"2090 transpose_21" -> "2091 reshape_47" [label="(4, 64, 12, 32)", style=solid]; -"2091 reshape_47" -> "2092 linear_65" [label="(4, 64, 384)", style=solid]; -"2092 linear_65" -> "2093 dropout_41" [label="(4, 64, 384)", style=solid]; -"2093 dropout_41" -> "2094 view_58" [label="(4, 64, 384)", style=solid]; -"2094 view_58" -> "2095 permute_49" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2095 permute_49" -> "2096 reshape_48" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2096 reshape_48" -> "2097 slice_162" [label="(1, 16, 16, 384)", style=solid]; -"2097 slice_162" -> "2098 slice_163" [label="(1, 16, 16, 384)", style=solid]; -"2098 slice_163" -> "2099 slice_164" [label="(1, 14, 16, 384)", style=solid]; -"2099 slice_164" -> "2100 slice_165" [label="(1, 14, 14, 384)", style=solid]; -"2100 slice_165" -> "2101 contiguous_19" [label="(1, 14, 14, 384)", style=solid]; -"2101 contiguous_19" -> "2102 layer_norm_23" [label="(1, 14, 14, 384)", style=solid]; -"2102 layer_norm_23" -> "2103 add_36" [label="(1, 14, 14, 384)", style=solid]; -"2103 add_36" -> "2104 quantize_per_tensor_default_67" [label="(1, 14, 14, 384)", style=solid]; -"2103 add_36" -> "2114 add_37" [label="(1, 14, 14, 384)", style=solid]; -"2104 quantize_per_tensor_default_67" -> "2105 dequantize_per_tensor_default_67" [label=None, style=solid]; -"2105 dequantize_per_tensor_default_67" -> "2106 linear_66" [label=None, style=solid]; -"2106 linear_66" -> "2107 gelu_10" [label="(1, 14, 14, 1536)", style=solid]; -"2107 gelu_10" -> "2108 quantize_per_tensor_default_68" [label="(1, 14, 14, 1536)", style=solid]; -"2108 quantize_per_tensor_default_68" -> "2109 dequantize_per_tensor_default_68" [label=None, style=solid]; -"2109 dequantize_per_tensor_default_68" -> "2110 dropout_42" [label=None, style=solid]; -"2110 dropout_42" -> "2111 linear_67" [label="(1, 14, 14, 1536)", style=solid]; -"2111 linear_67" -> "2112 dropout_43" [label="(1, 14, 14, 384)", style=solid]; -"2112 dropout_43" -> "2113 layer_norm_24" [label="(1, 14, 14, 384)", style=solid]; -"2113 layer_norm_24" -> "2114 add_37" [label="(1, 14, 14, 384)", style=solid]; -"2114 add_37" -> "2126 pad_13" [label="(1, 14, 14, 384)", style=solid]; -"2114 add_37" -> "2231 add_40" [label="(1, 14, 14, 384)", style=solid]; -"2115 linear_68" -> "2116 relu__11" [label="(1, 15, 15, 512)", style=solid]; -"2116 relu__11" -> "2117 linear_69" [label="(1, 15, 15, 512)", style=solid]; -"2117 linear_69" -> "2118 view_59" [label="(1, 15, 15, 12)", style=solid]; -"2118 view_59" -> "2119 index_11" [label="(225, 12)", style=solid]; -"2119 index_11" -> "2120 view_60" [label="(4096, 12)", style=solid]; -"2120 view_60" -> "2121 permute_50" [label="(64, 64, 12)", style=solid]; -"2121 permute_50" -> "2122 contiguous_20" [label="(12, 64, 64)", style=solid]; -"2122 contiguous_20" -> "2123 unsqueeze_31" [label="(12, 64, 64)", style=solid]; -"2123 unsqueeze_31" -> "2124 sigmoid_11" [label="(1, 12, 64, 64)", style=solid]; -"2124 sigmoid_11" -> "2125 mul_22" [label="(1, 12, 64, 64)", style=solid]; -"2125 mul_22" -> "2159 add_38" [label="(1, 12, 64, 64)", style=solid]; -"2126 pad_13" -> "2127 roll_10" [label="(1, 16, 16, 384)", style=solid]; -"2127 roll_10" -> "2128 view_61" [label="(1, 16, 16, 384)", style=solid]; -"2128 view_61" -> "2129 permute_51" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2129 permute_51" -> "2130 reshape_49" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2130 reshape_49" -> "2131 quantize_per_tensor_default_69" [label="(4, 64, 384)", style=solid]; -"2130 reshape_49" -> "2160 new_zeros_5" [label="(4, 64, 384)", style=solid]; -"2131 quantize_per_tensor_default_69" -> "2132 dequantize_per_tensor_default_69" [label=None, style=solid]; -"2132 dequantize_per_tensor_default_69" -> "2136 linear_70" [label=None, style=solid]; -"2133 clone_11" -> "2134 slice_166" [label="(1152,)", style=solid]; -"2133 clone_11" -> "2136 linear_70" [label="(1152,)", style=solid]; -"2134 slice_166" -> "2135 zero__11" [label="(384,)", style=solid]; -"2136 linear_70" -> "2137 reshape_50" [label="(4, 64, 1152)", style=solid]; -"2137 reshape_50" -> "2138 permute_52" [label="(4, 64, 3, 12, 32)", style=solid]; -"2138 permute_52" -> "2139 select_33" [label="(3, 4, 12, 64, 32)", style=solid]; -"2138 permute_52" -> "2140 select_34" [label="(3, 4, 12, 64, 32)", style=solid]; -"2138 permute_52" -> "2141 select_35" [label="(3, 4, 12, 64, 32)", style=solid]; -"2139 select_33" -> "2142 linalg_vector_norm_22" [label="(4, 12, 64, 32)", style=solid]; -"2139 select_33" -> "2144 expand_as_22" [label="(4, 12, 64, 32)", style=solid]; -"2139 select_33" -> "2145 div_22" [label="(4, 12, 64, 32)", style=solid]; -"2140 select_34" -> "2148 linalg_vector_norm_23" [label="(4, 12, 64, 32)", style=solid]; -"2140 select_34" -> "2150 expand_as_23" [label="(4, 12, 64, 32)", style=solid]; -"2140 select_34" -> "2151 div_23" [label="(4, 12, 64, 32)", style=solid]; -"2141 select_35" -> "2214 matmul_23" [label="(4, 12, 64, 32)", style=solid]; -"2142 linalg_vector_norm_22" -> "2143 clamp_min_22" [label="(4, 12, 64, 1)", style=solid]; -"2143 clamp_min_22" -> "2144 expand_as_22" [label="(4, 12, 64, 1)", style=solid]; -"2144 expand_as_22" -> "2145 div_22" [label="(4, 12, 64, 32)", style=solid]; -"2145 div_22" -> "2146 quantize_per_tensor_default_70" [label="(4, 12, 64, 32)", style=solid]; -"2146 quantize_per_tensor_default_70" -> "2147 dequantize_per_tensor_default_70" [label=None, style=solid]; -"2147 dequantize_per_tensor_default_70" -> "2155 matmul_22" [label=None, style=solid]; -"2148 linalg_vector_norm_23" -> "2149 clamp_min_23" [label="(4, 12, 64, 1)", style=solid]; -"2149 clamp_min_23" -> "2150 expand_as_23" [label="(4, 12, 64, 1)", style=solid]; -"2150 expand_as_23" -> "2151 div_23" [label="(4, 12, 64, 32)", style=solid]; -"2151 div_23" -> "2152 quantize_per_tensor_default_71" [label="(4, 12, 64, 32)", style=solid]; -"2152 quantize_per_tensor_default_71" -> "2153 dequantize_per_tensor_default_71" [label=None, style=solid]; -"2153 dequantize_per_tensor_default_71" -> "2154 transpose_22" [label=None, style=solid]; -"2154 transpose_22" -> "2155 matmul_22" [label="(4, 12, 32, 64)", style=solid]; -"2155 matmul_22" -> "2158 mul_23" [label="(4, 12, 64, 64)", style=solid]; -"2156 clamp_11" -> "2157 exp_11" [label="(12, 1, 1)", style=solid]; -"2157 exp_11" -> "2158 mul_23" [label="(12, 1, 1)", style=solid]; -"2158 mul_23" -> "2159 add_38" [label="(4, 12, 64, 64)", style=solid]; -"2159 add_38" -> "2207 view_63" [label="(4, 12, 64, 64)", style=solid]; -"2160 new_zeros_5" -> "2162 slice_167" [label="(16, 16)", style=solid]; -"2160 new_zeros_5" -> "2166 slice_169" [label="(16, 16)", style=solid]; -"2160 new_zeros_5" -> "2170 slice_171" [label="(16, 16)", style=solid]; -"2160 new_zeros_5" -> "2174 slice_173" [label="(16, 16)", style=solid]; -"2160 new_zeros_5" -> "2178 slice_175" [label="(16, 16)", style=solid]; -"2160 new_zeros_5" -> "2182 slice_177" [label="(16, 16)", style=solid]; -"2160 new_zeros_5" -> "2186 slice_179" [label="(16, 16)", style=solid]; -"2160 new_zeros_5" -> "2190 slice_181" [label="(16, 16)", style=solid]; -"2160 new_zeros_5" -> "2194 slice_183" [label="(16, 16)", style=solid]; -"2160 new_zeros_5" -> "2197 view_62" [label="(16, 16)", style=solid]; -"2161 lift_fresh_copy_45" -> "2164 fill__45" [label="()", style=solid]; -"2162 slice_167" -> "2163 slice_168" [label="(8, 16)", style=solid]; -"2163 slice_168" -> "2164 fill__45" [label="(8, 8)", style=solid]; -"2165 lift_fresh_copy_46" -> "2168 fill__46" [label="()", style=solid]; -"2166 slice_169" -> "2167 slice_170" [label="(8, 16)", style=solid]; -"2167 slice_170" -> "2168 fill__46" [label="(8, 4)", style=solid]; -"2169 lift_fresh_copy_47" -> "2172 fill__47" [label="()", style=solid]; -"2170 slice_171" -> "2171 slice_172" [label="(8, 16)", style=solid]; -"2171 slice_172" -> "2172 fill__47" [label="(8, 4)", style=solid]; -"2173 lift_fresh_copy_48" -> "2176 fill__48" [label="()", style=solid]; -"2174 slice_173" -> "2175 slice_174" [label="(4, 16)", style=solid]; -"2175 slice_174" -> "2176 fill__48" [label="(4, 8)", style=solid]; -"2177 lift_fresh_copy_49" -> "2180 fill__49" [label="()", style=solid]; -"2178 slice_175" -> "2179 slice_176" [label="(4, 16)", style=solid]; -"2179 slice_176" -> "2180 fill__49" [label="(4, 4)", style=solid]; -"2181 lift_fresh_copy_50" -> "2184 fill__50" [label="()", style=solid]; -"2182 slice_177" -> "2183 slice_178" [label="(4, 16)", style=solid]; -"2183 slice_178" -> "2184 fill__50" [label="(4, 4)", style=solid]; -"2185 lift_fresh_copy_51" -> "2188 fill__51" [label="()", style=solid]; -"2186 slice_179" -> "2187 slice_180" [label="(4, 16)", style=solid]; -"2187 slice_180" -> "2188 fill__51" [label="(4, 8)", style=solid]; -"2189 lift_fresh_copy_52" -> "2192 fill__52" [label="()", style=solid]; -"2190 slice_181" -> "2191 slice_182" [label="(4, 16)", style=solid]; -"2191 slice_182" -> "2192 fill__52" [label="(4, 4)", style=solid]; -"2193 lift_fresh_copy_53" -> "2196 fill__53" [label="()", style=solid]; -"2194 slice_183" -> "2195 slice_184" [label="(4, 16)", style=solid]; -"2195 slice_184" -> "2196 fill__53" [label="(4, 4)", style=solid]; -"2197 view_62" -> "2198 permute_53" [label="(2, 8, 2, 8)", style=solid]; -"2198 permute_53" -> "2199 reshape_51" [label="(2, 2, 8, 8)", style=solid]; -"2199 reshape_51" -> "2200 unsqueeze_32" [label="(4, 64)", style=solid]; -"2199 reshape_51" -> "2201 unsqueeze_33" [label="(4, 64)", style=solid]; -"2200 unsqueeze_32" -> "2202 sub_5" [label="(4, 1, 64)", style=solid]; -"2201 unsqueeze_33" -> "2202 sub_5" [label="(4, 64, 1)", style=solid]; -"2202 sub_5" -> "2203 ne_5" [label="(4, 64, 64)", style=solid]; -"2202 sub_5" -> "2204 masked_fill_10" [label="(4, 64, 64)", style=solid]; -"2202 sub_5" -> "2205 eq_5" [label="(4, 64, 64)", style=solid]; -"2203 ne_5" -> "2204 masked_fill_10" [label="(4, 64, 64)", style=solid]; -"2204 masked_fill_10" -> "2206 masked_fill_11" [label="(4, 64, 64)", style=solid]; -"2205 eq_5" -> "2206 masked_fill_11" [label="(4, 64, 64)", style=solid]; -"2206 masked_fill_11" -> "2208 unsqueeze_34" [label="(4, 64, 64)", style=solid]; -"2207 view_63" -> "2210 add_39" [label="(1, 4, 12, 64, 64)", style=solid]; -"2208 unsqueeze_34" -> "2209 unsqueeze_35" [label="(4, 1, 64, 64)", style=solid]; -"2209 unsqueeze_35" -> "2210 add_39" [label="(1, 4, 1, 64, 64)", style=solid]; -"2210 add_39" -> "2211 view_64" [label="(1, 4, 12, 64, 64)", style=solid]; -"2211 view_64" -> "2212 softmax_11" [label="(4, 12, 64, 64)", style=solid]; -"2212 softmax_11" -> "2213 dropout_44" [label="(4, 12, 64, 64)", style=solid]; -"2213 dropout_44" -> "2214 matmul_23" [label="(4, 12, 64, 64)", style=solid]; -"2214 matmul_23" -> "2215 quantize_per_tensor_default_72" [label="(4, 12, 64, 32)", style=solid]; -"2215 quantize_per_tensor_default_72" -> "2216 dequantize_per_tensor_default_72" [label=None, style=solid]; -"2216 dequantize_per_tensor_default_72" -> "2217 transpose_23" [label=None, style=solid]; -"2217 transpose_23" -> "2218 reshape_52" [label="(4, 64, 12, 32)", style=solid]; -"2218 reshape_52" -> "2219 linear_71" [label="(4, 64, 384)", style=solid]; -"2219 linear_71" -> "2220 dropout_45" [label="(4, 64, 384)", style=solid]; -"2220 dropout_45" -> "2221 view_65" [label="(4, 64, 384)", style=solid]; -"2221 view_65" -> "2222 permute_54" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2222 permute_54" -> "2223 reshape_53" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2223 reshape_53" -> "2224 roll_11" [label="(1, 16, 16, 384)", style=solid]; -"2224 roll_11" -> "2225 slice_185" [label="(1, 16, 16, 384)", style=solid]; -"2225 slice_185" -> "2226 slice_186" [label="(1, 16, 16, 384)", style=solid]; -"2226 slice_186" -> "2227 slice_187" [label="(1, 14, 16, 384)", style=solid]; -"2227 slice_187" -> "2228 slice_188" [label="(1, 14, 14, 384)", style=solid]; -"2228 slice_188" -> "2229 contiguous_21" [label="(1, 14, 14, 384)", style=solid]; -"2229 contiguous_21" -> "2230 layer_norm_25" [label="(1, 14, 14, 384)", style=solid]; -"2230 layer_norm_25" -> "2231 add_40" [label="(1, 14, 14, 384)", style=solid]; -"2231 add_40" -> "2232 quantize_per_tensor_default_73" [label="(1, 14, 14, 384)", style=solid]; -"2231 add_40" -> "2242 add_41" [label="(1, 14, 14, 384)", style=solid]; -"2232 quantize_per_tensor_default_73" -> "2233 dequantize_per_tensor_default_73" [label=None, style=solid]; -"2233 dequantize_per_tensor_default_73" -> "2234 linear_72" [label=None, style=solid]; -"2234 linear_72" -> "2235 gelu_11" [label="(1, 14, 14, 1536)", style=solid]; -"2235 gelu_11" -> "2236 quantize_per_tensor_default_74" [label="(1, 14, 14, 1536)", style=solid]; -"2236 quantize_per_tensor_default_74" -> "2237 dequantize_per_tensor_default_74" [label=None, style=solid]; -"2237 dequantize_per_tensor_default_74" -> "2238 dropout_46" [label=None, style=solid]; -"2238 dropout_46" -> "2239 linear_73" [label="(1, 14, 14, 1536)", style=solid]; -"2239 linear_73" -> "2240 dropout_47" [label="(1, 14, 14, 384)", style=solid]; -"2240 dropout_47" -> "2241 layer_norm_26" [label="(1, 14, 14, 384)", style=solid]; -"2241 layer_norm_26" -> "2242 add_41" [label="(1, 14, 14, 384)", style=solid]; -"2242 add_41" -> "2243 quantize_per_tensor_default_75" [label="(1, 14, 14, 384)", style=solid]; -"2242 add_41" -> "2305 add_43" [label="(1, 14, 14, 384)", style=solid]; -"2243 quantize_per_tensor_default_75" -> "2244 dequantize_per_tensor_default_75" [label=None, style=solid]; -"2244 dequantize_per_tensor_default_75" -> "2256 pad_14" [label=None, style=solid]; -"2245 linear_74" -> "2246 relu__12" [label="(1, 15, 15, 512)", style=solid]; -"2246 relu__12" -> "2247 linear_75" [label="(1, 15, 15, 512)", style=solid]; -"2247 linear_75" -> "2248 view_66" [label="(1, 15, 15, 12)", style=solid]; -"2248 view_66" -> "2249 index_12" [label="(225, 12)", style=solid]; -"2249 index_12" -> "2250 view_67" [label="(4096, 12)", style=solid]; -"2250 view_67" -> "2251 permute_55" [label="(64, 64, 12)", style=solid]; -"2251 permute_55" -> "2252 contiguous_22" [label="(12, 64, 64)", style=solid]; -"2252 contiguous_22" -> "2253 unsqueeze_36" [label="(12, 64, 64)", style=solid]; -"2253 unsqueeze_36" -> "2254 sigmoid_12" [label="(1, 12, 64, 64)", style=solid]; -"2254 sigmoid_12" -> "2255 mul_24" [label="(1, 12, 64, 64)", style=solid]; -"2255 mul_24" -> "2286 add_42" [label="(1, 12, 64, 64)", style=solid]; -"2256 pad_14" -> "2257 view_68" [label="(1, 16, 16, 384)", style=solid]; -"2257 view_68" -> "2258 permute_56" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2258 permute_56" -> "2259 reshape_54" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2259 reshape_54" -> "2263 linear_76" [label="(4, 64, 384)", style=solid]; -"2260 clone_12" -> "2261 slice_189" [label="(1152,)", style=solid]; -"2260 clone_12" -> "2263 linear_76" [label="(1152,)", style=solid]; -"2261 slice_189" -> "2262 zero__12" [label="(384,)", style=solid]; -"2263 linear_76" -> "2264 reshape_55" [label="(4, 64, 1152)", style=solid]; -"2264 reshape_55" -> "2265 permute_57" [label="(4, 64, 3, 12, 32)", style=solid]; -"2265 permute_57" -> "2266 select_36" [label="(3, 4, 12, 64, 32)", style=solid]; -"2265 permute_57" -> "2267 select_37" [label="(3, 4, 12, 64, 32)", style=solid]; -"2265 permute_57" -> "2268 select_38" [label="(3, 4, 12, 64, 32)", style=solid]; -"2266 select_36" -> "2269 linalg_vector_norm_24" [label="(4, 12, 64, 32)", style=solid]; -"2266 select_36" -> "2271 expand_as_24" [label="(4, 12, 64, 32)", style=solid]; -"2266 select_36" -> "2272 div_24" [label="(4, 12, 64, 32)", style=solid]; -"2267 select_37" -> "2275 linalg_vector_norm_25" [label="(4, 12, 64, 32)", style=solid]; -"2267 select_37" -> "2277 expand_as_25" [label="(4, 12, 64, 32)", style=solid]; -"2267 select_37" -> "2278 div_25" [label="(4, 12, 64, 32)", style=solid]; -"2268 select_38" -> "2289 matmul_25" [label="(4, 12, 64, 32)", style=solid]; -"2269 linalg_vector_norm_24" -> "2270 clamp_min_24" [label="(4, 12, 64, 1)", style=solid]; -"2270 clamp_min_24" -> "2271 expand_as_24" [label="(4, 12, 64, 1)", style=solid]; -"2271 expand_as_24" -> "2272 div_24" [label="(4, 12, 64, 32)", style=solid]; -"2272 div_24" -> "2273 quantize_per_tensor_default_76" [label="(4, 12, 64, 32)", style=solid]; -"2273 quantize_per_tensor_default_76" -> "2274 dequantize_per_tensor_default_76" [label=None, style=solid]; -"2274 dequantize_per_tensor_default_76" -> "2282 matmul_24" [label=None, style=solid]; -"2275 linalg_vector_norm_25" -> "2276 clamp_min_25" [label="(4, 12, 64, 1)", style=solid]; -"2276 clamp_min_25" -> "2277 expand_as_25" [label="(4, 12, 64, 1)", style=solid]; -"2277 expand_as_25" -> "2278 div_25" [label="(4, 12, 64, 32)", style=solid]; -"2278 div_25" -> "2279 quantize_per_tensor_default_77" [label="(4, 12, 64, 32)", style=solid]; -"2279 quantize_per_tensor_default_77" -> "2280 dequantize_per_tensor_default_77" [label=None, style=solid]; -"2280 dequantize_per_tensor_default_77" -> "2281 transpose_24" [label=None, style=solid]; -"2281 transpose_24" -> "2282 matmul_24" [label="(4, 12, 32, 64)", style=solid]; -"2282 matmul_24" -> "2285 mul_25" [label="(4, 12, 64, 64)", style=solid]; -"2283 clamp_12" -> "2284 exp_12" [label="(12, 1, 1)", style=solid]; -"2284 exp_12" -> "2285 mul_25" [label="(12, 1, 1)", style=solid]; -"2285 mul_25" -> "2286 add_42" [label="(4, 12, 64, 64)", style=solid]; -"2286 add_42" -> "2287 softmax_12" [label="(4, 12, 64, 64)", style=solid]; -"2287 softmax_12" -> "2288 dropout_48" [label="(4, 12, 64, 64)", style=solid]; -"2288 dropout_48" -> "2289 matmul_25" [label="(4, 12, 64, 64)", style=solid]; -"2289 matmul_25" -> "2290 quantize_per_tensor_default_78" [label="(4, 12, 64, 32)", style=solid]; -"2290 quantize_per_tensor_default_78" -> "2291 dequantize_per_tensor_default_78" [label=None, style=solid]; -"2291 dequantize_per_tensor_default_78" -> "2292 transpose_25" [label=None, style=solid]; -"2292 transpose_25" -> "2293 reshape_56" [label="(4, 64, 12, 32)", style=solid]; -"2293 reshape_56" -> "2294 linear_77" [label="(4, 64, 384)", style=solid]; -"2294 linear_77" -> "2295 dropout_49" [label="(4, 64, 384)", style=solid]; -"2295 dropout_49" -> "2296 view_69" [label="(4, 64, 384)", style=solid]; -"2296 view_69" -> "2297 permute_58" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2297 permute_58" -> "2298 reshape_57" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2298 reshape_57" -> "2299 slice_190" [label="(1, 16, 16, 384)", style=solid]; -"2299 slice_190" -> "2300 slice_191" [label="(1, 16, 16, 384)", style=solid]; -"2300 slice_191" -> "2301 slice_192" [label="(1, 14, 16, 384)", style=solid]; -"2301 slice_192" -> "2302 slice_193" [label="(1, 14, 14, 384)", style=solid]; -"2302 slice_193" -> "2303 contiguous_23" [label="(1, 14, 14, 384)", style=solid]; -"2303 contiguous_23" -> "2304 layer_norm_27" [label="(1, 14, 14, 384)", style=solid]; -"2304 layer_norm_27" -> "2305 add_43" [label="(1, 14, 14, 384)", style=solid]; -"2305 add_43" -> "2306 quantize_per_tensor_default_79" [label="(1, 14, 14, 384)", style=solid]; -"2305 add_43" -> "2316 add_44" [label="(1, 14, 14, 384)", style=solid]; -"2306 quantize_per_tensor_default_79" -> "2307 dequantize_per_tensor_default_79" [label=None, style=solid]; -"2307 dequantize_per_tensor_default_79" -> "2308 linear_78" [label=None, style=solid]; -"2308 linear_78" -> "2309 gelu_12" [label="(1, 14, 14, 1536)", style=solid]; -"2309 gelu_12" -> "2310 quantize_per_tensor_default_80" [label="(1, 14, 14, 1536)", style=solid]; -"2310 quantize_per_tensor_default_80" -> "2311 dequantize_per_tensor_default_80" [label=None, style=solid]; -"2311 dequantize_per_tensor_default_80" -> "2312 dropout_50" [label=None, style=solid]; -"2312 dropout_50" -> "2313 linear_79" [label="(1, 14, 14, 1536)", style=solid]; -"2313 linear_79" -> "2314 dropout_51" [label="(1, 14, 14, 384)", style=solid]; -"2314 dropout_51" -> "2315 layer_norm_28" [label="(1, 14, 14, 384)", style=solid]; -"2315 layer_norm_28" -> "2316 add_44" [label="(1, 14, 14, 384)", style=solid]; -"2316 add_44" -> "2328 pad_15" [label="(1, 14, 14, 384)", style=solid]; -"2316 add_44" -> "2433 add_47" [label="(1, 14, 14, 384)", style=solid]; -"2317 linear_80" -> "2318 relu__13" [label="(1, 15, 15, 512)", style=solid]; -"2318 relu__13" -> "2319 linear_81" [label="(1, 15, 15, 512)", style=solid]; -"2319 linear_81" -> "2320 view_70" [label="(1, 15, 15, 12)", style=solid]; -"2320 view_70" -> "2321 index_13" [label="(225, 12)", style=solid]; -"2321 index_13" -> "2322 view_71" [label="(4096, 12)", style=solid]; -"2322 view_71" -> "2323 permute_59" [label="(64, 64, 12)", style=solid]; -"2323 permute_59" -> "2324 contiguous_24" [label="(12, 64, 64)", style=solid]; -"2324 contiguous_24" -> "2325 unsqueeze_37" [label="(12, 64, 64)", style=solid]; -"2325 unsqueeze_37" -> "2326 sigmoid_13" [label="(1, 12, 64, 64)", style=solid]; -"2326 sigmoid_13" -> "2327 mul_26" [label="(1, 12, 64, 64)", style=solid]; -"2327 mul_26" -> "2361 add_45" [label="(1, 12, 64, 64)", style=solid]; -"2328 pad_15" -> "2329 roll_12" [label="(1, 16, 16, 384)", style=solid]; -"2329 roll_12" -> "2330 view_72" [label="(1, 16, 16, 384)", style=solid]; -"2330 view_72" -> "2331 permute_60" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2331 permute_60" -> "2332 reshape_58" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2332 reshape_58" -> "2333 quantize_per_tensor_default_81" [label="(4, 64, 384)", style=solid]; -"2332 reshape_58" -> "2362 new_zeros_6" [label="(4, 64, 384)", style=solid]; -"2333 quantize_per_tensor_default_81" -> "2334 dequantize_per_tensor_default_81" [label=None, style=solid]; -"2334 dequantize_per_tensor_default_81" -> "2338 linear_82" [label=None, style=solid]; -"2335 clone_13" -> "2336 slice_194" [label="(1152,)", style=solid]; -"2335 clone_13" -> "2338 linear_82" [label="(1152,)", style=solid]; -"2336 slice_194" -> "2337 zero__13" [label="(384,)", style=solid]; -"2338 linear_82" -> "2339 reshape_59" [label="(4, 64, 1152)", style=solid]; -"2339 reshape_59" -> "2340 permute_61" [label="(4, 64, 3, 12, 32)", style=solid]; -"2340 permute_61" -> "2341 select_39" [label="(3, 4, 12, 64, 32)", style=solid]; -"2340 permute_61" -> "2342 select_40" [label="(3, 4, 12, 64, 32)", style=solid]; -"2340 permute_61" -> "2343 select_41" [label="(3, 4, 12, 64, 32)", style=solid]; -"2341 select_39" -> "2344 linalg_vector_norm_26" [label="(4, 12, 64, 32)", style=solid]; -"2341 select_39" -> "2346 expand_as_26" [label="(4, 12, 64, 32)", style=solid]; -"2341 select_39" -> "2347 div_26" [label="(4, 12, 64, 32)", style=solid]; -"2342 select_40" -> "2350 linalg_vector_norm_27" [label="(4, 12, 64, 32)", style=solid]; -"2342 select_40" -> "2352 expand_as_27" [label="(4, 12, 64, 32)", style=solid]; -"2342 select_40" -> "2353 div_27" [label="(4, 12, 64, 32)", style=solid]; -"2343 select_41" -> "2416 matmul_27" [label="(4, 12, 64, 32)", style=solid]; -"2344 linalg_vector_norm_26" -> "2345 clamp_min_26" [label="(4, 12, 64, 1)", style=solid]; -"2345 clamp_min_26" -> "2346 expand_as_26" [label="(4, 12, 64, 1)", style=solid]; -"2346 expand_as_26" -> "2347 div_26" [label="(4, 12, 64, 32)", style=solid]; -"2347 div_26" -> "2348 quantize_per_tensor_default_82" [label="(4, 12, 64, 32)", style=solid]; -"2348 quantize_per_tensor_default_82" -> "2349 dequantize_per_tensor_default_82" [label=None, style=solid]; -"2349 dequantize_per_tensor_default_82" -> "2357 matmul_26" [label=None, style=solid]; -"2350 linalg_vector_norm_27" -> "2351 clamp_min_27" [label="(4, 12, 64, 1)", style=solid]; -"2351 clamp_min_27" -> "2352 expand_as_27" [label="(4, 12, 64, 1)", style=solid]; -"2352 expand_as_27" -> "2353 div_27" [label="(4, 12, 64, 32)", style=solid]; -"2353 div_27" -> "2354 quantize_per_tensor_default_83" [label="(4, 12, 64, 32)", style=solid]; -"2354 quantize_per_tensor_default_83" -> "2355 dequantize_per_tensor_default_83" [label=None, style=solid]; -"2355 dequantize_per_tensor_default_83" -> "2356 transpose_26" [label=None, style=solid]; -"2356 transpose_26" -> "2357 matmul_26" [label="(4, 12, 32, 64)", style=solid]; -"2357 matmul_26" -> "2360 mul_27" [label="(4, 12, 64, 64)", style=solid]; -"2358 clamp_13" -> "2359 exp_13" [label="(12, 1, 1)", style=solid]; -"2359 exp_13" -> "2360 mul_27" [label="(12, 1, 1)", style=solid]; -"2360 mul_27" -> "2361 add_45" [label="(4, 12, 64, 64)", style=solid]; -"2361 add_45" -> "2409 view_74" [label="(4, 12, 64, 64)", style=solid]; -"2362 new_zeros_6" -> "2364 slice_195" [label="(16, 16)", style=solid]; -"2362 new_zeros_6" -> "2368 slice_197" [label="(16, 16)", style=solid]; -"2362 new_zeros_6" -> "2372 slice_199" [label="(16, 16)", style=solid]; -"2362 new_zeros_6" -> "2376 slice_201" [label="(16, 16)", style=solid]; -"2362 new_zeros_6" -> "2380 slice_203" [label="(16, 16)", style=solid]; -"2362 new_zeros_6" -> "2384 slice_205" [label="(16, 16)", style=solid]; -"2362 new_zeros_6" -> "2388 slice_207" [label="(16, 16)", style=solid]; -"2362 new_zeros_6" -> "2392 slice_209" [label="(16, 16)", style=solid]; -"2362 new_zeros_6" -> "2396 slice_211" [label="(16, 16)", style=solid]; -"2362 new_zeros_6" -> "2399 view_73" [label="(16, 16)", style=solid]; -"2363 lift_fresh_copy_54" -> "2366 fill__54" [label="()", style=solid]; -"2364 slice_195" -> "2365 slice_196" [label="(8, 16)", style=solid]; -"2365 slice_196" -> "2366 fill__54" [label="(8, 8)", style=solid]; -"2367 lift_fresh_copy_55" -> "2370 fill__55" [label="()", style=solid]; -"2368 slice_197" -> "2369 slice_198" [label="(8, 16)", style=solid]; -"2369 slice_198" -> "2370 fill__55" [label="(8, 4)", style=solid]; -"2371 lift_fresh_copy_56" -> "2374 fill__56" [label="()", style=solid]; -"2372 slice_199" -> "2373 slice_200" [label="(8, 16)", style=solid]; -"2373 slice_200" -> "2374 fill__56" [label="(8, 4)", style=solid]; -"2375 lift_fresh_copy_57" -> "2378 fill__57" [label="()", style=solid]; -"2376 slice_201" -> "2377 slice_202" [label="(4, 16)", style=solid]; -"2377 slice_202" -> "2378 fill__57" [label="(4, 8)", style=solid]; -"2379 lift_fresh_copy_58" -> "2382 fill__58" [label="()", style=solid]; -"2380 slice_203" -> "2381 slice_204" [label="(4, 16)", style=solid]; -"2381 slice_204" -> "2382 fill__58" [label="(4, 4)", style=solid]; -"2383 lift_fresh_copy_59" -> "2386 fill__59" [label="()", style=solid]; -"2384 slice_205" -> "2385 slice_206" [label="(4, 16)", style=solid]; -"2385 slice_206" -> "2386 fill__59" [label="(4, 4)", style=solid]; -"2387 lift_fresh_copy_60" -> "2390 fill__60" [label="()", style=solid]; -"2388 slice_207" -> "2389 slice_208" [label="(4, 16)", style=solid]; -"2389 slice_208" -> "2390 fill__60" [label="(4, 8)", style=solid]; -"2391 lift_fresh_copy_61" -> "2394 fill__61" [label="()", style=solid]; -"2392 slice_209" -> "2393 slice_210" [label="(4, 16)", style=solid]; -"2393 slice_210" -> "2394 fill__61" [label="(4, 4)", style=solid]; -"2395 lift_fresh_copy_62" -> "2398 fill__62" [label="()", style=solid]; -"2396 slice_211" -> "2397 slice_212" [label="(4, 16)", style=solid]; -"2397 slice_212" -> "2398 fill__62" [label="(4, 4)", style=solid]; -"2399 view_73" -> "2400 permute_62" [label="(2, 8, 2, 8)", style=solid]; -"2400 permute_62" -> "2401 reshape_60" [label="(2, 2, 8, 8)", style=solid]; -"2401 reshape_60" -> "2402 unsqueeze_38" [label="(4, 64)", style=solid]; -"2401 reshape_60" -> "2403 unsqueeze_39" [label="(4, 64)", style=solid]; -"2402 unsqueeze_38" -> "2404 sub_6" [label="(4, 1, 64)", style=solid]; -"2403 unsqueeze_39" -> "2404 sub_6" [label="(4, 64, 1)", style=solid]; -"2404 sub_6" -> "2405 ne_6" [label="(4, 64, 64)", style=solid]; -"2404 sub_6" -> "2406 masked_fill_12" [label="(4, 64, 64)", style=solid]; -"2404 sub_6" -> "2407 eq_6" [label="(4, 64, 64)", style=solid]; -"2405 ne_6" -> "2406 masked_fill_12" [label="(4, 64, 64)", style=solid]; -"2406 masked_fill_12" -> "2408 masked_fill_13" [label="(4, 64, 64)", style=solid]; -"2407 eq_6" -> "2408 masked_fill_13" [label="(4, 64, 64)", style=solid]; -"2408 masked_fill_13" -> "2410 unsqueeze_40" [label="(4, 64, 64)", style=solid]; -"2409 view_74" -> "2412 add_46" [label="(1, 4, 12, 64, 64)", style=solid]; -"2410 unsqueeze_40" -> "2411 unsqueeze_41" [label="(4, 1, 64, 64)", style=solid]; -"2411 unsqueeze_41" -> "2412 add_46" [label="(1, 4, 1, 64, 64)", style=solid]; -"2412 add_46" -> "2413 view_75" [label="(1, 4, 12, 64, 64)", style=solid]; -"2413 view_75" -> "2414 softmax_13" [label="(4, 12, 64, 64)", style=solid]; -"2414 softmax_13" -> "2415 dropout_52" [label="(4, 12, 64, 64)", style=solid]; -"2415 dropout_52" -> "2416 matmul_27" [label="(4, 12, 64, 64)", style=solid]; -"2416 matmul_27" -> "2417 quantize_per_tensor_default_84" [label="(4, 12, 64, 32)", style=solid]; -"2417 quantize_per_tensor_default_84" -> "2418 dequantize_per_tensor_default_84" [label=None, style=solid]; -"2418 dequantize_per_tensor_default_84" -> "2419 transpose_27" [label=None, style=solid]; -"2419 transpose_27" -> "2420 reshape_61" [label="(4, 64, 12, 32)", style=solid]; -"2420 reshape_61" -> "2421 linear_83" [label="(4, 64, 384)", style=solid]; -"2421 linear_83" -> "2422 dropout_53" [label="(4, 64, 384)", style=solid]; -"2422 dropout_53" -> "2423 view_76" [label="(4, 64, 384)", style=solid]; -"2423 view_76" -> "2424 permute_63" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2424 permute_63" -> "2425 reshape_62" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2425 reshape_62" -> "2426 roll_13" [label="(1, 16, 16, 384)", style=solid]; -"2426 roll_13" -> "2427 slice_213" [label="(1, 16, 16, 384)", style=solid]; -"2427 slice_213" -> "2428 slice_214" [label="(1, 16, 16, 384)", style=solid]; -"2428 slice_214" -> "2429 slice_215" [label="(1, 14, 16, 384)", style=solid]; -"2429 slice_215" -> "2430 slice_216" [label="(1, 14, 14, 384)", style=solid]; -"2430 slice_216" -> "2431 contiguous_25" [label="(1, 14, 14, 384)", style=solid]; -"2431 contiguous_25" -> "2432 layer_norm_29" [label="(1, 14, 14, 384)", style=solid]; -"2432 layer_norm_29" -> "2433 add_47" [label="(1, 14, 14, 384)", style=solid]; -"2433 add_47" -> "2434 quantize_per_tensor_default_85" [label="(1, 14, 14, 384)", style=solid]; -"2433 add_47" -> "2444 add_48" [label="(1, 14, 14, 384)", style=solid]; -"2434 quantize_per_tensor_default_85" -> "2435 dequantize_per_tensor_default_85" [label=None, style=solid]; -"2435 dequantize_per_tensor_default_85" -> "2436 linear_84" [label=None, style=solid]; -"2436 linear_84" -> "2437 gelu_13" [label="(1, 14, 14, 1536)", style=solid]; -"2437 gelu_13" -> "2438 quantize_per_tensor_default_86" [label="(1, 14, 14, 1536)", style=solid]; -"2438 quantize_per_tensor_default_86" -> "2439 dequantize_per_tensor_default_86" [label=None, style=solid]; -"2439 dequantize_per_tensor_default_86" -> "2440 dropout_54" [label=None, style=solid]; -"2440 dropout_54" -> "2441 linear_85" [label="(1, 14, 14, 1536)", style=solid]; -"2441 linear_85" -> "2442 dropout_55" [label="(1, 14, 14, 384)", style=solid]; -"2442 dropout_55" -> "2443 layer_norm_30" [label="(1, 14, 14, 384)", style=solid]; -"2443 layer_norm_30" -> "2444 add_48" [label="(1, 14, 14, 384)", style=solid]; -"2444 add_48" -> "2445 quantize_per_tensor_default_87" [label="(1, 14, 14, 384)", style=solid]; -"2444 add_48" -> "2507 add_50" [label="(1, 14, 14, 384)", style=solid]; -"2445 quantize_per_tensor_default_87" -> "2446 dequantize_per_tensor_default_87" [label=None, style=solid]; -"2446 dequantize_per_tensor_default_87" -> "2458 pad_16" [label=None, style=solid]; -"2447 linear_86" -> "2448 relu__14" [label="(1, 15, 15, 512)", style=solid]; -"2448 relu__14" -> "2449 linear_87" [label="(1, 15, 15, 512)", style=solid]; -"2449 linear_87" -> "2450 view_77" [label="(1, 15, 15, 12)", style=solid]; -"2450 view_77" -> "2451 index_14" [label="(225, 12)", style=solid]; -"2451 index_14" -> "2452 view_78" [label="(4096, 12)", style=solid]; -"2452 view_78" -> "2453 permute_64" [label="(64, 64, 12)", style=solid]; -"2453 permute_64" -> "2454 contiguous_26" [label="(12, 64, 64)", style=solid]; -"2454 contiguous_26" -> "2455 unsqueeze_42" [label="(12, 64, 64)", style=solid]; -"2455 unsqueeze_42" -> "2456 sigmoid_14" [label="(1, 12, 64, 64)", style=solid]; -"2456 sigmoid_14" -> "2457 mul_28" [label="(1, 12, 64, 64)", style=solid]; -"2457 mul_28" -> "2488 add_49" [label="(1, 12, 64, 64)", style=solid]; -"2458 pad_16" -> "2459 view_79" [label="(1, 16, 16, 384)", style=solid]; -"2459 view_79" -> "2460 permute_65" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2460 permute_65" -> "2461 reshape_63" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2461 reshape_63" -> "2465 linear_88" [label="(4, 64, 384)", style=solid]; -"2462 clone_14" -> "2463 slice_217" [label="(1152,)", style=solid]; -"2462 clone_14" -> "2465 linear_88" [label="(1152,)", style=solid]; -"2463 slice_217" -> "2464 zero__14" [label="(384,)", style=solid]; -"2465 linear_88" -> "2466 reshape_64" [label="(4, 64, 1152)", style=solid]; -"2466 reshape_64" -> "2467 permute_66" [label="(4, 64, 3, 12, 32)", style=solid]; -"2467 permute_66" -> "2468 select_42" [label="(3, 4, 12, 64, 32)", style=solid]; -"2467 permute_66" -> "2469 select_43" [label="(3, 4, 12, 64, 32)", style=solid]; -"2467 permute_66" -> "2470 select_44" [label="(3, 4, 12, 64, 32)", style=solid]; -"2468 select_42" -> "2471 linalg_vector_norm_28" [label="(4, 12, 64, 32)", style=solid]; -"2468 select_42" -> "2473 expand_as_28" [label="(4, 12, 64, 32)", style=solid]; -"2468 select_42" -> "2474 div_28" [label="(4, 12, 64, 32)", style=solid]; -"2469 select_43" -> "2477 linalg_vector_norm_29" [label="(4, 12, 64, 32)", style=solid]; -"2469 select_43" -> "2479 expand_as_29" [label="(4, 12, 64, 32)", style=solid]; -"2469 select_43" -> "2480 div_29" [label="(4, 12, 64, 32)", style=solid]; -"2470 select_44" -> "2491 matmul_29" [label="(4, 12, 64, 32)", style=solid]; -"2471 linalg_vector_norm_28" -> "2472 clamp_min_28" [label="(4, 12, 64, 1)", style=solid]; -"2472 clamp_min_28" -> "2473 expand_as_28" [label="(4, 12, 64, 1)", style=solid]; -"2473 expand_as_28" -> "2474 div_28" [label="(4, 12, 64, 32)", style=solid]; -"2474 div_28" -> "2475 quantize_per_tensor_default_88" [label="(4, 12, 64, 32)", style=solid]; -"2475 quantize_per_tensor_default_88" -> "2476 dequantize_per_tensor_default_88" [label=None, style=solid]; -"2476 dequantize_per_tensor_default_88" -> "2484 matmul_28" [label=None, style=solid]; -"2477 linalg_vector_norm_29" -> "2478 clamp_min_29" [label="(4, 12, 64, 1)", style=solid]; -"2478 clamp_min_29" -> "2479 expand_as_29" [label="(4, 12, 64, 1)", style=solid]; -"2479 expand_as_29" -> "2480 div_29" [label="(4, 12, 64, 32)", style=solid]; -"2480 div_29" -> "2481 quantize_per_tensor_default_89" [label="(4, 12, 64, 32)", style=solid]; -"2481 quantize_per_tensor_default_89" -> "2482 dequantize_per_tensor_default_89" [label=None, style=solid]; -"2482 dequantize_per_tensor_default_89" -> "2483 transpose_28" [label=None, style=solid]; -"2483 transpose_28" -> "2484 matmul_28" [label="(4, 12, 32, 64)", style=solid]; -"2484 matmul_28" -> "2487 mul_29" [label="(4, 12, 64, 64)", style=solid]; -"2485 clamp_14" -> "2486 exp_14" [label="(12, 1, 1)", style=solid]; -"2486 exp_14" -> "2487 mul_29" [label="(12, 1, 1)", style=solid]; -"2487 mul_29" -> "2488 add_49" [label="(4, 12, 64, 64)", style=solid]; -"2488 add_49" -> "2489 softmax_14" [label="(4, 12, 64, 64)", style=solid]; -"2489 softmax_14" -> "2490 dropout_56" [label="(4, 12, 64, 64)", style=solid]; -"2490 dropout_56" -> "2491 matmul_29" [label="(4, 12, 64, 64)", style=solid]; -"2491 matmul_29" -> "2492 quantize_per_tensor_default_90" [label="(4, 12, 64, 32)", style=solid]; -"2492 quantize_per_tensor_default_90" -> "2493 dequantize_per_tensor_default_90" [label=None, style=solid]; -"2493 dequantize_per_tensor_default_90" -> "2494 transpose_29" [label=None, style=solid]; -"2494 transpose_29" -> "2495 reshape_65" [label="(4, 64, 12, 32)", style=solid]; -"2495 reshape_65" -> "2496 linear_89" [label="(4, 64, 384)", style=solid]; -"2496 linear_89" -> "2497 dropout_57" [label="(4, 64, 384)", style=solid]; -"2497 dropout_57" -> "2498 view_80" [label="(4, 64, 384)", style=solid]; -"2498 view_80" -> "2499 permute_67" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2499 permute_67" -> "2500 reshape_66" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2500 reshape_66" -> "2501 slice_218" [label="(1, 16, 16, 384)", style=solid]; -"2501 slice_218" -> "2502 slice_219" [label="(1, 16, 16, 384)", style=solid]; -"2502 slice_219" -> "2503 slice_220" [label="(1, 14, 16, 384)", style=solid]; -"2503 slice_220" -> "2504 slice_221" [label="(1, 14, 14, 384)", style=solid]; -"2504 slice_221" -> "2505 contiguous_27" [label="(1, 14, 14, 384)", style=solid]; -"2505 contiguous_27" -> "2506 layer_norm_31" [label="(1, 14, 14, 384)", style=solid]; -"2506 layer_norm_31" -> "2507 add_50" [label="(1, 14, 14, 384)", style=solid]; -"2507 add_50" -> "2508 quantize_per_tensor_default_91" [label="(1, 14, 14, 384)", style=solid]; -"2507 add_50" -> "2518 add_51" [label="(1, 14, 14, 384)", style=solid]; -"2508 quantize_per_tensor_default_91" -> "2509 dequantize_per_tensor_default_91" [label=None, style=solid]; -"2509 dequantize_per_tensor_default_91" -> "2510 linear_90" [label=None, style=solid]; -"2510 linear_90" -> "2511 gelu_14" [label="(1, 14, 14, 1536)", style=solid]; -"2511 gelu_14" -> "2512 quantize_per_tensor_default_92" [label="(1, 14, 14, 1536)", style=solid]; -"2512 quantize_per_tensor_default_92" -> "2513 dequantize_per_tensor_default_92" [label=None, style=solid]; -"2513 dequantize_per_tensor_default_92" -> "2514 dropout_58" [label=None, style=solid]; -"2514 dropout_58" -> "2515 linear_91" [label="(1, 14, 14, 1536)", style=solid]; -"2515 linear_91" -> "2516 dropout_59" [label="(1, 14, 14, 384)", style=solid]; -"2516 dropout_59" -> "2517 layer_norm_32" [label="(1, 14, 14, 384)", style=solid]; -"2517 layer_norm_32" -> "2518 add_51" [label="(1, 14, 14, 384)", style=solid]; -"2518 add_51" -> "2530 pad_17" [label="(1, 14, 14, 384)", style=solid]; -"2518 add_51" -> "2635 add_54" [label="(1, 14, 14, 384)", style=solid]; -"2519 linear_92" -> "2520 relu__15" [label="(1, 15, 15, 512)", style=solid]; -"2520 relu__15" -> "2521 linear_93" [label="(1, 15, 15, 512)", style=solid]; -"2521 linear_93" -> "2522 view_81" [label="(1, 15, 15, 12)", style=solid]; -"2522 view_81" -> "2523 index_15" [label="(225, 12)", style=solid]; -"2523 index_15" -> "2524 view_82" [label="(4096, 12)", style=solid]; -"2524 view_82" -> "2525 permute_68" [label="(64, 64, 12)", style=solid]; -"2525 permute_68" -> "2526 contiguous_28" [label="(12, 64, 64)", style=solid]; -"2526 contiguous_28" -> "2527 unsqueeze_43" [label="(12, 64, 64)", style=solid]; -"2527 unsqueeze_43" -> "2528 sigmoid_15" [label="(1, 12, 64, 64)", style=solid]; -"2528 sigmoid_15" -> "2529 mul_30" [label="(1, 12, 64, 64)", style=solid]; -"2529 mul_30" -> "2563 add_52" [label="(1, 12, 64, 64)", style=solid]; -"2530 pad_17" -> "2531 roll_14" [label="(1, 16, 16, 384)", style=solid]; -"2531 roll_14" -> "2532 view_83" [label="(1, 16, 16, 384)", style=solid]; -"2532 view_83" -> "2533 permute_69" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2533 permute_69" -> "2534 reshape_67" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2534 reshape_67" -> "2535 quantize_per_tensor_default_93" [label="(4, 64, 384)", style=solid]; -"2534 reshape_67" -> "2564 new_zeros_7" [label="(4, 64, 384)", style=solid]; -"2535 quantize_per_tensor_default_93" -> "2536 dequantize_per_tensor_default_93" [label=None, style=solid]; -"2536 dequantize_per_tensor_default_93" -> "2540 linear_94" [label=None, style=solid]; -"2537 clone_15" -> "2538 slice_222" [label="(1152,)", style=solid]; -"2537 clone_15" -> "2540 linear_94" [label="(1152,)", style=solid]; -"2538 slice_222" -> "2539 zero__15" [label="(384,)", style=solid]; -"2540 linear_94" -> "2541 reshape_68" [label="(4, 64, 1152)", style=solid]; -"2541 reshape_68" -> "2542 permute_70" [label="(4, 64, 3, 12, 32)", style=solid]; -"2542 permute_70" -> "2543 select_45" [label="(3, 4, 12, 64, 32)", style=solid]; -"2542 permute_70" -> "2544 select_46" [label="(3, 4, 12, 64, 32)", style=solid]; -"2542 permute_70" -> "2545 select_47" [label="(3, 4, 12, 64, 32)", style=solid]; -"2543 select_45" -> "2546 linalg_vector_norm_30" [label="(4, 12, 64, 32)", style=solid]; -"2543 select_45" -> "2548 expand_as_30" [label="(4, 12, 64, 32)", style=solid]; -"2543 select_45" -> "2549 div_30" [label="(4, 12, 64, 32)", style=solid]; -"2544 select_46" -> "2552 linalg_vector_norm_31" [label="(4, 12, 64, 32)", style=solid]; -"2544 select_46" -> "2554 expand_as_31" [label="(4, 12, 64, 32)", style=solid]; -"2544 select_46" -> "2555 div_31" [label="(4, 12, 64, 32)", style=solid]; -"2545 select_47" -> "2618 matmul_31" [label="(4, 12, 64, 32)", style=solid]; -"2546 linalg_vector_norm_30" -> "2547 clamp_min_30" [label="(4, 12, 64, 1)", style=solid]; -"2547 clamp_min_30" -> "2548 expand_as_30" [label="(4, 12, 64, 1)", style=solid]; -"2548 expand_as_30" -> "2549 div_30" [label="(4, 12, 64, 32)", style=solid]; -"2549 div_30" -> "2550 quantize_per_tensor_default_94" [label="(4, 12, 64, 32)", style=solid]; -"2550 quantize_per_tensor_default_94" -> "2551 dequantize_per_tensor_default_94" [label=None, style=solid]; -"2551 dequantize_per_tensor_default_94" -> "2559 matmul_30" [label=None, style=solid]; -"2552 linalg_vector_norm_31" -> "2553 clamp_min_31" [label="(4, 12, 64, 1)", style=solid]; -"2553 clamp_min_31" -> "2554 expand_as_31" [label="(4, 12, 64, 1)", style=solid]; -"2554 expand_as_31" -> "2555 div_31" [label="(4, 12, 64, 32)", style=solid]; -"2555 div_31" -> "2556 quantize_per_tensor_default_95" [label="(4, 12, 64, 32)", style=solid]; -"2556 quantize_per_tensor_default_95" -> "2557 dequantize_per_tensor_default_95" [label=None, style=solid]; -"2557 dequantize_per_tensor_default_95" -> "2558 transpose_30" [label=None, style=solid]; -"2558 transpose_30" -> "2559 matmul_30" [label="(4, 12, 32, 64)", style=solid]; -"2559 matmul_30" -> "2562 mul_31" [label="(4, 12, 64, 64)", style=solid]; -"2560 clamp_15" -> "2561 exp_15" [label="(12, 1, 1)", style=solid]; -"2561 exp_15" -> "2562 mul_31" [label="(12, 1, 1)", style=solid]; -"2562 mul_31" -> "2563 add_52" [label="(4, 12, 64, 64)", style=solid]; -"2563 add_52" -> "2611 view_85" [label="(4, 12, 64, 64)", style=solid]; -"2564 new_zeros_7" -> "2566 slice_223" [label="(16, 16)", style=solid]; -"2564 new_zeros_7" -> "2570 slice_225" [label="(16, 16)", style=solid]; -"2564 new_zeros_7" -> "2574 slice_227" [label="(16, 16)", style=solid]; -"2564 new_zeros_7" -> "2578 slice_229" [label="(16, 16)", style=solid]; -"2564 new_zeros_7" -> "2582 slice_231" [label="(16, 16)", style=solid]; -"2564 new_zeros_7" -> "2586 slice_233" [label="(16, 16)", style=solid]; -"2564 new_zeros_7" -> "2590 slice_235" [label="(16, 16)", style=solid]; -"2564 new_zeros_7" -> "2594 slice_237" [label="(16, 16)", style=solid]; -"2564 new_zeros_7" -> "2598 slice_239" [label="(16, 16)", style=solid]; -"2564 new_zeros_7" -> "2601 view_84" [label="(16, 16)", style=solid]; -"2565 lift_fresh_copy_63" -> "2568 fill__63" [label="()", style=solid]; -"2566 slice_223" -> "2567 slice_224" [label="(8, 16)", style=solid]; -"2567 slice_224" -> "2568 fill__63" [label="(8, 8)", style=solid]; -"2569 lift_fresh_copy_64" -> "2572 fill__64" [label="()", style=solid]; -"2570 slice_225" -> "2571 slice_226" [label="(8, 16)", style=solid]; -"2571 slice_226" -> "2572 fill__64" [label="(8, 4)", style=solid]; -"2573 lift_fresh_copy_65" -> "2576 fill__65" [label="()", style=solid]; -"2574 slice_227" -> "2575 slice_228" [label="(8, 16)", style=solid]; -"2575 slice_228" -> "2576 fill__65" [label="(8, 4)", style=solid]; -"2577 lift_fresh_copy_66" -> "2580 fill__66" [label="()", style=solid]; -"2578 slice_229" -> "2579 slice_230" [label="(4, 16)", style=solid]; -"2579 slice_230" -> "2580 fill__66" [label="(4, 8)", style=solid]; -"2581 lift_fresh_copy_67" -> "2584 fill__67" [label="()", style=solid]; -"2582 slice_231" -> "2583 slice_232" [label="(4, 16)", style=solid]; -"2583 slice_232" -> "2584 fill__67" [label="(4, 4)", style=solid]; -"2585 lift_fresh_copy_68" -> "2588 fill__68" [label="()", style=solid]; -"2586 slice_233" -> "2587 slice_234" [label="(4, 16)", style=solid]; -"2587 slice_234" -> "2588 fill__68" [label="(4, 4)", style=solid]; -"2589 lift_fresh_copy_69" -> "2592 fill__69" [label="()", style=solid]; -"2590 slice_235" -> "2591 slice_236" [label="(4, 16)", style=solid]; -"2591 slice_236" -> "2592 fill__69" [label="(4, 8)", style=solid]; -"2593 lift_fresh_copy_70" -> "2596 fill__70" [label="()", style=solid]; -"2594 slice_237" -> "2595 slice_238" [label="(4, 16)", style=solid]; -"2595 slice_238" -> "2596 fill__70" [label="(4, 4)", style=solid]; -"2597 lift_fresh_copy_71" -> "2600 fill__71" [label="()", style=solid]; -"2598 slice_239" -> "2599 slice_240" [label="(4, 16)", style=solid]; -"2599 slice_240" -> "2600 fill__71" [label="(4, 4)", style=solid]; -"2601 view_84" -> "2602 permute_71" [label="(2, 8, 2, 8)", style=solid]; -"2602 permute_71" -> "2603 reshape_69" [label="(2, 2, 8, 8)", style=solid]; -"2603 reshape_69" -> "2604 unsqueeze_44" [label="(4, 64)", style=solid]; -"2603 reshape_69" -> "2605 unsqueeze_45" [label="(4, 64)", style=solid]; -"2604 unsqueeze_44" -> "2606 sub_7" [label="(4, 1, 64)", style=solid]; -"2605 unsqueeze_45" -> "2606 sub_7" [label="(4, 64, 1)", style=solid]; -"2606 sub_7" -> "2607 ne_7" [label="(4, 64, 64)", style=solid]; -"2606 sub_7" -> "2608 masked_fill_14" [label="(4, 64, 64)", style=solid]; -"2606 sub_7" -> "2609 eq_7" [label="(4, 64, 64)", style=solid]; -"2607 ne_7" -> "2608 masked_fill_14" [label="(4, 64, 64)", style=solid]; -"2608 masked_fill_14" -> "2610 masked_fill_15" [label="(4, 64, 64)", style=solid]; -"2609 eq_7" -> "2610 masked_fill_15" [label="(4, 64, 64)", style=solid]; -"2610 masked_fill_15" -> "2612 unsqueeze_46" [label="(4, 64, 64)", style=solid]; -"2611 view_85" -> "2614 add_53" [label="(1, 4, 12, 64, 64)", style=solid]; -"2612 unsqueeze_46" -> "2613 unsqueeze_47" [label="(4, 1, 64, 64)", style=solid]; -"2613 unsqueeze_47" -> "2614 add_53" [label="(1, 4, 1, 64, 64)", style=solid]; -"2614 add_53" -> "2615 view_86" [label="(1, 4, 12, 64, 64)", style=solid]; -"2615 view_86" -> "2616 softmax_15" [label="(4, 12, 64, 64)", style=solid]; -"2616 softmax_15" -> "2617 dropout_60" [label="(4, 12, 64, 64)", style=solid]; -"2617 dropout_60" -> "2618 matmul_31" [label="(4, 12, 64, 64)", style=solid]; -"2618 matmul_31" -> "2619 quantize_per_tensor_default_96" [label="(4, 12, 64, 32)", style=solid]; -"2619 quantize_per_tensor_default_96" -> "2620 dequantize_per_tensor_default_96" [label=None, style=solid]; -"2620 dequantize_per_tensor_default_96" -> "2621 transpose_31" [label=None, style=solid]; -"2621 transpose_31" -> "2622 reshape_70" [label="(4, 64, 12, 32)", style=solid]; -"2622 reshape_70" -> "2623 linear_95" [label="(4, 64, 384)", style=solid]; -"2623 linear_95" -> "2624 dropout_61" [label="(4, 64, 384)", style=solid]; -"2624 dropout_61" -> "2625 view_87" [label="(4, 64, 384)", style=solid]; -"2625 view_87" -> "2626 permute_72" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2626 permute_72" -> "2627 reshape_71" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2627 reshape_71" -> "2628 roll_15" [label="(1, 16, 16, 384)", style=solid]; -"2628 roll_15" -> "2629 slice_241" [label="(1, 16, 16, 384)", style=solid]; -"2629 slice_241" -> "2630 slice_242" [label="(1, 16, 16, 384)", style=solid]; -"2630 slice_242" -> "2631 slice_243" [label="(1, 14, 16, 384)", style=solid]; -"2631 slice_243" -> "2632 slice_244" [label="(1, 14, 14, 384)", style=solid]; -"2632 slice_244" -> "2633 contiguous_29" [label="(1, 14, 14, 384)", style=solid]; -"2633 contiguous_29" -> "2634 layer_norm_33" [label="(1, 14, 14, 384)", style=solid]; -"2634 layer_norm_33" -> "2635 add_54" [label="(1, 14, 14, 384)", style=solid]; -"2635 add_54" -> "2636 quantize_per_tensor_default_97" [label="(1, 14, 14, 384)", style=solid]; -"2635 add_54" -> "2646 add_55" [label="(1, 14, 14, 384)", style=solid]; -"2636 quantize_per_tensor_default_97" -> "2637 dequantize_per_tensor_default_97" [label=None, style=solid]; -"2637 dequantize_per_tensor_default_97" -> "2638 linear_96" [label=None, style=solid]; -"2638 linear_96" -> "2639 gelu_15" [label="(1, 14, 14, 1536)", style=solid]; -"2639 gelu_15" -> "2640 quantize_per_tensor_default_98" [label="(1, 14, 14, 1536)", style=solid]; -"2640 quantize_per_tensor_default_98" -> "2641 dequantize_per_tensor_default_98" [label=None, style=solid]; -"2641 dequantize_per_tensor_default_98" -> "2642 dropout_62" [label=None, style=solid]; -"2642 dropout_62" -> "2643 linear_97" [label="(1, 14, 14, 1536)", style=solid]; -"2643 linear_97" -> "2644 dropout_63" [label="(1, 14, 14, 384)", style=solid]; -"2644 dropout_63" -> "2645 layer_norm_34" [label="(1, 14, 14, 384)", style=solid]; -"2645 layer_norm_34" -> "2646 add_55" [label="(1, 14, 14, 384)", style=solid]; -"2646 add_55" -> "2647 quantize_per_tensor_default_99" [label="(1, 14, 14, 384)", style=solid]; -"2646 add_55" -> "2709 add_57" [label="(1, 14, 14, 384)", style=solid]; -"2647 quantize_per_tensor_default_99" -> "2648 dequantize_per_tensor_default_99" [label=None, style=solid]; -"2648 dequantize_per_tensor_default_99" -> "2660 pad_18" [label=None, style=solid]; -"2649 linear_98" -> "2650 relu__16" [label="(1, 15, 15, 512)", style=solid]; -"2650 relu__16" -> "2651 linear_99" [label="(1, 15, 15, 512)", style=solid]; -"2651 linear_99" -> "2652 view_88" [label="(1, 15, 15, 12)", style=solid]; -"2652 view_88" -> "2653 index_16" [label="(225, 12)", style=solid]; -"2653 index_16" -> "2654 view_89" [label="(4096, 12)", style=solid]; -"2654 view_89" -> "2655 permute_73" [label="(64, 64, 12)", style=solid]; -"2655 permute_73" -> "2656 contiguous_30" [label="(12, 64, 64)", style=solid]; -"2656 contiguous_30" -> "2657 unsqueeze_48" [label="(12, 64, 64)", style=solid]; -"2657 unsqueeze_48" -> "2658 sigmoid_16" [label="(1, 12, 64, 64)", style=solid]; -"2658 sigmoid_16" -> "2659 mul_32" [label="(1, 12, 64, 64)", style=solid]; -"2659 mul_32" -> "2690 add_56" [label="(1, 12, 64, 64)", style=solid]; -"2660 pad_18" -> "2661 view_90" [label="(1, 16, 16, 384)", style=solid]; -"2661 view_90" -> "2662 permute_74" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2662 permute_74" -> "2663 reshape_72" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2663 reshape_72" -> "2667 linear_100" [label="(4, 64, 384)", style=solid]; -"2664 clone_16" -> "2665 slice_245" [label="(1152,)", style=solid]; -"2664 clone_16" -> "2667 linear_100" [label="(1152,)", style=solid]; -"2665 slice_245" -> "2666 zero__16" [label="(384,)", style=solid]; -"2667 linear_100" -> "2668 reshape_73" [label="(4, 64, 1152)", style=solid]; -"2668 reshape_73" -> "2669 permute_75" [label="(4, 64, 3, 12, 32)", style=solid]; -"2669 permute_75" -> "2670 select_48" [label="(3, 4, 12, 64, 32)", style=solid]; -"2669 permute_75" -> "2671 select_49" [label="(3, 4, 12, 64, 32)", style=solid]; -"2669 permute_75" -> "2672 select_50" [label="(3, 4, 12, 64, 32)", style=solid]; -"2670 select_48" -> "2673 linalg_vector_norm_32" [label="(4, 12, 64, 32)", style=solid]; -"2670 select_48" -> "2675 expand_as_32" [label="(4, 12, 64, 32)", style=solid]; -"2670 select_48" -> "2676 div_32" [label="(4, 12, 64, 32)", style=solid]; -"2671 select_49" -> "2679 linalg_vector_norm_33" [label="(4, 12, 64, 32)", style=solid]; -"2671 select_49" -> "2681 expand_as_33" [label="(4, 12, 64, 32)", style=solid]; -"2671 select_49" -> "2682 div_33" [label="(4, 12, 64, 32)", style=solid]; -"2672 select_50" -> "2693 matmul_33" [label="(4, 12, 64, 32)", style=solid]; -"2673 linalg_vector_norm_32" -> "2674 clamp_min_32" [label="(4, 12, 64, 1)", style=solid]; -"2674 clamp_min_32" -> "2675 expand_as_32" [label="(4, 12, 64, 1)", style=solid]; -"2675 expand_as_32" -> "2676 div_32" [label="(4, 12, 64, 32)", style=solid]; -"2676 div_32" -> "2677 quantize_per_tensor_default_100" [label="(4, 12, 64, 32)", style=solid]; -"2677 quantize_per_tensor_default_100" -> "2678 dequantize_per_tensor_default_100" [label=None, style=solid]; -"2678 dequantize_per_tensor_default_100" -> "2686 matmul_32" [label=None, style=solid]; -"2679 linalg_vector_norm_33" -> "2680 clamp_min_33" [label="(4, 12, 64, 1)", style=solid]; -"2680 clamp_min_33" -> "2681 expand_as_33" [label="(4, 12, 64, 1)", style=solid]; -"2681 expand_as_33" -> "2682 div_33" [label="(4, 12, 64, 32)", style=solid]; -"2682 div_33" -> "2683 quantize_per_tensor_default_101" [label="(4, 12, 64, 32)", style=solid]; -"2683 quantize_per_tensor_default_101" -> "2684 dequantize_per_tensor_default_101" [label=None, style=solid]; -"2684 dequantize_per_tensor_default_101" -> "2685 transpose_32" [label=None, style=solid]; -"2685 transpose_32" -> "2686 matmul_32" [label="(4, 12, 32, 64)", style=solid]; -"2686 matmul_32" -> "2689 mul_33" [label="(4, 12, 64, 64)", style=solid]; -"2687 clamp_16" -> "2688 exp_16" [label="(12, 1, 1)", style=solid]; -"2688 exp_16" -> "2689 mul_33" [label="(12, 1, 1)", style=solid]; -"2689 mul_33" -> "2690 add_56" [label="(4, 12, 64, 64)", style=solid]; -"2690 add_56" -> "2691 softmax_16" [label="(4, 12, 64, 64)", style=solid]; -"2691 softmax_16" -> "2692 dropout_64" [label="(4, 12, 64, 64)", style=solid]; -"2692 dropout_64" -> "2693 matmul_33" [label="(4, 12, 64, 64)", style=solid]; -"2693 matmul_33" -> "2694 quantize_per_tensor_default_102" [label="(4, 12, 64, 32)", style=solid]; -"2694 quantize_per_tensor_default_102" -> "2695 dequantize_per_tensor_default_102" [label=None, style=solid]; -"2695 dequantize_per_tensor_default_102" -> "2696 transpose_33" [label=None, style=solid]; -"2696 transpose_33" -> "2697 reshape_74" [label="(4, 64, 12, 32)", style=solid]; -"2697 reshape_74" -> "2698 linear_101" [label="(4, 64, 384)", style=solid]; -"2698 linear_101" -> "2699 dropout_65" [label="(4, 64, 384)", style=solid]; -"2699 dropout_65" -> "2700 view_91" [label="(4, 64, 384)", style=solid]; -"2700 view_91" -> "2701 permute_76" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2701 permute_76" -> "2702 reshape_75" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2702 reshape_75" -> "2703 slice_246" [label="(1, 16, 16, 384)", style=solid]; -"2703 slice_246" -> "2704 slice_247" [label="(1, 16, 16, 384)", style=solid]; -"2704 slice_247" -> "2705 slice_248" [label="(1, 14, 16, 384)", style=solid]; -"2705 slice_248" -> "2706 slice_249" [label="(1, 14, 14, 384)", style=solid]; -"2706 slice_249" -> "2707 contiguous_31" [label="(1, 14, 14, 384)", style=solid]; -"2707 contiguous_31" -> "2708 layer_norm_35" [label="(1, 14, 14, 384)", style=solid]; -"2708 layer_norm_35" -> "2709 add_57" [label="(1, 14, 14, 384)", style=solid]; -"2709 add_57" -> "2710 quantize_per_tensor_default_103" [label="(1, 14, 14, 384)", style=solid]; -"2709 add_57" -> "2720 add_58" [label="(1, 14, 14, 384)", style=solid]; -"2710 quantize_per_tensor_default_103" -> "2711 dequantize_per_tensor_default_103" [label=None, style=solid]; -"2711 dequantize_per_tensor_default_103" -> "2712 linear_102" [label=None, style=solid]; -"2712 linear_102" -> "2713 gelu_16" [label="(1, 14, 14, 1536)", style=solid]; -"2713 gelu_16" -> "2714 quantize_per_tensor_default_104" [label="(1, 14, 14, 1536)", style=solid]; -"2714 quantize_per_tensor_default_104" -> "2715 dequantize_per_tensor_default_104" [label=None, style=solid]; -"2715 dequantize_per_tensor_default_104" -> "2716 dropout_66" [label=None, style=solid]; -"2716 dropout_66" -> "2717 linear_103" [label="(1, 14, 14, 1536)", style=solid]; -"2717 linear_103" -> "2718 dropout_67" [label="(1, 14, 14, 384)", style=solid]; -"2718 dropout_67" -> "2719 layer_norm_36" [label="(1, 14, 14, 384)", style=solid]; -"2719 layer_norm_36" -> "2720 add_58" [label="(1, 14, 14, 384)", style=solid]; -"2720 add_58" -> "2732 pad_19" [label="(1, 14, 14, 384)", style=solid]; -"2720 add_58" -> "2837 add_61" [label="(1, 14, 14, 384)", style=solid]; -"2721 linear_104" -> "2722 relu__17" [label="(1, 15, 15, 512)", style=solid]; -"2722 relu__17" -> "2723 linear_105" [label="(1, 15, 15, 512)", style=solid]; -"2723 linear_105" -> "2724 view_92" [label="(1, 15, 15, 12)", style=solid]; -"2724 view_92" -> "2725 index_17" [label="(225, 12)", style=solid]; -"2725 index_17" -> "2726 view_93" [label="(4096, 12)", style=solid]; -"2726 view_93" -> "2727 permute_77" [label="(64, 64, 12)", style=solid]; -"2727 permute_77" -> "2728 contiguous_32" [label="(12, 64, 64)", style=solid]; -"2728 contiguous_32" -> "2729 unsqueeze_49" [label="(12, 64, 64)", style=solid]; -"2729 unsqueeze_49" -> "2730 sigmoid_17" [label="(1, 12, 64, 64)", style=solid]; -"2730 sigmoid_17" -> "2731 mul_34" [label="(1, 12, 64, 64)", style=solid]; -"2731 mul_34" -> "2765 add_59" [label="(1, 12, 64, 64)", style=solid]; -"2732 pad_19" -> "2733 roll_16" [label="(1, 16, 16, 384)", style=solid]; -"2733 roll_16" -> "2734 view_94" [label="(1, 16, 16, 384)", style=solid]; -"2734 view_94" -> "2735 permute_78" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2735 permute_78" -> "2736 reshape_76" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2736 reshape_76" -> "2737 quantize_per_tensor_default_105" [label="(4, 64, 384)", style=solid]; -"2736 reshape_76" -> "2766 new_zeros_8" [label="(4, 64, 384)", style=solid]; -"2737 quantize_per_tensor_default_105" -> "2738 dequantize_per_tensor_default_105" [label=None, style=solid]; -"2738 dequantize_per_tensor_default_105" -> "2742 linear_106" [label=None, style=solid]; -"2739 clone_17" -> "2740 slice_250" [label="(1152,)", style=solid]; -"2739 clone_17" -> "2742 linear_106" [label="(1152,)", style=solid]; -"2740 slice_250" -> "2741 zero__17" [label="(384,)", style=solid]; -"2742 linear_106" -> "2743 reshape_77" [label="(4, 64, 1152)", style=solid]; -"2743 reshape_77" -> "2744 permute_79" [label="(4, 64, 3, 12, 32)", style=solid]; -"2744 permute_79" -> "2745 select_51" [label="(3, 4, 12, 64, 32)", style=solid]; -"2744 permute_79" -> "2746 select_52" [label="(3, 4, 12, 64, 32)", style=solid]; -"2744 permute_79" -> "2747 select_53" [label="(3, 4, 12, 64, 32)", style=solid]; -"2745 select_51" -> "2748 linalg_vector_norm_34" [label="(4, 12, 64, 32)", style=solid]; -"2745 select_51" -> "2750 expand_as_34" [label="(4, 12, 64, 32)", style=solid]; -"2745 select_51" -> "2751 div_34" [label="(4, 12, 64, 32)", style=solid]; -"2746 select_52" -> "2754 linalg_vector_norm_35" [label="(4, 12, 64, 32)", style=solid]; -"2746 select_52" -> "2756 expand_as_35" [label="(4, 12, 64, 32)", style=solid]; -"2746 select_52" -> "2757 div_35" [label="(4, 12, 64, 32)", style=solid]; -"2747 select_53" -> "2820 matmul_35" [label="(4, 12, 64, 32)", style=solid]; -"2748 linalg_vector_norm_34" -> "2749 clamp_min_34" [label="(4, 12, 64, 1)", style=solid]; -"2749 clamp_min_34" -> "2750 expand_as_34" [label="(4, 12, 64, 1)", style=solid]; -"2750 expand_as_34" -> "2751 div_34" [label="(4, 12, 64, 32)", style=solid]; -"2751 div_34" -> "2752 quantize_per_tensor_default_106" [label="(4, 12, 64, 32)", style=solid]; -"2752 quantize_per_tensor_default_106" -> "2753 dequantize_per_tensor_default_106" [label=None, style=solid]; -"2753 dequantize_per_tensor_default_106" -> "2761 matmul_34" [label=None, style=solid]; -"2754 linalg_vector_norm_35" -> "2755 clamp_min_35" [label="(4, 12, 64, 1)", style=solid]; -"2755 clamp_min_35" -> "2756 expand_as_35" [label="(4, 12, 64, 1)", style=solid]; -"2756 expand_as_35" -> "2757 div_35" [label="(4, 12, 64, 32)", style=solid]; -"2757 div_35" -> "2758 quantize_per_tensor_default_107" [label="(4, 12, 64, 32)", style=solid]; -"2758 quantize_per_tensor_default_107" -> "2759 dequantize_per_tensor_default_107" [label=None, style=solid]; -"2759 dequantize_per_tensor_default_107" -> "2760 transpose_34" [label=None, style=solid]; -"2760 transpose_34" -> "2761 matmul_34" [label="(4, 12, 32, 64)", style=solid]; -"2761 matmul_34" -> "2764 mul_35" [label="(4, 12, 64, 64)", style=solid]; -"2762 clamp_17" -> "2763 exp_17" [label="(12, 1, 1)", style=solid]; -"2763 exp_17" -> "2764 mul_35" [label="(12, 1, 1)", style=solid]; -"2764 mul_35" -> "2765 add_59" [label="(4, 12, 64, 64)", style=solid]; -"2765 add_59" -> "2813 view_96" [label="(4, 12, 64, 64)", style=solid]; -"2766 new_zeros_8" -> "2768 slice_251" [label="(16, 16)", style=solid]; -"2766 new_zeros_8" -> "2772 slice_253" [label="(16, 16)", style=solid]; -"2766 new_zeros_8" -> "2776 slice_255" [label="(16, 16)", style=solid]; -"2766 new_zeros_8" -> "2780 slice_257" [label="(16, 16)", style=solid]; -"2766 new_zeros_8" -> "2784 slice_259" [label="(16, 16)", style=solid]; -"2766 new_zeros_8" -> "2788 slice_261" [label="(16, 16)", style=solid]; -"2766 new_zeros_8" -> "2792 slice_263" [label="(16, 16)", style=solid]; -"2766 new_zeros_8" -> "2796 slice_265" [label="(16, 16)", style=solid]; -"2766 new_zeros_8" -> "2800 slice_267" [label="(16, 16)", style=solid]; -"2766 new_zeros_8" -> "2803 view_95" [label="(16, 16)", style=solid]; -"2767 lift_fresh_copy_72" -> "2770 fill__72" [label="()", style=solid]; -"2768 slice_251" -> "2769 slice_252" [label="(8, 16)", style=solid]; -"2769 slice_252" -> "2770 fill__72" [label="(8, 8)", style=solid]; -"2771 lift_fresh_copy_73" -> "2774 fill__73" [label="()", style=solid]; -"2772 slice_253" -> "2773 slice_254" [label="(8, 16)", style=solid]; -"2773 slice_254" -> "2774 fill__73" [label="(8, 4)", style=solid]; -"2775 lift_fresh_copy_74" -> "2778 fill__74" [label="()", style=solid]; -"2776 slice_255" -> "2777 slice_256" [label="(8, 16)", style=solid]; -"2777 slice_256" -> "2778 fill__74" [label="(8, 4)", style=solid]; -"2779 lift_fresh_copy_75" -> "2782 fill__75" [label="()", style=solid]; -"2780 slice_257" -> "2781 slice_258" [label="(4, 16)", style=solid]; -"2781 slice_258" -> "2782 fill__75" [label="(4, 8)", style=solid]; -"2783 lift_fresh_copy_76" -> "2786 fill__76" [label="()", style=solid]; -"2784 slice_259" -> "2785 slice_260" [label="(4, 16)", style=solid]; -"2785 slice_260" -> "2786 fill__76" [label="(4, 4)", style=solid]; -"2787 lift_fresh_copy_77" -> "2790 fill__77" [label="()", style=solid]; -"2788 slice_261" -> "2789 slice_262" [label="(4, 16)", style=solid]; -"2789 slice_262" -> "2790 fill__77" [label="(4, 4)", style=solid]; -"2791 lift_fresh_copy_78" -> "2794 fill__78" [label="()", style=solid]; -"2792 slice_263" -> "2793 slice_264" [label="(4, 16)", style=solid]; -"2793 slice_264" -> "2794 fill__78" [label="(4, 8)", style=solid]; -"2795 lift_fresh_copy_79" -> "2798 fill__79" [label="()", style=solid]; -"2796 slice_265" -> "2797 slice_266" [label="(4, 16)", style=solid]; -"2797 slice_266" -> "2798 fill__79" [label="(4, 4)", style=solid]; -"2799 lift_fresh_copy_80" -> "2802 fill__80" [label="()", style=solid]; -"2800 slice_267" -> "2801 slice_268" [label="(4, 16)", style=solid]; -"2801 slice_268" -> "2802 fill__80" [label="(4, 4)", style=solid]; -"2803 view_95" -> "2804 permute_80" [label="(2, 8, 2, 8)", style=solid]; -"2804 permute_80" -> "2805 reshape_78" [label="(2, 2, 8, 8)", style=solid]; -"2805 reshape_78" -> "2806 unsqueeze_50" [label="(4, 64)", style=solid]; -"2805 reshape_78" -> "2807 unsqueeze_51" [label="(4, 64)", style=solid]; -"2806 unsqueeze_50" -> "2808 sub_8" [label="(4, 1, 64)", style=solid]; -"2807 unsqueeze_51" -> "2808 sub_8" [label="(4, 64, 1)", style=solid]; -"2808 sub_8" -> "2809 ne_8" [label="(4, 64, 64)", style=solid]; -"2808 sub_8" -> "2810 masked_fill_16" [label="(4, 64, 64)", style=solid]; -"2808 sub_8" -> "2811 eq_8" [label="(4, 64, 64)", style=solid]; -"2809 ne_8" -> "2810 masked_fill_16" [label="(4, 64, 64)", style=solid]; -"2810 masked_fill_16" -> "2812 masked_fill_17" [label="(4, 64, 64)", style=solid]; -"2811 eq_8" -> "2812 masked_fill_17" [label="(4, 64, 64)", style=solid]; -"2812 masked_fill_17" -> "2814 unsqueeze_52" [label="(4, 64, 64)", style=solid]; -"2813 view_96" -> "2816 add_60" [label="(1, 4, 12, 64, 64)", style=solid]; -"2814 unsqueeze_52" -> "2815 unsqueeze_53" [label="(4, 1, 64, 64)", style=solid]; -"2815 unsqueeze_53" -> "2816 add_60" [label="(1, 4, 1, 64, 64)", style=solid]; -"2816 add_60" -> "2817 view_97" [label="(1, 4, 12, 64, 64)", style=solid]; -"2817 view_97" -> "2818 softmax_17" [label="(4, 12, 64, 64)", style=solid]; -"2818 softmax_17" -> "2819 dropout_68" [label="(4, 12, 64, 64)", style=solid]; -"2819 dropout_68" -> "2820 matmul_35" [label="(4, 12, 64, 64)", style=solid]; -"2820 matmul_35" -> "2821 quantize_per_tensor_default_108" [label="(4, 12, 64, 32)", style=solid]; -"2821 quantize_per_tensor_default_108" -> "2822 dequantize_per_tensor_default_108" [label=None, style=solid]; -"2822 dequantize_per_tensor_default_108" -> "2823 transpose_35" [label=None, style=solid]; -"2823 transpose_35" -> "2824 reshape_79" [label="(4, 64, 12, 32)", style=solid]; -"2824 reshape_79" -> "2825 linear_107" [label="(4, 64, 384)", style=solid]; -"2825 linear_107" -> "2826 dropout_69" [label="(4, 64, 384)", style=solid]; -"2826 dropout_69" -> "2827 view_98" [label="(4, 64, 384)", style=solid]; -"2827 view_98" -> "2828 permute_81" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2828 permute_81" -> "2829 reshape_80" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2829 reshape_80" -> "2830 roll_17" [label="(1, 16, 16, 384)", style=solid]; -"2830 roll_17" -> "2831 slice_269" [label="(1, 16, 16, 384)", style=solid]; -"2831 slice_269" -> "2832 slice_270" [label="(1, 16, 16, 384)", style=solid]; -"2832 slice_270" -> "2833 slice_271" [label="(1, 14, 16, 384)", style=solid]; -"2833 slice_271" -> "2834 slice_272" [label="(1, 14, 14, 384)", style=solid]; -"2834 slice_272" -> "2835 contiguous_33" [label="(1, 14, 14, 384)", style=solid]; -"2835 contiguous_33" -> "2836 layer_norm_37" [label="(1, 14, 14, 384)", style=solid]; -"2836 layer_norm_37" -> "2837 add_61" [label="(1, 14, 14, 384)", style=solid]; -"2837 add_61" -> "2838 quantize_per_tensor_default_109" [label="(1, 14, 14, 384)", style=solid]; -"2837 add_61" -> "2848 add_62" [label="(1, 14, 14, 384)", style=solid]; -"2838 quantize_per_tensor_default_109" -> "2839 dequantize_per_tensor_default_109" [label=None, style=solid]; -"2839 dequantize_per_tensor_default_109" -> "2840 linear_108" [label=None, style=solid]; -"2840 linear_108" -> "2841 gelu_17" [label="(1, 14, 14, 1536)", style=solid]; -"2841 gelu_17" -> "2842 quantize_per_tensor_default_110" [label="(1, 14, 14, 1536)", style=solid]; -"2842 quantize_per_tensor_default_110" -> "2843 dequantize_per_tensor_default_110" [label=None, style=solid]; -"2843 dequantize_per_tensor_default_110" -> "2844 dropout_70" [label=None, style=solid]; -"2844 dropout_70" -> "2845 linear_109" [label="(1, 14, 14, 1536)", style=solid]; -"2845 linear_109" -> "2846 dropout_71" [label="(1, 14, 14, 384)", style=solid]; -"2846 dropout_71" -> "2847 layer_norm_38" [label="(1, 14, 14, 384)", style=solid]; -"2847 layer_norm_38" -> "2848 add_62" [label="(1, 14, 14, 384)", style=solid]; -"2848 add_62" -> "2849 quantize_per_tensor_default_111" [label="(1, 14, 14, 384)", style=solid]; -"2848 add_62" -> "2911 add_64" [label="(1, 14, 14, 384)", style=solid]; -"2849 quantize_per_tensor_default_111" -> "2850 dequantize_per_tensor_default_111" [label=None, style=solid]; -"2850 dequantize_per_tensor_default_111" -> "2862 pad_20" [label=None, style=solid]; -"2851 linear_110" -> "2852 relu__18" [label="(1, 15, 15, 512)", style=solid]; -"2852 relu__18" -> "2853 linear_111" [label="(1, 15, 15, 512)", style=solid]; -"2853 linear_111" -> "2854 view_99" [label="(1, 15, 15, 12)", style=solid]; -"2854 view_99" -> "2855 index_18" [label="(225, 12)", style=solid]; -"2855 index_18" -> "2856 view_100" [label="(4096, 12)", style=solid]; -"2856 view_100" -> "2857 permute_82" [label="(64, 64, 12)", style=solid]; -"2857 permute_82" -> "2858 contiguous_34" [label="(12, 64, 64)", style=solid]; -"2858 contiguous_34" -> "2859 unsqueeze_54" [label="(12, 64, 64)", style=solid]; -"2859 unsqueeze_54" -> "2860 sigmoid_18" [label="(1, 12, 64, 64)", style=solid]; -"2860 sigmoid_18" -> "2861 mul_36" [label="(1, 12, 64, 64)", style=solid]; -"2861 mul_36" -> "2892 add_63" [label="(1, 12, 64, 64)", style=solid]; -"2862 pad_20" -> "2863 view_101" [label="(1, 16, 16, 384)", style=solid]; -"2863 view_101" -> "2864 permute_83" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2864 permute_83" -> "2865 reshape_81" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2865 reshape_81" -> "2869 linear_112" [label="(4, 64, 384)", style=solid]; -"2866 clone_18" -> "2867 slice_273" [label="(1152,)", style=solid]; -"2866 clone_18" -> "2869 linear_112" [label="(1152,)", style=solid]; -"2867 slice_273" -> "2868 zero__18" [label="(384,)", style=solid]; -"2869 linear_112" -> "2870 reshape_82" [label="(4, 64, 1152)", style=solid]; -"2870 reshape_82" -> "2871 permute_84" [label="(4, 64, 3, 12, 32)", style=solid]; -"2871 permute_84" -> "2872 select_54" [label="(3, 4, 12, 64, 32)", style=solid]; -"2871 permute_84" -> "2873 select_55" [label="(3, 4, 12, 64, 32)", style=solid]; -"2871 permute_84" -> "2874 select_56" [label="(3, 4, 12, 64, 32)", style=solid]; -"2872 select_54" -> "2875 linalg_vector_norm_36" [label="(4, 12, 64, 32)", style=solid]; -"2872 select_54" -> "2877 expand_as_36" [label="(4, 12, 64, 32)", style=solid]; -"2872 select_54" -> "2878 div_36" [label="(4, 12, 64, 32)", style=solid]; -"2873 select_55" -> "2881 linalg_vector_norm_37" [label="(4, 12, 64, 32)", style=solid]; -"2873 select_55" -> "2883 expand_as_37" [label="(4, 12, 64, 32)", style=solid]; -"2873 select_55" -> "2884 div_37" [label="(4, 12, 64, 32)", style=solid]; -"2874 select_56" -> "2895 matmul_37" [label="(4, 12, 64, 32)", style=solid]; -"2875 linalg_vector_norm_36" -> "2876 clamp_min_36" [label="(4, 12, 64, 1)", style=solid]; -"2876 clamp_min_36" -> "2877 expand_as_36" [label="(4, 12, 64, 1)", style=solid]; -"2877 expand_as_36" -> "2878 div_36" [label="(4, 12, 64, 32)", style=solid]; -"2878 div_36" -> "2879 quantize_per_tensor_default_112" [label="(4, 12, 64, 32)", style=solid]; -"2879 quantize_per_tensor_default_112" -> "2880 dequantize_per_tensor_default_112" [label=None, style=solid]; -"2880 dequantize_per_tensor_default_112" -> "2888 matmul_36" [label=None, style=solid]; -"2881 linalg_vector_norm_37" -> "2882 clamp_min_37" [label="(4, 12, 64, 1)", style=solid]; -"2882 clamp_min_37" -> "2883 expand_as_37" [label="(4, 12, 64, 1)", style=solid]; -"2883 expand_as_37" -> "2884 div_37" [label="(4, 12, 64, 32)", style=solid]; -"2884 div_37" -> "2885 quantize_per_tensor_default_113" [label="(4, 12, 64, 32)", style=solid]; -"2885 quantize_per_tensor_default_113" -> "2886 dequantize_per_tensor_default_113" [label=None, style=solid]; -"2886 dequantize_per_tensor_default_113" -> "2887 transpose_36" [label=None, style=solid]; -"2887 transpose_36" -> "2888 matmul_36" [label="(4, 12, 32, 64)", style=solid]; -"2888 matmul_36" -> "2891 mul_37" [label="(4, 12, 64, 64)", style=solid]; -"2889 clamp_18" -> "2890 exp_18" [label="(12, 1, 1)", style=solid]; -"2890 exp_18" -> "2891 mul_37" [label="(12, 1, 1)", style=solid]; -"2891 mul_37" -> "2892 add_63" [label="(4, 12, 64, 64)", style=solid]; -"2892 add_63" -> "2893 softmax_18" [label="(4, 12, 64, 64)", style=solid]; -"2893 softmax_18" -> "2894 dropout_72" [label="(4, 12, 64, 64)", style=solid]; -"2894 dropout_72" -> "2895 matmul_37" [label="(4, 12, 64, 64)", style=solid]; -"2895 matmul_37" -> "2896 quantize_per_tensor_default_114" [label="(4, 12, 64, 32)", style=solid]; -"2896 quantize_per_tensor_default_114" -> "2897 dequantize_per_tensor_default_114" [label=None, style=solid]; -"2897 dequantize_per_tensor_default_114" -> "2898 transpose_37" [label=None, style=solid]; -"2898 transpose_37" -> "2899 reshape_83" [label="(4, 64, 12, 32)", style=solid]; -"2899 reshape_83" -> "2900 linear_113" [label="(4, 64, 384)", style=solid]; -"2900 linear_113" -> "2901 dropout_73" [label="(4, 64, 384)", style=solid]; -"2901 dropout_73" -> "2902 view_102" [label="(4, 64, 384)", style=solid]; -"2902 view_102" -> "2903 permute_85" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2903 permute_85" -> "2904 reshape_84" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2904 reshape_84" -> "2905 slice_274" [label="(1, 16, 16, 384)", style=solid]; -"2905 slice_274" -> "2906 slice_275" [label="(1, 16, 16, 384)", style=solid]; -"2906 slice_275" -> "2907 slice_276" [label="(1, 14, 16, 384)", style=solid]; -"2907 slice_276" -> "2908 slice_277" [label="(1, 14, 14, 384)", style=solid]; -"2908 slice_277" -> "2909 contiguous_35" [label="(1, 14, 14, 384)", style=solid]; -"2909 contiguous_35" -> "2910 layer_norm_39" [label="(1, 14, 14, 384)", style=solid]; -"2910 layer_norm_39" -> "2911 add_64" [label="(1, 14, 14, 384)", style=solid]; -"2911 add_64" -> "2912 quantize_per_tensor_default_115" [label="(1, 14, 14, 384)", style=solid]; -"2911 add_64" -> "2922 add_65" [label="(1, 14, 14, 384)", style=solid]; -"2912 quantize_per_tensor_default_115" -> "2913 dequantize_per_tensor_default_115" [label=None, style=solid]; -"2913 dequantize_per_tensor_default_115" -> "2914 linear_114" [label=None, style=solid]; -"2914 linear_114" -> "2915 gelu_18" [label="(1, 14, 14, 1536)", style=solid]; -"2915 gelu_18" -> "2916 quantize_per_tensor_default_116" [label="(1, 14, 14, 1536)", style=solid]; -"2916 quantize_per_tensor_default_116" -> "2917 dequantize_per_tensor_default_116" [label=None, style=solid]; -"2917 dequantize_per_tensor_default_116" -> "2918 dropout_74" [label=None, style=solid]; -"2918 dropout_74" -> "2919 linear_115" [label="(1, 14, 14, 1536)", style=solid]; -"2919 linear_115" -> "2920 dropout_75" [label="(1, 14, 14, 384)", style=solid]; -"2920 dropout_75" -> "2921 layer_norm_40" [label="(1, 14, 14, 384)", style=solid]; -"2921 layer_norm_40" -> "2922 add_65" [label="(1, 14, 14, 384)", style=solid]; -"2922 add_65" -> "2934 pad_21" [label="(1, 14, 14, 384)", style=solid]; -"2922 add_65" -> "3039 add_68" [label="(1, 14, 14, 384)", style=solid]; -"2923 linear_116" -> "2924 relu__19" [label="(1, 15, 15, 512)", style=solid]; -"2924 relu__19" -> "2925 linear_117" [label="(1, 15, 15, 512)", style=solid]; -"2925 linear_117" -> "2926 view_103" [label="(1, 15, 15, 12)", style=solid]; -"2926 view_103" -> "2927 index_19" [label="(225, 12)", style=solid]; -"2927 index_19" -> "2928 view_104" [label="(4096, 12)", style=solid]; -"2928 view_104" -> "2929 permute_86" [label="(64, 64, 12)", style=solid]; -"2929 permute_86" -> "2930 contiguous_36" [label="(12, 64, 64)", style=solid]; -"2930 contiguous_36" -> "2931 unsqueeze_55" [label="(12, 64, 64)", style=solid]; -"2931 unsqueeze_55" -> "2932 sigmoid_19" [label="(1, 12, 64, 64)", style=solid]; -"2932 sigmoid_19" -> "2933 mul_38" [label="(1, 12, 64, 64)", style=solid]; -"2933 mul_38" -> "2967 add_66" [label="(1, 12, 64, 64)", style=solid]; -"2934 pad_21" -> "2935 roll_18" [label="(1, 16, 16, 384)", style=solid]; -"2935 roll_18" -> "2936 view_105" [label="(1, 16, 16, 384)", style=solid]; -"2936 view_105" -> "2937 permute_87" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"2937 permute_87" -> "2938 reshape_85" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"2938 reshape_85" -> "2939 quantize_per_tensor_default_117" [label="(4, 64, 384)", style=solid]; -"2938 reshape_85" -> "2968 new_zeros_9" [label="(4, 64, 384)", style=solid]; -"2939 quantize_per_tensor_default_117" -> "2940 dequantize_per_tensor_default_117" [label=None, style=solid]; -"2940 dequantize_per_tensor_default_117" -> "2944 linear_118" [label=None, style=solid]; -"2941 clone_19" -> "2942 slice_278" [label="(1152,)", style=solid]; -"2941 clone_19" -> "2944 linear_118" [label="(1152,)", style=solid]; -"2942 slice_278" -> "2943 zero__19" [label="(384,)", style=solid]; -"2944 linear_118" -> "2945 reshape_86" [label="(4, 64, 1152)", style=solid]; -"2945 reshape_86" -> "2946 permute_88" [label="(4, 64, 3, 12, 32)", style=solid]; -"2946 permute_88" -> "2947 select_57" [label="(3, 4, 12, 64, 32)", style=solid]; -"2946 permute_88" -> "2948 select_58" [label="(3, 4, 12, 64, 32)", style=solid]; -"2946 permute_88" -> "2949 select_59" [label="(3, 4, 12, 64, 32)", style=solid]; -"2947 select_57" -> "2950 linalg_vector_norm_38" [label="(4, 12, 64, 32)", style=solid]; -"2947 select_57" -> "2952 expand_as_38" [label="(4, 12, 64, 32)", style=solid]; -"2947 select_57" -> "2953 div_38" [label="(4, 12, 64, 32)", style=solid]; -"2948 select_58" -> "2956 linalg_vector_norm_39" [label="(4, 12, 64, 32)", style=solid]; -"2948 select_58" -> "2958 expand_as_39" [label="(4, 12, 64, 32)", style=solid]; -"2948 select_58" -> "2959 div_39" [label="(4, 12, 64, 32)", style=solid]; -"2949 select_59" -> "3022 matmul_39" [label="(4, 12, 64, 32)", style=solid]; -"2950 linalg_vector_norm_38" -> "2951 clamp_min_38" [label="(4, 12, 64, 1)", style=solid]; -"2951 clamp_min_38" -> "2952 expand_as_38" [label="(4, 12, 64, 1)", style=solid]; -"2952 expand_as_38" -> "2953 div_38" [label="(4, 12, 64, 32)", style=solid]; -"2953 div_38" -> "2954 quantize_per_tensor_default_118" [label="(4, 12, 64, 32)", style=solid]; -"2954 quantize_per_tensor_default_118" -> "2955 dequantize_per_tensor_default_118" [label=None, style=solid]; -"2955 dequantize_per_tensor_default_118" -> "2963 matmul_38" [label=None, style=solid]; -"2956 linalg_vector_norm_39" -> "2957 clamp_min_39" [label="(4, 12, 64, 1)", style=solid]; -"2957 clamp_min_39" -> "2958 expand_as_39" [label="(4, 12, 64, 1)", style=solid]; -"2958 expand_as_39" -> "2959 div_39" [label="(4, 12, 64, 32)", style=solid]; -"2959 div_39" -> "2960 quantize_per_tensor_default_119" [label="(4, 12, 64, 32)", style=solid]; -"2960 quantize_per_tensor_default_119" -> "2961 dequantize_per_tensor_default_119" [label=None, style=solid]; -"2961 dequantize_per_tensor_default_119" -> "2962 transpose_38" [label=None, style=solid]; -"2962 transpose_38" -> "2963 matmul_38" [label="(4, 12, 32, 64)", style=solid]; -"2963 matmul_38" -> "2966 mul_39" [label="(4, 12, 64, 64)", style=solid]; -"2964 clamp_19" -> "2965 exp_19" [label="(12, 1, 1)", style=solid]; -"2965 exp_19" -> "2966 mul_39" [label="(12, 1, 1)", style=solid]; -"2966 mul_39" -> "2967 add_66" [label="(4, 12, 64, 64)", style=solid]; -"2967 add_66" -> "3015 view_107" [label="(4, 12, 64, 64)", style=solid]; -"2968 new_zeros_9" -> "2970 slice_279" [label="(16, 16)", style=solid]; -"2968 new_zeros_9" -> "2974 slice_281" [label="(16, 16)", style=solid]; -"2968 new_zeros_9" -> "2978 slice_283" [label="(16, 16)", style=solid]; -"2968 new_zeros_9" -> "2982 slice_285" [label="(16, 16)", style=solid]; -"2968 new_zeros_9" -> "2986 slice_287" [label="(16, 16)", style=solid]; -"2968 new_zeros_9" -> "2990 slice_289" [label="(16, 16)", style=solid]; -"2968 new_zeros_9" -> "2994 slice_291" [label="(16, 16)", style=solid]; -"2968 new_zeros_9" -> "2998 slice_293" [label="(16, 16)", style=solid]; -"2968 new_zeros_9" -> "3002 slice_295" [label="(16, 16)", style=solid]; -"2968 new_zeros_9" -> "3005 view_106" [label="(16, 16)", style=solid]; -"2969 lift_fresh_copy_81" -> "2972 fill__81" [label="()", style=solid]; -"2970 slice_279" -> "2971 slice_280" [label="(8, 16)", style=solid]; -"2971 slice_280" -> "2972 fill__81" [label="(8, 8)", style=solid]; -"2973 lift_fresh_copy_82" -> "2976 fill__82" [label="()", style=solid]; -"2974 slice_281" -> "2975 slice_282" [label="(8, 16)", style=solid]; -"2975 slice_282" -> "2976 fill__82" [label="(8, 4)", style=solid]; -"2977 lift_fresh_copy_83" -> "2980 fill__83" [label="()", style=solid]; -"2978 slice_283" -> "2979 slice_284" [label="(8, 16)", style=solid]; -"2979 slice_284" -> "2980 fill__83" [label="(8, 4)", style=solid]; -"2981 lift_fresh_copy_84" -> "2984 fill__84" [label="()", style=solid]; -"2982 slice_285" -> "2983 slice_286" [label="(4, 16)", style=solid]; -"2983 slice_286" -> "2984 fill__84" [label="(4, 8)", style=solid]; -"2985 lift_fresh_copy_85" -> "2988 fill__85" [label="()", style=solid]; -"2986 slice_287" -> "2987 slice_288" [label="(4, 16)", style=solid]; -"2987 slice_288" -> "2988 fill__85" [label="(4, 4)", style=solid]; -"2989 lift_fresh_copy_86" -> "2992 fill__86" [label="()", style=solid]; -"2990 slice_289" -> "2991 slice_290" [label="(4, 16)", style=solid]; -"2991 slice_290" -> "2992 fill__86" [label="(4, 4)", style=solid]; -"2993 lift_fresh_copy_87" -> "2996 fill__87" [label="()", style=solid]; -"2994 slice_291" -> "2995 slice_292" [label="(4, 16)", style=solid]; -"2995 slice_292" -> "2996 fill__87" [label="(4, 8)", style=solid]; -"2997 lift_fresh_copy_88" -> "3000 fill__88" [label="()", style=solid]; -"2998 slice_293" -> "2999 slice_294" [label="(4, 16)", style=solid]; -"2999 slice_294" -> "3000 fill__88" [label="(4, 4)", style=solid]; -"3001 lift_fresh_copy_89" -> "3004 fill__89" [label="()", style=solid]; -"3002 slice_295" -> "3003 slice_296" [label="(4, 16)", style=solid]; -"3003 slice_296" -> "3004 fill__89" [label="(4, 4)", style=solid]; -"3005 view_106" -> "3006 permute_89" [label="(2, 8, 2, 8)", style=solid]; -"3006 permute_89" -> "3007 reshape_87" [label="(2, 2, 8, 8)", style=solid]; -"3007 reshape_87" -> "3008 unsqueeze_56" [label="(4, 64)", style=solid]; -"3007 reshape_87" -> "3009 unsqueeze_57" [label="(4, 64)", style=solid]; -"3008 unsqueeze_56" -> "3010 sub_9" [label="(4, 1, 64)", style=solid]; -"3009 unsqueeze_57" -> "3010 sub_9" [label="(4, 64, 1)", style=solid]; -"3010 sub_9" -> "3011 ne_9" [label="(4, 64, 64)", style=solid]; -"3010 sub_9" -> "3012 masked_fill_18" [label="(4, 64, 64)", style=solid]; -"3010 sub_9" -> "3013 eq_9" [label="(4, 64, 64)", style=solid]; -"3011 ne_9" -> "3012 masked_fill_18" [label="(4, 64, 64)", style=solid]; -"3012 masked_fill_18" -> "3014 masked_fill_19" [label="(4, 64, 64)", style=solid]; -"3013 eq_9" -> "3014 masked_fill_19" [label="(4, 64, 64)", style=solid]; -"3014 masked_fill_19" -> "3016 unsqueeze_58" [label="(4, 64, 64)", style=solid]; -"3015 view_107" -> "3018 add_67" [label="(1, 4, 12, 64, 64)", style=solid]; -"3016 unsqueeze_58" -> "3017 unsqueeze_59" [label="(4, 1, 64, 64)", style=solid]; -"3017 unsqueeze_59" -> "3018 add_67" [label="(1, 4, 1, 64, 64)", style=solid]; -"3018 add_67" -> "3019 view_108" [label="(1, 4, 12, 64, 64)", style=solid]; -"3019 view_108" -> "3020 softmax_19" [label="(4, 12, 64, 64)", style=solid]; -"3020 softmax_19" -> "3021 dropout_76" [label="(4, 12, 64, 64)", style=solid]; -"3021 dropout_76" -> "3022 matmul_39" [label="(4, 12, 64, 64)", style=solid]; -"3022 matmul_39" -> "3023 quantize_per_tensor_default_120" [label="(4, 12, 64, 32)", style=solid]; -"3023 quantize_per_tensor_default_120" -> "3024 dequantize_per_tensor_default_120" [label=None, style=solid]; -"3024 dequantize_per_tensor_default_120" -> "3025 transpose_39" [label=None, style=solid]; -"3025 transpose_39" -> "3026 reshape_88" [label="(4, 64, 12, 32)", style=solid]; -"3026 reshape_88" -> "3027 linear_119" [label="(4, 64, 384)", style=solid]; -"3027 linear_119" -> "3028 dropout_77" [label="(4, 64, 384)", style=solid]; -"3028 dropout_77" -> "3029 view_109" [label="(4, 64, 384)", style=solid]; -"3029 view_109" -> "3030 permute_90" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"3030 permute_90" -> "3031 reshape_89" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"3031 reshape_89" -> "3032 roll_19" [label="(1, 16, 16, 384)", style=solid]; -"3032 roll_19" -> "3033 slice_297" [label="(1, 16, 16, 384)", style=solid]; -"3033 slice_297" -> "3034 slice_298" [label="(1, 16, 16, 384)", style=solid]; -"3034 slice_298" -> "3035 slice_299" [label="(1, 14, 16, 384)", style=solid]; -"3035 slice_299" -> "3036 slice_300" [label="(1, 14, 14, 384)", style=solid]; -"3036 slice_300" -> "3037 contiguous_37" [label="(1, 14, 14, 384)", style=solid]; -"3037 contiguous_37" -> "3038 layer_norm_41" [label="(1, 14, 14, 384)", style=solid]; -"3038 layer_norm_41" -> "3039 add_68" [label="(1, 14, 14, 384)", style=solid]; -"3039 add_68" -> "3040 quantize_per_tensor_default_121" [label="(1, 14, 14, 384)", style=solid]; -"3039 add_68" -> "3050 add_69" [label="(1, 14, 14, 384)", style=solid]; -"3040 quantize_per_tensor_default_121" -> "3041 dequantize_per_tensor_default_121" [label=None, style=solid]; -"3041 dequantize_per_tensor_default_121" -> "3042 linear_120" [label=None, style=solid]; -"3042 linear_120" -> "3043 gelu_19" [label="(1, 14, 14, 1536)", style=solid]; -"3043 gelu_19" -> "3044 quantize_per_tensor_default_122" [label="(1, 14, 14, 1536)", style=solid]; -"3044 quantize_per_tensor_default_122" -> "3045 dequantize_per_tensor_default_122" [label=None, style=solid]; -"3045 dequantize_per_tensor_default_122" -> "3046 dropout_78" [label=None, style=solid]; -"3046 dropout_78" -> "3047 linear_121" [label="(1, 14, 14, 1536)", style=solid]; -"3047 linear_121" -> "3048 dropout_79" [label="(1, 14, 14, 384)", style=solid]; -"3048 dropout_79" -> "3049 layer_norm_42" [label="(1, 14, 14, 384)", style=solid]; -"3049 layer_norm_42" -> "3050 add_69" [label="(1, 14, 14, 384)", style=solid]; -"3050 add_69" -> "3051 quantize_per_tensor_default_123" [label="(1, 14, 14, 384)", style=solid]; -"3050 add_69" -> "3113 add_71" [label="(1, 14, 14, 384)", style=solid]; -"3051 quantize_per_tensor_default_123" -> "3052 dequantize_per_tensor_default_123" [label=None, style=solid]; -"3052 dequantize_per_tensor_default_123" -> "3064 pad_22" [label=None, style=solid]; -"3053 linear_122" -> "3054 relu__20" [label="(1, 15, 15, 512)", style=solid]; -"3054 relu__20" -> "3055 linear_123" [label="(1, 15, 15, 512)", style=solid]; -"3055 linear_123" -> "3056 view_110" [label="(1, 15, 15, 12)", style=solid]; -"3056 view_110" -> "3057 index_20" [label="(225, 12)", style=solid]; -"3057 index_20" -> "3058 view_111" [label="(4096, 12)", style=solid]; -"3058 view_111" -> "3059 permute_91" [label="(64, 64, 12)", style=solid]; -"3059 permute_91" -> "3060 contiguous_38" [label="(12, 64, 64)", style=solid]; -"3060 contiguous_38" -> "3061 unsqueeze_60" [label="(12, 64, 64)", style=solid]; -"3061 unsqueeze_60" -> "3062 sigmoid_20" [label="(1, 12, 64, 64)", style=solid]; -"3062 sigmoid_20" -> "3063 mul_40" [label="(1, 12, 64, 64)", style=solid]; -"3063 mul_40" -> "3094 add_70" [label="(1, 12, 64, 64)", style=solid]; -"3064 pad_22" -> "3065 view_112" [label="(1, 16, 16, 384)", style=solid]; -"3065 view_112" -> "3066 permute_92" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"3066 permute_92" -> "3067 reshape_90" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"3067 reshape_90" -> "3071 linear_124" [label="(4, 64, 384)", style=solid]; -"3068 clone_20" -> "3069 slice_301" [label="(1152,)", style=solid]; -"3068 clone_20" -> "3071 linear_124" [label="(1152,)", style=solid]; -"3069 slice_301" -> "3070 zero__20" [label="(384,)", style=solid]; -"3071 linear_124" -> "3072 reshape_91" [label="(4, 64, 1152)", style=solid]; -"3072 reshape_91" -> "3073 permute_93" [label="(4, 64, 3, 12, 32)", style=solid]; -"3073 permute_93" -> "3074 select_60" [label="(3, 4, 12, 64, 32)", style=solid]; -"3073 permute_93" -> "3075 select_61" [label="(3, 4, 12, 64, 32)", style=solid]; -"3073 permute_93" -> "3076 select_62" [label="(3, 4, 12, 64, 32)", style=solid]; -"3074 select_60" -> "3077 linalg_vector_norm_40" [label="(4, 12, 64, 32)", style=solid]; -"3074 select_60" -> "3079 expand_as_40" [label="(4, 12, 64, 32)", style=solid]; -"3074 select_60" -> "3080 div_40" [label="(4, 12, 64, 32)", style=solid]; -"3075 select_61" -> "3083 linalg_vector_norm_41" [label="(4, 12, 64, 32)", style=solid]; -"3075 select_61" -> "3085 expand_as_41" [label="(4, 12, 64, 32)", style=solid]; -"3075 select_61" -> "3086 div_41" [label="(4, 12, 64, 32)", style=solid]; -"3076 select_62" -> "3097 matmul_41" [label="(4, 12, 64, 32)", style=solid]; -"3077 linalg_vector_norm_40" -> "3078 clamp_min_40" [label="(4, 12, 64, 1)", style=solid]; -"3078 clamp_min_40" -> "3079 expand_as_40" [label="(4, 12, 64, 1)", style=solid]; -"3079 expand_as_40" -> "3080 div_40" [label="(4, 12, 64, 32)", style=solid]; -"3080 div_40" -> "3081 quantize_per_tensor_default_124" [label="(4, 12, 64, 32)", style=solid]; -"3081 quantize_per_tensor_default_124" -> "3082 dequantize_per_tensor_default_124" [label=None, style=solid]; -"3082 dequantize_per_tensor_default_124" -> "3090 matmul_40" [label=None, style=solid]; -"3083 linalg_vector_norm_41" -> "3084 clamp_min_41" [label="(4, 12, 64, 1)", style=solid]; -"3084 clamp_min_41" -> "3085 expand_as_41" [label="(4, 12, 64, 1)", style=solid]; -"3085 expand_as_41" -> "3086 div_41" [label="(4, 12, 64, 32)", style=solid]; -"3086 div_41" -> "3087 quantize_per_tensor_default_125" [label="(4, 12, 64, 32)", style=solid]; -"3087 quantize_per_tensor_default_125" -> "3088 dequantize_per_tensor_default_125" [label=None, style=solid]; -"3088 dequantize_per_tensor_default_125" -> "3089 transpose_40" [label=None, style=solid]; -"3089 transpose_40" -> "3090 matmul_40" [label="(4, 12, 32, 64)", style=solid]; -"3090 matmul_40" -> "3093 mul_41" [label="(4, 12, 64, 64)", style=solid]; -"3091 clamp_20" -> "3092 exp_20" [label="(12, 1, 1)", style=solid]; -"3092 exp_20" -> "3093 mul_41" [label="(12, 1, 1)", style=solid]; -"3093 mul_41" -> "3094 add_70" [label="(4, 12, 64, 64)", style=solid]; -"3094 add_70" -> "3095 softmax_20" [label="(4, 12, 64, 64)", style=solid]; -"3095 softmax_20" -> "3096 dropout_80" [label="(4, 12, 64, 64)", style=solid]; -"3096 dropout_80" -> "3097 matmul_41" [label="(4, 12, 64, 64)", style=solid]; -"3097 matmul_41" -> "3098 quantize_per_tensor_default_126" [label="(4, 12, 64, 32)", style=solid]; -"3098 quantize_per_tensor_default_126" -> "3099 dequantize_per_tensor_default_126" [label=None, style=solid]; -"3099 dequantize_per_tensor_default_126" -> "3100 transpose_41" [label=None, style=solid]; -"3100 transpose_41" -> "3101 reshape_92" [label="(4, 64, 12, 32)", style=solid]; -"3101 reshape_92" -> "3102 linear_125" [label="(4, 64, 384)", style=solid]; -"3102 linear_125" -> "3103 dropout_81" [label="(4, 64, 384)", style=solid]; -"3103 dropout_81" -> "3104 view_113" [label="(4, 64, 384)", style=solid]; -"3104 view_113" -> "3105 permute_94" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"3105 permute_94" -> "3106 reshape_93" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"3106 reshape_93" -> "3107 slice_302" [label="(1, 16, 16, 384)", style=solid]; -"3107 slice_302" -> "3108 slice_303" [label="(1, 16, 16, 384)", style=solid]; -"3108 slice_303" -> "3109 slice_304" [label="(1, 14, 16, 384)", style=solid]; -"3109 slice_304" -> "3110 slice_305" [label="(1, 14, 14, 384)", style=solid]; -"3110 slice_305" -> "3111 contiguous_39" [label="(1, 14, 14, 384)", style=solid]; -"3111 contiguous_39" -> "3112 layer_norm_43" [label="(1, 14, 14, 384)", style=solid]; -"3112 layer_norm_43" -> "3113 add_71" [label="(1, 14, 14, 384)", style=solid]; -"3113 add_71" -> "3114 quantize_per_tensor_default_127" [label="(1, 14, 14, 384)", style=solid]; -"3113 add_71" -> "3124 add_72" [label="(1, 14, 14, 384)", style=solid]; -"3114 quantize_per_tensor_default_127" -> "3115 dequantize_per_tensor_default_127" [label=None, style=solid]; -"3115 dequantize_per_tensor_default_127" -> "3116 linear_126" [label=None, style=solid]; -"3116 linear_126" -> "3117 gelu_20" [label="(1, 14, 14, 1536)", style=solid]; -"3117 gelu_20" -> "3118 quantize_per_tensor_default_128" [label="(1, 14, 14, 1536)", style=solid]; -"3118 quantize_per_tensor_default_128" -> "3119 dequantize_per_tensor_default_128" [label=None, style=solid]; -"3119 dequantize_per_tensor_default_128" -> "3120 dropout_82" [label=None, style=solid]; -"3120 dropout_82" -> "3121 linear_127" [label="(1, 14, 14, 1536)", style=solid]; -"3121 linear_127" -> "3122 dropout_83" [label="(1, 14, 14, 384)", style=solid]; -"3122 dropout_83" -> "3123 layer_norm_44" [label="(1, 14, 14, 384)", style=solid]; -"3123 layer_norm_44" -> "3124 add_72" [label="(1, 14, 14, 384)", style=solid]; -"3124 add_72" -> "3136 pad_23" [label="(1, 14, 14, 384)", style=solid]; -"3124 add_72" -> "3241 add_75" [label="(1, 14, 14, 384)", style=solid]; -"3125 linear_128" -> "3126 relu__21" [label="(1, 15, 15, 512)", style=solid]; -"3126 relu__21" -> "3127 linear_129" [label="(1, 15, 15, 512)", style=solid]; -"3127 linear_129" -> "3128 view_114" [label="(1, 15, 15, 12)", style=solid]; -"3128 view_114" -> "3129 index_21" [label="(225, 12)", style=solid]; -"3129 index_21" -> "3130 view_115" [label="(4096, 12)", style=solid]; -"3130 view_115" -> "3131 permute_95" [label="(64, 64, 12)", style=solid]; -"3131 permute_95" -> "3132 contiguous_40" [label="(12, 64, 64)", style=solid]; -"3132 contiguous_40" -> "3133 unsqueeze_61" [label="(12, 64, 64)", style=solid]; -"3133 unsqueeze_61" -> "3134 sigmoid_21" [label="(1, 12, 64, 64)", style=solid]; -"3134 sigmoid_21" -> "3135 mul_42" [label="(1, 12, 64, 64)", style=solid]; -"3135 mul_42" -> "3169 add_73" [label="(1, 12, 64, 64)", style=solid]; -"3136 pad_23" -> "3137 roll_20" [label="(1, 16, 16, 384)", style=solid]; -"3137 roll_20" -> "3138 view_116" [label="(1, 16, 16, 384)", style=solid]; -"3138 view_116" -> "3139 permute_96" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"3139 permute_96" -> "3140 reshape_94" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"3140 reshape_94" -> "3141 quantize_per_tensor_default_129" [label="(4, 64, 384)", style=solid]; -"3140 reshape_94" -> "3170 new_zeros_10" [label="(4, 64, 384)", style=solid]; -"3141 quantize_per_tensor_default_129" -> "3142 dequantize_per_tensor_default_129" [label=None, style=solid]; -"3142 dequantize_per_tensor_default_129" -> "3146 linear_130" [label=None, style=solid]; -"3143 clone_21" -> "3144 slice_306" [label="(1152,)", style=solid]; -"3143 clone_21" -> "3146 linear_130" [label="(1152,)", style=solid]; -"3144 slice_306" -> "3145 zero__21" [label="(384,)", style=solid]; -"3146 linear_130" -> "3147 reshape_95" [label="(4, 64, 1152)", style=solid]; -"3147 reshape_95" -> "3148 permute_97" [label="(4, 64, 3, 12, 32)", style=solid]; -"3148 permute_97" -> "3149 select_63" [label="(3, 4, 12, 64, 32)", style=solid]; -"3148 permute_97" -> "3150 select_64" [label="(3, 4, 12, 64, 32)", style=solid]; -"3148 permute_97" -> "3151 select_65" [label="(3, 4, 12, 64, 32)", style=solid]; -"3149 select_63" -> "3152 linalg_vector_norm_42" [label="(4, 12, 64, 32)", style=solid]; -"3149 select_63" -> "3154 expand_as_42" [label="(4, 12, 64, 32)", style=solid]; -"3149 select_63" -> "3155 div_42" [label="(4, 12, 64, 32)", style=solid]; -"3150 select_64" -> "3158 linalg_vector_norm_43" [label="(4, 12, 64, 32)", style=solid]; -"3150 select_64" -> "3160 expand_as_43" [label="(4, 12, 64, 32)", style=solid]; -"3150 select_64" -> "3161 div_43" [label="(4, 12, 64, 32)", style=solid]; -"3151 select_65" -> "3224 matmul_43" [label="(4, 12, 64, 32)", style=solid]; -"3152 linalg_vector_norm_42" -> "3153 clamp_min_42" [label="(4, 12, 64, 1)", style=solid]; -"3153 clamp_min_42" -> "3154 expand_as_42" [label="(4, 12, 64, 1)", style=solid]; -"3154 expand_as_42" -> "3155 div_42" [label="(4, 12, 64, 32)", style=solid]; -"3155 div_42" -> "3156 quantize_per_tensor_default_130" [label="(4, 12, 64, 32)", style=solid]; -"3156 quantize_per_tensor_default_130" -> "3157 dequantize_per_tensor_default_130" [label=None, style=solid]; -"3157 dequantize_per_tensor_default_130" -> "3165 matmul_42" [label=None, style=solid]; -"3158 linalg_vector_norm_43" -> "3159 clamp_min_43" [label="(4, 12, 64, 1)", style=solid]; -"3159 clamp_min_43" -> "3160 expand_as_43" [label="(4, 12, 64, 1)", style=solid]; -"3160 expand_as_43" -> "3161 div_43" [label="(4, 12, 64, 32)", style=solid]; -"3161 div_43" -> "3162 quantize_per_tensor_default_131" [label="(4, 12, 64, 32)", style=solid]; -"3162 quantize_per_tensor_default_131" -> "3163 dequantize_per_tensor_default_131" [label=None, style=solid]; -"3163 dequantize_per_tensor_default_131" -> "3164 transpose_42" [label=None, style=solid]; -"3164 transpose_42" -> "3165 matmul_42" [label="(4, 12, 32, 64)", style=solid]; -"3165 matmul_42" -> "3168 mul_43" [label="(4, 12, 64, 64)", style=solid]; -"3166 clamp_21" -> "3167 exp_21" [label="(12, 1, 1)", style=solid]; -"3167 exp_21" -> "3168 mul_43" [label="(12, 1, 1)", style=solid]; -"3168 mul_43" -> "3169 add_73" [label="(4, 12, 64, 64)", style=solid]; -"3169 add_73" -> "3217 view_118" [label="(4, 12, 64, 64)", style=solid]; -"3170 new_zeros_10" -> "3172 slice_307" [label="(16, 16)", style=solid]; -"3170 new_zeros_10" -> "3176 slice_309" [label="(16, 16)", style=solid]; -"3170 new_zeros_10" -> "3180 slice_311" [label="(16, 16)", style=solid]; -"3170 new_zeros_10" -> "3184 slice_313" [label="(16, 16)", style=solid]; -"3170 new_zeros_10" -> "3188 slice_315" [label="(16, 16)", style=solid]; -"3170 new_zeros_10" -> "3192 slice_317" [label="(16, 16)", style=solid]; -"3170 new_zeros_10" -> "3196 slice_319" [label="(16, 16)", style=solid]; -"3170 new_zeros_10" -> "3200 slice_321" [label="(16, 16)", style=solid]; -"3170 new_zeros_10" -> "3204 slice_323" [label="(16, 16)", style=solid]; -"3170 new_zeros_10" -> "3207 view_117" [label="(16, 16)", style=solid]; -"3171 lift_fresh_copy_90" -> "3174 fill__90" [label="()", style=solid]; -"3172 slice_307" -> "3173 slice_308" [label="(8, 16)", style=solid]; -"3173 slice_308" -> "3174 fill__90" [label="(8, 8)", style=solid]; -"3175 lift_fresh_copy_91" -> "3178 fill__91" [label="()", style=solid]; -"3176 slice_309" -> "3177 slice_310" [label="(8, 16)", style=solid]; -"3177 slice_310" -> "3178 fill__91" [label="(8, 4)", style=solid]; -"3179 lift_fresh_copy_92" -> "3182 fill__92" [label="()", style=solid]; -"3180 slice_311" -> "3181 slice_312" [label="(8, 16)", style=solid]; -"3181 slice_312" -> "3182 fill__92" [label="(8, 4)", style=solid]; -"3183 lift_fresh_copy_93" -> "3186 fill__93" [label="()", style=solid]; -"3184 slice_313" -> "3185 slice_314" [label="(4, 16)", style=solid]; -"3185 slice_314" -> "3186 fill__93" [label="(4, 8)", style=solid]; -"3187 lift_fresh_copy_94" -> "3190 fill__94" [label="()", style=solid]; -"3188 slice_315" -> "3189 slice_316" [label="(4, 16)", style=solid]; -"3189 slice_316" -> "3190 fill__94" [label="(4, 4)", style=solid]; -"3191 lift_fresh_copy_95" -> "3194 fill__95" [label="()", style=solid]; -"3192 slice_317" -> "3193 slice_318" [label="(4, 16)", style=solid]; -"3193 slice_318" -> "3194 fill__95" [label="(4, 4)", style=solid]; -"3195 lift_fresh_copy_96" -> "3198 fill__96" [label="()", style=solid]; -"3196 slice_319" -> "3197 slice_320" [label="(4, 16)", style=solid]; -"3197 slice_320" -> "3198 fill__96" [label="(4, 8)", style=solid]; -"3199 lift_fresh_copy_97" -> "3202 fill__97" [label="()", style=solid]; -"3200 slice_321" -> "3201 slice_322" [label="(4, 16)", style=solid]; -"3201 slice_322" -> "3202 fill__97" [label="(4, 4)", style=solid]; -"3203 lift_fresh_copy_98" -> "3206 fill__98" [label="()", style=solid]; -"3204 slice_323" -> "3205 slice_324" [label="(4, 16)", style=solid]; -"3205 slice_324" -> "3206 fill__98" [label="(4, 4)", style=solid]; -"3207 view_117" -> "3208 permute_98" [label="(2, 8, 2, 8)", style=solid]; -"3208 permute_98" -> "3209 reshape_96" [label="(2, 2, 8, 8)", style=solid]; -"3209 reshape_96" -> "3210 unsqueeze_62" [label="(4, 64)", style=solid]; -"3209 reshape_96" -> "3211 unsqueeze_63" [label="(4, 64)", style=solid]; -"3210 unsqueeze_62" -> "3212 sub_10" [label="(4, 1, 64)", style=solid]; -"3211 unsqueeze_63" -> "3212 sub_10" [label="(4, 64, 1)", style=solid]; -"3212 sub_10" -> "3213 ne_10" [label="(4, 64, 64)", style=solid]; -"3212 sub_10" -> "3214 masked_fill_20" [label="(4, 64, 64)", style=solid]; -"3212 sub_10" -> "3215 eq_10" [label="(4, 64, 64)", style=solid]; -"3213 ne_10" -> "3214 masked_fill_20" [label="(4, 64, 64)", style=solid]; -"3214 masked_fill_20" -> "3216 masked_fill_21" [label="(4, 64, 64)", style=solid]; -"3215 eq_10" -> "3216 masked_fill_21" [label="(4, 64, 64)", style=solid]; -"3216 masked_fill_21" -> "3218 unsqueeze_64" [label="(4, 64, 64)", style=solid]; -"3217 view_118" -> "3220 add_74" [label="(1, 4, 12, 64, 64)", style=solid]; -"3218 unsqueeze_64" -> "3219 unsqueeze_65" [label="(4, 1, 64, 64)", style=solid]; -"3219 unsqueeze_65" -> "3220 add_74" [label="(1, 4, 1, 64, 64)", style=solid]; -"3220 add_74" -> "3221 view_119" [label="(1, 4, 12, 64, 64)", style=solid]; -"3221 view_119" -> "3222 softmax_21" [label="(4, 12, 64, 64)", style=solid]; -"3222 softmax_21" -> "3223 dropout_84" [label="(4, 12, 64, 64)", style=solid]; -"3223 dropout_84" -> "3224 matmul_43" [label="(4, 12, 64, 64)", style=solid]; -"3224 matmul_43" -> "3225 quantize_per_tensor_default_132" [label="(4, 12, 64, 32)", style=solid]; -"3225 quantize_per_tensor_default_132" -> "3226 dequantize_per_tensor_default_132" [label=None, style=solid]; -"3226 dequantize_per_tensor_default_132" -> "3227 transpose_43" [label=None, style=solid]; -"3227 transpose_43" -> "3228 reshape_97" [label="(4, 64, 12, 32)", style=solid]; -"3228 reshape_97" -> "3229 linear_131" [label="(4, 64, 384)", style=solid]; -"3229 linear_131" -> "3230 dropout_85" [label="(4, 64, 384)", style=solid]; -"3230 dropout_85" -> "3231 view_120" [label="(4, 64, 384)", style=solid]; -"3231 view_120" -> "3232 permute_99" [label="(1, 2, 2, 8, 8, 384)", style=solid]; -"3232 permute_99" -> "3233 reshape_98" [label="(1, 2, 8, 2, 8, 384)", style=solid]; -"3233 reshape_98" -> "3234 roll_21" [label="(1, 16, 16, 384)", style=solid]; -"3234 roll_21" -> "3235 slice_325" [label="(1, 16, 16, 384)", style=solid]; -"3235 slice_325" -> "3236 slice_326" [label="(1, 16, 16, 384)", style=solid]; -"3236 slice_326" -> "3237 slice_327" [label="(1, 14, 16, 384)", style=solid]; -"3237 slice_327" -> "3238 slice_328" [label="(1, 14, 14, 384)", style=solid]; -"3238 slice_328" -> "3239 contiguous_41" [label="(1, 14, 14, 384)", style=solid]; -"3239 contiguous_41" -> "3240 layer_norm_45" [label="(1, 14, 14, 384)", style=solid]; -"3240 layer_norm_45" -> "3241 add_75" [label="(1, 14, 14, 384)", style=solid]; -"3241 add_75" -> "3242 quantize_per_tensor_default_133" [label="(1, 14, 14, 384)", style=solid]; -"3241 add_75" -> "3252 add_76" [label="(1, 14, 14, 384)", style=solid]; -"3242 quantize_per_tensor_default_133" -> "3243 dequantize_per_tensor_default_133" [label=None, style=solid]; -"3243 dequantize_per_tensor_default_133" -> "3244 linear_132" [label=None, style=solid]; -"3244 linear_132" -> "3245 gelu_21" [label="(1, 14, 14, 1536)", style=solid]; -"3245 gelu_21" -> "3246 quantize_per_tensor_default_134" [label="(1, 14, 14, 1536)", style=solid]; -"3246 quantize_per_tensor_default_134" -> "3247 dequantize_per_tensor_default_134" [label=None, style=solid]; -"3247 dequantize_per_tensor_default_134" -> "3248 dropout_86" [label=None, style=solid]; -"3248 dropout_86" -> "3249 linear_133" [label="(1, 14, 14, 1536)", style=solid]; -"3249 linear_133" -> "3250 dropout_87" [label="(1, 14, 14, 384)", style=solid]; -"3250 dropout_87" -> "3251 layer_norm_46" [label="(1, 14, 14, 384)", style=solid]; -"3251 layer_norm_46" -> "3252 add_76" [label="(1, 14, 14, 384)", style=solid]; -"3252 add_76" -> "3253 quantize_per_tensor_default_135" [label="(1, 14, 14, 384)", style=solid]; -"3253 quantize_per_tensor_default_135" -> "3254 dequantize_per_tensor_default_135" [label=None, style=solid]; -"3254 dequantize_per_tensor_default_135" -> "3255 pad_24" [label=None, style=solid]; -"3255 pad_24" -> "3256 slice_329" [label="(1, 14, 14, 384)", style=solid]; -"3255 pad_24" -> "3259 slice_332" [label="(1, 14, 14, 384)", style=solid]; -"3255 pad_24" -> "3262 slice_335" [label="(1, 14, 14, 384)", style=solid]; -"3255 pad_24" -> "3265 slice_338" [label="(1, 14, 14, 384)", style=solid]; -"3256 slice_329" -> "3257 slice_330" [label="(1, 7, 14, 384)", style=solid]; -"3257 slice_330" -> "3258 slice_331" [label="(1, 7, 7, 384)", style=solid]; -"3258 slice_331" -> "3268 cat_2" [label="(1, 7, 7, 384)", style=solid]; -"3259 slice_332" -> "3260 slice_333" [label="(1, 7, 14, 384)", style=solid]; -"3260 slice_333" -> "3261 slice_334" [label="(1, 7, 7, 384)", style=solid]; -"3261 slice_334" -> "3268 cat_2" [label="(1, 7, 7, 384)", style=solid]; -"3262 slice_335" -> "3263 slice_336" [label="(1, 7, 14, 384)", style=solid]; -"3263 slice_336" -> "3264 slice_337" [label="(1, 7, 7, 384)", style=solid]; -"3264 slice_337" -> "3268 cat_2" [label="(1, 7, 7, 384)", style=solid]; -"3265 slice_338" -> "3266 slice_339" [label="(1, 7, 14, 384)", style=solid]; -"3266 slice_339" -> "3267 slice_340" [label="(1, 7, 7, 384)", style=solid]; -"3267 slice_340" -> "3268 cat_2" [label="(1, 7, 7, 384)", style=solid]; -"3268 cat_2" -> "3269 linear_134" [label="(1, 7, 7, 1536)", style=solid]; -"3269 linear_134" -> "3270 layer_norm_47" [label="(1, 7, 7, 768)", style=solid]; -"3270 layer_norm_47" -> "3271 quantize_per_tensor_default_136" [label="(1, 7, 7, 768)", style=solid]; -"3270 layer_norm_47" -> "3333 add_78" [label="(1, 7, 7, 768)", style=solid]; -"3271 quantize_per_tensor_default_136" -> "3272 dequantize_per_tensor_default_136" [label=None, style=solid]; -"3272 dequantize_per_tensor_default_136" -> "3284 pad_25" [label=None, style=solid]; -"3273 linear_135" -> "3274 relu__22" [label="(1, 15, 15, 512)", style=solid]; -"3274 relu__22" -> "3275 linear_136" [label="(1, 15, 15, 512)", style=solid]; -"3275 linear_136" -> "3276 view_121" [label="(1, 15, 15, 24)", style=solid]; -"3276 view_121" -> "3277 index_22" [label="(225, 24)", style=solid]; -"3277 index_22" -> "3278 view_122" [label="(4096, 24)", style=solid]; -"3278 view_122" -> "3279 permute_100" [label="(64, 64, 24)", style=solid]; -"3279 permute_100" -> "3280 contiguous_42" [label="(24, 64, 64)", style=solid]; -"3280 contiguous_42" -> "3281 unsqueeze_66" [label="(24, 64, 64)", style=solid]; -"3281 unsqueeze_66" -> "3282 sigmoid_22" [label="(1, 24, 64, 64)", style=solid]; -"3282 sigmoid_22" -> "3283 mul_44" [label="(1, 24, 64, 64)", style=solid]; -"3283 mul_44" -> "3314 add_77" [label="(1, 24, 64, 64)", style=solid]; -"3284 pad_25" -> "3285 view_123" [label="(1, 8, 8, 768)", style=solid]; -"3285 view_123" -> "3286 permute_101" [label="(1, 1, 8, 1, 8, 768)", style=solid]; -"3286 permute_101" -> "3287 reshape_99" [label="(1, 1, 1, 8, 8, 768)", style=solid]; -"3287 reshape_99" -> "3291 linear_137" [label="(1, 64, 768)", style=solid]; -"3288 clone_22" -> "3289 slice_341" [label="(2304,)", style=solid]; -"3288 clone_22" -> "3291 linear_137" [label="(2304,)", style=solid]; -"3289 slice_341" -> "3290 zero__22" [label="(768,)", style=solid]; -"3291 linear_137" -> "3292 reshape_100" [label="(1, 64, 2304)", style=solid]; -"3292 reshape_100" -> "3293 permute_102" [label="(1, 64, 3, 24, 32)", style=solid]; -"3293 permute_102" -> "3294 select_66" [label="(3, 1, 24, 64, 32)", style=solid]; -"3293 permute_102" -> "3295 select_67" [label="(3, 1, 24, 64, 32)", style=solid]; -"3293 permute_102" -> "3296 select_68" [label="(3, 1, 24, 64, 32)", style=solid]; -"3294 select_66" -> "3297 linalg_vector_norm_44" [label="(1, 24, 64, 32)", style=solid]; -"3294 select_66" -> "3299 expand_as_44" [label="(1, 24, 64, 32)", style=solid]; -"3294 select_66" -> "3300 div_44" [label="(1, 24, 64, 32)", style=solid]; -"3295 select_67" -> "3303 linalg_vector_norm_45" [label="(1, 24, 64, 32)", style=solid]; -"3295 select_67" -> "3305 expand_as_45" [label="(1, 24, 64, 32)", style=solid]; -"3295 select_67" -> "3306 div_45" [label="(1, 24, 64, 32)", style=solid]; -"3296 select_68" -> "3317 matmul_45" [label="(1, 24, 64, 32)", style=solid]; -"3297 linalg_vector_norm_44" -> "3298 clamp_min_44" [label="(1, 24, 64, 1)", style=solid]; -"3298 clamp_min_44" -> "3299 expand_as_44" [label="(1, 24, 64, 1)", style=solid]; -"3299 expand_as_44" -> "3300 div_44" [label="(1, 24, 64, 32)", style=solid]; -"3300 div_44" -> "3301 quantize_per_tensor_default_137" [label="(1, 24, 64, 32)", style=solid]; -"3301 quantize_per_tensor_default_137" -> "3302 dequantize_per_tensor_default_137" [label=None, style=solid]; -"3302 dequantize_per_tensor_default_137" -> "3310 matmul_44" [label=None, style=solid]; -"3303 linalg_vector_norm_45" -> "3304 clamp_min_45" [label="(1, 24, 64, 1)", style=solid]; -"3304 clamp_min_45" -> "3305 expand_as_45" [label="(1, 24, 64, 1)", style=solid]; -"3305 expand_as_45" -> "3306 div_45" [label="(1, 24, 64, 32)", style=solid]; -"3306 div_45" -> "3307 quantize_per_tensor_default_138" [label="(1, 24, 64, 32)", style=solid]; -"3307 quantize_per_tensor_default_138" -> "3308 dequantize_per_tensor_default_138" [label=None, style=solid]; -"3308 dequantize_per_tensor_default_138" -> "3309 transpose_44" [label=None, style=solid]; -"3309 transpose_44" -> "3310 matmul_44" [label="(1, 24, 32, 64)", style=solid]; -"3310 matmul_44" -> "3313 mul_45" [label="(1, 24, 64, 64)", style=solid]; -"3311 clamp_22" -> "3312 exp_22" [label="(24, 1, 1)", style=solid]; -"3312 exp_22" -> "3313 mul_45" [label="(24, 1, 1)", style=solid]; -"3313 mul_45" -> "3314 add_77" [label="(1, 24, 64, 64)", style=solid]; -"3314 add_77" -> "3315 softmax_22" [label="(1, 24, 64, 64)", style=solid]; -"3315 softmax_22" -> "3316 dropout_88" [label="(1, 24, 64, 64)", style=solid]; -"3316 dropout_88" -> "3317 matmul_45" [label="(1, 24, 64, 64)", style=solid]; -"3317 matmul_45" -> "3318 quantize_per_tensor_default_139" [label="(1, 24, 64, 32)", style=solid]; -"3318 quantize_per_tensor_default_139" -> "3319 dequantize_per_tensor_default_139" [label=None, style=solid]; -"3319 dequantize_per_tensor_default_139" -> "3320 transpose_45" [label=None, style=solid]; -"3320 transpose_45" -> "3321 reshape_101" [label="(1, 64, 24, 32)", style=solid]; -"3321 reshape_101" -> "3322 linear_138" [label="(1, 64, 768)", style=solid]; -"3322 linear_138" -> "3323 dropout_89" [label="(1, 64, 768)", style=solid]; -"3323 dropout_89" -> "3324 view_124" [label="(1, 64, 768)", style=solid]; -"3324 view_124" -> "3325 permute_103" [label="(1, 1, 1, 8, 8, 768)", style=solid]; -"3325 permute_103" -> "3326 reshape_102" [label="(1, 1, 8, 1, 8, 768)", style=solid]; -"3326 reshape_102" -> "3327 slice_342" [label="(1, 8, 8, 768)", style=solid]; -"3327 slice_342" -> "3328 slice_343" [label="(1, 8, 8, 768)", style=solid]; -"3328 slice_343" -> "3329 slice_344" [label="(1, 7, 8, 768)", style=solid]; -"3329 slice_344" -> "3330 slice_345" [label="(1, 7, 7, 768)", style=solid]; -"3330 slice_345" -> "3331 contiguous_43" [label="(1, 7, 7, 768)", style=solid]; -"3331 contiguous_43" -> "3332 layer_norm_48" [label="(1, 7, 7, 768)", style=solid]; -"3332 layer_norm_48" -> "3333 add_78" [label="(1, 7, 7, 768)", style=solid]; -"3333 add_78" -> "3334 quantize_per_tensor_default_140" [label="(1, 7, 7, 768)", style=solid]; -"3333 add_78" -> "3344 add_79" [label="(1, 7, 7, 768)", style=solid]; -"3334 quantize_per_tensor_default_140" -> "3335 dequantize_per_tensor_default_140" [label=None, style=solid]; -"3335 dequantize_per_tensor_default_140" -> "3336 linear_139" [label=None, style=solid]; -"3336 linear_139" -> "3337 gelu_22" [label="(1, 7, 7, 3072)", style=solid]; -"3337 gelu_22" -> "3338 quantize_per_tensor_default_141" [label="(1, 7, 7, 3072)", style=solid]; -"3338 quantize_per_tensor_default_141" -> "3339 dequantize_per_tensor_default_141" [label=None, style=solid]; -"3339 dequantize_per_tensor_default_141" -> "3340 dropout_90" [label=None, style=solid]; -"3340 dropout_90" -> "3341 linear_140" [label="(1, 7, 7, 3072)", style=solid]; -"3341 linear_140" -> "3342 dropout_91" [label="(1, 7, 7, 768)", style=solid]; -"3342 dropout_91" -> "3343 layer_norm_49" [label="(1, 7, 7, 768)", style=solid]; -"3343 layer_norm_49" -> "3344 add_79" [label="(1, 7, 7, 768)", style=solid]; -"3344 add_79" -> "3345 quantize_per_tensor_default_142" [label="(1, 7, 7, 768)", style=solid]; -"3344 add_79" -> "3407 add_81" [label="(1, 7, 7, 768)", style=solid]; -"3345 quantize_per_tensor_default_142" -> "3346 dequantize_per_tensor_default_142" [label=None, style=solid]; -"3346 dequantize_per_tensor_default_142" -> "3358 pad_26" [label=None, style=solid]; -"3347 linear_141" -> "3348 relu__23" [label="(1, 15, 15, 512)", style=solid]; -"3348 relu__23" -> "3349 linear_142" [label="(1, 15, 15, 512)", style=solid]; -"3349 linear_142" -> "3350 view_125" [label="(1, 15, 15, 24)", style=solid]; -"3350 view_125" -> "3351 index_23" [label="(225, 24)", style=solid]; -"3351 index_23" -> "3352 view_126" [label="(4096, 24)", style=solid]; -"3352 view_126" -> "3353 permute_104" [label="(64, 64, 24)", style=solid]; -"3353 permute_104" -> "3354 contiguous_44" [label="(24, 64, 64)", style=solid]; -"3354 contiguous_44" -> "3355 unsqueeze_67" [label="(24, 64, 64)", style=solid]; -"3355 unsqueeze_67" -> "3356 sigmoid_23" [label="(1, 24, 64, 64)", style=solid]; -"3356 sigmoid_23" -> "3357 mul_46" [label="(1, 24, 64, 64)", style=solid]; -"3357 mul_46" -> "3388 add_80" [label="(1, 24, 64, 64)", style=solid]; -"3358 pad_26" -> "3359 view_127" [label="(1, 8, 8, 768)", style=solid]; -"3359 view_127" -> "3360 permute_105" [label="(1, 1, 8, 1, 8, 768)", style=solid]; -"3360 permute_105" -> "3361 reshape_103" [label="(1, 1, 1, 8, 8, 768)", style=solid]; -"3361 reshape_103" -> "3365 linear_143" [label="(1, 64, 768)", style=solid]; -"3362 clone_23" -> "3363 slice_346" [label="(2304,)", style=solid]; -"3362 clone_23" -> "3365 linear_143" [label="(2304,)", style=solid]; -"3363 slice_346" -> "3364 zero__23" [label="(768,)", style=solid]; -"3365 linear_143" -> "3366 reshape_104" [label="(1, 64, 2304)", style=solid]; -"3366 reshape_104" -> "3367 permute_106" [label="(1, 64, 3, 24, 32)", style=solid]; -"3367 permute_106" -> "3368 select_69" [label="(3, 1, 24, 64, 32)", style=solid]; -"3367 permute_106" -> "3369 select_70" [label="(3, 1, 24, 64, 32)", style=solid]; -"3367 permute_106" -> "3370 select_71" [label="(3, 1, 24, 64, 32)", style=solid]; -"3368 select_69" -> "3371 linalg_vector_norm_46" [label="(1, 24, 64, 32)", style=solid]; -"3368 select_69" -> "3373 expand_as_46" [label="(1, 24, 64, 32)", style=solid]; -"3368 select_69" -> "3374 div_46" [label="(1, 24, 64, 32)", style=solid]; -"3369 select_70" -> "3377 linalg_vector_norm_47" [label="(1, 24, 64, 32)", style=solid]; -"3369 select_70" -> "3379 expand_as_47" [label="(1, 24, 64, 32)", style=solid]; -"3369 select_70" -> "3380 div_47" [label="(1, 24, 64, 32)", style=solid]; -"3370 select_71" -> "3391 matmul_47" [label="(1, 24, 64, 32)", style=solid]; -"3371 linalg_vector_norm_46" -> "3372 clamp_min_46" [label="(1, 24, 64, 1)", style=solid]; -"3372 clamp_min_46" -> "3373 expand_as_46" [label="(1, 24, 64, 1)", style=solid]; -"3373 expand_as_46" -> "3374 div_46" [label="(1, 24, 64, 32)", style=solid]; -"3374 div_46" -> "3375 quantize_per_tensor_default_143" [label="(1, 24, 64, 32)", style=solid]; -"3375 quantize_per_tensor_default_143" -> "3376 dequantize_per_tensor_default_143" [label=None, style=solid]; -"3376 dequantize_per_tensor_default_143" -> "3384 matmul_46" [label=None, style=solid]; -"3377 linalg_vector_norm_47" -> "3378 clamp_min_47" [label="(1, 24, 64, 1)", style=solid]; -"3378 clamp_min_47" -> "3379 expand_as_47" [label="(1, 24, 64, 1)", style=solid]; -"3379 expand_as_47" -> "3380 div_47" [label="(1, 24, 64, 32)", style=solid]; -"3380 div_47" -> "3381 quantize_per_tensor_default_144" [label="(1, 24, 64, 32)", style=solid]; -"3381 quantize_per_tensor_default_144" -> "3382 dequantize_per_tensor_default_144" [label=None, style=solid]; -"3382 dequantize_per_tensor_default_144" -> "3383 transpose_46" [label=None, style=solid]; -"3383 transpose_46" -> "3384 matmul_46" [label="(1, 24, 32, 64)", style=solid]; -"3384 matmul_46" -> "3387 mul_47" [label="(1, 24, 64, 64)", style=solid]; -"3385 clamp_23" -> "3386 exp_23" [label="(24, 1, 1)", style=solid]; -"3386 exp_23" -> "3387 mul_47" [label="(24, 1, 1)", style=solid]; -"3387 mul_47" -> "3388 add_80" [label="(1, 24, 64, 64)", style=solid]; -"3388 add_80" -> "3389 softmax_23" [label="(1, 24, 64, 64)", style=solid]; -"3389 softmax_23" -> "3390 dropout_92" [label="(1, 24, 64, 64)", style=solid]; -"3390 dropout_92" -> "3391 matmul_47" [label="(1, 24, 64, 64)", style=solid]; -"3391 matmul_47" -> "3392 quantize_per_tensor_default_145" [label="(1, 24, 64, 32)", style=solid]; -"3392 quantize_per_tensor_default_145" -> "3393 dequantize_per_tensor_default_145" [label=None, style=solid]; -"3393 dequantize_per_tensor_default_145" -> "3394 transpose_47" [label=None, style=solid]; -"3394 transpose_47" -> "3395 reshape_105" [label="(1, 64, 24, 32)", style=solid]; -"3395 reshape_105" -> "3396 linear_144" [label="(1, 64, 768)", style=solid]; -"3396 linear_144" -> "3397 dropout_93" [label="(1, 64, 768)", style=solid]; -"3397 dropout_93" -> "3398 view_128" [label="(1, 64, 768)", style=solid]; -"3398 view_128" -> "3399 permute_107" [label="(1, 1, 1, 8, 8, 768)", style=solid]; -"3399 permute_107" -> "3400 reshape_106" [label="(1, 1, 8, 1, 8, 768)", style=solid]; -"3400 reshape_106" -> "3401 slice_347" [label="(1, 8, 8, 768)", style=solid]; -"3401 slice_347" -> "3402 slice_348" [label="(1, 8, 8, 768)", style=solid]; -"3402 slice_348" -> "3403 slice_349" [label="(1, 7, 8, 768)", style=solid]; -"3403 slice_349" -> "3404 slice_350" [label="(1, 7, 7, 768)", style=solid]; -"3404 slice_350" -> "3405 contiguous_45" [label="(1, 7, 7, 768)", style=solid]; -"3405 contiguous_45" -> "3406 layer_norm_50" [label="(1, 7, 7, 768)", style=solid]; -"3406 layer_norm_50" -> "3407 add_81" [label="(1, 7, 7, 768)", style=solid]; -"3407 add_81" -> "3408 quantize_per_tensor_default_146" [label="(1, 7, 7, 768)", style=solid]; -"3407 add_81" -> "3418 add_82" [label="(1, 7, 7, 768)", style=solid]; -"3408 quantize_per_tensor_default_146" -> "3409 dequantize_per_tensor_default_146" [label=None, style=solid]; -"3409 dequantize_per_tensor_default_146" -> "3410 linear_145" [label=None, style=solid]; -"3410 linear_145" -> "3411 gelu_23" [label="(1, 7, 7, 3072)", style=solid]; -"3411 gelu_23" -> "3412 quantize_per_tensor_default_147" [label="(1, 7, 7, 3072)", style=solid]; -"3412 quantize_per_tensor_default_147" -> "3413 dequantize_per_tensor_default_147" [label=None, style=solid]; -"3413 dequantize_per_tensor_default_147" -> "3414 dropout_94" [label=None, style=solid]; -"3414 dropout_94" -> "3415 linear_146" [label="(1, 7, 7, 3072)", style=solid]; -"3415 linear_146" -> "3416 dropout_95" [label="(1, 7, 7, 768)", style=solid]; -"3416 dropout_95" -> "3417 layer_norm_51" [label="(1, 7, 7, 768)", style=solid]; -"3417 layer_norm_51" -> "3418 add_82" [label="(1, 7, 7, 768)", style=solid]; -"3418 add_82" -> "3419 layer_norm_52" [label="(1, 7, 7, 768)", style=solid]; -"3419 layer_norm_52" -> "3420 permute_108" [label="(1, 7, 7, 768)", style=solid]; -"3420 permute_108" -> "3421 adaptive_avg_pool2d" [label="(1, 768, 7, 7)", style=solid]; -"3421 adaptive_avg_pool2d" -> "3422 quantize_per_tensor_default_148" [label="(1, 768, 1, 1)", style=solid]; -"3422 quantize_per_tensor_default_148" -> "3423 dequantize_per_tensor_default_148" [label=None, style=solid]; -"3423 dequantize_per_tensor_default_148" -> "3424 flatten" [label=None, style=solid]; -"3424 flatten" -> "3425 linear_147" [label="(1, 768)", style=solid]; -"3425 linear_147" -> "3426 output_1" [label="(1, 1000)", style=solid]; +"6 _frozen_param6_1" -> "7 dequantize_per_channel_default_1" [label="(96, 96)", style=solid]; +"7 dequantize_per_channel_default_1" -> "628 linear_3" [label=None, style=solid]; +"8 features_1_0_attn_proj_bias" -> "628 linear_3" [label="(96,)", style=solid]; +"9 _scale_2" -> "12 dequantize_per_channel_default_2" [label="(288,)", style=solid]; +"10 _zero_point_2" -> "12 dequantize_per_channel_default_2" [label="(288,)", style=solid]; +"11 _frozen_param8_1" -> "12 dequantize_per_channel_default_2" [label="(288, 96)", style=solid]; +"12 dequantize_per_channel_default_2" -> "657 linear_8" [label=None, style=solid]; +"13 _scale_3" -> "16 dequantize_per_channel_default_3" [label="(96,)", style=solid]; +"14 _zero_point_3" -> "16 dequantize_per_channel_default_3" [label="(96,)", style=solid]; +"15 _frozen_param9_1" -> "16 dequantize_per_channel_default_3" [label="(96, 96)", style=solid]; +"16 dequantize_per_channel_default_3" -> "703 linear_9" [label=None, style=solid]; +"17 features_1_1_attn_proj_bias" -> "703 linear_9" [label="(96,)", style=solid]; +"18 _scale_4" -> "21 dequantize_per_channel_default_4" [label="(576,)", style=solid]; +"19 _zero_point_4" -> "21 dequantize_per_channel_default_4" [label="(576,)", style=solid]; +"20 _frozen_param10_1" -> "21 dequantize_per_channel_default_4" [label="(576, 192)", style=solid]; +"21 dequantize_per_channel_default_4" -> "750 linear_15" [label=None, style=solid]; +"22 _scale_5" -> "25 dequantize_per_channel_default_5" [label="(192,)", style=solid]; +"23 _zero_point_5" -> "25 dequantize_per_channel_default_5" [label="(192,)", style=solid]; +"24 _frozen_param11_1" -> "25 dequantize_per_channel_default_5" [label="(192, 192)", style=solid]; +"25 dequantize_per_channel_default_5" -> "780 linear_16" [label=None, style=solid]; +"26 features_3_0_attn_proj_bias" -> "780 linear_16" [label="(192,)", style=solid]; +"27 _scale_6" -> "30 dequantize_per_channel_default_6" [label="(576,)", style=solid]; +"28 _zero_point_6" -> "30 dequantize_per_channel_default_6" [label="(576,)", style=solid]; +"29 _frozen_param12_1" -> "30 dequantize_per_channel_default_6" [label="(576, 192)", style=solid]; +"30 dequantize_per_channel_default_6" -> "812 linear_21" [label=None, style=solid]; +"31 _scale_7" -> "34 dequantize_per_channel_default_7" [label="(192,)", style=solid]; +"32 _zero_point_7" -> "34 dequantize_per_channel_default_7" [label="(192,)", style=solid]; +"33 _frozen_param13_1" -> "34 dequantize_per_channel_default_7" [label="(192, 192)", style=solid]; +"34 dequantize_per_channel_default_7" -> "858 linear_22" [label=None, style=solid]; +"35 features_3_1_attn_proj_bias" -> "858 linear_22" [label="(192,)", style=solid]; +"36 _scale_8" -> "39 dequantize_per_channel_default_8" [label="(1152,)", style=solid]; +"37 _zero_point_8" -> "39 dequantize_per_channel_default_8" [label="(1152,)", style=solid]; +"38 _frozen_param14_1" -> "39 dequantize_per_channel_default_8" [label="(1152, 384)", style=solid]; +"39 dequantize_per_channel_default_8" -> "908 linear_28" [label=None, style=solid]; +"40 _scale_9" -> "43 dequantize_per_channel_default_9" [label="(384,)", style=solid]; +"41 _zero_point_9" -> "43 dequantize_per_channel_default_9" [label="(384,)", style=solid]; +"42 _frozen_param15_1" -> "43 dequantize_per_channel_default_9" [label="(384, 384)", style=solid]; +"43 dequantize_per_channel_default_9" -> "938 linear_29" [label=None, style=solid]; +"44 features_5_0_attn_proj_bias" -> "938 linear_29" [label="(384,)", style=solid]; +"45 _scale_10" -> "48 dequantize_per_channel_default_10" [label="(1152,)", style=solid]; +"46 _zero_point_10" -> "48 dequantize_per_channel_default_10" [label="(1152,)", style=solid]; +"47 _frozen_param16_1" -> "48 dequantize_per_channel_default_10" [label="(1152, 384)", style=solid]; +"48 dequantize_per_channel_default_10" -> "970 linear_34" [label=None, style=solid]; +"49 _scale_11" -> "52 dequantize_per_channel_default_11" [label="(384,)", style=solid]; +"50 _zero_point_11" -> "52 dequantize_per_channel_default_11" [label="(384,)", style=solid]; +"51 _frozen_param19_1" -> "52 dequantize_per_channel_default_11" [label="(384, 384)", style=solid]; +"52 dequantize_per_channel_default_11" -> "1016 linear_35" [label=None, style=solid]; +"53 features_5_1_attn_proj_bias" -> "1016 linear_35" [label="(384,)", style=solid]; +"54 _scale_12" -> "57 dequantize_per_channel_default_12" [label="(1152,)", style=solid]; +"55 _zero_point_12" -> "57 dequantize_per_channel_default_12" [label="(1152,)", style=solid]; +"56 _frozen_param23_1" -> "57 dequantize_per_channel_default_12" [label="(1152, 384)", style=solid]; +"57 dequantize_per_channel_default_12" -> "1048 linear_40" [label=None, style=solid]; +"58 _scale_13" -> "61 dequantize_per_channel_default_13" [label="(384,)", style=solid]; +"59 _zero_point_13" -> "61 dequantize_per_channel_default_13" [label="(384,)", style=solid]; +"60 _frozen_param25_1" -> "61 dequantize_per_channel_default_13" [label="(384, 384)", style=solid]; +"61 dequantize_per_channel_default_13" -> "1078 linear_41" [label=None, style=solid]; +"62 features_5_2_attn_proj_bias" -> "1078 linear_41" [label="(384,)", style=solid]; +"63 _scale_14" -> "66 dequantize_per_channel_default_14" [label="(1152,)", style=solid]; +"64 _zero_point_14" -> "66 dequantize_per_channel_default_14" [label="(1152,)", style=solid]; +"65 _frozen_param26_1" -> "66 dequantize_per_channel_default_14" [label="(1152, 384)", style=solid]; +"66 dequantize_per_channel_default_14" -> "1110 linear_46" [label=None, style=solid]; +"67 _scale_15" -> "70 dequantize_per_channel_default_15" [label="(384,)", style=solid]; +"68 _zero_point_15" -> "70 dequantize_per_channel_default_15" [label="(384,)", style=solid]; +"69 _frozen_param27_1" -> "70 dequantize_per_channel_default_15" [label="(384, 384)", style=solid]; +"70 dequantize_per_channel_default_15" -> "1156 linear_47" [label=None, style=solid]; +"71 features_5_3_attn_proj_bias" -> "1156 linear_47" [label="(384,)", style=solid]; +"72 _scale_16" -> "75 dequantize_per_channel_default_16" [label="(1152,)", style=solid]; +"73 _zero_point_16" -> "75 dequantize_per_channel_default_16" [label="(1152,)", style=solid]; +"74 _frozen_param28_1" -> "75 dequantize_per_channel_default_16" [label="(1152, 384)", style=solid]; +"75 dequantize_per_channel_default_16" -> "1188 linear_52" [label=None, style=solid]; +"76 _scale_17" -> "79 dequantize_per_channel_default_17" [label="(384,)", style=solid]; +"77 _zero_point_17" -> "79 dequantize_per_channel_default_17" [label="(384,)", style=solid]; +"78 _frozen_param29_1" -> "79 dequantize_per_channel_default_17" [label="(384, 384)", style=solid]; +"79 dequantize_per_channel_default_17" -> "1218 linear_53" [label=None, style=solid]; +"80 features_5_4_attn_proj_bias" -> "1218 linear_53" [label="(384,)", style=solid]; +"81 _scale_18" -> "84 dequantize_per_channel_default_18" [label="(1152,)", style=solid]; +"82 _zero_point_18" -> "84 dequantize_per_channel_default_18" [label="(1152,)", style=solid]; +"83 _frozen_param30_1" -> "84 dequantize_per_channel_default_18" [label="(1152, 384)", style=solid]; +"84 dequantize_per_channel_default_18" -> "1250 linear_58" [label=None, style=solid]; +"85 _scale_19" -> "88 dequantize_per_channel_default_19" [label="(384,)", style=solid]; +"86 _zero_point_19" -> "88 dequantize_per_channel_default_19" [label="(384,)", style=solid]; +"87 _frozen_param31_1" -> "88 dequantize_per_channel_default_19" [label="(384, 384)", style=solid]; +"88 dequantize_per_channel_default_19" -> "1296 linear_59" [label=None, style=solid]; +"89 features_5_5_attn_proj_bias" -> "1296 linear_59" [label="(384,)", style=solid]; +"90 _scale_20" -> "93 dequantize_per_channel_default_20" [label="(1152,)", style=solid]; +"91 _zero_point_20" -> "93 dequantize_per_channel_default_20" [label="(1152,)", style=solid]; +"92 _frozen_param32_1" -> "93 dequantize_per_channel_default_20" [label="(1152, 384)", style=solid]; +"93 dequantize_per_channel_default_20" -> "1328 linear_64" [label=None, style=solid]; +"94 _scale_21" -> "97 dequantize_per_channel_default_21" [label="(384,)", style=solid]; +"95 _zero_point_21" -> "97 dequantize_per_channel_default_21" [label="(384,)", style=solid]; +"96 _frozen_param33_1" -> "97 dequantize_per_channel_default_21" [label="(384, 384)", style=solid]; +"97 dequantize_per_channel_default_21" -> "1358 linear_65" [label=None, style=solid]; +"98 features_5_6_attn_proj_bias" -> "1358 linear_65" [label="(384,)", style=solid]; +"99 _scale_22" -> "102 dequantize_per_channel_default_22" [label="(1152,)", style=solid]; +"100 _zero_point_22" -> "102 dequantize_per_channel_default_22" [label="(1152,)", style=solid]; +"101 _frozen_param36_1" -> "102 dequantize_per_channel_default_22" [label="(1152, 384)", style=solid]; +"102 dequantize_per_channel_default_22" -> "1390 linear_70" [label=None, style=solid]; +"103 _scale_23" -> "106 dequantize_per_channel_default_23" [label="(384,)", style=solid]; +"104 _zero_point_23" -> "106 dequantize_per_channel_default_23" [label="(384,)", style=solid]; +"105 _frozen_param40_1" -> "106 dequantize_per_channel_default_23" [label="(384, 384)", style=solid]; +"106 dequantize_per_channel_default_23" -> "1436 linear_71" [label=None, style=solid]; +"107 features_5_7_attn_proj_bias" -> "1436 linear_71" [label="(384,)", style=solid]; +"108 _scale_24" -> "111 dequantize_per_channel_default_24" [label="(1152,)", style=solid]; +"109 _zero_point_24" -> "111 dequantize_per_channel_default_24" [label="(1152,)", style=solid]; +"110 _frozen_param42_1" -> "111 dequantize_per_channel_default_24" [label="(1152, 384)", style=solid]; +"111 dequantize_per_channel_default_24" -> "1468 linear_76" [label=None, style=solid]; +"112 _scale_25" -> "115 dequantize_per_channel_default_25" [label="(384,)", style=solid]; +"113 _zero_point_25" -> "115 dequantize_per_channel_default_25" [label="(384,)", style=solid]; +"114 _frozen_param43_1" -> "115 dequantize_per_channel_default_25" [label="(384, 384)", style=solid]; +"115 dequantize_per_channel_default_25" -> "1498 linear_77" [label=None, style=solid]; +"116 features_5_8_attn_proj_bias" -> "1498 linear_77" [label="(384,)", style=solid]; +"117 _scale_26" -> "120 dequantize_per_channel_default_26" [label="(1152,)", style=solid]; +"118 _zero_point_26" -> "120 dequantize_per_channel_default_26" [label="(1152,)", style=solid]; +"119 _frozen_param44_1" -> "120 dequantize_per_channel_default_26" [label="(1152, 384)", style=solid]; +"120 dequantize_per_channel_default_26" -> "1530 linear_82" [label=None, style=solid]; +"121 _scale_27" -> "124 dequantize_per_channel_default_27" [label="(384,)", style=solid]; +"122 _zero_point_27" -> "124 dequantize_per_channel_default_27" [label="(384,)", style=solid]; +"123 _frozen_param45_1" -> "124 dequantize_per_channel_default_27" [label="(384, 384)", style=solid]; +"124 dequantize_per_channel_default_27" -> "1576 linear_83" [label=None, style=solid]; +"125 features_5_9_attn_proj_bias" -> "1576 linear_83" [label="(384,)", style=solid]; +"126 _scale_28" -> "129 dequantize_per_channel_default_28" [label="(1152,)", style=solid]; +"127 _zero_point_28" -> "129 dequantize_per_channel_default_28" [label="(1152,)", style=solid]; +"128 _frozen_param46_1" -> "129 dequantize_per_channel_default_28" [label="(1152, 384)", style=solid]; +"129 dequantize_per_channel_default_28" -> "1608 linear_88" [label=None, style=solid]; +"130 _scale_29" -> "133 dequantize_per_channel_default_29" [label="(384,)", style=solid]; +"131 _zero_point_29" -> "133 dequantize_per_channel_default_29" [label="(384,)", style=solid]; +"132 _frozen_param47_1" -> "133 dequantize_per_channel_default_29" [label="(384, 384)", style=solid]; +"133 dequantize_per_channel_default_29" -> "1638 linear_89" [label=None, style=solid]; +"134 features_5_10_attn_proj_bias" -> "1638 linear_89" [label="(384,)", style=solid]; +"135 _scale_30" -> "138 dequantize_per_channel_default_30" [label="(1152,)", style=solid]; +"136 _zero_point_30" -> "138 dequantize_per_channel_default_30" [label="(1152,)", style=solid]; +"137 _frozen_param48_1" -> "138 dequantize_per_channel_default_30" [label="(1152, 384)", style=solid]; +"138 dequantize_per_channel_default_30" -> "1670 linear_94" [label=None, style=solid]; +"139 _scale_31" -> "142 dequantize_per_channel_default_31" [label="(384,)", style=solid]; +"140 _zero_point_31" -> "142 dequantize_per_channel_default_31" [label="(384,)", style=solid]; +"141 _frozen_param49_1" -> "142 dequantize_per_channel_default_31" [label="(384, 384)", style=solid]; +"142 dequantize_per_channel_default_31" -> "1716 linear_95" [label=None, style=solid]; +"143 features_5_11_attn_proj_bias" -> "1716 linear_95" [label="(384,)", style=solid]; +"144 _scale_32" -> "147 dequantize_per_channel_default_32" [label="(1152,)", style=solid]; +"145 _zero_point_32" -> "147 dequantize_per_channel_default_32" [label="(1152,)", style=solid]; +"146 _frozen_param50_1" -> "147 dequantize_per_channel_default_32" [label="(1152, 384)", style=solid]; +"147 dequantize_per_channel_default_32" -> "1748 linear_100" [label=None, style=solid]; +"148 _scale_33" -> "151 dequantize_per_channel_default_33" [label="(384,)", style=solid]; +"149 _zero_point_33" -> "151 dequantize_per_channel_default_33" [label="(384,)", style=solid]; +"150 _frozen_param53_1" -> "151 dequantize_per_channel_default_33" [label="(384, 384)", style=solid]; +"151 dequantize_per_channel_default_33" -> "1778 linear_101" [label=None, style=solid]; +"152 features_5_12_attn_proj_bias" -> "1778 linear_101" [label="(384,)", style=solid]; +"153 _scale_34" -> "156 dequantize_per_channel_default_34" [label="(1152,)", style=solid]; +"154 _zero_point_34" -> "156 dequantize_per_channel_default_34" [label="(1152,)", style=solid]; +"155 _frozen_param57_1" -> "156 dequantize_per_channel_default_34" [label="(1152, 384)", style=solid]; +"156 dequantize_per_channel_default_34" -> "1810 linear_106" [label=None, style=solid]; +"157 _scale_35" -> "160 dequantize_per_channel_default_35" [label="(384,)", style=solid]; +"158 _zero_point_35" -> "160 dequantize_per_channel_default_35" [label="(384,)", style=solid]; +"159 _frozen_param59_1" -> "160 dequantize_per_channel_default_35" [label="(384, 384)", style=solid]; +"160 dequantize_per_channel_default_35" -> "1856 linear_107" [label=None, style=solid]; +"161 features_5_13_attn_proj_bias" -> "1856 linear_107" [label="(384,)", style=solid]; +"162 _scale_36" -> "165 dequantize_per_channel_default_36" [label="(1152,)", style=solid]; +"163 _zero_point_36" -> "165 dequantize_per_channel_default_36" [label="(1152,)", style=solid]; +"164 _frozen_param60_1" -> "165 dequantize_per_channel_default_36" [label="(1152, 384)", style=solid]; +"165 dequantize_per_channel_default_36" -> "1888 linear_112" [label=None, style=solid]; +"166 _scale_37" -> "169 dequantize_per_channel_default_37" [label="(384,)", style=solid]; +"167 _zero_point_37" -> "169 dequantize_per_channel_default_37" [label="(384,)", style=solid]; +"168 _frozen_param61_1" -> "169 dequantize_per_channel_default_37" [label="(384, 384)", style=solid]; +"169 dequantize_per_channel_default_37" -> "1918 linear_113" [label=None, style=solid]; +"170 features_5_14_attn_proj_bias" -> "1918 linear_113" [label="(384,)", style=solid]; +"171 _scale_38" -> "174 dequantize_per_channel_default_38" [label="(1152,)", style=solid]; +"172 _zero_point_38" -> "174 dequantize_per_channel_default_38" [label="(1152,)", style=solid]; +"173 _frozen_param62_1" -> "174 dequantize_per_channel_default_38" [label="(1152, 384)", style=solid]; +"174 dequantize_per_channel_default_38" -> "1950 linear_118" [label=None, style=solid]; +"175 _scale_39" -> "178 dequantize_per_channel_default_39" [label="(384,)", style=solid]; +"176 _zero_point_39" -> "178 dequantize_per_channel_default_39" [label="(384,)", style=solid]; +"177 _frozen_param63_1" -> "178 dequantize_per_channel_default_39" [label="(384, 384)", style=solid]; +"178 dequantize_per_channel_default_39" -> "1996 linear_119" [label=None, style=solid]; +"179 features_5_15_attn_proj_bias" -> "1996 linear_119" [label="(384,)", style=solid]; +"180 _scale_40" -> "183 dequantize_per_channel_default_40" [label="(1152,)", style=solid]; +"181 _zero_point_40" -> "183 dequantize_per_channel_default_40" [label="(1152,)", style=solid]; +"182 _frozen_param64_1" -> "183 dequantize_per_channel_default_40" [label="(1152, 384)", style=solid]; +"183 dequantize_per_channel_default_40" -> "2028 linear_124" [label=None, style=solid]; +"184 _scale_41" -> "187 dequantize_per_channel_default_41" [label="(384,)", style=solid]; +"185 _zero_point_41" -> "187 dequantize_per_channel_default_41" [label="(384,)", style=solid]; +"186 _frozen_param65_1" -> "187 dequantize_per_channel_default_41" [label="(384, 384)", style=solid]; +"187 dequantize_per_channel_default_41" -> "2058 linear_125" [label=None, style=solid]; +"188 features_5_16_attn_proj_bias" -> "2058 linear_125" [label="(384,)", style=solid]; +"189 _scale_42" -> "192 dequantize_per_channel_default_42" [label="(1152,)", style=solid]; +"190 _zero_point_42" -> "192 dequantize_per_channel_default_42" [label="(1152,)", style=solid]; +"191 _frozen_param66_1" -> "192 dequantize_per_channel_default_42" [label="(1152, 384)", style=solid]; +"192 dequantize_per_channel_default_42" -> "2090 linear_130" [label=None, style=solid]; +"193 _scale_43" -> "196 dequantize_per_channel_default_43" [label="(384,)", style=solid]; +"194 _zero_point_43" -> "196 dequantize_per_channel_default_43" [label="(384,)", style=solid]; +"195 _frozen_param67_1" -> "196 dequantize_per_channel_default_43" [label="(384, 384)", style=solid]; +"196 dequantize_per_channel_default_43" -> "2136 linear_131" [label=None, style=solid]; +"197 features_5_17_attn_proj_bias" -> "2136 linear_131" [label="(384,)", style=solid]; +"198 _scale_44" -> "201 dequantize_per_channel_default_44" [label="(2304,)", style=solid]; +"199 _zero_point_44" -> "201 dequantize_per_channel_default_44" [label="(2304,)", style=solid]; +"200 _frozen_param70_1" -> "201 dequantize_per_channel_default_44" [label="(2304, 768)", style=solid]; +"201 dequantize_per_channel_default_44" -> "2186 linear_137" [label=None, style=solid]; +"202 _scale_45" -> "205 dequantize_per_channel_default_45" [label="(768,)", style=solid]; +"203 _zero_point_45" -> "205 dequantize_per_channel_default_45" [label="(768,)", style=solid]; +"204 _frozen_param74_1" -> "205 dequantize_per_channel_default_45" [label="(768, 768)", style=solid]; +"205 dequantize_per_channel_default_45" -> "2216 linear_138" [label=None, style=solid]; +"206 features_7_0_attn_proj_bias" -> "2216 linear_138" [label="(768,)", style=solid]; +"207 _scale_46" -> "210 dequantize_per_channel_default_46" [label="(2304,)", style=solid]; +"208 _zero_point_46" -> "210 dequantize_per_channel_default_46" [label="(2304,)", style=solid]; +"209 _frozen_param76_1" -> "210 dequantize_per_channel_default_46" [label="(2304, 768)", style=solid]; +"210 dequantize_per_channel_default_46" -> "2247 linear_143" [label=None, style=solid]; +"211 _scale_47" -> "214 dequantize_per_channel_default_47" [label="(768,)", style=solid]; +"212 _zero_point_47" -> "214 dequantize_per_channel_default_47" [label="(768,)", style=solid]; +"213 _frozen_param77_1" -> "214 dequantize_per_channel_default_47" [label="(768, 768)", style=solid]; +"214 dequantize_per_channel_default_47" -> "2277 linear_144" [label=None, style=solid]; +"215 features_7_1_attn_proj_bias" -> "2277 linear_144" [label="(768,)", style=solid]; +"216 _scale_48" -> "219 dequantize_per_channel_default_48" [label="(96,)", style=solid]; +"217 _zero_point_48" -> "219 dequantize_per_channel_default_48" [label="(96,)", style=solid]; +"218 _frozen_param78_1" -> "219 dequantize_per_channel_default_48" [label="(96, 3, 4, 4)", style=solid]; +"219 dequantize_per_channel_default_48" -> "587 conv2d" [label=None, style=solid]; +"220 features_0_0_bias" -> "587 conv2d" [label="(96,)", style=solid]; +"221 features_0_2_weight" -> "589 layer_norm" [label="(96,)", style=solid]; +"222 features_0_2_bias" -> "589 layer_norm" [label="(96,)", style=solid]; +"223 features_1_0_norm1_weight" -> "635 layer_norm_1" [label="(96,)", style=solid]; +"224 features_1_0_norm1_bias" -> "635 layer_norm_1" [label="(96,)", style=solid]; +"225 _scale_49" -> "228 dequantize_per_channel_default_49" [label="(384,)", style=solid]; +"226 _zero_point_49" -> "228 dequantize_per_channel_default_49" [label="(384,)", style=solid]; +"227 _frozen_param79_1" -> "228 dequantize_per_channel_default_49" [label="(384, 96)", style=solid]; +"228 dequantize_per_channel_default_49" -> "639 linear_4" [label=None, style=solid]; +"229 features_1_0_mlp_0_bias" -> "639 linear_4" [label="(384,)", style=solid]; +"230 _scale_50" -> "233 dequantize_per_channel_default_50" [label="(96,)", style=solid]; +"231 _zero_point_50" -> "233 dequantize_per_channel_default_50" [label="(96,)", style=solid]; +"232 _frozen_param80_1" -> "233 dequantize_per_channel_default_50" [label="(96, 384)", style=solid]; +"233 dequantize_per_channel_default_50" -> "644 linear_5" [label=None, style=solid]; +"234 features_1_0_mlp_3_bias" -> "644 linear_5" [label="(96,)", style=solid]; +"235 features_1_0_norm2_weight" -> "646 layer_norm_2" [label="(96,)", style=solid]; +"236 features_1_0_norm2_bias" -> "646 layer_norm_2" [label="(96,)", style=solid]; +"237 features_1_1_norm1_weight" -> "711 layer_norm_3" [label="(96,)", style=solid]; +"238 features_1_1_norm1_bias" -> "711 layer_norm_3" [label="(96,)", style=solid]; +"239 _scale_51" -> "242 dequantize_per_channel_default_51" [label="(384,)", style=solid]; +"240 _zero_point_51" -> "242 dequantize_per_channel_default_51" [label="(384,)", style=solid]; +"241 _frozen_param81_1" -> "242 dequantize_per_channel_default_51" [label="(384, 96)", style=solid]; +"242 dequantize_per_channel_default_51" -> "715 linear_10" [label=None, style=solid]; +"243 features_1_1_mlp_0_bias" -> "715 linear_10" [label="(384,)", style=solid]; +"244 _scale_52" -> "247 dequantize_per_channel_default_52" [label="(96,)", style=solid]; +"245 _zero_point_52" -> "247 dequantize_per_channel_default_52" [label="(96,)", style=solid]; +"246 _frozen_param82_1" -> "247 dequantize_per_channel_default_52" [label="(96, 384)", style=solid]; +"247 dequantize_per_channel_default_52" -> "720 linear_11" [label=None, style=solid]; +"248 features_1_1_mlp_3_bias" -> "720 linear_11" [label="(96,)", style=solid]; +"249 features_1_1_norm2_weight" -> "722 layer_norm_4" [label="(96,)", style=solid]; +"250 features_1_1_norm2_bias" -> "722 layer_norm_4" [label="(96,)", style=solid]; +"251 _scale_53" -> "254 dequantize_per_channel_default_53" [label="(192,)", style=solid]; +"252 _zero_point_53" -> "254 dequantize_per_channel_default_53" [label="(192,)", style=solid]; +"253 _frozen_param83_1" -> "254 dequantize_per_channel_default_53" [label="(192, 384)", style=solid]; +"254 dequantize_per_channel_default_53" -> "740 linear_12" [label=None, style=solid]; +"255 features_2_norm_weight" -> "741 layer_norm_5" [label="(192,)", style=solid]; +"256 features_2_norm_bias" -> "741 layer_norm_5" [label="(192,)", style=solid]; +"257 features_3_0_norm1_weight" -> "790 layer_norm_6" [label="(192,)", style=solid]; +"258 features_3_0_norm1_bias" -> "790 layer_norm_6" [label="(192,)", style=solid]; +"259 _scale_54" -> "262 dequantize_per_channel_default_54" [label="(768,)", style=solid]; +"260 _zero_point_54" -> "262 dequantize_per_channel_default_54" [label="(768,)", style=solid]; +"261 _frozen_param84_1" -> "262 dequantize_per_channel_default_54" [label="(768, 192)", style=solid]; +"262 dequantize_per_channel_default_54" -> "794 linear_17" [label=None, style=solid]; +"263 features_3_0_mlp_0_bias" -> "794 linear_17" [label="(768,)", style=solid]; +"264 _scale_55" -> "267 dequantize_per_channel_default_55" [label="(192,)", style=solid]; +"265 _zero_point_55" -> "267 dequantize_per_channel_default_55" [label="(192,)", style=solid]; +"266 _frozen_param87_1" -> "267 dequantize_per_channel_default_55" [label="(192, 768)", style=solid]; +"267 dequantize_per_channel_default_55" -> "799 linear_18" [label=None, style=solid]; +"268 features_3_0_mlp_3_bias" -> "799 linear_18" [label="(192,)", style=solid]; +"269 features_3_0_norm2_weight" -> "801 layer_norm_7" [label="(192,)", style=solid]; +"270 features_3_0_norm2_bias" -> "801 layer_norm_7" [label="(192,)", style=solid]; +"271 features_3_1_norm1_weight" -> "869 layer_norm_8" [label="(192,)", style=solid]; +"272 features_3_1_norm1_bias" -> "869 layer_norm_8" [label="(192,)", style=solid]; +"273 _scale_56" -> "276 dequantize_per_channel_default_56" [label="(768,)", style=solid]; +"274 _zero_point_56" -> "276 dequantize_per_channel_default_56" [label="(768,)", style=solid]; +"275 _frozen_param91_1" -> "276 dequantize_per_channel_default_56" [label="(768, 192)", style=solid]; +"276 dequantize_per_channel_default_56" -> "873 linear_23" [label=None, style=solid]; +"277 features_3_1_mlp_0_bias" -> "873 linear_23" [label="(768,)", style=solid]; +"278 _scale_57" -> "281 dequantize_per_channel_default_57" [label="(192,)", style=solid]; +"279 _zero_point_57" -> "281 dequantize_per_channel_default_57" [label="(192,)", style=solid]; +"280 _frozen_param93_1" -> "281 dequantize_per_channel_default_57" [label="(192, 768)", style=solid]; +"281 dequantize_per_channel_default_57" -> "878 linear_24" [label=None, style=solid]; +"282 features_3_1_mlp_3_bias" -> "878 linear_24" [label="(192,)", style=solid]; +"283 features_3_1_norm2_weight" -> "880 layer_norm_9" [label="(192,)", style=solid]; +"284 features_3_1_norm2_bias" -> "880 layer_norm_9" [label="(192,)", style=solid]; +"285 _scale_58" -> "288 dequantize_per_channel_default_58" [label="(384,)", style=solid]; +"286 _zero_point_58" -> "288 dequantize_per_channel_default_58" [label="(384,)", style=solid]; +"287 _frozen_param94_1" -> "288 dequantize_per_channel_default_58" [label="(384, 768)", style=solid]; +"288 dequantize_per_channel_default_58" -> "898 linear_25" [label=None, style=solid]; +"289 features_4_norm_weight" -> "899 layer_norm_10" [label="(384,)", style=solid]; +"290 features_4_norm_bias" -> "899 layer_norm_10" [label="(384,)", style=solid]; +"291 features_5_0_norm1_weight" -> "948 layer_norm_11" [label="(384,)", style=solid]; +"292 features_5_0_norm1_bias" -> "948 layer_norm_11" [label="(384,)", style=solid]; +"293 _scale_59" -> "296 dequantize_per_channel_default_59" [label="(1536,)", style=solid]; +"294 _zero_point_59" -> "296 dequantize_per_channel_default_59" [label="(1536,)", style=solid]; +"295 _frozen_param95_1" -> "296 dequantize_per_channel_default_59" [label="(1536, 384)", style=solid]; +"296 dequantize_per_channel_default_59" -> "952 linear_30" [label=None, style=solid]; +"297 features_5_0_mlp_0_bias" -> "952 linear_30" [label="(1536,)", style=solid]; +"298 _scale_60" -> "301 dequantize_per_channel_default_60" [label="(384,)", style=solid]; +"299 _zero_point_60" -> "301 dequantize_per_channel_default_60" [label="(384,)", style=solid]; +"300 _frozen_param96_1" -> "301 dequantize_per_channel_default_60" [label="(384, 1536)", style=solid]; +"301 dequantize_per_channel_default_60" -> "957 linear_31" [label=None, style=solid]; +"302 features_5_0_mlp_3_bias" -> "957 linear_31" [label="(384,)", style=solid]; +"303 features_5_0_norm2_weight" -> "959 layer_norm_12" [label="(384,)", style=solid]; +"304 features_5_0_norm2_bias" -> "959 layer_norm_12" [label="(384,)", style=solid]; +"305 features_5_1_norm1_weight" -> "1027 layer_norm_13" [label="(384,)", style=solid]; +"306 features_5_1_norm1_bias" -> "1027 layer_norm_13" [label="(384,)", style=solid]; +"307 _scale_61" -> "310 dequantize_per_channel_default_61" [label="(1536,)", style=solid]; +"308 _zero_point_61" -> "310 dequantize_per_channel_default_61" [label="(1536,)", style=solid]; +"309 _frozen_param97_1" -> "310 dequantize_per_channel_default_61" [label="(1536, 384)", style=solid]; +"310 dequantize_per_channel_default_61" -> "1031 linear_36" [label=None, style=solid]; +"311 features_5_1_mlp_0_bias" -> "1031 linear_36" [label="(1536,)", style=solid]; +"312 _scale_62" -> "315 dequantize_per_channel_default_62" [label="(384,)", style=solid]; +"313 _zero_point_62" -> "315 dequantize_per_channel_default_62" [label="(384,)", style=solid]; +"314 _frozen_param98_1" -> "315 dequantize_per_channel_default_62" [label="(384, 1536)", style=solid]; +"315 dequantize_per_channel_default_62" -> "1036 linear_37" [label=None, style=solid]; +"316 features_5_1_mlp_3_bias" -> "1036 linear_37" [label="(384,)", style=solid]; +"317 features_5_1_norm2_weight" -> "1038 layer_norm_14" [label="(384,)", style=solid]; +"318 features_5_1_norm2_bias" -> "1038 layer_norm_14" [label="(384,)", style=solid]; +"319 features_5_2_norm1_weight" -> "1088 layer_norm_15" [label="(384,)", style=solid]; +"320 features_5_2_norm1_bias" -> "1088 layer_norm_15" [label="(384,)", style=solid]; +"321 _scale_63" -> "324 dequantize_per_channel_default_63" [label="(1536,)", style=solid]; +"322 _zero_point_63" -> "324 dequantize_per_channel_default_63" [label="(1536,)", style=solid]; +"323 _frozen_param99_1" -> "324 dequantize_per_channel_default_63" [label="(1536, 384)", style=solid]; +"324 dequantize_per_channel_default_63" -> "1092 linear_42" [label=None, style=solid]; +"325 features_5_2_mlp_0_bias" -> "1092 linear_42" [label="(1536,)", style=solid]; +"326 _scale_64" -> "329 dequantize_per_channel_default_64" [label="(384,)", style=solid]; +"327 _zero_point_64" -> "329 dequantize_per_channel_default_64" [label="(384,)", style=solid]; +"328 _frozen_param100_1" -> "329 dequantize_per_channel_default_64" [label="(384, 1536)", style=solid]; +"329 dequantize_per_channel_default_64" -> "1097 linear_43" [label=None, style=solid]; +"330 features_5_2_mlp_3_bias" -> "1097 linear_43" [label="(384,)", style=solid]; +"331 features_5_2_norm2_weight" -> "1099 layer_norm_16" [label="(384,)", style=solid]; +"332 features_5_2_norm2_bias" -> "1099 layer_norm_16" [label="(384,)", style=solid]; +"333 features_5_3_norm1_weight" -> "1167 layer_norm_17" [label="(384,)", style=solid]; +"334 features_5_3_norm1_bias" -> "1167 layer_norm_17" [label="(384,)", style=solid]; +"335 _scale_65" -> "338 dequantize_per_channel_default_65" [label="(1536,)", style=solid]; +"336 _zero_point_65" -> "338 dequantize_per_channel_default_65" [label="(1536,)", style=solid]; +"337 _frozen_param101_1" -> "338 dequantize_per_channel_default_65" [label="(1536, 384)", style=solid]; +"338 dequantize_per_channel_default_65" -> "1171 linear_48" [label=None, style=solid]; +"339 features_5_3_mlp_0_bias" -> "1171 linear_48" [label="(1536,)", style=solid]; +"340 _scale_66" -> "343 dequantize_per_channel_default_66" [label="(384,)", style=solid]; +"341 _zero_point_66" -> "343 dequantize_per_channel_default_66" [label="(384,)", style=solid]; +"342 _frozen_param104_1" -> "343 dequantize_per_channel_default_66" [label="(384, 1536)", style=solid]; +"343 dequantize_per_channel_default_66" -> "1176 linear_49" [label=None, style=solid]; +"344 features_5_3_mlp_3_bias" -> "1176 linear_49" [label="(384,)", style=solid]; +"345 features_5_3_norm2_weight" -> "1178 layer_norm_18" [label="(384,)", style=solid]; +"346 features_5_3_norm2_bias" -> "1178 layer_norm_18" [label="(384,)", style=solid]; +"347 features_5_4_norm1_weight" -> "1228 layer_norm_19" [label="(384,)", style=solid]; +"348 features_5_4_norm1_bias" -> "1228 layer_norm_19" [label="(384,)", style=solid]; +"349 _scale_67" -> "352 dequantize_per_channel_default_67" [label="(1536,)", style=solid]; +"350 _zero_point_67" -> "352 dequantize_per_channel_default_67" [label="(1536,)", style=solid]; +"351 _frozen_param108_1" -> "352 dequantize_per_channel_default_67" [label="(1536, 384)", style=solid]; +"352 dequantize_per_channel_default_67" -> "1232 linear_54" [label=None, style=solid]; +"353 features_5_4_mlp_0_bias" -> "1232 linear_54" [label="(1536,)", style=solid]; +"354 _scale_68" -> "357 dequantize_per_channel_default_68" [label="(384,)", style=solid]; +"355 _zero_point_68" -> "357 dequantize_per_channel_default_68" [label="(384,)", style=solid]; +"356 _frozen_param110_1" -> "357 dequantize_per_channel_default_68" [label="(384, 1536)", style=solid]; +"357 dequantize_per_channel_default_68" -> "1237 linear_55" [label=None, style=solid]; +"358 features_5_4_mlp_3_bias" -> "1237 linear_55" [label="(384,)", style=solid]; +"359 features_5_4_norm2_weight" -> "1239 layer_norm_20" [label="(384,)", style=solid]; +"360 features_5_4_norm2_bias" -> "1239 layer_norm_20" [label="(384,)", style=solid]; +"361 features_5_5_norm1_weight" -> "1307 layer_norm_21" [label="(384,)", style=solid]; +"362 features_5_5_norm1_bias" -> "1307 layer_norm_21" [label="(384,)", style=solid]; +"363 _scale_69" -> "366 dequantize_per_channel_default_69" [label="(1536,)", style=solid]; +"364 _zero_point_69" -> "366 dequantize_per_channel_default_69" [label="(1536,)", style=solid]; +"365 _frozen_param111_1" -> "366 dequantize_per_channel_default_69" [label="(1536, 384)", style=solid]; +"366 dequantize_per_channel_default_69" -> "1311 linear_60" [label=None, style=solid]; +"367 features_5_5_mlp_0_bias" -> "1311 linear_60" [label="(1536,)", style=solid]; +"368 _scale_70" -> "371 dequantize_per_channel_default_70" [label="(384,)", style=solid]; +"369 _zero_point_70" -> "371 dequantize_per_channel_default_70" [label="(384,)", style=solid]; +"370 _frozen_param112_1" -> "371 dequantize_per_channel_default_70" [label="(384, 1536)", style=solid]; +"371 dequantize_per_channel_default_70" -> "1316 linear_61" [label=None, style=solid]; +"372 features_5_5_mlp_3_bias" -> "1316 linear_61" [label="(384,)", style=solid]; +"373 features_5_5_norm2_weight" -> "1318 layer_norm_22" [label="(384,)", style=solid]; +"374 features_5_5_norm2_bias" -> "1318 layer_norm_22" [label="(384,)", style=solid]; +"375 features_5_6_norm1_weight" -> "1368 layer_norm_23" [label="(384,)", style=solid]; +"376 features_5_6_norm1_bias" -> "1368 layer_norm_23" [label="(384,)", style=solid]; +"377 _scale_71" -> "380 dequantize_per_channel_default_71" [label="(1536,)", style=solid]; +"378 _zero_point_71" -> "380 dequantize_per_channel_default_71" [label="(1536,)", style=solid]; +"379 _frozen_param113_1" -> "380 dequantize_per_channel_default_71" [label="(1536, 384)", style=solid]; +"380 dequantize_per_channel_default_71" -> "1372 linear_66" [label=None, style=solid]; +"381 features_5_6_mlp_0_bias" -> "1372 linear_66" [label="(1536,)", style=solid]; +"382 _scale_72" -> "385 dequantize_per_channel_default_72" [label="(384,)", style=solid]; +"383 _zero_point_72" -> "385 dequantize_per_channel_default_72" [label="(384,)", style=solid]; +"384 _frozen_param114_1" -> "385 dequantize_per_channel_default_72" [label="(384, 1536)", style=solid]; +"385 dequantize_per_channel_default_72" -> "1377 linear_67" [label=None, style=solid]; +"386 features_5_6_mlp_3_bias" -> "1377 linear_67" [label="(384,)", style=solid]; +"387 features_5_6_norm2_weight" -> "1379 layer_norm_24" [label="(384,)", style=solid]; +"388 features_5_6_norm2_bias" -> "1379 layer_norm_24" [label="(384,)", style=solid]; +"389 features_5_7_norm1_weight" -> "1447 layer_norm_25" [label="(384,)", style=solid]; +"390 features_5_7_norm1_bias" -> "1447 layer_norm_25" [label="(384,)", style=solid]; +"391 _scale_73" -> "394 dequantize_per_channel_default_73" [label="(1536,)", style=solid]; +"392 _zero_point_73" -> "394 dequantize_per_channel_default_73" [label="(1536,)", style=solid]; +"393 _frozen_param115_1" -> "394 dequantize_per_channel_default_73" [label="(1536, 384)", style=solid]; +"394 dequantize_per_channel_default_73" -> "1451 linear_72" [label=None, style=solid]; +"395 features_5_7_mlp_0_bias" -> "1451 linear_72" [label="(1536,)", style=solid]; +"396 _scale_74" -> "399 dequantize_per_channel_default_74" [label="(384,)", style=solid]; +"397 _zero_point_74" -> "399 dequantize_per_channel_default_74" [label="(384,)", style=solid]; +"398 _frozen_param116_1" -> "399 dequantize_per_channel_default_74" [label="(384, 1536)", style=solid]; +"399 dequantize_per_channel_default_74" -> "1456 linear_73" [label=None, style=solid]; +"400 features_5_7_mlp_3_bias" -> "1456 linear_73" [label="(384,)", style=solid]; +"401 features_5_7_norm2_weight" -> "1458 layer_norm_26" [label="(384,)", style=solid]; +"402 features_5_7_norm2_bias" -> "1458 layer_norm_26" [label="(384,)", style=solid]; +"403 features_5_8_norm1_weight" -> "1508 layer_norm_27" [label="(384,)", style=solid]; +"404 features_5_8_norm1_bias" -> "1508 layer_norm_27" [label="(384,)", style=solid]; +"405 _scale_75" -> "408 dequantize_per_channel_default_75" [label="(1536,)", style=solid]; +"406 _zero_point_75" -> "408 dequantize_per_channel_default_75" [label="(1536,)", style=solid]; +"407 _frozen_param117_1" -> "408 dequantize_per_channel_default_75" [label="(1536, 384)", style=solid]; +"408 dequantize_per_channel_default_75" -> "1512 linear_78" [label=None, style=solid]; +"409 features_5_8_mlp_0_bias" -> "1512 linear_78" [label="(1536,)", style=solid]; +"410 _scale_76" -> "413 dequantize_per_channel_default_76" [label="(384,)", style=solid]; +"411 _zero_point_76" -> "413 dequantize_per_channel_default_76" [label="(384,)", style=solid]; +"412 _frozen_param118_1" -> "413 dequantize_per_channel_default_76" [label="(384, 1536)", style=solid]; +"413 dequantize_per_channel_default_76" -> "1517 linear_79" [label=None, style=solid]; +"414 features_5_8_mlp_3_bias" -> "1517 linear_79" [label="(384,)", style=solid]; +"415 features_5_8_norm2_weight" -> "1519 layer_norm_28" [label="(384,)", style=solid]; +"416 features_5_8_norm2_bias" -> "1519 layer_norm_28" [label="(384,)", style=solid]; +"417 features_5_9_norm1_weight" -> "1587 layer_norm_29" [label="(384,)", style=solid]; +"418 features_5_9_norm1_bias" -> "1587 layer_norm_29" [label="(384,)", style=solid]; +"419 _scale_77" -> "422 dequantize_per_channel_default_77" [label="(1536,)", style=solid]; +"420 _zero_point_77" -> "422 dequantize_per_channel_default_77" [label="(1536,)", style=solid]; +"421 _frozen_param121_1" -> "422 dequantize_per_channel_default_77" [label="(1536, 384)", style=solid]; +"422 dequantize_per_channel_default_77" -> "1591 linear_84" [label=None, style=solid]; +"423 features_5_9_mlp_0_bias" -> "1591 linear_84" [label="(1536,)", style=solid]; +"424 _scale_78" -> "427 dequantize_per_channel_default_78" [label="(384,)", style=solid]; +"425 _zero_point_78" -> "427 dequantize_per_channel_default_78" [label="(384,)", style=solid]; +"426 _frozen_param125_1" -> "427 dequantize_per_channel_default_78" [label="(384, 1536)", style=solid]; +"427 dequantize_per_channel_default_78" -> "1596 linear_85" [label=None, style=solid]; +"428 features_5_9_mlp_3_bias" -> "1596 linear_85" [label="(384,)", style=solid]; +"429 features_5_9_norm2_weight" -> "1598 layer_norm_30" [label="(384,)", style=solid]; +"430 features_5_9_norm2_bias" -> "1598 layer_norm_30" [label="(384,)", style=solid]; +"431 features_5_10_norm1_weight" -> "1648 layer_norm_31" [label="(384,)", style=solid]; +"432 features_5_10_norm1_bias" -> "1648 layer_norm_31" [label="(384,)", style=solid]; +"433 _scale_79" -> "436 dequantize_per_channel_default_79" [label="(1536,)", style=solid]; +"434 _zero_point_79" -> "436 dequantize_per_channel_default_79" [label="(1536,)", style=solid]; +"435 _frozen_param127_1" -> "436 dequantize_per_channel_default_79" [label="(1536, 384)", style=solid]; +"436 dequantize_per_channel_default_79" -> "1652 linear_90" [label=None, style=solid]; +"437 features_5_10_mlp_0_bias" -> "1652 linear_90" [label="(1536,)", style=solid]; +"438 _scale_80" -> "441 dequantize_per_channel_default_80" [label="(384,)", style=solid]; +"439 _zero_point_80" -> "441 dequantize_per_channel_default_80" [label="(384,)", style=solid]; +"440 _frozen_param128_1" -> "441 dequantize_per_channel_default_80" [label="(384, 1536)", style=solid]; +"441 dequantize_per_channel_default_80" -> "1657 linear_91" [label=None, style=solid]; +"442 features_5_10_mlp_3_bias" -> "1657 linear_91" [label="(384,)", style=solid]; +"443 features_5_10_norm2_weight" -> "1659 layer_norm_32" [label="(384,)", style=solid]; +"444 features_5_10_norm2_bias" -> "1659 layer_norm_32" [label="(384,)", style=solid]; +"445 features_5_11_norm1_weight" -> "1727 layer_norm_33" [label="(384,)", style=solid]; +"446 features_5_11_norm1_bias" -> "1727 layer_norm_33" [label="(384,)", style=solid]; +"447 _scale_81" -> "450 dequantize_per_channel_default_81" [label="(1536,)", style=solid]; +"448 _zero_point_81" -> "450 dequantize_per_channel_default_81" [label="(1536,)", style=solid]; +"449 _frozen_param129_1" -> "450 dequantize_per_channel_default_81" [label="(1536, 384)", style=solid]; +"450 dequantize_per_channel_default_81" -> "1731 linear_96" [label=None, style=solid]; +"451 features_5_11_mlp_0_bias" -> "1731 linear_96" [label="(1536,)", style=solid]; +"452 _scale_82" -> "455 dequantize_per_channel_default_82" [label="(384,)", style=solid]; +"453 _zero_point_82" -> "455 dequantize_per_channel_default_82" [label="(384,)", style=solid]; +"454 _frozen_param130_1" -> "455 dequantize_per_channel_default_82" [label="(384, 1536)", style=solid]; +"455 dequantize_per_channel_default_82" -> "1736 linear_97" [label=None, style=solid]; +"456 features_5_11_mlp_3_bias" -> "1736 linear_97" [label="(384,)", style=solid]; +"457 features_5_11_norm2_weight" -> "1738 layer_norm_34" [label="(384,)", style=solid]; +"458 features_5_11_norm2_bias" -> "1738 layer_norm_34" [label="(384,)", style=solid]; +"459 features_5_12_norm1_weight" -> "1788 layer_norm_35" [label="(384,)", style=solid]; +"460 features_5_12_norm1_bias" -> "1788 layer_norm_35" [label="(384,)", style=solid]; +"461 _scale_83" -> "464 dequantize_per_channel_default_83" [label="(1536,)", style=solid]; +"462 _zero_point_83" -> "464 dequantize_per_channel_default_83" [label="(1536,)", style=solid]; +"463 _frozen_param131_1" -> "464 dequantize_per_channel_default_83" [label="(1536, 384)", style=solid]; +"464 dequantize_per_channel_default_83" -> "1792 linear_102" [label=None, style=solid]; +"465 features_5_12_mlp_0_bias" -> "1792 linear_102" [label="(1536,)", style=solid]; +"466 _scale_84" -> "469 dequantize_per_channel_default_84" [label="(384,)", style=solid]; +"467 _zero_point_84" -> "469 dequantize_per_channel_default_84" [label="(384,)", style=solid]; +"468 _frozen_param132_1" -> "469 dequantize_per_channel_default_84" [label="(384, 1536)", style=solid]; +"469 dequantize_per_channel_default_84" -> "1797 linear_103" [label=None, style=solid]; +"470 features_5_12_mlp_3_bias" -> "1797 linear_103" [label="(384,)", style=solid]; +"471 features_5_12_norm2_weight" -> "1799 layer_norm_36" [label="(384,)", style=solid]; +"472 features_5_12_norm2_bias" -> "1799 layer_norm_36" [label="(384,)", style=solid]; +"473 features_5_13_norm1_weight" -> "1867 layer_norm_37" [label="(384,)", style=solid]; +"474 features_5_13_norm1_bias" -> "1867 layer_norm_37" [label="(384,)", style=solid]; +"475 _scale_85" -> "478 dequantize_per_channel_default_85" [label="(1536,)", style=solid]; +"476 _zero_point_85" -> "478 dequantize_per_channel_default_85" [label="(1536,)", style=solid]; +"477 _frozen_param133_1" -> "478 dequantize_per_channel_default_85" [label="(1536, 384)", style=solid]; +"478 dequantize_per_channel_default_85" -> "1871 linear_108" [label=None, style=solid]; +"479 features_5_13_mlp_0_bias" -> "1871 linear_108" [label="(1536,)", style=solid]; +"480 _scale_86" -> "483 dequantize_per_channel_default_86" [label="(384,)", style=solid]; +"481 _zero_point_86" -> "483 dequantize_per_channel_default_86" [label="(384,)", style=solid]; +"482 _frozen_param134_1" -> "483 dequantize_per_channel_default_86" [label="(384, 1536)", style=solid]; +"483 dequantize_per_channel_default_86" -> "1876 linear_109" [label=None, style=solid]; +"484 features_5_13_mlp_3_bias" -> "1876 linear_109" [label="(384,)", style=solid]; +"485 features_5_13_norm2_weight" -> "1878 layer_norm_38" [label="(384,)", style=solid]; +"486 features_5_13_norm2_bias" -> "1878 layer_norm_38" [label="(384,)", style=solid]; +"487 features_5_14_norm1_weight" -> "1928 layer_norm_39" [label="(384,)", style=solid]; +"488 features_5_14_norm1_bias" -> "1928 layer_norm_39" [label="(384,)", style=solid]; +"489 _scale_87" -> "492 dequantize_per_channel_default_87" [label="(1536,)", style=solid]; +"490 _zero_point_87" -> "492 dequantize_per_channel_default_87" [label="(1536,)", style=solid]; +"491 _frozen_param135_1" -> "492 dequantize_per_channel_default_87" [label="(1536, 384)", style=solid]; +"492 dequantize_per_channel_default_87" -> "1932 linear_114" [label=None, style=solid]; +"493 features_5_14_mlp_0_bias" -> "1932 linear_114" [label="(1536,)", style=solid]; +"494 _scale_88" -> "497 dequantize_per_channel_default_88" [label="(384,)", style=solid]; +"495 _zero_point_88" -> "497 dequantize_per_channel_default_88" [label="(384,)", style=solid]; +"496 _frozen_param138_1" -> "497 dequantize_per_channel_default_88" [label="(384, 1536)", style=solid]; +"497 dequantize_per_channel_default_88" -> "1937 linear_115" [label=None, style=solid]; +"498 features_5_14_mlp_3_bias" -> "1937 linear_115" [label="(384,)", style=solid]; +"499 features_5_14_norm2_weight" -> "1939 layer_norm_40" [label="(384,)", style=solid]; +"500 features_5_14_norm2_bias" -> "1939 layer_norm_40" [label="(384,)", style=solid]; +"501 features_5_15_norm1_weight" -> "2007 layer_norm_41" [label="(384,)", style=solid]; +"502 features_5_15_norm1_bias" -> "2007 layer_norm_41" [label="(384,)", style=solid]; +"503 _scale_89" -> "506 dequantize_per_channel_default_89" [label="(1536,)", style=solid]; +"504 _zero_point_89" -> "506 dequantize_per_channel_default_89" [label="(1536,)", style=solid]; +"505 _frozen_param142_1" -> "506 dequantize_per_channel_default_89" [label="(1536, 384)", style=solid]; +"506 dequantize_per_channel_default_89" -> "2011 linear_120" [label=None, style=solid]; +"507 features_5_15_mlp_0_bias" -> "2011 linear_120" [label="(1536,)", style=solid]; +"508 _scale_90" -> "511 dequantize_per_channel_default_90" [label="(384,)", style=solid]; +"509 _zero_point_90" -> "511 dequantize_per_channel_default_90" [label="(384,)", style=solid]; +"510 _frozen_param144_1" -> "511 dequantize_per_channel_default_90" [label="(384, 1536)", style=solid]; +"511 dequantize_per_channel_default_90" -> "2016 linear_121" [label=None, style=solid]; +"512 features_5_15_mlp_3_bias" -> "2016 linear_121" [label="(384,)", style=solid]; +"513 features_5_15_norm2_weight" -> "2018 layer_norm_42" [label="(384,)", style=solid]; +"514 features_5_15_norm2_bias" -> "2018 layer_norm_42" [label="(384,)", style=solid]; +"515 features_5_16_norm1_weight" -> "2068 layer_norm_43" [label="(384,)", style=solid]; +"516 features_5_16_norm1_bias" -> "2068 layer_norm_43" [label="(384,)", style=solid]; +"517 _scale_91" -> "520 dequantize_per_channel_default_91" [label="(1536,)", style=solid]; +"518 _zero_point_91" -> "520 dequantize_per_channel_default_91" [label="(1536,)", style=solid]; +"519 _frozen_param145_1" -> "520 dequantize_per_channel_default_91" [label="(1536, 384)", style=solid]; +"520 dequantize_per_channel_default_91" -> "2072 linear_126" [label=None, style=solid]; +"521 features_5_16_mlp_0_bias" -> "2072 linear_126" [label="(1536,)", style=solid]; +"522 _scale_92" -> "525 dequantize_per_channel_default_92" [label="(384,)", style=solid]; +"523 _zero_point_92" -> "525 dequantize_per_channel_default_92" [label="(384,)", style=solid]; +"524 _frozen_param146_1" -> "525 dequantize_per_channel_default_92" [label="(384, 1536)", style=solid]; +"525 dequantize_per_channel_default_92" -> "2077 linear_127" [label=None, style=solid]; +"526 features_5_16_mlp_3_bias" -> "2077 linear_127" [label="(384,)", style=solid]; +"527 features_5_16_norm2_weight" -> "2079 layer_norm_44" [label="(384,)", style=solid]; +"528 features_5_16_norm2_bias" -> "2079 layer_norm_44" [label="(384,)", style=solid]; +"529 features_5_17_norm1_weight" -> "2147 layer_norm_45" [label="(384,)", style=solid]; +"530 features_5_17_norm1_bias" -> "2147 layer_norm_45" [label="(384,)", style=solid]; +"531 _scale_93" -> "534 dequantize_per_channel_default_93" [label="(1536,)", style=solid]; +"532 _zero_point_93" -> "534 dequantize_per_channel_default_93" [label="(1536,)", style=solid]; +"533 _frozen_param147_1" -> "534 dequantize_per_channel_default_93" [label="(1536, 384)", style=solid]; +"534 dequantize_per_channel_default_93" -> "2151 linear_132" [label=None, style=solid]; +"535 features_5_17_mlp_0_bias" -> "2151 linear_132" [label="(1536,)", style=solid]; +"536 _scale_94" -> "539 dequantize_per_channel_default_94" [label="(384,)", style=solid]; +"537 _zero_point_94" -> "539 dequantize_per_channel_default_94" [label="(384,)", style=solid]; +"538 _frozen_param148_1" -> "539 dequantize_per_channel_default_94" [label="(384, 1536)", style=solid]; +"539 dequantize_per_channel_default_94" -> "2156 linear_133" [label=None, style=solid]; +"540 features_5_17_mlp_3_bias" -> "2156 linear_133" [label="(384,)", style=solid]; +"541 features_5_17_norm2_weight" -> "2158 layer_norm_46" [label="(384,)", style=solid]; +"542 features_5_17_norm2_bias" -> "2158 layer_norm_46" [label="(384,)", style=solid]; +"543 _scale_95" -> "546 dequantize_per_channel_default_95" [label="(768,)", style=solid]; +"544 _zero_point_95" -> "546 dequantize_per_channel_default_95" [label="(768,)", style=solid]; +"545 _frozen_param149_1" -> "546 dequantize_per_channel_default_95" [label="(768, 1536)", style=solid]; +"546 dequantize_per_channel_default_95" -> "2176 linear_134" [label=None, style=solid]; +"547 features_6_norm_weight" -> "2177 layer_norm_47" [label="(768,)", style=solid]; +"548 features_6_norm_bias" -> "2177 layer_norm_47" [label="(768,)", style=solid]; +"549 features_7_0_norm1_weight" -> "2226 layer_norm_48" [label="(768,)", style=solid]; +"550 features_7_0_norm1_bias" -> "2226 layer_norm_48" [label="(768,)", style=solid]; +"551 _scale_96" -> "554 dequantize_per_channel_default_96" [label="(3072,)", style=solid]; +"552 _zero_point_96" -> "554 dequantize_per_channel_default_96" [label="(3072,)", style=solid]; +"553 _frozen_param150_1" -> "554 dequantize_per_channel_default_96" [label="(3072, 768)", style=solid]; +"554 dequantize_per_channel_default_96" -> "2230 linear_139" [label=None, style=solid]; +"555 features_7_0_mlp_0_bias" -> "2230 linear_139" [label="(3072,)", style=solid]; +"556 _scale_97" -> "559 dequantize_per_channel_default_97" [label="(768,)", style=solid]; +"557 _zero_point_97" -> "559 dequantize_per_channel_default_97" [label="(768,)", style=solid]; +"558 _frozen_param151_1" -> "559 dequantize_per_channel_default_97" [label="(768, 3072)", style=solid]; +"559 dequantize_per_channel_default_97" -> "2235 linear_140" [label=None, style=solid]; +"560 features_7_0_mlp_3_bias" -> "2235 linear_140" [label="(768,)", style=solid]; +"561 features_7_0_norm2_weight" -> "2237 layer_norm_49" [label="(768,)", style=solid]; +"562 features_7_0_norm2_bias" -> "2237 layer_norm_49" [label="(768,)", style=solid]; +"563 features_7_1_norm1_weight" -> "2287 layer_norm_50" [label="(768,)", style=solid]; +"564 features_7_1_norm1_bias" -> "2287 layer_norm_50" [label="(768,)", style=solid]; +"565 _scale_98" -> "568 dequantize_per_channel_default_98" [label="(3072,)", style=solid]; +"566 _zero_point_98" -> "568 dequantize_per_channel_default_98" [label="(3072,)", style=solid]; +"567 _frozen_param152_1" -> "568 dequantize_per_channel_default_98" [label="(3072, 768)", style=solid]; +"568 dequantize_per_channel_default_98" -> "2291 linear_145" [label=None, style=solid]; +"569 features_7_1_mlp_0_bias" -> "2291 linear_145" [label="(3072,)", style=solid]; +"570 _scale_99" -> "573 dequantize_per_channel_default_99" [label="(768,)", style=solid]; +"571 _zero_point_99" -> "573 dequantize_per_channel_default_99" [label="(768,)", style=solid]; +"572 _frozen_param155_1" -> "573 dequantize_per_channel_default_99" [label="(768, 3072)", style=solid]; +"573 dequantize_per_channel_default_99" -> "2296 linear_146" [label=None, style=solid]; +"574 features_7_1_mlp_3_bias" -> "2296 linear_146" [label="(768,)", style=solid]; +"575 features_7_1_norm2_weight" -> "2298 layer_norm_51" [label="(768,)", style=solid]; +"576 features_7_1_norm2_bias" -> "2298 layer_norm_51" [label="(768,)", style=solid]; +"577 norm_weight" -> "2300 layer_norm_52" [label="(768,)", style=solid]; +"578 norm_bias" -> "2300 layer_norm_52" [label="(768,)", style=solid]; +"579 _scale_100" -> "582 dequantize_per_channel_default_100" [label="(1000,)", style=solid]; +"580 _zero_point_100" -> "582 dequantize_per_channel_default_100" [label="(1000,)", style=solid]; +"581 _frozen_param159_1" -> "582 dequantize_per_channel_default_100" [label="(1000, 768)", style=solid]; +"582 dequantize_per_channel_default_100" -> "2306 linear_147" [label=None, style=solid]; +"583 head_bias" -> "2306 linear_147" [label="(1000,)", style=solid]; +"584 x" -> "585 quantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; +"585 quantize_per_tensor_default" -> "586 dequantize_per_tensor_default" [label=None, style=solid]; +"586 dequantize_per_tensor_default" -> "587 conv2d" [label=None, style=solid]; +"587 conv2d" -> "588 permute" [label="(1, 96, 56, 56)", style=solid]; +"588 permute" -> "589 layer_norm" [label="(1, 56, 56, 96)", style=solid]; +"589 layer_norm" -> "590 quantize_per_tensor_default_1" [label="(1, 56, 56, 96)", style=solid]; +"589 layer_norm" -> "636 add_1" [label="(1, 56, 56, 96)", style=solid]; +"590 quantize_per_tensor_default_1" -> "591 dequantize_per_tensor_default_1" [label=None, style=solid]; +"591 dequantize_per_tensor_default_1" -> "593 pad" [label=None, style=solid]; +"592 _frozen_param0" -> "620 add" [label="(1, 3, 64, 64)", style=solid]; +"593 pad" -> "594 view_2" [label="(1, 56, 56, 96)", style=solid]; +"594 view_2" -> "595 permute_2" [label="(1, 7, 8, 7, 8, 96)", style=solid]; +"595 permute_2" -> "596 reshape" [label="(1, 7, 7, 8, 8, 96)", style=solid]; +"596 reshape" -> "598 linear_2" [label="(49, 64, 96)", style=solid]; +"597 _frozen_param1" -> "598 linear_2" [label="(288,)", style=solid]; +"598 linear_2" -> "599 reshape_1" [label="(49, 64, 288)", style=solid]; +"599 reshape_1" -> "600 permute_3" [label="(49, 64, 3, 3, 32)", style=solid]; +"600 permute_3" -> "601 select" [label="(3, 49, 3, 64, 32)", style=solid]; +"600 permute_3" -> "602 select_1" [label="(3, 49, 3, 64, 32)", style=solid]; +"600 permute_3" -> "603 select_2" [label="(3, 49, 3, 64, 32)", style=solid]; +"601 select" -> "604 linalg_vector_norm" [label="(49, 3, 64, 32)", style=solid]; +"601 select" -> "606 expand_as" [label="(49, 3, 64, 32)", style=solid]; +"601 select" -> "607 div" [label="(49, 3, 64, 32)", style=solid]; +"602 select_1" -> "610 linalg_vector_norm_1" [label="(49, 3, 64, 32)", style=solid]; +"602 select_1" -> "612 expand_as_1" [label="(49, 3, 64, 32)", style=solid]; +"602 select_1" -> "613 div_1" [label="(49, 3, 64, 32)", style=solid]; +"603 select_2" -> "623 matmul_1" [label="(49, 3, 64, 32)", style=solid]; +"604 linalg_vector_norm" -> "605 clamp_min" [label="(49, 3, 64, 1)", style=solid]; +"605 clamp_min" -> "606 expand_as" [label="(49, 3, 64, 1)", style=solid]; +"606 expand_as" -> "607 div" [label="(49, 3, 64, 32)", style=solid]; +"607 div" -> "608 quantize_per_tensor_default_2" [label="(49, 3, 64, 32)", style=solid]; +"608 quantize_per_tensor_default_2" -> "609 dequantize_per_tensor_default_2" [label=None, style=solid]; +"609 dequantize_per_tensor_default_2" -> "617 matmul" [label=None, style=solid]; +"610 linalg_vector_norm_1" -> "611 clamp_min_1" [label="(49, 3, 64, 1)", style=solid]; +"611 clamp_min_1" -> "612 expand_as_1" [label="(49, 3, 64, 1)", style=solid]; +"612 expand_as_1" -> "613 div_1" [label="(49, 3, 64, 32)", style=solid]; +"613 div_1" -> "614 quantize_per_tensor_default_3" [label="(49, 3, 64, 32)", style=solid]; +"614 quantize_per_tensor_default_3" -> "615 dequantize_per_tensor_default_3" [label=None, style=solid]; +"615 dequantize_per_tensor_default_3" -> "616 transpose" [label=None, style=solid]; +"616 transpose" -> "617 matmul" [label="(49, 3, 32, 64)", style=solid]; +"617 matmul" -> "619 mul_1" [label="(49, 3, 64, 64)", style=solid]; +"618 _frozen_param3" -> "619 mul_1" [label="(3, 1, 1)", style=solid]; +"619 mul_1" -> "620 add" [label="(49, 3, 64, 64)", style=solid]; +"620 add" -> "621 softmax" [label="(49, 3, 64, 64)", style=solid]; +"621 softmax" -> "622 dropout" [label="(49, 3, 64, 64)", style=solid]; +"622 dropout" -> "623 matmul_1" [label="(49, 3, 64, 64)", style=solid]; +"623 matmul_1" -> "624 quantize_per_tensor_default_4" [label="(49, 3, 64, 32)", style=solid]; +"624 quantize_per_tensor_default_4" -> "625 dequantize_per_tensor_default_4" [label=None, style=solid]; +"625 dequantize_per_tensor_default_4" -> "626 transpose_1" [label=None, style=solid]; +"626 transpose_1" -> "627 reshape_2" [label="(49, 64, 3, 32)", style=solid]; +"627 reshape_2" -> "628 linear_3" [label="(49, 64, 96)", style=solid]; +"628 linear_3" -> "629 dropout_1" [label="(49, 64, 96)", style=solid]; +"629 dropout_1" -> "630 view_3" [label="(49, 64, 96)", style=solid]; +"630 view_3" -> "631 permute_4" [label="(1, 7, 7, 8, 8, 96)", style=solid]; +"631 permute_4" -> "632 reshape_3" [label="(1, 7, 8, 7, 8, 96)", style=solid]; +"632 reshape_3" -> "633 slice_2" [label="(1, 56, 56, 96)", style=solid]; +"633 slice_2" -> "634 slice_3" [label="(1, 56, 56, 96)", style=solid]; +"634 slice_3" -> "635 layer_norm_1" [label="(1, 56, 56, 96)", style=solid]; +"635 layer_norm_1" -> "636 add_1" [label="(1, 56, 56, 96)", style=solid]; +"636 add_1" -> "637 quantize_per_tensor_default_5" [label="(1, 56, 56, 96)", style=solid]; +"636 add_1" -> "647 add_2" [label="(1, 56, 56, 96)", style=solid]; +"637 quantize_per_tensor_default_5" -> "638 dequantize_per_tensor_default_5" [label=None, style=solid]; +"638 dequantize_per_tensor_default_5" -> "639 linear_4" [label=None, style=solid]; +"639 linear_4" -> "640 gelu" [label="(1, 56, 56, 384)", style=solid]; +"640 gelu" -> "641 quantize_per_tensor_default_6" [label="(1, 56, 56, 384)", style=solid]; +"641 quantize_per_tensor_default_6" -> "642 dequantize_per_tensor_default_6" [label=None, style=solid]; +"642 dequantize_per_tensor_default_6" -> "643 dropout_2" [label=None, style=solid]; +"643 dropout_2" -> "644 linear_5" [label="(1, 56, 56, 384)", style=solid]; +"644 linear_5" -> "645 dropout_3" [label="(1, 56, 56, 96)", style=solid]; +"645 dropout_3" -> "646 layer_norm_2" [label="(1, 56, 56, 96)", style=solid]; +"646 layer_norm_2" -> "647 add_2" [label="(1, 56, 56, 96)", style=solid]; +"647 add_2" -> "649 pad_1" [label="(1, 56, 56, 96)", style=solid]; +"647 add_2" -> "712 add_5" [label="(1, 56, 56, 96)", style=solid]; +"648 _frozen_param4" -> "679 add_3" [label="(1, 3, 64, 64)", style=solid]; +"649 pad_1" -> "650 roll" [label="(1, 56, 56, 96)", style=solid]; +"650 roll" -> "651 view_6" [label="(1, 56, 56, 96)", style=solid]; +"651 view_6" -> "652 permute_6" [label="(1, 7, 8, 7, 8, 96)", style=solid]; +"652 permute_6" -> "653 reshape_4" [label="(1, 7, 7, 8, 8, 96)", style=solid]; +"653 reshape_4" -> "654 quantize_per_tensor_default_7" [label="(49, 64, 96)", style=solid]; +"653 reshape_4" -> "680 new_zeros" [label="(49, 64, 96)", style=solid]; +"654 quantize_per_tensor_default_7" -> "655 dequantize_per_tensor_default_7" [label=None, style=solid]; +"655 dequantize_per_tensor_default_7" -> "657 linear_8" [label=None, style=solid]; +"656 _frozen_param5" -> "657 linear_8" [label="(288,)", style=solid]; +"657 linear_8" -> "658 reshape_5" [label="(49, 64, 288)", style=solid]; +"658 reshape_5" -> "659 permute_7" [label="(49, 64, 3, 3, 32)", style=solid]; +"659 permute_7" -> "660 select_3" [label="(3, 49, 3, 64, 32)", style=solid]; +"659 permute_7" -> "661 select_4" [label="(3, 49, 3, 64, 32)", style=solid]; +"659 permute_7" -> "662 select_5" [label="(3, 49, 3, 64, 32)", style=solid]; +"660 select_3" -> "663 linalg_vector_norm_2" [label="(49, 3, 64, 32)", style=solid]; +"660 select_3" -> "665 expand_as_2" [label="(49, 3, 64, 32)", style=solid]; +"660 select_3" -> "666 div_2" [label="(49, 3, 64, 32)", style=solid]; +"661 select_4" -> "669 linalg_vector_norm_3" [label="(49, 3, 64, 32)", style=solid]; +"661 select_4" -> "671 expand_as_3" [label="(49, 3, 64, 32)", style=solid]; +"661 select_4" -> "672 div_3" [label="(49, 3, 64, 32)", style=solid]; +"662 select_5" -> "698 matmul_3" [label="(49, 3, 64, 32)", style=solid]; +"663 linalg_vector_norm_2" -> "664 clamp_min_2" [label="(49, 3, 64, 1)", style=solid]; +"664 clamp_min_2" -> "665 expand_as_2" [label="(49, 3, 64, 1)", style=solid]; +"665 expand_as_2" -> "666 div_2" [label="(49, 3, 64, 32)", style=solid]; +"666 div_2" -> "667 quantize_per_tensor_default_8" [label="(49, 3, 64, 32)", style=solid]; +"667 quantize_per_tensor_default_8" -> "668 dequantize_per_tensor_default_8" [label=None, style=solid]; +"668 dequantize_per_tensor_default_8" -> "676 matmul_2" [label=None, style=solid]; +"669 linalg_vector_norm_3" -> "670 clamp_min_3" [label="(49, 3, 64, 1)", style=solid]; +"670 clamp_min_3" -> "671 expand_as_3" [label="(49, 3, 64, 1)", style=solid]; +"671 expand_as_3" -> "672 div_3" [label="(49, 3, 64, 32)", style=solid]; +"672 div_3" -> "673 quantize_per_tensor_default_9" [label="(49, 3, 64, 32)", style=solid]; +"673 quantize_per_tensor_default_9" -> "674 dequantize_per_tensor_default_9" [label=None, style=solid]; +"674 dequantize_per_tensor_default_9" -> "675 transpose_2" [label=None, style=solid]; +"675 transpose_2" -> "676 matmul_2" [label="(49, 3, 32, 64)", style=solid]; +"676 matmul_2" -> "678 mul_3" [label="(49, 3, 64, 64)", style=solid]; +"677 _frozen_param7" -> "678 mul_3" [label="(3, 1, 1)", style=solid]; +"678 mul_3" -> "679 add_3" [label="(49, 3, 64, 64)", style=solid]; +"679 add_3" -> "691 view_8" [label="(49, 3, 64, 64)", style=solid]; +"680 new_zeros" -> "681 view_7" [label="(56, 56)", style=solid]; +"681 view_7" -> "682 permute_8" [label="(7, 8, 7, 8)", style=solid]; +"682 permute_8" -> "683 reshape_6" [label="(7, 7, 8, 8)", style=solid]; +"683 reshape_6" -> "684 unsqueeze_2" [label="(49, 64)", style=solid]; +"683 reshape_6" -> "685 unsqueeze_3" [label="(49, 64)", style=solid]; +"684 unsqueeze_2" -> "686 sub" [label="(49, 1, 64)", style=solid]; +"685 unsqueeze_3" -> "686 sub" [label="(49, 64, 1)", style=solid]; +"686 sub" -> "687 ne" [label="(49, 64, 64)", style=solid]; +"686 sub" -> "688 masked_fill" [label="(49, 64, 64)", style=solid]; +"686 sub" -> "689 eq" [label="(49, 64, 64)", style=solid]; +"687 ne" -> "688 masked_fill" [label="(49, 64, 64)", style=solid]; +"688 masked_fill" -> "690 masked_fill_1" [label="(49, 64, 64)", style=solid]; +"689 eq" -> "690 masked_fill_1" [label="(49, 64, 64)", style=solid]; +"690 masked_fill_1" -> "692 unsqueeze_4" [label="(49, 64, 64)", style=solid]; +"691 view_8" -> "694 add_4" [label="(1, 49, 3, 64, 64)", style=solid]; +"692 unsqueeze_4" -> "693 unsqueeze_5" [label="(49, 1, 64, 64)", style=solid]; +"693 unsqueeze_5" -> "694 add_4" [label="(1, 49, 1, 64, 64)", style=solid]; +"694 add_4" -> "695 view_9" [label="(1, 49, 3, 64, 64)", style=solid]; +"695 view_9" -> "696 softmax_1" [label="(49, 3, 64, 64)", style=solid]; +"696 softmax_1" -> "697 dropout_4" [label="(49, 3, 64, 64)", style=solid]; +"697 dropout_4" -> "698 matmul_3" [label="(49, 3, 64, 64)", style=solid]; +"698 matmul_3" -> "699 quantize_per_tensor_default_10" [label="(49, 3, 64, 32)", style=solid]; +"699 quantize_per_tensor_default_10" -> "700 dequantize_per_tensor_default_10" [label=None, style=solid]; +"700 dequantize_per_tensor_default_10" -> "701 transpose_3" [label=None, style=solid]; +"701 transpose_3" -> "702 reshape_7" [label="(49, 64, 3, 32)", style=solid]; +"702 reshape_7" -> "703 linear_9" [label="(49, 64, 96)", style=solid]; +"703 linear_9" -> "704 dropout_5" [label="(49, 64, 96)", style=solid]; +"704 dropout_5" -> "705 view_10" [label="(49, 64, 96)", style=solid]; +"705 view_10" -> "706 permute_9" [label="(1, 7, 7, 8, 8, 96)", style=solid]; +"706 permute_9" -> "707 reshape_8" [label="(1, 7, 8, 7, 8, 96)", style=solid]; +"707 reshape_8" -> "708 roll_1" [label="(1, 56, 56, 96)", style=solid]; +"708 roll_1" -> "709 slice_23" [label="(1, 56, 56, 96)", style=solid]; +"709 slice_23" -> "710 slice_24" [label="(1, 56, 56, 96)", style=solid]; +"710 slice_24" -> "711 layer_norm_3" [label="(1, 56, 56, 96)", style=solid]; +"711 layer_norm_3" -> "712 add_5" [label="(1, 56, 56, 96)", style=solid]; +"712 add_5" -> "713 quantize_per_tensor_default_11" [label="(1, 56, 56, 96)", style=solid]; +"712 add_5" -> "723 add_6" [label="(1, 56, 56, 96)", style=solid]; +"713 quantize_per_tensor_default_11" -> "714 dequantize_per_tensor_default_11" [label=None, style=solid]; +"714 dequantize_per_tensor_default_11" -> "715 linear_10" [label=None, style=solid]; +"715 linear_10" -> "716 gelu_1" [label="(1, 56, 56, 384)", style=solid]; +"716 gelu_1" -> "717 quantize_per_tensor_default_12" [label="(1, 56, 56, 384)", style=solid]; +"717 quantize_per_tensor_default_12" -> "718 dequantize_per_tensor_default_12" [label=None, style=solid]; +"718 dequantize_per_tensor_default_12" -> "719 dropout_6" [label=None, style=solid]; +"719 dropout_6" -> "720 linear_11" [label="(1, 56, 56, 384)", style=solid]; +"720 linear_11" -> "721 dropout_7" [label="(1, 56, 56, 96)", style=solid]; +"721 dropout_7" -> "722 layer_norm_4" [label="(1, 56, 56, 96)", style=solid]; +"722 layer_norm_4" -> "723 add_6" [label="(1, 56, 56, 96)", style=solid]; +"723 add_6" -> "724 quantize_per_tensor_default_13" [label="(1, 56, 56, 96)", style=solid]; +"724 quantize_per_tensor_default_13" -> "725 dequantize_per_tensor_default_13" [label=None, style=solid]; +"725 dequantize_per_tensor_default_13" -> "726 pad_2" [label=None, style=solid]; +"726 pad_2" -> "727 slice_25" [label="(1, 56, 56, 96)", style=solid]; +"726 pad_2" -> "730 slice_28" [label="(1, 56, 56, 96)", style=solid]; +"726 pad_2" -> "733 slice_31" [label="(1, 56, 56, 96)", style=solid]; +"726 pad_2" -> "736 slice_34" [label="(1, 56, 56, 96)", style=solid]; +"727 slice_25" -> "728 slice_26" [label="(1, 28, 56, 96)", style=solid]; +"728 slice_26" -> "729 slice_27" [label="(1, 28, 28, 96)", style=solid]; +"729 slice_27" -> "739 cat" [label="(1, 28, 28, 96)", style=solid]; +"730 slice_28" -> "731 slice_29" [label="(1, 28, 56, 96)", style=solid]; +"731 slice_29" -> "732 slice_30" [label="(1, 28, 28, 96)", style=solid]; +"732 slice_30" -> "739 cat" [label="(1, 28, 28, 96)", style=solid]; +"733 slice_31" -> "734 slice_32" [label="(1, 28, 56, 96)", style=solid]; +"734 slice_32" -> "735 slice_33" [label="(1, 28, 28, 96)", style=solid]; +"735 slice_33" -> "739 cat" [label="(1, 28, 28, 96)", style=solid]; +"736 slice_34" -> "737 slice_35" [label="(1, 28, 56, 96)", style=solid]; +"737 slice_35" -> "738 slice_36" [label="(1, 28, 28, 96)", style=solid]; +"738 slice_36" -> "739 cat" [label="(1, 28, 28, 96)", style=solid]; +"739 cat" -> "740 linear_12" [label="(1, 28, 28, 384)", style=solid]; +"740 linear_12" -> "741 layer_norm_5" [label="(1, 28, 28, 192)", style=solid]; +"741 layer_norm_5" -> "742 quantize_per_tensor_default_14" [label="(1, 28, 28, 192)", style=solid]; +"741 layer_norm_5" -> "791 add_8" [label="(1, 28, 28, 192)", style=solid]; +"742 quantize_per_tensor_default_14" -> "743 dequantize_per_tensor_default_14" [label=None, style=solid]; +"743 dequantize_per_tensor_default_14" -> "745 pad_3" [label=None, style=solid]; +"744 _frozen_param17" -> "772 add_7" [label="(1, 6, 64, 64)", style=solid]; +"745 pad_3" -> "746 view_13" [label="(1, 32, 32, 192)", style=solid]; +"746 view_13" -> "747 permute_11" [label="(1, 4, 8, 4, 8, 192)", style=solid]; +"747 permute_11" -> "748 reshape_9" [label="(1, 4, 4, 8, 8, 192)", style=solid]; +"748 reshape_9" -> "750 linear_15" [label="(16, 64, 192)", style=solid]; +"749 _frozen_param18" -> "750 linear_15" [label="(576,)", style=solid]; +"750 linear_15" -> "751 reshape_10" [label="(16, 64, 576)", style=solid]; +"751 reshape_10" -> "752 permute_12" [label="(16, 64, 3, 6, 32)", style=solid]; +"752 permute_12" -> "753 select_6" [label="(3, 16, 6, 64, 32)", style=solid]; +"752 permute_12" -> "754 select_7" [label="(3, 16, 6, 64, 32)", style=solid]; +"752 permute_12" -> "755 select_8" [label="(3, 16, 6, 64, 32)", style=solid]; +"753 select_6" -> "756 linalg_vector_norm_4" [label="(16, 6, 64, 32)", style=solid]; +"753 select_6" -> "758 expand_as_4" [label="(16, 6, 64, 32)", style=solid]; +"753 select_6" -> "759 div_4" [label="(16, 6, 64, 32)", style=solid]; +"754 select_7" -> "762 linalg_vector_norm_5" [label="(16, 6, 64, 32)", style=solid]; +"754 select_7" -> "764 expand_as_5" [label="(16, 6, 64, 32)", style=solid]; +"754 select_7" -> "765 div_5" [label="(16, 6, 64, 32)", style=solid]; +"755 select_8" -> "775 matmul_5" [label="(16, 6, 64, 32)", style=solid]; +"756 linalg_vector_norm_4" -> "757 clamp_min_4" [label="(16, 6, 64, 1)", style=solid]; +"757 clamp_min_4" -> "758 expand_as_4" [label="(16, 6, 64, 1)", style=solid]; +"758 expand_as_4" -> "759 div_4" [label="(16, 6, 64, 32)", style=solid]; +"759 div_4" -> "760 quantize_per_tensor_default_15" [label="(16, 6, 64, 32)", style=solid]; +"760 quantize_per_tensor_default_15" -> "761 dequantize_per_tensor_default_15" [label=None, style=solid]; +"761 dequantize_per_tensor_default_15" -> "769 matmul_4" [label=None, style=solid]; +"762 linalg_vector_norm_5" -> "763 clamp_min_5" [label="(16, 6, 64, 1)", style=solid]; +"763 clamp_min_5" -> "764 expand_as_5" [label="(16, 6, 64, 1)", style=solid]; +"764 expand_as_5" -> "765 div_5" [label="(16, 6, 64, 32)", style=solid]; +"765 div_5" -> "766 quantize_per_tensor_default_16" [label="(16, 6, 64, 32)", style=solid]; +"766 quantize_per_tensor_default_16" -> "767 dequantize_per_tensor_default_16" [label=None, style=solid]; +"767 dequantize_per_tensor_default_16" -> "768 transpose_4" [label=None, style=solid]; +"768 transpose_4" -> "769 matmul_4" [label="(16, 6, 32, 64)", style=solid]; +"769 matmul_4" -> "771 mul_5" [label="(16, 6, 64, 64)", style=solid]; +"770 _frozen_param20" -> "771 mul_5" [label="(6, 1, 1)", style=solid]; +"771 mul_5" -> "772 add_7" [label="(16, 6, 64, 64)", style=solid]; +"772 add_7" -> "773 softmax_2" [label="(16, 6, 64, 64)", style=solid]; +"773 softmax_2" -> "774 dropout_8" [label="(16, 6, 64, 64)", style=solid]; +"774 dropout_8" -> "775 matmul_5" [label="(16, 6, 64, 64)", style=solid]; +"775 matmul_5" -> "776 quantize_per_tensor_default_17" [label="(16, 6, 64, 32)", style=solid]; +"776 quantize_per_tensor_default_17" -> "777 dequantize_per_tensor_default_17" [label=None, style=solid]; +"777 dequantize_per_tensor_default_17" -> "778 transpose_5" [label=None, style=solid]; +"778 transpose_5" -> "779 reshape_11" [label="(16, 64, 6, 32)", style=solid]; +"779 reshape_11" -> "780 linear_16" [label="(16, 64, 192)", style=solid]; +"780 linear_16" -> "781 dropout_9" [label="(16, 64, 192)", style=solid]; +"781 dropout_9" -> "782 view_14" [label="(16, 64, 192)", style=solid]; +"782 view_14" -> "783 permute_13" [label="(1, 4, 4, 8, 8, 192)", style=solid]; +"783 permute_13" -> "784 reshape_12" [label="(1, 4, 8, 4, 8, 192)", style=solid]; +"784 reshape_12" -> "785 slice_38" [label="(1, 32, 32, 192)", style=solid]; +"785 slice_38" -> "786 slice_39" [label="(1, 32, 32, 192)", style=solid]; +"786 slice_39" -> "787 slice_40" [label="(1, 28, 32, 192)", style=solid]; +"787 slice_40" -> "788 slice_41" [label="(1, 28, 28, 192)", style=solid]; +"788 slice_41" -> "789 contiguous_3" [label="(1, 28, 28, 192)", style=solid]; +"789 contiguous_3" -> "790 layer_norm_6" [label="(1, 28, 28, 192)", style=solid]; +"790 layer_norm_6" -> "791 add_8" [label="(1, 28, 28, 192)", style=solid]; +"791 add_8" -> "792 quantize_per_tensor_default_18" [label="(1, 28, 28, 192)", style=solid]; +"791 add_8" -> "802 add_9" [label="(1, 28, 28, 192)", style=solid]; +"792 quantize_per_tensor_default_18" -> "793 dequantize_per_tensor_default_18" [label=None, style=solid]; +"793 dequantize_per_tensor_default_18" -> "794 linear_17" [label=None, style=solid]; +"794 linear_17" -> "795 gelu_2" [label="(1, 28, 28, 768)", style=solid]; +"795 gelu_2" -> "796 quantize_per_tensor_default_19" [label="(1, 28, 28, 768)", style=solid]; +"796 quantize_per_tensor_default_19" -> "797 dequantize_per_tensor_default_19" [label=None, style=solid]; +"797 dequantize_per_tensor_default_19" -> "798 dropout_10" [label=None, style=solid]; +"798 dropout_10" -> "799 linear_18" [label="(1, 28, 28, 768)", style=solid]; +"799 linear_18" -> "800 dropout_11" [label="(1, 28, 28, 192)", style=solid]; +"800 dropout_11" -> "801 layer_norm_7" [label="(1, 28, 28, 192)", style=solid]; +"801 layer_norm_7" -> "802 add_9" [label="(1, 28, 28, 192)", style=solid]; +"802 add_9" -> "804 pad_4" [label="(1, 28, 28, 192)", style=solid]; +"802 add_9" -> "870 add_12" [label="(1, 28, 28, 192)", style=solid]; +"803 _frozen_param21" -> "834 add_10" [label="(1, 6, 64, 64)", style=solid]; +"804 pad_4" -> "805 roll_2" [label="(1, 32, 32, 192)", style=solid]; +"805 roll_2" -> "806 view_17" [label="(1, 32, 32, 192)", style=solid]; +"806 view_17" -> "807 permute_15" [label="(1, 4, 8, 4, 8, 192)", style=solid]; +"807 permute_15" -> "808 reshape_13" [label="(1, 4, 4, 8, 8, 192)", style=solid]; +"808 reshape_13" -> "809 quantize_per_tensor_default_20" [label="(16, 64, 192)", style=solid]; +"808 reshape_13" -> "835 new_zeros_1" [label="(16, 64, 192)", style=solid]; +"809 quantize_per_tensor_default_20" -> "810 dequantize_per_tensor_default_20" [label=None, style=solid]; +"810 dequantize_per_tensor_default_20" -> "812 linear_21" [label=None, style=solid]; +"811 _frozen_param22" -> "812 linear_21" [label="(576,)", style=solid]; +"812 linear_21" -> "813 reshape_14" [label="(16, 64, 576)", style=solid]; +"813 reshape_14" -> "814 permute_16" [label="(16, 64, 3, 6, 32)", style=solid]; +"814 permute_16" -> "815 select_9" [label="(3, 16, 6, 64, 32)", style=solid]; +"814 permute_16" -> "816 select_10" [label="(3, 16, 6, 64, 32)", style=solid]; +"814 permute_16" -> "817 select_11" [label="(3, 16, 6, 64, 32)", style=solid]; +"815 select_9" -> "818 linalg_vector_norm_6" [label="(16, 6, 64, 32)", style=solid]; +"815 select_9" -> "820 expand_as_6" [label="(16, 6, 64, 32)", style=solid]; +"815 select_9" -> "821 div_6" [label="(16, 6, 64, 32)", style=solid]; +"816 select_10" -> "824 linalg_vector_norm_7" [label="(16, 6, 64, 32)", style=solid]; +"816 select_10" -> "826 expand_as_7" [label="(16, 6, 64, 32)", style=solid]; +"816 select_10" -> "827 div_7" [label="(16, 6, 64, 32)", style=solid]; +"817 select_11" -> "853 matmul_7" [label="(16, 6, 64, 32)", style=solid]; +"818 linalg_vector_norm_6" -> "819 clamp_min_6" [label="(16, 6, 64, 1)", style=solid]; +"819 clamp_min_6" -> "820 expand_as_6" [label="(16, 6, 64, 1)", style=solid]; +"820 expand_as_6" -> "821 div_6" [label="(16, 6, 64, 32)", style=solid]; +"821 div_6" -> "822 quantize_per_tensor_default_21" [label="(16, 6, 64, 32)", style=solid]; +"822 quantize_per_tensor_default_21" -> "823 dequantize_per_tensor_default_21" [label=None, style=solid]; +"823 dequantize_per_tensor_default_21" -> "831 matmul_6" [label=None, style=solid]; +"824 linalg_vector_norm_7" -> "825 clamp_min_7" [label="(16, 6, 64, 1)", style=solid]; +"825 clamp_min_7" -> "826 expand_as_7" [label="(16, 6, 64, 1)", style=solid]; +"826 expand_as_7" -> "827 div_7" [label="(16, 6, 64, 32)", style=solid]; +"827 div_7" -> "828 quantize_per_tensor_default_22" [label="(16, 6, 64, 32)", style=solid]; +"828 quantize_per_tensor_default_22" -> "829 dequantize_per_tensor_default_22" [label=None, style=solid]; +"829 dequantize_per_tensor_default_22" -> "830 transpose_6" [label=None, style=solid]; +"830 transpose_6" -> "831 matmul_6" [label="(16, 6, 32, 64)", style=solid]; +"831 matmul_6" -> "833 mul_7" [label="(16, 6, 64, 64)", style=solid]; +"832 _frozen_param24" -> "833 mul_7" [label="(6, 1, 1)", style=solid]; +"833 mul_7" -> "834 add_10" [label="(16, 6, 64, 64)", style=solid]; +"834 add_10" -> "846 view_19" [label="(16, 6, 64, 64)", style=solid]; +"835 new_zeros_1" -> "836 view_18" [label="(32, 32)", style=solid]; +"836 view_18" -> "837 permute_17" [label="(4, 8, 4, 8)", style=solid]; +"837 permute_17" -> "838 reshape_15" [label="(4, 4, 8, 8)", style=solid]; +"838 reshape_15" -> "839 unsqueeze_8" [label="(16, 64)", style=solid]; +"838 reshape_15" -> "840 unsqueeze_9" [label="(16, 64)", style=solid]; +"839 unsqueeze_8" -> "841 sub_1" [label="(16, 1, 64)", style=solid]; +"840 unsqueeze_9" -> "841 sub_1" [label="(16, 64, 1)", style=solid]; +"841 sub_1" -> "842 ne_1" [label="(16, 64, 64)", style=solid]; +"841 sub_1" -> "843 masked_fill_2" [label="(16, 64, 64)", style=solid]; +"841 sub_1" -> "844 eq_1" [label="(16, 64, 64)", style=solid]; +"842 ne_1" -> "843 masked_fill_2" [label="(16, 64, 64)", style=solid]; +"843 masked_fill_2" -> "845 masked_fill_3" [label="(16, 64, 64)", style=solid]; +"844 eq_1" -> "845 masked_fill_3" [label="(16, 64, 64)", style=solid]; +"845 masked_fill_3" -> "847 unsqueeze_10" [label="(16, 64, 64)", style=solid]; +"846 view_19" -> "849 add_11" [label="(1, 16, 6, 64, 64)", style=solid]; +"847 unsqueeze_10" -> "848 unsqueeze_11" [label="(16, 1, 64, 64)", style=solid]; +"848 unsqueeze_11" -> "849 add_11" [label="(1, 16, 1, 64, 64)", style=solid]; +"849 add_11" -> "850 view_20" [label="(1, 16, 6, 64, 64)", style=solid]; +"850 view_20" -> "851 softmax_3" [label="(16, 6, 64, 64)", style=solid]; +"851 softmax_3" -> "852 dropout_12" [label="(16, 6, 64, 64)", style=solid]; +"852 dropout_12" -> "853 matmul_7" [label="(16, 6, 64, 64)", style=solid]; +"853 matmul_7" -> "854 quantize_per_tensor_default_23" [label="(16, 6, 64, 32)", style=solid]; +"854 quantize_per_tensor_default_23" -> "855 dequantize_per_tensor_default_23" [label=None, style=solid]; +"855 dequantize_per_tensor_default_23" -> "856 transpose_7" [label=None, style=solid]; +"856 transpose_7" -> "857 reshape_16" [label="(16, 64, 6, 32)", style=solid]; +"857 reshape_16" -> "858 linear_22" [label="(16, 64, 192)", style=solid]; +"858 linear_22" -> "859 dropout_13" [label="(16, 64, 192)", style=solid]; +"859 dropout_13" -> "860 view_21" [label="(16, 64, 192)", style=solid]; +"860 view_21" -> "861 permute_18" [label="(1, 4, 4, 8, 8, 192)", style=solid]; +"861 permute_18" -> "862 reshape_17" [label="(1, 4, 8, 4, 8, 192)", style=solid]; +"862 reshape_17" -> "863 roll_3" [label="(1, 32, 32, 192)", style=solid]; +"863 roll_3" -> "864 slice_61" [label="(1, 32, 32, 192)", style=solid]; +"864 slice_61" -> "865 slice_62" [label="(1, 32, 32, 192)", style=solid]; +"865 slice_62" -> "866 slice_63" [label="(1, 28, 32, 192)", style=solid]; +"866 slice_63" -> "867 slice_64" [label="(1, 28, 28, 192)", style=solid]; +"867 slice_64" -> "868 contiguous_5" [label="(1, 28, 28, 192)", style=solid]; +"868 contiguous_5" -> "869 layer_norm_8" [label="(1, 28, 28, 192)", style=solid]; +"869 layer_norm_8" -> "870 add_12" [label="(1, 28, 28, 192)", style=solid]; +"870 add_12" -> "871 quantize_per_tensor_default_24" [label="(1, 28, 28, 192)", style=solid]; +"870 add_12" -> "881 add_13" [label="(1, 28, 28, 192)", style=solid]; +"871 quantize_per_tensor_default_24" -> "872 dequantize_per_tensor_default_24" [label=None, style=solid]; +"872 dequantize_per_tensor_default_24" -> "873 linear_23" [label=None, style=solid]; +"873 linear_23" -> "874 gelu_3" [label="(1, 28, 28, 768)", style=solid]; +"874 gelu_3" -> "875 quantize_per_tensor_default_25" [label="(1, 28, 28, 768)", style=solid]; +"875 quantize_per_tensor_default_25" -> "876 dequantize_per_tensor_default_25" [label=None, style=solid]; +"876 dequantize_per_tensor_default_25" -> "877 dropout_14" [label=None, style=solid]; +"877 dropout_14" -> "878 linear_24" [label="(1, 28, 28, 768)", style=solid]; +"878 linear_24" -> "879 dropout_15" [label="(1, 28, 28, 192)", style=solid]; +"879 dropout_15" -> "880 layer_norm_9" [label="(1, 28, 28, 192)", style=solid]; +"880 layer_norm_9" -> "881 add_13" [label="(1, 28, 28, 192)", style=solid]; +"881 add_13" -> "882 quantize_per_tensor_default_26" [label="(1, 28, 28, 192)", style=solid]; +"882 quantize_per_tensor_default_26" -> "883 dequantize_per_tensor_default_26" [label=None, style=solid]; +"883 dequantize_per_tensor_default_26" -> "884 pad_5" [label=None, style=solid]; +"884 pad_5" -> "885 slice_65" [label="(1, 28, 28, 192)", style=solid]; +"884 pad_5" -> "888 slice_68" [label="(1, 28, 28, 192)", style=solid]; +"884 pad_5" -> "891 slice_71" [label="(1, 28, 28, 192)", style=solid]; +"884 pad_5" -> "894 slice_74" [label="(1, 28, 28, 192)", style=solid]; +"885 slice_65" -> "886 slice_66" [label="(1, 14, 28, 192)", style=solid]; +"886 slice_66" -> "887 slice_67" [label="(1, 14, 14, 192)", style=solid]; +"887 slice_67" -> "897 cat_1" [label="(1, 14, 14, 192)", style=solid]; +"888 slice_68" -> "889 slice_69" [label="(1, 14, 28, 192)", style=solid]; +"889 slice_69" -> "890 slice_70" [label="(1, 14, 14, 192)", style=solid]; +"890 slice_70" -> "897 cat_1" [label="(1, 14, 14, 192)", style=solid]; +"891 slice_71" -> "892 slice_72" [label="(1, 14, 28, 192)", style=solid]; +"892 slice_72" -> "893 slice_73" [label="(1, 14, 14, 192)", style=solid]; +"893 slice_73" -> "897 cat_1" [label="(1, 14, 14, 192)", style=solid]; +"894 slice_74" -> "895 slice_75" [label="(1, 14, 28, 192)", style=solid]; +"895 slice_75" -> "896 slice_76" [label="(1, 14, 14, 192)", style=solid]; +"896 slice_76" -> "897 cat_1" [label="(1, 14, 14, 192)", style=solid]; +"897 cat_1" -> "898 linear_25" [label="(1, 14, 14, 768)", style=solid]; +"898 linear_25" -> "899 layer_norm_10" [label="(1, 14, 14, 384)", style=solid]; +"899 layer_norm_10" -> "900 quantize_per_tensor_default_27" [label="(1, 14, 14, 384)", style=solid]; +"899 layer_norm_10" -> "949 add_15" [label="(1, 14, 14, 384)", style=solid]; +"900 quantize_per_tensor_default_27" -> "901 dequantize_per_tensor_default_27" [label=None, style=solid]; +"901 dequantize_per_tensor_default_27" -> "903 pad_6" [label=None, style=solid]; +"902 _frozen_param34" -> "930 add_14" [label="(1, 12, 64, 64)", style=solid]; +"903 pad_6" -> "904 view_24" [label="(1, 16, 16, 384)", style=solid]; +"904 view_24" -> "905 permute_20" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"905 permute_20" -> "906 reshape_18" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"906 reshape_18" -> "908 linear_28" [label="(4, 64, 384)", style=solid]; +"907 _frozen_param35" -> "908 linear_28" [label="(1152,)", style=solid]; +"908 linear_28" -> "909 reshape_19" [label="(4, 64, 1152)", style=solid]; +"909 reshape_19" -> "910 permute_21" [label="(4, 64, 3, 12, 32)", style=solid]; +"910 permute_21" -> "911 select_12" [label="(3, 4, 12, 64, 32)", style=solid]; +"910 permute_21" -> "912 select_13" [label="(3, 4, 12, 64, 32)", style=solid]; +"910 permute_21" -> "913 select_14" [label="(3, 4, 12, 64, 32)", style=solid]; +"911 select_12" -> "914 linalg_vector_norm_8" [label="(4, 12, 64, 32)", style=solid]; +"911 select_12" -> "916 expand_as_8" [label="(4, 12, 64, 32)", style=solid]; +"911 select_12" -> "917 div_8" [label="(4, 12, 64, 32)", style=solid]; +"912 select_13" -> "920 linalg_vector_norm_9" [label="(4, 12, 64, 32)", style=solid]; +"912 select_13" -> "922 expand_as_9" [label="(4, 12, 64, 32)", style=solid]; +"912 select_13" -> "923 div_9" [label="(4, 12, 64, 32)", style=solid]; +"913 select_14" -> "933 matmul_9" [label="(4, 12, 64, 32)", style=solid]; +"914 linalg_vector_norm_8" -> "915 clamp_min_8" [label="(4, 12, 64, 1)", style=solid]; +"915 clamp_min_8" -> "916 expand_as_8" [label="(4, 12, 64, 1)", style=solid]; +"916 expand_as_8" -> "917 div_8" [label="(4, 12, 64, 32)", style=solid]; +"917 div_8" -> "918 quantize_per_tensor_default_28" [label="(4, 12, 64, 32)", style=solid]; +"918 quantize_per_tensor_default_28" -> "919 dequantize_per_tensor_default_28" [label=None, style=solid]; +"919 dequantize_per_tensor_default_28" -> "927 matmul_8" [label=None, style=solid]; +"920 linalg_vector_norm_9" -> "921 clamp_min_9" [label="(4, 12, 64, 1)", style=solid]; +"921 clamp_min_9" -> "922 expand_as_9" [label="(4, 12, 64, 1)", style=solid]; +"922 expand_as_9" -> "923 div_9" [label="(4, 12, 64, 32)", style=solid]; +"923 div_9" -> "924 quantize_per_tensor_default_29" [label="(4, 12, 64, 32)", style=solid]; +"924 quantize_per_tensor_default_29" -> "925 dequantize_per_tensor_default_29" [label=None, style=solid]; +"925 dequantize_per_tensor_default_29" -> "926 transpose_8" [label=None, style=solid]; +"926 transpose_8" -> "927 matmul_8" [label="(4, 12, 32, 64)", style=solid]; +"927 matmul_8" -> "929 mul_9" [label="(4, 12, 64, 64)", style=solid]; +"928 _frozen_param37" -> "929 mul_9" [label="(12, 1, 1)", style=solid]; +"929 mul_9" -> "930 add_14" [label="(4, 12, 64, 64)", style=solid]; +"930 add_14" -> "931 softmax_4" [label="(4, 12, 64, 64)", style=solid]; +"931 softmax_4" -> "932 dropout_16" [label="(4, 12, 64, 64)", style=solid]; +"932 dropout_16" -> "933 matmul_9" [label="(4, 12, 64, 64)", style=solid]; +"933 matmul_9" -> "934 quantize_per_tensor_default_30" [label="(4, 12, 64, 32)", style=solid]; +"934 quantize_per_tensor_default_30" -> "935 dequantize_per_tensor_default_30" [label=None, style=solid]; +"935 dequantize_per_tensor_default_30" -> "936 transpose_9" [label=None, style=solid]; +"936 transpose_9" -> "937 reshape_20" [label="(4, 64, 12, 32)", style=solid]; +"937 reshape_20" -> "938 linear_29" [label="(4, 64, 384)", style=solid]; +"938 linear_29" -> "939 dropout_17" [label="(4, 64, 384)", style=solid]; +"939 dropout_17" -> "940 view_25" [label="(4, 64, 384)", style=solid]; +"940 view_25" -> "941 permute_22" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"941 permute_22" -> "942 reshape_21" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"942 reshape_21" -> "943 slice_78" [label="(1, 16, 16, 384)", style=solid]; +"943 slice_78" -> "944 slice_79" [label="(1, 16, 16, 384)", style=solid]; +"944 slice_79" -> "945 slice_80" [label="(1, 14, 16, 384)", style=solid]; +"945 slice_80" -> "946 slice_81" [label="(1, 14, 14, 384)", style=solid]; +"946 slice_81" -> "947 contiguous_7" [label="(1, 14, 14, 384)", style=solid]; +"947 contiguous_7" -> "948 layer_norm_11" [label="(1, 14, 14, 384)", style=solid]; +"948 layer_norm_11" -> "949 add_15" [label="(1, 14, 14, 384)", style=solid]; +"949 add_15" -> "950 quantize_per_tensor_default_31" [label="(1, 14, 14, 384)", style=solid]; +"949 add_15" -> "960 add_16" [label="(1, 14, 14, 384)", style=solid]; +"950 quantize_per_tensor_default_31" -> "951 dequantize_per_tensor_default_31" [label=None, style=solid]; +"951 dequantize_per_tensor_default_31" -> "952 linear_30" [label=None, style=solid]; +"952 linear_30" -> "953 gelu_4" [label="(1, 14, 14, 1536)", style=solid]; +"953 gelu_4" -> "954 quantize_per_tensor_default_32" [label="(1, 14, 14, 1536)", style=solid]; +"954 quantize_per_tensor_default_32" -> "955 dequantize_per_tensor_default_32" [label=None, style=solid]; +"955 dequantize_per_tensor_default_32" -> "956 dropout_18" [label=None, style=solid]; +"956 dropout_18" -> "957 linear_31" [label="(1, 14, 14, 1536)", style=solid]; +"957 linear_31" -> "958 dropout_19" [label="(1, 14, 14, 384)", style=solid]; +"958 dropout_19" -> "959 layer_norm_12" [label="(1, 14, 14, 384)", style=solid]; +"959 layer_norm_12" -> "960 add_16" [label="(1, 14, 14, 384)", style=solid]; +"960 add_16" -> "962 pad_7" [label="(1, 14, 14, 384)", style=solid]; +"960 add_16" -> "1028 add_19" [label="(1, 14, 14, 384)", style=solid]; +"961 _frozen_param38" -> "992 add_17" [label="(1, 12, 64, 64)", style=solid]; +"962 pad_7" -> "963 roll_4" [label="(1, 16, 16, 384)", style=solid]; +"963 roll_4" -> "964 view_28" [label="(1, 16, 16, 384)", style=solid]; +"964 view_28" -> "965 permute_24" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"965 permute_24" -> "966 reshape_22" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"966 reshape_22" -> "967 quantize_per_tensor_default_33" [label="(4, 64, 384)", style=solid]; +"966 reshape_22" -> "993 new_zeros_2" [label="(4, 64, 384)", style=solid]; +"967 quantize_per_tensor_default_33" -> "968 dequantize_per_tensor_default_33" [label=None, style=solid]; +"968 dequantize_per_tensor_default_33" -> "970 linear_34" [label=None, style=solid]; +"969 _frozen_param39" -> "970 linear_34" [label="(1152,)", style=solid]; +"970 linear_34" -> "971 reshape_23" [label="(4, 64, 1152)", style=solid]; +"971 reshape_23" -> "972 permute_25" [label="(4, 64, 3, 12, 32)", style=solid]; +"972 permute_25" -> "973 select_15" [label="(3, 4, 12, 64, 32)", style=solid]; +"972 permute_25" -> "974 select_16" [label="(3, 4, 12, 64, 32)", style=solid]; +"972 permute_25" -> "975 select_17" [label="(3, 4, 12, 64, 32)", style=solid]; +"973 select_15" -> "976 linalg_vector_norm_10" [label="(4, 12, 64, 32)", style=solid]; +"973 select_15" -> "978 expand_as_10" [label="(4, 12, 64, 32)", style=solid]; +"973 select_15" -> "979 div_10" [label="(4, 12, 64, 32)", style=solid]; +"974 select_16" -> "982 linalg_vector_norm_11" [label="(4, 12, 64, 32)", style=solid]; +"974 select_16" -> "984 expand_as_11" [label="(4, 12, 64, 32)", style=solid]; +"974 select_16" -> "985 div_11" [label="(4, 12, 64, 32)", style=solid]; +"975 select_17" -> "1011 matmul_11" [label="(4, 12, 64, 32)", style=solid]; +"976 linalg_vector_norm_10" -> "977 clamp_min_10" [label="(4, 12, 64, 1)", style=solid]; +"977 clamp_min_10" -> "978 expand_as_10" [label="(4, 12, 64, 1)", style=solid]; +"978 expand_as_10" -> "979 div_10" [label="(4, 12, 64, 32)", style=solid]; +"979 div_10" -> "980 quantize_per_tensor_default_34" [label="(4, 12, 64, 32)", style=solid]; +"980 quantize_per_tensor_default_34" -> "981 dequantize_per_tensor_default_34" [label=None, style=solid]; +"981 dequantize_per_tensor_default_34" -> "989 matmul_10" [label=None, style=solid]; +"982 linalg_vector_norm_11" -> "983 clamp_min_11" [label="(4, 12, 64, 1)", style=solid]; +"983 clamp_min_11" -> "984 expand_as_11" [label="(4, 12, 64, 1)", style=solid]; +"984 expand_as_11" -> "985 div_11" [label="(4, 12, 64, 32)", style=solid]; +"985 div_11" -> "986 quantize_per_tensor_default_35" [label="(4, 12, 64, 32)", style=solid]; +"986 quantize_per_tensor_default_35" -> "987 dequantize_per_tensor_default_35" [label=None, style=solid]; +"987 dequantize_per_tensor_default_35" -> "988 transpose_10" [label=None, style=solid]; +"988 transpose_10" -> "989 matmul_10" [label="(4, 12, 32, 64)", style=solid]; +"989 matmul_10" -> "991 mul_11" [label="(4, 12, 64, 64)", style=solid]; +"990 _frozen_param41" -> "991 mul_11" [label="(12, 1, 1)", style=solid]; +"991 mul_11" -> "992 add_17" [label="(4, 12, 64, 64)", style=solid]; +"992 add_17" -> "1004 view_30" [label="(4, 12, 64, 64)", style=solid]; +"993 new_zeros_2" -> "994 view_29" [label="(16, 16)", style=solid]; +"994 view_29" -> "995 permute_26" [label="(2, 8, 2, 8)", style=solid]; +"995 permute_26" -> "996 reshape_24" [label="(2, 2, 8, 8)", style=solid]; +"996 reshape_24" -> "997 unsqueeze_14" [label="(4, 64)", style=solid]; +"996 reshape_24" -> "998 unsqueeze_15" [label="(4, 64)", style=solid]; +"997 unsqueeze_14" -> "999 sub_2" [label="(4, 1, 64)", style=solid]; +"998 unsqueeze_15" -> "999 sub_2" [label="(4, 64, 1)", style=solid]; +"999 sub_2" -> "1000 ne_2" [label="(4, 64, 64)", style=solid]; +"999 sub_2" -> "1001 masked_fill_4" [label="(4, 64, 64)", style=solid]; +"999 sub_2" -> "1002 eq_2" [label="(4, 64, 64)", style=solid]; +"1000 ne_2" -> "1001 masked_fill_4" [label="(4, 64, 64)", style=solid]; +"1001 masked_fill_4" -> "1003 masked_fill_5" [label="(4, 64, 64)", style=solid]; +"1002 eq_2" -> "1003 masked_fill_5" [label="(4, 64, 64)", style=solid]; +"1003 masked_fill_5" -> "1005 unsqueeze_16" [label="(4, 64, 64)", style=solid]; +"1004 view_30" -> "1007 add_18" [label="(1, 4, 12, 64, 64)", style=solid]; +"1005 unsqueeze_16" -> "1006 unsqueeze_17" [label="(4, 1, 64, 64)", style=solid]; +"1006 unsqueeze_17" -> "1007 add_18" [label="(1, 4, 1, 64, 64)", style=solid]; +"1007 add_18" -> "1008 view_31" [label="(1, 4, 12, 64, 64)", style=solid]; +"1008 view_31" -> "1009 softmax_5" [label="(4, 12, 64, 64)", style=solid]; +"1009 softmax_5" -> "1010 dropout_20" [label="(4, 12, 64, 64)", style=solid]; +"1010 dropout_20" -> "1011 matmul_11" [label="(4, 12, 64, 64)", style=solid]; +"1011 matmul_11" -> "1012 quantize_per_tensor_default_36" [label="(4, 12, 64, 32)", style=solid]; +"1012 quantize_per_tensor_default_36" -> "1013 dequantize_per_tensor_default_36" [label=None, style=solid]; +"1013 dequantize_per_tensor_default_36" -> "1014 transpose_11" [label=None, style=solid]; +"1014 transpose_11" -> "1015 reshape_25" [label="(4, 64, 12, 32)", style=solid]; +"1015 reshape_25" -> "1016 linear_35" [label="(4, 64, 384)", style=solid]; +"1016 linear_35" -> "1017 dropout_21" [label="(4, 64, 384)", style=solid]; +"1017 dropout_21" -> "1018 view_32" [label="(4, 64, 384)", style=solid]; +"1018 view_32" -> "1019 permute_27" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1019 permute_27" -> "1020 reshape_26" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1020 reshape_26" -> "1021 roll_5" [label="(1, 16, 16, 384)", style=solid]; +"1021 roll_5" -> "1022 slice_101" [label="(1, 16, 16, 384)", style=solid]; +"1022 slice_101" -> "1023 slice_102" [label="(1, 16, 16, 384)", style=solid]; +"1023 slice_102" -> "1024 slice_103" [label="(1, 14, 16, 384)", style=solid]; +"1024 slice_103" -> "1025 slice_104" [label="(1, 14, 14, 384)", style=solid]; +"1025 slice_104" -> "1026 contiguous_9" [label="(1, 14, 14, 384)", style=solid]; +"1026 contiguous_9" -> "1027 layer_norm_13" [label="(1, 14, 14, 384)", style=solid]; +"1027 layer_norm_13" -> "1028 add_19" [label="(1, 14, 14, 384)", style=solid]; +"1028 add_19" -> "1029 quantize_per_tensor_default_37" [label="(1, 14, 14, 384)", style=solid]; +"1028 add_19" -> "1039 add_20" [label="(1, 14, 14, 384)", style=solid]; +"1029 quantize_per_tensor_default_37" -> "1030 dequantize_per_tensor_default_37" [label=None, style=solid]; +"1030 dequantize_per_tensor_default_37" -> "1031 linear_36" [label=None, style=solid]; +"1031 linear_36" -> "1032 gelu_5" [label="(1, 14, 14, 1536)", style=solid]; +"1032 gelu_5" -> "1033 quantize_per_tensor_default_38" [label="(1, 14, 14, 1536)", style=solid]; +"1033 quantize_per_tensor_default_38" -> "1034 dequantize_per_tensor_default_38" [label=None, style=solid]; +"1034 dequantize_per_tensor_default_38" -> "1035 dropout_22" [label=None, style=solid]; +"1035 dropout_22" -> "1036 linear_37" [label="(1, 14, 14, 1536)", style=solid]; +"1036 linear_37" -> "1037 dropout_23" [label="(1, 14, 14, 384)", style=solid]; +"1037 dropout_23" -> "1038 layer_norm_14" [label="(1, 14, 14, 384)", style=solid]; +"1038 layer_norm_14" -> "1039 add_20" [label="(1, 14, 14, 384)", style=solid]; +"1039 add_20" -> "1040 quantize_per_tensor_default_39" [label="(1, 14, 14, 384)", style=solid]; +"1039 add_20" -> "1089 add_22" [label="(1, 14, 14, 384)", style=solid]; +"1040 quantize_per_tensor_default_39" -> "1041 dequantize_per_tensor_default_39" [label=None, style=solid]; +"1041 dequantize_per_tensor_default_39" -> "1043 pad_8" [label=None, style=solid]; +"1042 _frozen_param51" -> "1070 add_21" [label="(1, 12, 64, 64)", style=solid]; +"1043 pad_8" -> "1044 view_35" [label="(1, 16, 16, 384)", style=solid]; +"1044 view_35" -> "1045 permute_29" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1045 permute_29" -> "1046 reshape_27" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1046 reshape_27" -> "1048 linear_40" [label="(4, 64, 384)", style=solid]; +"1047 _frozen_param52" -> "1048 linear_40" [label="(1152,)", style=solid]; +"1048 linear_40" -> "1049 reshape_28" [label="(4, 64, 1152)", style=solid]; +"1049 reshape_28" -> "1050 permute_30" [label="(4, 64, 3, 12, 32)", style=solid]; +"1050 permute_30" -> "1051 select_18" [label="(3, 4, 12, 64, 32)", style=solid]; +"1050 permute_30" -> "1052 select_19" [label="(3, 4, 12, 64, 32)", style=solid]; +"1050 permute_30" -> "1053 select_20" [label="(3, 4, 12, 64, 32)", style=solid]; +"1051 select_18" -> "1054 linalg_vector_norm_12" [label="(4, 12, 64, 32)", style=solid]; +"1051 select_18" -> "1056 expand_as_12" [label="(4, 12, 64, 32)", style=solid]; +"1051 select_18" -> "1057 div_12" [label="(4, 12, 64, 32)", style=solid]; +"1052 select_19" -> "1060 linalg_vector_norm_13" [label="(4, 12, 64, 32)", style=solid]; +"1052 select_19" -> "1062 expand_as_13" [label="(4, 12, 64, 32)", style=solid]; +"1052 select_19" -> "1063 div_13" [label="(4, 12, 64, 32)", style=solid]; +"1053 select_20" -> "1073 matmul_13" [label="(4, 12, 64, 32)", style=solid]; +"1054 linalg_vector_norm_12" -> "1055 clamp_min_12" [label="(4, 12, 64, 1)", style=solid]; +"1055 clamp_min_12" -> "1056 expand_as_12" [label="(4, 12, 64, 1)", style=solid]; +"1056 expand_as_12" -> "1057 div_12" [label="(4, 12, 64, 32)", style=solid]; +"1057 div_12" -> "1058 quantize_per_tensor_default_40" [label="(4, 12, 64, 32)", style=solid]; +"1058 quantize_per_tensor_default_40" -> "1059 dequantize_per_tensor_default_40" [label=None, style=solid]; +"1059 dequantize_per_tensor_default_40" -> "1067 matmul_12" [label=None, style=solid]; +"1060 linalg_vector_norm_13" -> "1061 clamp_min_13" [label="(4, 12, 64, 1)", style=solid]; +"1061 clamp_min_13" -> "1062 expand_as_13" [label="(4, 12, 64, 1)", style=solid]; +"1062 expand_as_13" -> "1063 div_13" [label="(4, 12, 64, 32)", style=solid]; +"1063 div_13" -> "1064 quantize_per_tensor_default_41" [label="(4, 12, 64, 32)", style=solid]; +"1064 quantize_per_tensor_default_41" -> "1065 dequantize_per_tensor_default_41" [label=None, style=solid]; +"1065 dequantize_per_tensor_default_41" -> "1066 transpose_12" [label=None, style=solid]; +"1066 transpose_12" -> "1067 matmul_12" [label="(4, 12, 32, 64)", style=solid]; +"1067 matmul_12" -> "1069 mul_13" [label="(4, 12, 64, 64)", style=solid]; +"1068 _frozen_param54" -> "1069 mul_13" [label="(12, 1, 1)", style=solid]; +"1069 mul_13" -> "1070 add_21" [label="(4, 12, 64, 64)", style=solid]; +"1070 add_21" -> "1071 softmax_6" [label="(4, 12, 64, 64)", style=solid]; +"1071 softmax_6" -> "1072 dropout_24" [label="(4, 12, 64, 64)", style=solid]; +"1072 dropout_24" -> "1073 matmul_13" [label="(4, 12, 64, 64)", style=solid]; +"1073 matmul_13" -> "1074 quantize_per_tensor_default_42" [label="(4, 12, 64, 32)", style=solid]; +"1074 quantize_per_tensor_default_42" -> "1075 dequantize_per_tensor_default_42" [label=None, style=solid]; +"1075 dequantize_per_tensor_default_42" -> "1076 transpose_13" [label=None, style=solid]; +"1076 transpose_13" -> "1077 reshape_29" [label="(4, 64, 12, 32)", style=solid]; +"1077 reshape_29" -> "1078 linear_41" [label="(4, 64, 384)", style=solid]; +"1078 linear_41" -> "1079 dropout_25" [label="(4, 64, 384)", style=solid]; +"1079 dropout_25" -> "1080 view_36" [label="(4, 64, 384)", style=solid]; +"1080 view_36" -> "1081 permute_31" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1081 permute_31" -> "1082 reshape_30" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1082 reshape_30" -> "1083 slice_106" [label="(1, 16, 16, 384)", style=solid]; +"1083 slice_106" -> "1084 slice_107" [label="(1, 16, 16, 384)", style=solid]; +"1084 slice_107" -> "1085 slice_108" [label="(1, 14, 16, 384)", style=solid]; +"1085 slice_108" -> "1086 slice_109" [label="(1, 14, 14, 384)", style=solid]; +"1086 slice_109" -> "1087 contiguous_11" [label="(1, 14, 14, 384)", style=solid]; +"1087 contiguous_11" -> "1088 layer_norm_15" [label="(1, 14, 14, 384)", style=solid]; +"1088 layer_norm_15" -> "1089 add_22" [label="(1, 14, 14, 384)", style=solid]; +"1089 add_22" -> "1090 quantize_per_tensor_default_43" [label="(1, 14, 14, 384)", style=solid]; +"1089 add_22" -> "1100 add_23" [label="(1, 14, 14, 384)", style=solid]; +"1090 quantize_per_tensor_default_43" -> "1091 dequantize_per_tensor_default_43" [label=None, style=solid]; +"1091 dequantize_per_tensor_default_43" -> "1092 linear_42" [label=None, style=solid]; +"1092 linear_42" -> "1093 gelu_6" [label="(1, 14, 14, 1536)", style=solid]; +"1093 gelu_6" -> "1094 quantize_per_tensor_default_44" [label="(1, 14, 14, 1536)", style=solid]; +"1094 quantize_per_tensor_default_44" -> "1095 dequantize_per_tensor_default_44" [label=None, style=solid]; +"1095 dequantize_per_tensor_default_44" -> "1096 dropout_26" [label=None, style=solid]; +"1096 dropout_26" -> "1097 linear_43" [label="(1, 14, 14, 1536)", style=solid]; +"1097 linear_43" -> "1098 dropout_27" [label="(1, 14, 14, 384)", style=solid]; +"1098 dropout_27" -> "1099 layer_norm_16" [label="(1, 14, 14, 384)", style=solid]; +"1099 layer_norm_16" -> "1100 add_23" [label="(1, 14, 14, 384)", style=solid]; +"1100 add_23" -> "1102 pad_9" [label="(1, 14, 14, 384)", style=solid]; +"1100 add_23" -> "1168 add_26" [label="(1, 14, 14, 384)", style=solid]; +"1101 _frozen_param55" -> "1132 add_24" [label="(1, 12, 64, 64)", style=solid]; +"1102 pad_9" -> "1103 roll_6" [label="(1, 16, 16, 384)", style=solid]; +"1103 roll_6" -> "1104 view_39" [label="(1, 16, 16, 384)", style=solid]; +"1104 view_39" -> "1105 permute_33" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1105 permute_33" -> "1106 reshape_31" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1106 reshape_31" -> "1107 quantize_per_tensor_default_45" [label="(4, 64, 384)", style=solid]; +"1106 reshape_31" -> "1133 new_zeros_3" [label="(4, 64, 384)", style=solid]; +"1107 quantize_per_tensor_default_45" -> "1108 dequantize_per_tensor_default_45" [label=None, style=solid]; +"1108 dequantize_per_tensor_default_45" -> "1110 linear_46" [label=None, style=solid]; +"1109 _frozen_param56" -> "1110 linear_46" [label="(1152,)", style=solid]; +"1110 linear_46" -> "1111 reshape_32" [label="(4, 64, 1152)", style=solid]; +"1111 reshape_32" -> "1112 permute_34" [label="(4, 64, 3, 12, 32)", style=solid]; +"1112 permute_34" -> "1113 select_21" [label="(3, 4, 12, 64, 32)", style=solid]; +"1112 permute_34" -> "1114 select_22" [label="(3, 4, 12, 64, 32)", style=solid]; +"1112 permute_34" -> "1115 select_23" [label="(3, 4, 12, 64, 32)", style=solid]; +"1113 select_21" -> "1116 linalg_vector_norm_14" [label="(4, 12, 64, 32)", style=solid]; +"1113 select_21" -> "1118 expand_as_14" [label="(4, 12, 64, 32)", style=solid]; +"1113 select_21" -> "1119 div_14" [label="(4, 12, 64, 32)", style=solid]; +"1114 select_22" -> "1122 linalg_vector_norm_15" [label="(4, 12, 64, 32)", style=solid]; +"1114 select_22" -> "1124 expand_as_15" [label="(4, 12, 64, 32)", style=solid]; +"1114 select_22" -> "1125 div_15" [label="(4, 12, 64, 32)", style=solid]; +"1115 select_23" -> "1151 matmul_15" [label="(4, 12, 64, 32)", style=solid]; +"1116 linalg_vector_norm_14" -> "1117 clamp_min_14" [label="(4, 12, 64, 1)", style=solid]; +"1117 clamp_min_14" -> "1118 expand_as_14" [label="(4, 12, 64, 1)", style=solid]; +"1118 expand_as_14" -> "1119 div_14" [label="(4, 12, 64, 32)", style=solid]; +"1119 div_14" -> "1120 quantize_per_tensor_default_46" [label="(4, 12, 64, 32)", style=solid]; +"1120 quantize_per_tensor_default_46" -> "1121 dequantize_per_tensor_default_46" [label=None, style=solid]; +"1121 dequantize_per_tensor_default_46" -> "1129 matmul_14" [label=None, style=solid]; +"1122 linalg_vector_norm_15" -> "1123 clamp_min_15" [label="(4, 12, 64, 1)", style=solid]; +"1123 clamp_min_15" -> "1124 expand_as_15" [label="(4, 12, 64, 1)", style=solid]; +"1124 expand_as_15" -> "1125 div_15" [label="(4, 12, 64, 32)", style=solid]; +"1125 div_15" -> "1126 quantize_per_tensor_default_47" [label="(4, 12, 64, 32)", style=solid]; +"1126 quantize_per_tensor_default_47" -> "1127 dequantize_per_tensor_default_47" [label=None, style=solid]; +"1127 dequantize_per_tensor_default_47" -> "1128 transpose_14" [label=None, style=solid]; +"1128 transpose_14" -> "1129 matmul_14" [label="(4, 12, 32, 64)", style=solid]; +"1129 matmul_14" -> "1131 mul_15" [label="(4, 12, 64, 64)", style=solid]; +"1130 _frozen_param58" -> "1131 mul_15" [label="(12, 1, 1)", style=solid]; +"1131 mul_15" -> "1132 add_24" [label="(4, 12, 64, 64)", style=solid]; +"1132 add_24" -> "1144 view_41" [label="(4, 12, 64, 64)", style=solid]; +"1133 new_zeros_3" -> "1134 view_40" [label="(16, 16)", style=solid]; +"1134 view_40" -> "1135 permute_35" [label="(2, 8, 2, 8)", style=solid]; +"1135 permute_35" -> "1136 reshape_33" [label="(2, 2, 8, 8)", style=solid]; +"1136 reshape_33" -> "1137 unsqueeze_20" [label="(4, 64)", style=solid]; +"1136 reshape_33" -> "1138 unsqueeze_21" [label="(4, 64)", style=solid]; +"1137 unsqueeze_20" -> "1139 sub_3" [label="(4, 1, 64)", style=solid]; +"1138 unsqueeze_21" -> "1139 sub_3" [label="(4, 64, 1)", style=solid]; +"1139 sub_3" -> "1140 ne_3" [label="(4, 64, 64)", style=solid]; +"1139 sub_3" -> "1141 masked_fill_6" [label="(4, 64, 64)", style=solid]; +"1139 sub_3" -> "1142 eq_3" [label="(4, 64, 64)", style=solid]; +"1140 ne_3" -> "1141 masked_fill_6" [label="(4, 64, 64)", style=solid]; +"1141 masked_fill_6" -> "1143 masked_fill_7" [label="(4, 64, 64)", style=solid]; +"1142 eq_3" -> "1143 masked_fill_7" [label="(4, 64, 64)", style=solid]; +"1143 masked_fill_7" -> "1145 unsqueeze_22" [label="(4, 64, 64)", style=solid]; +"1144 view_41" -> "1147 add_25" [label="(1, 4, 12, 64, 64)", style=solid]; +"1145 unsqueeze_22" -> "1146 unsqueeze_23" [label="(4, 1, 64, 64)", style=solid]; +"1146 unsqueeze_23" -> "1147 add_25" [label="(1, 4, 1, 64, 64)", style=solid]; +"1147 add_25" -> "1148 view_42" [label="(1, 4, 12, 64, 64)", style=solid]; +"1148 view_42" -> "1149 softmax_7" [label="(4, 12, 64, 64)", style=solid]; +"1149 softmax_7" -> "1150 dropout_28" [label="(4, 12, 64, 64)", style=solid]; +"1150 dropout_28" -> "1151 matmul_15" [label="(4, 12, 64, 64)", style=solid]; +"1151 matmul_15" -> "1152 quantize_per_tensor_default_48" [label="(4, 12, 64, 32)", style=solid]; +"1152 quantize_per_tensor_default_48" -> "1153 dequantize_per_tensor_default_48" [label=None, style=solid]; +"1153 dequantize_per_tensor_default_48" -> "1154 transpose_15" [label=None, style=solid]; +"1154 transpose_15" -> "1155 reshape_34" [label="(4, 64, 12, 32)", style=solid]; +"1155 reshape_34" -> "1156 linear_47" [label="(4, 64, 384)", style=solid]; +"1156 linear_47" -> "1157 dropout_29" [label="(4, 64, 384)", style=solid]; +"1157 dropout_29" -> "1158 view_43" [label="(4, 64, 384)", style=solid]; +"1158 view_43" -> "1159 permute_36" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1159 permute_36" -> "1160 reshape_35" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1160 reshape_35" -> "1161 roll_7" [label="(1, 16, 16, 384)", style=solid]; +"1161 roll_7" -> "1162 slice_129" [label="(1, 16, 16, 384)", style=solid]; +"1162 slice_129" -> "1163 slice_130" [label="(1, 16, 16, 384)", style=solid]; +"1163 slice_130" -> "1164 slice_131" [label="(1, 14, 16, 384)", style=solid]; +"1164 slice_131" -> "1165 slice_132" [label="(1, 14, 14, 384)", style=solid]; +"1165 slice_132" -> "1166 contiguous_13" [label="(1, 14, 14, 384)", style=solid]; +"1166 contiguous_13" -> "1167 layer_norm_17" [label="(1, 14, 14, 384)", style=solid]; +"1167 layer_norm_17" -> "1168 add_26" [label="(1, 14, 14, 384)", style=solid]; +"1168 add_26" -> "1169 quantize_per_tensor_default_49" [label="(1, 14, 14, 384)", style=solid]; +"1168 add_26" -> "1179 add_27" [label="(1, 14, 14, 384)", style=solid]; +"1169 quantize_per_tensor_default_49" -> "1170 dequantize_per_tensor_default_49" [label=None, style=solid]; +"1170 dequantize_per_tensor_default_49" -> "1171 linear_48" [label=None, style=solid]; +"1171 linear_48" -> "1172 gelu_7" [label="(1, 14, 14, 1536)", style=solid]; +"1172 gelu_7" -> "1173 quantize_per_tensor_default_50" [label="(1, 14, 14, 1536)", style=solid]; +"1173 quantize_per_tensor_default_50" -> "1174 dequantize_per_tensor_default_50" [label=None, style=solid]; +"1174 dequantize_per_tensor_default_50" -> "1175 dropout_30" [label=None, style=solid]; +"1175 dropout_30" -> "1176 linear_49" [label="(1, 14, 14, 1536)", style=solid]; +"1176 linear_49" -> "1177 dropout_31" [label="(1, 14, 14, 384)", style=solid]; +"1177 dropout_31" -> "1178 layer_norm_18" [label="(1, 14, 14, 384)", style=solid]; +"1178 layer_norm_18" -> "1179 add_27" [label="(1, 14, 14, 384)", style=solid]; +"1179 add_27" -> "1180 quantize_per_tensor_default_51" [label="(1, 14, 14, 384)", style=solid]; +"1179 add_27" -> "1229 add_29" [label="(1, 14, 14, 384)", style=solid]; +"1180 quantize_per_tensor_default_51" -> "1181 dequantize_per_tensor_default_51" [label=None, style=solid]; +"1181 dequantize_per_tensor_default_51" -> "1183 pad_10" [label=None, style=solid]; +"1182 _frozen_param68" -> "1210 add_28" [label="(1, 12, 64, 64)", style=solid]; +"1183 pad_10" -> "1184 view_46" [label="(1, 16, 16, 384)", style=solid]; +"1184 view_46" -> "1185 permute_38" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1185 permute_38" -> "1186 reshape_36" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1186 reshape_36" -> "1188 linear_52" [label="(4, 64, 384)", style=solid]; +"1187 _frozen_param69" -> "1188 linear_52" [label="(1152,)", style=solid]; +"1188 linear_52" -> "1189 reshape_37" [label="(4, 64, 1152)", style=solid]; +"1189 reshape_37" -> "1190 permute_39" [label="(4, 64, 3, 12, 32)", style=solid]; +"1190 permute_39" -> "1191 select_24" [label="(3, 4, 12, 64, 32)", style=solid]; +"1190 permute_39" -> "1192 select_25" [label="(3, 4, 12, 64, 32)", style=solid]; +"1190 permute_39" -> "1193 select_26" [label="(3, 4, 12, 64, 32)", style=solid]; +"1191 select_24" -> "1194 linalg_vector_norm_16" [label="(4, 12, 64, 32)", style=solid]; +"1191 select_24" -> "1196 expand_as_16" [label="(4, 12, 64, 32)", style=solid]; +"1191 select_24" -> "1197 div_16" [label="(4, 12, 64, 32)", style=solid]; +"1192 select_25" -> "1200 linalg_vector_norm_17" [label="(4, 12, 64, 32)", style=solid]; +"1192 select_25" -> "1202 expand_as_17" [label="(4, 12, 64, 32)", style=solid]; +"1192 select_25" -> "1203 div_17" [label="(4, 12, 64, 32)", style=solid]; +"1193 select_26" -> "1213 matmul_17" [label="(4, 12, 64, 32)", style=solid]; +"1194 linalg_vector_norm_16" -> "1195 clamp_min_16" [label="(4, 12, 64, 1)", style=solid]; +"1195 clamp_min_16" -> "1196 expand_as_16" [label="(4, 12, 64, 1)", style=solid]; +"1196 expand_as_16" -> "1197 div_16" [label="(4, 12, 64, 32)", style=solid]; +"1197 div_16" -> "1198 quantize_per_tensor_default_52" [label="(4, 12, 64, 32)", style=solid]; +"1198 quantize_per_tensor_default_52" -> "1199 dequantize_per_tensor_default_52" [label=None, style=solid]; +"1199 dequantize_per_tensor_default_52" -> "1207 matmul_16" [label=None, style=solid]; +"1200 linalg_vector_norm_17" -> "1201 clamp_min_17" [label="(4, 12, 64, 1)", style=solid]; +"1201 clamp_min_17" -> "1202 expand_as_17" [label="(4, 12, 64, 1)", style=solid]; +"1202 expand_as_17" -> "1203 div_17" [label="(4, 12, 64, 32)", style=solid]; +"1203 div_17" -> "1204 quantize_per_tensor_default_53" [label="(4, 12, 64, 32)", style=solid]; +"1204 quantize_per_tensor_default_53" -> "1205 dequantize_per_tensor_default_53" [label=None, style=solid]; +"1205 dequantize_per_tensor_default_53" -> "1206 transpose_16" [label=None, style=solid]; +"1206 transpose_16" -> "1207 matmul_16" [label="(4, 12, 32, 64)", style=solid]; +"1207 matmul_16" -> "1209 mul_17" [label="(4, 12, 64, 64)", style=solid]; +"1208 _frozen_param71" -> "1209 mul_17" [label="(12, 1, 1)", style=solid]; +"1209 mul_17" -> "1210 add_28" [label="(4, 12, 64, 64)", style=solid]; +"1210 add_28" -> "1211 softmax_8" [label="(4, 12, 64, 64)", style=solid]; +"1211 softmax_8" -> "1212 dropout_32" [label="(4, 12, 64, 64)", style=solid]; +"1212 dropout_32" -> "1213 matmul_17" [label="(4, 12, 64, 64)", style=solid]; +"1213 matmul_17" -> "1214 quantize_per_tensor_default_54" [label="(4, 12, 64, 32)", style=solid]; +"1214 quantize_per_tensor_default_54" -> "1215 dequantize_per_tensor_default_54" [label=None, style=solid]; +"1215 dequantize_per_tensor_default_54" -> "1216 transpose_17" [label=None, style=solid]; +"1216 transpose_17" -> "1217 reshape_38" [label="(4, 64, 12, 32)", style=solid]; +"1217 reshape_38" -> "1218 linear_53" [label="(4, 64, 384)", style=solid]; +"1218 linear_53" -> "1219 dropout_33" [label="(4, 64, 384)", style=solid]; +"1219 dropout_33" -> "1220 view_47" [label="(4, 64, 384)", style=solid]; +"1220 view_47" -> "1221 permute_40" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1221 permute_40" -> "1222 reshape_39" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1222 reshape_39" -> "1223 slice_134" [label="(1, 16, 16, 384)", style=solid]; +"1223 slice_134" -> "1224 slice_135" [label="(1, 16, 16, 384)", style=solid]; +"1224 slice_135" -> "1225 slice_136" [label="(1, 14, 16, 384)", style=solid]; +"1225 slice_136" -> "1226 slice_137" [label="(1, 14, 14, 384)", style=solid]; +"1226 slice_137" -> "1227 contiguous_15" [label="(1, 14, 14, 384)", style=solid]; +"1227 contiguous_15" -> "1228 layer_norm_19" [label="(1, 14, 14, 384)", style=solid]; +"1228 layer_norm_19" -> "1229 add_29" [label="(1, 14, 14, 384)", style=solid]; +"1229 add_29" -> "1230 quantize_per_tensor_default_55" [label="(1, 14, 14, 384)", style=solid]; +"1229 add_29" -> "1240 add_30" [label="(1, 14, 14, 384)", style=solid]; +"1230 quantize_per_tensor_default_55" -> "1231 dequantize_per_tensor_default_55" [label=None, style=solid]; +"1231 dequantize_per_tensor_default_55" -> "1232 linear_54" [label=None, style=solid]; +"1232 linear_54" -> "1233 gelu_8" [label="(1, 14, 14, 1536)", style=solid]; +"1233 gelu_8" -> "1234 quantize_per_tensor_default_56" [label="(1, 14, 14, 1536)", style=solid]; +"1234 quantize_per_tensor_default_56" -> "1235 dequantize_per_tensor_default_56" [label=None, style=solid]; +"1235 dequantize_per_tensor_default_56" -> "1236 dropout_34" [label=None, style=solid]; +"1236 dropout_34" -> "1237 linear_55" [label="(1, 14, 14, 1536)", style=solid]; +"1237 linear_55" -> "1238 dropout_35" [label="(1, 14, 14, 384)", style=solid]; +"1238 dropout_35" -> "1239 layer_norm_20" [label="(1, 14, 14, 384)", style=solid]; +"1239 layer_norm_20" -> "1240 add_30" [label="(1, 14, 14, 384)", style=solid]; +"1240 add_30" -> "1242 pad_11" [label="(1, 14, 14, 384)", style=solid]; +"1240 add_30" -> "1308 add_33" [label="(1, 14, 14, 384)", style=solid]; +"1241 _frozen_param72" -> "1272 add_31" [label="(1, 12, 64, 64)", style=solid]; +"1242 pad_11" -> "1243 roll_8" [label="(1, 16, 16, 384)", style=solid]; +"1243 roll_8" -> "1244 view_50" [label="(1, 16, 16, 384)", style=solid]; +"1244 view_50" -> "1245 permute_42" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1245 permute_42" -> "1246 reshape_40" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1246 reshape_40" -> "1247 quantize_per_tensor_default_57" [label="(4, 64, 384)", style=solid]; +"1246 reshape_40" -> "1273 new_zeros_4" [label="(4, 64, 384)", style=solid]; +"1247 quantize_per_tensor_default_57" -> "1248 dequantize_per_tensor_default_57" [label=None, style=solid]; +"1248 dequantize_per_tensor_default_57" -> "1250 linear_58" [label=None, style=solid]; +"1249 _frozen_param73" -> "1250 linear_58" [label="(1152,)", style=solid]; +"1250 linear_58" -> "1251 reshape_41" [label="(4, 64, 1152)", style=solid]; +"1251 reshape_41" -> "1252 permute_43" [label="(4, 64, 3, 12, 32)", style=solid]; +"1252 permute_43" -> "1253 select_27" [label="(3, 4, 12, 64, 32)", style=solid]; +"1252 permute_43" -> "1254 select_28" [label="(3, 4, 12, 64, 32)", style=solid]; +"1252 permute_43" -> "1255 select_29" [label="(3, 4, 12, 64, 32)", style=solid]; +"1253 select_27" -> "1256 linalg_vector_norm_18" [label="(4, 12, 64, 32)", style=solid]; +"1253 select_27" -> "1258 expand_as_18" [label="(4, 12, 64, 32)", style=solid]; +"1253 select_27" -> "1259 div_18" [label="(4, 12, 64, 32)", style=solid]; +"1254 select_28" -> "1262 linalg_vector_norm_19" [label="(4, 12, 64, 32)", style=solid]; +"1254 select_28" -> "1264 expand_as_19" [label="(4, 12, 64, 32)", style=solid]; +"1254 select_28" -> "1265 div_19" [label="(4, 12, 64, 32)", style=solid]; +"1255 select_29" -> "1291 matmul_19" [label="(4, 12, 64, 32)", style=solid]; +"1256 linalg_vector_norm_18" -> "1257 clamp_min_18" [label="(4, 12, 64, 1)", style=solid]; +"1257 clamp_min_18" -> "1258 expand_as_18" [label="(4, 12, 64, 1)", style=solid]; +"1258 expand_as_18" -> "1259 div_18" [label="(4, 12, 64, 32)", style=solid]; +"1259 div_18" -> "1260 quantize_per_tensor_default_58" [label="(4, 12, 64, 32)", style=solid]; +"1260 quantize_per_tensor_default_58" -> "1261 dequantize_per_tensor_default_58" [label=None, style=solid]; +"1261 dequantize_per_tensor_default_58" -> "1269 matmul_18" [label=None, style=solid]; +"1262 linalg_vector_norm_19" -> "1263 clamp_min_19" [label="(4, 12, 64, 1)", style=solid]; +"1263 clamp_min_19" -> "1264 expand_as_19" [label="(4, 12, 64, 1)", style=solid]; +"1264 expand_as_19" -> "1265 div_19" [label="(4, 12, 64, 32)", style=solid]; +"1265 div_19" -> "1266 quantize_per_tensor_default_59" [label="(4, 12, 64, 32)", style=solid]; +"1266 quantize_per_tensor_default_59" -> "1267 dequantize_per_tensor_default_59" [label=None, style=solid]; +"1267 dequantize_per_tensor_default_59" -> "1268 transpose_18" [label=None, style=solid]; +"1268 transpose_18" -> "1269 matmul_18" [label="(4, 12, 32, 64)", style=solid]; +"1269 matmul_18" -> "1271 mul_19" [label="(4, 12, 64, 64)", style=solid]; +"1270 _frozen_param75" -> "1271 mul_19" [label="(12, 1, 1)", style=solid]; +"1271 mul_19" -> "1272 add_31" [label="(4, 12, 64, 64)", style=solid]; +"1272 add_31" -> "1284 view_52" [label="(4, 12, 64, 64)", style=solid]; +"1273 new_zeros_4" -> "1274 view_51" [label="(16, 16)", style=solid]; +"1274 view_51" -> "1275 permute_44" [label="(2, 8, 2, 8)", style=solid]; +"1275 permute_44" -> "1276 reshape_42" [label="(2, 2, 8, 8)", style=solid]; +"1276 reshape_42" -> "1277 unsqueeze_26" [label="(4, 64)", style=solid]; +"1276 reshape_42" -> "1278 unsqueeze_27" [label="(4, 64)", style=solid]; +"1277 unsqueeze_26" -> "1279 sub_4" [label="(4, 1, 64)", style=solid]; +"1278 unsqueeze_27" -> "1279 sub_4" [label="(4, 64, 1)", style=solid]; +"1279 sub_4" -> "1280 ne_4" [label="(4, 64, 64)", style=solid]; +"1279 sub_4" -> "1281 masked_fill_8" [label="(4, 64, 64)", style=solid]; +"1279 sub_4" -> "1282 eq_4" [label="(4, 64, 64)", style=solid]; +"1280 ne_4" -> "1281 masked_fill_8" [label="(4, 64, 64)", style=solid]; +"1281 masked_fill_8" -> "1283 masked_fill_9" [label="(4, 64, 64)", style=solid]; +"1282 eq_4" -> "1283 masked_fill_9" [label="(4, 64, 64)", style=solid]; +"1283 masked_fill_9" -> "1285 unsqueeze_28" [label="(4, 64, 64)", style=solid]; +"1284 view_52" -> "1287 add_32" [label="(1, 4, 12, 64, 64)", style=solid]; +"1285 unsqueeze_28" -> "1286 unsqueeze_29" [label="(4, 1, 64, 64)", style=solid]; +"1286 unsqueeze_29" -> "1287 add_32" [label="(1, 4, 1, 64, 64)", style=solid]; +"1287 add_32" -> "1288 view_53" [label="(1, 4, 12, 64, 64)", style=solid]; +"1288 view_53" -> "1289 softmax_9" [label="(4, 12, 64, 64)", style=solid]; +"1289 softmax_9" -> "1290 dropout_36" [label="(4, 12, 64, 64)", style=solid]; +"1290 dropout_36" -> "1291 matmul_19" [label="(4, 12, 64, 64)", style=solid]; +"1291 matmul_19" -> "1292 quantize_per_tensor_default_60" [label="(4, 12, 64, 32)", style=solid]; +"1292 quantize_per_tensor_default_60" -> "1293 dequantize_per_tensor_default_60" [label=None, style=solid]; +"1293 dequantize_per_tensor_default_60" -> "1294 transpose_19" [label=None, style=solid]; +"1294 transpose_19" -> "1295 reshape_43" [label="(4, 64, 12, 32)", style=solid]; +"1295 reshape_43" -> "1296 linear_59" [label="(4, 64, 384)", style=solid]; +"1296 linear_59" -> "1297 dropout_37" [label="(4, 64, 384)", style=solid]; +"1297 dropout_37" -> "1298 view_54" [label="(4, 64, 384)", style=solid]; +"1298 view_54" -> "1299 permute_45" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1299 permute_45" -> "1300 reshape_44" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1300 reshape_44" -> "1301 roll_9" [label="(1, 16, 16, 384)", style=solid]; +"1301 roll_9" -> "1302 slice_157" [label="(1, 16, 16, 384)", style=solid]; +"1302 slice_157" -> "1303 slice_158" [label="(1, 16, 16, 384)", style=solid]; +"1303 slice_158" -> "1304 slice_159" [label="(1, 14, 16, 384)", style=solid]; +"1304 slice_159" -> "1305 slice_160" [label="(1, 14, 14, 384)", style=solid]; +"1305 slice_160" -> "1306 contiguous_17" [label="(1, 14, 14, 384)", style=solid]; +"1306 contiguous_17" -> "1307 layer_norm_21" [label="(1, 14, 14, 384)", style=solid]; +"1307 layer_norm_21" -> "1308 add_33" [label="(1, 14, 14, 384)", style=solid]; +"1308 add_33" -> "1309 quantize_per_tensor_default_61" [label="(1, 14, 14, 384)", style=solid]; +"1308 add_33" -> "1319 add_34" [label="(1, 14, 14, 384)", style=solid]; +"1309 quantize_per_tensor_default_61" -> "1310 dequantize_per_tensor_default_61" [label=None, style=solid]; +"1310 dequantize_per_tensor_default_61" -> "1311 linear_60" [label=None, style=solid]; +"1311 linear_60" -> "1312 gelu_9" [label="(1, 14, 14, 1536)", style=solid]; +"1312 gelu_9" -> "1313 quantize_per_tensor_default_62" [label="(1, 14, 14, 1536)", style=solid]; +"1313 quantize_per_tensor_default_62" -> "1314 dequantize_per_tensor_default_62" [label=None, style=solid]; +"1314 dequantize_per_tensor_default_62" -> "1315 dropout_38" [label=None, style=solid]; +"1315 dropout_38" -> "1316 linear_61" [label="(1, 14, 14, 1536)", style=solid]; +"1316 linear_61" -> "1317 dropout_39" [label="(1, 14, 14, 384)", style=solid]; +"1317 dropout_39" -> "1318 layer_norm_22" [label="(1, 14, 14, 384)", style=solid]; +"1318 layer_norm_22" -> "1319 add_34" [label="(1, 14, 14, 384)", style=solid]; +"1319 add_34" -> "1320 quantize_per_tensor_default_63" [label="(1, 14, 14, 384)", style=solid]; +"1319 add_34" -> "1369 add_36" [label="(1, 14, 14, 384)", style=solid]; +"1320 quantize_per_tensor_default_63" -> "1321 dequantize_per_tensor_default_63" [label=None, style=solid]; +"1321 dequantize_per_tensor_default_63" -> "1323 pad_12" [label=None, style=solid]; +"1322 _frozen_param85" -> "1350 add_35" [label="(1, 12, 64, 64)", style=solid]; +"1323 pad_12" -> "1324 view_57" [label="(1, 16, 16, 384)", style=solid]; +"1324 view_57" -> "1325 permute_47" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1325 permute_47" -> "1326 reshape_45" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1326 reshape_45" -> "1328 linear_64" [label="(4, 64, 384)", style=solid]; +"1327 _frozen_param86" -> "1328 linear_64" [label="(1152,)", style=solid]; +"1328 linear_64" -> "1329 reshape_46" [label="(4, 64, 1152)", style=solid]; +"1329 reshape_46" -> "1330 permute_48" [label="(4, 64, 3, 12, 32)", style=solid]; +"1330 permute_48" -> "1331 select_30" [label="(3, 4, 12, 64, 32)", style=solid]; +"1330 permute_48" -> "1332 select_31" [label="(3, 4, 12, 64, 32)", style=solid]; +"1330 permute_48" -> "1333 select_32" [label="(3, 4, 12, 64, 32)", style=solid]; +"1331 select_30" -> "1334 linalg_vector_norm_20" [label="(4, 12, 64, 32)", style=solid]; +"1331 select_30" -> "1336 expand_as_20" [label="(4, 12, 64, 32)", style=solid]; +"1331 select_30" -> "1337 div_20" [label="(4, 12, 64, 32)", style=solid]; +"1332 select_31" -> "1340 linalg_vector_norm_21" [label="(4, 12, 64, 32)", style=solid]; +"1332 select_31" -> "1342 expand_as_21" [label="(4, 12, 64, 32)", style=solid]; +"1332 select_31" -> "1343 div_21" [label="(4, 12, 64, 32)", style=solid]; +"1333 select_32" -> "1353 matmul_21" [label="(4, 12, 64, 32)", style=solid]; +"1334 linalg_vector_norm_20" -> "1335 clamp_min_20" [label="(4, 12, 64, 1)", style=solid]; +"1335 clamp_min_20" -> "1336 expand_as_20" [label="(4, 12, 64, 1)", style=solid]; +"1336 expand_as_20" -> "1337 div_20" [label="(4, 12, 64, 32)", style=solid]; +"1337 div_20" -> "1338 quantize_per_tensor_default_64" [label="(4, 12, 64, 32)", style=solid]; +"1338 quantize_per_tensor_default_64" -> "1339 dequantize_per_tensor_default_64" [label=None, style=solid]; +"1339 dequantize_per_tensor_default_64" -> "1347 matmul_20" [label=None, style=solid]; +"1340 linalg_vector_norm_21" -> "1341 clamp_min_21" [label="(4, 12, 64, 1)", style=solid]; +"1341 clamp_min_21" -> "1342 expand_as_21" [label="(4, 12, 64, 1)", style=solid]; +"1342 expand_as_21" -> "1343 div_21" [label="(4, 12, 64, 32)", style=solid]; +"1343 div_21" -> "1344 quantize_per_tensor_default_65" [label="(4, 12, 64, 32)", style=solid]; +"1344 quantize_per_tensor_default_65" -> "1345 dequantize_per_tensor_default_65" [label=None, style=solid]; +"1345 dequantize_per_tensor_default_65" -> "1346 transpose_20" [label=None, style=solid]; +"1346 transpose_20" -> "1347 matmul_20" [label="(4, 12, 32, 64)", style=solid]; +"1347 matmul_20" -> "1349 mul_21" [label="(4, 12, 64, 64)", style=solid]; +"1348 _frozen_param88" -> "1349 mul_21" [label="(12, 1, 1)", style=solid]; +"1349 mul_21" -> "1350 add_35" [label="(4, 12, 64, 64)", style=solid]; +"1350 add_35" -> "1351 softmax_10" [label="(4, 12, 64, 64)", style=solid]; +"1351 softmax_10" -> "1352 dropout_40" [label="(4, 12, 64, 64)", style=solid]; +"1352 dropout_40" -> "1353 matmul_21" [label="(4, 12, 64, 64)", style=solid]; +"1353 matmul_21" -> "1354 quantize_per_tensor_default_66" [label="(4, 12, 64, 32)", style=solid]; +"1354 quantize_per_tensor_default_66" -> "1355 dequantize_per_tensor_default_66" [label=None, style=solid]; +"1355 dequantize_per_tensor_default_66" -> "1356 transpose_21" [label=None, style=solid]; +"1356 transpose_21" -> "1357 reshape_47" [label="(4, 64, 12, 32)", style=solid]; +"1357 reshape_47" -> "1358 linear_65" [label="(4, 64, 384)", style=solid]; +"1358 linear_65" -> "1359 dropout_41" [label="(4, 64, 384)", style=solid]; +"1359 dropout_41" -> "1360 view_58" [label="(4, 64, 384)", style=solid]; +"1360 view_58" -> "1361 permute_49" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1361 permute_49" -> "1362 reshape_48" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1362 reshape_48" -> "1363 slice_162" [label="(1, 16, 16, 384)", style=solid]; +"1363 slice_162" -> "1364 slice_163" [label="(1, 16, 16, 384)", style=solid]; +"1364 slice_163" -> "1365 slice_164" [label="(1, 14, 16, 384)", style=solid]; +"1365 slice_164" -> "1366 slice_165" [label="(1, 14, 14, 384)", style=solid]; +"1366 slice_165" -> "1367 contiguous_19" [label="(1, 14, 14, 384)", style=solid]; +"1367 contiguous_19" -> "1368 layer_norm_23" [label="(1, 14, 14, 384)", style=solid]; +"1368 layer_norm_23" -> "1369 add_36" [label="(1, 14, 14, 384)", style=solid]; +"1369 add_36" -> "1370 quantize_per_tensor_default_67" [label="(1, 14, 14, 384)", style=solid]; +"1369 add_36" -> "1380 add_37" [label="(1, 14, 14, 384)", style=solid]; +"1370 quantize_per_tensor_default_67" -> "1371 dequantize_per_tensor_default_67" [label=None, style=solid]; +"1371 dequantize_per_tensor_default_67" -> "1372 linear_66" [label=None, style=solid]; +"1372 linear_66" -> "1373 gelu_10" [label="(1, 14, 14, 1536)", style=solid]; +"1373 gelu_10" -> "1374 quantize_per_tensor_default_68" [label="(1, 14, 14, 1536)", style=solid]; +"1374 quantize_per_tensor_default_68" -> "1375 dequantize_per_tensor_default_68" [label=None, style=solid]; +"1375 dequantize_per_tensor_default_68" -> "1376 dropout_42" [label=None, style=solid]; +"1376 dropout_42" -> "1377 linear_67" [label="(1, 14, 14, 1536)", style=solid]; +"1377 linear_67" -> "1378 dropout_43" [label="(1, 14, 14, 384)", style=solid]; +"1378 dropout_43" -> "1379 layer_norm_24" [label="(1, 14, 14, 384)", style=solid]; +"1379 layer_norm_24" -> "1380 add_37" [label="(1, 14, 14, 384)", style=solid]; +"1380 add_37" -> "1382 pad_13" [label="(1, 14, 14, 384)", style=solid]; +"1380 add_37" -> "1448 add_40" [label="(1, 14, 14, 384)", style=solid]; +"1381 _frozen_param89" -> "1412 add_38" [label="(1, 12, 64, 64)", style=solid]; +"1382 pad_13" -> "1383 roll_10" [label="(1, 16, 16, 384)", style=solid]; +"1383 roll_10" -> "1384 view_61" [label="(1, 16, 16, 384)", style=solid]; +"1384 view_61" -> "1385 permute_51" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1385 permute_51" -> "1386 reshape_49" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1386 reshape_49" -> "1387 quantize_per_tensor_default_69" [label="(4, 64, 384)", style=solid]; +"1386 reshape_49" -> "1413 new_zeros_5" [label="(4, 64, 384)", style=solid]; +"1387 quantize_per_tensor_default_69" -> "1388 dequantize_per_tensor_default_69" [label=None, style=solid]; +"1388 dequantize_per_tensor_default_69" -> "1390 linear_70" [label=None, style=solid]; +"1389 _frozen_param90" -> "1390 linear_70" [label="(1152,)", style=solid]; +"1390 linear_70" -> "1391 reshape_50" [label="(4, 64, 1152)", style=solid]; +"1391 reshape_50" -> "1392 permute_52" [label="(4, 64, 3, 12, 32)", style=solid]; +"1392 permute_52" -> "1393 select_33" [label="(3, 4, 12, 64, 32)", style=solid]; +"1392 permute_52" -> "1394 select_34" [label="(3, 4, 12, 64, 32)", style=solid]; +"1392 permute_52" -> "1395 select_35" [label="(3, 4, 12, 64, 32)", style=solid]; +"1393 select_33" -> "1396 linalg_vector_norm_22" [label="(4, 12, 64, 32)", style=solid]; +"1393 select_33" -> "1398 expand_as_22" [label="(4, 12, 64, 32)", style=solid]; +"1393 select_33" -> "1399 div_22" [label="(4, 12, 64, 32)", style=solid]; +"1394 select_34" -> "1402 linalg_vector_norm_23" [label="(4, 12, 64, 32)", style=solid]; +"1394 select_34" -> "1404 expand_as_23" [label="(4, 12, 64, 32)", style=solid]; +"1394 select_34" -> "1405 div_23" [label="(4, 12, 64, 32)", style=solid]; +"1395 select_35" -> "1431 matmul_23" [label="(4, 12, 64, 32)", style=solid]; +"1396 linalg_vector_norm_22" -> "1397 clamp_min_22" [label="(4, 12, 64, 1)", style=solid]; +"1397 clamp_min_22" -> "1398 expand_as_22" [label="(4, 12, 64, 1)", style=solid]; +"1398 expand_as_22" -> "1399 div_22" [label="(4, 12, 64, 32)", style=solid]; +"1399 div_22" -> "1400 quantize_per_tensor_default_70" [label="(4, 12, 64, 32)", style=solid]; +"1400 quantize_per_tensor_default_70" -> "1401 dequantize_per_tensor_default_70" [label=None, style=solid]; +"1401 dequantize_per_tensor_default_70" -> "1409 matmul_22" [label=None, style=solid]; +"1402 linalg_vector_norm_23" -> "1403 clamp_min_23" [label="(4, 12, 64, 1)", style=solid]; +"1403 clamp_min_23" -> "1404 expand_as_23" [label="(4, 12, 64, 1)", style=solid]; +"1404 expand_as_23" -> "1405 div_23" [label="(4, 12, 64, 32)", style=solid]; +"1405 div_23" -> "1406 quantize_per_tensor_default_71" [label="(4, 12, 64, 32)", style=solid]; +"1406 quantize_per_tensor_default_71" -> "1407 dequantize_per_tensor_default_71" [label=None, style=solid]; +"1407 dequantize_per_tensor_default_71" -> "1408 transpose_22" [label=None, style=solid]; +"1408 transpose_22" -> "1409 matmul_22" [label="(4, 12, 32, 64)", style=solid]; +"1409 matmul_22" -> "1411 mul_23" [label="(4, 12, 64, 64)", style=solid]; +"1410 _frozen_param92" -> "1411 mul_23" [label="(12, 1, 1)", style=solid]; +"1411 mul_23" -> "1412 add_38" [label="(4, 12, 64, 64)", style=solid]; +"1412 add_38" -> "1424 view_63" [label="(4, 12, 64, 64)", style=solid]; +"1413 new_zeros_5" -> "1414 view_62" [label="(16, 16)", style=solid]; +"1414 view_62" -> "1415 permute_53" [label="(2, 8, 2, 8)", style=solid]; +"1415 permute_53" -> "1416 reshape_51" [label="(2, 2, 8, 8)", style=solid]; +"1416 reshape_51" -> "1417 unsqueeze_32" [label="(4, 64)", style=solid]; +"1416 reshape_51" -> "1418 unsqueeze_33" [label="(4, 64)", style=solid]; +"1417 unsqueeze_32" -> "1419 sub_5" [label="(4, 1, 64)", style=solid]; +"1418 unsqueeze_33" -> "1419 sub_5" [label="(4, 64, 1)", style=solid]; +"1419 sub_5" -> "1420 ne_5" [label="(4, 64, 64)", style=solid]; +"1419 sub_5" -> "1421 masked_fill_10" [label="(4, 64, 64)", style=solid]; +"1419 sub_5" -> "1422 eq_5" [label="(4, 64, 64)", style=solid]; +"1420 ne_5" -> "1421 masked_fill_10" [label="(4, 64, 64)", style=solid]; +"1421 masked_fill_10" -> "1423 masked_fill_11" [label="(4, 64, 64)", style=solid]; +"1422 eq_5" -> "1423 masked_fill_11" [label="(4, 64, 64)", style=solid]; +"1423 masked_fill_11" -> "1425 unsqueeze_34" [label="(4, 64, 64)", style=solid]; +"1424 view_63" -> "1427 add_39" [label="(1, 4, 12, 64, 64)", style=solid]; +"1425 unsqueeze_34" -> "1426 unsqueeze_35" [label="(4, 1, 64, 64)", style=solid]; +"1426 unsqueeze_35" -> "1427 add_39" [label="(1, 4, 1, 64, 64)", style=solid]; +"1427 add_39" -> "1428 view_64" [label="(1, 4, 12, 64, 64)", style=solid]; +"1428 view_64" -> "1429 softmax_11" [label="(4, 12, 64, 64)", style=solid]; +"1429 softmax_11" -> "1430 dropout_44" [label="(4, 12, 64, 64)", style=solid]; +"1430 dropout_44" -> "1431 matmul_23" [label="(4, 12, 64, 64)", style=solid]; +"1431 matmul_23" -> "1432 quantize_per_tensor_default_72" [label="(4, 12, 64, 32)", style=solid]; +"1432 quantize_per_tensor_default_72" -> "1433 dequantize_per_tensor_default_72" [label=None, style=solid]; +"1433 dequantize_per_tensor_default_72" -> "1434 transpose_23" [label=None, style=solid]; +"1434 transpose_23" -> "1435 reshape_52" [label="(4, 64, 12, 32)", style=solid]; +"1435 reshape_52" -> "1436 linear_71" [label="(4, 64, 384)", style=solid]; +"1436 linear_71" -> "1437 dropout_45" [label="(4, 64, 384)", style=solid]; +"1437 dropout_45" -> "1438 view_65" [label="(4, 64, 384)", style=solid]; +"1438 view_65" -> "1439 permute_54" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1439 permute_54" -> "1440 reshape_53" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1440 reshape_53" -> "1441 roll_11" [label="(1, 16, 16, 384)", style=solid]; +"1441 roll_11" -> "1442 slice_185" [label="(1, 16, 16, 384)", style=solid]; +"1442 slice_185" -> "1443 slice_186" [label="(1, 16, 16, 384)", style=solid]; +"1443 slice_186" -> "1444 slice_187" [label="(1, 14, 16, 384)", style=solid]; +"1444 slice_187" -> "1445 slice_188" [label="(1, 14, 14, 384)", style=solid]; +"1445 slice_188" -> "1446 contiguous_21" [label="(1, 14, 14, 384)", style=solid]; +"1446 contiguous_21" -> "1447 layer_norm_25" [label="(1, 14, 14, 384)", style=solid]; +"1447 layer_norm_25" -> "1448 add_40" [label="(1, 14, 14, 384)", style=solid]; +"1448 add_40" -> "1449 quantize_per_tensor_default_73" [label="(1, 14, 14, 384)", style=solid]; +"1448 add_40" -> "1459 add_41" [label="(1, 14, 14, 384)", style=solid]; +"1449 quantize_per_tensor_default_73" -> "1450 dequantize_per_tensor_default_73" [label=None, style=solid]; +"1450 dequantize_per_tensor_default_73" -> "1451 linear_72" [label=None, style=solid]; +"1451 linear_72" -> "1452 gelu_11" [label="(1, 14, 14, 1536)", style=solid]; +"1452 gelu_11" -> "1453 quantize_per_tensor_default_74" [label="(1, 14, 14, 1536)", style=solid]; +"1453 quantize_per_tensor_default_74" -> "1454 dequantize_per_tensor_default_74" [label=None, style=solid]; +"1454 dequantize_per_tensor_default_74" -> "1455 dropout_46" [label=None, style=solid]; +"1455 dropout_46" -> "1456 linear_73" [label="(1, 14, 14, 1536)", style=solid]; +"1456 linear_73" -> "1457 dropout_47" [label="(1, 14, 14, 384)", style=solid]; +"1457 dropout_47" -> "1458 layer_norm_26" [label="(1, 14, 14, 384)", style=solid]; +"1458 layer_norm_26" -> "1459 add_41" [label="(1, 14, 14, 384)", style=solid]; +"1459 add_41" -> "1460 quantize_per_tensor_default_75" [label="(1, 14, 14, 384)", style=solid]; +"1459 add_41" -> "1509 add_43" [label="(1, 14, 14, 384)", style=solid]; +"1460 quantize_per_tensor_default_75" -> "1461 dequantize_per_tensor_default_75" [label=None, style=solid]; +"1461 dequantize_per_tensor_default_75" -> "1463 pad_14" [label=None, style=solid]; +"1462 _frozen_param102" -> "1490 add_42" [label="(1, 12, 64, 64)", style=solid]; +"1463 pad_14" -> "1464 view_68" [label="(1, 16, 16, 384)", style=solid]; +"1464 view_68" -> "1465 permute_56" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1465 permute_56" -> "1466 reshape_54" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1466 reshape_54" -> "1468 linear_76" [label="(4, 64, 384)", style=solid]; +"1467 _frozen_param103" -> "1468 linear_76" [label="(1152,)", style=solid]; +"1468 linear_76" -> "1469 reshape_55" [label="(4, 64, 1152)", style=solid]; +"1469 reshape_55" -> "1470 permute_57" [label="(4, 64, 3, 12, 32)", style=solid]; +"1470 permute_57" -> "1471 select_36" [label="(3, 4, 12, 64, 32)", style=solid]; +"1470 permute_57" -> "1472 select_37" [label="(3, 4, 12, 64, 32)", style=solid]; +"1470 permute_57" -> "1473 select_38" [label="(3, 4, 12, 64, 32)", style=solid]; +"1471 select_36" -> "1474 linalg_vector_norm_24" [label="(4, 12, 64, 32)", style=solid]; +"1471 select_36" -> "1476 expand_as_24" [label="(4, 12, 64, 32)", style=solid]; +"1471 select_36" -> "1477 div_24" [label="(4, 12, 64, 32)", style=solid]; +"1472 select_37" -> "1480 linalg_vector_norm_25" [label="(4, 12, 64, 32)", style=solid]; +"1472 select_37" -> "1482 expand_as_25" [label="(4, 12, 64, 32)", style=solid]; +"1472 select_37" -> "1483 div_25" [label="(4, 12, 64, 32)", style=solid]; +"1473 select_38" -> "1493 matmul_25" [label="(4, 12, 64, 32)", style=solid]; +"1474 linalg_vector_norm_24" -> "1475 clamp_min_24" [label="(4, 12, 64, 1)", style=solid]; +"1475 clamp_min_24" -> "1476 expand_as_24" [label="(4, 12, 64, 1)", style=solid]; +"1476 expand_as_24" -> "1477 div_24" [label="(4, 12, 64, 32)", style=solid]; +"1477 div_24" -> "1478 quantize_per_tensor_default_76" [label="(4, 12, 64, 32)", style=solid]; +"1478 quantize_per_tensor_default_76" -> "1479 dequantize_per_tensor_default_76" [label=None, style=solid]; +"1479 dequantize_per_tensor_default_76" -> "1487 matmul_24" [label=None, style=solid]; +"1480 linalg_vector_norm_25" -> "1481 clamp_min_25" [label="(4, 12, 64, 1)", style=solid]; +"1481 clamp_min_25" -> "1482 expand_as_25" [label="(4, 12, 64, 1)", style=solid]; +"1482 expand_as_25" -> "1483 div_25" [label="(4, 12, 64, 32)", style=solid]; +"1483 div_25" -> "1484 quantize_per_tensor_default_77" [label="(4, 12, 64, 32)", style=solid]; +"1484 quantize_per_tensor_default_77" -> "1485 dequantize_per_tensor_default_77" [label=None, style=solid]; +"1485 dequantize_per_tensor_default_77" -> "1486 transpose_24" [label=None, style=solid]; +"1486 transpose_24" -> "1487 matmul_24" [label="(4, 12, 32, 64)", style=solid]; +"1487 matmul_24" -> "1489 mul_25" [label="(4, 12, 64, 64)", style=solid]; +"1488 _frozen_param105" -> "1489 mul_25" [label="(12, 1, 1)", style=solid]; +"1489 mul_25" -> "1490 add_42" [label="(4, 12, 64, 64)", style=solid]; +"1490 add_42" -> "1491 softmax_12" [label="(4, 12, 64, 64)", style=solid]; +"1491 softmax_12" -> "1492 dropout_48" [label="(4, 12, 64, 64)", style=solid]; +"1492 dropout_48" -> "1493 matmul_25" [label="(4, 12, 64, 64)", style=solid]; +"1493 matmul_25" -> "1494 quantize_per_tensor_default_78" [label="(4, 12, 64, 32)", style=solid]; +"1494 quantize_per_tensor_default_78" -> "1495 dequantize_per_tensor_default_78" [label=None, style=solid]; +"1495 dequantize_per_tensor_default_78" -> "1496 transpose_25" [label=None, style=solid]; +"1496 transpose_25" -> "1497 reshape_56" [label="(4, 64, 12, 32)", style=solid]; +"1497 reshape_56" -> "1498 linear_77" [label="(4, 64, 384)", style=solid]; +"1498 linear_77" -> "1499 dropout_49" [label="(4, 64, 384)", style=solid]; +"1499 dropout_49" -> "1500 view_69" [label="(4, 64, 384)", style=solid]; +"1500 view_69" -> "1501 permute_58" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1501 permute_58" -> "1502 reshape_57" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1502 reshape_57" -> "1503 slice_190" [label="(1, 16, 16, 384)", style=solid]; +"1503 slice_190" -> "1504 slice_191" [label="(1, 16, 16, 384)", style=solid]; +"1504 slice_191" -> "1505 slice_192" [label="(1, 14, 16, 384)", style=solid]; +"1505 slice_192" -> "1506 slice_193" [label="(1, 14, 14, 384)", style=solid]; +"1506 slice_193" -> "1507 contiguous_23" [label="(1, 14, 14, 384)", style=solid]; +"1507 contiguous_23" -> "1508 layer_norm_27" [label="(1, 14, 14, 384)", style=solid]; +"1508 layer_norm_27" -> "1509 add_43" [label="(1, 14, 14, 384)", style=solid]; +"1509 add_43" -> "1510 quantize_per_tensor_default_79" [label="(1, 14, 14, 384)", style=solid]; +"1509 add_43" -> "1520 add_44" [label="(1, 14, 14, 384)", style=solid]; +"1510 quantize_per_tensor_default_79" -> "1511 dequantize_per_tensor_default_79" [label=None, style=solid]; +"1511 dequantize_per_tensor_default_79" -> "1512 linear_78" [label=None, style=solid]; +"1512 linear_78" -> "1513 gelu_12" [label="(1, 14, 14, 1536)", style=solid]; +"1513 gelu_12" -> "1514 quantize_per_tensor_default_80" [label="(1, 14, 14, 1536)", style=solid]; +"1514 quantize_per_tensor_default_80" -> "1515 dequantize_per_tensor_default_80" [label=None, style=solid]; +"1515 dequantize_per_tensor_default_80" -> "1516 dropout_50" [label=None, style=solid]; +"1516 dropout_50" -> "1517 linear_79" [label="(1, 14, 14, 1536)", style=solid]; +"1517 linear_79" -> "1518 dropout_51" [label="(1, 14, 14, 384)", style=solid]; +"1518 dropout_51" -> "1519 layer_norm_28" [label="(1, 14, 14, 384)", style=solid]; +"1519 layer_norm_28" -> "1520 add_44" [label="(1, 14, 14, 384)", style=solid]; +"1520 add_44" -> "1522 pad_15" [label="(1, 14, 14, 384)", style=solid]; +"1520 add_44" -> "1588 add_47" [label="(1, 14, 14, 384)", style=solid]; +"1521 _frozen_param106" -> "1552 add_45" [label="(1, 12, 64, 64)", style=solid]; +"1522 pad_15" -> "1523 roll_12" [label="(1, 16, 16, 384)", style=solid]; +"1523 roll_12" -> "1524 view_72" [label="(1, 16, 16, 384)", style=solid]; +"1524 view_72" -> "1525 permute_60" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1525 permute_60" -> "1526 reshape_58" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1526 reshape_58" -> "1527 quantize_per_tensor_default_81" [label="(4, 64, 384)", style=solid]; +"1526 reshape_58" -> "1553 new_zeros_6" [label="(4, 64, 384)", style=solid]; +"1527 quantize_per_tensor_default_81" -> "1528 dequantize_per_tensor_default_81" [label=None, style=solid]; +"1528 dequantize_per_tensor_default_81" -> "1530 linear_82" [label=None, style=solid]; +"1529 _frozen_param107" -> "1530 linear_82" [label="(1152,)", style=solid]; +"1530 linear_82" -> "1531 reshape_59" [label="(4, 64, 1152)", style=solid]; +"1531 reshape_59" -> "1532 permute_61" [label="(4, 64, 3, 12, 32)", style=solid]; +"1532 permute_61" -> "1533 select_39" [label="(3, 4, 12, 64, 32)", style=solid]; +"1532 permute_61" -> "1534 select_40" [label="(3, 4, 12, 64, 32)", style=solid]; +"1532 permute_61" -> "1535 select_41" [label="(3, 4, 12, 64, 32)", style=solid]; +"1533 select_39" -> "1536 linalg_vector_norm_26" [label="(4, 12, 64, 32)", style=solid]; +"1533 select_39" -> "1538 expand_as_26" [label="(4, 12, 64, 32)", style=solid]; +"1533 select_39" -> "1539 div_26" [label="(4, 12, 64, 32)", style=solid]; +"1534 select_40" -> "1542 linalg_vector_norm_27" [label="(4, 12, 64, 32)", style=solid]; +"1534 select_40" -> "1544 expand_as_27" [label="(4, 12, 64, 32)", style=solid]; +"1534 select_40" -> "1545 div_27" [label="(4, 12, 64, 32)", style=solid]; +"1535 select_41" -> "1571 matmul_27" [label="(4, 12, 64, 32)", style=solid]; +"1536 linalg_vector_norm_26" -> "1537 clamp_min_26" [label="(4, 12, 64, 1)", style=solid]; +"1537 clamp_min_26" -> "1538 expand_as_26" [label="(4, 12, 64, 1)", style=solid]; +"1538 expand_as_26" -> "1539 div_26" [label="(4, 12, 64, 32)", style=solid]; +"1539 div_26" -> "1540 quantize_per_tensor_default_82" [label="(4, 12, 64, 32)", style=solid]; +"1540 quantize_per_tensor_default_82" -> "1541 dequantize_per_tensor_default_82" [label=None, style=solid]; +"1541 dequantize_per_tensor_default_82" -> "1549 matmul_26" [label=None, style=solid]; +"1542 linalg_vector_norm_27" -> "1543 clamp_min_27" [label="(4, 12, 64, 1)", style=solid]; +"1543 clamp_min_27" -> "1544 expand_as_27" [label="(4, 12, 64, 1)", style=solid]; +"1544 expand_as_27" -> "1545 div_27" [label="(4, 12, 64, 32)", style=solid]; +"1545 div_27" -> "1546 quantize_per_tensor_default_83" [label="(4, 12, 64, 32)", style=solid]; +"1546 quantize_per_tensor_default_83" -> "1547 dequantize_per_tensor_default_83" [label=None, style=solid]; +"1547 dequantize_per_tensor_default_83" -> "1548 transpose_26" [label=None, style=solid]; +"1548 transpose_26" -> "1549 matmul_26" [label="(4, 12, 32, 64)", style=solid]; +"1549 matmul_26" -> "1551 mul_27" [label="(4, 12, 64, 64)", style=solid]; +"1550 _frozen_param109" -> "1551 mul_27" [label="(12, 1, 1)", style=solid]; +"1551 mul_27" -> "1552 add_45" [label="(4, 12, 64, 64)", style=solid]; +"1552 add_45" -> "1564 view_74" [label="(4, 12, 64, 64)", style=solid]; +"1553 new_zeros_6" -> "1554 view_73" [label="(16, 16)", style=solid]; +"1554 view_73" -> "1555 permute_62" [label="(2, 8, 2, 8)", style=solid]; +"1555 permute_62" -> "1556 reshape_60" [label="(2, 2, 8, 8)", style=solid]; +"1556 reshape_60" -> "1557 unsqueeze_38" [label="(4, 64)", style=solid]; +"1556 reshape_60" -> "1558 unsqueeze_39" [label="(4, 64)", style=solid]; +"1557 unsqueeze_38" -> "1559 sub_6" [label="(4, 1, 64)", style=solid]; +"1558 unsqueeze_39" -> "1559 sub_6" [label="(4, 64, 1)", style=solid]; +"1559 sub_6" -> "1560 ne_6" [label="(4, 64, 64)", style=solid]; +"1559 sub_6" -> "1561 masked_fill_12" [label="(4, 64, 64)", style=solid]; +"1559 sub_6" -> "1562 eq_6" [label="(4, 64, 64)", style=solid]; +"1560 ne_6" -> "1561 masked_fill_12" [label="(4, 64, 64)", style=solid]; +"1561 masked_fill_12" -> "1563 masked_fill_13" [label="(4, 64, 64)", style=solid]; +"1562 eq_6" -> "1563 masked_fill_13" [label="(4, 64, 64)", style=solid]; +"1563 masked_fill_13" -> "1565 unsqueeze_40" [label="(4, 64, 64)", style=solid]; +"1564 view_74" -> "1567 add_46" [label="(1, 4, 12, 64, 64)", style=solid]; +"1565 unsqueeze_40" -> "1566 unsqueeze_41" [label="(4, 1, 64, 64)", style=solid]; +"1566 unsqueeze_41" -> "1567 add_46" [label="(1, 4, 1, 64, 64)", style=solid]; +"1567 add_46" -> "1568 view_75" [label="(1, 4, 12, 64, 64)", style=solid]; +"1568 view_75" -> "1569 softmax_13" [label="(4, 12, 64, 64)", style=solid]; +"1569 softmax_13" -> "1570 dropout_52" [label="(4, 12, 64, 64)", style=solid]; +"1570 dropout_52" -> "1571 matmul_27" [label="(4, 12, 64, 64)", style=solid]; +"1571 matmul_27" -> "1572 quantize_per_tensor_default_84" [label="(4, 12, 64, 32)", style=solid]; +"1572 quantize_per_tensor_default_84" -> "1573 dequantize_per_tensor_default_84" [label=None, style=solid]; +"1573 dequantize_per_tensor_default_84" -> "1574 transpose_27" [label=None, style=solid]; +"1574 transpose_27" -> "1575 reshape_61" [label="(4, 64, 12, 32)", style=solid]; +"1575 reshape_61" -> "1576 linear_83" [label="(4, 64, 384)", style=solid]; +"1576 linear_83" -> "1577 dropout_53" [label="(4, 64, 384)", style=solid]; +"1577 dropout_53" -> "1578 view_76" [label="(4, 64, 384)", style=solid]; +"1578 view_76" -> "1579 permute_63" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1579 permute_63" -> "1580 reshape_62" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1580 reshape_62" -> "1581 roll_13" [label="(1, 16, 16, 384)", style=solid]; +"1581 roll_13" -> "1582 slice_213" [label="(1, 16, 16, 384)", style=solid]; +"1582 slice_213" -> "1583 slice_214" [label="(1, 16, 16, 384)", style=solid]; +"1583 slice_214" -> "1584 slice_215" [label="(1, 14, 16, 384)", style=solid]; +"1584 slice_215" -> "1585 slice_216" [label="(1, 14, 14, 384)", style=solid]; +"1585 slice_216" -> "1586 contiguous_25" [label="(1, 14, 14, 384)", style=solid]; +"1586 contiguous_25" -> "1587 layer_norm_29" [label="(1, 14, 14, 384)", style=solid]; +"1587 layer_norm_29" -> "1588 add_47" [label="(1, 14, 14, 384)", style=solid]; +"1588 add_47" -> "1589 quantize_per_tensor_default_85" [label="(1, 14, 14, 384)", style=solid]; +"1588 add_47" -> "1599 add_48" [label="(1, 14, 14, 384)", style=solid]; +"1589 quantize_per_tensor_default_85" -> "1590 dequantize_per_tensor_default_85" [label=None, style=solid]; +"1590 dequantize_per_tensor_default_85" -> "1591 linear_84" [label=None, style=solid]; +"1591 linear_84" -> "1592 gelu_13" [label="(1, 14, 14, 1536)", style=solid]; +"1592 gelu_13" -> "1593 quantize_per_tensor_default_86" [label="(1, 14, 14, 1536)", style=solid]; +"1593 quantize_per_tensor_default_86" -> "1594 dequantize_per_tensor_default_86" [label=None, style=solid]; +"1594 dequantize_per_tensor_default_86" -> "1595 dropout_54" [label=None, style=solid]; +"1595 dropout_54" -> "1596 linear_85" [label="(1, 14, 14, 1536)", style=solid]; +"1596 linear_85" -> "1597 dropout_55" [label="(1, 14, 14, 384)", style=solid]; +"1597 dropout_55" -> "1598 layer_norm_30" [label="(1, 14, 14, 384)", style=solid]; +"1598 layer_norm_30" -> "1599 add_48" [label="(1, 14, 14, 384)", style=solid]; +"1599 add_48" -> "1600 quantize_per_tensor_default_87" [label="(1, 14, 14, 384)", style=solid]; +"1599 add_48" -> "1649 add_50" [label="(1, 14, 14, 384)", style=solid]; +"1600 quantize_per_tensor_default_87" -> "1601 dequantize_per_tensor_default_87" [label=None, style=solid]; +"1601 dequantize_per_tensor_default_87" -> "1603 pad_16" [label=None, style=solid]; +"1602 _frozen_param119" -> "1630 add_49" [label="(1, 12, 64, 64)", style=solid]; +"1603 pad_16" -> "1604 view_79" [label="(1, 16, 16, 384)", style=solid]; +"1604 view_79" -> "1605 permute_65" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1605 permute_65" -> "1606 reshape_63" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1606 reshape_63" -> "1608 linear_88" [label="(4, 64, 384)", style=solid]; +"1607 _frozen_param120" -> "1608 linear_88" [label="(1152,)", style=solid]; +"1608 linear_88" -> "1609 reshape_64" [label="(4, 64, 1152)", style=solid]; +"1609 reshape_64" -> "1610 permute_66" [label="(4, 64, 3, 12, 32)", style=solid]; +"1610 permute_66" -> "1611 select_42" [label="(3, 4, 12, 64, 32)", style=solid]; +"1610 permute_66" -> "1612 select_43" [label="(3, 4, 12, 64, 32)", style=solid]; +"1610 permute_66" -> "1613 select_44" [label="(3, 4, 12, 64, 32)", style=solid]; +"1611 select_42" -> "1614 linalg_vector_norm_28" [label="(4, 12, 64, 32)", style=solid]; +"1611 select_42" -> "1616 expand_as_28" [label="(4, 12, 64, 32)", style=solid]; +"1611 select_42" -> "1617 div_28" [label="(4, 12, 64, 32)", style=solid]; +"1612 select_43" -> "1620 linalg_vector_norm_29" [label="(4, 12, 64, 32)", style=solid]; +"1612 select_43" -> "1622 expand_as_29" [label="(4, 12, 64, 32)", style=solid]; +"1612 select_43" -> "1623 div_29" [label="(4, 12, 64, 32)", style=solid]; +"1613 select_44" -> "1633 matmul_29" [label="(4, 12, 64, 32)", style=solid]; +"1614 linalg_vector_norm_28" -> "1615 clamp_min_28" [label="(4, 12, 64, 1)", style=solid]; +"1615 clamp_min_28" -> "1616 expand_as_28" [label="(4, 12, 64, 1)", style=solid]; +"1616 expand_as_28" -> "1617 div_28" [label="(4, 12, 64, 32)", style=solid]; +"1617 div_28" -> "1618 quantize_per_tensor_default_88" [label="(4, 12, 64, 32)", style=solid]; +"1618 quantize_per_tensor_default_88" -> "1619 dequantize_per_tensor_default_88" [label=None, style=solid]; +"1619 dequantize_per_tensor_default_88" -> "1627 matmul_28" [label=None, style=solid]; +"1620 linalg_vector_norm_29" -> "1621 clamp_min_29" [label="(4, 12, 64, 1)", style=solid]; +"1621 clamp_min_29" -> "1622 expand_as_29" [label="(4, 12, 64, 1)", style=solid]; +"1622 expand_as_29" -> "1623 div_29" [label="(4, 12, 64, 32)", style=solid]; +"1623 div_29" -> "1624 quantize_per_tensor_default_89" [label="(4, 12, 64, 32)", style=solid]; +"1624 quantize_per_tensor_default_89" -> "1625 dequantize_per_tensor_default_89" [label=None, style=solid]; +"1625 dequantize_per_tensor_default_89" -> "1626 transpose_28" [label=None, style=solid]; +"1626 transpose_28" -> "1627 matmul_28" [label="(4, 12, 32, 64)", style=solid]; +"1627 matmul_28" -> "1629 mul_29" [label="(4, 12, 64, 64)", style=solid]; +"1628 _frozen_param122" -> "1629 mul_29" [label="(12, 1, 1)", style=solid]; +"1629 mul_29" -> "1630 add_49" [label="(4, 12, 64, 64)", style=solid]; +"1630 add_49" -> "1631 softmax_14" [label="(4, 12, 64, 64)", style=solid]; +"1631 softmax_14" -> "1632 dropout_56" [label="(4, 12, 64, 64)", style=solid]; +"1632 dropout_56" -> "1633 matmul_29" [label="(4, 12, 64, 64)", style=solid]; +"1633 matmul_29" -> "1634 quantize_per_tensor_default_90" [label="(4, 12, 64, 32)", style=solid]; +"1634 quantize_per_tensor_default_90" -> "1635 dequantize_per_tensor_default_90" [label=None, style=solid]; +"1635 dequantize_per_tensor_default_90" -> "1636 transpose_29" [label=None, style=solid]; +"1636 transpose_29" -> "1637 reshape_65" [label="(4, 64, 12, 32)", style=solid]; +"1637 reshape_65" -> "1638 linear_89" [label="(4, 64, 384)", style=solid]; +"1638 linear_89" -> "1639 dropout_57" [label="(4, 64, 384)", style=solid]; +"1639 dropout_57" -> "1640 view_80" [label="(4, 64, 384)", style=solid]; +"1640 view_80" -> "1641 permute_67" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1641 permute_67" -> "1642 reshape_66" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1642 reshape_66" -> "1643 slice_218" [label="(1, 16, 16, 384)", style=solid]; +"1643 slice_218" -> "1644 slice_219" [label="(1, 16, 16, 384)", style=solid]; +"1644 slice_219" -> "1645 slice_220" [label="(1, 14, 16, 384)", style=solid]; +"1645 slice_220" -> "1646 slice_221" [label="(1, 14, 14, 384)", style=solid]; +"1646 slice_221" -> "1647 contiguous_27" [label="(1, 14, 14, 384)", style=solid]; +"1647 contiguous_27" -> "1648 layer_norm_31" [label="(1, 14, 14, 384)", style=solid]; +"1648 layer_norm_31" -> "1649 add_50" [label="(1, 14, 14, 384)", style=solid]; +"1649 add_50" -> "1650 quantize_per_tensor_default_91" [label="(1, 14, 14, 384)", style=solid]; +"1649 add_50" -> "1660 add_51" [label="(1, 14, 14, 384)", style=solid]; +"1650 quantize_per_tensor_default_91" -> "1651 dequantize_per_tensor_default_91" [label=None, style=solid]; +"1651 dequantize_per_tensor_default_91" -> "1652 linear_90" [label=None, style=solid]; +"1652 linear_90" -> "1653 gelu_14" [label="(1, 14, 14, 1536)", style=solid]; +"1653 gelu_14" -> "1654 quantize_per_tensor_default_92" [label="(1, 14, 14, 1536)", style=solid]; +"1654 quantize_per_tensor_default_92" -> "1655 dequantize_per_tensor_default_92" [label=None, style=solid]; +"1655 dequantize_per_tensor_default_92" -> "1656 dropout_58" [label=None, style=solid]; +"1656 dropout_58" -> "1657 linear_91" [label="(1, 14, 14, 1536)", style=solid]; +"1657 linear_91" -> "1658 dropout_59" [label="(1, 14, 14, 384)", style=solid]; +"1658 dropout_59" -> "1659 layer_norm_32" [label="(1, 14, 14, 384)", style=solid]; +"1659 layer_norm_32" -> "1660 add_51" [label="(1, 14, 14, 384)", style=solid]; +"1660 add_51" -> "1662 pad_17" [label="(1, 14, 14, 384)", style=solid]; +"1660 add_51" -> "1728 add_54" [label="(1, 14, 14, 384)", style=solid]; +"1661 _frozen_param123" -> "1692 add_52" [label="(1, 12, 64, 64)", style=solid]; +"1662 pad_17" -> "1663 roll_14" [label="(1, 16, 16, 384)", style=solid]; +"1663 roll_14" -> "1664 view_83" [label="(1, 16, 16, 384)", style=solid]; +"1664 view_83" -> "1665 permute_69" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1665 permute_69" -> "1666 reshape_67" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1666 reshape_67" -> "1667 quantize_per_tensor_default_93" [label="(4, 64, 384)", style=solid]; +"1666 reshape_67" -> "1693 new_zeros_7" [label="(4, 64, 384)", style=solid]; +"1667 quantize_per_tensor_default_93" -> "1668 dequantize_per_tensor_default_93" [label=None, style=solid]; +"1668 dequantize_per_tensor_default_93" -> "1670 linear_94" [label=None, style=solid]; +"1669 _frozen_param124" -> "1670 linear_94" [label="(1152,)", style=solid]; +"1670 linear_94" -> "1671 reshape_68" [label="(4, 64, 1152)", style=solid]; +"1671 reshape_68" -> "1672 permute_70" [label="(4, 64, 3, 12, 32)", style=solid]; +"1672 permute_70" -> "1673 select_45" [label="(3, 4, 12, 64, 32)", style=solid]; +"1672 permute_70" -> "1674 select_46" [label="(3, 4, 12, 64, 32)", style=solid]; +"1672 permute_70" -> "1675 select_47" [label="(3, 4, 12, 64, 32)", style=solid]; +"1673 select_45" -> "1676 linalg_vector_norm_30" [label="(4, 12, 64, 32)", style=solid]; +"1673 select_45" -> "1678 expand_as_30" [label="(4, 12, 64, 32)", style=solid]; +"1673 select_45" -> "1679 div_30" [label="(4, 12, 64, 32)", style=solid]; +"1674 select_46" -> "1682 linalg_vector_norm_31" [label="(4, 12, 64, 32)", style=solid]; +"1674 select_46" -> "1684 expand_as_31" [label="(4, 12, 64, 32)", style=solid]; +"1674 select_46" -> "1685 div_31" [label="(4, 12, 64, 32)", style=solid]; +"1675 select_47" -> "1711 matmul_31" [label="(4, 12, 64, 32)", style=solid]; +"1676 linalg_vector_norm_30" -> "1677 clamp_min_30" [label="(4, 12, 64, 1)", style=solid]; +"1677 clamp_min_30" -> "1678 expand_as_30" [label="(4, 12, 64, 1)", style=solid]; +"1678 expand_as_30" -> "1679 div_30" [label="(4, 12, 64, 32)", style=solid]; +"1679 div_30" -> "1680 quantize_per_tensor_default_94" [label="(4, 12, 64, 32)", style=solid]; +"1680 quantize_per_tensor_default_94" -> "1681 dequantize_per_tensor_default_94" [label=None, style=solid]; +"1681 dequantize_per_tensor_default_94" -> "1689 matmul_30" [label=None, style=solid]; +"1682 linalg_vector_norm_31" -> "1683 clamp_min_31" [label="(4, 12, 64, 1)", style=solid]; +"1683 clamp_min_31" -> "1684 expand_as_31" [label="(4, 12, 64, 1)", style=solid]; +"1684 expand_as_31" -> "1685 div_31" [label="(4, 12, 64, 32)", style=solid]; +"1685 div_31" -> "1686 quantize_per_tensor_default_95" [label="(4, 12, 64, 32)", style=solid]; +"1686 quantize_per_tensor_default_95" -> "1687 dequantize_per_tensor_default_95" [label=None, style=solid]; +"1687 dequantize_per_tensor_default_95" -> "1688 transpose_30" [label=None, style=solid]; +"1688 transpose_30" -> "1689 matmul_30" [label="(4, 12, 32, 64)", style=solid]; +"1689 matmul_30" -> "1691 mul_31" [label="(4, 12, 64, 64)", style=solid]; +"1690 _frozen_param126" -> "1691 mul_31" [label="(12, 1, 1)", style=solid]; +"1691 mul_31" -> "1692 add_52" [label="(4, 12, 64, 64)", style=solid]; +"1692 add_52" -> "1704 view_85" [label="(4, 12, 64, 64)", style=solid]; +"1693 new_zeros_7" -> "1694 view_84" [label="(16, 16)", style=solid]; +"1694 view_84" -> "1695 permute_71" [label="(2, 8, 2, 8)", style=solid]; +"1695 permute_71" -> "1696 reshape_69" [label="(2, 2, 8, 8)", style=solid]; +"1696 reshape_69" -> "1697 unsqueeze_44" [label="(4, 64)", style=solid]; +"1696 reshape_69" -> "1698 unsqueeze_45" [label="(4, 64)", style=solid]; +"1697 unsqueeze_44" -> "1699 sub_7" [label="(4, 1, 64)", style=solid]; +"1698 unsqueeze_45" -> "1699 sub_7" [label="(4, 64, 1)", style=solid]; +"1699 sub_7" -> "1700 ne_7" [label="(4, 64, 64)", style=solid]; +"1699 sub_7" -> "1701 masked_fill_14" [label="(4, 64, 64)", style=solid]; +"1699 sub_7" -> "1702 eq_7" [label="(4, 64, 64)", style=solid]; +"1700 ne_7" -> "1701 masked_fill_14" [label="(4, 64, 64)", style=solid]; +"1701 masked_fill_14" -> "1703 masked_fill_15" [label="(4, 64, 64)", style=solid]; +"1702 eq_7" -> "1703 masked_fill_15" [label="(4, 64, 64)", style=solid]; +"1703 masked_fill_15" -> "1705 unsqueeze_46" [label="(4, 64, 64)", style=solid]; +"1704 view_85" -> "1707 add_53" [label="(1, 4, 12, 64, 64)", style=solid]; +"1705 unsqueeze_46" -> "1706 unsqueeze_47" [label="(4, 1, 64, 64)", style=solid]; +"1706 unsqueeze_47" -> "1707 add_53" [label="(1, 4, 1, 64, 64)", style=solid]; +"1707 add_53" -> "1708 view_86" [label="(1, 4, 12, 64, 64)", style=solid]; +"1708 view_86" -> "1709 softmax_15" [label="(4, 12, 64, 64)", style=solid]; +"1709 softmax_15" -> "1710 dropout_60" [label="(4, 12, 64, 64)", style=solid]; +"1710 dropout_60" -> "1711 matmul_31" [label="(4, 12, 64, 64)", style=solid]; +"1711 matmul_31" -> "1712 quantize_per_tensor_default_96" [label="(4, 12, 64, 32)", style=solid]; +"1712 quantize_per_tensor_default_96" -> "1713 dequantize_per_tensor_default_96" [label=None, style=solid]; +"1713 dequantize_per_tensor_default_96" -> "1714 transpose_31" [label=None, style=solid]; +"1714 transpose_31" -> "1715 reshape_70" [label="(4, 64, 12, 32)", style=solid]; +"1715 reshape_70" -> "1716 linear_95" [label="(4, 64, 384)", style=solid]; +"1716 linear_95" -> "1717 dropout_61" [label="(4, 64, 384)", style=solid]; +"1717 dropout_61" -> "1718 view_87" [label="(4, 64, 384)", style=solid]; +"1718 view_87" -> "1719 permute_72" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1719 permute_72" -> "1720 reshape_71" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1720 reshape_71" -> "1721 roll_15" [label="(1, 16, 16, 384)", style=solid]; +"1721 roll_15" -> "1722 slice_241" [label="(1, 16, 16, 384)", style=solid]; +"1722 slice_241" -> "1723 slice_242" [label="(1, 16, 16, 384)", style=solid]; +"1723 slice_242" -> "1724 slice_243" [label="(1, 14, 16, 384)", style=solid]; +"1724 slice_243" -> "1725 slice_244" [label="(1, 14, 14, 384)", style=solid]; +"1725 slice_244" -> "1726 contiguous_29" [label="(1, 14, 14, 384)", style=solid]; +"1726 contiguous_29" -> "1727 layer_norm_33" [label="(1, 14, 14, 384)", style=solid]; +"1727 layer_norm_33" -> "1728 add_54" [label="(1, 14, 14, 384)", style=solid]; +"1728 add_54" -> "1729 quantize_per_tensor_default_97" [label="(1, 14, 14, 384)", style=solid]; +"1728 add_54" -> "1739 add_55" [label="(1, 14, 14, 384)", style=solid]; +"1729 quantize_per_tensor_default_97" -> "1730 dequantize_per_tensor_default_97" [label=None, style=solid]; +"1730 dequantize_per_tensor_default_97" -> "1731 linear_96" [label=None, style=solid]; +"1731 linear_96" -> "1732 gelu_15" [label="(1, 14, 14, 1536)", style=solid]; +"1732 gelu_15" -> "1733 quantize_per_tensor_default_98" [label="(1, 14, 14, 1536)", style=solid]; +"1733 quantize_per_tensor_default_98" -> "1734 dequantize_per_tensor_default_98" [label=None, style=solid]; +"1734 dequantize_per_tensor_default_98" -> "1735 dropout_62" [label=None, style=solid]; +"1735 dropout_62" -> "1736 linear_97" [label="(1, 14, 14, 1536)", style=solid]; +"1736 linear_97" -> "1737 dropout_63" [label="(1, 14, 14, 384)", style=solid]; +"1737 dropout_63" -> "1738 layer_norm_34" [label="(1, 14, 14, 384)", style=solid]; +"1738 layer_norm_34" -> "1739 add_55" [label="(1, 14, 14, 384)", style=solid]; +"1739 add_55" -> "1740 quantize_per_tensor_default_99" [label="(1, 14, 14, 384)", style=solid]; +"1739 add_55" -> "1789 add_57" [label="(1, 14, 14, 384)", style=solid]; +"1740 quantize_per_tensor_default_99" -> "1741 dequantize_per_tensor_default_99" [label=None, style=solid]; +"1741 dequantize_per_tensor_default_99" -> "1743 pad_18" [label=None, style=solid]; +"1742 _frozen_param136" -> "1770 add_56" [label="(1, 12, 64, 64)", style=solid]; +"1743 pad_18" -> "1744 view_90" [label="(1, 16, 16, 384)", style=solid]; +"1744 view_90" -> "1745 permute_74" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1745 permute_74" -> "1746 reshape_72" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1746 reshape_72" -> "1748 linear_100" [label="(4, 64, 384)", style=solid]; +"1747 _frozen_param137" -> "1748 linear_100" [label="(1152,)", style=solid]; +"1748 linear_100" -> "1749 reshape_73" [label="(4, 64, 1152)", style=solid]; +"1749 reshape_73" -> "1750 permute_75" [label="(4, 64, 3, 12, 32)", style=solid]; +"1750 permute_75" -> "1751 select_48" [label="(3, 4, 12, 64, 32)", style=solid]; +"1750 permute_75" -> "1752 select_49" [label="(3, 4, 12, 64, 32)", style=solid]; +"1750 permute_75" -> "1753 select_50" [label="(3, 4, 12, 64, 32)", style=solid]; +"1751 select_48" -> "1754 linalg_vector_norm_32" [label="(4, 12, 64, 32)", style=solid]; +"1751 select_48" -> "1756 expand_as_32" [label="(4, 12, 64, 32)", style=solid]; +"1751 select_48" -> "1757 div_32" [label="(4, 12, 64, 32)", style=solid]; +"1752 select_49" -> "1760 linalg_vector_norm_33" [label="(4, 12, 64, 32)", style=solid]; +"1752 select_49" -> "1762 expand_as_33" [label="(4, 12, 64, 32)", style=solid]; +"1752 select_49" -> "1763 div_33" [label="(4, 12, 64, 32)", style=solid]; +"1753 select_50" -> "1773 matmul_33" [label="(4, 12, 64, 32)", style=solid]; +"1754 linalg_vector_norm_32" -> "1755 clamp_min_32" [label="(4, 12, 64, 1)", style=solid]; +"1755 clamp_min_32" -> "1756 expand_as_32" [label="(4, 12, 64, 1)", style=solid]; +"1756 expand_as_32" -> "1757 div_32" [label="(4, 12, 64, 32)", style=solid]; +"1757 div_32" -> "1758 quantize_per_tensor_default_100" [label="(4, 12, 64, 32)", style=solid]; +"1758 quantize_per_tensor_default_100" -> "1759 dequantize_per_tensor_default_100" [label=None, style=solid]; +"1759 dequantize_per_tensor_default_100" -> "1767 matmul_32" [label=None, style=solid]; +"1760 linalg_vector_norm_33" -> "1761 clamp_min_33" [label="(4, 12, 64, 1)", style=solid]; +"1761 clamp_min_33" -> "1762 expand_as_33" [label="(4, 12, 64, 1)", style=solid]; +"1762 expand_as_33" -> "1763 div_33" [label="(4, 12, 64, 32)", style=solid]; +"1763 div_33" -> "1764 quantize_per_tensor_default_101" [label="(4, 12, 64, 32)", style=solid]; +"1764 quantize_per_tensor_default_101" -> "1765 dequantize_per_tensor_default_101" [label=None, style=solid]; +"1765 dequantize_per_tensor_default_101" -> "1766 transpose_32" [label=None, style=solid]; +"1766 transpose_32" -> "1767 matmul_32" [label="(4, 12, 32, 64)", style=solid]; +"1767 matmul_32" -> "1769 mul_33" [label="(4, 12, 64, 64)", style=solid]; +"1768 _frozen_param139" -> "1769 mul_33" [label="(12, 1, 1)", style=solid]; +"1769 mul_33" -> "1770 add_56" [label="(4, 12, 64, 64)", style=solid]; +"1770 add_56" -> "1771 softmax_16" [label="(4, 12, 64, 64)", style=solid]; +"1771 softmax_16" -> "1772 dropout_64" [label="(4, 12, 64, 64)", style=solid]; +"1772 dropout_64" -> "1773 matmul_33" [label="(4, 12, 64, 64)", style=solid]; +"1773 matmul_33" -> "1774 quantize_per_tensor_default_102" [label="(4, 12, 64, 32)", style=solid]; +"1774 quantize_per_tensor_default_102" -> "1775 dequantize_per_tensor_default_102" [label=None, style=solid]; +"1775 dequantize_per_tensor_default_102" -> "1776 transpose_33" [label=None, style=solid]; +"1776 transpose_33" -> "1777 reshape_74" [label="(4, 64, 12, 32)", style=solid]; +"1777 reshape_74" -> "1778 linear_101" [label="(4, 64, 384)", style=solid]; +"1778 linear_101" -> "1779 dropout_65" [label="(4, 64, 384)", style=solid]; +"1779 dropout_65" -> "1780 view_91" [label="(4, 64, 384)", style=solid]; +"1780 view_91" -> "1781 permute_76" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1781 permute_76" -> "1782 reshape_75" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1782 reshape_75" -> "1783 slice_246" [label="(1, 16, 16, 384)", style=solid]; +"1783 slice_246" -> "1784 slice_247" [label="(1, 16, 16, 384)", style=solid]; +"1784 slice_247" -> "1785 slice_248" [label="(1, 14, 16, 384)", style=solid]; +"1785 slice_248" -> "1786 slice_249" [label="(1, 14, 14, 384)", style=solid]; +"1786 slice_249" -> "1787 contiguous_31" [label="(1, 14, 14, 384)", style=solid]; +"1787 contiguous_31" -> "1788 layer_norm_35" [label="(1, 14, 14, 384)", style=solid]; +"1788 layer_norm_35" -> "1789 add_57" [label="(1, 14, 14, 384)", style=solid]; +"1789 add_57" -> "1790 quantize_per_tensor_default_103" [label="(1, 14, 14, 384)", style=solid]; +"1789 add_57" -> "1800 add_58" [label="(1, 14, 14, 384)", style=solid]; +"1790 quantize_per_tensor_default_103" -> "1791 dequantize_per_tensor_default_103" [label=None, style=solid]; +"1791 dequantize_per_tensor_default_103" -> "1792 linear_102" [label=None, style=solid]; +"1792 linear_102" -> "1793 gelu_16" [label="(1, 14, 14, 1536)", style=solid]; +"1793 gelu_16" -> "1794 quantize_per_tensor_default_104" [label="(1, 14, 14, 1536)", style=solid]; +"1794 quantize_per_tensor_default_104" -> "1795 dequantize_per_tensor_default_104" [label=None, style=solid]; +"1795 dequantize_per_tensor_default_104" -> "1796 dropout_66" [label=None, style=solid]; +"1796 dropout_66" -> "1797 linear_103" [label="(1, 14, 14, 1536)", style=solid]; +"1797 linear_103" -> "1798 dropout_67" [label="(1, 14, 14, 384)", style=solid]; +"1798 dropout_67" -> "1799 layer_norm_36" [label="(1, 14, 14, 384)", style=solid]; +"1799 layer_norm_36" -> "1800 add_58" [label="(1, 14, 14, 384)", style=solid]; +"1800 add_58" -> "1802 pad_19" [label="(1, 14, 14, 384)", style=solid]; +"1800 add_58" -> "1868 add_61" [label="(1, 14, 14, 384)", style=solid]; +"1801 _frozen_param140" -> "1832 add_59" [label="(1, 12, 64, 64)", style=solid]; +"1802 pad_19" -> "1803 roll_16" [label="(1, 16, 16, 384)", style=solid]; +"1803 roll_16" -> "1804 view_94" [label="(1, 16, 16, 384)", style=solid]; +"1804 view_94" -> "1805 permute_78" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1805 permute_78" -> "1806 reshape_76" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1806 reshape_76" -> "1807 quantize_per_tensor_default_105" [label="(4, 64, 384)", style=solid]; +"1806 reshape_76" -> "1833 new_zeros_8" [label="(4, 64, 384)", style=solid]; +"1807 quantize_per_tensor_default_105" -> "1808 dequantize_per_tensor_default_105" [label=None, style=solid]; +"1808 dequantize_per_tensor_default_105" -> "1810 linear_106" [label=None, style=solid]; +"1809 _frozen_param141" -> "1810 linear_106" [label="(1152,)", style=solid]; +"1810 linear_106" -> "1811 reshape_77" [label="(4, 64, 1152)", style=solid]; +"1811 reshape_77" -> "1812 permute_79" [label="(4, 64, 3, 12, 32)", style=solid]; +"1812 permute_79" -> "1813 select_51" [label="(3, 4, 12, 64, 32)", style=solid]; +"1812 permute_79" -> "1814 select_52" [label="(3, 4, 12, 64, 32)", style=solid]; +"1812 permute_79" -> "1815 select_53" [label="(3, 4, 12, 64, 32)", style=solid]; +"1813 select_51" -> "1816 linalg_vector_norm_34" [label="(4, 12, 64, 32)", style=solid]; +"1813 select_51" -> "1818 expand_as_34" [label="(4, 12, 64, 32)", style=solid]; +"1813 select_51" -> "1819 div_34" [label="(4, 12, 64, 32)", style=solid]; +"1814 select_52" -> "1822 linalg_vector_norm_35" [label="(4, 12, 64, 32)", style=solid]; +"1814 select_52" -> "1824 expand_as_35" [label="(4, 12, 64, 32)", style=solid]; +"1814 select_52" -> "1825 div_35" [label="(4, 12, 64, 32)", style=solid]; +"1815 select_53" -> "1851 matmul_35" [label="(4, 12, 64, 32)", style=solid]; +"1816 linalg_vector_norm_34" -> "1817 clamp_min_34" [label="(4, 12, 64, 1)", style=solid]; +"1817 clamp_min_34" -> "1818 expand_as_34" [label="(4, 12, 64, 1)", style=solid]; +"1818 expand_as_34" -> "1819 div_34" [label="(4, 12, 64, 32)", style=solid]; +"1819 div_34" -> "1820 quantize_per_tensor_default_106" [label="(4, 12, 64, 32)", style=solid]; +"1820 quantize_per_tensor_default_106" -> "1821 dequantize_per_tensor_default_106" [label=None, style=solid]; +"1821 dequantize_per_tensor_default_106" -> "1829 matmul_34" [label=None, style=solid]; +"1822 linalg_vector_norm_35" -> "1823 clamp_min_35" [label="(4, 12, 64, 1)", style=solid]; +"1823 clamp_min_35" -> "1824 expand_as_35" [label="(4, 12, 64, 1)", style=solid]; +"1824 expand_as_35" -> "1825 div_35" [label="(4, 12, 64, 32)", style=solid]; +"1825 div_35" -> "1826 quantize_per_tensor_default_107" [label="(4, 12, 64, 32)", style=solid]; +"1826 quantize_per_tensor_default_107" -> "1827 dequantize_per_tensor_default_107" [label=None, style=solid]; +"1827 dequantize_per_tensor_default_107" -> "1828 transpose_34" [label=None, style=solid]; +"1828 transpose_34" -> "1829 matmul_34" [label="(4, 12, 32, 64)", style=solid]; +"1829 matmul_34" -> "1831 mul_35" [label="(4, 12, 64, 64)", style=solid]; +"1830 _frozen_param143" -> "1831 mul_35" [label="(12, 1, 1)", style=solid]; +"1831 mul_35" -> "1832 add_59" [label="(4, 12, 64, 64)", style=solid]; +"1832 add_59" -> "1844 view_96" [label="(4, 12, 64, 64)", style=solid]; +"1833 new_zeros_8" -> "1834 view_95" [label="(16, 16)", style=solid]; +"1834 view_95" -> "1835 permute_80" [label="(2, 8, 2, 8)", style=solid]; +"1835 permute_80" -> "1836 reshape_78" [label="(2, 2, 8, 8)", style=solid]; +"1836 reshape_78" -> "1837 unsqueeze_50" [label="(4, 64)", style=solid]; +"1836 reshape_78" -> "1838 unsqueeze_51" [label="(4, 64)", style=solid]; +"1837 unsqueeze_50" -> "1839 sub_8" [label="(4, 1, 64)", style=solid]; +"1838 unsqueeze_51" -> "1839 sub_8" [label="(4, 64, 1)", style=solid]; +"1839 sub_8" -> "1840 ne_8" [label="(4, 64, 64)", style=solid]; +"1839 sub_8" -> "1841 masked_fill_16" [label="(4, 64, 64)", style=solid]; +"1839 sub_8" -> "1842 eq_8" [label="(4, 64, 64)", style=solid]; +"1840 ne_8" -> "1841 masked_fill_16" [label="(4, 64, 64)", style=solid]; +"1841 masked_fill_16" -> "1843 masked_fill_17" [label="(4, 64, 64)", style=solid]; +"1842 eq_8" -> "1843 masked_fill_17" [label="(4, 64, 64)", style=solid]; +"1843 masked_fill_17" -> "1845 unsqueeze_52" [label="(4, 64, 64)", style=solid]; +"1844 view_96" -> "1847 add_60" [label="(1, 4, 12, 64, 64)", style=solid]; +"1845 unsqueeze_52" -> "1846 unsqueeze_53" [label="(4, 1, 64, 64)", style=solid]; +"1846 unsqueeze_53" -> "1847 add_60" [label="(1, 4, 1, 64, 64)", style=solid]; +"1847 add_60" -> "1848 view_97" [label="(1, 4, 12, 64, 64)", style=solid]; +"1848 view_97" -> "1849 softmax_17" [label="(4, 12, 64, 64)", style=solid]; +"1849 softmax_17" -> "1850 dropout_68" [label="(4, 12, 64, 64)", style=solid]; +"1850 dropout_68" -> "1851 matmul_35" [label="(4, 12, 64, 64)", style=solid]; +"1851 matmul_35" -> "1852 quantize_per_tensor_default_108" [label="(4, 12, 64, 32)", style=solid]; +"1852 quantize_per_tensor_default_108" -> "1853 dequantize_per_tensor_default_108" [label=None, style=solid]; +"1853 dequantize_per_tensor_default_108" -> "1854 transpose_35" [label=None, style=solid]; +"1854 transpose_35" -> "1855 reshape_79" [label="(4, 64, 12, 32)", style=solid]; +"1855 reshape_79" -> "1856 linear_107" [label="(4, 64, 384)", style=solid]; +"1856 linear_107" -> "1857 dropout_69" [label="(4, 64, 384)", style=solid]; +"1857 dropout_69" -> "1858 view_98" [label="(4, 64, 384)", style=solid]; +"1858 view_98" -> "1859 permute_81" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1859 permute_81" -> "1860 reshape_80" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1860 reshape_80" -> "1861 roll_17" [label="(1, 16, 16, 384)", style=solid]; +"1861 roll_17" -> "1862 slice_269" [label="(1, 16, 16, 384)", style=solid]; +"1862 slice_269" -> "1863 slice_270" [label="(1, 16, 16, 384)", style=solid]; +"1863 slice_270" -> "1864 slice_271" [label="(1, 14, 16, 384)", style=solid]; +"1864 slice_271" -> "1865 slice_272" [label="(1, 14, 14, 384)", style=solid]; +"1865 slice_272" -> "1866 contiguous_33" [label="(1, 14, 14, 384)", style=solid]; +"1866 contiguous_33" -> "1867 layer_norm_37" [label="(1, 14, 14, 384)", style=solid]; +"1867 layer_norm_37" -> "1868 add_61" [label="(1, 14, 14, 384)", style=solid]; +"1868 add_61" -> "1869 quantize_per_tensor_default_109" [label="(1, 14, 14, 384)", style=solid]; +"1868 add_61" -> "1879 add_62" [label="(1, 14, 14, 384)", style=solid]; +"1869 quantize_per_tensor_default_109" -> "1870 dequantize_per_tensor_default_109" [label=None, style=solid]; +"1870 dequantize_per_tensor_default_109" -> "1871 linear_108" [label=None, style=solid]; +"1871 linear_108" -> "1872 gelu_17" [label="(1, 14, 14, 1536)", style=solid]; +"1872 gelu_17" -> "1873 quantize_per_tensor_default_110" [label="(1, 14, 14, 1536)", style=solid]; +"1873 quantize_per_tensor_default_110" -> "1874 dequantize_per_tensor_default_110" [label=None, style=solid]; +"1874 dequantize_per_tensor_default_110" -> "1875 dropout_70" [label=None, style=solid]; +"1875 dropout_70" -> "1876 linear_109" [label="(1, 14, 14, 1536)", style=solid]; +"1876 linear_109" -> "1877 dropout_71" [label="(1, 14, 14, 384)", style=solid]; +"1877 dropout_71" -> "1878 layer_norm_38" [label="(1, 14, 14, 384)", style=solid]; +"1878 layer_norm_38" -> "1879 add_62" [label="(1, 14, 14, 384)", style=solid]; +"1879 add_62" -> "1880 quantize_per_tensor_default_111" [label="(1, 14, 14, 384)", style=solid]; +"1879 add_62" -> "1929 add_64" [label="(1, 14, 14, 384)", style=solid]; +"1880 quantize_per_tensor_default_111" -> "1881 dequantize_per_tensor_default_111" [label=None, style=solid]; +"1881 dequantize_per_tensor_default_111" -> "1883 pad_20" [label=None, style=solid]; +"1882 _frozen_param153" -> "1910 add_63" [label="(1, 12, 64, 64)", style=solid]; +"1883 pad_20" -> "1884 view_101" [label="(1, 16, 16, 384)", style=solid]; +"1884 view_101" -> "1885 permute_83" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1885 permute_83" -> "1886 reshape_81" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1886 reshape_81" -> "1888 linear_112" [label="(4, 64, 384)", style=solid]; +"1887 _frozen_param154" -> "1888 linear_112" [label="(1152,)", style=solid]; +"1888 linear_112" -> "1889 reshape_82" [label="(4, 64, 1152)", style=solid]; +"1889 reshape_82" -> "1890 permute_84" [label="(4, 64, 3, 12, 32)", style=solid]; +"1890 permute_84" -> "1891 select_54" [label="(3, 4, 12, 64, 32)", style=solid]; +"1890 permute_84" -> "1892 select_55" [label="(3, 4, 12, 64, 32)", style=solid]; +"1890 permute_84" -> "1893 select_56" [label="(3, 4, 12, 64, 32)", style=solid]; +"1891 select_54" -> "1894 linalg_vector_norm_36" [label="(4, 12, 64, 32)", style=solid]; +"1891 select_54" -> "1896 expand_as_36" [label="(4, 12, 64, 32)", style=solid]; +"1891 select_54" -> "1897 div_36" [label="(4, 12, 64, 32)", style=solid]; +"1892 select_55" -> "1900 linalg_vector_norm_37" [label="(4, 12, 64, 32)", style=solid]; +"1892 select_55" -> "1902 expand_as_37" [label="(4, 12, 64, 32)", style=solid]; +"1892 select_55" -> "1903 div_37" [label="(4, 12, 64, 32)", style=solid]; +"1893 select_56" -> "1913 matmul_37" [label="(4, 12, 64, 32)", style=solid]; +"1894 linalg_vector_norm_36" -> "1895 clamp_min_36" [label="(4, 12, 64, 1)", style=solid]; +"1895 clamp_min_36" -> "1896 expand_as_36" [label="(4, 12, 64, 1)", style=solid]; +"1896 expand_as_36" -> "1897 div_36" [label="(4, 12, 64, 32)", style=solid]; +"1897 div_36" -> "1898 quantize_per_tensor_default_112" [label="(4, 12, 64, 32)", style=solid]; +"1898 quantize_per_tensor_default_112" -> "1899 dequantize_per_tensor_default_112" [label=None, style=solid]; +"1899 dequantize_per_tensor_default_112" -> "1907 matmul_36" [label=None, style=solid]; +"1900 linalg_vector_norm_37" -> "1901 clamp_min_37" [label="(4, 12, 64, 1)", style=solid]; +"1901 clamp_min_37" -> "1902 expand_as_37" [label="(4, 12, 64, 1)", style=solid]; +"1902 expand_as_37" -> "1903 div_37" [label="(4, 12, 64, 32)", style=solid]; +"1903 div_37" -> "1904 quantize_per_tensor_default_113" [label="(4, 12, 64, 32)", style=solid]; +"1904 quantize_per_tensor_default_113" -> "1905 dequantize_per_tensor_default_113" [label=None, style=solid]; +"1905 dequantize_per_tensor_default_113" -> "1906 transpose_36" [label=None, style=solid]; +"1906 transpose_36" -> "1907 matmul_36" [label="(4, 12, 32, 64)", style=solid]; +"1907 matmul_36" -> "1909 mul_37" [label="(4, 12, 64, 64)", style=solid]; +"1908 _frozen_param156" -> "1909 mul_37" [label="(12, 1, 1)", style=solid]; +"1909 mul_37" -> "1910 add_63" [label="(4, 12, 64, 64)", style=solid]; +"1910 add_63" -> "1911 softmax_18" [label="(4, 12, 64, 64)", style=solid]; +"1911 softmax_18" -> "1912 dropout_72" [label="(4, 12, 64, 64)", style=solid]; +"1912 dropout_72" -> "1913 matmul_37" [label="(4, 12, 64, 64)", style=solid]; +"1913 matmul_37" -> "1914 quantize_per_tensor_default_114" [label="(4, 12, 64, 32)", style=solid]; +"1914 quantize_per_tensor_default_114" -> "1915 dequantize_per_tensor_default_114" [label=None, style=solid]; +"1915 dequantize_per_tensor_default_114" -> "1916 transpose_37" [label=None, style=solid]; +"1916 transpose_37" -> "1917 reshape_83" [label="(4, 64, 12, 32)", style=solid]; +"1917 reshape_83" -> "1918 linear_113" [label="(4, 64, 384)", style=solid]; +"1918 linear_113" -> "1919 dropout_73" [label="(4, 64, 384)", style=solid]; +"1919 dropout_73" -> "1920 view_102" [label="(4, 64, 384)", style=solid]; +"1920 view_102" -> "1921 permute_85" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1921 permute_85" -> "1922 reshape_84" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1922 reshape_84" -> "1923 slice_274" [label="(1, 16, 16, 384)", style=solid]; +"1923 slice_274" -> "1924 slice_275" [label="(1, 16, 16, 384)", style=solid]; +"1924 slice_275" -> "1925 slice_276" [label="(1, 14, 16, 384)", style=solid]; +"1925 slice_276" -> "1926 slice_277" [label="(1, 14, 14, 384)", style=solid]; +"1926 slice_277" -> "1927 contiguous_35" [label="(1, 14, 14, 384)", style=solid]; +"1927 contiguous_35" -> "1928 layer_norm_39" [label="(1, 14, 14, 384)", style=solid]; +"1928 layer_norm_39" -> "1929 add_64" [label="(1, 14, 14, 384)", style=solid]; +"1929 add_64" -> "1930 quantize_per_tensor_default_115" [label="(1, 14, 14, 384)", style=solid]; +"1929 add_64" -> "1940 add_65" [label="(1, 14, 14, 384)", style=solid]; +"1930 quantize_per_tensor_default_115" -> "1931 dequantize_per_tensor_default_115" [label=None, style=solid]; +"1931 dequantize_per_tensor_default_115" -> "1932 linear_114" [label=None, style=solid]; +"1932 linear_114" -> "1933 gelu_18" [label="(1, 14, 14, 1536)", style=solid]; +"1933 gelu_18" -> "1934 quantize_per_tensor_default_116" [label="(1, 14, 14, 1536)", style=solid]; +"1934 quantize_per_tensor_default_116" -> "1935 dequantize_per_tensor_default_116" [label=None, style=solid]; +"1935 dequantize_per_tensor_default_116" -> "1936 dropout_74" [label=None, style=solid]; +"1936 dropout_74" -> "1937 linear_115" [label="(1, 14, 14, 1536)", style=solid]; +"1937 linear_115" -> "1938 dropout_75" [label="(1, 14, 14, 384)", style=solid]; +"1938 dropout_75" -> "1939 layer_norm_40" [label="(1, 14, 14, 384)", style=solid]; +"1939 layer_norm_40" -> "1940 add_65" [label="(1, 14, 14, 384)", style=solid]; +"1940 add_65" -> "1942 pad_21" [label="(1, 14, 14, 384)", style=solid]; +"1940 add_65" -> "2008 add_68" [label="(1, 14, 14, 384)", style=solid]; +"1941 _frozen_param157" -> "1972 add_66" [label="(1, 12, 64, 64)", style=solid]; +"1942 pad_21" -> "1943 roll_18" [label="(1, 16, 16, 384)", style=solid]; +"1943 roll_18" -> "1944 view_105" [label="(1, 16, 16, 384)", style=solid]; +"1944 view_105" -> "1945 permute_87" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"1945 permute_87" -> "1946 reshape_85" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1946 reshape_85" -> "1947 quantize_per_tensor_default_117" [label="(4, 64, 384)", style=solid]; +"1946 reshape_85" -> "1973 new_zeros_9" [label="(4, 64, 384)", style=solid]; +"1947 quantize_per_tensor_default_117" -> "1948 dequantize_per_tensor_default_117" [label=None, style=solid]; +"1948 dequantize_per_tensor_default_117" -> "1950 linear_118" [label=None, style=solid]; +"1949 _frozen_param158" -> "1950 linear_118" [label="(1152,)", style=solid]; +"1950 linear_118" -> "1951 reshape_86" [label="(4, 64, 1152)", style=solid]; +"1951 reshape_86" -> "1952 permute_88" [label="(4, 64, 3, 12, 32)", style=solid]; +"1952 permute_88" -> "1953 select_57" [label="(3, 4, 12, 64, 32)", style=solid]; +"1952 permute_88" -> "1954 select_58" [label="(3, 4, 12, 64, 32)", style=solid]; +"1952 permute_88" -> "1955 select_59" [label="(3, 4, 12, 64, 32)", style=solid]; +"1953 select_57" -> "1956 linalg_vector_norm_38" [label="(4, 12, 64, 32)", style=solid]; +"1953 select_57" -> "1958 expand_as_38" [label="(4, 12, 64, 32)", style=solid]; +"1953 select_57" -> "1959 div_38" [label="(4, 12, 64, 32)", style=solid]; +"1954 select_58" -> "1962 linalg_vector_norm_39" [label="(4, 12, 64, 32)", style=solid]; +"1954 select_58" -> "1964 expand_as_39" [label="(4, 12, 64, 32)", style=solid]; +"1954 select_58" -> "1965 div_39" [label="(4, 12, 64, 32)", style=solid]; +"1955 select_59" -> "1991 matmul_39" [label="(4, 12, 64, 32)", style=solid]; +"1956 linalg_vector_norm_38" -> "1957 clamp_min_38" [label="(4, 12, 64, 1)", style=solid]; +"1957 clamp_min_38" -> "1958 expand_as_38" [label="(4, 12, 64, 1)", style=solid]; +"1958 expand_as_38" -> "1959 div_38" [label="(4, 12, 64, 32)", style=solid]; +"1959 div_38" -> "1960 quantize_per_tensor_default_118" [label="(4, 12, 64, 32)", style=solid]; +"1960 quantize_per_tensor_default_118" -> "1961 dequantize_per_tensor_default_118" [label=None, style=solid]; +"1961 dequantize_per_tensor_default_118" -> "1969 matmul_38" [label=None, style=solid]; +"1962 linalg_vector_norm_39" -> "1963 clamp_min_39" [label="(4, 12, 64, 1)", style=solid]; +"1963 clamp_min_39" -> "1964 expand_as_39" [label="(4, 12, 64, 1)", style=solid]; +"1964 expand_as_39" -> "1965 div_39" [label="(4, 12, 64, 32)", style=solid]; +"1965 div_39" -> "1966 quantize_per_tensor_default_119" [label="(4, 12, 64, 32)", style=solid]; +"1966 quantize_per_tensor_default_119" -> "1967 dequantize_per_tensor_default_119" [label=None, style=solid]; +"1967 dequantize_per_tensor_default_119" -> "1968 transpose_38" [label=None, style=solid]; +"1968 transpose_38" -> "1969 matmul_38" [label="(4, 12, 32, 64)", style=solid]; +"1969 matmul_38" -> "1971 mul_39" [label="(4, 12, 64, 64)", style=solid]; +"1970 _frozen_param160" -> "1971 mul_39" [label="(12, 1, 1)", style=solid]; +"1971 mul_39" -> "1972 add_66" [label="(4, 12, 64, 64)", style=solid]; +"1972 add_66" -> "1984 view_107" [label="(4, 12, 64, 64)", style=solid]; +"1973 new_zeros_9" -> "1974 view_106" [label="(16, 16)", style=solid]; +"1974 view_106" -> "1975 permute_89" [label="(2, 8, 2, 8)", style=solid]; +"1975 permute_89" -> "1976 reshape_87" [label="(2, 2, 8, 8)", style=solid]; +"1976 reshape_87" -> "1977 unsqueeze_56" [label="(4, 64)", style=solid]; +"1976 reshape_87" -> "1978 unsqueeze_57" [label="(4, 64)", style=solid]; +"1977 unsqueeze_56" -> "1979 sub_9" [label="(4, 1, 64)", style=solid]; +"1978 unsqueeze_57" -> "1979 sub_9" [label="(4, 64, 1)", style=solid]; +"1979 sub_9" -> "1980 ne_9" [label="(4, 64, 64)", style=solid]; +"1979 sub_9" -> "1981 masked_fill_18" [label="(4, 64, 64)", style=solid]; +"1979 sub_9" -> "1982 eq_9" [label="(4, 64, 64)", style=solid]; +"1980 ne_9" -> "1981 masked_fill_18" [label="(4, 64, 64)", style=solid]; +"1981 masked_fill_18" -> "1983 masked_fill_19" [label="(4, 64, 64)", style=solid]; +"1982 eq_9" -> "1983 masked_fill_19" [label="(4, 64, 64)", style=solid]; +"1983 masked_fill_19" -> "1985 unsqueeze_58" [label="(4, 64, 64)", style=solid]; +"1984 view_107" -> "1987 add_67" [label="(1, 4, 12, 64, 64)", style=solid]; +"1985 unsqueeze_58" -> "1986 unsqueeze_59" [label="(4, 1, 64, 64)", style=solid]; +"1986 unsqueeze_59" -> "1987 add_67" [label="(1, 4, 1, 64, 64)", style=solid]; +"1987 add_67" -> "1988 view_108" [label="(1, 4, 12, 64, 64)", style=solid]; +"1988 view_108" -> "1989 softmax_19" [label="(4, 12, 64, 64)", style=solid]; +"1989 softmax_19" -> "1990 dropout_76" [label="(4, 12, 64, 64)", style=solid]; +"1990 dropout_76" -> "1991 matmul_39" [label="(4, 12, 64, 64)", style=solid]; +"1991 matmul_39" -> "1992 quantize_per_tensor_default_120" [label="(4, 12, 64, 32)", style=solid]; +"1992 quantize_per_tensor_default_120" -> "1993 dequantize_per_tensor_default_120" [label=None, style=solid]; +"1993 dequantize_per_tensor_default_120" -> "1994 transpose_39" [label=None, style=solid]; +"1994 transpose_39" -> "1995 reshape_88" [label="(4, 64, 12, 32)", style=solid]; +"1995 reshape_88" -> "1996 linear_119" [label="(4, 64, 384)", style=solid]; +"1996 linear_119" -> "1997 dropout_77" [label="(4, 64, 384)", style=solid]; +"1997 dropout_77" -> "1998 view_109" [label="(4, 64, 384)", style=solid]; +"1998 view_109" -> "1999 permute_90" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"1999 permute_90" -> "2000 reshape_89" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2000 reshape_89" -> "2001 roll_19" [label="(1, 16, 16, 384)", style=solid]; +"2001 roll_19" -> "2002 slice_297" [label="(1, 16, 16, 384)", style=solid]; +"2002 slice_297" -> "2003 slice_298" [label="(1, 16, 16, 384)", style=solid]; +"2003 slice_298" -> "2004 slice_299" [label="(1, 14, 16, 384)", style=solid]; +"2004 slice_299" -> "2005 slice_300" [label="(1, 14, 14, 384)", style=solid]; +"2005 slice_300" -> "2006 contiguous_37" [label="(1, 14, 14, 384)", style=solid]; +"2006 contiguous_37" -> "2007 layer_norm_41" [label="(1, 14, 14, 384)", style=solid]; +"2007 layer_norm_41" -> "2008 add_68" [label="(1, 14, 14, 384)", style=solid]; +"2008 add_68" -> "2009 quantize_per_tensor_default_121" [label="(1, 14, 14, 384)", style=solid]; +"2008 add_68" -> "2019 add_69" [label="(1, 14, 14, 384)", style=solid]; +"2009 quantize_per_tensor_default_121" -> "2010 dequantize_per_tensor_default_121" [label=None, style=solid]; +"2010 dequantize_per_tensor_default_121" -> "2011 linear_120" [label=None, style=solid]; +"2011 linear_120" -> "2012 gelu_19" [label="(1, 14, 14, 1536)", style=solid]; +"2012 gelu_19" -> "2013 quantize_per_tensor_default_122" [label="(1, 14, 14, 1536)", style=solid]; +"2013 quantize_per_tensor_default_122" -> "2014 dequantize_per_tensor_default_122" [label=None, style=solid]; +"2014 dequantize_per_tensor_default_122" -> "2015 dropout_78" [label=None, style=solid]; +"2015 dropout_78" -> "2016 linear_121" [label="(1, 14, 14, 1536)", style=solid]; +"2016 linear_121" -> "2017 dropout_79" [label="(1, 14, 14, 384)", style=solid]; +"2017 dropout_79" -> "2018 layer_norm_42" [label="(1, 14, 14, 384)", style=solid]; +"2018 layer_norm_42" -> "2019 add_69" [label="(1, 14, 14, 384)", style=solid]; +"2019 add_69" -> "2020 quantize_per_tensor_default_123" [label="(1, 14, 14, 384)", style=solid]; +"2019 add_69" -> "2069 add_71" [label="(1, 14, 14, 384)", style=solid]; +"2020 quantize_per_tensor_default_123" -> "2021 dequantize_per_tensor_default_123" [label=None, style=solid]; +"2021 dequantize_per_tensor_default_123" -> "2023 pad_22" [label=None, style=solid]; +"2022 _frozen_param170" -> "2050 add_70" [label="(1, 12, 64, 64)", style=solid]; +"2023 pad_22" -> "2024 view_112" [label="(1, 16, 16, 384)", style=solid]; +"2024 view_112" -> "2025 permute_92" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2025 permute_92" -> "2026 reshape_90" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2026 reshape_90" -> "2028 linear_124" [label="(4, 64, 384)", style=solid]; +"2027 _frozen_param171" -> "2028 linear_124" [label="(1152,)", style=solid]; +"2028 linear_124" -> "2029 reshape_91" [label="(4, 64, 1152)", style=solid]; +"2029 reshape_91" -> "2030 permute_93" [label="(4, 64, 3, 12, 32)", style=solid]; +"2030 permute_93" -> "2031 select_60" [label="(3, 4, 12, 64, 32)", style=solid]; +"2030 permute_93" -> "2032 select_61" [label="(3, 4, 12, 64, 32)", style=solid]; +"2030 permute_93" -> "2033 select_62" [label="(3, 4, 12, 64, 32)", style=solid]; +"2031 select_60" -> "2034 linalg_vector_norm_40" [label="(4, 12, 64, 32)", style=solid]; +"2031 select_60" -> "2036 expand_as_40" [label="(4, 12, 64, 32)", style=solid]; +"2031 select_60" -> "2037 div_40" [label="(4, 12, 64, 32)", style=solid]; +"2032 select_61" -> "2040 linalg_vector_norm_41" [label="(4, 12, 64, 32)", style=solid]; +"2032 select_61" -> "2042 expand_as_41" [label="(4, 12, 64, 32)", style=solid]; +"2032 select_61" -> "2043 div_41" [label="(4, 12, 64, 32)", style=solid]; +"2033 select_62" -> "2053 matmul_41" [label="(4, 12, 64, 32)", style=solid]; +"2034 linalg_vector_norm_40" -> "2035 clamp_min_40" [label="(4, 12, 64, 1)", style=solid]; +"2035 clamp_min_40" -> "2036 expand_as_40" [label="(4, 12, 64, 1)", style=solid]; +"2036 expand_as_40" -> "2037 div_40" [label="(4, 12, 64, 32)", style=solid]; +"2037 div_40" -> "2038 quantize_per_tensor_default_124" [label="(4, 12, 64, 32)", style=solid]; +"2038 quantize_per_tensor_default_124" -> "2039 dequantize_per_tensor_default_124" [label=None, style=solid]; +"2039 dequantize_per_tensor_default_124" -> "2047 matmul_40" [label=None, style=solid]; +"2040 linalg_vector_norm_41" -> "2041 clamp_min_41" [label="(4, 12, 64, 1)", style=solid]; +"2041 clamp_min_41" -> "2042 expand_as_41" [label="(4, 12, 64, 1)", style=solid]; +"2042 expand_as_41" -> "2043 div_41" [label="(4, 12, 64, 32)", style=solid]; +"2043 div_41" -> "2044 quantize_per_tensor_default_125" [label="(4, 12, 64, 32)", style=solid]; +"2044 quantize_per_tensor_default_125" -> "2045 dequantize_per_tensor_default_125" [label=None, style=solid]; +"2045 dequantize_per_tensor_default_125" -> "2046 transpose_40" [label=None, style=solid]; +"2046 transpose_40" -> "2047 matmul_40" [label="(4, 12, 32, 64)", style=solid]; +"2047 matmul_40" -> "2049 mul_41" [label="(4, 12, 64, 64)", style=solid]; +"2048 _frozen_param173" -> "2049 mul_41" [label="(12, 1, 1)", style=solid]; +"2049 mul_41" -> "2050 add_70" [label="(4, 12, 64, 64)", style=solid]; +"2050 add_70" -> "2051 softmax_20" [label="(4, 12, 64, 64)", style=solid]; +"2051 softmax_20" -> "2052 dropout_80" [label="(4, 12, 64, 64)", style=solid]; +"2052 dropout_80" -> "2053 matmul_41" [label="(4, 12, 64, 64)", style=solid]; +"2053 matmul_41" -> "2054 quantize_per_tensor_default_126" [label="(4, 12, 64, 32)", style=solid]; +"2054 quantize_per_tensor_default_126" -> "2055 dequantize_per_tensor_default_126" [label=None, style=solid]; +"2055 dequantize_per_tensor_default_126" -> "2056 transpose_41" [label=None, style=solid]; +"2056 transpose_41" -> "2057 reshape_92" [label="(4, 64, 12, 32)", style=solid]; +"2057 reshape_92" -> "2058 linear_125" [label="(4, 64, 384)", style=solid]; +"2058 linear_125" -> "2059 dropout_81" [label="(4, 64, 384)", style=solid]; +"2059 dropout_81" -> "2060 view_113" [label="(4, 64, 384)", style=solid]; +"2060 view_113" -> "2061 permute_94" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2061 permute_94" -> "2062 reshape_93" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2062 reshape_93" -> "2063 slice_302" [label="(1, 16, 16, 384)", style=solid]; +"2063 slice_302" -> "2064 slice_303" [label="(1, 16, 16, 384)", style=solid]; +"2064 slice_303" -> "2065 slice_304" [label="(1, 14, 16, 384)", style=solid]; +"2065 slice_304" -> "2066 slice_305" [label="(1, 14, 14, 384)", style=solid]; +"2066 slice_305" -> "2067 contiguous_39" [label="(1, 14, 14, 384)", style=solid]; +"2067 contiguous_39" -> "2068 layer_norm_43" [label="(1, 14, 14, 384)", style=solid]; +"2068 layer_norm_43" -> "2069 add_71" [label="(1, 14, 14, 384)", style=solid]; +"2069 add_71" -> "2070 quantize_per_tensor_default_127" [label="(1, 14, 14, 384)", style=solid]; +"2069 add_71" -> "2080 add_72" [label="(1, 14, 14, 384)", style=solid]; +"2070 quantize_per_tensor_default_127" -> "2071 dequantize_per_tensor_default_127" [label=None, style=solid]; +"2071 dequantize_per_tensor_default_127" -> "2072 linear_126" [label=None, style=solid]; +"2072 linear_126" -> "2073 gelu_20" [label="(1, 14, 14, 1536)", style=solid]; +"2073 gelu_20" -> "2074 quantize_per_tensor_default_128" [label="(1, 14, 14, 1536)", style=solid]; +"2074 quantize_per_tensor_default_128" -> "2075 dequantize_per_tensor_default_128" [label=None, style=solid]; +"2075 dequantize_per_tensor_default_128" -> "2076 dropout_82" [label=None, style=solid]; +"2076 dropout_82" -> "2077 linear_127" [label="(1, 14, 14, 1536)", style=solid]; +"2077 linear_127" -> "2078 dropout_83" [label="(1, 14, 14, 384)", style=solid]; +"2078 dropout_83" -> "2079 layer_norm_44" [label="(1, 14, 14, 384)", style=solid]; +"2079 layer_norm_44" -> "2080 add_72" [label="(1, 14, 14, 384)", style=solid]; +"2080 add_72" -> "2082 pad_23" [label="(1, 14, 14, 384)", style=solid]; +"2080 add_72" -> "2148 add_75" [label="(1, 14, 14, 384)", style=solid]; +"2081 _frozen_param174" -> "2112 add_73" [label="(1, 12, 64, 64)", style=solid]; +"2082 pad_23" -> "2083 roll_20" [label="(1, 16, 16, 384)", style=solid]; +"2083 roll_20" -> "2084 view_116" [label="(1, 16, 16, 384)", style=solid]; +"2084 view_116" -> "2085 permute_96" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2085 permute_96" -> "2086 reshape_94" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2086 reshape_94" -> "2087 quantize_per_tensor_default_129" [label="(4, 64, 384)", style=solid]; +"2086 reshape_94" -> "2113 new_zeros_10" [label="(4, 64, 384)", style=solid]; +"2087 quantize_per_tensor_default_129" -> "2088 dequantize_per_tensor_default_129" [label=None, style=solid]; +"2088 dequantize_per_tensor_default_129" -> "2090 linear_130" [label=None, style=solid]; +"2089 _frozen_param175" -> "2090 linear_130" [label="(1152,)", style=solid]; +"2090 linear_130" -> "2091 reshape_95" [label="(4, 64, 1152)", style=solid]; +"2091 reshape_95" -> "2092 permute_97" [label="(4, 64, 3, 12, 32)", style=solid]; +"2092 permute_97" -> "2093 select_63" [label="(3, 4, 12, 64, 32)", style=solid]; +"2092 permute_97" -> "2094 select_64" [label="(3, 4, 12, 64, 32)", style=solid]; +"2092 permute_97" -> "2095 select_65" [label="(3, 4, 12, 64, 32)", style=solid]; +"2093 select_63" -> "2096 linalg_vector_norm_42" [label="(4, 12, 64, 32)", style=solid]; +"2093 select_63" -> "2098 expand_as_42" [label="(4, 12, 64, 32)", style=solid]; +"2093 select_63" -> "2099 div_42" [label="(4, 12, 64, 32)", style=solid]; +"2094 select_64" -> "2102 linalg_vector_norm_43" [label="(4, 12, 64, 32)", style=solid]; +"2094 select_64" -> "2104 expand_as_43" [label="(4, 12, 64, 32)", style=solid]; +"2094 select_64" -> "2105 div_43" [label="(4, 12, 64, 32)", style=solid]; +"2095 select_65" -> "2131 matmul_43" [label="(4, 12, 64, 32)", style=solid]; +"2096 linalg_vector_norm_42" -> "2097 clamp_min_42" [label="(4, 12, 64, 1)", style=solid]; +"2097 clamp_min_42" -> "2098 expand_as_42" [label="(4, 12, 64, 1)", style=solid]; +"2098 expand_as_42" -> "2099 div_42" [label="(4, 12, 64, 32)", style=solid]; +"2099 div_42" -> "2100 quantize_per_tensor_default_130" [label="(4, 12, 64, 32)", style=solid]; +"2100 quantize_per_tensor_default_130" -> "2101 dequantize_per_tensor_default_130" [label=None, style=solid]; +"2101 dequantize_per_tensor_default_130" -> "2109 matmul_42" [label=None, style=solid]; +"2102 linalg_vector_norm_43" -> "2103 clamp_min_43" [label="(4, 12, 64, 1)", style=solid]; +"2103 clamp_min_43" -> "2104 expand_as_43" [label="(4, 12, 64, 1)", style=solid]; +"2104 expand_as_43" -> "2105 div_43" [label="(4, 12, 64, 32)", style=solid]; +"2105 div_43" -> "2106 quantize_per_tensor_default_131" [label="(4, 12, 64, 32)", style=solid]; +"2106 quantize_per_tensor_default_131" -> "2107 dequantize_per_tensor_default_131" [label=None, style=solid]; +"2107 dequantize_per_tensor_default_131" -> "2108 transpose_42" [label=None, style=solid]; +"2108 transpose_42" -> "2109 matmul_42" [label="(4, 12, 32, 64)", style=solid]; +"2109 matmul_42" -> "2111 mul_43" [label="(4, 12, 64, 64)", style=solid]; +"2110 _frozen_param177" -> "2111 mul_43" [label="(12, 1, 1)", style=solid]; +"2111 mul_43" -> "2112 add_73" [label="(4, 12, 64, 64)", style=solid]; +"2112 add_73" -> "2124 view_118" [label="(4, 12, 64, 64)", style=solid]; +"2113 new_zeros_10" -> "2114 view_117" [label="(16, 16)", style=solid]; +"2114 view_117" -> "2115 permute_98" [label="(2, 8, 2, 8)", style=solid]; +"2115 permute_98" -> "2116 reshape_96" [label="(2, 2, 8, 8)", style=solid]; +"2116 reshape_96" -> "2117 unsqueeze_62" [label="(4, 64)", style=solid]; +"2116 reshape_96" -> "2118 unsqueeze_63" [label="(4, 64)", style=solid]; +"2117 unsqueeze_62" -> "2119 sub_10" [label="(4, 1, 64)", style=solid]; +"2118 unsqueeze_63" -> "2119 sub_10" [label="(4, 64, 1)", style=solid]; +"2119 sub_10" -> "2120 ne_10" [label="(4, 64, 64)", style=solid]; +"2119 sub_10" -> "2121 masked_fill_20" [label="(4, 64, 64)", style=solid]; +"2119 sub_10" -> "2122 eq_10" [label="(4, 64, 64)", style=solid]; +"2120 ne_10" -> "2121 masked_fill_20" [label="(4, 64, 64)", style=solid]; +"2121 masked_fill_20" -> "2123 masked_fill_21" [label="(4, 64, 64)", style=solid]; +"2122 eq_10" -> "2123 masked_fill_21" [label="(4, 64, 64)", style=solid]; +"2123 masked_fill_21" -> "2125 unsqueeze_64" [label="(4, 64, 64)", style=solid]; +"2124 view_118" -> "2127 add_74" [label="(1, 4, 12, 64, 64)", style=solid]; +"2125 unsqueeze_64" -> "2126 unsqueeze_65" [label="(4, 1, 64, 64)", style=solid]; +"2126 unsqueeze_65" -> "2127 add_74" [label="(1, 4, 1, 64, 64)", style=solid]; +"2127 add_74" -> "2128 view_119" [label="(1, 4, 12, 64, 64)", style=solid]; +"2128 view_119" -> "2129 softmax_21" [label="(4, 12, 64, 64)", style=solid]; +"2129 softmax_21" -> "2130 dropout_84" [label="(4, 12, 64, 64)", style=solid]; +"2130 dropout_84" -> "2131 matmul_43" [label="(4, 12, 64, 64)", style=solid]; +"2131 matmul_43" -> "2132 quantize_per_tensor_default_132" [label="(4, 12, 64, 32)", style=solid]; +"2132 quantize_per_tensor_default_132" -> "2133 dequantize_per_tensor_default_132" [label=None, style=solid]; +"2133 dequantize_per_tensor_default_132" -> "2134 transpose_43" [label=None, style=solid]; +"2134 transpose_43" -> "2135 reshape_97" [label="(4, 64, 12, 32)", style=solid]; +"2135 reshape_97" -> "2136 linear_131" [label="(4, 64, 384)", style=solid]; +"2136 linear_131" -> "2137 dropout_85" [label="(4, 64, 384)", style=solid]; +"2137 dropout_85" -> "2138 view_120" [label="(4, 64, 384)", style=solid]; +"2138 view_120" -> "2139 permute_99" [label="(1, 2, 2, 8, 8, 384)", style=solid]; +"2139 permute_99" -> "2140 reshape_98" [label="(1, 2, 8, 2, 8, 384)", style=solid]; +"2140 reshape_98" -> "2141 roll_21" [label="(1, 16, 16, 384)", style=solid]; +"2141 roll_21" -> "2142 slice_325" [label="(1, 16, 16, 384)", style=solid]; +"2142 slice_325" -> "2143 slice_326" [label="(1, 16, 16, 384)", style=solid]; +"2143 slice_326" -> "2144 slice_327" [label="(1, 14, 16, 384)", style=solid]; +"2144 slice_327" -> "2145 slice_328" [label="(1, 14, 14, 384)", style=solid]; +"2145 slice_328" -> "2146 contiguous_41" [label="(1, 14, 14, 384)", style=solid]; +"2146 contiguous_41" -> "2147 layer_norm_45" [label="(1, 14, 14, 384)", style=solid]; +"2147 layer_norm_45" -> "2148 add_75" [label="(1, 14, 14, 384)", style=solid]; +"2148 add_75" -> "2149 quantize_per_tensor_default_133" [label="(1, 14, 14, 384)", style=solid]; +"2148 add_75" -> "2159 add_76" [label="(1, 14, 14, 384)", style=solid]; +"2149 quantize_per_tensor_default_133" -> "2150 dequantize_per_tensor_default_133" [label=None, style=solid]; +"2150 dequantize_per_tensor_default_133" -> "2151 linear_132" [label=None, style=solid]; +"2151 linear_132" -> "2152 gelu_21" [label="(1, 14, 14, 1536)", style=solid]; +"2152 gelu_21" -> "2153 quantize_per_tensor_default_134" [label="(1, 14, 14, 1536)", style=solid]; +"2153 quantize_per_tensor_default_134" -> "2154 dequantize_per_tensor_default_134" [label=None, style=solid]; +"2154 dequantize_per_tensor_default_134" -> "2155 dropout_86" [label=None, style=solid]; +"2155 dropout_86" -> "2156 linear_133" [label="(1, 14, 14, 1536)", style=solid]; +"2156 linear_133" -> "2157 dropout_87" [label="(1, 14, 14, 384)", style=solid]; +"2157 dropout_87" -> "2158 layer_norm_46" [label="(1, 14, 14, 384)", style=solid]; +"2158 layer_norm_46" -> "2159 add_76" [label="(1, 14, 14, 384)", style=solid]; +"2159 add_76" -> "2160 quantize_per_tensor_default_135" [label="(1, 14, 14, 384)", style=solid]; +"2160 quantize_per_tensor_default_135" -> "2161 dequantize_per_tensor_default_135" [label=None, style=solid]; +"2161 dequantize_per_tensor_default_135" -> "2162 pad_24" [label=None, style=solid]; +"2162 pad_24" -> "2163 slice_329" [label="(1, 14, 14, 384)", style=solid]; +"2162 pad_24" -> "2166 slice_332" [label="(1, 14, 14, 384)", style=solid]; +"2162 pad_24" -> "2169 slice_335" [label="(1, 14, 14, 384)", style=solid]; +"2162 pad_24" -> "2172 slice_338" [label="(1, 14, 14, 384)", style=solid]; +"2163 slice_329" -> "2164 slice_330" [label="(1, 7, 14, 384)", style=solid]; +"2164 slice_330" -> "2165 slice_331" [label="(1, 7, 7, 384)", style=solid]; +"2165 slice_331" -> "2175 cat_2" [label="(1, 7, 7, 384)", style=solid]; +"2166 slice_332" -> "2167 slice_333" [label="(1, 7, 14, 384)", style=solid]; +"2167 slice_333" -> "2168 slice_334" [label="(1, 7, 7, 384)", style=solid]; +"2168 slice_334" -> "2175 cat_2" [label="(1, 7, 7, 384)", style=solid]; +"2169 slice_335" -> "2170 slice_336" [label="(1, 7, 14, 384)", style=solid]; +"2170 slice_336" -> "2171 slice_337" [label="(1, 7, 7, 384)", style=solid]; +"2171 slice_337" -> "2175 cat_2" [label="(1, 7, 7, 384)", style=solid]; +"2172 slice_338" -> "2173 slice_339" [label="(1, 7, 14, 384)", style=solid]; +"2173 slice_339" -> "2174 slice_340" [label="(1, 7, 7, 384)", style=solid]; +"2174 slice_340" -> "2175 cat_2" [label="(1, 7, 7, 384)", style=solid]; +"2175 cat_2" -> "2176 linear_134" [label="(1, 7, 7, 1536)", style=solid]; +"2176 linear_134" -> "2177 layer_norm_47" [label="(1, 7, 7, 768)", style=solid]; +"2177 layer_norm_47" -> "2178 quantize_per_tensor_default_136" [label="(1, 7, 7, 768)", style=solid]; +"2177 layer_norm_47" -> "2227 add_78" [label="(1, 7, 7, 768)", style=solid]; +"2178 quantize_per_tensor_default_136" -> "2179 dequantize_per_tensor_default_136" [label=None, style=solid]; +"2179 dequantize_per_tensor_default_136" -> "2181 pad_25" [label=None, style=solid]; +"2180 _frozen_param187" -> "2208 add_77" [label="(1, 24, 64, 64)", style=solid]; +"2181 pad_25" -> "2182 view_123" [label="(1, 8, 8, 768)", style=solid]; +"2182 view_123" -> "2183 permute_101" [label="(1, 1, 8, 1, 8, 768)", style=solid]; +"2183 permute_101" -> "2184 reshape_99" [label="(1, 1, 1, 8, 8, 768)", style=solid]; +"2184 reshape_99" -> "2186 linear_137" [label="(1, 64, 768)", style=solid]; +"2185 _frozen_param188" -> "2186 linear_137" [label="(2304,)", style=solid]; +"2186 linear_137" -> "2187 reshape_100" [label="(1, 64, 2304)", style=solid]; +"2187 reshape_100" -> "2188 permute_102" [label="(1, 64, 3, 24, 32)", style=solid]; +"2188 permute_102" -> "2189 select_66" [label="(3, 1, 24, 64, 32)", style=solid]; +"2188 permute_102" -> "2190 select_67" [label="(3, 1, 24, 64, 32)", style=solid]; +"2188 permute_102" -> "2191 select_68" [label="(3, 1, 24, 64, 32)", style=solid]; +"2189 select_66" -> "2192 linalg_vector_norm_44" [label="(1, 24, 64, 32)", style=solid]; +"2189 select_66" -> "2194 expand_as_44" [label="(1, 24, 64, 32)", style=solid]; +"2189 select_66" -> "2195 div_44" [label="(1, 24, 64, 32)", style=solid]; +"2190 select_67" -> "2198 linalg_vector_norm_45" [label="(1, 24, 64, 32)", style=solid]; +"2190 select_67" -> "2200 expand_as_45" [label="(1, 24, 64, 32)", style=solid]; +"2190 select_67" -> "2201 div_45" [label="(1, 24, 64, 32)", style=solid]; +"2191 select_68" -> "2211 matmul_45" [label="(1, 24, 64, 32)", style=solid]; +"2192 linalg_vector_norm_44" -> "2193 clamp_min_44" [label="(1, 24, 64, 1)", style=solid]; +"2193 clamp_min_44" -> "2194 expand_as_44" [label="(1, 24, 64, 1)", style=solid]; +"2194 expand_as_44" -> "2195 div_44" [label="(1, 24, 64, 32)", style=solid]; +"2195 div_44" -> "2196 quantize_per_tensor_default_137" [label="(1, 24, 64, 32)", style=solid]; +"2196 quantize_per_tensor_default_137" -> "2197 dequantize_per_tensor_default_137" [label=None, style=solid]; +"2197 dequantize_per_tensor_default_137" -> "2205 matmul_44" [label=None, style=solid]; +"2198 linalg_vector_norm_45" -> "2199 clamp_min_45" [label="(1, 24, 64, 1)", style=solid]; +"2199 clamp_min_45" -> "2200 expand_as_45" [label="(1, 24, 64, 1)", style=solid]; +"2200 expand_as_45" -> "2201 div_45" [label="(1, 24, 64, 32)", style=solid]; +"2201 div_45" -> "2202 quantize_per_tensor_default_138" [label="(1, 24, 64, 32)", style=solid]; +"2202 quantize_per_tensor_default_138" -> "2203 dequantize_per_tensor_default_138" [label=None, style=solid]; +"2203 dequantize_per_tensor_default_138" -> "2204 transpose_44" [label=None, style=solid]; +"2204 transpose_44" -> "2205 matmul_44" [label="(1, 24, 32, 64)", style=solid]; +"2205 matmul_44" -> "2207 mul_45" [label="(1, 24, 64, 64)", style=solid]; +"2206 _frozen_param190" -> "2207 mul_45" [label="(24, 1, 1)", style=solid]; +"2207 mul_45" -> "2208 add_77" [label="(1, 24, 64, 64)", style=solid]; +"2208 add_77" -> "2209 softmax_22" [label="(1, 24, 64, 64)", style=solid]; +"2209 softmax_22" -> "2210 dropout_88" [label="(1, 24, 64, 64)", style=solid]; +"2210 dropout_88" -> "2211 matmul_45" [label="(1, 24, 64, 64)", style=solid]; +"2211 matmul_45" -> "2212 quantize_per_tensor_default_139" [label="(1, 24, 64, 32)", style=solid]; +"2212 quantize_per_tensor_default_139" -> "2213 dequantize_per_tensor_default_139" [label=None, style=solid]; +"2213 dequantize_per_tensor_default_139" -> "2214 transpose_45" [label=None, style=solid]; +"2214 transpose_45" -> "2215 reshape_101" [label="(1, 64, 24, 32)", style=solid]; +"2215 reshape_101" -> "2216 linear_138" [label="(1, 64, 768)", style=solid]; +"2216 linear_138" -> "2217 dropout_89" [label="(1, 64, 768)", style=solid]; +"2217 dropout_89" -> "2218 view_124" [label="(1, 64, 768)", style=solid]; +"2218 view_124" -> "2219 permute_103" [label="(1, 1, 1, 8, 8, 768)", style=solid]; +"2219 permute_103" -> "2220 reshape_102" [label="(1, 1, 8, 1, 8, 768)", style=solid]; +"2220 reshape_102" -> "2221 slice_342" [label="(1, 8, 8, 768)", style=solid]; +"2221 slice_342" -> "2222 slice_343" [label="(1, 8, 8, 768)", style=solid]; +"2222 slice_343" -> "2223 slice_344" [label="(1, 7, 8, 768)", style=solid]; +"2223 slice_344" -> "2224 slice_345" [label="(1, 7, 7, 768)", style=solid]; +"2224 slice_345" -> "2225 contiguous_43" [label="(1, 7, 7, 768)", style=solid]; +"2225 contiguous_43" -> "2226 layer_norm_48" [label="(1, 7, 7, 768)", style=solid]; +"2226 layer_norm_48" -> "2227 add_78" [label="(1, 7, 7, 768)", style=solid]; +"2227 add_78" -> "2228 quantize_per_tensor_default_140" [label="(1, 7, 7, 768)", style=solid]; +"2227 add_78" -> "2238 add_79" [label="(1, 7, 7, 768)", style=solid]; +"2228 quantize_per_tensor_default_140" -> "2229 dequantize_per_tensor_default_140" [label=None, style=solid]; +"2229 dequantize_per_tensor_default_140" -> "2230 linear_139" [label=None, style=solid]; +"2230 linear_139" -> "2231 gelu_22" [label="(1, 7, 7, 3072)", style=solid]; +"2231 gelu_22" -> "2232 quantize_per_tensor_default_141" [label="(1, 7, 7, 3072)", style=solid]; +"2232 quantize_per_tensor_default_141" -> "2233 dequantize_per_tensor_default_141" [label=None, style=solid]; +"2233 dequantize_per_tensor_default_141" -> "2234 dropout_90" [label=None, style=solid]; +"2234 dropout_90" -> "2235 linear_140" [label="(1, 7, 7, 3072)", style=solid]; +"2235 linear_140" -> "2236 dropout_91" [label="(1, 7, 7, 768)", style=solid]; +"2236 dropout_91" -> "2237 layer_norm_49" [label="(1, 7, 7, 768)", style=solid]; +"2237 layer_norm_49" -> "2238 add_79" [label="(1, 7, 7, 768)", style=solid]; +"2238 add_79" -> "2239 quantize_per_tensor_default_142" [label="(1, 7, 7, 768)", style=solid]; +"2238 add_79" -> "2288 add_81" [label="(1, 7, 7, 768)", style=solid]; +"2239 quantize_per_tensor_default_142" -> "2240 dequantize_per_tensor_default_142" [label=None, style=solid]; +"2240 dequantize_per_tensor_default_142" -> "2242 pad_26" [label=None, style=solid]; +"2241 _frozen_param191" -> "2269 add_80" [label="(1, 24, 64, 64)", style=solid]; +"2242 pad_26" -> "2243 view_127" [label="(1, 8, 8, 768)", style=solid]; +"2243 view_127" -> "2244 permute_105" [label="(1, 1, 8, 1, 8, 768)", style=solid]; +"2244 permute_105" -> "2245 reshape_103" [label="(1, 1, 1, 8, 8, 768)", style=solid]; +"2245 reshape_103" -> "2247 linear_143" [label="(1, 64, 768)", style=solid]; +"2246 _frozen_param192" -> "2247 linear_143" [label="(2304,)", style=solid]; +"2247 linear_143" -> "2248 reshape_104" [label="(1, 64, 2304)", style=solid]; +"2248 reshape_104" -> "2249 permute_106" [label="(1, 64, 3, 24, 32)", style=solid]; +"2249 permute_106" -> "2250 select_69" [label="(3, 1, 24, 64, 32)", style=solid]; +"2249 permute_106" -> "2251 select_70" [label="(3, 1, 24, 64, 32)", style=solid]; +"2249 permute_106" -> "2252 select_71" [label="(3, 1, 24, 64, 32)", style=solid]; +"2250 select_69" -> "2253 linalg_vector_norm_46" [label="(1, 24, 64, 32)", style=solid]; +"2250 select_69" -> "2255 expand_as_46" [label="(1, 24, 64, 32)", style=solid]; +"2250 select_69" -> "2256 div_46" [label="(1, 24, 64, 32)", style=solid]; +"2251 select_70" -> "2259 linalg_vector_norm_47" [label="(1, 24, 64, 32)", style=solid]; +"2251 select_70" -> "2261 expand_as_47" [label="(1, 24, 64, 32)", style=solid]; +"2251 select_70" -> "2262 div_47" [label="(1, 24, 64, 32)", style=solid]; +"2252 select_71" -> "2272 matmul_47" [label="(1, 24, 64, 32)", style=solid]; +"2253 linalg_vector_norm_46" -> "2254 clamp_min_46" [label="(1, 24, 64, 1)", style=solid]; +"2254 clamp_min_46" -> "2255 expand_as_46" [label="(1, 24, 64, 1)", style=solid]; +"2255 expand_as_46" -> "2256 div_46" [label="(1, 24, 64, 32)", style=solid]; +"2256 div_46" -> "2257 quantize_per_tensor_default_143" [label="(1, 24, 64, 32)", style=solid]; +"2257 quantize_per_tensor_default_143" -> "2258 dequantize_per_tensor_default_143" [label=None, style=solid]; +"2258 dequantize_per_tensor_default_143" -> "2266 matmul_46" [label=None, style=solid]; +"2259 linalg_vector_norm_47" -> "2260 clamp_min_47" [label="(1, 24, 64, 1)", style=solid]; +"2260 clamp_min_47" -> "2261 expand_as_47" [label="(1, 24, 64, 1)", style=solid]; +"2261 expand_as_47" -> "2262 div_47" [label="(1, 24, 64, 32)", style=solid]; +"2262 div_47" -> "2263 quantize_per_tensor_default_144" [label="(1, 24, 64, 32)", style=solid]; +"2263 quantize_per_tensor_default_144" -> "2264 dequantize_per_tensor_default_144" [label=None, style=solid]; +"2264 dequantize_per_tensor_default_144" -> "2265 transpose_46" [label=None, style=solid]; +"2265 transpose_46" -> "2266 matmul_46" [label="(1, 24, 32, 64)", style=solid]; +"2266 matmul_46" -> "2268 mul_47" [label="(1, 24, 64, 64)", style=solid]; +"2267 _frozen_param194" -> "2268 mul_47" [label="(24, 1, 1)", style=solid]; +"2268 mul_47" -> "2269 add_80" [label="(1, 24, 64, 64)", style=solid]; +"2269 add_80" -> "2270 softmax_23" [label="(1, 24, 64, 64)", style=solid]; +"2270 softmax_23" -> "2271 dropout_92" [label="(1, 24, 64, 64)", style=solid]; +"2271 dropout_92" -> "2272 matmul_47" [label="(1, 24, 64, 64)", style=solid]; +"2272 matmul_47" -> "2273 quantize_per_tensor_default_145" [label="(1, 24, 64, 32)", style=solid]; +"2273 quantize_per_tensor_default_145" -> "2274 dequantize_per_tensor_default_145" [label=None, style=solid]; +"2274 dequantize_per_tensor_default_145" -> "2275 transpose_47" [label=None, style=solid]; +"2275 transpose_47" -> "2276 reshape_105" [label="(1, 64, 24, 32)", style=solid]; +"2276 reshape_105" -> "2277 linear_144" [label="(1, 64, 768)", style=solid]; +"2277 linear_144" -> "2278 dropout_93" [label="(1, 64, 768)", style=solid]; +"2278 dropout_93" -> "2279 view_128" [label="(1, 64, 768)", style=solid]; +"2279 view_128" -> "2280 permute_107" [label="(1, 1, 1, 8, 8, 768)", style=solid]; +"2280 permute_107" -> "2281 reshape_106" [label="(1, 1, 8, 1, 8, 768)", style=solid]; +"2281 reshape_106" -> "2282 slice_347" [label="(1, 8, 8, 768)", style=solid]; +"2282 slice_347" -> "2283 slice_348" [label="(1, 8, 8, 768)", style=solid]; +"2283 slice_348" -> "2284 slice_349" [label="(1, 7, 8, 768)", style=solid]; +"2284 slice_349" -> "2285 slice_350" [label="(1, 7, 7, 768)", style=solid]; +"2285 slice_350" -> "2286 contiguous_45" [label="(1, 7, 7, 768)", style=solid]; +"2286 contiguous_45" -> "2287 layer_norm_50" [label="(1, 7, 7, 768)", style=solid]; +"2287 layer_norm_50" -> "2288 add_81" [label="(1, 7, 7, 768)", style=solid]; +"2288 add_81" -> "2289 quantize_per_tensor_default_146" [label="(1, 7, 7, 768)", style=solid]; +"2288 add_81" -> "2299 add_82" [label="(1, 7, 7, 768)", style=solid]; +"2289 quantize_per_tensor_default_146" -> "2290 dequantize_per_tensor_default_146" [label=None, style=solid]; +"2290 dequantize_per_tensor_default_146" -> "2291 linear_145" [label=None, style=solid]; +"2291 linear_145" -> "2292 gelu_23" [label="(1, 7, 7, 3072)", style=solid]; +"2292 gelu_23" -> "2293 quantize_per_tensor_default_147" [label="(1, 7, 7, 3072)", style=solid]; +"2293 quantize_per_tensor_default_147" -> "2294 dequantize_per_tensor_default_147" [label=None, style=solid]; +"2294 dequantize_per_tensor_default_147" -> "2295 dropout_94" [label=None, style=solid]; +"2295 dropout_94" -> "2296 linear_146" [label="(1, 7, 7, 3072)", style=solid]; +"2296 linear_146" -> "2297 dropout_95" [label="(1, 7, 7, 768)", style=solid]; +"2297 dropout_95" -> "2298 layer_norm_51" [label="(1, 7, 7, 768)", style=solid]; +"2298 layer_norm_51" -> "2299 add_82" [label="(1, 7, 7, 768)", style=solid]; +"2299 add_82" -> "2300 layer_norm_52" [label="(1, 7, 7, 768)", style=solid]; +"2300 layer_norm_52" -> "2301 permute_108" [label="(1, 7, 7, 768)", style=solid]; +"2301 permute_108" -> "2302 adaptive_avg_pool2d" [label="(1, 768, 7, 7)", style=solid]; +"2302 adaptive_avg_pool2d" -> "2303 quantize_per_tensor_default_148" [label="(1, 768, 1, 1)", style=solid]; +"2303 quantize_per_tensor_default_148" -> "2304 dequantize_per_tensor_default_148" [label=None, style=solid]; +"2304 dequantize_per_tensor_default_148" -> "2305 flatten" [label=None, style=solid]; +"2305 flatten" -> "2306 linear_147" [label="(1, 768)", style=solid]; +"2306 linear_147" -> "2307 output_1" [label="(1, 1000)", style=solid]; } diff --git a/tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/vit_b_16.dot b/tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/vit_b_16.dot index d3c40ce0d67..c54eee713de 100644 --- a/tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/vit_b_16.dot +++ b/tests/torch/data/reference_graphs/fx/experimental/ao_export_quantization_OpenVINOQuantizer/vit_b_16.dot @@ -1,1791 +1,1789 @@ strict digraph { -"0 class_token" [id=0, type=get_attr]; -"1 encoder_pos_embedding" [id=1, type=get_attr]; -"2 _scale_0" [id=2, type=get_attr]; -"3 _zero_point_0" [id=3, type=get_attr]; -"4 _frozen_param0" [id=4, type=get_attr]; -"5 dequantize_per_channel_default" [id=5, type=dequantize_per_channel]; -"6 conv_proj_bias" [id=6, type=get_attr]; -"7 encoder_layers_encoder_layer_0_ln_1_weight" [id=7, type=get_attr]; -"8 encoder_layers_encoder_layer_0_ln_1_bias" [id=8, type=get_attr]; -"9 _scale_1" [id=9, type=get_attr]; -"10 _zero_point_1" [id=10, type=get_attr]; -"11 _frozen_param1" [id=11, type=get_attr]; -"12 dequantize_per_channel_default_1" [id=12, type=dequantize_per_channel]; -"13 encoder_layers_encoder_layer_0_self_attention_in_proj_bias" [id=13, type=get_attr]; -"14 _scale_2" [id=14, type=get_attr]; -"15 _zero_point_2" [id=15, type=get_attr]; -"16 _frozen_param2" [id=16, type=get_attr]; -"17 dequantize_per_channel_default_2" [id=17, type=dequantize_per_channel]; -"18 encoder_layers_encoder_layer_0_self_attention_out_proj_bias" [id=18, type=get_attr]; -"19 encoder_layers_encoder_layer_0_ln_2_weight" [id=19, type=get_attr]; -"20 encoder_layers_encoder_layer_0_ln_2_bias" [id=20, type=get_attr]; -"21 _scale_3" [id=21, type=get_attr]; -"22 _zero_point_3" [id=22, type=get_attr]; -"23 _frozen_param3" [id=23, type=get_attr]; -"24 dequantize_per_channel_default_3" [id=24, type=dequantize_per_channel]; -"25 encoder_layers_encoder_layer_0_mlp_0_bias" [id=25, type=get_attr]; -"26 _scale_4" [id=26, type=get_attr]; -"27 _zero_point_4" [id=27, type=get_attr]; -"28 _frozen_param4" [id=28, type=get_attr]; -"29 dequantize_per_channel_default_4" [id=29, type=dequantize_per_channel]; -"30 encoder_layers_encoder_layer_0_mlp_3_bias" [id=30, type=get_attr]; -"31 encoder_layers_encoder_layer_1_ln_1_weight" [id=31, type=get_attr]; -"32 encoder_layers_encoder_layer_1_ln_1_bias" [id=32, type=get_attr]; -"33 _scale_5" [id=33, type=get_attr]; -"34 _zero_point_5" [id=34, type=get_attr]; -"35 _frozen_param5" [id=35, type=get_attr]; -"36 dequantize_per_channel_default_5" [id=36, type=dequantize_per_channel]; -"37 encoder_layers_encoder_layer_1_self_attention_in_proj_bias" [id=37, type=get_attr]; -"38 _scale_6" [id=38, type=get_attr]; -"39 _zero_point_6" [id=39, type=get_attr]; -"40 _frozen_param6" [id=40, type=get_attr]; -"41 dequantize_per_channel_default_6" [id=41, type=dequantize_per_channel]; -"42 encoder_layers_encoder_layer_1_self_attention_out_proj_bias" [id=42, type=get_attr]; -"43 encoder_layers_encoder_layer_1_ln_2_weight" [id=43, type=get_attr]; -"44 encoder_layers_encoder_layer_1_ln_2_bias" [id=44, type=get_attr]; -"45 _scale_7" [id=45, type=get_attr]; -"46 _zero_point_7" [id=46, type=get_attr]; -"47 _frozen_param7" [id=47, type=get_attr]; -"48 dequantize_per_channel_default_7" [id=48, type=dequantize_per_channel]; -"49 encoder_layers_encoder_layer_1_mlp_0_bias" [id=49, type=get_attr]; -"50 _scale_8" [id=50, type=get_attr]; -"51 _zero_point_8" [id=51, type=get_attr]; -"52 _frozen_param8" [id=52, type=get_attr]; -"53 dequantize_per_channel_default_8" [id=53, type=dequantize_per_channel]; -"54 encoder_layers_encoder_layer_1_mlp_3_bias" [id=54, type=get_attr]; -"55 encoder_layers_encoder_layer_2_ln_1_weight" [id=55, type=get_attr]; -"56 encoder_layers_encoder_layer_2_ln_1_bias" [id=56, type=get_attr]; -"57 _scale_9" [id=57, type=get_attr]; -"58 _zero_point_9" [id=58, type=get_attr]; -"59 _frozen_param9" [id=59, type=get_attr]; -"60 dequantize_per_channel_default_9" [id=60, type=dequantize_per_channel]; -"61 encoder_layers_encoder_layer_2_self_attention_in_proj_bias" [id=61, type=get_attr]; -"62 _scale_10" [id=62, type=get_attr]; -"63 _zero_point_10" [id=63, type=get_attr]; -"64 _frozen_param10" [id=64, type=get_attr]; -"65 dequantize_per_channel_default_10" [id=65, type=dequantize_per_channel]; -"66 encoder_layers_encoder_layer_2_self_attention_out_proj_bias" [id=66, type=get_attr]; -"67 encoder_layers_encoder_layer_2_ln_2_weight" [id=67, type=get_attr]; -"68 encoder_layers_encoder_layer_2_ln_2_bias" [id=68, type=get_attr]; -"69 _scale_11" [id=69, type=get_attr]; -"70 _zero_point_11" [id=70, type=get_attr]; -"71 _frozen_param11" [id=71, type=get_attr]; -"72 dequantize_per_channel_default_11" [id=72, type=dequantize_per_channel]; -"73 encoder_layers_encoder_layer_2_mlp_0_bias" [id=73, type=get_attr]; -"74 _scale_12" [id=74, type=get_attr]; -"75 _zero_point_12" [id=75, type=get_attr]; -"76 _frozen_param12" [id=76, type=get_attr]; -"77 dequantize_per_channel_default_12" [id=77, type=dequantize_per_channel]; -"78 encoder_layers_encoder_layer_2_mlp_3_bias" [id=78, type=get_attr]; -"79 encoder_layers_encoder_layer_3_ln_1_weight" [id=79, type=get_attr]; -"80 encoder_layers_encoder_layer_3_ln_1_bias" [id=80, type=get_attr]; -"81 _scale_13" [id=81, type=get_attr]; -"82 _zero_point_13" [id=82, type=get_attr]; -"83 _frozen_param13" [id=83, type=get_attr]; -"84 dequantize_per_channel_default_13" [id=84, type=dequantize_per_channel]; -"85 encoder_layers_encoder_layer_3_self_attention_in_proj_bias" [id=85, type=get_attr]; -"86 _scale_14" [id=86, type=get_attr]; -"87 _zero_point_14" [id=87, type=get_attr]; -"88 _frozen_param14" [id=88, type=get_attr]; -"89 dequantize_per_channel_default_14" [id=89, type=dequantize_per_channel]; -"90 encoder_layers_encoder_layer_3_self_attention_out_proj_bias" [id=90, type=get_attr]; -"91 encoder_layers_encoder_layer_3_ln_2_weight" [id=91, type=get_attr]; -"92 encoder_layers_encoder_layer_3_ln_2_bias" [id=92, type=get_attr]; -"93 _scale_15" [id=93, type=get_attr]; -"94 _zero_point_15" [id=94, type=get_attr]; -"95 _frozen_param15" [id=95, type=get_attr]; -"96 dequantize_per_channel_default_15" [id=96, type=dequantize_per_channel]; -"97 encoder_layers_encoder_layer_3_mlp_0_bias" [id=97, type=get_attr]; -"98 _scale_16" [id=98, type=get_attr]; -"99 _zero_point_16" [id=99, type=get_attr]; -"100 _frozen_param16" [id=100, type=get_attr]; -"101 dequantize_per_channel_default_16" [id=101, type=dequantize_per_channel]; -"102 encoder_layers_encoder_layer_3_mlp_3_bias" [id=102, type=get_attr]; -"103 encoder_layers_encoder_layer_4_ln_1_weight" [id=103, type=get_attr]; -"104 encoder_layers_encoder_layer_4_ln_1_bias" [id=104, type=get_attr]; -"105 _scale_17" [id=105, type=get_attr]; -"106 _zero_point_17" [id=106, type=get_attr]; -"107 _frozen_param17" [id=107, type=get_attr]; -"108 dequantize_per_channel_default_17" [id=108, type=dequantize_per_channel]; -"109 encoder_layers_encoder_layer_4_self_attention_in_proj_bias" [id=109, type=get_attr]; -"110 _scale_18" [id=110, type=get_attr]; -"111 _zero_point_18" [id=111, type=get_attr]; -"112 _frozen_param18" [id=112, type=get_attr]; -"113 dequantize_per_channel_default_18" [id=113, type=dequantize_per_channel]; -"114 encoder_layers_encoder_layer_4_self_attention_out_proj_bias" [id=114, type=get_attr]; -"115 encoder_layers_encoder_layer_4_ln_2_weight" [id=115, type=get_attr]; -"116 encoder_layers_encoder_layer_4_ln_2_bias" [id=116, type=get_attr]; -"117 _scale_19" [id=117, type=get_attr]; -"118 _zero_point_19" [id=118, type=get_attr]; -"119 _frozen_param19" [id=119, type=get_attr]; -"120 dequantize_per_channel_default_19" [id=120, type=dequantize_per_channel]; -"121 encoder_layers_encoder_layer_4_mlp_0_bias" [id=121, type=get_attr]; -"122 _scale_20" [id=122, type=get_attr]; -"123 _zero_point_20" [id=123, type=get_attr]; -"124 _frozen_param20" [id=124, type=get_attr]; -"125 dequantize_per_channel_default_20" [id=125, type=dequantize_per_channel]; -"126 encoder_layers_encoder_layer_4_mlp_3_bias" [id=126, type=get_attr]; -"127 encoder_layers_encoder_layer_5_ln_1_weight" [id=127, type=get_attr]; -"128 encoder_layers_encoder_layer_5_ln_1_bias" [id=128, type=get_attr]; -"129 _scale_21" [id=129, type=get_attr]; -"130 _zero_point_21" [id=130, type=get_attr]; -"131 _frozen_param21" [id=131, type=get_attr]; -"132 dequantize_per_channel_default_21" [id=132, type=dequantize_per_channel]; -"133 encoder_layers_encoder_layer_5_self_attention_in_proj_bias" [id=133, type=get_attr]; -"134 _scale_22" [id=134, type=get_attr]; -"135 _zero_point_22" [id=135, type=get_attr]; -"136 _frozen_param22" [id=136, type=get_attr]; -"137 dequantize_per_channel_default_22" [id=137, type=dequantize_per_channel]; -"138 encoder_layers_encoder_layer_5_self_attention_out_proj_bias" [id=138, type=get_attr]; -"139 encoder_layers_encoder_layer_5_ln_2_weight" [id=139, type=get_attr]; -"140 encoder_layers_encoder_layer_5_ln_2_bias" [id=140, type=get_attr]; -"141 _scale_23" [id=141, type=get_attr]; -"142 _zero_point_23" [id=142, type=get_attr]; -"143 _frozen_param23" [id=143, type=get_attr]; -"144 dequantize_per_channel_default_23" [id=144, type=dequantize_per_channel]; -"145 encoder_layers_encoder_layer_5_mlp_0_bias" [id=145, type=get_attr]; -"146 _scale_24" [id=146, type=get_attr]; -"147 _zero_point_24" [id=147, type=get_attr]; -"148 _frozen_param24" [id=148, type=get_attr]; -"149 dequantize_per_channel_default_24" [id=149, type=dequantize_per_channel]; -"150 encoder_layers_encoder_layer_5_mlp_3_bias" [id=150, type=get_attr]; -"151 encoder_layers_encoder_layer_6_ln_1_weight" [id=151, type=get_attr]; -"152 encoder_layers_encoder_layer_6_ln_1_bias" [id=152, type=get_attr]; -"153 _scale_25" [id=153, type=get_attr]; -"154 _zero_point_25" [id=154, type=get_attr]; -"155 _frozen_param25" [id=155, type=get_attr]; -"156 dequantize_per_channel_default_25" [id=156, type=dequantize_per_channel]; -"157 encoder_layers_encoder_layer_6_self_attention_in_proj_bias" [id=157, type=get_attr]; -"158 _scale_26" [id=158, type=get_attr]; -"159 _zero_point_26" [id=159, type=get_attr]; -"160 _frozen_param26" [id=160, type=get_attr]; -"161 dequantize_per_channel_default_26" [id=161, type=dequantize_per_channel]; -"162 encoder_layers_encoder_layer_6_self_attention_out_proj_bias" [id=162, type=get_attr]; -"163 encoder_layers_encoder_layer_6_ln_2_weight" [id=163, type=get_attr]; -"164 encoder_layers_encoder_layer_6_ln_2_bias" [id=164, type=get_attr]; -"165 _scale_27" [id=165, type=get_attr]; -"166 _zero_point_27" [id=166, type=get_attr]; -"167 _frozen_param27" [id=167, type=get_attr]; -"168 dequantize_per_channel_default_27" [id=168, type=dequantize_per_channel]; -"169 encoder_layers_encoder_layer_6_mlp_0_bias" [id=169, type=get_attr]; -"170 _scale_28" [id=170, type=get_attr]; -"171 _zero_point_28" [id=171, type=get_attr]; -"172 _frozen_param28" [id=172, type=get_attr]; -"173 dequantize_per_channel_default_28" [id=173, type=dequantize_per_channel]; -"174 encoder_layers_encoder_layer_6_mlp_3_bias" [id=174, type=get_attr]; -"175 encoder_layers_encoder_layer_7_ln_1_weight" [id=175, type=get_attr]; -"176 encoder_layers_encoder_layer_7_ln_1_bias" [id=176, type=get_attr]; -"177 _scale_29" [id=177, type=get_attr]; -"178 _zero_point_29" [id=178, type=get_attr]; -"179 _frozen_param29" [id=179, type=get_attr]; -"180 dequantize_per_channel_default_29" [id=180, type=dequantize_per_channel]; -"181 encoder_layers_encoder_layer_7_self_attention_in_proj_bias" [id=181, type=get_attr]; -"182 _scale_30" [id=182, type=get_attr]; -"183 _zero_point_30" [id=183, type=get_attr]; -"184 _frozen_param30" [id=184, type=get_attr]; -"185 dequantize_per_channel_default_30" [id=185, type=dequantize_per_channel]; -"186 encoder_layers_encoder_layer_7_self_attention_out_proj_bias" [id=186, type=get_attr]; -"187 encoder_layers_encoder_layer_7_ln_2_weight" [id=187, type=get_attr]; -"188 encoder_layers_encoder_layer_7_ln_2_bias" [id=188, type=get_attr]; -"189 _scale_31" [id=189, type=get_attr]; -"190 _zero_point_31" [id=190, type=get_attr]; -"191 _frozen_param31" [id=191, type=get_attr]; -"192 dequantize_per_channel_default_31" [id=192, type=dequantize_per_channel]; -"193 encoder_layers_encoder_layer_7_mlp_0_bias" [id=193, type=get_attr]; -"194 _scale_32" [id=194, type=get_attr]; -"195 _zero_point_32" [id=195, type=get_attr]; -"196 _frozen_param32" [id=196, type=get_attr]; -"197 dequantize_per_channel_default_32" [id=197, type=dequantize_per_channel]; -"198 encoder_layers_encoder_layer_7_mlp_3_bias" [id=198, type=get_attr]; -"199 encoder_layers_encoder_layer_8_ln_1_weight" [id=199, type=get_attr]; -"200 encoder_layers_encoder_layer_8_ln_1_bias" [id=200, type=get_attr]; -"201 _scale_33" [id=201, type=get_attr]; -"202 _zero_point_33" [id=202, type=get_attr]; -"203 _frozen_param33" [id=203, type=get_attr]; -"204 dequantize_per_channel_default_33" [id=204, type=dequantize_per_channel]; -"205 encoder_layers_encoder_layer_8_self_attention_in_proj_bias" [id=205, type=get_attr]; -"206 _scale_34" [id=206, type=get_attr]; -"207 _zero_point_34" [id=207, type=get_attr]; -"208 _frozen_param34" [id=208, type=get_attr]; -"209 dequantize_per_channel_default_34" [id=209, type=dequantize_per_channel]; -"210 encoder_layers_encoder_layer_8_self_attention_out_proj_bias" [id=210, type=get_attr]; -"211 encoder_layers_encoder_layer_8_ln_2_weight" [id=211, type=get_attr]; -"212 encoder_layers_encoder_layer_8_ln_2_bias" [id=212, type=get_attr]; -"213 _scale_35" [id=213, type=get_attr]; -"214 _zero_point_35" [id=214, type=get_attr]; -"215 _frozen_param35" [id=215, type=get_attr]; -"216 dequantize_per_channel_default_35" [id=216, type=dequantize_per_channel]; -"217 encoder_layers_encoder_layer_8_mlp_0_bias" [id=217, type=get_attr]; -"218 _scale_36" [id=218, type=get_attr]; -"219 _zero_point_36" [id=219, type=get_attr]; -"220 _frozen_param36" [id=220, type=get_attr]; -"221 dequantize_per_channel_default_36" [id=221, type=dequantize_per_channel]; -"222 encoder_layers_encoder_layer_8_mlp_3_bias" [id=222, type=get_attr]; -"223 encoder_layers_encoder_layer_9_ln_1_weight" [id=223, type=get_attr]; -"224 encoder_layers_encoder_layer_9_ln_1_bias" [id=224, type=get_attr]; -"225 _scale_37" [id=225, type=get_attr]; -"226 _zero_point_37" [id=226, type=get_attr]; -"227 _frozen_param37" [id=227, type=get_attr]; -"228 dequantize_per_channel_default_37" [id=228, type=dequantize_per_channel]; -"229 encoder_layers_encoder_layer_9_self_attention_in_proj_bias" [id=229, type=get_attr]; -"230 _scale_38" [id=230, type=get_attr]; -"231 _zero_point_38" [id=231, type=get_attr]; -"232 _frozen_param38" [id=232, type=get_attr]; -"233 dequantize_per_channel_default_38" [id=233, type=dequantize_per_channel]; -"234 encoder_layers_encoder_layer_9_self_attention_out_proj_bias" [id=234, type=get_attr]; -"235 encoder_layers_encoder_layer_9_ln_2_weight" [id=235, type=get_attr]; -"236 encoder_layers_encoder_layer_9_ln_2_bias" [id=236, type=get_attr]; -"237 _scale_39" [id=237, type=get_attr]; -"238 _zero_point_39" [id=238, type=get_attr]; -"239 _frozen_param39" [id=239, type=get_attr]; -"240 dequantize_per_channel_default_39" [id=240, type=dequantize_per_channel]; -"241 encoder_layers_encoder_layer_9_mlp_0_bias" [id=241, type=get_attr]; -"242 _scale_40" [id=242, type=get_attr]; -"243 _zero_point_40" [id=243, type=get_attr]; -"244 _frozen_param40" [id=244, type=get_attr]; -"245 dequantize_per_channel_default_40" [id=245, type=dequantize_per_channel]; -"246 encoder_layers_encoder_layer_9_mlp_3_bias" [id=246, type=get_attr]; -"247 encoder_layers_encoder_layer_10_ln_1_weight" [id=247, type=get_attr]; -"248 encoder_layers_encoder_layer_10_ln_1_bias" [id=248, type=get_attr]; -"249 _scale_41" [id=249, type=get_attr]; -"250 _zero_point_41" [id=250, type=get_attr]; -"251 _frozen_param41" [id=251, type=get_attr]; -"252 dequantize_per_channel_default_41" [id=252, type=dequantize_per_channel]; -"253 encoder_layers_encoder_layer_10_self_attention_in_proj_bias" [id=253, type=get_attr]; -"254 _scale_42" [id=254, type=get_attr]; -"255 _zero_point_42" [id=255, type=get_attr]; -"256 _frozen_param42" [id=256, type=get_attr]; -"257 dequantize_per_channel_default_42" [id=257, type=dequantize_per_channel]; -"258 encoder_layers_encoder_layer_10_self_attention_out_proj_bias" [id=258, type=get_attr]; -"259 encoder_layers_encoder_layer_10_ln_2_weight" [id=259, type=get_attr]; -"260 encoder_layers_encoder_layer_10_ln_2_bias" [id=260, type=get_attr]; -"261 _scale_43" [id=261, type=get_attr]; -"262 _zero_point_43" [id=262, type=get_attr]; -"263 _frozen_param43" [id=263, type=get_attr]; -"264 dequantize_per_channel_default_43" [id=264, type=dequantize_per_channel]; -"265 encoder_layers_encoder_layer_10_mlp_0_bias" [id=265, type=get_attr]; -"266 _scale_44" [id=266, type=get_attr]; -"267 _zero_point_44" [id=267, type=get_attr]; -"268 _frozen_param44" [id=268, type=get_attr]; -"269 dequantize_per_channel_default_44" [id=269, type=dequantize_per_channel]; -"270 encoder_layers_encoder_layer_10_mlp_3_bias" [id=270, type=get_attr]; -"271 encoder_layers_encoder_layer_11_ln_1_weight" [id=271, type=get_attr]; -"272 encoder_layers_encoder_layer_11_ln_1_bias" [id=272, type=get_attr]; -"273 _scale_45" [id=273, type=get_attr]; -"274 _zero_point_45" [id=274, type=get_attr]; -"275 _frozen_param45" [id=275, type=get_attr]; -"276 dequantize_per_channel_default_45" [id=276, type=dequantize_per_channel]; -"277 encoder_layers_encoder_layer_11_self_attention_in_proj_bias" [id=277, type=get_attr]; -"278 _scale_46" [id=278, type=get_attr]; -"279 _zero_point_46" [id=279, type=get_attr]; -"280 _frozen_param46" [id=280, type=get_attr]; -"281 dequantize_per_channel_default_46" [id=281, type=dequantize_per_channel]; -"282 encoder_layers_encoder_layer_11_self_attention_out_proj_bias" [id=282, type=get_attr]; -"283 encoder_layers_encoder_layer_11_ln_2_weight" [id=283, type=get_attr]; -"284 encoder_layers_encoder_layer_11_ln_2_bias" [id=284, type=get_attr]; -"285 _scale_47" [id=285, type=get_attr]; -"286 _zero_point_47" [id=286, type=get_attr]; -"287 _frozen_param47" [id=287, type=get_attr]; -"288 dequantize_per_channel_default_47" [id=288, type=dequantize_per_channel]; -"289 encoder_layers_encoder_layer_11_mlp_0_bias" [id=289, type=get_attr]; -"290 _scale_48" [id=290, type=get_attr]; -"291 _zero_point_48" [id=291, type=get_attr]; -"292 _frozen_param48" [id=292, type=get_attr]; -"293 dequantize_per_channel_default_48" [id=293, type=dequantize_per_channel]; -"294 encoder_layers_encoder_layer_11_mlp_3_bias" [id=294, type=get_attr]; -"295 encoder_ln_weight" [id=295, type=get_attr]; -"296 encoder_ln_bias" [id=296, type=get_attr]; -"297 _scale_49" [id=297, type=get_attr]; -"298 _zero_point_49" [id=298, type=get_attr]; -"299 _frozen_param49" [id=299, type=get_attr]; -"300 dequantize_per_channel_default_49" [id=300, type=dequantize_per_channel]; -"301 heads_head_bias" [id=301, type=get_attr]; -"302 x" [id=302, type=input]; -"303 quantize_per_tensor_default" [id=303, type=quantize_per_tensor]; -"304 dequantize_per_tensor_default" [id=304, type=dequantize_per_tensor]; -"305 conv2d" [id=305, type=conv2d]; -"306 reshape" [id=306, type=reshape]; -"307 permute" [id=307, type=permute]; -"308 expand" [id=308, type=expand]; -"309 cat" [id=309, type=cat]; -"310 add" [id=310, type=add]; -"311 dropout" [id=311, type=dropout]; -"312 layer_norm" [id=312, type=layer_norm]; -"313 quantize_per_tensor_default_1" [id=313, type=quantize_per_tensor]; -"314 dequantize_per_tensor_default_1" [id=314, type=dequantize_per_tensor]; -"315 transpose" [id=315, type=transpose]; -"316 linear" [id=316, type=linear]; -"317 unflatten" [id=317, type=unflatten]; -"318 unsqueeze" [id=318, type=unsqueeze]; -"319 transpose_1" [id=319, type=transpose]; -"320 squeeze" [id=320, type=squeeze]; -"321 contiguous" [id=321, type=contiguous]; -"322 quantize_per_tensor_default_2" [id=322, type=quantize_per_tensor]; -"323 dequantize_per_tensor_default_63" [id=323, type=dequantize_per_tensor]; -"324 dequantize_per_tensor_default_62" [id=324, type=dequantize_per_tensor]; -"325 select" [id=325, type=select]; -"326 select_1" [id=326, type=select]; -"327 select_2" [id=327, type=select]; -"328 view" [id=328, type=view]; -"329 transpose_2" [id=329, type=transpose]; -"330 view_1" [id=330, type=view]; -"331 transpose_3" [id=331, type=transpose]; -"332 view_2" [id=332, type=view]; -"333 transpose_4" [id=333, type=transpose]; -"334 view_3" [id=334, type=view]; -"335 view_4" [id=335, type=view]; -"336 view_5" [id=336, type=view]; -"337 scaled_dot_product_attention" [id=337, type=scaled_dot_product_attention]; -"338 quantize_per_tensor_default_3" [id=338, type=quantize_per_tensor]; -"339 dequantize_per_tensor_default_3" [id=339, type=dequantize_per_tensor]; -"340 permute_1" [id=340, type=permute]; -"341 view_6" [id=341, type=view]; -"342 linear_1" [id=342, type=linear]; -"343 view_7" [id=343, type=view]; -"344 transpose_5" [id=344, type=transpose]; -"345 dropout_1" [id=345, type=dropout]; -"346 add_1" [id=346, type=add]; -"347 layer_norm_1" [id=347, type=layer_norm]; -"348 quantize_per_tensor_default_4" [id=348, type=quantize_per_tensor]; -"349 dequantize_per_tensor_default_4" [id=349, type=dequantize_per_tensor]; -"350 linear_2" [id=350, type=linear]; -"351 gelu" [id=351, type=gelu]; -"352 quantize_per_tensor_default_5" [id=352, type=quantize_per_tensor]; -"353 dequantize_per_tensor_default_5" [id=353, type=dequantize_per_tensor]; -"354 dropout_2" [id=354, type=dropout]; -"355 linear_3" [id=355, type=linear]; -"356 dropout_3" [id=356, type=dropout]; -"357 add_2" [id=357, type=add]; -"358 layer_norm_2" [id=358, type=layer_norm]; -"359 quantize_per_tensor_default_6" [id=359, type=quantize_per_tensor]; -"360 dequantize_per_tensor_default_6" [id=360, type=dequantize_per_tensor]; -"361 transpose_6" [id=361, type=transpose]; -"362 linear_4" [id=362, type=linear]; -"363 unflatten_1" [id=363, type=unflatten]; -"364 unsqueeze_1" [id=364, type=unsqueeze]; -"365 transpose_7" [id=365, type=transpose]; -"366 squeeze_1" [id=366, type=squeeze]; -"367 contiguous_1" [id=367, type=contiguous]; -"368 quantize_per_tensor_default_7" [id=368, type=quantize_per_tensor]; -"369 dequantize_per_tensor_default_65" [id=369, type=dequantize_per_tensor]; -"370 dequantize_per_tensor_default_64" [id=370, type=dequantize_per_tensor]; -"371 select_3" [id=371, type=select]; -"372 select_4" [id=372, type=select]; -"373 select_5" [id=373, type=select]; -"374 view_8" [id=374, type=view]; -"375 transpose_8" [id=375, type=transpose]; -"376 view_9" [id=376, type=view]; -"377 transpose_9" [id=377, type=transpose]; -"378 view_10" [id=378, type=view]; -"379 transpose_10" [id=379, type=transpose]; -"380 view_11" [id=380, type=view]; -"381 view_12" [id=381, type=view]; -"382 view_13" [id=382, type=view]; -"383 scaled_dot_product_attention_1" [id=383, type=scaled_dot_product_attention]; -"384 quantize_per_tensor_default_8" [id=384, type=quantize_per_tensor]; -"385 dequantize_per_tensor_default_8" [id=385, type=dequantize_per_tensor]; -"386 permute_2" [id=386, type=permute]; -"387 view_14" [id=387, type=view]; -"388 linear_5" [id=388, type=linear]; -"389 view_15" [id=389, type=view]; -"390 transpose_11" [id=390, type=transpose]; -"391 dropout_4" [id=391, type=dropout]; -"392 add_3" [id=392, type=add]; -"393 layer_norm_3" [id=393, type=layer_norm]; -"394 quantize_per_tensor_default_9" [id=394, type=quantize_per_tensor]; -"395 dequantize_per_tensor_default_9" [id=395, type=dequantize_per_tensor]; -"396 linear_6" [id=396, type=linear]; -"397 gelu_1" [id=397, type=gelu]; -"398 quantize_per_tensor_default_10" [id=398, type=quantize_per_tensor]; -"399 dequantize_per_tensor_default_10" [id=399, type=dequantize_per_tensor]; -"400 dropout_5" [id=400, type=dropout]; -"401 linear_7" [id=401, type=linear]; -"402 dropout_6" [id=402, type=dropout]; -"403 add_4" [id=403, type=add]; -"404 layer_norm_4" [id=404, type=layer_norm]; -"405 quantize_per_tensor_default_11" [id=405, type=quantize_per_tensor]; -"406 dequantize_per_tensor_default_11" [id=406, type=dequantize_per_tensor]; -"407 transpose_12" [id=407, type=transpose]; -"408 linear_8" [id=408, type=linear]; -"409 unflatten_2" [id=409, type=unflatten]; -"410 unsqueeze_2" [id=410, type=unsqueeze]; -"411 transpose_13" [id=411, type=transpose]; -"412 squeeze_2" [id=412, type=squeeze]; -"413 contiguous_2" [id=413, type=contiguous]; -"414 quantize_per_tensor_default_12" [id=414, type=quantize_per_tensor]; -"415 dequantize_per_tensor_default_67" [id=415, type=dequantize_per_tensor]; -"416 dequantize_per_tensor_default_66" [id=416, type=dequantize_per_tensor]; -"417 select_6" [id=417, type=select]; -"418 select_7" [id=418, type=select]; -"419 select_8" [id=419, type=select]; -"420 view_16" [id=420, type=view]; -"421 transpose_14" [id=421, type=transpose]; -"422 view_17" [id=422, type=view]; -"423 transpose_15" [id=423, type=transpose]; -"424 view_18" [id=424, type=view]; -"425 transpose_16" [id=425, type=transpose]; -"426 view_19" [id=426, type=view]; -"427 view_20" [id=427, type=view]; -"428 view_21" [id=428, type=view]; -"429 scaled_dot_product_attention_2" [id=429, type=scaled_dot_product_attention]; -"430 quantize_per_tensor_default_13" [id=430, type=quantize_per_tensor]; -"431 dequantize_per_tensor_default_13" [id=431, type=dequantize_per_tensor]; -"432 permute_3" [id=432, type=permute]; -"433 view_22" [id=433, type=view]; -"434 linear_9" [id=434, type=linear]; -"435 view_23" [id=435, type=view]; -"436 transpose_17" [id=436, type=transpose]; -"437 dropout_7" [id=437, type=dropout]; -"438 add_5" [id=438, type=add]; -"439 layer_norm_5" [id=439, type=layer_norm]; -"440 quantize_per_tensor_default_14" [id=440, type=quantize_per_tensor]; -"441 dequantize_per_tensor_default_14" [id=441, type=dequantize_per_tensor]; -"442 linear_10" [id=442, type=linear]; -"443 gelu_2" [id=443, type=gelu]; -"444 quantize_per_tensor_default_15" [id=444, type=quantize_per_tensor]; -"445 dequantize_per_tensor_default_15" [id=445, type=dequantize_per_tensor]; -"446 dropout_8" [id=446, type=dropout]; -"447 linear_11" [id=447, type=linear]; -"448 dropout_9" [id=448, type=dropout]; -"449 add_6" [id=449, type=add]; -"450 layer_norm_6" [id=450, type=layer_norm]; -"451 quantize_per_tensor_default_16" [id=451, type=quantize_per_tensor]; -"452 dequantize_per_tensor_default_16" [id=452, type=dequantize_per_tensor]; -"453 transpose_18" [id=453, type=transpose]; -"454 linear_12" [id=454, type=linear]; -"455 unflatten_3" [id=455, type=unflatten]; -"456 unsqueeze_3" [id=456, type=unsqueeze]; -"457 transpose_19" [id=457, type=transpose]; -"458 squeeze_3" [id=458, type=squeeze]; -"459 contiguous_3" [id=459, type=contiguous]; -"460 quantize_per_tensor_default_17" [id=460, type=quantize_per_tensor]; -"461 dequantize_per_tensor_default_69" [id=461, type=dequantize_per_tensor]; -"462 dequantize_per_tensor_default_68" [id=462, type=dequantize_per_tensor]; -"463 select_9" [id=463, type=select]; -"464 select_10" [id=464, type=select]; -"465 select_11" [id=465, type=select]; -"466 view_24" [id=466, type=view]; -"467 transpose_20" [id=467, type=transpose]; -"468 view_25" [id=468, type=view]; -"469 transpose_21" [id=469, type=transpose]; -"470 view_26" [id=470, type=view]; -"471 transpose_22" [id=471, type=transpose]; -"472 view_27" [id=472, type=view]; -"473 view_28" [id=473, type=view]; -"474 view_29" [id=474, type=view]; -"475 scaled_dot_product_attention_3" [id=475, type=scaled_dot_product_attention]; -"476 quantize_per_tensor_default_18" [id=476, type=quantize_per_tensor]; -"477 dequantize_per_tensor_default_18" [id=477, type=dequantize_per_tensor]; -"478 permute_4" [id=478, type=permute]; -"479 view_30" [id=479, type=view]; -"480 linear_13" [id=480, type=linear]; -"481 view_31" [id=481, type=view]; -"482 transpose_23" [id=482, type=transpose]; -"483 dropout_10" [id=483, type=dropout]; -"484 add_7" [id=484, type=add]; -"485 layer_norm_7" [id=485, type=layer_norm]; -"486 quantize_per_tensor_default_19" [id=486, type=quantize_per_tensor]; -"487 dequantize_per_tensor_default_19" [id=487, type=dequantize_per_tensor]; -"488 linear_14" [id=488, type=linear]; -"489 gelu_3" [id=489, type=gelu]; -"490 quantize_per_tensor_default_20" [id=490, type=quantize_per_tensor]; -"491 dequantize_per_tensor_default_20" [id=491, type=dequantize_per_tensor]; -"492 dropout_11" [id=492, type=dropout]; -"493 linear_15" [id=493, type=linear]; -"494 dropout_12" [id=494, type=dropout]; -"495 add_8" [id=495, type=add]; -"496 layer_norm_8" [id=496, type=layer_norm]; -"497 quantize_per_tensor_default_21" [id=497, type=quantize_per_tensor]; -"498 dequantize_per_tensor_default_21" [id=498, type=dequantize_per_tensor]; -"499 transpose_24" [id=499, type=transpose]; -"500 linear_16" [id=500, type=linear]; -"501 unflatten_4" [id=501, type=unflatten]; -"502 unsqueeze_4" [id=502, type=unsqueeze]; -"503 transpose_25" [id=503, type=transpose]; -"504 squeeze_4" [id=504, type=squeeze]; -"505 contiguous_4" [id=505, type=contiguous]; -"506 quantize_per_tensor_default_22" [id=506, type=quantize_per_tensor]; -"507 dequantize_per_tensor_default_71" [id=507, type=dequantize_per_tensor]; -"508 dequantize_per_tensor_default_70" [id=508, type=dequantize_per_tensor]; -"509 select_12" [id=509, type=select]; -"510 select_13" [id=510, type=select]; -"511 select_14" [id=511, type=select]; -"512 view_32" [id=512, type=view]; -"513 transpose_26" [id=513, type=transpose]; -"514 view_33" [id=514, type=view]; -"515 transpose_27" [id=515, type=transpose]; -"516 view_34" [id=516, type=view]; -"517 transpose_28" [id=517, type=transpose]; -"518 view_35" [id=518, type=view]; -"519 view_36" [id=519, type=view]; -"520 view_37" [id=520, type=view]; -"521 scaled_dot_product_attention_4" [id=521, type=scaled_dot_product_attention]; -"522 quantize_per_tensor_default_23" [id=522, type=quantize_per_tensor]; -"523 dequantize_per_tensor_default_23" [id=523, type=dequantize_per_tensor]; -"524 permute_5" [id=524, type=permute]; -"525 view_38" [id=525, type=view]; -"526 linear_17" [id=526, type=linear]; -"527 view_39" [id=527, type=view]; -"528 transpose_29" [id=528, type=transpose]; -"529 dropout_13" [id=529, type=dropout]; -"530 add_9" [id=530, type=add]; -"531 layer_norm_9" [id=531, type=layer_norm]; -"532 quantize_per_tensor_default_24" [id=532, type=quantize_per_tensor]; -"533 dequantize_per_tensor_default_24" [id=533, type=dequantize_per_tensor]; -"534 linear_18" [id=534, type=linear]; -"535 gelu_4" [id=535, type=gelu]; -"536 quantize_per_tensor_default_25" [id=536, type=quantize_per_tensor]; -"537 dequantize_per_tensor_default_25" [id=537, type=dequantize_per_tensor]; -"538 dropout_14" [id=538, type=dropout]; -"539 linear_19" [id=539, type=linear]; -"540 dropout_15" [id=540, type=dropout]; -"541 add_10" [id=541, type=add]; -"542 layer_norm_10" [id=542, type=layer_norm]; -"543 quantize_per_tensor_default_26" [id=543, type=quantize_per_tensor]; -"544 dequantize_per_tensor_default_26" [id=544, type=dequantize_per_tensor]; -"545 transpose_30" [id=545, type=transpose]; -"546 linear_20" [id=546, type=linear]; -"547 unflatten_5" [id=547, type=unflatten]; -"548 unsqueeze_5" [id=548, type=unsqueeze]; -"549 transpose_31" [id=549, type=transpose]; -"550 squeeze_5" [id=550, type=squeeze]; -"551 contiguous_5" [id=551, type=contiguous]; -"552 quantize_per_tensor_default_27" [id=552, type=quantize_per_tensor]; -"553 dequantize_per_tensor_default_73" [id=553, type=dequantize_per_tensor]; -"554 dequantize_per_tensor_default_72" [id=554, type=dequantize_per_tensor]; -"555 select_15" [id=555, type=select]; -"556 select_16" [id=556, type=select]; -"557 select_17" [id=557, type=select]; -"558 view_40" [id=558, type=view]; -"559 transpose_32" [id=559, type=transpose]; -"560 view_41" [id=560, type=view]; -"561 transpose_33" [id=561, type=transpose]; -"562 view_42" [id=562, type=view]; -"563 transpose_34" [id=563, type=transpose]; -"564 view_43" [id=564, type=view]; -"565 view_44" [id=565, type=view]; -"566 view_45" [id=566, type=view]; -"567 scaled_dot_product_attention_5" [id=567, type=scaled_dot_product_attention]; -"568 quantize_per_tensor_default_28" [id=568, type=quantize_per_tensor]; -"569 dequantize_per_tensor_default_28" [id=569, type=dequantize_per_tensor]; -"570 permute_6" [id=570, type=permute]; -"571 view_46" [id=571, type=view]; -"572 linear_21" [id=572, type=linear]; -"573 view_47" [id=573, type=view]; -"574 transpose_35" [id=574, type=transpose]; -"575 dropout_16" [id=575, type=dropout]; -"576 add_11" [id=576, type=add]; -"577 layer_norm_11" [id=577, type=layer_norm]; -"578 quantize_per_tensor_default_29" [id=578, type=quantize_per_tensor]; -"579 dequantize_per_tensor_default_29" [id=579, type=dequantize_per_tensor]; -"580 linear_22" [id=580, type=linear]; -"581 gelu_5" [id=581, type=gelu]; -"582 quantize_per_tensor_default_30" [id=582, type=quantize_per_tensor]; -"583 dequantize_per_tensor_default_30" [id=583, type=dequantize_per_tensor]; -"584 dropout_17" [id=584, type=dropout]; -"585 linear_23" [id=585, type=linear]; -"586 dropout_18" [id=586, type=dropout]; -"587 add_12" [id=587, type=add]; -"588 layer_norm_12" [id=588, type=layer_norm]; -"589 quantize_per_tensor_default_31" [id=589, type=quantize_per_tensor]; -"590 dequantize_per_tensor_default_31" [id=590, type=dequantize_per_tensor]; -"591 transpose_36" [id=591, type=transpose]; -"592 linear_24" [id=592, type=linear]; -"593 unflatten_6" [id=593, type=unflatten]; -"594 unsqueeze_6" [id=594, type=unsqueeze]; -"595 transpose_37" [id=595, type=transpose]; -"596 squeeze_6" [id=596, type=squeeze]; -"597 contiguous_6" [id=597, type=contiguous]; -"598 quantize_per_tensor_default_32" [id=598, type=quantize_per_tensor]; -"599 dequantize_per_tensor_default_75" [id=599, type=dequantize_per_tensor]; -"600 dequantize_per_tensor_default_74" [id=600, type=dequantize_per_tensor]; -"601 select_18" [id=601, type=select]; -"602 select_19" [id=602, type=select]; -"603 select_20" [id=603, type=select]; -"604 view_48" [id=604, type=view]; -"605 transpose_38" [id=605, type=transpose]; -"606 view_49" [id=606, type=view]; -"607 transpose_39" [id=607, type=transpose]; -"608 view_50" [id=608, type=view]; -"609 transpose_40" [id=609, type=transpose]; -"610 view_51" [id=610, type=view]; -"611 view_52" [id=611, type=view]; -"612 view_53" [id=612, type=view]; -"613 scaled_dot_product_attention_6" [id=613, type=scaled_dot_product_attention]; -"614 quantize_per_tensor_default_33" [id=614, type=quantize_per_tensor]; -"615 dequantize_per_tensor_default_33" [id=615, type=dequantize_per_tensor]; -"616 permute_7" [id=616, type=permute]; -"617 view_54" [id=617, type=view]; -"618 linear_25" [id=618, type=linear]; -"619 view_55" [id=619, type=view]; -"620 transpose_41" [id=620, type=transpose]; -"621 dropout_19" [id=621, type=dropout]; -"622 add_13" [id=622, type=add]; -"623 layer_norm_13" [id=623, type=layer_norm]; -"624 quantize_per_tensor_default_34" [id=624, type=quantize_per_tensor]; -"625 dequantize_per_tensor_default_34" [id=625, type=dequantize_per_tensor]; -"626 linear_26" [id=626, type=linear]; -"627 gelu_6" [id=627, type=gelu]; -"628 quantize_per_tensor_default_35" [id=628, type=quantize_per_tensor]; -"629 dequantize_per_tensor_default_35" [id=629, type=dequantize_per_tensor]; -"630 dropout_20" [id=630, type=dropout]; -"631 linear_27" [id=631, type=linear]; -"632 dropout_21" [id=632, type=dropout]; -"633 add_14" [id=633, type=add]; -"634 layer_norm_14" [id=634, type=layer_norm]; -"635 quantize_per_tensor_default_36" [id=635, type=quantize_per_tensor]; -"636 dequantize_per_tensor_default_36" [id=636, type=dequantize_per_tensor]; -"637 transpose_42" [id=637, type=transpose]; -"638 linear_28" [id=638, type=linear]; -"639 unflatten_7" [id=639, type=unflatten]; -"640 unsqueeze_7" [id=640, type=unsqueeze]; -"641 transpose_43" [id=641, type=transpose]; -"642 squeeze_7" [id=642, type=squeeze]; -"643 contiguous_7" [id=643, type=contiguous]; -"644 quantize_per_tensor_default_37" [id=644, type=quantize_per_tensor]; -"645 dequantize_per_tensor_default_77" [id=645, type=dequantize_per_tensor]; -"646 dequantize_per_tensor_default_76" [id=646, type=dequantize_per_tensor]; -"647 select_21" [id=647, type=select]; -"648 select_22" [id=648, type=select]; -"649 select_23" [id=649, type=select]; -"650 view_56" [id=650, type=view]; -"651 transpose_44" [id=651, type=transpose]; -"652 view_57" [id=652, type=view]; -"653 transpose_45" [id=653, type=transpose]; -"654 view_58" [id=654, type=view]; -"655 transpose_46" [id=655, type=transpose]; -"656 view_59" [id=656, type=view]; -"657 view_60" [id=657, type=view]; -"658 view_61" [id=658, type=view]; -"659 scaled_dot_product_attention_7" [id=659, type=scaled_dot_product_attention]; -"660 quantize_per_tensor_default_38" [id=660, type=quantize_per_tensor]; -"661 dequantize_per_tensor_default_38" [id=661, type=dequantize_per_tensor]; -"662 permute_8" [id=662, type=permute]; -"663 view_62" [id=663, type=view]; -"664 linear_29" [id=664, type=linear]; -"665 view_63" [id=665, type=view]; -"666 transpose_47" [id=666, type=transpose]; -"667 dropout_22" [id=667, type=dropout]; -"668 add_15" [id=668, type=add]; -"669 layer_norm_15" [id=669, type=layer_norm]; -"670 quantize_per_tensor_default_39" [id=670, type=quantize_per_tensor]; -"671 dequantize_per_tensor_default_39" [id=671, type=dequantize_per_tensor]; -"672 linear_30" [id=672, type=linear]; -"673 gelu_7" [id=673, type=gelu]; -"674 quantize_per_tensor_default_40" [id=674, type=quantize_per_tensor]; -"675 dequantize_per_tensor_default_40" [id=675, type=dequantize_per_tensor]; -"676 dropout_23" [id=676, type=dropout]; -"677 linear_31" [id=677, type=linear]; -"678 dropout_24" [id=678, type=dropout]; -"679 add_16" [id=679, type=add]; -"680 layer_norm_16" [id=680, type=layer_norm]; -"681 quantize_per_tensor_default_41" [id=681, type=quantize_per_tensor]; -"682 dequantize_per_tensor_default_41" [id=682, type=dequantize_per_tensor]; -"683 transpose_48" [id=683, type=transpose]; -"684 linear_32" [id=684, type=linear]; -"685 unflatten_8" [id=685, type=unflatten]; -"686 unsqueeze_8" [id=686, type=unsqueeze]; -"687 transpose_49" [id=687, type=transpose]; -"688 squeeze_8" [id=688, type=squeeze]; -"689 contiguous_8" [id=689, type=contiguous]; -"690 quantize_per_tensor_default_42" [id=690, type=quantize_per_tensor]; -"691 dequantize_per_tensor_default_79" [id=691, type=dequantize_per_tensor]; -"692 dequantize_per_tensor_default_78" [id=692, type=dequantize_per_tensor]; -"693 select_24" [id=693, type=select]; -"694 select_25" [id=694, type=select]; -"695 select_26" [id=695, type=select]; -"696 view_64" [id=696, type=view]; -"697 transpose_50" [id=697, type=transpose]; -"698 view_65" [id=698, type=view]; -"699 transpose_51" [id=699, type=transpose]; -"700 view_66" [id=700, type=view]; -"701 transpose_52" [id=701, type=transpose]; -"702 view_67" [id=702, type=view]; -"703 view_68" [id=703, type=view]; -"704 view_69" [id=704, type=view]; -"705 scaled_dot_product_attention_8" [id=705, type=scaled_dot_product_attention]; -"706 quantize_per_tensor_default_43" [id=706, type=quantize_per_tensor]; -"707 dequantize_per_tensor_default_43" [id=707, type=dequantize_per_tensor]; -"708 permute_9" [id=708, type=permute]; -"709 view_70" [id=709, type=view]; -"710 linear_33" [id=710, type=linear]; -"711 view_71" [id=711, type=view]; -"712 transpose_53" [id=712, type=transpose]; -"713 dropout_25" [id=713, type=dropout]; -"714 add_17" [id=714, type=add]; -"715 layer_norm_17" [id=715, type=layer_norm]; -"716 quantize_per_tensor_default_44" [id=716, type=quantize_per_tensor]; -"717 dequantize_per_tensor_default_44" [id=717, type=dequantize_per_tensor]; -"718 linear_34" [id=718, type=linear]; -"719 gelu_8" [id=719, type=gelu]; -"720 quantize_per_tensor_default_45" [id=720, type=quantize_per_tensor]; -"721 dequantize_per_tensor_default_45" [id=721, type=dequantize_per_tensor]; -"722 dropout_26" [id=722, type=dropout]; -"723 linear_35" [id=723, type=linear]; -"724 dropout_27" [id=724, type=dropout]; -"725 add_18" [id=725, type=add]; -"726 layer_norm_18" [id=726, type=layer_norm]; -"727 quantize_per_tensor_default_46" [id=727, type=quantize_per_tensor]; -"728 dequantize_per_tensor_default_46" [id=728, type=dequantize_per_tensor]; -"729 transpose_54" [id=729, type=transpose]; -"730 linear_36" [id=730, type=linear]; -"731 unflatten_9" [id=731, type=unflatten]; -"732 unsqueeze_9" [id=732, type=unsqueeze]; -"733 transpose_55" [id=733, type=transpose]; -"734 squeeze_9" [id=734, type=squeeze]; -"735 contiguous_9" [id=735, type=contiguous]; -"736 quantize_per_tensor_default_47" [id=736, type=quantize_per_tensor]; -"737 dequantize_per_tensor_default_81" [id=737, type=dequantize_per_tensor]; -"738 dequantize_per_tensor_default_80" [id=738, type=dequantize_per_tensor]; -"739 select_27" [id=739, type=select]; -"740 select_28" [id=740, type=select]; -"741 select_29" [id=741, type=select]; -"742 view_72" [id=742, type=view]; -"743 transpose_56" [id=743, type=transpose]; -"744 view_73" [id=744, type=view]; -"745 transpose_57" [id=745, type=transpose]; -"746 view_74" [id=746, type=view]; -"747 transpose_58" [id=747, type=transpose]; -"748 view_75" [id=748, type=view]; -"749 view_76" [id=749, type=view]; -"750 view_77" [id=750, type=view]; -"751 scaled_dot_product_attention_9" [id=751, type=scaled_dot_product_attention]; -"752 quantize_per_tensor_default_48" [id=752, type=quantize_per_tensor]; -"753 dequantize_per_tensor_default_48" [id=753, type=dequantize_per_tensor]; -"754 permute_10" [id=754, type=permute]; -"755 view_78" [id=755, type=view]; -"756 linear_37" [id=756, type=linear]; -"757 view_79" [id=757, type=view]; -"758 transpose_59" [id=758, type=transpose]; -"759 dropout_28" [id=759, type=dropout]; -"760 add_19" [id=760, type=add]; -"761 layer_norm_19" [id=761, type=layer_norm]; -"762 quantize_per_tensor_default_49" [id=762, type=quantize_per_tensor]; -"763 dequantize_per_tensor_default_49" [id=763, type=dequantize_per_tensor]; -"764 linear_38" [id=764, type=linear]; -"765 gelu_9" [id=765, type=gelu]; -"766 quantize_per_tensor_default_50" [id=766, type=quantize_per_tensor]; -"767 dequantize_per_tensor_default_50" [id=767, type=dequantize_per_tensor]; -"768 dropout_29" [id=768, type=dropout]; -"769 linear_39" [id=769, type=linear]; -"770 dropout_30" [id=770, type=dropout]; -"771 add_20" [id=771, type=add]; -"772 layer_norm_20" [id=772, type=layer_norm]; -"773 quantize_per_tensor_default_51" [id=773, type=quantize_per_tensor]; -"774 dequantize_per_tensor_default_51" [id=774, type=dequantize_per_tensor]; -"775 transpose_60" [id=775, type=transpose]; -"776 linear_40" [id=776, type=linear]; -"777 unflatten_10" [id=777, type=unflatten]; -"778 unsqueeze_10" [id=778, type=unsqueeze]; -"779 transpose_61" [id=779, type=transpose]; -"780 squeeze_10" [id=780, type=squeeze]; -"781 contiguous_10" [id=781, type=contiguous]; -"782 quantize_per_tensor_default_52" [id=782, type=quantize_per_tensor]; -"783 dequantize_per_tensor_default_83" [id=783, type=dequantize_per_tensor]; -"784 dequantize_per_tensor_default_82" [id=784, type=dequantize_per_tensor]; -"785 select_30" [id=785, type=select]; -"786 select_31" [id=786, type=select]; -"787 select_32" [id=787, type=select]; -"788 view_80" [id=788, type=view]; -"789 transpose_62" [id=789, type=transpose]; -"790 view_81" [id=790, type=view]; -"791 transpose_63" [id=791, type=transpose]; -"792 view_82" [id=792, type=view]; -"793 transpose_64" [id=793, type=transpose]; -"794 view_83" [id=794, type=view]; -"795 view_84" [id=795, type=view]; -"796 view_85" [id=796, type=view]; -"797 scaled_dot_product_attention_10" [id=797, type=scaled_dot_product_attention]; -"798 quantize_per_tensor_default_53" [id=798, type=quantize_per_tensor]; -"799 dequantize_per_tensor_default_53" [id=799, type=dequantize_per_tensor]; -"800 permute_11" [id=800, type=permute]; -"801 view_86" [id=801, type=view]; -"802 linear_41" [id=802, type=linear]; -"803 view_87" [id=803, type=view]; -"804 transpose_65" [id=804, type=transpose]; -"805 dropout_31" [id=805, type=dropout]; -"806 add_21" [id=806, type=add]; -"807 layer_norm_21" [id=807, type=layer_norm]; -"808 quantize_per_tensor_default_54" [id=808, type=quantize_per_tensor]; -"809 dequantize_per_tensor_default_54" [id=809, type=dequantize_per_tensor]; -"810 linear_42" [id=810, type=linear]; -"811 gelu_10" [id=811, type=gelu]; -"812 quantize_per_tensor_default_55" [id=812, type=quantize_per_tensor]; -"813 dequantize_per_tensor_default_55" [id=813, type=dequantize_per_tensor]; -"814 dropout_32" [id=814, type=dropout]; -"815 linear_43" [id=815, type=linear]; -"816 dropout_33" [id=816, type=dropout]; -"817 add_22" [id=817, type=add]; -"818 layer_norm_22" [id=818, type=layer_norm]; -"819 quantize_per_tensor_default_56" [id=819, type=quantize_per_tensor]; -"820 dequantize_per_tensor_default_56" [id=820, type=dequantize_per_tensor]; -"821 transpose_66" [id=821, type=transpose]; -"822 linear_44" [id=822, type=linear]; -"823 unflatten_11" [id=823, type=unflatten]; -"824 unsqueeze_11" [id=824, type=unsqueeze]; -"825 transpose_67" [id=825, type=transpose]; -"826 squeeze_11" [id=826, type=squeeze]; -"827 contiguous_11" [id=827, type=contiguous]; -"828 quantize_per_tensor_default_57" [id=828, type=quantize_per_tensor]; -"829 dequantize_per_tensor_default_85" [id=829, type=dequantize_per_tensor]; -"830 dequantize_per_tensor_default_84" [id=830, type=dequantize_per_tensor]; -"831 select_33" [id=831, type=select]; -"832 select_34" [id=832, type=select]; -"833 select_35" [id=833, type=select]; -"834 view_88" [id=834, type=view]; -"835 transpose_68" [id=835, type=transpose]; -"836 view_89" [id=836, type=view]; -"837 transpose_69" [id=837, type=transpose]; -"838 view_90" [id=838, type=view]; -"839 transpose_70" [id=839, type=transpose]; -"840 view_91" [id=840, type=view]; -"841 view_92" [id=841, type=view]; -"842 view_93" [id=842, type=view]; -"843 scaled_dot_product_attention_11" [id=843, type=scaled_dot_product_attention]; -"844 quantize_per_tensor_default_58" [id=844, type=quantize_per_tensor]; -"845 dequantize_per_tensor_default_58" [id=845, type=dequantize_per_tensor]; -"846 permute_12" [id=846, type=permute]; -"847 view_94" [id=847, type=view]; -"848 linear_45" [id=848, type=linear]; -"849 view_95" [id=849, type=view]; -"850 transpose_71" [id=850, type=transpose]; -"851 dropout_34" [id=851, type=dropout]; -"852 add_23" [id=852, type=add]; -"853 layer_norm_23" [id=853, type=layer_norm]; -"854 quantize_per_tensor_default_59" [id=854, type=quantize_per_tensor]; -"855 dequantize_per_tensor_default_59" [id=855, type=dequantize_per_tensor]; -"856 linear_46" [id=856, type=linear]; -"857 gelu_11" [id=857, type=gelu]; -"858 quantize_per_tensor_default_60" [id=858, type=quantize_per_tensor]; -"859 dequantize_per_tensor_default_60" [id=859, type=dequantize_per_tensor]; -"860 dropout_35" [id=860, type=dropout]; -"861 linear_47" [id=861, type=linear]; -"862 dropout_36" [id=862, type=dropout]; -"863 add_24" [id=863, type=add]; -"864 layer_norm_24" [id=864, type=layer_norm]; -"865 quantize_per_tensor_default_61" [id=865, type=quantize_per_tensor]; -"866 dequantize_per_tensor_default_61" [id=866, type=dequantize_per_tensor]; -"867 slice_1" [id=867, type=slice]; -"868 select_36" [id=868, type=select]; -"869 linear_48" [id=869, type=linear]; -"870 output_1" [id=870, type=output]; -"0 class_token" -> "308 expand" [label="(1, 1, 768)", style=solid]; -"1 encoder_pos_embedding" -> "310 add" [label="(1, 197, 768)", style=solid]; -"2 _scale_0" -> "5 dequantize_per_channel_default" [label="(768,)", style=solid]; -"3 _zero_point_0" -> "5 dequantize_per_channel_default" [label="(768,)", style=solid]; -"4 _frozen_param0" -> "5 dequantize_per_channel_default" [label="(768, 3, 16, 16)", style=solid]; -"5 dequantize_per_channel_default" -> "305 conv2d" [label=None, style=solid]; -"6 conv_proj_bias" -> "305 conv2d" [label="(768,)", style=solid]; -"7 encoder_layers_encoder_layer_0_ln_1_weight" -> "312 layer_norm" [label="(768,)", style=solid]; -"8 encoder_layers_encoder_layer_0_ln_1_bias" -> "312 layer_norm" [label="(768,)", style=solid]; -"9 _scale_1" -> "12 dequantize_per_channel_default_1" [label="(2304,)", style=solid]; -"10 _zero_point_1" -> "12 dequantize_per_channel_default_1" [label="(2304,)", style=solid]; -"11 _frozen_param1" -> "12 dequantize_per_channel_default_1" [label="(2304, 768)", style=solid]; -"12 dequantize_per_channel_default_1" -> "316 linear" [label=None, style=solid]; -"13 encoder_layers_encoder_layer_0_self_attention_in_proj_bias" -> "316 linear" [label="(2304,)", style=solid]; -"14 _scale_2" -> "17 dequantize_per_channel_default_2" [label="(768,)", style=solid]; -"15 _zero_point_2" -> "17 dequantize_per_channel_default_2" [label="(768,)", style=solid]; -"16 _frozen_param2" -> "17 dequantize_per_channel_default_2" [label="(768, 768)", style=solid]; -"17 dequantize_per_channel_default_2" -> "342 linear_1" [label=None, style=solid]; -"18 encoder_layers_encoder_layer_0_self_attention_out_proj_bias" -> "342 linear_1" [label="(768,)", style=solid]; -"19 encoder_layers_encoder_layer_0_ln_2_weight" -> "347 layer_norm_1" [label="(768,)", style=solid]; -"20 encoder_layers_encoder_layer_0_ln_2_bias" -> "347 layer_norm_1" [label="(768,)", style=solid]; -"21 _scale_3" -> "24 dequantize_per_channel_default_3" [label="(3072,)", style=solid]; -"22 _zero_point_3" -> "24 dequantize_per_channel_default_3" [label="(3072,)", style=solid]; -"23 _frozen_param3" -> "24 dequantize_per_channel_default_3" [label="(3072, 768)", style=solid]; -"24 dequantize_per_channel_default_3" -> "350 linear_2" [label=None, style=solid]; -"25 encoder_layers_encoder_layer_0_mlp_0_bias" -> "350 linear_2" [label="(3072,)", style=solid]; -"26 _scale_4" -> "29 dequantize_per_channel_default_4" [label="(768,)", style=solid]; -"27 _zero_point_4" -> "29 dequantize_per_channel_default_4" [label="(768,)", style=solid]; -"28 _frozen_param4" -> "29 dequantize_per_channel_default_4" [label="(768, 3072)", style=solid]; -"29 dequantize_per_channel_default_4" -> "355 linear_3" [label=None, style=solid]; -"30 encoder_layers_encoder_layer_0_mlp_3_bias" -> "355 linear_3" [label="(768,)", style=solid]; -"31 encoder_layers_encoder_layer_1_ln_1_weight" -> "358 layer_norm_2" [label="(768,)", style=solid]; -"32 encoder_layers_encoder_layer_1_ln_1_bias" -> "358 layer_norm_2" [label="(768,)", style=solid]; -"33 _scale_5" -> "36 dequantize_per_channel_default_5" [label="(2304,)", style=solid]; -"34 _zero_point_5" -> "36 dequantize_per_channel_default_5" [label="(2304,)", style=solid]; -"35 _frozen_param5" -> "36 dequantize_per_channel_default_5" [label="(2304, 768)", style=solid]; -"36 dequantize_per_channel_default_5" -> "362 linear_4" [label=None, style=solid]; -"37 encoder_layers_encoder_layer_1_self_attention_in_proj_bias" -> "362 linear_4" [label="(2304,)", style=solid]; -"38 _scale_6" -> "41 dequantize_per_channel_default_6" [label="(768,)", style=solid]; -"39 _zero_point_6" -> "41 dequantize_per_channel_default_6" [label="(768,)", style=solid]; -"40 _frozen_param6" -> "41 dequantize_per_channel_default_6" [label="(768, 768)", style=solid]; -"41 dequantize_per_channel_default_6" -> "388 linear_5" [label=None, style=solid]; -"42 encoder_layers_encoder_layer_1_self_attention_out_proj_bias" -> "388 linear_5" [label="(768,)", style=solid]; -"43 encoder_layers_encoder_layer_1_ln_2_weight" -> "393 layer_norm_3" [label="(768,)", style=solid]; -"44 encoder_layers_encoder_layer_1_ln_2_bias" -> "393 layer_norm_3" [label="(768,)", style=solid]; -"45 _scale_7" -> "48 dequantize_per_channel_default_7" [label="(3072,)", style=solid]; -"46 _zero_point_7" -> "48 dequantize_per_channel_default_7" [label="(3072,)", style=solid]; -"47 _frozen_param7" -> "48 dequantize_per_channel_default_7" [label="(3072, 768)", style=solid]; -"48 dequantize_per_channel_default_7" -> "396 linear_6" [label=None, style=solid]; -"49 encoder_layers_encoder_layer_1_mlp_0_bias" -> "396 linear_6" [label="(3072,)", style=solid]; -"50 _scale_8" -> "53 dequantize_per_channel_default_8" [label="(768,)", style=solid]; -"51 _zero_point_8" -> "53 dequantize_per_channel_default_8" [label="(768,)", style=solid]; -"52 _frozen_param8" -> "53 dequantize_per_channel_default_8" [label="(768, 3072)", style=solid]; -"53 dequantize_per_channel_default_8" -> "401 linear_7" [label=None, style=solid]; -"54 encoder_layers_encoder_layer_1_mlp_3_bias" -> "401 linear_7" [label="(768,)", style=solid]; -"55 encoder_layers_encoder_layer_2_ln_1_weight" -> "404 layer_norm_4" [label="(768,)", style=solid]; -"56 encoder_layers_encoder_layer_2_ln_1_bias" -> "404 layer_norm_4" [label="(768,)", style=solid]; -"57 _scale_9" -> "60 dequantize_per_channel_default_9" [label="(2304,)", style=solid]; -"58 _zero_point_9" -> "60 dequantize_per_channel_default_9" [label="(2304,)", style=solid]; -"59 _frozen_param9" -> "60 dequantize_per_channel_default_9" [label="(2304, 768)", style=solid]; -"60 dequantize_per_channel_default_9" -> "408 linear_8" [label=None, style=solid]; -"61 encoder_layers_encoder_layer_2_self_attention_in_proj_bias" -> "408 linear_8" [label="(2304,)", style=solid]; -"62 _scale_10" -> "65 dequantize_per_channel_default_10" [label="(768,)", style=solid]; -"63 _zero_point_10" -> "65 dequantize_per_channel_default_10" [label="(768,)", style=solid]; -"64 _frozen_param10" -> "65 dequantize_per_channel_default_10" [label="(768, 768)", style=solid]; -"65 dequantize_per_channel_default_10" -> "434 linear_9" [label=None, style=solid]; -"66 encoder_layers_encoder_layer_2_self_attention_out_proj_bias" -> "434 linear_9" [label="(768,)", style=solid]; -"67 encoder_layers_encoder_layer_2_ln_2_weight" -> "439 layer_norm_5" [label="(768,)", style=solid]; -"68 encoder_layers_encoder_layer_2_ln_2_bias" -> "439 layer_norm_5" [label="(768,)", style=solid]; -"69 _scale_11" -> "72 dequantize_per_channel_default_11" [label="(3072,)", style=solid]; -"70 _zero_point_11" -> "72 dequantize_per_channel_default_11" [label="(3072,)", style=solid]; -"71 _frozen_param11" -> "72 dequantize_per_channel_default_11" [label="(3072, 768)", style=solid]; -"72 dequantize_per_channel_default_11" -> "442 linear_10" [label=None, style=solid]; -"73 encoder_layers_encoder_layer_2_mlp_0_bias" -> "442 linear_10" [label="(3072,)", style=solid]; -"74 _scale_12" -> "77 dequantize_per_channel_default_12" [label="(768,)", style=solid]; -"75 _zero_point_12" -> "77 dequantize_per_channel_default_12" [label="(768,)", style=solid]; -"76 _frozen_param12" -> "77 dequantize_per_channel_default_12" [label="(768, 3072)", style=solid]; -"77 dequantize_per_channel_default_12" -> "447 linear_11" [label=None, style=solid]; -"78 encoder_layers_encoder_layer_2_mlp_3_bias" -> "447 linear_11" [label="(768,)", style=solid]; -"79 encoder_layers_encoder_layer_3_ln_1_weight" -> "450 layer_norm_6" [label="(768,)", style=solid]; -"80 encoder_layers_encoder_layer_3_ln_1_bias" -> "450 layer_norm_6" [label="(768,)", style=solid]; -"81 _scale_13" -> "84 dequantize_per_channel_default_13" [label="(2304,)", style=solid]; -"82 _zero_point_13" -> "84 dequantize_per_channel_default_13" [label="(2304,)", style=solid]; -"83 _frozen_param13" -> "84 dequantize_per_channel_default_13" [label="(2304, 768)", style=solid]; -"84 dequantize_per_channel_default_13" -> "454 linear_12" [label=None, style=solid]; -"85 encoder_layers_encoder_layer_3_self_attention_in_proj_bias" -> "454 linear_12" [label="(2304,)", style=solid]; -"86 _scale_14" -> "89 dequantize_per_channel_default_14" [label="(768,)", style=solid]; -"87 _zero_point_14" -> "89 dequantize_per_channel_default_14" [label="(768,)", style=solid]; -"88 _frozen_param14" -> "89 dequantize_per_channel_default_14" [label="(768, 768)", style=solid]; -"89 dequantize_per_channel_default_14" -> "480 linear_13" [label=None, style=solid]; -"90 encoder_layers_encoder_layer_3_self_attention_out_proj_bias" -> "480 linear_13" [label="(768,)", style=solid]; -"91 encoder_layers_encoder_layer_3_ln_2_weight" -> "485 layer_norm_7" [label="(768,)", style=solid]; -"92 encoder_layers_encoder_layer_3_ln_2_bias" -> "485 layer_norm_7" [label="(768,)", style=solid]; -"93 _scale_15" -> "96 dequantize_per_channel_default_15" [label="(3072,)", style=solid]; -"94 _zero_point_15" -> "96 dequantize_per_channel_default_15" [label="(3072,)", style=solid]; -"95 _frozen_param15" -> "96 dequantize_per_channel_default_15" [label="(3072, 768)", style=solid]; -"96 dequantize_per_channel_default_15" -> "488 linear_14" [label=None, style=solid]; -"97 encoder_layers_encoder_layer_3_mlp_0_bias" -> "488 linear_14" [label="(3072,)", style=solid]; -"98 _scale_16" -> "101 dequantize_per_channel_default_16" [label="(768,)", style=solid]; -"99 _zero_point_16" -> "101 dequantize_per_channel_default_16" [label="(768,)", style=solid]; -"100 _frozen_param16" -> "101 dequantize_per_channel_default_16" [label="(768, 3072)", style=solid]; -"101 dequantize_per_channel_default_16" -> "493 linear_15" [label=None, style=solid]; -"102 encoder_layers_encoder_layer_3_mlp_3_bias" -> "493 linear_15" [label="(768,)", style=solid]; -"103 encoder_layers_encoder_layer_4_ln_1_weight" -> "496 layer_norm_8" [label="(768,)", style=solid]; -"104 encoder_layers_encoder_layer_4_ln_1_bias" -> "496 layer_norm_8" [label="(768,)", style=solid]; -"105 _scale_17" -> "108 dequantize_per_channel_default_17" [label="(2304,)", style=solid]; -"106 _zero_point_17" -> "108 dequantize_per_channel_default_17" [label="(2304,)", style=solid]; -"107 _frozen_param17" -> "108 dequantize_per_channel_default_17" [label="(2304, 768)", style=solid]; -"108 dequantize_per_channel_default_17" -> "500 linear_16" [label=None, style=solid]; -"109 encoder_layers_encoder_layer_4_self_attention_in_proj_bias" -> "500 linear_16" [label="(2304,)", style=solid]; -"110 _scale_18" -> "113 dequantize_per_channel_default_18" [label="(768,)", style=solid]; -"111 _zero_point_18" -> "113 dequantize_per_channel_default_18" [label="(768,)", style=solid]; -"112 _frozen_param18" -> "113 dequantize_per_channel_default_18" [label="(768, 768)", style=solid]; -"113 dequantize_per_channel_default_18" -> "526 linear_17" [label=None, style=solid]; -"114 encoder_layers_encoder_layer_4_self_attention_out_proj_bias" -> "526 linear_17" [label="(768,)", style=solid]; -"115 encoder_layers_encoder_layer_4_ln_2_weight" -> "531 layer_norm_9" [label="(768,)", style=solid]; -"116 encoder_layers_encoder_layer_4_ln_2_bias" -> "531 layer_norm_9" [label="(768,)", style=solid]; -"117 _scale_19" -> "120 dequantize_per_channel_default_19" [label="(3072,)", style=solid]; -"118 _zero_point_19" -> "120 dequantize_per_channel_default_19" [label="(3072,)", style=solid]; -"119 _frozen_param19" -> "120 dequantize_per_channel_default_19" [label="(3072, 768)", style=solid]; -"120 dequantize_per_channel_default_19" -> "534 linear_18" [label=None, style=solid]; -"121 encoder_layers_encoder_layer_4_mlp_0_bias" -> "534 linear_18" [label="(3072,)", style=solid]; -"122 _scale_20" -> "125 dequantize_per_channel_default_20" [label="(768,)", style=solid]; -"123 _zero_point_20" -> "125 dequantize_per_channel_default_20" [label="(768,)", style=solid]; -"124 _frozen_param20" -> "125 dequantize_per_channel_default_20" [label="(768, 3072)", style=solid]; -"125 dequantize_per_channel_default_20" -> "539 linear_19" [label=None, style=solid]; -"126 encoder_layers_encoder_layer_4_mlp_3_bias" -> "539 linear_19" [label="(768,)", style=solid]; -"127 encoder_layers_encoder_layer_5_ln_1_weight" -> "542 layer_norm_10" [label="(768,)", style=solid]; -"128 encoder_layers_encoder_layer_5_ln_1_bias" -> "542 layer_norm_10" [label="(768,)", style=solid]; -"129 _scale_21" -> "132 dequantize_per_channel_default_21" [label="(2304,)", style=solid]; -"130 _zero_point_21" -> "132 dequantize_per_channel_default_21" [label="(2304,)", style=solid]; -"131 _frozen_param21" -> "132 dequantize_per_channel_default_21" [label="(2304, 768)", style=solid]; -"132 dequantize_per_channel_default_21" -> "546 linear_20" [label=None, style=solid]; -"133 encoder_layers_encoder_layer_5_self_attention_in_proj_bias" -> "546 linear_20" [label="(2304,)", style=solid]; -"134 _scale_22" -> "137 dequantize_per_channel_default_22" [label="(768,)", style=solid]; -"135 _zero_point_22" -> "137 dequantize_per_channel_default_22" [label="(768,)", style=solid]; -"136 _frozen_param22" -> "137 dequantize_per_channel_default_22" [label="(768, 768)", style=solid]; -"137 dequantize_per_channel_default_22" -> "572 linear_21" [label=None, style=solid]; -"138 encoder_layers_encoder_layer_5_self_attention_out_proj_bias" -> "572 linear_21" [label="(768,)", style=solid]; -"139 encoder_layers_encoder_layer_5_ln_2_weight" -> "577 layer_norm_11" [label="(768,)", style=solid]; -"140 encoder_layers_encoder_layer_5_ln_2_bias" -> "577 layer_norm_11" [label="(768,)", style=solid]; -"141 _scale_23" -> "144 dequantize_per_channel_default_23" [label="(3072,)", style=solid]; -"142 _zero_point_23" -> "144 dequantize_per_channel_default_23" [label="(3072,)", style=solid]; -"143 _frozen_param23" -> "144 dequantize_per_channel_default_23" [label="(3072, 768)", style=solid]; -"144 dequantize_per_channel_default_23" -> "580 linear_22" [label=None, style=solid]; -"145 encoder_layers_encoder_layer_5_mlp_0_bias" -> "580 linear_22" [label="(3072,)", style=solid]; -"146 _scale_24" -> "149 dequantize_per_channel_default_24" [label="(768,)", style=solid]; -"147 _zero_point_24" -> "149 dequantize_per_channel_default_24" [label="(768,)", style=solid]; -"148 _frozen_param24" -> "149 dequantize_per_channel_default_24" [label="(768, 3072)", style=solid]; -"149 dequantize_per_channel_default_24" -> "585 linear_23" [label=None, style=solid]; -"150 encoder_layers_encoder_layer_5_mlp_3_bias" -> "585 linear_23" [label="(768,)", style=solid]; -"151 encoder_layers_encoder_layer_6_ln_1_weight" -> "588 layer_norm_12" [label="(768,)", style=solid]; -"152 encoder_layers_encoder_layer_6_ln_1_bias" -> "588 layer_norm_12" [label="(768,)", style=solid]; -"153 _scale_25" -> "156 dequantize_per_channel_default_25" [label="(2304,)", style=solid]; -"154 _zero_point_25" -> "156 dequantize_per_channel_default_25" [label="(2304,)", style=solid]; -"155 _frozen_param25" -> "156 dequantize_per_channel_default_25" [label="(2304, 768)", style=solid]; -"156 dequantize_per_channel_default_25" -> "592 linear_24" [label=None, style=solid]; -"157 encoder_layers_encoder_layer_6_self_attention_in_proj_bias" -> "592 linear_24" [label="(2304,)", style=solid]; -"158 _scale_26" -> "161 dequantize_per_channel_default_26" [label="(768,)", style=solid]; -"159 _zero_point_26" -> "161 dequantize_per_channel_default_26" [label="(768,)", style=solid]; -"160 _frozen_param26" -> "161 dequantize_per_channel_default_26" [label="(768, 768)", style=solid]; -"161 dequantize_per_channel_default_26" -> "618 linear_25" [label=None, style=solid]; -"162 encoder_layers_encoder_layer_6_self_attention_out_proj_bias" -> "618 linear_25" [label="(768,)", style=solid]; -"163 encoder_layers_encoder_layer_6_ln_2_weight" -> "623 layer_norm_13" [label="(768,)", style=solid]; -"164 encoder_layers_encoder_layer_6_ln_2_bias" -> "623 layer_norm_13" [label="(768,)", style=solid]; -"165 _scale_27" -> "168 dequantize_per_channel_default_27" [label="(3072,)", style=solid]; -"166 _zero_point_27" -> "168 dequantize_per_channel_default_27" [label="(3072,)", style=solid]; -"167 _frozen_param27" -> "168 dequantize_per_channel_default_27" [label="(3072, 768)", style=solid]; -"168 dequantize_per_channel_default_27" -> "626 linear_26" [label=None, style=solid]; -"169 encoder_layers_encoder_layer_6_mlp_0_bias" -> "626 linear_26" [label="(3072,)", style=solid]; -"170 _scale_28" -> "173 dequantize_per_channel_default_28" [label="(768,)", style=solid]; -"171 _zero_point_28" -> "173 dequantize_per_channel_default_28" [label="(768,)", style=solid]; -"172 _frozen_param28" -> "173 dequantize_per_channel_default_28" [label="(768, 3072)", style=solid]; -"173 dequantize_per_channel_default_28" -> "631 linear_27" [label=None, style=solid]; -"174 encoder_layers_encoder_layer_6_mlp_3_bias" -> "631 linear_27" [label="(768,)", style=solid]; -"175 encoder_layers_encoder_layer_7_ln_1_weight" -> "634 layer_norm_14" [label="(768,)", style=solid]; -"176 encoder_layers_encoder_layer_7_ln_1_bias" -> "634 layer_norm_14" [label="(768,)", style=solid]; -"177 _scale_29" -> "180 dequantize_per_channel_default_29" [label="(2304,)", style=solid]; -"178 _zero_point_29" -> "180 dequantize_per_channel_default_29" [label="(2304,)", style=solid]; -"179 _frozen_param29" -> "180 dequantize_per_channel_default_29" [label="(2304, 768)", style=solid]; -"180 dequantize_per_channel_default_29" -> "638 linear_28" [label=None, style=solid]; -"181 encoder_layers_encoder_layer_7_self_attention_in_proj_bias" -> "638 linear_28" [label="(2304,)", style=solid]; -"182 _scale_30" -> "185 dequantize_per_channel_default_30" [label="(768,)", style=solid]; -"183 _zero_point_30" -> "185 dequantize_per_channel_default_30" [label="(768,)", style=solid]; -"184 _frozen_param30" -> "185 dequantize_per_channel_default_30" [label="(768, 768)", style=solid]; -"185 dequantize_per_channel_default_30" -> "664 linear_29" [label=None, style=solid]; -"186 encoder_layers_encoder_layer_7_self_attention_out_proj_bias" -> "664 linear_29" [label="(768,)", style=solid]; -"187 encoder_layers_encoder_layer_7_ln_2_weight" -> "669 layer_norm_15" [label="(768,)", style=solid]; -"188 encoder_layers_encoder_layer_7_ln_2_bias" -> "669 layer_norm_15" [label="(768,)", style=solid]; -"189 _scale_31" -> "192 dequantize_per_channel_default_31" [label="(3072,)", style=solid]; -"190 _zero_point_31" -> "192 dequantize_per_channel_default_31" [label="(3072,)", style=solid]; -"191 _frozen_param31" -> "192 dequantize_per_channel_default_31" [label="(3072, 768)", style=solid]; -"192 dequantize_per_channel_default_31" -> "672 linear_30" [label=None, style=solid]; -"193 encoder_layers_encoder_layer_7_mlp_0_bias" -> "672 linear_30" [label="(3072,)", style=solid]; -"194 _scale_32" -> "197 dequantize_per_channel_default_32" [label="(768,)", style=solid]; -"195 _zero_point_32" -> "197 dequantize_per_channel_default_32" [label="(768,)", style=solid]; -"196 _frozen_param32" -> "197 dequantize_per_channel_default_32" [label="(768, 3072)", style=solid]; -"197 dequantize_per_channel_default_32" -> "677 linear_31" [label=None, style=solid]; -"198 encoder_layers_encoder_layer_7_mlp_3_bias" -> "677 linear_31" [label="(768,)", style=solid]; -"199 encoder_layers_encoder_layer_8_ln_1_weight" -> "680 layer_norm_16" [label="(768,)", style=solid]; -"200 encoder_layers_encoder_layer_8_ln_1_bias" -> "680 layer_norm_16" [label="(768,)", style=solid]; -"201 _scale_33" -> "204 dequantize_per_channel_default_33" [label="(2304,)", style=solid]; -"202 _zero_point_33" -> "204 dequantize_per_channel_default_33" [label="(2304,)", style=solid]; -"203 _frozen_param33" -> "204 dequantize_per_channel_default_33" [label="(2304, 768)", style=solid]; -"204 dequantize_per_channel_default_33" -> "684 linear_32" [label=None, style=solid]; -"205 encoder_layers_encoder_layer_8_self_attention_in_proj_bias" -> "684 linear_32" [label="(2304,)", style=solid]; -"206 _scale_34" -> "209 dequantize_per_channel_default_34" [label="(768,)", style=solid]; -"207 _zero_point_34" -> "209 dequantize_per_channel_default_34" [label="(768,)", style=solid]; -"208 _frozen_param34" -> "209 dequantize_per_channel_default_34" [label="(768, 768)", style=solid]; -"209 dequantize_per_channel_default_34" -> "710 linear_33" [label=None, style=solid]; -"210 encoder_layers_encoder_layer_8_self_attention_out_proj_bias" -> "710 linear_33" [label="(768,)", style=solid]; -"211 encoder_layers_encoder_layer_8_ln_2_weight" -> "715 layer_norm_17" [label="(768,)", style=solid]; -"212 encoder_layers_encoder_layer_8_ln_2_bias" -> "715 layer_norm_17" [label="(768,)", style=solid]; -"213 _scale_35" -> "216 dequantize_per_channel_default_35" [label="(3072,)", style=solid]; -"214 _zero_point_35" -> "216 dequantize_per_channel_default_35" [label="(3072,)", style=solid]; -"215 _frozen_param35" -> "216 dequantize_per_channel_default_35" [label="(3072, 768)", style=solid]; -"216 dequantize_per_channel_default_35" -> "718 linear_34" [label=None, style=solid]; -"217 encoder_layers_encoder_layer_8_mlp_0_bias" -> "718 linear_34" [label="(3072,)", style=solid]; -"218 _scale_36" -> "221 dequantize_per_channel_default_36" [label="(768,)", style=solid]; -"219 _zero_point_36" -> "221 dequantize_per_channel_default_36" [label="(768,)", style=solid]; -"220 _frozen_param36" -> "221 dequantize_per_channel_default_36" [label="(768, 3072)", style=solid]; -"221 dequantize_per_channel_default_36" -> "723 linear_35" [label=None, style=solid]; -"222 encoder_layers_encoder_layer_8_mlp_3_bias" -> "723 linear_35" [label="(768,)", style=solid]; -"223 encoder_layers_encoder_layer_9_ln_1_weight" -> "726 layer_norm_18" [label="(768,)", style=solid]; -"224 encoder_layers_encoder_layer_9_ln_1_bias" -> "726 layer_norm_18" [label="(768,)", style=solid]; -"225 _scale_37" -> "228 dequantize_per_channel_default_37" [label="(2304,)", style=solid]; -"226 _zero_point_37" -> "228 dequantize_per_channel_default_37" [label="(2304,)", style=solid]; -"227 _frozen_param37" -> "228 dequantize_per_channel_default_37" [label="(2304, 768)", style=solid]; -"228 dequantize_per_channel_default_37" -> "730 linear_36" [label=None, style=solid]; -"229 encoder_layers_encoder_layer_9_self_attention_in_proj_bias" -> "730 linear_36" [label="(2304,)", style=solid]; -"230 _scale_38" -> "233 dequantize_per_channel_default_38" [label="(768,)", style=solid]; -"231 _zero_point_38" -> "233 dequantize_per_channel_default_38" [label="(768,)", style=solid]; -"232 _frozen_param38" -> "233 dequantize_per_channel_default_38" [label="(768, 768)", style=solid]; -"233 dequantize_per_channel_default_38" -> "756 linear_37" [label=None, style=solid]; -"234 encoder_layers_encoder_layer_9_self_attention_out_proj_bias" -> "756 linear_37" [label="(768,)", style=solid]; -"235 encoder_layers_encoder_layer_9_ln_2_weight" -> "761 layer_norm_19" [label="(768,)", style=solid]; -"236 encoder_layers_encoder_layer_9_ln_2_bias" -> "761 layer_norm_19" [label="(768,)", style=solid]; -"237 _scale_39" -> "240 dequantize_per_channel_default_39" [label="(3072,)", style=solid]; -"238 _zero_point_39" -> "240 dequantize_per_channel_default_39" [label="(3072,)", style=solid]; -"239 _frozen_param39" -> "240 dequantize_per_channel_default_39" [label="(3072, 768)", style=solid]; -"240 dequantize_per_channel_default_39" -> "764 linear_38" [label=None, style=solid]; -"241 encoder_layers_encoder_layer_9_mlp_0_bias" -> "764 linear_38" [label="(3072,)", style=solid]; -"242 _scale_40" -> "245 dequantize_per_channel_default_40" [label="(768,)", style=solid]; -"243 _zero_point_40" -> "245 dequantize_per_channel_default_40" [label="(768,)", style=solid]; -"244 _frozen_param40" -> "245 dequantize_per_channel_default_40" [label="(768, 3072)", style=solid]; -"245 dequantize_per_channel_default_40" -> "769 linear_39" [label=None, style=solid]; -"246 encoder_layers_encoder_layer_9_mlp_3_bias" -> "769 linear_39" [label="(768,)", style=solid]; -"247 encoder_layers_encoder_layer_10_ln_1_weight" -> "772 layer_norm_20" [label="(768,)", style=solid]; -"248 encoder_layers_encoder_layer_10_ln_1_bias" -> "772 layer_norm_20" [label="(768,)", style=solid]; -"249 _scale_41" -> "252 dequantize_per_channel_default_41" [label="(2304,)", style=solid]; -"250 _zero_point_41" -> "252 dequantize_per_channel_default_41" [label="(2304,)", style=solid]; -"251 _frozen_param41" -> "252 dequantize_per_channel_default_41" [label="(2304, 768)", style=solid]; -"252 dequantize_per_channel_default_41" -> "776 linear_40" [label=None, style=solid]; -"253 encoder_layers_encoder_layer_10_self_attention_in_proj_bias" -> "776 linear_40" [label="(2304,)", style=solid]; -"254 _scale_42" -> "257 dequantize_per_channel_default_42" [label="(768,)", style=solid]; -"255 _zero_point_42" -> "257 dequantize_per_channel_default_42" [label="(768,)", style=solid]; -"256 _frozen_param42" -> "257 dequantize_per_channel_default_42" [label="(768, 768)", style=solid]; -"257 dequantize_per_channel_default_42" -> "802 linear_41" [label=None, style=solid]; -"258 encoder_layers_encoder_layer_10_self_attention_out_proj_bias" -> "802 linear_41" [label="(768,)", style=solid]; -"259 encoder_layers_encoder_layer_10_ln_2_weight" -> "807 layer_norm_21" [label="(768,)", style=solid]; -"260 encoder_layers_encoder_layer_10_ln_2_bias" -> "807 layer_norm_21" [label="(768,)", style=solid]; -"261 _scale_43" -> "264 dequantize_per_channel_default_43" [label="(3072,)", style=solid]; -"262 _zero_point_43" -> "264 dequantize_per_channel_default_43" [label="(3072,)", style=solid]; -"263 _frozen_param43" -> "264 dequantize_per_channel_default_43" [label="(3072, 768)", style=solid]; -"264 dequantize_per_channel_default_43" -> "810 linear_42" [label=None, style=solid]; -"265 encoder_layers_encoder_layer_10_mlp_0_bias" -> "810 linear_42" [label="(3072,)", style=solid]; -"266 _scale_44" -> "269 dequantize_per_channel_default_44" [label="(768,)", style=solid]; -"267 _zero_point_44" -> "269 dequantize_per_channel_default_44" [label="(768,)", style=solid]; -"268 _frozen_param44" -> "269 dequantize_per_channel_default_44" [label="(768, 3072)", style=solid]; -"269 dequantize_per_channel_default_44" -> "815 linear_43" [label=None, style=solid]; -"270 encoder_layers_encoder_layer_10_mlp_3_bias" -> "815 linear_43" [label="(768,)", style=solid]; -"271 encoder_layers_encoder_layer_11_ln_1_weight" -> "818 layer_norm_22" [label="(768,)", style=solid]; -"272 encoder_layers_encoder_layer_11_ln_1_bias" -> "818 layer_norm_22" [label="(768,)", style=solid]; -"273 _scale_45" -> "276 dequantize_per_channel_default_45" [label="(2304,)", style=solid]; -"274 _zero_point_45" -> "276 dequantize_per_channel_default_45" [label="(2304,)", style=solid]; -"275 _frozen_param45" -> "276 dequantize_per_channel_default_45" [label="(2304, 768)", style=solid]; -"276 dequantize_per_channel_default_45" -> "822 linear_44" [label=None, style=solid]; -"277 encoder_layers_encoder_layer_11_self_attention_in_proj_bias" -> "822 linear_44" [label="(2304,)", style=solid]; -"278 _scale_46" -> "281 dequantize_per_channel_default_46" [label="(768,)", style=solid]; -"279 _zero_point_46" -> "281 dequantize_per_channel_default_46" [label="(768,)", style=solid]; -"280 _frozen_param46" -> "281 dequantize_per_channel_default_46" [label="(768, 768)", style=solid]; -"281 dequantize_per_channel_default_46" -> "848 linear_45" [label=None, style=solid]; -"282 encoder_layers_encoder_layer_11_self_attention_out_proj_bias" -> "848 linear_45" [label="(768,)", style=solid]; -"283 encoder_layers_encoder_layer_11_ln_2_weight" -> "853 layer_norm_23" [label="(768,)", style=solid]; -"284 encoder_layers_encoder_layer_11_ln_2_bias" -> "853 layer_norm_23" [label="(768,)", style=solid]; -"285 _scale_47" -> "288 dequantize_per_channel_default_47" [label="(3072,)", style=solid]; -"286 _zero_point_47" -> "288 dequantize_per_channel_default_47" [label="(3072,)", style=solid]; -"287 _frozen_param47" -> "288 dequantize_per_channel_default_47" [label="(3072, 768)", style=solid]; -"288 dequantize_per_channel_default_47" -> "856 linear_46" [label=None, style=solid]; -"289 encoder_layers_encoder_layer_11_mlp_0_bias" -> "856 linear_46" [label="(3072,)", style=solid]; -"290 _scale_48" -> "293 dequantize_per_channel_default_48" [label="(768,)", style=solid]; -"291 _zero_point_48" -> "293 dequantize_per_channel_default_48" [label="(768,)", style=solid]; -"292 _frozen_param48" -> "293 dequantize_per_channel_default_48" [label="(768, 3072)", style=solid]; -"293 dequantize_per_channel_default_48" -> "861 linear_47" [label=None, style=solid]; -"294 encoder_layers_encoder_layer_11_mlp_3_bias" -> "861 linear_47" [label="(768,)", style=solid]; -"295 encoder_ln_weight" -> "864 layer_norm_24" [label="(768,)", style=solid]; -"296 encoder_ln_bias" -> "864 layer_norm_24" [label="(768,)", style=solid]; -"297 _scale_49" -> "300 dequantize_per_channel_default_49" [label="(1000,)", style=solid]; -"298 _zero_point_49" -> "300 dequantize_per_channel_default_49" [label="(1000,)", style=solid]; -"299 _frozen_param49" -> "300 dequantize_per_channel_default_49" [label="(1000, 768)", style=solid]; -"300 dequantize_per_channel_default_49" -> "869 linear_48" [label=None, style=solid]; -"301 heads_head_bias" -> "869 linear_48" [label="(1000,)", style=solid]; -"302 x" -> "303 quantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; -"303 quantize_per_tensor_default" -> "304 dequantize_per_tensor_default" [label=None, style=solid]; -"304 dequantize_per_tensor_default" -> "305 conv2d" [label=None, style=solid]; -"305 conv2d" -> "306 reshape" [label="(1, 768, 14, 14)", style=solid]; -"306 reshape" -> "307 permute" [label="(1, 768, 196)", style=solid]; -"307 permute" -> "309 cat" [label="(1, 196, 768)", style=solid]; -"308 expand" -> "309 cat" [label="(1, 1, 768)", style=solid]; -"309 cat" -> "310 add" [label="(1, 197, 768)", style=solid]; -"310 add" -> "311 dropout" [label="(1, 197, 768)", style=solid]; -"311 dropout" -> "312 layer_norm" [label="(1, 197, 768)", style=solid]; -"311 dropout" -> "346 add_1" [label="(1, 197, 768)", style=solid]; -"312 layer_norm" -> "313 quantize_per_tensor_default_1" [label="(1, 197, 768)", style=solid]; -"313 quantize_per_tensor_default_1" -> "314 dequantize_per_tensor_default_1" [label=None, style=solid]; -"314 dequantize_per_tensor_default_1" -> "315 transpose" [label=None, style=solid]; -"315 transpose" -> "316 linear" [label="(197, 1, 768)", style=solid]; -"316 linear" -> "317 unflatten" [label="(197, 1, 2304)", style=solid]; -"317 unflatten" -> "318 unsqueeze" [label="(197, 1, 3, 768)", style=solid]; -"318 unsqueeze" -> "319 transpose_1" [label="(1, 197, 1, 3, 768)", style=solid]; -"319 transpose_1" -> "320 squeeze" [label="(3, 197, 1, 1, 768)", style=solid]; -"320 squeeze" -> "321 contiguous" [label="(3, 197, 1, 768)", style=solid]; -"321 contiguous" -> "322 quantize_per_tensor_default_2" [label="(3, 197, 1, 768)", style=solid]; -"321 contiguous" -> "327 select_2" [label="(3, 197, 1, 768)", style=solid]; -"322 quantize_per_tensor_default_2" -> "323 dequantize_per_tensor_default_63" [label=None, style=solid]; -"322 quantize_per_tensor_default_2" -> "324 dequantize_per_tensor_default_62" [label=None, style=solid]; -"323 dequantize_per_tensor_default_63" -> "326 select_1" [label=None, style=solid]; -"324 dequantize_per_tensor_default_62" -> "325 select" [label=None, style=solid]; -"325 select" -> "328 view" [label="(197, 1, 768)", style=solid]; -"326 select_1" -> "330 view_1" [label="(197, 1, 768)", style=solid]; -"327 select_2" -> "332 view_2" [label="(197, 1, 768)", style=solid]; -"328 view" -> "329 transpose_2" [label="(197, 12, 64)", style=solid]; -"329 transpose_2" -> "334 view_3" [label="(12, 197, 64)", style=solid]; -"330 view_1" -> "331 transpose_3" [label="(197, 12, 64)", style=solid]; -"331 transpose_3" -> "335 view_4" [label="(12, 197, 64)", style=solid]; -"332 view_2" -> "333 transpose_4" [label="(197, 12, 64)", style=solid]; -"333 transpose_4" -> "336 view_5" [label="(12, 197, 64)", style=solid]; -"334 view_3" -> "337 scaled_dot_product_attention" [label="(1, 12, 197, 64)", style=solid]; -"335 view_4" -> "337 scaled_dot_product_attention" [label="(1, 12, 197, 64)", style=solid]; -"336 view_5" -> "337 scaled_dot_product_attention" [label="(1, 12, 197, 64)", style=solid]; -"337 scaled_dot_product_attention" -> "338 quantize_per_tensor_default_3" [label="(1, 12, 197, 64)", style=solid]; -"338 quantize_per_tensor_default_3" -> "339 dequantize_per_tensor_default_3" [label=None, style=solid]; -"339 dequantize_per_tensor_default_3" -> "340 permute_1" [label=None, style=solid]; -"340 permute_1" -> "341 view_6" [label="(197, 1, 12, 64)", style=solid]; -"341 view_6" -> "342 linear_1" [label="(197, 768)", style=solid]; -"342 linear_1" -> "343 view_7" [label="(197, 768)", style=solid]; -"343 view_7" -> "344 transpose_5" [label="(197, 1, 768)", style=solid]; -"344 transpose_5" -> "345 dropout_1" [label="(1, 197, 768)", style=solid]; -"345 dropout_1" -> "346 add_1" [label="(1, 197, 768)", style=solid]; -"346 add_1" -> "347 layer_norm_1" [label="(1, 197, 768)", style=solid]; -"346 add_1" -> "357 add_2" [label="(1, 197, 768)", style=solid]; -"347 layer_norm_1" -> "348 quantize_per_tensor_default_4" [label="(1, 197, 768)", style=solid]; -"348 quantize_per_tensor_default_4" -> "349 dequantize_per_tensor_default_4" [label=None, style=solid]; -"349 dequantize_per_tensor_default_4" -> "350 linear_2" [label=None, style=solid]; -"350 linear_2" -> "351 gelu" [label="(1, 197, 3072)", style=solid]; -"351 gelu" -> "352 quantize_per_tensor_default_5" [label="(1, 197, 3072)", style=solid]; -"352 quantize_per_tensor_default_5" -> "353 dequantize_per_tensor_default_5" [label=None, style=solid]; -"353 dequantize_per_tensor_default_5" -> "354 dropout_2" [label=None, style=solid]; -"354 dropout_2" -> "355 linear_3" [label="(1, 197, 3072)", style=solid]; -"355 linear_3" -> "356 dropout_3" [label="(1, 197, 768)", style=solid]; -"356 dropout_3" -> "357 add_2" [label="(1, 197, 768)", style=solid]; -"357 add_2" -> "358 layer_norm_2" [label="(1, 197, 768)", style=solid]; -"357 add_2" -> "392 add_3" [label="(1, 197, 768)", style=solid]; -"358 layer_norm_2" -> "359 quantize_per_tensor_default_6" [label="(1, 197, 768)", style=solid]; -"359 quantize_per_tensor_default_6" -> "360 dequantize_per_tensor_default_6" [label=None, style=solid]; -"360 dequantize_per_tensor_default_6" -> "361 transpose_6" [label=None, style=solid]; -"361 transpose_6" -> "362 linear_4" [label="(197, 1, 768)", style=solid]; -"362 linear_4" -> "363 unflatten_1" [label="(197, 1, 2304)", style=solid]; -"363 unflatten_1" -> "364 unsqueeze_1" [label="(197, 1, 3, 768)", style=solid]; -"364 unsqueeze_1" -> "365 transpose_7" [label="(1, 197, 1, 3, 768)", style=solid]; -"365 transpose_7" -> "366 squeeze_1" [label="(3, 197, 1, 1, 768)", style=solid]; -"366 squeeze_1" -> "367 contiguous_1" [label="(3, 197, 1, 768)", style=solid]; -"367 contiguous_1" -> "368 quantize_per_tensor_default_7" [label="(3, 197, 1, 768)", style=solid]; -"367 contiguous_1" -> "373 select_5" [label="(3, 197, 1, 768)", style=solid]; -"368 quantize_per_tensor_default_7" -> "369 dequantize_per_tensor_default_65" [label=None, style=solid]; -"368 quantize_per_tensor_default_7" -> "370 dequantize_per_tensor_default_64" [label=None, style=solid]; -"369 dequantize_per_tensor_default_65" -> "372 select_4" [label=None, style=solid]; -"370 dequantize_per_tensor_default_64" -> "371 select_3" [label=None, style=solid]; -"371 select_3" -> "374 view_8" [label="(197, 1, 768)", style=solid]; -"372 select_4" -> "376 view_9" [label="(197, 1, 768)", style=solid]; -"373 select_5" -> "378 view_10" [label="(197, 1, 768)", style=solid]; -"374 view_8" -> "375 transpose_8" [label="(197, 12, 64)", style=solid]; -"375 transpose_8" -> "380 view_11" [label="(12, 197, 64)", style=solid]; -"376 view_9" -> "377 transpose_9" [label="(197, 12, 64)", style=solid]; -"377 transpose_9" -> "381 view_12" [label="(12, 197, 64)", style=solid]; -"378 view_10" -> "379 transpose_10" [label="(197, 12, 64)", style=solid]; -"379 transpose_10" -> "382 view_13" [label="(12, 197, 64)", style=solid]; -"380 view_11" -> "383 scaled_dot_product_attention_1" [label="(1, 12, 197, 64)", style=solid]; -"381 view_12" -> "383 scaled_dot_product_attention_1" [label="(1, 12, 197, 64)", style=solid]; -"382 view_13" -> "383 scaled_dot_product_attention_1" [label="(1, 12, 197, 64)", style=solid]; -"383 scaled_dot_product_attention_1" -> "384 quantize_per_tensor_default_8" [label="(1, 12, 197, 64)", style=solid]; -"384 quantize_per_tensor_default_8" -> "385 dequantize_per_tensor_default_8" [label=None, style=solid]; -"385 dequantize_per_tensor_default_8" -> "386 permute_2" [label=None, style=solid]; -"386 permute_2" -> "387 view_14" [label="(197, 1, 12, 64)", style=solid]; -"387 view_14" -> "388 linear_5" [label="(197, 768)", style=solid]; -"388 linear_5" -> "389 view_15" [label="(197, 768)", style=solid]; -"389 view_15" -> "390 transpose_11" [label="(197, 1, 768)", style=solid]; -"390 transpose_11" -> "391 dropout_4" [label="(1, 197, 768)", style=solid]; -"391 dropout_4" -> "392 add_3" [label="(1, 197, 768)", style=solid]; -"392 add_3" -> "393 layer_norm_3" [label="(1, 197, 768)", style=solid]; -"392 add_3" -> "403 add_4" [label="(1, 197, 768)", style=solid]; -"393 layer_norm_3" -> "394 quantize_per_tensor_default_9" [label="(1, 197, 768)", style=solid]; -"394 quantize_per_tensor_default_9" -> "395 dequantize_per_tensor_default_9" [label=None, style=solid]; -"395 dequantize_per_tensor_default_9" -> "396 linear_6" [label=None, style=solid]; -"396 linear_6" -> "397 gelu_1" [label="(1, 197, 3072)", style=solid]; -"397 gelu_1" -> "398 quantize_per_tensor_default_10" [label="(1, 197, 3072)", style=solid]; -"398 quantize_per_tensor_default_10" -> "399 dequantize_per_tensor_default_10" [label=None, style=solid]; -"399 dequantize_per_tensor_default_10" -> "400 dropout_5" [label=None, style=solid]; -"400 dropout_5" -> "401 linear_7" [label="(1, 197, 3072)", style=solid]; -"401 linear_7" -> "402 dropout_6" [label="(1, 197, 768)", style=solid]; -"402 dropout_6" -> "403 add_4" [label="(1, 197, 768)", style=solid]; -"403 add_4" -> "404 layer_norm_4" [label="(1, 197, 768)", style=solid]; -"403 add_4" -> "438 add_5" [label="(1, 197, 768)", style=solid]; -"404 layer_norm_4" -> "405 quantize_per_tensor_default_11" [label="(1, 197, 768)", style=solid]; -"405 quantize_per_tensor_default_11" -> "406 dequantize_per_tensor_default_11" [label=None, style=solid]; -"406 dequantize_per_tensor_default_11" -> "407 transpose_12" [label=None, style=solid]; -"407 transpose_12" -> "408 linear_8" [label="(197, 1, 768)", style=solid]; -"408 linear_8" -> "409 unflatten_2" [label="(197, 1, 2304)", style=solid]; -"409 unflatten_2" -> "410 unsqueeze_2" [label="(197, 1, 3, 768)", style=solid]; -"410 unsqueeze_2" -> "411 transpose_13" [label="(1, 197, 1, 3, 768)", style=solid]; -"411 transpose_13" -> "412 squeeze_2" [label="(3, 197, 1, 1, 768)", style=solid]; -"412 squeeze_2" -> "413 contiguous_2" [label="(3, 197, 1, 768)", style=solid]; -"413 contiguous_2" -> "414 quantize_per_tensor_default_12" [label="(3, 197, 1, 768)", style=solid]; -"413 contiguous_2" -> "419 select_8" [label="(3, 197, 1, 768)", style=solid]; -"414 quantize_per_tensor_default_12" -> "415 dequantize_per_tensor_default_67" [label=None, style=solid]; -"414 quantize_per_tensor_default_12" -> "416 dequantize_per_tensor_default_66" [label=None, style=solid]; -"415 dequantize_per_tensor_default_67" -> "418 select_7" [label=None, style=solid]; -"416 dequantize_per_tensor_default_66" -> "417 select_6" [label=None, style=solid]; -"417 select_6" -> "420 view_16" [label="(197, 1, 768)", style=solid]; -"418 select_7" -> "422 view_17" [label="(197, 1, 768)", style=solid]; -"419 select_8" -> "424 view_18" [label="(197, 1, 768)", style=solid]; -"420 view_16" -> "421 transpose_14" [label="(197, 12, 64)", style=solid]; -"421 transpose_14" -> "426 view_19" [label="(12, 197, 64)", style=solid]; -"422 view_17" -> "423 transpose_15" [label="(197, 12, 64)", style=solid]; -"423 transpose_15" -> "427 view_20" [label="(12, 197, 64)", style=solid]; -"424 view_18" -> "425 transpose_16" [label="(197, 12, 64)", style=solid]; -"425 transpose_16" -> "428 view_21" [label="(12, 197, 64)", style=solid]; -"426 view_19" -> "429 scaled_dot_product_attention_2" [label="(1, 12, 197, 64)", style=solid]; -"427 view_20" -> "429 scaled_dot_product_attention_2" [label="(1, 12, 197, 64)", style=solid]; -"428 view_21" -> "429 scaled_dot_product_attention_2" [label="(1, 12, 197, 64)", style=solid]; -"429 scaled_dot_product_attention_2" -> "430 quantize_per_tensor_default_13" [label="(1, 12, 197, 64)", style=solid]; -"430 quantize_per_tensor_default_13" -> "431 dequantize_per_tensor_default_13" [label=None, style=solid]; -"431 dequantize_per_tensor_default_13" -> "432 permute_3" [label=None, style=solid]; -"432 permute_3" -> "433 view_22" [label="(197, 1, 12, 64)", style=solid]; -"433 view_22" -> "434 linear_9" [label="(197, 768)", style=solid]; -"434 linear_9" -> "435 view_23" [label="(197, 768)", style=solid]; -"435 view_23" -> "436 transpose_17" [label="(197, 1, 768)", style=solid]; -"436 transpose_17" -> "437 dropout_7" [label="(1, 197, 768)", style=solid]; -"437 dropout_7" -> "438 add_5" [label="(1, 197, 768)", style=solid]; -"438 add_5" -> "439 layer_norm_5" [label="(1, 197, 768)", style=solid]; -"438 add_5" -> "449 add_6" [label="(1, 197, 768)", style=solid]; -"439 layer_norm_5" -> "440 quantize_per_tensor_default_14" [label="(1, 197, 768)", style=solid]; -"440 quantize_per_tensor_default_14" -> "441 dequantize_per_tensor_default_14" [label=None, style=solid]; -"441 dequantize_per_tensor_default_14" -> "442 linear_10" [label=None, style=solid]; -"442 linear_10" -> "443 gelu_2" [label="(1, 197, 3072)", style=solid]; -"443 gelu_2" -> "444 quantize_per_tensor_default_15" [label="(1, 197, 3072)", style=solid]; -"444 quantize_per_tensor_default_15" -> "445 dequantize_per_tensor_default_15" [label=None, style=solid]; -"445 dequantize_per_tensor_default_15" -> "446 dropout_8" [label=None, style=solid]; -"446 dropout_8" -> "447 linear_11" [label="(1, 197, 3072)", style=solid]; -"447 linear_11" -> "448 dropout_9" [label="(1, 197, 768)", style=solid]; -"448 dropout_9" -> "449 add_6" [label="(1, 197, 768)", style=solid]; -"449 add_6" -> "450 layer_norm_6" [label="(1, 197, 768)", style=solid]; -"449 add_6" -> "484 add_7" [label="(1, 197, 768)", style=solid]; -"450 layer_norm_6" -> "451 quantize_per_tensor_default_16" [label="(1, 197, 768)", style=solid]; -"451 quantize_per_tensor_default_16" -> "452 dequantize_per_tensor_default_16" [label=None, style=solid]; -"452 dequantize_per_tensor_default_16" -> "453 transpose_18" [label=None, style=solid]; -"453 transpose_18" -> "454 linear_12" [label="(197, 1, 768)", style=solid]; -"454 linear_12" -> "455 unflatten_3" [label="(197, 1, 2304)", style=solid]; -"455 unflatten_3" -> "456 unsqueeze_3" [label="(197, 1, 3, 768)", style=solid]; -"456 unsqueeze_3" -> "457 transpose_19" [label="(1, 197, 1, 3, 768)", style=solid]; -"457 transpose_19" -> "458 squeeze_3" [label="(3, 197, 1, 1, 768)", style=solid]; -"458 squeeze_3" -> "459 contiguous_3" [label="(3, 197, 1, 768)", style=solid]; -"459 contiguous_3" -> "460 quantize_per_tensor_default_17" [label="(3, 197, 1, 768)", style=solid]; -"459 contiguous_3" -> "465 select_11" [label="(3, 197, 1, 768)", style=solid]; -"460 quantize_per_tensor_default_17" -> "461 dequantize_per_tensor_default_69" [label=None, style=solid]; -"460 quantize_per_tensor_default_17" -> "462 dequantize_per_tensor_default_68" [label=None, style=solid]; -"461 dequantize_per_tensor_default_69" -> "464 select_10" [label=None, style=solid]; -"462 dequantize_per_tensor_default_68" -> "463 select_9" [label=None, style=solid]; -"463 select_9" -> "466 view_24" [label="(197, 1, 768)", style=solid]; -"464 select_10" -> "468 view_25" [label="(197, 1, 768)", style=solid]; -"465 select_11" -> "470 view_26" [label="(197, 1, 768)", style=solid]; -"466 view_24" -> "467 transpose_20" [label="(197, 12, 64)", style=solid]; -"467 transpose_20" -> "472 view_27" [label="(12, 197, 64)", style=solid]; -"468 view_25" -> "469 transpose_21" [label="(197, 12, 64)", style=solid]; -"469 transpose_21" -> "473 view_28" [label="(12, 197, 64)", style=solid]; -"470 view_26" -> "471 transpose_22" [label="(197, 12, 64)", style=solid]; -"471 transpose_22" -> "474 view_29" [label="(12, 197, 64)", style=solid]; -"472 view_27" -> "475 scaled_dot_product_attention_3" [label="(1, 12, 197, 64)", style=solid]; -"473 view_28" -> "475 scaled_dot_product_attention_3" [label="(1, 12, 197, 64)", style=solid]; -"474 view_29" -> "475 scaled_dot_product_attention_3" [label="(1, 12, 197, 64)", style=solid]; -"475 scaled_dot_product_attention_3" -> "476 quantize_per_tensor_default_18" [label="(1, 12, 197, 64)", style=solid]; -"476 quantize_per_tensor_default_18" -> "477 dequantize_per_tensor_default_18" [label=None, style=solid]; -"477 dequantize_per_tensor_default_18" -> "478 permute_4" [label=None, style=solid]; -"478 permute_4" -> "479 view_30" [label="(197, 1, 12, 64)", style=solid]; -"479 view_30" -> "480 linear_13" [label="(197, 768)", style=solid]; -"480 linear_13" -> "481 view_31" [label="(197, 768)", style=solid]; -"481 view_31" -> "482 transpose_23" [label="(197, 1, 768)", style=solid]; -"482 transpose_23" -> "483 dropout_10" [label="(1, 197, 768)", style=solid]; -"483 dropout_10" -> "484 add_7" [label="(1, 197, 768)", style=solid]; -"484 add_7" -> "485 layer_norm_7" [label="(1, 197, 768)", style=solid]; -"484 add_7" -> "495 add_8" [label="(1, 197, 768)", style=solid]; -"485 layer_norm_7" -> "486 quantize_per_tensor_default_19" [label="(1, 197, 768)", style=solid]; -"486 quantize_per_tensor_default_19" -> "487 dequantize_per_tensor_default_19" [label=None, style=solid]; -"487 dequantize_per_tensor_default_19" -> "488 linear_14" [label=None, style=solid]; -"488 linear_14" -> "489 gelu_3" [label="(1, 197, 3072)", style=solid]; -"489 gelu_3" -> "490 quantize_per_tensor_default_20" [label="(1, 197, 3072)", style=solid]; -"490 quantize_per_tensor_default_20" -> "491 dequantize_per_tensor_default_20" [label=None, style=solid]; -"491 dequantize_per_tensor_default_20" -> "492 dropout_11" [label=None, style=solid]; -"492 dropout_11" -> "493 linear_15" [label="(1, 197, 3072)", style=solid]; -"493 linear_15" -> "494 dropout_12" [label="(1, 197, 768)", style=solid]; -"494 dropout_12" -> "495 add_8" [label="(1, 197, 768)", style=solid]; -"495 add_8" -> "496 layer_norm_8" [label="(1, 197, 768)", style=solid]; -"495 add_8" -> "530 add_9" [label="(1, 197, 768)", style=solid]; -"496 layer_norm_8" -> "497 quantize_per_tensor_default_21" [label="(1, 197, 768)", style=solid]; -"497 quantize_per_tensor_default_21" -> "498 dequantize_per_tensor_default_21" [label=None, style=solid]; -"498 dequantize_per_tensor_default_21" -> "499 transpose_24" [label=None, style=solid]; -"499 transpose_24" -> "500 linear_16" [label="(197, 1, 768)", style=solid]; -"500 linear_16" -> "501 unflatten_4" [label="(197, 1, 2304)", style=solid]; -"501 unflatten_4" -> "502 unsqueeze_4" [label="(197, 1, 3, 768)", style=solid]; -"502 unsqueeze_4" -> "503 transpose_25" [label="(1, 197, 1, 3, 768)", style=solid]; -"503 transpose_25" -> "504 squeeze_4" [label="(3, 197, 1, 1, 768)", style=solid]; -"504 squeeze_4" -> "505 contiguous_4" [label="(3, 197, 1, 768)", style=solid]; -"505 contiguous_4" -> "506 quantize_per_tensor_default_22" [label="(3, 197, 1, 768)", style=solid]; -"505 contiguous_4" -> "511 select_14" [label="(3, 197, 1, 768)", style=solid]; -"506 quantize_per_tensor_default_22" -> "507 dequantize_per_tensor_default_71" [label=None, style=solid]; -"506 quantize_per_tensor_default_22" -> "508 dequantize_per_tensor_default_70" [label=None, style=solid]; -"507 dequantize_per_tensor_default_71" -> "510 select_13" [label=None, style=solid]; -"508 dequantize_per_tensor_default_70" -> "509 select_12" [label=None, style=solid]; -"509 select_12" -> "512 view_32" [label="(197, 1, 768)", style=solid]; -"510 select_13" -> "514 view_33" [label="(197, 1, 768)", style=solid]; -"511 select_14" -> "516 view_34" [label="(197, 1, 768)", style=solid]; -"512 view_32" -> "513 transpose_26" [label="(197, 12, 64)", style=solid]; -"513 transpose_26" -> "518 view_35" [label="(12, 197, 64)", style=solid]; -"514 view_33" -> "515 transpose_27" [label="(197, 12, 64)", style=solid]; -"515 transpose_27" -> "519 view_36" [label="(12, 197, 64)", style=solid]; -"516 view_34" -> "517 transpose_28" [label="(197, 12, 64)", style=solid]; -"517 transpose_28" -> "520 view_37" [label="(12, 197, 64)", style=solid]; -"518 view_35" -> "521 scaled_dot_product_attention_4" [label="(1, 12, 197, 64)", style=solid]; -"519 view_36" -> "521 scaled_dot_product_attention_4" [label="(1, 12, 197, 64)", style=solid]; -"520 view_37" -> "521 scaled_dot_product_attention_4" [label="(1, 12, 197, 64)", style=solid]; -"521 scaled_dot_product_attention_4" -> "522 quantize_per_tensor_default_23" [label="(1, 12, 197, 64)", style=solid]; -"522 quantize_per_tensor_default_23" -> "523 dequantize_per_tensor_default_23" [label=None, style=solid]; -"523 dequantize_per_tensor_default_23" -> "524 permute_5" [label=None, style=solid]; -"524 permute_5" -> "525 view_38" [label="(197, 1, 12, 64)", style=solid]; -"525 view_38" -> "526 linear_17" [label="(197, 768)", style=solid]; -"526 linear_17" -> "527 view_39" [label="(197, 768)", style=solid]; -"527 view_39" -> "528 transpose_29" [label="(197, 1, 768)", style=solid]; -"528 transpose_29" -> "529 dropout_13" [label="(1, 197, 768)", style=solid]; -"529 dropout_13" -> "530 add_9" [label="(1, 197, 768)", style=solid]; -"530 add_9" -> "531 layer_norm_9" [label="(1, 197, 768)", style=solid]; -"530 add_9" -> "541 add_10" [label="(1, 197, 768)", style=solid]; -"531 layer_norm_9" -> "532 quantize_per_tensor_default_24" [label="(1, 197, 768)", style=solid]; -"532 quantize_per_tensor_default_24" -> "533 dequantize_per_tensor_default_24" [label=None, style=solid]; -"533 dequantize_per_tensor_default_24" -> "534 linear_18" [label=None, style=solid]; -"534 linear_18" -> "535 gelu_4" [label="(1, 197, 3072)", style=solid]; -"535 gelu_4" -> "536 quantize_per_tensor_default_25" [label="(1, 197, 3072)", style=solid]; -"536 quantize_per_tensor_default_25" -> "537 dequantize_per_tensor_default_25" [label=None, style=solid]; -"537 dequantize_per_tensor_default_25" -> "538 dropout_14" [label=None, style=solid]; -"538 dropout_14" -> "539 linear_19" [label="(1, 197, 3072)", style=solid]; -"539 linear_19" -> "540 dropout_15" [label="(1, 197, 768)", style=solid]; -"540 dropout_15" -> "541 add_10" [label="(1, 197, 768)", style=solid]; -"541 add_10" -> "542 layer_norm_10" [label="(1, 197, 768)", style=solid]; -"541 add_10" -> "576 add_11" [label="(1, 197, 768)", style=solid]; -"542 layer_norm_10" -> "543 quantize_per_tensor_default_26" [label="(1, 197, 768)", style=solid]; -"543 quantize_per_tensor_default_26" -> "544 dequantize_per_tensor_default_26" [label=None, style=solid]; -"544 dequantize_per_tensor_default_26" -> "545 transpose_30" [label=None, style=solid]; -"545 transpose_30" -> "546 linear_20" [label="(197, 1, 768)", style=solid]; -"546 linear_20" -> "547 unflatten_5" [label="(197, 1, 2304)", style=solid]; -"547 unflatten_5" -> "548 unsqueeze_5" [label="(197, 1, 3, 768)", style=solid]; -"548 unsqueeze_5" -> "549 transpose_31" [label="(1, 197, 1, 3, 768)", style=solid]; -"549 transpose_31" -> "550 squeeze_5" [label="(3, 197, 1, 1, 768)", style=solid]; -"550 squeeze_5" -> "551 contiguous_5" [label="(3, 197, 1, 768)", style=solid]; -"551 contiguous_5" -> "552 quantize_per_tensor_default_27" [label="(3, 197, 1, 768)", style=solid]; -"551 contiguous_5" -> "557 select_17" [label="(3, 197, 1, 768)", style=solid]; -"552 quantize_per_tensor_default_27" -> "553 dequantize_per_tensor_default_73" [label=None, style=solid]; -"552 quantize_per_tensor_default_27" -> "554 dequantize_per_tensor_default_72" [label=None, style=solid]; -"553 dequantize_per_tensor_default_73" -> "556 select_16" [label=None, style=solid]; -"554 dequantize_per_tensor_default_72" -> "555 select_15" [label=None, style=solid]; -"555 select_15" -> "558 view_40" [label="(197, 1, 768)", style=solid]; -"556 select_16" -> "560 view_41" [label="(197, 1, 768)", style=solid]; -"557 select_17" -> "562 view_42" [label="(197, 1, 768)", style=solid]; -"558 view_40" -> "559 transpose_32" [label="(197, 12, 64)", style=solid]; -"559 transpose_32" -> "564 view_43" [label="(12, 197, 64)", style=solid]; -"560 view_41" -> "561 transpose_33" [label="(197, 12, 64)", style=solid]; -"561 transpose_33" -> "565 view_44" [label="(12, 197, 64)", style=solid]; -"562 view_42" -> "563 transpose_34" [label="(197, 12, 64)", style=solid]; -"563 transpose_34" -> "566 view_45" [label="(12, 197, 64)", style=solid]; -"564 view_43" -> "567 scaled_dot_product_attention_5" [label="(1, 12, 197, 64)", style=solid]; -"565 view_44" -> "567 scaled_dot_product_attention_5" [label="(1, 12, 197, 64)", style=solid]; -"566 view_45" -> "567 scaled_dot_product_attention_5" [label="(1, 12, 197, 64)", style=solid]; -"567 scaled_dot_product_attention_5" -> "568 quantize_per_tensor_default_28" [label="(1, 12, 197, 64)", style=solid]; -"568 quantize_per_tensor_default_28" -> "569 dequantize_per_tensor_default_28" [label=None, style=solid]; -"569 dequantize_per_tensor_default_28" -> "570 permute_6" [label=None, style=solid]; -"570 permute_6" -> "571 view_46" [label="(197, 1, 12, 64)", style=solid]; -"571 view_46" -> "572 linear_21" [label="(197, 768)", style=solid]; -"572 linear_21" -> "573 view_47" [label="(197, 768)", style=solid]; -"573 view_47" -> "574 transpose_35" [label="(197, 1, 768)", style=solid]; -"574 transpose_35" -> "575 dropout_16" [label="(1, 197, 768)", style=solid]; -"575 dropout_16" -> "576 add_11" [label="(1, 197, 768)", style=solid]; -"576 add_11" -> "577 layer_norm_11" [label="(1, 197, 768)", style=solid]; -"576 add_11" -> "587 add_12" [label="(1, 197, 768)", style=solid]; -"577 layer_norm_11" -> "578 quantize_per_tensor_default_29" [label="(1, 197, 768)", style=solid]; -"578 quantize_per_tensor_default_29" -> "579 dequantize_per_tensor_default_29" [label=None, style=solid]; -"579 dequantize_per_tensor_default_29" -> "580 linear_22" [label=None, style=solid]; -"580 linear_22" -> "581 gelu_5" [label="(1, 197, 3072)", style=solid]; -"581 gelu_5" -> "582 quantize_per_tensor_default_30" [label="(1, 197, 3072)", style=solid]; -"582 quantize_per_tensor_default_30" -> "583 dequantize_per_tensor_default_30" [label=None, style=solid]; -"583 dequantize_per_tensor_default_30" -> "584 dropout_17" [label=None, style=solid]; -"584 dropout_17" -> "585 linear_23" [label="(1, 197, 3072)", style=solid]; -"585 linear_23" -> "586 dropout_18" [label="(1, 197, 768)", style=solid]; -"586 dropout_18" -> "587 add_12" [label="(1, 197, 768)", style=solid]; -"587 add_12" -> "588 layer_norm_12" [label="(1, 197, 768)", style=solid]; -"587 add_12" -> "622 add_13" [label="(1, 197, 768)", style=solid]; -"588 layer_norm_12" -> "589 quantize_per_tensor_default_31" [label="(1, 197, 768)", style=solid]; -"589 quantize_per_tensor_default_31" -> "590 dequantize_per_tensor_default_31" [label=None, style=solid]; -"590 dequantize_per_tensor_default_31" -> "591 transpose_36" [label=None, style=solid]; -"591 transpose_36" -> "592 linear_24" [label="(197, 1, 768)", style=solid]; -"592 linear_24" -> "593 unflatten_6" [label="(197, 1, 2304)", style=solid]; -"593 unflatten_6" -> "594 unsqueeze_6" [label="(197, 1, 3, 768)", style=solid]; -"594 unsqueeze_6" -> "595 transpose_37" [label="(1, 197, 1, 3, 768)", style=solid]; -"595 transpose_37" -> "596 squeeze_6" [label="(3, 197, 1, 1, 768)", style=solid]; -"596 squeeze_6" -> "597 contiguous_6" [label="(3, 197, 1, 768)", style=solid]; -"597 contiguous_6" -> "598 quantize_per_tensor_default_32" [label="(3, 197, 1, 768)", style=solid]; -"597 contiguous_6" -> "603 select_20" [label="(3, 197, 1, 768)", style=solid]; -"598 quantize_per_tensor_default_32" -> "599 dequantize_per_tensor_default_75" [label=None, style=solid]; -"598 quantize_per_tensor_default_32" -> "600 dequantize_per_tensor_default_74" [label=None, style=solid]; -"599 dequantize_per_tensor_default_75" -> "602 select_19" [label=None, style=solid]; -"600 dequantize_per_tensor_default_74" -> "601 select_18" [label=None, style=solid]; -"601 select_18" -> "604 view_48" [label="(197, 1, 768)", style=solid]; -"602 select_19" -> "606 view_49" [label="(197, 1, 768)", style=solid]; -"603 select_20" -> "608 view_50" [label="(197, 1, 768)", style=solid]; -"604 view_48" -> "605 transpose_38" [label="(197, 12, 64)", style=solid]; -"605 transpose_38" -> "610 view_51" [label="(12, 197, 64)", style=solid]; -"606 view_49" -> "607 transpose_39" [label="(197, 12, 64)", style=solid]; -"607 transpose_39" -> "611 view_52" [label="(12, 197, 64)", style=solid]; -"608 view_50" -> "609 transpose_40" [label="(197, 12, 64)", style=solid]; -"609 transpose_40" -> "612 view_53" [label="(12, 197, 64)", style=solid]; -"610 view_51" -> "613 scaled_dot_product_attention_6" [label="(1, 12, 197, 64)", style=solid]; -"611 view_52" -> "613 scaled_dot_product_attention_6" [label="(1, 12, 197, 64)", style=solid]; -"612 view_53" -> "613 scaled_dot_product_attention_6" [label="(1, 12, 197, 64)", style=solid]; -"613 scaled_dot_product_attention_6" -> "614 quantize_per_tensor_default_33" [label="(1, 12, 197, 64)", style=solid]; -"614 quantize_per_tensor_default_33" -> "615 dequantize_per_tensor_default_33" [label=None, style=solid]; -"615 dequantize_per_tensor_default_33" -> "616 permute_7" [label=None, style=solid]; -"616 permute_7" -> "617 view_54" [label="(197, 1, 12, 64)", style=solid]; -"617 view_54" -> "618 linear_25" [label="(197, 768)", style=solid]; -"618 linear_25" -> "619 view_55" [label="(197, 768)", style=solid]; -"619 view_55" -> "620 transpose_41" [label="(197, 1, 768)", style=solid]; -"620 transpose_41" -> "621 dropout_19" [label="(1, 197, 768)", style=solid]; -"621 dropout_19" -> "622 add_13" [label="(1, 197, 768)", style=solid]; -"622 add_13" -> "623 layer_norm_13" [label="(1, 197, 768)", style=solid]; -"622 add_13" -> "633 add_14" [label="(1, 197, 768)", style=solid]; -"623 layer_norm_13" -> "624 quantize_per_tensor_default_34" [label="(1, 197, 768)", style=solid]; -"624 quantize_per_tensor_default_34" -> "625 dequantize_per_tensor_default_34" [label=None, style=solid]; -"625 dequantize_per_tensor_default_34" -> "626 linear_26" [label=None, style=solid]; -"626 linear_26" -> "627 gelu_6" [label="(1, 197, 3072)", style=solid]; -"627 gelu_6" -> "628 quantize_per_tensor_default_35" [label="(1, 197, 3072)", style=solid]; -"628 quantize_per_tensor_default_35" -> "629 dequantize_per_tensor_default_35" [label=None, style=solid]; -"629 dequantize_per_tensor_default_35" -> "630 dropout_20" [label=None, style=solid]; -"630 dropout_20" -> "631 linear_27" [label="(1, 197, 3072)", style=solid]; -"631 linear_27" -> "632 dropout_21" [label="(1, 197, 768)", style=solid]; -"632 dropout_21" -> "633 add_14" [label="(1, 197, 768)", style=solid]; -"633 add_14" -> "634 layer_norm_14" [label="(1, 197, 768)", style=solid]; -"633 add_14" -> "668 add_15" [label="(1, 197, 768)", style=solid]; -"634 layer_norm_14" -> "635 quantize_per_tensor_default_36" [label="(1, 197, 768)", style=solid]; -"635 quantize_per_tensor_default_36" -> "636 dequantize_per_tensor_default_36" [label=None, style=solid]; -"636 dequantize_per_tensor_default_36" -> "637 transpose_42" [label=None, style=solid]; -"637 transpose_42" -> "638 linear_28" [label="(197, 1, 768)", style=solid]; -"638 linear_28" -> "639 unflatten_7" [label="(197, 1, 2304)", style=solid]; -"639 unflatten_7" -> "640 unsqueeze_7" [label="(197, 1, 3, 768)", style=solid]; -"640 unsqueeze_7" -> "641 transpose_43" [label="(1, 197, 1, 3, 768)", style=solid]; -"641 transpose_43" -> "642 squeeze_7" [label="(3, 197, 1, 1, 768)", style=solid]; -"642 squeeze_7" -> "643 contiguous_7" [label="(3, 197, 1, 768)", style=solid]; -"643 contiguous_7" -> "644 quantize_per_tensor_default_37" [label="(3, 197, 1, 768)", style=solid]; -"643 contiguous_7" -> "649 select_23" [label="(3, 197, 1, 768)", style=solid]; -"644 quantize_per_tensor_default_37" -> "645 dequantize_per_tensor_default_77" [label=None, style=solid]; -"644 quantize_per_tensor_default_37" -> "646 dequantize_per_tensor_default_76" [label=None, style=solid]; -"645 dequantize_per_tensor_default_77" -> "648 select_22" [label=None, style=solid]; -"646 dequantize_per_tensor_default_76" -> "647 select_21" [label=None, style=solid]; -"647 select_21" -> "650 view_56" [label="(197, 1, 768)", style=solid]; -"648 select_22" -> "652 view_57" [label="(197, 1, 768)", style=solid]; -"649 select_23" -> "654 view_58" [label="(197, 1, 768)", style=solid]; -"650 view_56" -> "651 transpose_44" [label="(197, 12, 64)", style=solid]; -"651 transpose_44" -> "656 view_59" [label="(12, 197, 64)", style=solid]; -"652 view_57" -> "653 transpose_45" [label="(197, 12, 64)", style=solid]; -"653 transpose_45" -> "657 view_60" [label="(12, 197, 64)", style=solid]; -"654 view_58" -> "655 transpose_46" [label="(197, 12, 64)", style=solid]; -"655 transpose_46" -> "658 view_61" [label="(12, 197, 64)", style=solid]; -"656 view_59" -> "659 scaled_dot_product_attention_7" [label="(1, 12, 197, 64)", style=solid]; -"657 view_60" -> "659 scaled_dot_product_attention_7" [label="(1, 12, 197, 64)", style=solid]; -"658 view_61" -> "659 scaled_dot_product_attention_7" [label="(1, 12, 197, 64)", style=solid]; -"659 scaled_dot_product_attention_7" -> "660 quantize_per_tensor_default_38" [label="(1, 12, 197, 64)", style=solid]; -"660 quantize_per_tensor_default_38" -> "661 dequantize_per_tensor_default_38" [label=None, style=solid]; -"661 dequantize_per_tensor_default_38" -> "662 permute_8" [label=None, style=solid]; -"662 permute_8" -> "663 view_62" [label="(197, 1, 12, 64)", style=solid]; -"663 view_62" -> "664 linear_29" [label="(197, 768)", style=solid]; -"664 linear_29" -> "665 view_63" [label="(197, 768)", style=solid]; -"665 view_63" -> "666 transpose_47" [label="(197, 1, 768)", style=solid]; -"666 transpose_47" -> "667 dropout_22" [label="(1, 197, 768)", style=solid]; -"667 dropout_22" -> "668 add_15" [label="(1, 197, 768)", style=solid]; -"668 add_15" -> "669 layer_norm_15" [label="(1, 197, 768)", style=solid]; -"668 add_15" -> "679 add_16" [label="(1, 197, 768)", style=solid]; -"669 layer_norm_15" -> "670 quantize_per_tensor_default_39" [label="(1, 197, 768)", style=solid]; -"670 quantize_per_tensor_default_39" -> "671 dequantize_per_tensor_default_39" [label=None, style=solid]; -"671 dequantize_per_tensor_default_39" -> "672 linear_30" [label=None, style=solid]; -"672 linear_30" -> "673 gelu_7" [label="(1, 197, 3072)", style=solid]; -"673 gelu_7" -> "674 quantize_per_tensor_default_40" [label="(1, 197, 3072)", style=solid]; -"674 quantize_per_tensor_default_40" -> "675 dequantize_per_tensor_default_40" [label=None, style=solid]; -"675 dequantize_per_tensor_default_40" -> "676 dropout_23" [label=None, style=solid]; -"676 dropout_23" -> "677 linear_31" [label="(1, 197, 3072)", style=solid]; -"677 linear_31" -> "678 dropout_24" [label="(1, 197, 768)", style=solid]; -"678 dropout_24" -> "679 add_16" [label="(1, 197, 768)", style=solid]; -"679 add_16" -> "680 layer_norm_16" [label="(1, 197, 768)", style=solid]; -"679 add_16" -> "714 add_17" [label="(1, 197, 768)", style=solid]; -"680 layer_norm_16" -> "681 quantize_per_tensor_default_41" [label="(1, 197, 768)", style=solid]; -"681 quantize_per_tensor_default_41" -> "682 dequantize_per_tensor_default_41" [label=None, style=solid]; -"682 dequantize_per_tensor_default_41" -> "683 transpose_48" [label=None, style=solid]; -"683 transpose_48" -> "684 linear_32" [label="(197, 1, 768)", style=solid]; -"684 linear_32" -> "685 unflatten_8" [label="(197, 1, 2304)", style=solid]; -"685 unflatten_8" -> "686 unsqueeze_8" [label="(197, 1, 3, 768)", style=solid]; -"686 unsqueeze_8" -> "687 transpose_49" [label="(1, 197, 1, 3, 768)", style=solid]; -"687 transpose_49" -> "688 squeeze_8" [label="(3, 197, 1, 1, 768)", style=solid]; -"688 squeeze_8" -> "689 contiguous_8" [label="(3, 197, 1, 768)", style=solid]; -"689 contiguous_8" -> "690 quantize_per_tensor_default_42" [label="(3, 197, 1, 768)", style=solid]; -"689 contiguous_8" -> "695 select_26" [label="(3, 197, 1, 768)", style=solid]; -"690 quantize_per_tensor_default_42" -> "691 dequantize_per_tensor_default_79" [label=None, style=solid]; -"690 quantize_per_tensor_default_42" -> "692 dequantize_per_tensor_default_78" [label=None, style=solid]; -"691 dequantize_per_tensor_default_79" -> "694 select_25" [label=None, style=solid]; -"692 dequantize_per_tensor_default_78" -> "693 select_24" [label=None, style=solid]; -"693 select_24" -> "696 view_64" [label="(197, 1, 768)", style=solid]; -"694 select_25" -> "698 view_65" [label="(197, 1, 768)", style=solid]; -"695 select_26" -> "700 view_66" [label="(197, 1, 768)", style=solid]; -"696 view_64" -> "697 transpose_50" [label="(197, 12, 64)", style=solid]; -"697 transpose_50" -> "702 view_67" [label="(12, 197, 64)", style=solid]; -"698 view_65" -> "699 transpose_51" [label="(197, 12, 64)", style=solid]; -"699 transpose_51" -> "703 view_68" [label="(12, 197, 64)", style=solid]; -"700 view_66" -> "701 transpose_52" [label="(197, 12, 64)", style=solid]; -"701 transpose_52" -> "704 view_69" [label="(12, 197, 64)", style=solid]; -"702 view_67" -> "705 scaled_dot_product_attention_8" [label="(1, 12, 197, 64)", style=solid]; -"703 view_68" -> "705 scaled_dot_product_attention_8" [label="(1, 12, 197, 64)", style=solid]; -"704 view_69" -> "705 scaled_dot_product_attention_8" [label="(1, 12, 197, 64)", style=solid]; -"705 scaled_dot_product_attention_8" -> "706 quantize_per_tensor_default_43" [label="(1, 12, 197, 64)", style=solid]; -"706 quantize_per_tensor_default_43" -> "707 dequantize_per_tensor_default_43" [label=None, style=solid]; -"707 dequantize_per_tensor_default_43" -> "708 permute_9" [label=None, style=solid]; -"708 permute_9" -> "709 view_70" [label="(197, 1, 12, 64)", style=solid]; -"709 view_70" -> "710 linear_33" [label="(197, 768)", style=solid]; -"710 linear_33" -> "711 view_71" [label="(197, 768)", style=solid]; -"711 view_71" -> "712 transpose_53" [label="(197, 1, 768)", style=solid]; -"712 transpose_53" -> "713 dropout_25" [label="(1, 197, 768)", style=solid]; -"713 dropout_25" -> "714 add_17" [label="(1, 197, 768)", style=solid]; -"714 add_17" -> "715 layer_norm_17" [label="(1, 197, 768)", style=solid]; -"714 add_17" -> "725 add_18" [label="(1, 197, 768)", style=solid]; -"715 layer_norm_17" -> "716 quantize_per_tensor_default_44" [label="(1, 197, 768)", style=solid]; -"716 quantize_per_tensor_default_44" -> "717 dequantize_per_tensor_default_44" [label=None, style=solid]; -"717 dequantize_per_tensor_default_44" -> "718 linear_34" [label=None, style=solid]; -"718 linear_34" -> "719 gelu_8" [label="(1, 197, 3072)", style=solid]; -"719 gelu_8" -> "720 quantize_per_tensor_default_45" [label="(1, 197, 3072)", style=solid]; -"720 quantize_per_tensor_default_45" -> "721 dequantize_per_tensor_default_45" [label=None, style=solid]; -"721 dequantize_per_tensor_default_45" -> "722 dropout_26" [label=None, style=solid]; -"722 dropout_26" -> "723 linear_35" [label="(1, 197, 3072)", style=solid]; -"723 linear_35" -> "724 dropout_27" [label="(1, 197, 768)", style=solid]; -"724 dropout_27" -> "725 add_18" [label="(1, 197, 768)", style=solid]; -"725 add_18" -> "726 layer_norm_18" [label="(1, 197, 768)", style=solid]; -"725 add_18" -> "760 add_19" [label="(1, 197, 768)", style=solid]; -"726 layer_norm_18" -> "727 quantize_per_tensor_default_46" [label="(1, 197, 768)", style=solid]; -"727 quantize_per_tensor_default_46" -> "728 dequantize_per_tensor_default_46" [label=None, style=solid]; -"728 dequantize_per_tensor_default_46" -> "729 transpose_54" [label=None, style=solid]; -"729 transpose_54" -> "730 linear_36" [label="(197, 1, 768)", style=solid]; -"730 linear_36" -> "731 unflatten_9" [label="(197, 1, 2304)", style=solid]; -"731 unflatten_9" -> "732 unsqueeze_9" [label="(197, 1, 3, 768)", style=solid]; -"732 unsqueeze_9" -> "733 transpose_55" [label="(1, 197, 1, 3, 768)", style=solid]; -"733 transpose_55" -> "734 squeeze_9" [label="(3, 197, 1, 1, 768)", style=solid]; -"734 squeeze_9" -> "735 contiguous_9" [label="(3, 197, 1, 768)", style=solid]; -"735 contiguous_9" -> "736 quantize_per_tensor_default_47" [label="(3, 197, 1, 768)", style=solid]; -"735 contiguous_9" -> "741 select_29" [label="(3, 197, 1, 768)", style=solid]; -"736 quantize_per_tensor_default_47" -> "737 dequantize_per_tensor_default_81" [label=None, style=solid]; -"736 quantize_per_tensor_default_47" -> "738 dequantize_per_tensor_default_80" [label=None, style=solid]; -"737 dequantize_per_tensor_default_81" -> "740 select_28" [label=None, style=solid]; -"738 dequantize_per_tensor_default_80" -> "739 select_27" [label=None, style=solid]; -"739 select_27" -> "742 view_72" [label="(197, 1, 768)", style=solid]; -"740 select_28" -> "744 view_73" [label="(197, 1, 768)", style=solid]; -"741 select_29" -> "746 view_74" [label="(197, 1, 768)", style=solid]; -"742 view_72" -> "743 transpose_56" [label="(197, 12, 64)", style=solid]; -"743 transpose_56" -> "748 view_75" [label="(12, 197, 64)", style=solid]; -"744 view_73" -> "745 transpose_57" [label="(197, 12, 64)", style=solid]; -"745 transpose_57" -> "749 view_76" [label="(12, 197, 64)", style=solid]; -"746 view_74" -> "747 transpose_58" [label="(197, 12, 64)", style=solid]; -"747 transpose_58" -> "750 view_77" [label="(12, 197, 64)", style=solid]; -"748 view_75" -> "751 scaled_dot_product_attention_9" [label="(1, 12, 197, 64)", style=solid]; -"749 view_76" -> "751 scaled_dot_product_attention_9" [label="(1, 12, 197, 64)", style=solid]; -"750 view_77" -> "751 scaled_dot_product_attention_9" [label="(1, 12, 197, 64)", style=solid]; -"751 scaled_dot_product_attention_9" -> "752 quantize_per_tensor_default_48" [label="(1, 12, 197, 64)", style=solid]; -"752 quantize_per_tensor_default_48" -> "753 dequantize_per_tensor_default_48" [label=None, style=solid]; -"753 dequantize_per_tensor_default_48" -> "754 permute_10" [label=None, style=solid]; -"754 permute_10" -> "755 view_78" [label="(197, 1, 12, 64)", style=solid]; -"755 view_78" -> "756 linear_37" [label="(197, 768)", style=solid]; -"756 linear_37" -> "757 view_79" [label="(197, 768)", style=solid]; -"757 view_79" -> "758 transpose_59" [label="(197, 1, 768)", style=solid]; -"758 transpose_59" -> "759 dropout_28" [label="(1, 197, 768)", style=solid]; -"759 dropout_28" -> "760 add_19" [label="(1, 197, 768)", style=solid]; -"760 add_19" -> "761 layer_norm_19" [label="(1, 197, 768)", style=solid]; -"760 add_19" -> "771 add_20" [label="(1, 197, 768)", style=solid]; -"761 layer_norm_19" -> "762 quantize_per_tensor_default_49" [label="(1, 197, 768)", style=solid]; -"762 quantize_per_tensor_default_49" -> "763 dequantize_per_tensor_default_49" [label=None, style=solid]; -"763 dequantize_per_tensor_default_49" -> "764 linear_38" [label=None, style=solid]; -"764 linear_38" -> "765 gelu_9" [label="(1, 197, 3072)", style=solid]; -"765 gelu_9" -> "766 quantize_per_tensor_default_50" [label="(1, 197, 3072)", style=solid]; -"766 quantize_per_tensor_default_50" -> "767 dequantize_per_tensor_default_50" [label=None, style=solid]; -"767 dequantize_per_tensor_default_50" -> "768 dropout_29" [label=None, style=solid]; -"768 dropout_29" -> "769 linear_39" [label="(1, 197, 3072)", style=solid]; -"769 linear_39" -> "770 dropout_30" [label="(1, 197, 768)", style=solid]; -"770 dropout_30" -> "771 add_20" [label="(1, 197, 768)", style=solid]; -"771 add_20" -> "772 layer_norm_20" [label="(1, 197, 768)", style=solid]; -"771 add_20" -> "806 add_21" [label="(1, 197, 768)", style=solid]; -"772 layer_norm_20" -> "773 quantize_per_tensor_default_51" [label="(1, 197, 768)", style=solid]; -"773 quantize_per_tensor_default_51" -> "774 dequantize_per_tensor_default_51" [label=None, style=solid]; -"774 dequantize_per_tensor_default_51" -> "775 transpose_60" [label=None, style=solid]; -"775 transpose_60" -> "776 linear_40" [label="(197, 1, 768)", style=solid]; -"776 linear_40" -> "777 unflatten_10" [label="(197, 1, 2304)", style=solid]; -"777 unflatten_10" -> "778 unsqueeze_10" [label="(197, 1, 3, 768)", style=solid]; -"778 unsqueeze_10" -> "779 transpose_61" [label="(1, 197, 1, 3, 768)", style=solid]; -"779 transpose_61" -> "780 squeeze_10" [label="(3, 197, 1, 1, 768)", style=solid]; -"780 squeeze_10" -> "781 contiguous_10" [label="(3, 197, 1, 768)", style=solid]; -"781 contiguous_10" -> "782 quantize_per_tensor_default_52" [label="(3, 197, 1, 768)", style=solid]; -"781 contiguous_10" -> "787 select_32" [label="(3, 197, 1, 768)", style=solid]; -"782 quantize_per_tensor_default_52" -> "783 dequantize_per_tensor_default_83" [label=None, style=solid]; -"782 quantize_per_tensor_default_52" -> "784 dequantize_per_tensor_default_82" [label=None, style=solid]; -"783 dequantize_per_tensor_default_83" -> "786 select_31" [label=None, style=solid]; -"784 dequantize_per_tensor_default_82" -> "785 select_30" [label=None, style=solid]; -"785 select_30" -> "788 view_80" [label="(197, 1, 768)", style=solid]; -"786 select_31" -> "790 view_81" [label="(197, 1, 768)", style=solid]; -"787 select_32" -> "792 view_82" [label="(197, 1, 768)", style=solid]; -"788 view_80" -> "789 transpose_62" [label="(197, 12, 64)", style=solid]; -"789 transpose_62" -> "794 view_83" [label="(12, 197, 64)", style=solid]; -"790 view_81" -> "791 transpose_63" [label="(197, 12, 64)", style=solid]; -"791 transpose_63" -> "795 view_84" [label="(12, 197, 64)", style=solid]; -"792 view_82" -> "793 transpose_64" [label="(197, 12, 64)", style=solid]; -"793 transpose_64" -> "796 view_85" [label="(12, 197, 64)", style=solid]; -"794 view_83" -> "797 scaled_dot_product_attention_10" [label="(1, 12, 197, 64)", style=solid]; -"795 view_84" -> "797 scaled_dot_product_attention_10" [label="(1, 12, 197, 64)", style=solid]; -"796 view_85" -> "797 scaled_dot_product_attention_10" [label="(1, 12, 197, 64)", style=solid]; -"797 scaled_dot_product_attention_10" -> "798 quantize_per_tensor_default_53" [label="(1, 12, 197, 64)", style=solid]; -"798 quantize_per_tensor_default_53" -> "799 dequantize_per_tensor_default_53" [label=None, style=solid]; -"799 dequantize_per_tensor_default_53" -> "800 permute_11" [label=None, style=solid]; -"800 permute_11" -> "801 view_86" [label="(197, 1, 12, 64)", style=solid]; -"801 view_86" -> "802 linear_41" [label="(197, 768)", style=solid]; -"802 linear_41" -> "803 view_87" [label="(197, 768)", style=solid]; -"803 view_87" -> "804 transpose_65" [label="(197, 1, 768)", style=solid]; -"804 transpose_65" -> "805 dropout_31" [label="(1, 197, 768)", style=solid]; -"805 dropout_31" -> "806 add_21" [label="(1, 197, 768)", style=solid]; -"806 add_21" -> "807 layer_norm_21" [label="(1, 197, 768)", style=solid]; -"806 add_21" -> "817 add_22" [label="(1, 197, 768)", style=solid]; -"807 layer_norm_21" -> "808 quantize_per_tensor_default_54" [label="(1, 197, 768)", style=solid]; -"808 quantize_per_tensor_default_54" -> "809 dequantize_per_tensor_default_54" [label=None, style=solid]; -"809 dequantize_per_tensor_default_54" -> "810 linear_42" [label=None, style=solid]; -"810 linear_42" -> "811 gelu_10" [label="(1, 197, 3072)", style=solid]; -"811 gelu_10" -> "812 quantize_per_tensor_default_55" [label="(1, 197, 3072)", style=solid]; -"812 quantize_per_tensor_default_55" -> "813 dequantize_per_tensor_default_55" [label=None, style=solid]; -"813 dequantize_per_tensor_default_55" -> "814 dropout_32" [label=None, style=solid]; -"814 dropout_32" -> "815 linear_43" [label="(1, 197, 3072)", style=solid]; -"815 linear_43" -> "816 dropout_33" [label="(1, 197, 768)", style=solid]; -"816 dropout_33" -> "817 add_22" [label="(1, 197, 768)", style=solid]; -"817 add_22" -> "818 layer_norm_22" [label="(1, 197, 768)", style=solid]; -"817 add_22" -> "852 add_23" [label="(1, 197, 768)", style=solid]; -"818 layer_norm_22" -> "819 quantize_per_tensor_default_56" [label="(1, 197, 768)", style=solid]; -"819 quantize_per_tensor_default_56" -> "820 dequantize_per_tensor_default_56" [label=None, style=solid]; -"820 dequantize_per_tensor_default_56" -> "821 transpose_66" [label=None, style=solid]; -"821 transpose_66" -> "822 linear_44" [label="(197, 1, 768)", style=solid]; -"822 linear_44" -> "823 unflatten_11" [label="(197, 1, 2304)", style=solid]; -"823 unflatten_11" -> "824 unsqueeze_11" [label="(197, 1, 3, 768)", style=solid]; -"824 unsqueeze_11" -> "825 transpose_67" [label="(1, 197, 1, 3, 768)", style=solid]; -"825 transpose_67" -> "826 squeeze_11" [label="(3, 197, 1, 1, 768)", style=solid]; -"826 squeeze_11" -> "827 contiguous_11" [label="(3, 197, 1, 768)", style=solid]; -"827 contiguous_11" -> "828 quantize_per_tensor_default_57" [label="(3, 197, 1, 768)", style=solid]; -"827 contiguous_11" -> "833 select_35" [label="(3, 197, 1, 768)", style=solid]; -"828 quantize_per_tensor_default_57" -> "829 dequantize_per_tensor_default_85" [label=None, style=solid]; -"828 quantize_per_tensor_default_57" -> "830 dequantize_per_tensor_default_84" [label=None, style=solid]; -"829 dequantize_per_tensor_default_85" -> "832 select_34" [label=None, style=solid]; -"830 dequantize_per_tensor_default_84" -> "831 select_33" [label=None, style=solid]; -"831 select_33" -> "834 view_88" [label="(197, 1, 768)", style=solid]; -"832 select_34" -> "836 view_89" [label="(197, 1, 768)", style=solid]; -"833 select_35" -> "838 view_90" [label="(197, 1, 768)", style=solid]; -"834 view_88" -> "835 transpose_68" [label="(197, 12, 64)", style=solid]; -"835 transpose_68" -> "840 view_91" [label="(12, 197, 64)", style=solid]; -"836 view_89" -> "837 transpose_69" [label="(197, 12, 64)", style=solid]; -"837 transpose_69" -> "841 view_92" [label="(12, 197, 64)", style=solid]; -"838 view_90" -> "839 transpose_70" [label="(197, 12, 64)", style=solid]; -"839 transpose_70" -> "842 view_93" [label="(12, 197, 64)", style=solid]; -"840 view_91" -> "843 scaled_dot_product_attention_11" [label="(1, 12, 197, 64)", style=solid]; -"841 view_92" -> "843 scaled_dot_product_attention_11" [label="(1, 12, 197, 64)", style=solid]; -"842 view_93" -> "843 scaled_dot_product_attention_11" [label="(1, 12, 197, 64)", style=solid]; -"843 scaled_dot_product_attention_11" -> "844 quantize_per_tensor_default_58" [label="(1, 12, 197, 64)", style=solid]; -"844 quantize_per_tensor_default_58" -> "845 dequantize_per_tensor_default_58" [label=None, style=solid]; -"845 dequantize_per_tensor_default_58" -> "846 permute_12" [label=None, style=solid]; -"846 permute_12" -> "847 view_94" [label="(197, 1, 12, 64)", style=solid]; -"847 view_94" -> "848 linear_45" [label="(197, 768)", style=solid]; -"848 linear_45" -> "849 view_95" [label="(197, 768)", style=solid]; -"849 view_95" -> "850 transpose_71" [label="(197, 1, 768)", style=solid]; -"850 transpose_71" -> "851 dropout_34" [label="(1, 197, 768)", style=solid]; -"851 dropout_34" -> "852 add_23" [label="(1, 197, 768)", style=solid]; -"852 add_23" -> "853 layer_norm_23" [label="(1, 197, 768)", style=solid]; -"852 add_23" -> "863 add_24" [label="(1, 197, 768)", style=solid]; -"853 layer_norm_23" -> "854 quantize_per_tensor_default_59" [label="(1, 197, 768)", style=solid]; -"854 quantize_per_tensor_default_59" -> "855 dequantize_per_tensor_default_59" [label=None, style=solid]; -"855 dequantize_per_tensor_default_59" -> "856 linear_46" [label=None, style=solid]; -"856 linear_46" -> "857 gelu_11" [label="(1, 197, 3072)", style=solid]; -"857 gelu_11" -> "858 quantize_per_tensor_default_60" [label="(1, 197, 3072)", style=solid]; -"858 quantize_per_tensor_default_60" -> "859 dequantize_per_tensor_default_60" [label=None, style=solid]; -"859 dequantize_per_tensor_default_60" -> "860 dropout_35" [label=None, style=solid]; -"860 dropout_35" -> "861 linear_47" [label="(1, 197, 3072)", style=solid]; -"861 linear_47" -> "862 dropout_36" [label="(1, 197, 768)", style=solid]; -"862 dropout_36" -> "863 add_24" [label="(1, 197, 768)", style=solid]; -"863 add_24" -> "864 layer_norm_24" [label="(1, 197, 768)", style=solid]; -"864 layer_norm_24" -> "865 quantize_per_tensor_default_61" [label="(1, 197, 768)", style=solid]; -"865 quantize_per_tensor_default_61" -> "866 dequantize_per_tensor_default_61" [label=None, style=solid]; -"866 dequantize_per_tensor_default_61" -> "867 slice_1" [label=None, style=solid]; -"867 slice_1" -> "868 select_36" [label="(1, 197, 768)", style=solid]; -"868 select_36" -> "869 linear_48" [label="(1, 768)", style=solid]; -"869 linear_48" -> "870 output_1" [label="(1, 1000)", style=solid]; +"0 encoder_pos_embedding" [id=0, type=get_attr]; +"1 _scale_0" [id=1, type=get_attr]; +"2 _zero_point_0" [id=2, type=get_attr]; +"3 _frozen_param1" [id=3, type=get_attr]; +"4 dequantize_per_channel_default" [id=4, type=dequantize_per_channel]; +"5 conv_proj_bias" [id=5, type=get_attr]; +"6 encoder_layers_encoder_layer_0_ln_1_weight" [id=6, type=get_attr]; +"7 encoder_layers_encoder_layer_0_ln_1_bias" [id=7, type=get_attr]; +"8 _scale_1" [id=8, type=get_attr]; +"9 _zero_point_1" [id=9, type=get_attr]; +"10 _frozen_param2" [id=10, type=get_attr]; +"11 dequantize_per_channel_default_1" [id=11, type=dequantize_per_channel]; +"12 encoder_layers_encoder_layer_0_self_attention_in_proj_bias" [id=12, type=get_attr]; +"13 _scale_2" [id=13, type=get_attr]; +"14 _zero_point_2" [id=14, type=get_attr]; +"15 _frozen_param3" [id=15, type=get_attr]; +"16 dequantize_per_channel_default_2" [id=16, type=dequantize_per_channel]; +"17 encoder_layers_encoder_layer_0_self_attention_out_proj_bias" [id=17, type=get_attr]; +"18 encoder_layers_encoder_layer_0_ln_2_weight" [id=18, type=get_attr]; +"19 encoder_layers_encoder_layer_0_ln_2_bias" [id=19, type=get_attr]; +"20 _scale_3" [id=20, type=get_attr]; +"21 _zero_point_3" [id=21, type=get_attr]; +"22 _frozen_param4" [id=22, type=get_attr]; +"23 dequantize_per_channel_default_3" [id=23, type=dequantize_per_channel]; +"24 encoder_layers_encoder_layer_0_mlp_0_bias" [id=24, type=get_attr]; +"25 _scale_4" [id=25, type=get_attr]; +"26 _zero_point_4" [id=26, type=get_attr]; +"27 _frozen_param5" [id=27, type=get_attr]; +"28 dequantize_per_channel_default_4" [id=28, type=dequantize_per_channel]; +"29 encoder_layers_encoder_layer_0_mlp_3_bias" [id=29, type=get_attr]; +"30 encoder_layers_encoder_layer_1_ln_1_weight" [id=30, type=get_attr]; +"31 encoder_layers_encoder_layer_1_ln_1_bias" [id=31, type=get_attr]; +"32 _scale_5" [id=32, type=get_attr]; +"33 _zero_point_5" [id=33, type=get_attr]; +"34 _frozen_param6" [id=34, type=get_attr]; +"35 dequantize_per_channel_default_5" [id=35, type=dequantize_per_channel]; +"36 encoder_layers_encoder_layer_1_self_attention_in_proj_bias" [id=36, type=get_attr]; +"37 _scale_6" [id=37, type=get_attr]; +"38 _zero_point_6" [id=38, type=get_attr]; +"39 _frozen_param7" [id=39, type=get_attr]; +"40 dequantize_per_channel_default_6" [id=40, type=dequantize_per_channel]; +"41 encoder_layers_encoder_layer_1_self_attention_out_proj_bias" [id=41, type=get_attr]; +"42 encoder_layers_encoder_layer_1_ln_2_weight" [id=42, type=get_attr]; +"43 encoder_layers_encoder_layer_1_ln_2_bias" [id=43, type=get_attr]; +"44 _scale_7" [id=44, type=get_attr]; +"45 _zero_point_7" [id=45, type=get_attr]; +"46 _frozen_param8" [id=46, type=get_attr]; +"47 dequantize_per_channel_default_7" [id=47, type=dequantize_per_channel]; +"48 encoder_layers_encoder_layer_1_mlp_0_bias" [id=48, type=get_attr]; +"49 _scale_8" [id=49, type=get_attr]; +"50 _zero_point_8" [id=50, type=get_attr]; +"51 _frozen_param9" [id=51, type=get_attr]; +"52 dequantize_per_channel_default_8" [id=52, type=dequantize_per_channel]; +"53 encoder_layers_encoder_layer_1_mlp_3_bias" [id=53, type=get_attr]; +"54 encoder_layers_encoder_layer_2_ln_1_weight" [id=54, type=get_attr]; +"55 encoder_layers_encoder_layer_2_ln_1_bias" [id=55, type=get_attr]; +"56 _scale_9" [id=56, type=get_attr]; +"57 _zero_point_9" [id=57, type=get_attr]; +"58 _frozen_param10" [id=58, type=get_attr]; +"59 dequantize_per_channel_default_9" [id=59, type=dequantize_per_channel]; +"60 encoder_layers_encoder_layer_2_self_attention_in_proj_bias" [id=60, type=get_attr]; +"61 _scale_10" [id=61, type=get_attr]; +"62 _zero_point_10" [id=62, type=get_attr]; +"63 _frozen_param11" [id=63, type=get_attr]; +"64 dequantize_per_channel_default_10" [id=64, type=dequantize_per_channel]; +"65 encoder_layers_encoder_layer_2_self_attention_out_proj_bias" [id=65, type=get_attr]; +"66 encoder_layers_encoder_layer_2_ln_2_weight" [id=66, type=get_attr]; +"67 encoder_layers_encoder_layer_2_ln_2_bias" [id=67, type=get_attr]; +"68 _scale_11" [id=68, type=get_attr]; +"69 _zero_point_11" [id=69, type=get_attr]; +"70 _frozen_param12" [id=70, type=get_attr]; +"71 dequantize_per_channel_default_11" [id=71, type=dequantize_per_channel]; +"72 encoder_layers_encoder_layer_2_mlp_0_bias" [id=72, type=get_attr]; +"73 _scale_12" [id=73, type=get_attr]; +"74 _zero_point_12" [id=74, type=get_attr]; +"75 _frozen_param13" [id=75, type=get_attr]; +"76 dequantize_per_channel_default_12" [id=76, type=dequantize_per_channel]; +"77 encoder_layers_encoder_layer_2_mlp_3_bias" [id=77, type=get_attr]; +"78 encoder_layers_encoder_layer_3_ln_1_weight" [id=78, type=get_attr]; +"79 encoder_layers_encoder_layer_3_ln_1_bias" [id=79, type=get_attr]; +"80 _scale_13" [id=80, type=get_attr]; +"81 _zero_point_13" [id=81, type=get_attr]; +"82 _frozen_param14" [id=82, type=get_attr]; +"83 dequantize_per_channel_default_13" [id=83, type=dequantize_per_channel]; +"84 encoder_layers_encoder_layer_3_self_attention_in_proj_bias" [id=84, type=get_attr]; +"85 _scale_14" [id=85, type=get_attr]; +"86 _zero_point_14" [id=86, type=get_attr]; +"87 _frozen_param15" [id=87, type=get_attr]; +"88 dequantize_per_channel_default_14" [id=88, type=dequantize_per_channel]; +"89 encoder_layers_encoder_layer_3_self_attention_out_proj_bias" [id=89, type=get_attr]; +"90 encoder_layers_encoder_layer_3_ln_2_weight" [id=90, type=get_attr]; +"91 encoder_layers_encoder_layer_3_ln_2_bias" [id=91, type=get_attr]; +"92 _scale_15" [id=92, type=get_attr]; +"93 _zero_point_15" [id=93, type=get_attr]; +"94 _frozen_param16" [id=94, type=get_attr]; +"95 dequantize_per_channel_default_15" [id=95, type=dequantize_per_channel]; +"96 encoder_layers_encoder_layer_3_mlp_0_bias" [id=96, type=get_attr]; +"97 _scale_16" [id=97, type=get_attr]; +"98 _zero_point_16" [id=98, type=get_attr]; +"99 _frozen_param17" [id=99, type=get_attr]; +"100 dequantize_per_channel_default_16" [id=100, type=dequantize_per_channel]; +"101 encoder_layers_encoder_layer_3_mlp_3_bias" [id=101, type=get_attr]; +"102 encoder_layers_encoder_layer_4_ln_1_weight" [id=102, type=get_attr]; +"103 encoder_layers_encoder_layer_4_ln_1_bias" [id=103, type=get_attr]; +"104 _scale_17" [id=104, type=get_attr]; +"105 _zero_point_17" [id=105, type=get_attr]; +"106 _frozen_param18" [id=106, type=get_attr]; +"107 dequantize_per_channel_default_17" [id=107, type=dequantize_per_channel]; +"108 encoder_layers_encoder_layer_4_self_attention_in_proj_bias" [id=108, type=get_attr]; +"109 _scale_18" [id=109, type=get_attr]; +"110 _zero_point_18" [id=110, type=get_attr]; +"111 _frozen_param19" [id=111, type=get_attr]; +"112 dequantize_per_channel_default_18" [id=112, type=dequantize_per_channel]; +"113 encoder_layers_encoder_layer_4_self_attention_out_proj_bias" [id=113, type=get_attr]; +"114 encoder_layers_encoder_layer_4_ln_2_weight" [id=114, type=get_attr]; +"115 encoder_layers_encoder_layer_4_ln_2_bias" [id=115, type=get_attr]; +"116 _scale_19" [id=116, type=get_attr]; +"117 _zero_point_19" [id=117, type=get_attr]; +"118 _frozen_param20" [id=118, type=get_attr]; +"119 dequantize_per_channel_default_19" [id=119, type=dequantize_per_channel]; +"120 encoder_layers_encoder_layer_4_mlp_0_bias" [id=120, type=get_attr]; +"121 _scale_20" [id=121, type=get_attr]; +"122 _zero_point_20" [id=122, type=get_attr]; +"123 _frozen_param21" [id=123, type=get_attr]; +"124 dequantize_per_channel_default_20" [id=124, type=dequantize_per_channel]; +"125 encoder_layers_encoder_layer_4_mlp_3_bias" [id=125, type=get_attr]; +"126 encoder_layers_encoder_layer_5_ln_1_weight" [id=126, type=get_attr]; +"127 encoder_layers_encoder_layer_5_ln_1_bias" [id=127, type=get_attr]; +"128 _scale_21" [id=128, type=get_attr]; +"129 _zero_point_21" [id=129, type=get_attr]; +"130 _frozen_param22" [id=130, type=get_attr]; +"131 dequantize_per_channel_default_21" [id=131, type=dequantize_per_channel]; +"132 encoder_layers_encoder_layer_5_self_attention_in_proj_bias" [id=132, type=get_attr]; +"133 _scale_22" [id=133, type=get_attr]; +"134 _zero_point_22" [id=134, type=get_attr]; +"135 _frozen_param23" [id=135, type=get_attr]; +"136 dequantize_per_channel_default_22" [id=136, type=dequantize_per_channel]; +"137 encoder_layers_encoder_layer_5_self_attention_out_proj_bias" [id=137, type=get_attr]; +"138 encoder_layers_encoder_layer_5_ln_2_weight" [id=138, type=get_attr]; +"139 encoder_layers_encoder_layer_5_ln_2_bias" [id=139, type=get_attr]; +"140 _scale_23" [id=140, type=get_attr]; +"141 _zero_point_23" [id=141, type=get_attr]; +"142 _frozen_param24" [id=142, type=get_attr]; +"143 dequantize_per_channel_default_23" [id=143, type=dequantize_per_channel]; +"144 encoder_layers_encoder_layer_5_mlp_0_bias" [id=144, type=get_attr]; +"145 _scale_24" [id=145, type=get_attr]; +"146 _zero_point_24" [id=146, type=get_attr]; +"147 _frozen_param25" [id=147, type=get_attr]; +"148 dequantize_per_channel_default_24" [id=148, type=dequantize_per_channel]; +"149 encoder_layers_encoder_layer_5_mlp_3_bias" [id=149, type=get_attr]; +"150 encoder_layers_encoder_layer_6_ln_1_weight" [id=150, type=get_attr]; +"151 encoder_layers_encoder_layer_6_ln_1_bias" [id=151, type=get_attr]; +"152 _scale_25" [id=152, type=get_attr]; +"153 _zero_point_25" [id=153, type=get_attr]; +"154 _frozen_param26" [id=154, type=get_attr]; +"155 dequantize_per_channel_default_25" [id=155, type=dequantize_per_channel]; +"156 encoder_layers_encoder_layer_6_self_attention_in_proj_bias" [id=156, type=get_attr]; +"157 _scale_26" [id=157, type=get_attr]; +"158 _zero_point_26" [id=158, type=get_attr]; +"159 _frozen_param27" [id=159, type=get_attr]; +"160 dequantize_per_channel_default_26" [id=160, type=dequantize_per_channel]; +"161 encoder_layers_encoder_layer_6_self_attention_out_proj_bias" [id=161, type=get_attr]; +"162 encoder_layers_encoder_layer_6_ln_2_weight" [id=162, type=get_attr]; +"163 encoder_layers_encoder_layer_6_ln_2_bias" [id=163, type=get_attr]; +"164 _scale_27" [id=164, type=get_attr]; +"165 _zero_point_27" [id=165, type=get_attr]; +"166 _frozen_param28" [id=166, type=get_attr]; +"167 dequantize_per_channel_default_27" [id=167, type=dequantize_per_channel]; +"168 encoder_layers_encoder_layer_6_mlp_0_bias" [id=168, type=get_attr]; +"169 _scale_28" [id=169, type=get_attr]; +"170 _zero_point_28" [id=170, type=get_attr]; +"171 _frozen_param29" [id=171, type=get_attr]; +"172 dequantize_per_channel_default_28" [id=172, type=dequantize_per_channel]; +"173 encoder_layers_encoder_layer_6_mlp_3_bias" [id=173, type=get_attr]; +"174 encoder_layers_encoder_layer_7_ln_1_weight" [id=174, type=get_attr]; +"175 encoder_layers_encoder_layer_7_ln_1_bias" [id=175, type=get_attr]; +"176 _scale_29" [id=176, type=get_attr]; +"177 _zero_point_29" [id=177, type=get_attr]; +"178 _frozen_param30" [id=178, type=get_attr]; +"179 dequantize_per_channel_default_29" [id=179, type=dequantize_per_channel]; +"180 encoder_layers_encoder_layer_7_self_attention_in_proj_bias" [id=180, type=get_attr]; +"181 _scale_30" [id=181, type=get_attr]; +"182 _zero_point_30" [id=182, type=get_attr]; +"183 _frozen_param31" [id=183, type=get_attr]; +"184 dequantize_per_channel_default_30" [id=184, type=dequantize_per_channel]; +"185 encoder_layers_encoder_layer_7_self_attention_out_proj_bias" [id=185, type=get_attr]; +"186 encoder_layers_encoder_layer_7_ln_2_weight" [id=186, type=get_attr]; +"187 encoder_layers_encoder_layer_7_ln_2_bias" [id=187, type=get_attr]; +"188 _scale_31" [id=188, type=get_attr]; +"189 _zero_point_31" [id=189, type=get_attr]; +"190 _frozen_param32" [id=190, type=get_attr]; +"191 dequantize_per_channel_default_31" [id=191, type=dequantize_per_channel]; +"192 encoder_layers_encoder_layer_7_mlp_0_bias" [id=192, type=get_attr]; +"193 _scale_32" [id=193, type=get_attr]; +"194 _zero_point_32" [id=194, type=get_attr]; +"195 _frozen_param33" [id=195, type=get_attr]; +"196 dequantize_per_channel_default_32" [id=196, type=dequantize_per_channel]; +"197 encoder_layers_encoder_layer_7_mlp_3_bias" [id=197, type=get_attr]; +"198 encoder_layers_encoder_layer_8_ln_1_weight" [id=198, type=get_attr]; +"199 encoder_layers_encoder_layer_8_ln_1_bias" [id=199, type=get_attr]; +"200 _scale_33" [id=200, type=get_attr]; +"201 _zero_point_33" [id=201, type=get_attr]; +"202 _frozen_param34" [id=202, type=get_attr]; +"203 dequantize_per_channel_default_33" [id=203, type=dequantize_per_channel]; +"204 encoder_layers_encoder_layer_8_self_attention_in_proj_bias" [id=204, type=get_attr]; +"205 _scale_34" [id=205, type=get_attr]; +"206 _zero_point_34" [id=206, type=get_attr]; +"207 _frozen_param35" [id=207, type=get_attr]; +"208 dequantize_per_channel_default_34" [id=208, type=dequantize_per_channel]; +"209 encoder_layers_encoder_layer_8_self_attention_out_proj_bias" [id=209, type=get_attr]; +"210 encoder_layers_encoder_layer_8_ln_2_weight" [id=210, type=get_attr]; +"211 encoder_layers_encoder_layer_8_ln_2_bias" [id=211, type=get_attr]; +"212 _scale_35" [id=212, type=get_attr]; +"213 _zero_point_35" [id=213, type=get_attr]; +"214 _frozen_param36" [id=214, type=get_attr]; +"215 dequantize_per_channel_default_35" [id=215, type=dequantize_per_channel]; +"216 encoder_layers_encoder_layer_8_mlp_0_bias" [id=216, type=get_attr]; +"217 _scale_36" [id=217, type=get_attr]; +"218 _zero_point_36" [id=218, type=get_attr]; +"219 _frozen_param37" [id=219, type=get_attr]; +"220 dequantize_per_channel_default_36" [id=220, type=dequantize_per_channel]; +"221 encoder_layers_encoder_layer_8_mlp_3_bias" [id=221, type=get_attr]; +"222 encoder_layers_encoder_layer_9_ln_1_weight" [id=222, type=get_attr]; +"223 encoder_layers_encoder_layer_9_ln_1_bias" [id=223, type=get_attr]; +"224 _scale_37" [id=224, type=get_attr]; +"225 _zero_point_37" [id=225, type=get_attr]; +"226 _frozen_param38" [id=226, type=get_attr]; +"227 dequantize_per_channel_default_37" [id=227, type=dequantize_per_channel]; +"228 encoder_layers_encoder_layer_9_self_attention_in_proj_bias" [id=228, type=get_attr]; +"229 _scale_38" [id=229, type=get_attr]; +"230 _zero_point_38" [id=230, type=get_attr]; +"231 _frozen_param39" [id=231, type=get_attr]; +"232 dequantize_per_channel_default_38" [id=232, type=dequantize_per_channel]; +"233 encoder_layers_encoder_layer_9_self_attention_out_proj_bias" [id=233, type=get_attr]; +"234 encoder_layers_encoder_layer_9_ln_2_weight" [id=234, type=get_attr]; +"235 encoder_layers_encoder_layer_9_ln_2_bias" [id=235, type=get_attr]; +"236 _scale_39" [id=236, type=get_attr]; +"237 _zero_point_39" [id=237, type=get_attr]; +"238 _frozen_param40" [id=238, type=get_attr]; +"239 dequantize_per_channel_default_39" [id=239, type=dequantize_per_channel]; +"240 encoder_layers_encoder_layer_9_mlp_0_bias" [id=240, type=get_attr]; +"241 _scale_40" [id=241, type=get_attr]; +"242 _zero_point_40" [id=242, type=get_attr]; +"243 _frozen_param41" [id=243, type=get_attr]; +"244 dequantize_per_channel_default_40" [id=244, type=dequantize_per_channel]; +"245 encoder_layers_encoder_layer_9_mlp_3_bias" [id=245, type=get_attr]; +"246 encoder_layers_encoder_layer_10_ln_1_weight" [id=246, type=get_attr]; +"247 encoder_layers_encoder_layer_10_ln_1_bias" [id=247, type=get_attr]; +"248 _scale_41" [id=248, type=get_attr]; +"249 _zero_point_41" [id=249, type=get_attr]; +"250 _frozen_param42" [id=250, type=get_attr]; +"251 dequantize_per_channel_default_41" [id=251, type=dequantize_per_channel]; +"252 encoder_layers_encoder_layer_10_self_attention_in_proj_bias" [id=252, type=get_attr]; +"253 _scale_42" [id=253, type=get_attr]; +"254 _zero_point_42" [id=254, type=get_attr]; +"255 _frozen_param43" [id=255, type=get_attr]; +"256 dequantize_per_channel_default_42" [id=256, type=dequantize_per_channel]; +"257 encoder_layers_encoder_layer_10_self_attention_out_proj_bias" [id=257, type=get_attr]; +"258 encoder_layers_encoder_layer_10_ln_2_weight" [id=258, type=get_attr]; +"259 encoder_layers_encoder_layer_10_ln_2_bias" [id=259, type=get_attr]; +"260 _scale_43" [id=260, type=get_attr]; +"261 _zero_point_43" [id=261, type=get_attr]; +"262 _frozen_param44" [id=262, type=get_attr]; +"263 dequantize_per_channel_default_43" [id=263, type=dequantize_per_channel]; +"264 encoder_layers_encoder_layer_10_mlp_0_bias" [id=264, type=get_attr]; +"265 _scale_44" [id=265, type=get_attr]; +"266 _zero_point_44" [id=266, type=get_attr]; +"267 _frozen_param45" [id=267, type=get_attr]; +"268 dequantize_per_channel_default_44" [id=268, type=dequantize_per_channel]; +"269 encoder_layers_encoder_layer_10_mlp_3_bias" [id=269, type=get_attr]; +"270 encoder_layers_encoder_layer_11_ln_1_weight" [id=270, type=get_attr]; +"271 encoder_layers_encoder_layer_11_ln_1_bias" [id=271, type=get_attr]; +"272 _scale_45" [id=272, type=get_attr]; +"273 _zero_point_45" [id=273, type=get_attr]; +"274 _frozen_param46" [id=274, type=get_attr]; +"275 dequantize_per_channel_default_45" [id=275, type=dequantize_per_channel]; +"276 encoder_layers_encoder_layer_11_self_attention_in_proj_bias" [id=276, type=get_attr]; +"277 _scale_46" [id=277, type=get_attr]; +"278 _zero_point_46" [id=278, type=get_attr]; +"279 _frozen_param47" [id=279, type=get_attr]; +"280 dequantize_per_channel_default_46" [id=280, type=dequantize_per_channel]; +"281 encoder_layers_encoder_layer_11_self_attention_out_proj_bias" [id=281, type=get_attr]; +"282 encoder_layers_encoder_layer_11_ln_2_weight" [id=282, type=get_attr]; +"283 encoder_layers_encoder_layer_11_ln_2_bias" [id=283, type=get_attr]; +"284 _scale_47" [id=284, type=get_attr]; +"285 _zero_point_47" [id=285, type=get_attr]; +"286 _frozen_param48" [id=286, type=get_attr]; +"287 dequantize_per_channel_default_47" [id=287, type=dequantize_per_channel]; +"288 encoder_layers_encoder_layer_11_mlp_0_bias" [id=288, type=get_attr]; +"289 _scale_48" [id=289, type=get_attr]; +"290 _zero_point_48" [id=290, type=get_attr]; +"291 _frozen_param49" [id=291, type=get_attr]; +"292 dequantize_per_channel_default_48" [id=292, type=dequantize_per_channel]; +"293 encoder_layers_encoder_layer_11_mlp_3_bias" [id=293, type=get_attr]; +"294 encoder_ln_weight" [id=294, type=get_attr]; +"295 encoder_ln_bias" [id=295, type=get_attr]; +"296 _scale_49" [id=296, type=get_attr]; +"297 _zero_point_49" [id=297, type=get_attr]; +"298 _frozen_param50" [id=298, type=get_attr]; +"299 dequantize_per_channel_default_49" [id=299, type=dequantize_per_channel]; +"300 heads_head_bias" [id=300, type=get_attr]; +"301 x" [id=301, type=input]; +"302 quantize_per_tensor_default" [id=302, type=quantize_per_tensor]; +"303 dequantize_per_tensor_default" [id=303, type=dequantize_per_tensor]; +"304 conv2d" [id=304, type=conv2d]; +"305 reshape" [id=305, type=reshape]; +"306 permute" [id=306, type=permute]; +"307 _frozen_param0" [id=307, type=get_attr]; +"308 cat" [id=308, type=cat]; +"309 add" [id=309, type=add]; +"310 dropout" [id=310, type=dropout]; +"311 layer_norm" [id=311, type=layer_norm]; +"312 quantize_per_tensor_default_1" [id=312, type=quantize_per_tensor]; +"313 dequantize_per_tensor_default_1" [id=313, type=dequantize_per_tensor]; +"314 transpose" [id=314, type=transpose]; +"315 linear" [id=315, type=linear]; +"316 unflatten" [id=316, type=unflatten]; +"317 unsqueeze" [id=317, type=unsqueeze]; +"318 transpose_1" [id=318, type=transpose]; +"319 squeeze" [id=319, type=squeeze]; +"320 contiguous" [id=320, type=contiguous]; +"321 quantize_per_tensor_default_2" [id=321, type=quantize_per_tensor]; +"322 dequantize_per_tensor_default_63" [id=322, type=dequantize_per_tensor]; +"323 dequantize_per_tensor_default_62" [id=323, type=dequantize_per_tensor]; +"324 select" [id=324, type=select]; +"325 select_1" [id=325, type=select]; +"326 select_2" [id=326, type=select]; +"327 view" [id=327, type=view]; +"328 transpose_2" [id=328, type=transpose]; +"329 view_1" [id=329, type=view]; +"330 transpose_3" [id=330, type=transpose]; +"331 view_2" [id=331, type=view]; +"332 transpose_4" [id=332, type=transpose]; +"333 view_3" [id=333, type=view]; +"334 view_4" [id=334, type=view]; +"335 view_5" [id=335, type=view]; +"336 scaled_dot_product_attention" [id=336, type=scaled_dot_product_attention]; +"337 quantize_per_tensor_default_3" [id=337, type=quantize_per_tensor]; +"338 dequantize_per_tensor_default_3" [id=338, type=dequantize_per_tensor]; +"339 permute_1" [id=339, type=permute]; +"340 view_6" [id=340, type=view]; +"341 linear_1" [id=341, type=linear]; +"342 view_7" [id=342, type=view]; +"343 transpose_5" [id=343, type=transpose]; +"344 dropout_1" [id=344, type=dropout]; +"345 add_1" [id=345, type=add]; +"346 layer_norm_1" [id=346, type=layer_norm]; +"347 quantize_per_tensor_default_4" [id=347, type=quantize_per_tensor]; +"348 dequantize_per_tensor_default_4" [id=348, type=dequantize_per_tensor]; +"349 linear_2" [id=349, type=linear]; +"350 gelu" [id=350, type=gelu]; +"351 quantize_per_tensor_default_5" [id=351, type=quantize_per_tensor]; +"352 dequantize_per_tensor_default_5" [id=352, type=dequantize_per_tensor]; +"353 dropout_2" [id=353, type=dropout]; +"354 linear_3" [id=354, type=linear]; +"355 dropout_3" [id=355, type=dropout]; +"356 add_2" [id=356, type=add]; +"357 layer_norm_2" [id=357, type=layer_norm]; +"358 quantize_per_tensor_default_6" [id=358, type=quantize_per_tensor]; +"359 dequantize_per_tensor_default_6" [id=359, type=dequantize_per_tensor]; +"360 transpose_6" [id=360, type=transpose]; +"361 linear_4" [id=361, type=linear]; +"362 unflatten_1" [id=362, type=unflatten]; +"363 unsqueeze_1" [id=363, type=unsqueeze]; +"364 transpose_7" [id=364, type=transpose]; +"365 squeeze_1" [id=365, type=squeeze]; +"366 contiguous_1" [id=366, type=contiguous]; +"367 quantize_per_tensor_default_7" [id=367, type=quantize_per_tensor]; +"368 dequantize_per_tensor_default_65" [id=368, type=dequantize_per_tensor]; +"369 dequantize_per_tensor_default_64" [id=369, type=dequantize_per_tensor]; +"370 select_3" [id=370, type=select]; +"371 select_4" [id=371, type=select]; +"372 select_5" [id=372, type=select]; +"373 view_8" [id=373, type=view]; +"374 transpose_8" [id=374, type=transpose]; +"375 view_9" [id=375, type=view]; +"376 transpose_9" [id=376, type=transpose]; +"377 view_10" [id=377, type=view]; +"378 transpose_10" [id=378, type=transpose]; +"379 view_11" [id=379, type=view]; +"380 view_12" [id=380, type=view]; +"381 view_13" [id=381, type=view]; +"382 scaled_dot_product_attention_1" [id=382, type=scaled_dot_product_attention]; +"383 quantize_per_tensor_default_8" [id=383, type=quantize_per_tensor]; +"384 dequantize_per_tensor_default_8" [id=384, type=dequantize_per_tensor]; +"385 permute_2" [id=385, type=permute]; +"386 view_14" [id=386, type=view]; +"387 linear_5" [id=387, type=linear]; +"388 view_15" [id=388, type=view]; +"389 transpose_11" [id=389, type=transpose]; +"390 dropout_4" [id=390, type=dropout]; +"391 add_3" [id=391, type=add]; +"392 layer_norm_3" [id=392, type=layer_norm]; +"393 quantize_per_tensor_default_9" [id=393, type=quantize_per_tensor]; +"394 dequantize_per_tensor_default_9" [id=394, type=dequantize_per_tensor]; +"395 linear_6" [id=395, type=linear]; +"396 gelu_1" [id=396, type=gelu]; +"397 quantize_per_tensor_default_10" [id=397, type=quantize_per_tensor]; +"398 dequantize_per_tensor_default_10" [id=398, type=dequantize_per_tensor]; +"399 dropout_5" [id=399, type=dropout]; +"400 linear_7" [id=400, type=linear]; +"401 dropout_6" [id=401, type=dropout]; +"402 add_4" [id=402, type=add]; +"403 layer_norm_4" [id=403, type=layer_norm]; +"404 quantize_per_tensor_default_11" [id=404, type=quantize_per_tensor]; +"405 dequantize_per_tensor_default_11" [id=405, type=dequantize_per_tensor]; +"406 transpose_12" [id=406, type=transpose]; +"407 linear_8" [id=407, type=linear]; +"408 unflatten_2" [id=408, type=unflatten]; +"409 unsqueeze_2" [id=409, type=unsqueeze]; +"410 transpose_13" [id=410, type=transpose]; +"411 squeeze_2" [id=411, type=squeeze]; +"412 contiguous_2" [id=412, type=contiguous]; +"413 quantize_per_tensor_default_12" [id=413, type=quantize_per_tensor]; +"414 dequantize_per_tensor_default_67" [id=414, type=dequantize_per_tensor]; +"415 dequantize_per_tensor_default_66" [id=415, type=dequantize_per_tensor]; +"416 select_6" [id=416, type=select]; +"417 select_7" [id=417, type=select]; +"418 select_8" [id=418, type=select]; +"419 view_16" [id=419, type=view]; +"420 transpose_14" [id=420, type=transpose]; +"421 view_17" [id=421, type=view]; +"422 transpose_15" [id=422, type=transpose]; +"423 view_18" [id=423, type=view]; +"424 transpose_16" [id=424, type=transpose]; +"425 view_19" [id=425, type=view]; +"426 view_20" [id=426, type=view]; +"427 view_21" [id=427, type=view]; +"428 scaled_dot_product_attention_2" [id=428, type=scaled_dot_product_attention]; +"429 quantize_per_tensor_default_13" [id=429, type=quantize_per_tensor]; +"430 dequantize_per_tensor_default_13" [id=430, type=dequantize_per_tensor]; +"431 permute_3" [id=431, type=permute]; +"432 view_22" [id=432, type=view]; +"433 linear_9" [id=433, type=linear]; +"434 view_23" [id=434, type=view]; +"435 transpose_17" [id=435, type=transpose]; +"436 dropout_7" [id=436, type=dropout]; +"437 add_5" [id=437, type=add]; +"438 layer_norm_5" [id=438, type=layer_norm]; +"439 quantize_per_tensor_default_14" [id=439, type=quantize_per_tensor]; +"440 dequantize_per_tensor_default_14" [id=440, type=dequantize_per_tensor]; +"441 linear_10" [id=441, type=linear]; +"442 gelu_2" [id=442, type=gelu]; +"443 quantize_per_tensor_default_15" [id=443, type=quantize_per_tensor]; +"444 dequantize_per_tensor_default_15" [id=444, type=dequantize_per_tensor]; +"445 dropout_8" [id=445, type=dropout]; +"446 linear_11" [id=446, type=linear]; +"447 dropout_9" [id=447, type=dropout]; +"448 add_6" [id=448, type=add]; +"449 layer_norm_6" [id=449, type=layer_norm]; +"450 quantize_per_tensor_default_16" [id=450, type=quantize_per_tensor]; +"451 dequantize_per_tensor_default_16" [id=451, type=dequantize_per_tensor]; +"452 transpose_18" [id=452, type=transpose]; +"453 linear_12" [id=453, type=linear]; +"454 unflatten_3" [id=454, type=unflatten]; +"455 unsqueeze_3" [id=455, type=unsqueeze]; +"456 transpose_19" [id=456, type=transpose]; +"457 squeeze_3" [id=457, type=squeeze]; +"458 contiguous_3" [id=458, type=contiguous]; +"459 quantize_per_tensor_default_17" [id=459, type=quantize_per_tensor]; +"460 dequantize_per_tensor_default_69" [id=460, type=dequantize_per_tensor]; +"461 dequantize_per_tensor_default_68" [id=461, type=dequantize_per_tensor]; +"462 select_9" [id=462, type=select]; +"463 select_10" [id=463, type=select]; +"464 select_11" [id=464, type=select]; +"465 view_24" [id=465, type=view]; +"466 transpose_20" [id=466, type=transpose]; +"467 view_25" [id=467, type=view]; +"468 transpose_21" [id=468, type=transpose]; +"469 view_26" [id=469, type=view]; +"470 transpose_22" [id=470, type=transpose]; +"471 view_27" [id=471, type=view]; +"472 view_28" [id=472, type=view]; +"473 view_29" [id=473, type=view]; +"474 scaled_dot_product_attention_3" [id=474, type=scaled_dot_product_attention]; +"475 quantize_per_tensor_default_18" [id=475, type=quantize_per_tensor]; +"476 dequantize_per_tensor_default_18" [id=476, type=dequantize_per_tensor]; +"477 permute_4" [id=477, type=permute]; +"478 view_30" [id=478, type=view]; +"479 linear_13" [id=479, type=linear]; +"480 view_31" [id=480, type=view]; +"481 transpose_23" [id=481, type=transpose]; +"482 dropout_10" [id=482, type=dropout]; +"483 add_7" [id=483, type=add]; +"484 layer_norm_7" [id=484, type=layer_norm]; +"485 quantize_per_tensor_default_19" [id=485, type=quantize_per_tensor]; +"486 dequantize_per_tensor_default_19" [id=486, type=dequantize_per_tensor]; +"487 linear_14" [id=487, type=linear]; +"488 gelu_3" [id=488, type=gelu]; +"489 quantize_per_tensor_default_20" [id=489, type=quantize_per_tensor]; +"490 dequantize_per_tensor_default_20" [id=490, type=dequantize_per_tensor]; +"491 dropout_11" [id=491, type=dropout]; +"492 linear_15" [id=492, type=linear]; +"493 dropout_12" [id=493, type=dropout]; +"494 add_8" [id=494, type=add]; +"495 layer_norm_8" [id=495, type=layer_norm]; +"496 quantize_per_tensor_default_21" [id=496, type=quantize_per_tensor]; +"497 dequantize_per_tensor_default_21" [id=497, type=dequantize_per_tensor]; +"498 transpose_24" [id=498, type=transpose]; +"499 linear_16" [id=499, type=linear]; +"500 unflatten_4" [id=500, type=unflatten]; +"501 unsqueeze_4" [id=501, type=unsqueeze]; +"502 transpose_25" [id=502, type=transpose]; +"503 squeeze_4" [id=503, type=squeeze]; +"504 contiguous_4" [id=504, type=contiguous]; +"505 quantize_per_tensor_default_22" [id=505, type=quantize_per_tensor]; +"506 dequantize_per_tensor_default_71" [id=506, type=dequantize_per_tensor]; +"507 dequantize_per_tensor_default_70" [id=507, type=dequantize_per_tensor]; +"508 select_12" [id=508, type=select]; +"509 select_13" [id=509, type=select]; +"510 select_14" [id=510, type=select]; +"511 view_32" [id=511, type=view]; +"512 transpose_26" [id=512, type=transpose]; +"513 view_33" [id=513, type=view]; +"514 transpose_27" [id=514, type=transpose]; +"515 view_34" [id=515, type=view]; +"516 transpose_28" [id=516, type=transpose]; +"517 view_35" [id=517, type=view]; +"518 view_36" [id=518, type=view]; +"519 view_37" [id=519, type=view]; +"520 scaled_dot_product_attention_4" [id=520, type=scaled_dot_product_attention]; +"521 quantize_per_tensor_default_23" [id=521, type=quantize_per_tensor]; +"522 dequantize_per_tensor_default_23" [id=522, type=dequantize_per_tensor]; +"523 permute_5" [id=523, type=permute]; +"524 view_38" [id=524, type=view]; +"525 linear_17" [id=525, type=linear]; +"526 view_39" [id=526, type=view]; +"527 transpose_29" [id=527, type=transpose]; +"528 dropout_13" [id=528, type=dropout]; +"529 add_9" [id=529, type=add]; +"530 layer_norm_9" [id=530, type=layer_norm]; +"531 quantize_per_tensor_default_24" [id=531, type=quantize_per_tensor]; +"532 dequantize_per_tensor_default_24" [id=532, type=dequantize_per_tensor]; +"533 linear_18" [id=533, type=linear]; +"534 gelu_4" [id=534, type=gelu]; +"535 quantize_per_tensor_default_25" [id=535, type=quantize_per_tensor]; +"536 dequantize_per_tensor_default_25" [id=536, type=dequantize_per_tensor]; +"537 dropout_14" [id=537, type=dropout]; +"538 linear_19" [id=538, type=linear]; +"539 dropout_15" [id=539, type=dropout]; +"540 add_10" [id=540, type=add]; +"541 layer_norm_10" [id=541, type=layer_norm]; +"542 quantize_per_tensor_default_26" [id=542, type=quantize_per_tensor]; +"543 dequantize_per_tensor_default_26" [id=543, type=dequantize_per_tensor]; +"544 transpose_30" [id=544, type=transpose]; +"545 linear_20" [id=545, type=linear]; +"546 unflatten_5" [id=546, type=unflatten]; +"547 unsqueeze_5" [id=547, type=unsqueeze]; +"548 transpose_31" [id=548, type=transpose]; +"549 squeeze_5" [id=549, type=squeeze]; +"550 contiguous_5" [id=550, type=contiguous]; +"551 quantize_per_tensor_default_27" [id=551, type=quantize_per_tensor]; +"552 dequantize_per_tensor_default_73" [id=552, type=dequantize_per_tensor]; +"553 dequantize_per_tensor_default_72" [id=553, type=dequantize_per_tensor]; +"554 select_15" [id=554, type=select]; +"555 select_16" [id=555, type=select]; +"556 select_17" [id=556, type=select]; +"557 view_40" [id=557, type=view]; +"558 transpose_32" [id=558, type=transpose]; +"559 view_41" [id=559, type=view]; +"560 transpose_33" [id=560, type=transpose]; +"561 view_42" [id=561, type=view]; +"562 transpose_34" [id=562, type=transpose]; +"563 view_43" [id=563, type=view]; +"564 view_44" [id=564, type=view]; +"565 view_45" [id=565, type=view]; +"566 scaled_dot_product_attention_5" [id=566, type=scaled_dot_product_attention]; +"567 quantize_per_tensor_default_28" [id=567, type=quantize_per_tensor]; +"568 dequantize_per_tensor_default_28" [id=568, type=dequantize_per_tensor]; +"569 permute_6" [id=569, type=permute]; +"570 view_46" [id=570, type=view]; +"571 linear_21" [id=571, type=linear]; +"572 view_47" [id=572, type=view]; +"573 transpose_35" [id=573, type=transpose]; +"574 dropout_16" [id=574, type=dropout]; +"575 add_11" [id=575, type=add]; +"576 layer_norm_11" [id=576, type=layer_norm]; +"577 quantize_per_tensor_default_29" [id=577, type=quantize_per_tensor]; +"578 dequantize_per_tensor_default_29" [id=578, type=dequantize_per_tensor]; +"579 linear_22" [id=579, type=linear]; +"580 gelu_5" [id=580, type=gelu]; +"581 quantize_per_tensor_default_30" [id=581, type=quantize_per_tensor]; +"582 dequantize_per_tensor_default_30" [id=582, type=dequantize_per_tensor]; +"583 dropout_17" [id=583, type=dropout]; +"584 linear_23" [id=584, type=linear]; +"585 dropout_18" [id=585, type=dropout]; +"586 add_12" [id=586, type=add]; +"587 layer_norm_12" [id=587, type=layer_norm]; +"588 quantize_per_tensor_default_31" [id=588, type=quantize_per_tensor]; +"589 dequantize_per_tensor_default_31" [id=589, type=dequantize_per_tensor]; +"590 transpose_36" [id=590, type=transpose]; +"591 linear_24" [id=591, type=linear]; +"592 unflatten_6" [id=592, type=unflatten]; +"593 unsqueeze_6" [id=593, type=unsqueeze]; +"594 transpose_37" [id=594, type=transpose]; +"595 squeeze_6" [id=595, type=squeeze]; +"596 contiguous_6" [id=596, type=contiguous]; +"597 quantize_per_tensor_default_32" [id=597, type=quantize_per_tensor]; +"598 dequantize_per_tensor_default_75" [id=598, type=dequantize_per_tensor]; +"599 dequantize_per_tensor_default_74" [id=599, type=dequantize_per_tensor]; +"600 select_18" [id=600, type=select]; +"601 select_19" [id=601, type=select]; +"602 select_20" [id=602, type=select]; +"603 view_48" [id=603, type=view]; +"604 transpose_38" [id=604, type=transpose]; +"605 view_49" [id=605, type=view]; +"606 transpose_39" [id=606, type=transpose]; +"607 view_50" [id=607, type=view]; +"608 transpose_40" [id=608, type=transpose]; +"609 view_51" [id=609, type=view]; +"610 view_52" [id=610, type=view]; +"611 view_53" [id=611, type=view]; +"612 scaled_dot_product_attention_6" [id=612, type=scaled_dot_product_attention]; +"613 quantize_per_tensor_default_33" [id=613, type=quantize_per_tensor]; +"614 dequantize_per_tensor_default_33" [id=614, type=dequantize_per_tensor]; +"615 permute_7" [id=615, type=permute]; +"616 view_54" [id=616, type=view]; +"617 linear_25" [id=617, type=linear]; +"618 view_55" [id=618, type=view]; +"619 transpose_41" [id=619, type=transpose]; +"620 dropout_19" [id=620, type=dropout]; +"621 add_13" [id=621, type=add]; +"622 layer_norm_13" [id=622, type=layer_norm]; +"623 quantize_per_tensor_default_34" [id=623, type=quantize_per_tensor]; +"624 dequantize_per_tensor_default_34" [id=624, type=dequantize_per_tensor]; +"625 linear_26" [id=625, type=linear]; +"626 gelu_6" [id=626, type=gelu]; +"627 quantize_per_tensor_default_35" [id=627, type=quantize_per_tensor]; +"628 dequantize_per_tensor_default_35" [id=628, type=dequantize_per_tensor]; +"629 dropout_20" [id=629, type=dropout]; +"630 linear_27" [id=630, type=linear]; +"631 dropout_21" [id=631, type=dropout]; +"632 add_14" [id=632, type=add]; +"633 layer_norm_14" [id=633, type=layer_norm]; +"634 quantize_per_tensor_default_36" [id=634, type=quantize_per_tensor]; +"635 dequantize_per_tensor_default_36" [id=635, type=dequantize_per_tensor]; +"636 transpose_42" [id=636, type=transpose]; +"637 linear_28" [id=637, type=linear]; +"638 unflatten_7" [id=638, type=unflatten]; +"639 unsqueeze_7" [id=639, type=unsqueeze]; +"640 transpose_43" [id=640, type=transpose]; +"641 squeeze_7" [id=641, type=squeeze]; +"642 contiguous_7" [id=642, type=contiguous]; +"643 quantize_per_tensor_default_37" [id=643, type=quantize_per_tensor]; +"644 dequantize_per_tensor_default_77" [id=644, type=dequantize_per_tensor]; +"645 dequantize_per_tensor_default_76" [id=645, type=dequantize_per_tensor]; +"646 select_21" [id=646, type=select]; +"647 select_22" [id=647, type=select]; +"648 select_23" [id=648, type=select]; +"649 view_56" [id=649, type=view]; +"650 transpose_44" [id=650, type=transpose]; +"651 view_57" [id=651, type=view]; +"652 transpose_45" [id=652, type=transpose]; +"653 view_58" [id=653, type=view]; +"654 transpose_46" [id=654, type=transpose]; +"655 view_59" [id=655, type=view]; +"656 view_60" [id=656, type=view]; +"657 view_61" [id=657, type=view]; +"658 scaled_dot_product_attention_7" [id=658, type=scaled_dot_product_attention]; +"659 quantize_per_tensor_default_38" [id=659, type=quantize_per_tensor]; +"660 dequantize_per_tensor_default_38" [id=660, type=dequantize_per_tensor]; +"661 permute_8" [id=661, type=permute]; +"662 view_62" [id=662, type=view]; +"663 linear_29" [id=663, type=linear]; +"664 view_63" [id=664, type=view]; +"665 transpose_47" [id=665, type=transpose]; +"666 dropout_22" [id=666, type=dropout]; +"667 add_15" [id=667, type=add]; +"668 layer_norm_15" [id=668, type=layer_norm]; +"669 quantize_per_tensor_default_39" [id=669, type=quantize_per_tensor]; +"670 dequantize_per_tensor_default_39" [id=670, type=dequantize_per_tensor]; +"671 linear_30" [id=671, type=linear]; +"672 gelu_7" [id=672, type=gelu]; +"673 quantize_per_tensor_default_40" [id=673, type=quantize_per_tensor]; +"674 dequantize_per_tensor_default_40" [id=674, type=dequantize_per_tensor]; +"675 dropout_23" [id=675, type=dropout]; +"676 linear_31" [id=676, type=linear]; +"677 dropout_24" [id=677, type=dropout]; +"678 add_16" [id=678, type=add]; +"679 layer_norm_16" [id=679, type=layer_norm]; +"680 quantize_per_tensor_default_41" [id=680, type=quantize_per_tensor]; +"681 dequantize_per_tensor_default_41" [id=681, type=dequantize_per_tensor]; +"682 transpose_48" [id=682, type=transpose]; +"683 linear_32" [id=683, type=linear]; +"684 unflatten_8" [id=684, type=unflatten]; +"685 unsqueeze_8" [id=685, type=unsqueeze]; +"686 transpose_49" [id=686, type=transpose]; +"687 squeeze_8" [id=687, type=squeeze]; +"688 contiguous_8" [id=688, type=contiguous]; +"689 quantize_per_tensor_default_42" [id=689, type=quantize_per_tensor]; +"690 dequantize_per_tensor_default_79" [id=690, type=dequantize_per_tensor]; +"691 dequantize_per_tensor_default_78" [id=691, type=dequantize_per_tensor]; +"692 select_24" [id=692, type=select]; +"693 select_25" [id=693, type=select]; +"694 select_26" [id=694, type=select]; +"695 view_64" [id=695, type=view]; +"696 transpose_50" [id=696, type=transpose]; +"697 view_65" [id=697, type=view]; +"698 transpose_51" [id=698, type=transpose]; +"699 view_66" [id=699, type=view]; +"700 transpose_52" [id=700, type=transpose]; +"701 view_67" [id=701, type=view]; +"702 view_68" [id=702, type=view]; +"703 view_69" [id=703, type=view]; +"704 scaled_dot_product_attention_8" [id=704, type=scaled_dot_product_attention]; +"705 quantize_per_tensor_default_43" [id=705, type=quantize_per_tensor]; +"706 dequantize_per_tensor_default_43" [id=706, type=dequantize_per_tensor]; +"707 permute_9" [id=707, type=permute]; +"708 view_70" [id=708, type=view]; +"709 linear_33" [id=709, type=linear]; +"710 view_71" [id=710, type=view]; +"711 transpose_53" [id=711, type=transpose]; +"712 dropout_25" [id=712, type=dropout]; +"713 add_17" [id=713, type=add]; +"714 layer_norm_17" [id=714, type=layer_norm]; +"715 quantize_per_tensor_default_44" [id=715, type=quantize_per_tensor]; +"716 dequantize_per_tensor_default_44" [id=716, type=dequantize_per_tensor]; +"717 linear_34" [id=717, type=linear]; +"718 gelu_8" [id=718, type=gelu]; +"719 quantize_per_tensor_default_45" [id=719, type=quantize_per_tensor]; +"720 dequantize_per_tensor_default_45" [id=720, type=dequantize_per_tensor]; +"721 dropout_26" [id=721, type=dropout]; +"722 linear_35" [id=722, type=linear]; +"723 dropout_27" [id=723, type=dropout]; +"724 add_18" [id=724, type=add]; +"725 layer_norm_18" [id=725, type=layer_norm]; +"726 quantize_per_tensor_default_46" [id=726, type=quantize_per_tensor]; +"727 dequantize_per_tensor_default_46" [id=727, type=dequantize_per_tensor]; +"728 transpose_54" [id=728, type=transpose]; +"729 linear_36" [id=729, type=linear]; +"730 unflatten_9" [id=730, type=unflatten]; +"731 unsqueeze_9" [id=731, type=unsqueeze]; +"732 transpose_55" [id=732, type=transpose]; +"733 squeeze_9" [id=733, type=squeeze]; +"734 contiguous_9" [id=734, type=contiguous]; +"735 quantize_per_tensor_default_47" [id=735, type=quantize_per_tensor]; +"736 dequantize_per_tensor_default_81" [id=736, type=dequantize_per_tensor]; +"737 dequantize_per_tensor_default_80" [id=737, type=dequantize_per_tensor]; +"738 select_27" [id=738, type=select]; +"739 select_28" [id=739, type=select]; +"740 select_29" [id=740, type=select]; +"741 view_72" [id=741, type=view]; +"742 transpose_56" [id=742, type=transpose]; +"743 view_73" [id=743, type=view]; +"744 transpose_57" [id=744, type=transpose]; +"745 view_74" [id=745, type=view]; +"746 transpose_58" [id=746, type=transpose]; +"747 view_75" [id=747, type=view]; +"748 view_76" [id=748, type=view]; +"749 view_77" [id=749, type=view]; +"750 scaled_dot_product_attention_9" [id=750, type=scaled_dot_product_attention]; +"751 quantize_per_tensor_default_48" [id=751, type=quantize_per_tensor]; +"752 dequantize_per_tensor_default_48" [id=752, type=dequantize_per_tensor]; +"753 permute_10" [id=753, type=permute]; +"754 view_78" [id=754, type=view]; +"755 linear_37" [id=755, type=linear]; +"756 view_79" [id=756, type=view]; +"757 transpose_59" [id=757, type=transpose]; +"758 dropout_28" [id=758, type=dropout]; +"759 add_19" [id=759, type=add]; +"760 layer_norm_19" [id=760, type=layer_norm]; +"761 quantize_per_tensor_default_49" [id=761, type=quantize_per_tensor]; +"762 dequantize_per_tensor_default_49" [id=762, type=dequantize_per_tensor]; +"763 linear_38" [id=763, type=linear]; +"764 gelu_9" [id=764, type=gelu]; +"765 quantize_per_tensor_default_50" [id=765, type=quantize_per_tensor]; +"766 dequantize_per_tensor_default_50" [id=766, type=dequantize_per_tensor]; +"767 dropout_29" [id=767, type=dropout]; +"768 linear_39" [id=768, type=linear]; +"769 dropout_30" [id=769, type=dropout]; +"770 add_20" [id=770, type=add]; +"771 layer_norm_20" [id=771, type=layer_norm]; +"772 quantize_per_tensor_default_51" [id=772, type=quantize_per_tensor]; +"773 dequantize_per_tensor_default_51" [id=773, type=dequantize_per_tensor]; +"774 transpose_60" [id=774, type=transpose]; +"775 linear_40" [id=775, type=linear]; +"776 unflatten_10" [id=776, type=unflatten]; +"777 unsqueeze_10" [id=777, type=unsqueeze]; +"778 transpose_61" [id=778, type=transpose]; +"779 squeeze_10" [id=779, type=squeeze]; +"780 contiguous_10" [id=780, type=contiguous]; +"781 quantize_per_tensor_default_52" [id=781, type=quantize_per_tensor]; +"782 dequantize_per_tensor_default_83" [id=782, type=dequantize_per_tensor]; +"783 dequantize_per_tensor_default_82" [id=783, type=dequantize_per_tensor]; +"784 select_30" [id=784, type=select]; +"785 select_31" [id=785, type=select]; +"786 select_32" [id=786, type=select]; +"787 view_80" [id=787, type=view]; +"788 transpose_62" [id=788, type=transpose]; +"789 view_81" [id=789, type=view]; +"790 transpose_63" [id=790, type=transpose]; +"791 view_82" [id=791, type=view]; +"792 transpose_64" [id=792, type=transpose]; +"793 view_83" [id=793, type=view]; +"794 view_84" [id=794, type=view]; +"795 view_85" [id=795, type=view]; +"796 scaled_dot_product_attention_10" [id=796, type=scaled_dot_product_attention]; +"797 quantize_per_tensor_default_53" [id=797, type=quantize_per_tensor]; +"798 dequantize_per_tensor_default_53" [id=798, type=dequantize_per_tensor]; +"799 permute_11" [id=799, type=permute]; +"800 view_86" [id=800, type=view]; +"801 linear_41" [id=801, type=linear]; +"802 view_87" [id=802, type=view]; +"803 transpose_65" [id=803, type=transpose]; +"804 dropout_31" [id=804, type=dropout]; +"805 add_21" [id=805, type=add]; +"806 layer_norm_21" [id=806, type=layer_norm]; +"807 quantize_per_tensor_default_54" [id=807, type=quantize_per_tensor]; +"808 dequantize_per_tensor_default_54" [id=808, type=dequantize_per_tensor]; +"809 linear_42" [id=809, type=linear]; +"810 gelu_10" [id=810, type=gelu]; +"811 quantize_per_tensor_default_55" [id=811, type=quantize_per_tensor]; +"812 dequantize_per_tensor_default_55" [id=812, type=dequantize_per_tensor]; +"813 dropout_32" [id=813, type=dropout]; +"814 linear_43" [id=814, type=linear]; +"815 dropout_33" [id=815, type=dropout]; +"816 add_22" [id=816, type=add]; +"817 layer_norm_22" [id=817, type=layer_norm]; +"818 quantize_per_tensor_default_56" [id=818, type=quantize_per_tensor]; +"819 dequantize_per_tensor_default_56" [id=819, type=dequantize_per_tensor]; +"820 transpose_66" [id=820, type=transpose]; +"821 linear_44" [id=821, type=linear]; +"822 unflatten_11" [id=822, type=unflatten]; +"823 unsqueeze_11" [id=823, type=unsqueeze]; +"824 transpose_67" [id=824, type=transpose]; +"825 squeeze_11" [id=825, type=squeeze]; +"826 contiguous_11" [id=826, type=contiguous]; +"827 quantize_per_tensor_default_57" [id=827, type=quantize_per_tensor]; +"828 dequantize_per_tensor_default_85" [id=828, type=dequantize_per_tensor]; +"829 dequantize_per_tensor_default_84" [id=829, type=dequantize_per_tensor]; +"830 select_33" [id=830, type=select]; +"831 select_34" [id=831, type=select]; +"832 select_35" [id=832, type=select]; +"833 view_88" [id=833, type=view]; +"834 transpose_68" [id=834, type=transpose]; +"835 view_89" [id=835, type=view]; +"836 transpose_69" [id=836, type=transpose]; +"837 view_90" [id=837, type=view]; +"838 transpose_70" [id=838, type=transpose]; +"839 view_91" [id=839, type=view]; +"840 view_92" [id=840, type=view]; +"841 view_93" [id=841, type=view]; +"842 scaled_dot_product_attention_11" [id=842, type=scaled_dot_product_attention]; +"843 quantize_per_tensor_default_58" [id=843, type=quantize_per_tensor]; +"844 dequantize_per_tensor_default_58" [id=844, type=dequantize_per_tensor]; +"845 permute_12" [id=845, type=permute]; +"846 view_94" [id=846, type=view]; +"847 linear_45" [id=847, type=linear]; +"848 view_95" [id=848, type=view]; +"849 transpose_71" [id=849, type=transpose]; +"850 dropout_34" [id=850, type=dropout]; +"851 add_23" [id=851, type=add]; +"852 layer_norm_23" [id=852, type=layer_norm]; +"853 quantize_per_tensor_default_59" [id=853, type=quantize_per_tensor]; +"854 dequantize_per_tensor_default_59" [id=854, type=dequantize_per_tensor]; +"855 linear_46" [id=855, type=linear]; +"856 gelu_11" [id=856, type=gelu]; +"857 quantize_per_tensor_default_60" [id=857, type=quantize_per_tensor]; +"858 dequantize_per_tensor_default_60" [id=858, type=dequantize_per_tensor]; +"859 dropout_35" [id=859, type=dropout]; +"860 linear_47" [id=860, type=linear]; +"861 dropout_36" [id=861, type=dropout]; +"862 add_24" [id=862, type=add]; +"863 layer_norm_24" [id=863, type=layer_norm]; +"864 quantize_per_tensor_default_61" [id=864, type=quantize_per_tensor]; +"865 dequantize_per_tensor_default_61" [id=865, type=dequantize_per_tensor]; +"866 slice_1" [id=866, type=slice]; +"867 select_36" [id=867, type=select]; +"868 linear_48" [id=868, type=linear]; +"869 output_1" [id=869, type=output]; +"0 encoder_pos_embedding" -> "309 add" [label="(1, 197, 768)", style=solid]; +"1 _scale_0" -> "4 dequantize_per_channel_default" [label="(768,)", style=solid]; +"2 _zero_point_0" -> "4 dequantize_per_channel_default" [label="(768,)", style=solid]; +"3 _frozen_param1" -> "4 dequantize_per_channel_default" [label="(768, 3, 16, 16)", style=solid]; +"4 dequantize_per_channel_default" -> "304 conv2d" [label=None, style=solid]; +"5 conv_proj_bias" -> "304 conv2d" [label="(768,)", style=solid]; +"6 encoder_layers_encoder_layer_0_ln_1_weight" -> "311 layer_norm" [label="(768,)", style=solid]; +"7 encoder_layers_encoder_layer_0_ln_1_bias" -> "311 layer_norm" [label="(768,)", style=solid]; +"8 _scale_1" -> "11 dequantize_per_channel_default_1" [label="(2304,)", style=solid]; +"9 _zero_point_1" -> "11 dequantize_per_channel_default_1" [label="(2304,)", style=solid]; +"10 _frozen_param2" -> "11 dequantize_per_channel_default_1" [label="(2304, 768)", style=solid]; +"11 dequantize_per_channel_default_1" -> "315 linear" [label=None, style=solid]; +"12 encoder_layers_encoder_layer_0_self_attention_in_proj_bias" -> "315 linear" [label="(2304,)", style=solid]; +"13 _scale_2" -> "16 dequantize_per_channel_default_2" [label="(768,)", style=solid]; +"14 _zero_point_2" -> "16 dequantize_per_channel_default_2" [label="(768,)", style=solid]; +"15 _frozen_param3" -> "16 dequantize_per_channel_default_2" [label="(768, 768)", style=solid]; +"16 dequantize_per_channel_default_2" -> "341 linear_1" [label=None, style=solid]; +"17 encoder_layers_encoder_layer_0_self_attention_out_proj_bias" -> "341 linear_1" [label="(768,)", style=solid]; +"18 encoder_layers_encoder_layer_0_ln_2_weight" -> "346 layer_norm_1" [label="(768,)", style=solid]; +"19 encoder_layers_encoder_layer_0_ln_2_bias" -> "346 layer_norm_1" [label="(768,)", style=solid]; +"20 _scale_3" -> "23 dequantize_per_channel_default_3" [label="(3072,)", style=solid]; +"21 _zero_point_3" -> "23 dequantize_per_channel_default_3" [label="(3072,)", style=solid]; +"22 _frozen_param4" -> "23 dequantize_per_channel_default_3" [label="(3072, 768)", style=solid]; +"23 dequantize_per_channel_default_3" -> "349 linear_2" [label=None, style=solid]; +"24 encoder_layers_encoder_layer_0_mlp_0_bias" -> "349 linear_2" [label="(3072,)", style=solid]; +"25 _scale_4" -> "28 dequantize_per_channel_default_4" [label="(768,)", style=solid]; +"26 _zero_point_4" -> "28 dequantize_per_channel_default_4" [label="(768,)", style=solid]; +"27 _frozen_param5" -> "28 dequantize_per_channel_default_4" [label="(768, 3072)", style=solid]; +"28 dequantize_per_channel_default_4" -> "354 linear_3" [label=None, style=solid]; +"29 encoder_layers_encoder_layer_0_mlp_3_bias" -> "354 linear_3" [label="(768,)", style=solid]; +"30 encoder_layers_encoder_layer_1_ln_1_weight" -> "357 layer_norm_2" [label="(768,)", style=solid]; +"31 encoder_layers_encoder_layer_1_ln_1_bias" -> "357 layer_norm_2" [label="(768,)", style=solid]; +"32 _scale_5" -> "35 dequantize_per_channel_default_5" [label="(2304,)", style=solid]; +"33 _zero_point_5" -> "35 dequantize_per_channel_default_5" [label="(2304,)", style=solid]; +"34 _frozen_param6" -> "35 dequantize_per_channel_default_5" [label="(2304, 768)", style=solid]; +"35 dequantize_per_channel_default_5" -> "361 linear_4" [label=None, style=solid]; +"36 encoder_layers_encoder_layer_1_self_attention_in_proj_bias" -> "361 linear_4" [label="(2304,)", style=solid]; +"37 _scale_6" -> "40 dequantize_per_channel_default_6" [label="(768,)", style=solid]; +"38 _zero_point_6" -> "40 dequantize_per_channel_default_6" [label="(768,)", style=solid]; +"39 _frozen_param7" -> "40 dequantize_per_channel_default_6" [label="(768, 768)", style=solid]; +"40 dequantize_per_channel_default_6" -> "387 linear_5" [label=None, style=solid]; +"41 encoder_layers_encoder_layer_1_self_attention_out_proj_bias" -> "387 linear_5" [label="(768,)", style=solid]; +"42 encoder_layers_encoder_layer_1_ln_2_weight" -> "392 layer_norm_3" [label="(768,)", style=solid]; +"43 encoder_layers_encoder_layer_1_ln_2_bias" -> "392 layer_norm_3" [label="(768,)", style=solid]; +"44 _scale_7" -> "47 dequantize_per_channel_default_7" [label="(3072,)", style=solid]; +"45 _zero_point_7" -> "47 dequantize_per_channel_default_7" [label="(3072,)", style=solid]; +"46 _frozen_param8" -> "47 dequantize_per_channel_default_7" [label="(3072, 768)", style=solid]; +"47 dequantize_per_channel_default_7" -> "395 linear_6" [label=None, style=solid]; +"48 encoder_layers_encoder_layer_1_mlp_0_bias" -> "395 linear_6" [label="(3072,)", style=solid]; +"49 _scale_8" -> "52 dequantize_per_channel_default_8" [label="(768,)", style=solid]; +"50 _zero_point_8" -> "52 dequantize_per_channel_default_8" [label="(768,)", style=solid]; +"51 _frozen_param9" -> "52 dequantize_per_channel_default_8" [label="(768, 3072)", style=solid]; +"52 dequantize_per_channel_default_8" -> "400 linear_7" [label=None, style=solid]; +"53 encoder_layers_encoder_layer_1_mlp_3_bias" -> "400 linear_7" [label="(768,)", style=solid]; +"54 encoder_layers_encoder_layer_2_ln_1_weight" -> "403 layer_norm_4" [label="(768,)", style=solid]; +"55 encoder_layers_encoder_layer_2_ln_1_bias" -> "403 layer_norm_4" [label="(768,)", style=solid]; +"56 _scale_9" -> "59 dequantize_per_channel_default_9" [label="(2304,)", style=solid]; +"57 _zero_point_9" -> "59 dequantize_per_channel_default_9" [label="(2304,)", style=solid]; +"58 _frozen_param10" -> "59 dequantize_per_channel_default_9" [label="(2304, 768)", style=solid]; +"59 dequantize_per_channel_default_9" -> "407 linear_8" [label=None, style=solid]; +"60 encoder_layers_encoder_layer_2_self_attention_in_proj_bias" -> "407 linear_8" [label="(2304,)", style=solid]; +"61 _scale_10" -> "64 dequantize_per_channel_default_10" [label="(768,)", style=solid]; +"62 _zero_point_10" -> "64 dequantize_per_channel_default_10" [label="(768,)", style=solid]; +"63 _frozen_param11" -> "64 dequantize_per_channel_default_10" [label="(768, 768)", style=solid]; +"64 dequantize_per_channel_default_10" -> "433 linear_9" [label=None, style=solid]; +"65 encoder_layers_encoder_layer_2_self_attention_out_proj_bias" -> "433 linear_9" [label="(768,)", style=solid]; +"66 encoder_layers_encoder_layer_2_ln_2_weight" -> "438 layer_norm_5" [label="(768,)", style=solid]; +"67 encoder_layers_encoder_layer_2_ln_2_bias" -> "438 layer_norm_5" [label="(768,)", style=solid]; +"68 _scale_11" -> "71 dequantize_per_channel_default_11" [label="(3072,)", style=solid]; +"69 _zero_point_11" -> "71 dequantize_per_channel_default_11" [label="(3072,)", style=solid]; +"70 _frozen_param12" -> "71 dequantize_per_channel_default_11" [label="(3072, 768)", style=solid]; +"71 dequantize_per_channel_default_11" -> "441 linear_10" [label=None, style=solid]; +"72 encoder_layers_encoder_layer_2_mlp_0_bias" -> "441 linear_10" [label="(3072,)", style=solid]; +"73 _scale_12" -> "76 dequantize_per_channel_default_12" [label="(768,)", style=solid]; +"74 _zero_point_12" -> "76 dequantize_per_channel_default_12" [label="(768,)", style=solid]; +"75 _frozen_param13" -> "76 dequantize_per_channel_default_12" [label="(768, 3072)", style=solid]; +"76 dequantize_per_channel_default_12" -> "446 linear_11" [label=None, style=solid]; +"77 encoder_layers_encoder_layer_2_mlp_3_bias" -> "446 linear_11" [label="(768,)", style=solid]; +"78 encoder_layers_encoder_layer_3_ln_1_weight" -> "449 layer_norm_6" [label="(768,)", style=solid]; +"79 encoder_layers_encoder_layer_3_ln_1_bias" -> "449 layer_norm_6" [label="(768,)", style=solid]; +"80 _scale_13" -> "83 dequantize_per_channel_default_13" [label="(2304,)", style=solid]; +"81 _zero_point_13" -> "83 dequantize_per_channel_default_13" [label="(2304,)", style=solid]; +"82 _frozen_param14" -> "83 dequantize_per_channel_default_13" [label="(2304, 768)", style=solid]; +"83 dequantize_per_channel_default_13" -> "453 linear_12" [label=None, style=solid]; +"84 encoder_layers_encoder_layer_3_self_attention_in_proj_bias" -> "453 linear_12" [label="(2304,)", style=solid]; +"85 _scale_14" -> "88 dequantize_per_channel_default_14" [label="(768,)", style=solid]; +"86 _zero_point_14" -> "88 dequantize_per_channel_default_14" [label="(768,)", style=solid]; +"87 _frozen_param15" -> "88 dequantize_per_channel_default_14" [label="(768, 768)", style=solid]; +"88 dequantize_per_channel_default_14" -> "479 linear_13" [label=None, style=solid]; +"89 encoder_layers_encoder_layer_3_self_attention_out_proj_bias" -> "479 linear_13" [label="(768,)", style=solid]; +"90 encoder_layers_encoder_layer_3_ln_2_weight" -> "484 layer_norm_7" [label="(768,)", style=solid]; +"91 encoder_layers_encoder_layer_3_ln_2_bias" -> "484 layer_norm_7" [label="(768,)", style=solid]; +"92 _scale_15" -> "95 dequantize_per_channel_default_15" [label="(3072,)", style=solid]; +"93 _zero_point_15" -> "95 dequantize_per_channel_default_15" [label="(3072,)", style=solid]; +"94 _frozen_param16" -> "95 dequantize_per_channel_default_15" [label="(3072, 768)", style=solid]; +"95 dequantize_per_channel_default_15" -> "487 linear_14" [label=None, style=solid]; +"96 encoder_layers_encoder_layer_3_mlp_0_bias" -> "487 linear_14" [label="(3072,)", style=solid]; +"97 _scale_16" -> "100 dequantize_per_channel_default_16" [label="(768,)", style=solid]; +"98 _zero_point_16" -> "100 dequantize_per_channel_default_16" [label="(768,)", style=solid]; +"99 _frozen_param17" -> "100 dequantize_per_channel_default_16" [label="(768, 3072)", style=solid]; +"100 dequantize_per_channel_default_16" -> "492 linear_15" [label=None, style=solid]; +"101 encoder_layers_encoder_layer_3_mlp_3_bias" -> "492 linear_15" [label="(768,)", style=solid]; +"102 encoder_layers_encoder_layer_4_ln_1_weight" -> "495 layer_norm_8" [label="(768,)", style=solid]; +"103 encoder_layers_encoder_layer_4_ln_1_bias" -> "495 layer_norm_8" [label="(768,)", style=solid]; +"104 _scale_17" -> "107 dequantize_per_channel_default_17" [label="(2304,)", style=solid]; +"105 _zero_point_17" -> "107 dequantize_per_channel_default_17" [label="(2304,)", style=solid]; +"106 _frozen_param18" -> "107 dequantize_per_channel_default_17" [label="(2304, 768)", style=solid]; +"107 dequantize_per_channel_default_17" -> "499 linear_16" [label=None, style=solid]; +"108 encoder_layers_encoder_layer_4_self_attention_in_proj_bias" -> "499 linear_16" [label="(2304,)", style=solid]; +"109 _scale_18" -> "112 dequantize_per_channel_default_18" [label="(768,)", style=solid]; +"110 _zero_point_18" -> "112 dequantize_per_channel_default_18" [label="(768,)", style=solid]; +"111 _frozen_param19" -> "112 dequantize_per_channel_default_18" [label="(768, 768)", style=solid]; +"112 dequantize_per_channel_default_18" -> "525 linear_17" [label=None, style=solid]; +"113 encoder_layers_encoder_layer_4_self_attention_out_proj_bias" -> "525 linear_17" [label="(768,)", style=solid]; +"114 encoder_layers_encoder_layer_4_ln_2_weight" -> "530 layer_norm_9" [label="(768,)", style=solid]; +"115 encoder_layers_encoder_layer_4_ln_2_bias" -> "530 layer_norm_9" [label="(768,)", style=solid]; +"116 _scale_19" -> "119 dequantize_per_channel_default_19" [label="(3072,)", style=solid]; +"117 _zero_point_19" -> "119 dequantize_per_channel_default_19" [label="(3072,)", style=solid]; +"118 _frozen_param20" -> "119 dequantize_per_channel_default_19" [label="(3072, 768)", style=solid]; +"119 dequantize_per_channel_default_19" -> "533 linear_18" [label=None, style=solid]; +"120 encoder_layers_encoder_layer_4_mlp_0_bias" -> "533 linear_18" [label="(3072,)", style=solid]; +"121 _scale_20" -> "124 dequantize_per_channel_default_20" [label="(768,)", style=solid]; +"122 _zero_point_20" -> "124 dequantize_per_channel_default_20" [label="(768,)", style=solid]; +"123 _frozen_param21" -> "124 dequantize_per_channel_default_20" [label="(768, 3072)", style=solid]; +"124 dequantize_per_channel_default_20" -> "538 linear_19" [label=None, style=solid]; +"125 encoder_layers_encoder_layer_4_mlp_3_bias" -> "538 linear_19" [label="(768,)", style=solid]; +"126 encoder_layers_encoder_layer_5_ln_1_weight" -> "541 layer_norm_10" [label="(768,)", style=solid]; +"127 encoder_layers_encoder_layer_5_ln_1_bias" -> "541 layer_norm_10" [label="(768,)", style=solid]; +"128 _scale_21" -> "131 dequantize_per_channel_default_21" [label="(2304,)", style=solid]; +"129 _zero_point_21" -> "131 dequantize_per_channel_default_21" [label="(2304,)", style=solid]; +"130 _frozen_param22" -> "131 dequantize_per_channel_default_21" [label="(2304, 768)", style=solid]; +"131 dequantize_per_channel_default_21" -> "545 linear_20" [label=None, style=solid]; +"132 encoder_layers_encoder_layer_5_self_attention_in_proj_bias" -> "545 linear_20" [label="(2304,)", style=solid]; +"133 _scale_22" -> "136 dequantize_per_channel_default_22" [label="(768,)", style=solid]; +"134 _zero_point_22" -> "136 dequantize_per_channel_default_22" [label="(768,)", style=solid]; +"135 _frozen_param23" -> "136 dequantize_per_channel_default_22" [label="(768, 768)", style=solid]; +"136 dequantize_per_channel_default_22" -> "571 linear_21" [label=None, style=solid]; +"137 encoder_layers_encoder_layer_5_self_attention_out_proj_bias" -> "571 linear_21" [label="(768,)", style=solid]; +"138 encoder_layers_encoder_layer_5_ln_2_weight" -> "576 layer_norm_11" [label="(768,)", style=solid]; +"139 encoder_layers_encoder_layer_5_ln_2_bias" -> "576 layer_norm_11" [label="(768,)", style=solid]; +"140 _scale_23" -> "143 dequantize_per_channel_default_23" [label="(3072,)", style=solid]; +"141 _zero_point_23" -> "143 dequantize_per_channel_default_23" [label="(3072,)", style=solid]; +"142 _frozen_param24" -> "143 dequantize_per_channel_default_23" [label="(3072, 768)", style=solid]; +"143 dequantize_per_channel_default_23" -> "579 linear_22" [label=None, style=solid]; +"144 encoder_layers_encoder_layer_5_mlp_0_bias" -> "579 linear_22" [label="(3072,)", style=solid]; +"145 _scale_24" -> "148 dequantize_per_channel_default_24" [label="(768,)", style=solid]; +"146 _zero_point_24" -> "148 dequantize_per_channel_default_24" [label="(768,)", style=solid]; +"147 _frozen_param25" -> "148 dequantize_per_channel_default_24" [label="(768, 3072)", style=solid]; +"148 dequantize_per_channel_default_24" -> "584 linear_23" [label=None, style=solid]; +"149 encoder_layers_encoder_layer_5_mlp_3_bias" -> "584 linear_23" [label="(768,)", style=solid]; +"150 encoder_layers_encoder_layer_6_ln_1_weight" -> "587 layer_norm_12" [label="(768,)", style=solid]; +"151 encoder_layers_encoder_layer_6_ln_1_bias" -> "587 layer_norm_12" [label="(768,)", style=solid]; +"152 _scale_25" -> "155 dequantize_per_channel_default_25" [label="(2304,)", style=solid]; +"153 _zero_point_25" -> "155 dequantize_per_channel_default_25" [label="(2304,)", style=solid]; +"154 _frozen_param26" -> "155 dequantize_per_channel_default_25" [label="(2304, 768)", style=solid]; +"155 dequantize_per_channel_default_25" -> "591 linear_24" [label=None, style=solid]; +"156 encoder_layers_encoder_layer_6_self_attention_in_proj_bias" -> "591 linear_24" [label="(2304,)", style=solid]; +"157 _scale_26" -> "160 dequantize_per_channel_default_26" [label="(768,)", style=solid]; +"158 _zero_point_26" -> "160 dequantize_per_channel_default_26" [label="(768,)", style=solid]; +"159 _frozen_param27" -> "160 dequantize_per_channel_default_26" [label="(768, 768)", style=solid]; +"160 dequantize_per_channel_default_26" -> "617 linear_25" [label=None, style=solid]; +"161 encoder_layers_encoder_layer_6_self_attention_out_proj_bias" -> "617 linear_25" [label="(768,)", style=solid]; +"162 encoder_layers_encoder_layer_6_ln_2_weight" -> "622 layer_norm_13" [label="(768,)", style=solid]; +"163 encoder_layers_encoder_layer_6_ln_2_bias" -> "622 layer_norm_13" [label="(768,)", style=solid]; +"164 _scale_27" -> "167 dequantize_per_channel_default_27" [label="(3072,)", style=solid]; +"165 _zero_point_27" -> "167 dequantize_per_channel_default_27" [label="(3072,)", style=solid]; +"166 _frozen_param28" -> "167 dequantize_per_channel_default_27" [label="(3072, 768)", style=solid]; +"167 dequantize_per_channel_default_27" -> "625 linear_26" [label=None, style=solid]; +"168 encoder_layers_encoder_layer_6_mlp_0_bias" -> "625 linear_26" [label="(3072,)", style=solid]; +"169 _scale_28" -> "172 dequantize_per_channel_default_28" [label="(768,)", style=solid]; +"170 _zero_point_28" -> "172 dequantize_per_channel_default_28" [label="(768,)", style=solid]; +"171 _frozen_param29" -> "172 dequantize_per_channel_default_28" [label="(768, 3072)", style=solid]; +"172 dequantize_per_channel_default_28" -> "630 linear_27" [label=None, style=solid]; +"173 encoder_layers_encoder_layer_6_mlp_3_bias" -> "630 linear_27" [label="(768,)", style=solid]; +"174 encoder_layers_encoder_layer_7_ln_1_weight" -> "633 layer_norm_14" [label="(768,)", style=solid]; +"175 encoder_layers_encoder_layer_7_ln_1_bias" -> "633 layer_norm_14" [label="(768,)", style=solid]; +"176 _scale_29" -> "179 dequantize_per_channel_default_29" [label="(2304,)", style=solid]; +"177 _zero_point_29" -> "179 dequantize_per_channel_default_29" [label="(2304,)", style=solid]; +"178 _frozen_param30" -> "179 dequantize_per_channel_default_29" [label="(2304, 768)", style=solid]; +"179 dequantize_per_channel_default_29" -> "637 linear_28" [label=None, style=solid]; +"180 encoder_layers_encoder_layer_7_self_attention_in_proj_bias" -> "637 linear_28" [label="(2304,)", style=solid]; +"181 _scale_30" -> "184 dequantize_per_channel_default_30" [label="(768,)", style=solid]; +"182 _zero_point_30" -> "184 dequantize_per_channel_default_30" [label="(768,)", style=solid]; +"183 _frozen_param31" -> "184 dequantize_per_channel_default_30" [label="(768, 768)", style=solid]; +"184 dequantize_per_channel_default_30" -> "663 linear_29" [label=None, style=solid]; +"185 encoder_layers_encoder_layer_7_self_attention_out_proj_bias" -> "663 linear_29" [label="(768,)", style=solid]; +"186 encoder_layers_encoder_layer_7_ln_2_weight" -> "668 layer_norm_15" [label="(768,)", style=solid]; +"187 encoder_layers_encoder_layer_7_ln_2_bias" -> "668 layer_norm_15" [label="(768,)", style=solid]; +"188 _scale_31" -> "191 dequantize_per_channel_default_31" [label="(3072,)", style=solid]; +"189 _zero_point_31" -> "191 dequantize_per_channel_default_31" [label="(3072,)", style=solid]; +"190 _frozen_param32" -> "191 dequantize_per_channel_default_31" [label="(3072, 768)", style=solid]; +"191 dequantize_per_channel_default_31" -> "671 linear_30" [label=None, style=solid]; +"192 encoder_layers_encoder_layer_7_mlp_0_bias" -> "671 linear_30" [label="(3072,)", style=solid]; +"193 _scale_32" -> "196 dequantize_per_channel_default_32" [label="(768,)", style=solid]; +"194 _zero_point_32" -> "196 dequantize_per_channel_default_32" [label="(768,)", style=solid]; +"195 _frozen_param33" -> "196 dequantize_per_channel_default_32" [label="(768, 3072)", style=solid]; +"196 dequantize_per_channel_default_32" -> "676 linear_31" [label=None, style=solid]; +"197 encoder_layers_encoder_layer_7_mlp_3_bias" -> "676 linear_31" [label="(768,)", style=solid]; +"198 encoder_layers_encoder_layer_8_ln_1_weight" -> "679 layer_norm_16" [label="(768,)", style=solid]; +"199 encoder_layers_encoder_layer_8_ln_1_bias" -> "679 layer_norm_16" [label="(768,)", style=solid]; +"200 _scale_33" -> "203 dequantize_per_channel_default_33" [label="(2304,)", style=solid]; +"201 _zero_point_33" -> "203 dequantize_per_channel_default_33" [label="(2304,)", style=solid]; +"202 _frozen_param34" -> "203 dequantize_per_channel_default_33" [label="(2304, 768)", style=solid]; +"203 dequantize_per_channel_default_33" -> "683 linear_32" [label=None, style=solid]; +"204 encoder_layers_encoder_layer_8_self_attention_in_proj_bias" -> "683 linear_32" [label="(2304,)", style=solid]; +"205 _scale_34" -> "208 dequantize_per_channel_default_34" [label="(768,)", style=solid]; +"206 _zero_point_34" -> "208 dequantize_per_channel_default_34" [label="(768,)", style=solid]; +"207 _frozen_param35" -> "208 dequantize_per_channel_default_34" [label="(768, 768)", style=solid]; +"208 dequantize_per_channel_default_34" -> "709 linear_33" [label=None, style=solid]; +"209 encoder_layers_encoder_layer_8_self_attention_out_proj_bias" -> "709 linear_33" [label="(768,)", style=solid]; +"210 encoder_layers_encoder_layer_8_ln_2_weight" -> "714 layer_norm_17" [label="(768,)", style=solid]; +"211 encoder_layers_encoder_layer_8_ln_2_bias" -> "714 layer_norm_17" [label="(768,)", style=solid]; +"212 _scale_35" -> "215 dequantize_per_channel_default_35" [label="(3072,)", style=solid]; +"213 _zero_point_35" -> "215 dequantize_per_channel_default_35" [label="(3072,)", style=solid]; +"214 _frozen_param36" -> "215 dequantize_per_channel_default_35" [label="(3072, 768)", style=solid]; +"215 dequantize_per_channel_default_35" -> "717 linear_34" [label=None, style=solid]; +"216 encoder_layers_encoder_layer_8_mlp_0_bias" -> "717 linear_34" [label="(3072,)", style=solid]; +"217 _scale_36" -> "220 dequantize_per_channel_default_36" [label="(768,)", style=solid]; +"218 _zero_point_36" -> "220 dequantize_per_channel_default_36" [label="(768,)", style=solid]; +"219 _frozen_param37" -> "220 dequantize_per_channel_default_36" [label="(768, 3072)", style=solid]; +"220 dequantize_per_channel_default_36" -> "722 linear_35" [label=None, style=solid]; +"221 encoder_layers_encoder_layer_8_mlp_3_bias" -> "722 linear_35" [label="(768,)", style=solid]; +"222 encoder_layers_encoder_layer_9_ln_1_weight" -> "725 layer_norm_18" [label="(768,)", style=solid]; +"223 encoder_layers_encoder_layer_9_ln_1_bias" -> "725 layer_norm_18" [label="(768,)", style=solid]; +"224 _scale_37" -> "227 dequantize_per_channel_default_37" [label="(2304,)", style=solid]; +"225 _zero_point_37" -> "227 dequantize_per_channel_default_37" [label="(2304,)", style=solid]; +"226 _frozen_param38" -> "227 dequantize_per_channel_default_37" [label="(2304, 768)", style=solid]; +"227 dequantize_per_channel_default_37" -> "729 linear_36" [label=None, style=solid]; +"228 encoder_layers_encoder_layer_9_self_attention_in_proj_bias" -> "729 linear_36" [label="(2304,)", style=solid]; +"229 _scale_38" -> "232 dequantize_per_channel_default_38" [label="(768,)", style=solid]; +"230 _zero_point_38" -> "232 dequantize_per_channel_default_38" [label="(768,)", style=solid]; +"231 _frozen_param39" -> "232 dequantize_per_channel_default_38" [label="(768, 768)", style=solid]; +"232 dequantize_per_channel_default_38" -> "755 linear_37" [label=None, style=solid]; +"233 encoder_layers_encoder_layer_9_self_attention_out_proj_bias" -> "755 linear_37" [label="(768,)", style=solid]; +"234 encoder_layers_encoder_layer_9_ln_2_weight" -> "760 layer_norm_19" [label="(768,)", style=solid]; +"235 encoder_layers_encoder_layer_9_ln_2_bias" -> "760 layer_norm_19" [label="(768,)", style=solid]; +"236 _scale_39" -> "239 dequantize_per_channel_default_39" [label="(3072,)", style=solid]; +"237 _zero_point_39" -> "239 dequantize_per_channel_default_39" [label="(3072,)", style=solid]; +"238 _frozen_param40" -> "239 dequantize_per_channel_default_39" [label="(3072, 768)", style=solid]; +"239 dequantize_per_channel_default_39" -> "763 linear_38" [label=None, style=solid]; +"240 encoder_layers_encoder_layer_9_mlp_0_bias" -> "763 linear_38" [label="(3072,)", style=solid]; +"241 _scale_40" -> "244 dequantize_per_channel_default_40" [label="(768,)", style=solid]; +"242 _zero_point_40" -> "244 dequantize_per_channel_default_40" [label="(768,)", style=solid]; +"243 _frozen_param41" -> "244 dequantize_per_channel_default_40" [label="(768, 3072)", style=solid]; +"244 dequantize_per_channel_default_40" -> "768 linear_39" [label=None, style=solid]; +"245 encoder_layers_encoder_layer_9_mlp_3_bias" -> "768 linear_39" [label="(768,)", style=solid]; +"246 encoder_layers_encoder_layer_10_ln_1_weight" -> "771 layer_norm_20" [label="(768,)", style=solid]; +"247 encoder_layers_encoder_layer_10_ln_1_bias" -> "771 layer_norm_20" [label="(768,)", style=solid]; +"248 _scale_41" -> "251 dequantize_per_channel_default_41" [label="(2304,)", style=solid]; +"249 _zero_point_41" -> "251 dequantize_per_channel_default_41" [label="(2304,)", style=solid]; +"250 _frozen_param42" -> "251 dequantize_per_channel_default_41" [label="(2304, 768)", style=solid]; +"251 dequantize_per_channel_default_41" -> "775 linear_40" [label=None, style=solid]; +"252 encoder_layers_encoder_layer_10_self_attention_in_proj_bias" -> "775 linear_40" [label="(2304,)", style=solid]; +"253 _scale_42" -> "256 dequantize_per_channel_default_42" [label="(768,)", style=solid]; +"254 _zero_point_42" -> "256 dequantize_per_channel_default_42" [label="(768,)", style=solid]; +"255 _frozen_param43" -> "256 dequantize_per_channel_default_42" [label="(768, 768)", style=solid]; +"256 dequantize_per_channel_default_42" -> "801 linear_41" [label=None, style=solid]; +"257 encoder_layers_encoder_layer_10_self_attention_out_proj_bias" -> "801 linear_41" [label="(768,)", style=solid]; +"258 encoder_layers_encoder_layer_10_ln_2_weight" -> "806 layer_norm_21" [label="(768,)", style=solid]; +"259 encoder_layers_encoder_layer_10_ln_2_bias" -> "806 layer_norm_21" [label="(768,)", style=solid]; +"260 _scale_43" -> "263 dequantize_per_channel_default_43" [label="(3072,)", style=solid]; +"261 _zero_point_43" -> "263 dequantize_per_channel_default_43" [label="(3072,)", style=solid]; +"262 _frozen_param44" -> "263 dequantize_per_channel_default_43" [label="(3072, 768)", style=solid]; +"263 dequantize_per_channel_default_43" -> "809 linear_42" [label=None, style=solid]; +"264 encoder_layers_encoder_layer_10_mlp_0_bias" -> "809 linear_42" [label="(3072,)", style=solid]; +"265 _scale_44" -> "268 dequantize_per_channel_default_44" [label="(768,)", style=solid]; +"266 _zero_point_44" -> "268 dequantize_per_channel_default_44" [label="(768,)", style=solid]; +"267 _frozen_param45" -> "268 dequantize_per_channel_default_44" [label="(768, 3072)", style=solid]; +"268 dequantize_per_channel_default_44" -> "814 linear_43" [label=None, style=solid]; +"269 encoder_layers_encoder_layer_10_mlp_3_bias" -> "814 linear_43" [label="(768,)", style=solid]; +"270 encoder_layers_encoder_layer_11_ln_1_weight" -> "817 layer_norm_22" [label="(768,)", style=solid]; +"271 encoder_layers_encoder_layer_11_ln_1_bias" -> "817 layer_norm_22" [label="(768,)", style=solid]; +"272 _scale_45" -> "275 dequantize_per_channel_default_45" [label="(2304,)", style=solid]; +"273 _zero_point_45" -> "275 dequantize_per_channel_default_45" [label="(2304,)", style=solid]; +"274 _frozen_param46" -> "275 dequantize_per_channel_default_45" [label="(2304, 768)", style=solid]; +"275 dequantize_per_channel_default_45" -> "821 linear_44" [label=None, style=solid]; +"276 encoder_layers_encoder_layer_11_self_attention_in_proj_bias" -> "821 linear_44" [label="(2304,)", style=solid]; +"277 _scale_46" -> "280 dequantize_per_channel_default_46" [label="(768,)", style=solid]; +"278 _zero_point_46" -> "280 dequantize_per_channel_default_46" [label="(768,)", style=solid]; +"279 _frozen_param47" -> "280 dequantize_per_channel_default_46" [label="(768, 768)", style=solid]; +"280 dequantize_per_channel_default_46" -> "847 linear_45" [label=None, style=solid]; +"281 encoder_layers_encoder_layer_11_self_attention_out_proj_bias" -> "847 linear_45" [label="(768,)", style=solid]; +"282 encoder_layers_encoder_layer_11_ln_2_weight" -> "852 layer_norm_23" [label="(768,)", style=solid]; +"283 encoder_layers_encoder_layer_11_ln_2_bias" -> "852 layer_norm_23" [label="(768,)", style=solid]; +"284 _scale_47" -> "287 dequantize_per_channel_default_47" [label="(3072,)", style=solid]; +"285 _zero_point_47" -> "287 dequantize_per_channel_default_47" [label="(3072,)", style=solid]; +"286 _frozen_param48" -> "287 dequantize_per_channel_default_47" [label="(3072, 768)", style=solid]; +"287 dequantize_per_channel_default_47" -> "855 linear_46" [label=None, style=solid]; +"288 encoder_layers_encoder_layer_11_mlp_0_bias" -> "855 linear_46" [label="(3072,)", style=solid]; +"289 _scale_48" -> "292 dequantize_per_channel_default_48" [label="(768,)", style=solid]; +"290 _zero_point_48" -> "292 dequantize_per_channel_default_48" [label="(768,)", style=solid]; +"291 _frozen_param49" -> "292 dequantize_per_channel_default_48" [label="(768, 3072)", style=solid]; +"292 dequantize_per_channel_default_48" -> "860 linear_47" [label=None, style=solid]; +"293 encoder_layers_encoder_layer_11_mlp_3_bias" -> "860 linear_47" [label="(768,)", style=solid]; +"294 encoder_ln_weight" -> "863 layer_norm_24" [label="(768,)", style=solid]; +"295 encoder_ln_bias" -> "863 layer_norm_24" [label="(768,)", style=solid]; +"296 _scale_49" -> "299 dequantize_per_channel_default_49" [label="(1000,)", style=solid]; +"297 _zero_point_49" -> "299 dequantize_per_channel_default_49" [label="(1000,)", style=solid]; +"298 _frozen_param50" -> "299 dequantize_per_channel_default_49" [label="(1000, 768)", style=solid]; +"299 dequantize_per_channel_default_49" -> "868 linear_48" [label=None, style=solid]; +"300 heads_head_bias" -> "868 linear_48" [label="(1000,)", style=solid]; +"301 x" -> "302 quantize_per_tensor_default" [label="(1, 3, 224, 224)", style=solid]; +"302 quantize_per_tensor_default" -> "303 dequantize_per_tensor_default" [label=None, style=solid]; +"303 dequantize_per_tensor_default" -> "304 conv2d" [label=None, style=solid]; +"304 conv2d" -> "305 reshape" [label="(1, 768, 14, 14)", style=solid]; +"305 reshape" -> "306 permute" [label="(1, 768, 196)", style=solid]; +"306 permute" -> "308 cat" [label="(1, 196, 768)", style=solid]; +"307 _frozen_param0" -> "308 cat" [label="(1, 1, 768)", style=solid]; +"308 cat" -> "309 add" [label="(1, 197, 768)", style=solid]; +"309 add" -> "310 dropout" [label="(1, 197, 768)", style=solid]; +"310 dropout" -> "311 layer_norm" [label="(1, 197, 768)", style=solid]; +"310 dropout" -> "345 add_1" [label="(1, 197, 768)", style=solid]; +"311 layer_norm" -> "312 quantize_per_tensor_default_1" [label="(1, 197, 768)", style=solid]; +"312 quantize_per_tensor_default_1" -> "313 dequantize_per_tensor_default_1" [label=None, style=solid]; +"313 dequantize_per_tensor_default_1" -> "314 transpose" [label=None, style=solid]; +"314 transpose" -> "315 linear" [label="(197, 1, 768)", style=solid]; +"315 linear" -> "316 unflatten" [label="(197, 1, 2304)", style=solid]; +"316 unflatten" -> "317 unsqueeze" [label="(197, 1, 3, 768)", style=solid]; +"317 unsqueeze" -> "318 transpose_1" [label="(1, 197, 1, 3, 768)", style=solid]; +"318 transpose_1" -> "319 squeeze" [label="(3, 197, 1, 1, 768)", style=solid]; +"319 squeeze" -> "320 contiguous" [label="(3, 197, 1, 768)", style=solid]; +"320 contiguous" -> "321 quantize_per_tensor_default_2" [label="(3, 197, 1, 768)", style=solid]; +"320 contiguous" -> "326 select_2" [label="(3, 197, 1, 768)", style=solid]; +"321 quantize_per_tensor_default_2" -> "322 dequantize_per_tensor_default_63" [label=None, style=solid]; +"321 quantize_per_tensor_default_2" -> "323 dequantize_per_tensor_default_62" [label=None, style=solid]; +"322 dequantize_per_tensor_default_63" -> "325 select_1" [label=None, style=solid]; +"323 dequantize_per_tensor_default_62" -> "324 select" [label=None, style=solid]; +"324 select" -> "327 view" [label="(197, 1, 768)", style=solid]; +"325 select_1" -> "329 view_1" [label="(197, 1, 768)", style=solid]; +"326 select_2" -> "331 view_2" [label="(197, 1, 768)", style=solid]; +"327 view" -> "328 transpose_2" [label="(197, 12, 64)", style=solid]; +"328 transpose_2" -> "333 view_3" [label="(12, 197, 64)", style=solid]; +"329 view_1" -> "330 transpose_3" [label="(197, 12, 64)", style=solid]; +"330 transpose_3" -> "334 view_4" [label="(12, 197, 64)", style=solid]; +"331 view_2" -> "332 transpose_4" [label="(197, 12, 64)", style=solid]; +"332 transpose_4" -> "335 view_5" [label="(12, 197, 64)", style=solid]; +"333 view_3" -> "336 scaled_dot_product_attention" [label="(1, 12, 197, 64)", style=solid]; +"334 view_4" -> "336 scaled_dot_product_attention" [label="(1, 12, 197, 64)", style=solid]; +"335 view_5" -> "336 scaled_dot_product_attention" [label="(1, 12, 197, 64)", style=solid]; +"336 scaled_dot_product_attention" -> "337 quantize_per_tensor_default_3" [label="(1, 12, 197, 64)", style=solid]; +"337 quantize_per_tensor_default_3" -> "338 dequantize_per_tensor_default_3" [label=None, style=solid]; +"338 dequantize_per_tensor_default_3" -> "339 permute_1" [label=None, style=solid]; +"339 permute_1" -> "340 view_6" [label="(197, 1, 12, 64)", style=solid]; +"340 view_6" -> "341 linear_1" [label="(197, 768)", style=solid]; +"341 linear_1" -> "342 view_7" [label="(197, 768)", style=solid]; +"342 view_7" -> "343 transpose_5" [label="(197, 1, 768)", style=solid]; +"343 transpose_5" -> "344 dropout_1" [label="(1, 197, 768)", style=solid]; +"344 dropout_1" -> "345 add_1" [label="(1, 197, 768)", style=solid]; +"345 add_1" -> "346 layer_norm_1" [label="(1, 197, 768)", style=solid]; +"345 add_1" -> "356 add_2" [label="(1, 197, 768)", style=solid]; +"346 layer_norm_1" -> "347 quantize_per_tensor_default_4" [label="(1, 197, 768)", style=solid]; +"347 quantize_per_tensor_default_4" -> "348 dequantize_per_tensor_default_4" [label=None, style=solid]; +"348 dequantize_per_tensor_default_4" -> "349 linear_2" [label=None, style=solid]; +"349 linear_2" -> "350 gelu" [label="(1, 197, 3072)", style=solid]; +"350 gelu" -> "351 quantize_per_tensor_default_5" [label="(1, 197, 3072)", style=solid]; +"351 quantize_per_tensor_default_5" -> "352 dequantize_per_tensor_default_5" [label=None, style=solid]; +"352 dequantize_per_tensor_default_5" -> "353 dropout_2" [label=None, style=solid]; +"353 dropout_2" -> "354 linear_3" [label="(1, 197, 3072)", style=solid]; +"354 linear_3" -> "355 dropout_3" [label="(1, 197, 768)", style=solid]; +"355 dropout_3" -> "356 add_2" [label="(1, 197, 768)", style=solid]; +"356 add_2" -> "357 layer_norm_2" [label="(1, 197, 768)", style=solid]; +"356 add_2" -> "391 add_3" [label="(1, 197, 768)", style=solid]; +"357 layer_norm_2" -> "358 quantize_per_tensor_default_6" [label="(1, 197, 768)", style=solid]; +"358 quantize_per_tensor_default_6" -> "359 dequantize_per_tensor_default_6" [label=None, style=solid]; +"359 dequantize_per_tensor_default_6" -> "360 transpose_6" [label=None, style=solid]; +"360 transpose_6" -> "361 linear_4" [label="(197, 1, 768)", style=solid]; +"361 linear_4" -> "362 unflatten_1" [label="(197, 1, 2304)", style=solid]; +"362 unflatten_1" -> "363 unsqueeze_1" [label="(197, 1, 3, 768)", style=solid]; +"363 unsqueeze_1" -> "364 transpose_7" [label="(1, 197, 1, 3, 768)", style=solid]; +"364 transpose_7" -> "365 squeeze_1" [label="(3, 197, 1, 1, 768)", style=solid]; +"365 squeeze_1" -> "366 contiguous_1" [label="(3, 197, 1, 768)", style=solid]; +"366 contiguous_1" -> "367 quantize_per_tensor_default_7" [label="(3, 197, 1, 768)", style=solid]; +"366 contiguous_1" -> "372 select_5" [label="(3, 197, 1, 768)", style=solid]; +"367 quantize_per_tensor_default_7" -> "368 dequantize_per_tensor_default_65" [label=None, style=solid]; +"367 quantize_per_tensor_default_7" -> "369 dequantize_per_tensor_default_64" [label=None, style=solid]; +"368 dequantize_per_tensor_default_65" -> "371 select_4" [label=None, style=solid]; +"369 dequantize_per_tensor_default_64" -> "370 select_3" [label=None, style=solid]; +"370 select_3" -> "373 view_8" [label="(197, 1, 768)", style=solid]; +"371 select_4" -> "375 view_9" [label="(197, 1, 768)", style=solid]; +"372 select_5" -> "377 view_10" [label="(197, 1, 768)", style=solid]; +"373 view_8" -> "374 transpose_8" [label="(197, 12, 64)", style=solid]; +"374 transpose_8" -> "379 view_11" [label="(12, 197, 64)", style=solid]; +"375 view_9" -> "376 transpose_9" [label="(197, 12, 64)", style=solid]; +"376 transpose_9" -> "380 view_12" [label="(12, 197, 64)", style=solid]; +"377 view_10" -> "378 transpose_10" [label="(197, 12, 64)", style=solid]; +"378 transpose_10" -> "381 view_13" [label="(12, 197, 64)", style=solid]; +"379 view_11" -> "382 scaled_dot_product_attention_1" [label="(1, 12, 197, 64)", style=solid]; +"380 view_12" -> "382 scaled_dot_product_attention_1" [label="(1, 12, 197, 64)", style=solid]; +"381 view_13" -> "382 scaled_dot_product_attention_1" [label="(1, 12, 197, 64)", style=solid]; +"382 scaled_dot_product_attention_1" -> "383 quantize_per_tensor_default_8" [label="(1, 12, 197, 64)", style=solid]; +"383 quantize_per_tensor_default_8" -> "384 dequantize_per_tensor_default_8" [label=None, style=solid]; +"384 dequantize_per_tensor_default_8" -> "385 permute_2" [label=None, style=solid]; +"385 permute_2" -> "386 view_14" [label="(197, 1, 12, 64)", style=solid]; +"386 view_14" -> "387 linear_5" [label="(197, 768)", style=solid]; +"387 linear_5" -> "388 view_15" [label="(197, 768)", style=solid]; +"388 view_15" -> "389 transpose_11" [label="(197, 1, 768)", style=solid]; +"389 transpose_11" -> "390 dropout_4" [label="(1, 197, 768)", style=solid]; +"390 dropout_4" -> "391 add_3" [label="(1, 197, 768)", style=solid]; +"391 add_3" -> "392 layer_norm_3" [label="(1, 197, 768)", style=solid]; +"391 add_3" -> "402 add_4" [label="(1, 197, 768)", style=solid]; +"392 layer_norm_3" -> "393 quantize_per_tensor_default_9" [label="(1, 197, 768)", style=solid]; +"393 quantize_per_tensor_default_9" -> "394 dequantize_per_tensor_default_9" [label=None, style=solid]; +"394 dequantize_per_tensor_default_9" -> "395 linear_6" [label=None, style=solid]; +"395 linear_6" -> "396 gelu_1" [label="(1, 197, 3072)", style=solid]; +"396 gelu_1" -> "397 quantize_per_tensor_default_10" [label="(1, 197, 3072)", style=solid]; +"397 quantize_per_tensor_default_10" -> "398 dequantize_per_tensor_default_10" [label=None, style=solid]; +"398 dequantize_per_tensor_default_10" -> "399 dropout_5" [label=None, style=solid]; +"399 dropout_5" -> "400 linear_7" [label="(1, 197, 3072)", style=solid]; +"400 linear_7" -> "401 dropout_6" [label="(1, 197, 768)", style=solid]; +"401 dropout_6" -> "402 add_4" [label="(1, 197, 768)", style=solid]; +"402 add_4" -> "403 layer_norm_4" [label="(1, 197, 768)", style=solid]; +"402 add_4" -> "437 add_5" [label="(1, 197, 768)", style=solid]; +"403 layer_norm_4" -> "404 quantize_per_tensor_default_11" [label="(1, 197, 768)", style=solid]; +"404 quantize_per_tensor_default_11" -> "405 dequantize_per_tensor_default_11" [label=None, style=solid]; +"405 dequantize_per_tensor_default_11" -> "406 transpose_12" [label=None, style=solid]; +"406 transpose_12" -> "407 linear_8" [label="(197, 1, 768)", style=solid]; +"407 linear_8" -> "408 unflatten_2" [label="(197, 1, 2304)", style=solid]; +"408 unflatten_2" -> "409 unsqueeze_2" [label="(197, 1, 3, 768)", style=solid]; +"409 unsqueeze_2" -> "410 transpose_13" [label="(1, 197, 1, 3, 768)", style=solid]; +"410 transpose_13" -> "411 squeeze_2" [label="(3, 197, 1, 1, 768)", style=solid]; +"411 squeeze_2" -> "412 contiguous_2" [label="(3, 197, 1, 768)", style=solid]; +"412 contiguous_2" -> "413 quantize_per_tensor_default_12" [label="(3, 197, 1, 768)", style=solid]; +"412 contiguous_2" -> "418 select_8" [label="(3, 197, 1, 768)", style=solid]; +"413 quantize_per_tensor_default_12" -> "414 dequantize_per_tensor_default_67" [label=None, style=solid]; +"413 quantize_per_tensor_default_12" -> "415 dequantize_per_tensor_default_66" [label=None, style=solid]; +"414 dequantize_per_tensor_default_67" -> "417 select_7" [label=None, style=solid]; +"415 dequantize_per_tensor_default_66" -> "416 select_6" [label=None, style=solid]; +"416 select_6" -> "419 view_16" [label="(197, 1, 768)", style=solid]; +"417 select_7" -> "421 view_17" [label="(197, 1, 768)", style=solid]; +"418 select_8" -> "423 view_18" [label="(197, 1, 768)", style=solid]; +"419 view_16" -> "420 transpose_14" [label="(197, 12, 64)", style=solid]; +"420 transpose_14" -> "425 view_19" [label="(12, 197, 64)", style=solid]; +"421 view_17" -> "422 transpose_15" [label="(197, 12, 64)", style=solid]; +"422 transpose_15" -> "426 view_20" [label="(12, 197, 64)", style=solid]; +"423 view_18" -> "424 transpose_16" [label="(197, 12, 64)", style=solid]; +"424 transpose_16" -> "427 view_21" [label="(12, 197, 64)", style=solid]; +"425 view_19" -> "428 scaled_dot_product_attention_2" [label="(1, 12, 197, 64)", style=solid]; +"426 view_20" -> "428 scaled_dot_product_attention_2" [label="(1, 12, 197, 64)", style=solid]; +"427 view_21" -> "428 scaled_dot_product_attention_2" [label="(1, 12, 197, 64)", style=solid]; +"428 scaled_dot_product_attention_2" -> "429 quantize_per_tensor_default_13" [label="(1, 12, 197, 64)", style=solid]; +"429 quantize_per_tensor_default_13" -> "430 dequantize_per_tensor_default_13" [label=None, style=solid]; +"430 dequantize_per_tensor_default_13" -> "431 permute_3" [label=None, style=solid]; +"431 permute_3" -> "432 view_22" [label="(197, 1, 12, 64)", style=solid]; +"432 view_22" -> "433 linear_9" [label="(197, 768)", style=solid]; +"433 linear_9" -> "434 view_23" [label="(197, 768)", style=solid]; +"434 view_23" -> "435 transpose_17" [label="(197, 1, 768)", style=solid]; +"435 transpose_17" -> "436 dropout_7" [label="(1, 197, 768)", style=solid]; +"436 dropout_7" -> "437 add_5" [label="(1, 197, 768)", style=solid]; +"437 add_5" -> "438 layer_norm_5" [label="(1, 197, 768)", style=solid]; +"437 add_5" -> "448 add_6" [label="(1, 197, 768)", style=solid]; +"438 layer_norm_5" -> "439 quantize_per_tensor_default_14" [label="(1, 197, 768)", style=solid]; +"439 quantize_per_tensor_default_14" -> "440 dequantize_per_tensor_default_14" [label=None, style=solid]; +"440 dequantize_per_tensor_default_14" -> "441 linear_10" [label=None, style=solid]; +"441 linear_10" -> "442 gelu_2" [label="(1, 197, 3072)", style=solid]; +"442 gelu_2" -> "443 quantize_per_tensor_default_15" [label="(1, 197, 3072)", style=solid]; +"443 quantize_per_tensor_default_15" -> "444 dequantize_per_tensor_default_15" [label=None, style=solid]; +"444 dequantize_per_tensor_default_15" -> "445 dropout_8" [label=None, style=solid]; +"445 dropout_8" -> "446 linear_11" [label="(1, 197, 3072)", style=solid]; +"446 linear_11" -> "447 dropout_9" [label="(1, 197, 768)", style=solid]; +"447 dropout_9" -> "448 add_6" [label="(1, 197, 768)", style=solid]; +"448 add_6" -> "449 layer_norm_6" [label="(1, 197, 768)", style=solid]; +"448 add_6" -> "483 add_7" [label="(1, 197, 768)", style=solid]; +"449 layer_norm_6" -> "450 quantize_per_tensor_default_16" [label="(1, 197, 768)", style=solid]; +"450 quantize_per_tensor_default_16" -> "451 dequantize_per_tensor_default_16" [label=None, style=solid]; +"451 dequantize_per_tensor_default_16" -> "452 transpose_18" [label=None, style=solid]; +"452 transpose_18" -> "453 linear_12" [label="(197, 1, 768)", style=solid]; +"453 linear_12" -> "454 unflatten_3" [label="(197, 1, 2304)", style=solid]; +"454 unflatten_3" -> "455 unsqueeze_3" [label="(197, 1, 3, 768)", style=solid]; +"455 unsqueeze_3" -> "456 transpose_19" [label="(1, 197, 1, 3, 768)", style=solid]; +"456 transpose_19" -> "457 squeeze_3" [label="(3, 197, 1, 1, 768)", style=solid]; +"457 squeeze_3" -> "458 contiguous_3" [label="(3, 197, 1, 768)", style=solid]; +"458 contiguous_3" -> "459 quantize_per_tensor_default_17" [label="(3, 197, 1, 768)", style=solid]; +"458 contiguous_3" -> "464 select_11" [label="(3, 197, 1, 768)", style=solid]; +"459 quantize_per_tensor_default_17" -> "460 dequantize_per_tensor_default_69" [label=None, style=solid]; +"459 quantize_per_tensor_default_17" -> "461 dequantize_per_tensor_default_68" [label=None, style=solid]; +"460 dequantize_per_tensor_default_69" -> "463 select_10" [label=None, style=solid]; +"461 dequantize_per_tensor_default_68" -> "462 select_9" [label=None, style=solid]; +"462 select_9" -> "465 view_24" [label="(197, 1, 768)", style=solid]; +"463 select_10" -> "467 view_25" [label="(197, 1, 768)", style=solid]; +"464 select_11" -> "469 view_26" [label="(197, 1, 768)", style=solid]; +"465 view_24" -> "466 transpose_20" [label="(197, 12, 64)", style=solid]; +"466 transpose_20" -> "471 view_27" [label="(12, 197, 64)", style=solid]; +"467 view_25" -> "468 transpose_21" [label="(197, 12, 64)", style=solid]; +"468 transpose_21" -> "472 view_28" [label="(12, 197, 64)", style=solid]; +"469 view_26" -> "470 transpose_22" [label="(197, 12, 64)", style=solid]; +"470 transpose_22" -> "473 view_29" [label="(12, 197, 64)", style=solid]; +"471 view_27" -> "474 scaled_dot_product_attention_3" [label="(1, 12, 197, 64)", style=solid]; +"472 view_28" -> "474 scaled_dot_product_attention_3" [label="(1, 12, 197, 64)", style=solid]; +"473 view_29" -> "474 scaled_dot_product_attention_3" [label="(1, 12, 197, 64)", style=solid]; +"474 scaled_dot_product_attention_3" -> "475 quantize_per_tensor_default_18" [label="(1, 12, 197, 64)", style=solid]; +"475 quantize_per_tensor_default_18" -> "476 dequantize_per_tensor_default_18" [label=None, style=solid]; +"476 dequantize_per_tensor_default_18" -> "477 permute_4" [label=None, style=solid]; +"477 permute_4" -> "478 view_30" [label="(197, 1, 12, 64)", style=solid]; +"478 view_30" -> "479 linear_13" [label="(197, 768)", style=solid]; +"479 linear_13" -> "480 view_31" [label="(197, 768)", style=solid]; +"480 view_31" -> "481 transpose_23" [label="(197, 1, 768)", style=solid]; +"481 transpose_23" -> "482 dropout_10" [label="(1, 197, 768)", style=solid]; +"482 dropout_10" -> "483 add_7" [label="(1, 197, 768)", style=solid]; +"483 add_7" -> "484 layer_norm_7" [label="(1, 197, 768)", style=solid]; +"483 add_7" -> "494 add_8" [label="(1, 197, 768)", style=solid]; +"484 layer_norm_7" -> "485 quantize_per_tensor_default_19" [label="(1, 197, 768)", style=solid]; +"485 quantize_per_tensor_default_19" -> "486 dequantize_per_tensor_default_19" [label=None, style=solid]; +"486 dequantize_per_tensor_default_19" -> "487 linear_14" [label=None, style=solid]; +"487 linear_14" -> "488 gelu_3" [label="(1, 197, 3072)", style=solid]; +"488 gelu_3" -> "489 quantize_per_tensor_default_20" [label="(1, 197, 3072)", style=solid]; +"489 quantize_per_tensor_default_20" -> "490 dequantize_per_tensor_default_20" [label=None, style=solid]; +"490 dequantize_per_tensor_default_20" -> "491 dropout_11" [label=None, style=solid]; +"491 dropout_11" -> "492 linear_15" [label="(1, 197, 3072)", style=solid]; +"492 linear_15" -> "493 dropout_12" [label="(1, 197, 768)", style=solid]; +"493 dropout_12" -> "494 add_8" [label="(1, 197, 768)", style=solid]; +"494 add_8" -> "495 layer_norm_8" [label="(1, 197, 768)", style=solid]; +"494 add_8" -> "529 add_9" [label="(1, 197, 768)", style=solid]; +"495 layer_norm_8" -> "496 quantize_per_tensor_default_21" [label="(1, 197, 768)", style=solid]; +"496 quantize_per_tensor_default_21" -> "497 dequantize_per_tensor_default_21" [label=None, style=solid]; +"497 dequantize_per_tensor_default_21" -> "498 transpose_24" [label=None, style=solid]; +"498 transpose_24" -> "499 linear_16" [label="(197, 1, 768)", style=solid]; +"499 linear_16" -> "500 unflatten_4" [label="(197, 1, 2304)", style=solid]; +"500 unflatten_4" -> "501 unsqueeze_4" [label="(197, 1, 3, 768)", style=solid]; +"501 unsqueeze_4" -> "502 transpose_25" [label="(1, 197, 1, 3, 768)", style=solid]; +"502 transpose_25" -> "503 squeeze_4" [label="(3, 197, 1, 1, 768)", style=solid]; +"503 squeeze_4" -> "504 contiguous_4" [label="(3, 197, 1, 768)", style=solid]; +"504 contiguous_4" -> "505 quantize_per_tensor_default_22" [label="(3, 197, 1, 768)", style=solid]; +"504 contiguous_4" -> "510 select_14" [label="(3, 197, 1, 768)", style=solid]; +"505 quantize_per_tensor_default_22" -> "506 dequantize_per_tensor_default_71" [label=None, style=solid]; +"505 quantize_per_tensor_default_22" -> "507 dequantize_per_tensor_default_70" [label=None, style=solid]; +"506 dequantize_per_tensor_default_71" -> "509 select_13" [label=None, style=solid]; +"507 dequantize_per_tensor_default_70" -> "508 select_12" [label=None, style=solid]; +"508 select_12" -> "511 view_32" [label="(197, 1, 768)", style=solid]; +"509 select_13" -> "513 view_33" [label="(197, 1, 768)", style=solid]; +"510 select_14" -> "515 view_34" [label="(197, 1, 768)", style=solid]; +"511 view_32" -> "512 transpose_26" [label="(197, 12, 64)", style=solid]; +"512 transpose_26" -> "517 view_35" [label="(12, 197, 64)", style=solid]; +"513 view_33" -> "514 transpose_27" [label="(197, 12, 64)", style=solid]; +"514 transpose_27" -> "518 view_36" [label="(12, 197, 64)", style=solid]; +"515 view_34" -> "516 transpose_28" [label="(197, 12, 64)", style=solid]; +"516 transpose_28" -> "519 view_37" [label="(12, 197, 64)", style=solid]; +"517 view_35" -> "520 scaled_dot_product_attention_4" [label="(1, 12, 197, 64)", style=solid]; +"518 view_36" -> "520 scaled_dot_product_attention_4" [label="(1, 12, 197, 64)", style=solid]; +"519 view_37" -> "520 scaled_dot_product_attention_4" [label="(1, 12, 197, 64)", style=solid]; +"520 scaled_dot_product_attention_4" -> "521 quantize_per_tensor_default_23" [label="(1, 12, 197, 64)", style=solid]; +"521 quantize_per_tensor_default_23" -> "522 dequantize_per_tensor_default_23" [label=None, style=solid]; +"522 dequantize_per_tensor_default_23" -> "523 permute_5" [label=None, style=solid]; +"523 permute_5" -> "524 view_38" [label="(197, 1, 12, 64)", style=solid]; +"524 view_38" -> "525 linear_17" [label="(197, 768)", style=solid]; +"525 linear_17" -> "526 view_39" [label="(197, 768)", style=solid]; +"526 view_39" -> "527 transpose_29" [label="(197, 1, 768)", style=solid]; +"527 transpose_29" -> "528 dropout_13" [label="(1, 197, 768)", style=solid]; +"528 dropout_13" -> "529 add_9" [label="(1, 197, 768)", style=solid]; +"529 add_9" -> "530 layer_norm_9" [label="(1, 197, 768)", style=solid]; +"529 add_9" -> "540 add_10" [label="(1, 197, 768)", style=solid]; +"530 layer_norm_9" -> "531 quantize_per_tensor_default_24" [label="(1, 197, 768)", style=solid]; +"531 quantize_per_tensor_default_24" -> "532 dequantize_per_tensor_default_24" [label=None, style=solid]; +"532 dequantize_per_tensor_default_24" -> "533 linear_18" [label=None, style=solid]; +"533 linear_18" -> "534 gelu_4" [label="(1, 197, 3072)", style=solid]; +"534 gelu_4" -> "535 quantize_per_tensor_default_25" [label="(1, 197, 3072)", style=solid]; +"535 quantize_per_tensor_default_25" -> "536 dequantize_per_tensor_default_25" [label=None, style=solid]; +"536 dequantize_per_tensor_default_25" -> "537 dropout_14" [label=None, style=solid]; +"537 dropout_14" -> "538 linear_19" [label="(1, 197, 3072)", style=solid]; +"538 linear_19" -> "539 dropout_15" [label="(1, 197, 768)", style=solid]; +"539 dropout_15" -> "540 add_10" [label="(1, 197, 768)", style=solid]; +"540 add_10" -> "541 layer_norm_10" [label="(1, 197, 768)", style=solid]; +"540 add_10" -> "575 add_11" [label="(1, 197, 768)", style=solid]; +"541 layer_norm_10" -> "542 quantize_per_tensor_default_26" [label="(1, 197, 768)", style=solid]; +"542 quantize_per_tensor_default_26" -> "543 dequantize_per_tensor_default_26" [label=None, style=solid]; +"543 dequantize_per_tensor_default_26" -> "544 transpose_30" [label=None, style=solid]; +"544 transpose_30" -> "545 linear_20" [label="(197, 1, 768)", style=solid]; +"545 linear_20" -> "546 unflatten_5" [label="(197, 1, 2304)", style=solid]; +"546 unflatten_5" -> "547 unsqueeze_5" [label="(197, 1, 3, 768)", style=solid]; +"547 unsqueeze_5" -> "548 transpose_31" [label="(1, 197, 1, 3, 768)", style=solid]; +"548 transpose_31" -> "549 squeeze_5" [label="(3, 197, 1, 1, 768)", style=solid]; +"549 squeeze_5" -> "550 contiguous_5" [label="(3, 197, 1, 768)", style=solid]; +"550 contiguous_5" -> "551 quantize_per_tensor_default_27" [label="(3, 197, 1, 768)", style=solid]; +"550 contiguous_5" -> "556 select_17" [label="(3, 197, 1, 768)", style=solid]; +"551 quantize_per_tensor_default_27" -> "552 dequantize_per_tensor_default_73" [label=None, style=solid]; +"551 quantize_per_tensor_default_27" -> "553 dequantize_per_tensor_default_72" [label=None, style=solid]; +"552 dequantize_per_tensor_default_73" -> "555 select_16" [label=None, style=solid]; +"553 dequantize_per_tensor_default_72" -> "554 select_15" [label=None, style=solid]; +"554 select_15" -> "557 view_40" [label="(197, 1, 768)", style=solid]; +"555 select_16" -> "559 view_41" [label="(197, 1, 768)", style=solid]; +"556 select_17" -> "561 view_42" [label="(197, 1, 768)", style=solid]; +"557 view_40" -> "558 transpose_32" [label="(197, 12, 64)", style=solid]; +"558 transpose_32" -> "563 view_43" [label="(12, 197, 64)", style=solid]; +"559 view_41" -> "560 transpose_33" [label="(197, 12, 64)", style=solid]; +"560 transpose_33" -> "564 view_44" [label="(12, 197, 64)", style=solid]; +"561 view_42" -> "562 transpose_34" [label="(197, 12, 64)", style=solid]; +"562 transpose_34" -> "565 view_45" [label="(12, 197, 64)", style=solid]; +"563 view_43" -> "566 scaled_dot_product_attention_5" [label="(1, 12, 197, 64)", style=solid]; +"564 view_44" -> "566 scaled_dot_product_attention_5" [label="(1, 12, 197, 64)", style=solid]; +"565 view_45" -> "566 scaled_dot_product_attention_5" [label="(1, 12, 197, 64)", style=solid]; +"566 scaled_dot_product_attention_5" -> "567 quantize_per_tensor_default_28" [label="(1, 12, 197, 64)", style=solid]; +"567 quantize_per_tensor_default_28" -> "568 dequantize_per_tensor_default_28" [label=None, style=solid]; +"568 dequantize_per_tensor_default_28" -> "569 permute_6" [label=None, style=solid]; +"569 permute_6" -> "570 view_46" [label="(197, 1, 12, 64)", style=solid]; +"570 view_46" -> "571 linear_21" [label="(197, 768)", style=solid]; +"571 linear_21" -> "572 view_47" [label="(197, 768)", style=solid]; +"572 view_47" -> "573 transpose_35" [label="(197, 1, 768)", style=solid]; +"573 transpose_35" -> "574 dropout_16" [label="(1, 197, 768)", style=solid]; +"574 dropout_16" -> "575 add_11" [label="(1, 197, 768)", style=solid]; +"575 add_11" -> "576 layer_norm_11" [label="(1, 197, 768)", style=solid]; +"575 add_11" -> "586 add_12" [label="(1, 197, 768)", style=solid]; +"576 layer_norm_11" -> "577 quantize_per_tensor_default_29" [label="(1, 197, 768)", style=solid]; +"577 quantize_per_tensor_default_29" -> "578 dequantize_per_tensor_default_29" [label=None, style=solid]; +"578 dequantize_per_tensor_default_29" -> "579 linear_22" [label=None, style=solid]; +"579 linear_22" -> "580 gelu_5" [label="(1, 197, 3072)", style=solid]; +"580 gelu_5" -> "581 quantize_per_tensor_default_30" [label="(1, 197, 3072)", style=solid]; +"581 quantize_per_tensor_default_30" -> "582 dequantize_per_tensor_default_30" [label=None, style=solid]; +"582 dequantize_per_tensor_default_30" -> "583 dropout_17" [label=None, style=solid]; +"583 dropout_17" -> "584 linear_23" [label="(1, 197, 3072)", style=solid]; +"584 linear_23" -> "585 dropout_18" [label="(1, 197, 768)", style=solid]; +"585 dropout_18" -> "586 add_12" [label="(1, 197, 768)", style=solid]; +"586 add_12" -> "587 layer_norm_12" [label="(1, 197, 768)", style=solid]; +"586 add_12" -> "621 add_13" [label="(1, 197, 768)", style=solid]; +"587 layer_norm_12" -> "588 quantize_per_tensor_default_31" [label="(1, 197, 768)", style=solid]; +"588 quantize_per_tensor_default_31" -> "589 dequantize_per_tensor_default_31" [label=None, style=solid]; +"589 dequantize_per_tensor_default_31" -> "590 transpose_36" [label=None, style=solid]; +"590 transpose_36" -> "591 linear_24" [label="(197, 1, 768)", style=solid]; +"591 linear_24" -> "592 unflatten_6" [label="(197, 1, 2304)", style=solid]; +"592 unflatten_6" -> "593 unsqueeze_6" [label="(197, 1, 3, 768)", style=solid]; +"593 unsqueeze_6" -> "594 transpose_37" [label="(1, 197, 1, 3, 768)", style=solid]; +"594 transpose_37" -> "595 squeeze_6" [label="(3, 197, 1, 1, 768)", style=solid]; +"595 squeeze_6" -> "596 contiguous_6" [label="(3, 197, 1, 768)", style=solid]; +"596 contiguous_6" -> "597 quantize_per_tensor_default_32" [label="(3, 197, 1, 768)", style=solid]; +"596 contiguous_6" -> "602 select_20" [label="(3, 197, 1, 768)", style=solid]; +"597 quantize_per_tensor_default_32" -> "598 dequantize_per_tensor_default_75" [label=None, style=solid]; +"597 quantize_per_tensor_default_32" -> "599 dequantize_per_tensor_default_74" [label=None, style=solid]; +"598 dequantize_per_tensor_default_75" -> "601 select_19" [label=None, style=solid]; +"599 dequantize_per_tensor_default_74" -> "600 select_18" [label=None, style=solid]; +"600 select_18" -> "603 view_48" [label="(197, 1, 768)", style=solid]; +"601 select_19" -> "605 view_49" [label="(197, 1, 768)", style=solid]; +"602 select_20" -> "607 view_50" [label="(197, 1, 768)", style=solid]; +"603 view_48" -> "604 transpose_38" [label="(197, 12, 64)", style=solid]; +"604 transpose_38" -> "609 view_51" [label="(12, 197, 64)", style=solid]; +"605 view_49" -> "606 transpose_39" [label="(197, 12, 64)", style=solid]; +"606 transpose_39" -> "610 view_52" [label="(12, 197, 64)", style=solid]; +"607 view_50" -> "608 transpose_40" [label="(197, 12, 64)", style=solid]; +"608 transpose_40" -> "611 view_53" [label="(12, 197, 64)", style=solid]; +"609 view_51" -> "612 scaled_dot_product_attention_6" [label="(1, 12, 197, 64)", style=solid]; +"610 view_52" -> "612 scaled_dot_product_attention_6" [label="(1, 12, 197, 64)", style=solid]; +"611 view_53" -> "612 scaled_dot_product_attention_6" [label="(1, 12, 197, 64)", style=solid]; +"612 scaled_dot_product_attention_6" -> "613 quantize_per_tensor_default_33" [label="(1, 12, 197, 64)", style=solid]; +"613 quantize_per_tensor_default_33" -> "614 dequantize_per_tensor_default_33" [label=None, style=solid]; +"614 dequantize_per_tensor_default_33" -> "615 permute_7" [label=None, style=solid]; +"615 permute_7" -> "616 view_54" [label="(197, 1, 12, 64)", style=solid]; +"616 view_54" -> "617 linear_25" [label="(197, 768)", style=solid]; +"617 linear_25" -> "618 view_55" [label="(197, 768)", style=solid]; +"618 view_55" -> "619 transpose_41" [label="(197, 1, 768)", style=solid]; +"619 transpose_41" -> "620 dropout_19" [label="(1, 197, 768)", style=solid]; +"620 dropout_19" -> "621 add_13" [label="(1, 197, 768)", style=solid]; +"621 add_13" -> "622 layer_norm_13" [label="(1, 197, 768)", style=solid]; +"621 add_13" -> "632 add_14" [label="(1, 197, 768)", style=solid]; +"622 layer_norm_13" -> "623 quantize_per_tensor_default_34" [label="(1, 197, 768)", style=solid]; +"623 quantize_per_tensor_default_34" -> "624 dequantize_per_tensor_default_34" [label=None, style=solid]; +"624 dequantize_per_tensor_default_34" -> "625 linear_26" [label=None, style=solid]; +"625 linear_26" -> "626 gelu_6" [label="(1, 197, 3072)", style=solid]; +"626 gelu_6" -> "627 quantize_per_tensor_default_35" [label="(1, 197, 3072)", style=solid]; +"627 quantize_per_tensor_default_35" -> "628 dequantize_per_tensor_default_35" [label=None, style=solid]; +"628 dequantize_per_tensor_default_35" -> "629 dropout_20" [label=None, style=solid]; +"629 dropout_20" -> "630 linear_27" [label="(1, 197, 3072)", style=solid]; +"630 linear_27" -> "631 dropout_21" [label="(1, 197, 768)", style=solid]; +"631 dropout_21" -> "632 add_14" [label="(1, 197, 768)", style=solid]; +"632 add_14" -> "633 layer_norm_14" [label="(1, 197, 768)", style=solid]; +"632 add_14" -> "667 add_15" [label="(1, 197, 768)", style=solid]; +"633 layer_norm_14" -> "634 quantize_per_tensor_default_36" [label="(1, 197, 768)", style=solid]; +"634 quantize_per_tensor_default_36" -> "635 dequantize_per_tensor_default_36" [label=None, style=solid]; +"635 dequantize_per_tensor_default_36" -> "636 transpose_42" [label=None, style=solid]; +"636 transpose_42" -> "637 linear_28" [label="(197, 1, 768)", style=solid]; +"637 linear_28" -> "638 unflatten_7" [label="(197, 1, 2304)", style=solid]; +"638 unflatten_7" -> "639 unsqueeze_7" [label="(197, 1, 3, 768)", style=solid]; +"639 unsqueeze_7" -> "640 transpose_43" [label="(1, 197, 1, 3, 768)", style=solid]; +"640 transpose_43" -> "641 squeeze_7" [label="(3, 197, 1, 1, 768)", style=solid]; +"641 squeeze_7" -> "642 contiguous_7" [label="(3, 197, 1, 768)", style=solid]; +"642 contiguous_7" -> "643 quantize_per_tensor_default_37" [label="(3, 197, 1, 768)", style=solid]; +"642 contiguous_7" -> "648 select_23" [label="(3, 197, 1, 768)", style=solid]; +"643 quantize_per_tensor_default_37" -> "644 dequantize_per_tensor_default_77" [label=None, style=solid]; +"643 quantize_per_tensor_default_37" -> "645 dequantize_per_tensor_default_76" [label=None, style=solid]; +"644 dequantize_per_tensor_default_77" -> "647 select_22" [label=None, style=solid]; +"645 dequantize_per_tensor_default_76" -> "646 select_21" [label=None, style=solid]; +"646 select_21" -> "649 view_56" [label="(197, 1, 768)", style=solid]; +"647 select_22" -> "651 view_57" [label="(197, 1, 768)", style=solid]; +"648 select_23" -> "653 view_58" [label="(197, 1, 768)", style=solid]; +"649 view_56" -> "650 transpose_44" [label="(197, 12, 64)", style=solid]; +"650 transpose_44" -> "655 view_59" [label="(12, 197, 64)", style=solid]; +"651 view_57" -> "652 transpose_45" [label="(197, 12, 64)", style=solid]; +"652 transpose_45" -> "656 view_60" [label="(12, 197, 64)", style=solid]; +"653 view_58" -> "654 transpose_46" [label="(197, 12, 64)", style=solid]; +"654 transpose_46" -> "657 view_61" [label="(12, 197, 64)", style=solid]; +"655 view_59" -> "658 scaled_dot_product_attention_7" [label="(1, 12, 197, 64)", style=solid]; +"656 view_60" -> "658 scaled_dot_product_attention_7" [label="(1, 12, 197, 64)", style=solid]; +"657 view_61" -> "658 scaled_dot_product_attention_7" [label="(1, 12, 197, 64)", style=solid]; +"658 scaled_dot_product_attention_7" -> "659 quantize_per_tensor_default_38" [label="(1, 12, 197, 64)", style=solid]; +"659 quantize_per_tensor_default_38" -> "660 dequantize_per_tensor_default_38" [label=None, style=solid]; +"660 dequantize_per_tensor_default_38" -> "661 permute_8" [label=None, style=solid]; +"661 permute_8" -> "662 view_62" [label="(197, 1, 12, 64)", style=solid]; +"662 view_62" -> "663 linear_29" [label="(197, 768)", style=solid]; +"663 linear_29" -> "664 view_63" [label="(197, 768)", style=solid]; +"664 view_63" -> "665 transpose_47" [label="(197, 1, 768)", style=solid]; +"665 transpose_47" -> "666 dropout_22" [label="(1, 197, 768)", style=solid]; +"666 dropout_22" -> "667 add_15" [label="(1, 197, 768)", style=solid]; +"667 add_15" -> "668 layer_norm_15" [label="(1, 197, 768)", style=solid]; +"667 add_15" -> "678 add_16" [label="(1, 197, 768)", style=solid]; +"668 layer_norm_15" -> "669 quantize_per_tensor_default_39" [label="(1, 197, 768)", style=solid]; +"669 quantize_per_tensor_default_39" -> "670 dequantize_per_tensor_default_39" [label=None, style=solid]; +"670 dequantize_per_tensor_default_39" -> "671 linear_30" [label=None, style=solid]; +"671 linear_30" -> "672 gelu_7" [label="(1, 197, 3072)", style=solid]; +"672 gelu_7" -> "673 quantize_per_tensor_default_40" [label="(1, 197, 3072)", style=solid]; +"673 quantize_per_tensor_default_40" -> "674 dequantize_per_tensor_default_40" [label=None, style=solid]; +"674 dequantize_per_tensor_default_40" -> "675 dropout_23" [label=None, style=solid]; +"675 dropout_23" -> "676 linear_31" [label="(1, 197, 3072)", style=solid]; +"676 linear_31" -> "677 dropout_24" [label="(1, 197, 768)", style=solid]; +"677 dropout_24" -> "678 add_16" [label="(1, 197, 768)", style=solid]; +"678 add_16" -> "679 layer_norm_16" [label="(1, 197, 768)", style=solid]; +"678 add_16" -> "713 add_17" [label="(1, 197, 768)", style=solid]; +"679 layer_norm_16" -> "680 quantize_per_tensor_default_41" [label="(1, 197, 768)", style=solid]; +"680 quantize_per_tensor_default_41" -> "681 dequantize_per_tensor_default_41" [label=None, style=solid]; +"681 dequantize_per_tensor_default_41" -> "682 transpose_48" [label=None, style=solid]; +"682 transpose_48" -> "683 linear_32" [label="(197, 1, 768)", style=solid]; +"683 linear_32" -> "684 unflatten_8" [label="(197, 1, 2304)", style=solid]; +"684 unflatten_8" -> "685 unsqueeze_8" [label="(197, 1, 3, 768)", style=solid]; +"685 unsqueeze_8" -> "686 transpose_49" [label="(1, 197, 1, 3, 768)", style=solid]; +"686 transpose_49" -> "687 squeeze_8" [label="(3, 197, 1, 1, 768)", style=solid]; +"687 squeeze_8" -> "688 contiguous_8" [label="(3, 197, 1, 768)", style=solid]; +"688 contiguous_8" -> "689 quantize_per_tensor_default_42" [label="(3, 197, 1, 768)", style=solid]; +"688 contiguous_8" -> "694 select_26" [label="(3, 197, 1, 768)", style=solid]; +"689 quantize_per_tensor_default_42" -> "690 dequantize_per_tensor_default_79" [label=None, style=solid]; +"689 quantize_per_tensor_default_42" -> "691 dequantize_per_tensor_default_78" [label=None, style=solid]; +"690 dequantize_per_tensor_default_79" -> "693 select_25" [label=None, style=solid]; +"691 dequantize_per_tensor_default_78" -> "692 select_24" [label=None, style=solid]; +"692 select_24" -> "695 view_64" [label="(197, 1, 768)", style=solid]; +"693 select_25" -> "697 view_65" [label="(197, 1, 768)", style=solid]; +"694 select_26" -> "699 view_66" [label="(197, 1, 768)", style=solid]; +"695 view_64" -> "696 transpose_50" [label="(197, 12, 64)", style=solid]; +"696 transpose_50" -> "701 view_67" [label="(12, 197, 64)", style=solid]; +"697 view_65" -> "698 transpose_51" [label="(197, 12, 64)", style=solid]; +"698 transpose_51" -> "702 view_68" [label="(12, 197, 64)", style=solid]; +"699 view_66" -> "700 transpose_52" [label="(197, 12, 64)", style=solid]; +"700 transpose_52" -> "703 view_69" [label="(12, 197, 64)", style=solid]; +"701 view_67" -> "704 scaled_dot_product_attention_8" [label="(1, 12, 197, 64)", style=solid]; +"702 view_68" -> "704 scaled_dot_product_attention_8" [label="(1, 12, 197, 64)", style=solid]; +"703 view_69" -> "704 scaled_dot_product_attention_8" [label="(1, 12, 197, 64)", style=solid]; +"704 scaled_dot_product_attention_8" -> "705 quantize_per_tensor_default_43" [label="(1, 12, 197, 64)", style=solid]; +"705 quantize_per_tensor_default_43" -> "706 dequantize_per_tensor_default_43" [label=None, style=solid]; +"706 dequantize_per_tensor_default_43" -> "707 permute_9" [label=None, style=solid]; +"707 permute_9" -> "708 view_70" [label="(197, 1, 12, 64)", style=solid]; +"708 view_70" -> "709 linear_33" [label="(197, 768)", style=solid]; +"709 linear_33" -> "710 view_71" [label="(197, 768)", style=solid]; +"710 view_71" -> "711 transpose_53" [label="(197, 1, 768)", style=solid]; +"711 transpose_53" -> "712 dropout_25" [label="(1, 197, 768)", style=solid]; +"712 dropout_25" -> "713 add_17" [label="(1, 197, 768)", style=solid]; +"713 add_17" -> "714 layer_norm_17" [label="(1, 197, 768)", style=solid]; +"713 add_17" -> "724 add_18" [label="(1, 197, 768)", style=solid]; +"714 layer_norm_17" -> "715 quantize_per_tensor_default_44" [label="(1, 197, 768)", style=solid]; +"715 quantize_per_tensor_default_44" -> "716 dequantize_per_tensor_default_44" [label=None, style=solid]; +"716 dequantize_per_tensor_default_44" -> "717 linear_34" [label=None, style=solid]; +"717 linear_34" -> "718 gelu_8" [label="(1, 197, 3072)", style=solid]; +"718 gelu_8" -> "719 quantize_per_tensor_default_45" [label="(1, 197, 3072)", style=solid]; +"719 quantize_per_tensor_default_45" -> "720 dequantize_per_tensor_default_45" [label=None, style=solid]; +"720 dequantize_per_tensor_default_45" -> "721 dropout_26" [label=None, style=solid]; +"721 dropout_26" -> "722 linear_35" [label="(1, 197, 3072)", style=solid]; +"722 linear_35" -> "723 dropout_27" [label="(1, 197, 768)", style=solid]; +"723 dropout_27" -> "724 add_18" [label="(1, 197, 768)", style=solid]; +"724 add_18" -> "725 layer_norm_18" [label="(1, 197, 768)", style=solid]; +"724 add_18" -> "759 add_19" [label="(1, 197, 768)", style=solid]; +"725 layer_norm_18" -> "726 quantize_per_tensor_default_46" [label="(1, 197, 768)", style=solid]; +"726 quantize_per_tensor_default_46" -> "727 dequantize_per_tensor_default_46" [label=None, style=solid]; +"727 dequantize_per_tensor_default_46" -> "728 transpose_54" [label=None, style=solid]; +"728 transpose_54" -> "729 linear_36" [label="(197, 1, 768)", style=solid]; +"729 linear_36" -> "730 unflatten_9" [label="(197, 1, 2304)", style=solid]; +"730 unflatten_9" -> "731 unsqueeze_9" [label="(197, 1, 3, 768)", style=solid]; +"731 unsqueeze_9" -> "732 transpose_55" [label="(1, 197, 1, 3, 768)", style=solid]; +"732 transpose_55" -> "733 squeeze_9" [label="(3, 197, 1, 1, 768)", style=solid]; +"733 squeeze_9" -> "734 contiguous_9" [label="(3, 197, 1, 768)", style=solid]; +"734 contiguous_9" -> "735 quantize_per_tensor_default_47" [label="(3, 197, 1, 768)", style=solid]; +"734 contiguous_9" -> "740 select_29" [label="(3, 197, 1, 768)", style=solid]; +"735 quantize_per_tensor_default_47" -> "736 dequantize_per_tensor_default_81" [label=None, style=solid]; +"735 quantize_per_tensor_default_47" -> "737 dequantize_per_tensor_default_80" [label=None, style=solid]; +"736 dequantize_per_tensor_default_81" -> "739 select_28" [label=None, style=solid]; +"737 dequantize_per_tensor_default_80" -> "738 select_27" [label=None, style=solid]; +"738 select_27" -> "741 view_72" [label="(197, 1, 768)", style=solid]; +"739 select_28" -> "743 view_73" [label="(197, 1, 768)", style=solid]; +"740 select_29" -> "745 view_74" [label="(197, 1, 768)", style=solid]; +"741 view_72" -> "742 transpose_56" [label="(197, 12, 64)", style=solid]; +"742 transpose_56" -> "747 view_75" [label="(12, 197, 64)", style=solid]; +"743 view_73" -> "744 transpose_57" [label="(197, 12, 64)", style=solid]; +"744 transpose_57" -> "748 view_76" [label="(12, 197, 64)", style=solid]; +"745 view_74" -> "746 transpose_58" [label="(197, 12, 64)", style=solid]; +"746 transpose_58" -> "749 view_77" [label="(12, 197, 64)", style=solid]; +"747 view_75" -> "750 scaled_dot_product_attention_9" [label="(1, 12, 197, 64)", style=solid]; +"748 view_76" -> "750 scaled_dot_product_attention_9" [label="(1, 12, 197, 64)", style=solid]; +"749 view_77" -> "750 scaled_dot_product_attention_9" [label="(1, 12, 197, 64)", style=solid]; +"750 scaled_dot_product_attention_9" -> "751 quantize_per_tensor_default_48" [label="(1, 12, 197, 64)", style=solid]; +"751 quantize_per_tensor_default_48" -> "752 dequantize_per_tensor_default_48" [label=None, style=solid]; +"752 dequantize_per_tensor_default_48" -> "753 permute_10" [label=None, style=solid]; +"753 permute_10" -> "754 view_78" [label="(197, 1, 12, 64)", style=solid]; +"754 view_78" -> "755 linear_37" [label="(197, 768)", style=solid]; +"755 linear_37" -> "756 view_79" [label="(197, 768)", style=solid]; +"756 view_79" -> "757 transpose_59" [label="(197, 1, 768)", style=solid]; +"757 transpose_59" -> "758 dropout_28" [label="(1, 197, 768)", style=solid]; +"758 dropout_28" -> "759 add_19" [label="(1, 197, 768)", style=solid]; +"759 add_19" -> "760 layer_norm_19" [label="(1, 197, 768)", style=solid]; +"759 add_19" -> "770 add_20" [label="(1, 197, 768)", style=solid]; +"760 layer_norm_19" -> "761 quantize_per_tensor_default_49" [label="(1, 197, 768)", style=solid]; +"761 quantize_per_tensor_default_49" -> "762 dequantize_per_tensor_default_49" [label=None, style=solid]; +"762 dequantize_per_tensor_default_49" -> "763 linear_38" [label=None, style=solid]; +"763 linear_38" -> "764 gelu_9" [label="(1, 197, 3072)", style=solid]; +"764 gelu_9" -> "765 quantize_per_tensor_default_50" [label="(1, 197, 3072)", style=solid]; +"765 quantize_per_tensor_default_50" -> "766 dequantize_per_tensor_default_50" [label=None, style=solid]; +"766 dequantize_per_tensor_default_50" -> "767 dropout_29" [label=None, style=solid]; +"767 dropout_29" -> "768 linear_39" [label="(1, 197, 3072)", style=solid]; +"768 linear_39" -> "769 dropout_30" [label="(1, 197, 768)", style=solid]; +"769 dropout_30" -> "770 add_20" [label="(1, 197, 768)", style=solid]; +"770 add_20" -> "771 layer_norm_20" [label="(1, 197, 768)", style=solid]; +"770 add_20" -> "805 add_21" [label="(1, 197, 768)", style=solid]; +"771 layer_norm_20" -> "772 quantize_per_tensor_default_51" [label="(1, 197, 768)", style=solid]; +"772 quantize_per_tensor_default_51" -> "773 dequantize_per_tensor_default_51" [label=None, style=solid]; +"773 dequantize_per_tensor_default_51" -> "774 transpose_60" [label=None, style=solid]; +"774 transpose_60" -> "775 linear_40" [label="(197, 1, 768)", style=solid]; +"775 linear_40" -> "776 unflatten_10" [label="(197, 1, 2304)", style=solid]; +"776 unflatten_10" -> "777 unsqueeze_10" [label="(197, 1, 3, 768)", style=solid]; +"777 unsqueeze_10" -> "778 transpose_61" [label="(1, 197, 1, 3, 768)", style=solid]; +"778 transpose_61" -> "779 squeeze_10" [label="(3, 197, 1, 1, 768)", style=solid]; +"779 squeeze_10" -> "780 contiguous_10" [label="(3, 197, 1, 768)", style=solid]; +"780 contiguous_10" -> "781 quantize_per_tensor_default_52" [label="(3, 197, 1, 768)", style=solid]; +"780 contiguous_10" -> "786 select_32" [label="(3, 197, 1, 768)", style=solid]; +"781 quantize_per_tensor_default_52" -> "782 dequantize_per_tensor_default_83" [label=None, style=solid]; +"781 quantize_per_tensor_default_52" -> "783 dequantize_per_tensor_default_82" [label=None, style=solid]; +"782 dequantize_per_tensor_default_83" -> "785 select_31" [label=None, style=solid]; +"783 dequantize_per_tensor_default_82" -> "784 select_30" [label=None, style=solid]; +"784 select_30" -> "787 view_80" [label="(197, 1, 768)", style=solid]; +"785 select_31" -> "789 view_81" [label="(197, 1, 768)", style=solid]; +"786 select_32" -> "791 view_82" [label="(197, 1, 768)", style=solid]; +"787 view_80" -> "788 transpose_62" [label="(197, 12, 64)", style=solid]; +"788 transpose_62" -> "793 view_83" [label="(12, 197, 64)", style=solid]; +"789 view_81" -> "790 transpose_63" [label="(197, 12, 64)", style=solid]; +"790 transpose_63" -> "794 view_84" [label="(12, 197, 64)", style=solid]; +"791 view_82" -> "792 transpose_64" [label="(197, 12, 64)", style=solid]; +"792 transpose_64" -> "795 view_85" [label="(12, 197, 64)", style=solid]; +"793 view_83" -> "796 scaled_dot_product_attention_10" [label="(1, 12, 197, 64)", style=solid]; +"794 view_84" -> "796 scaled_dot_product_attention_10" [label="(1, 12, 197, 64)", style=solid]; +"795 view_85" -> "796 scaled_dot_product_attention_10" [label="(1, 12, 197, 64)", style=solid]; +"796 scaled_dot_product_attention_10" -> "797 quantize_per_tensor_default_53" [label="(1, 12, 197, 64)", style=solid]; +"797 quantize_per_tensor_default_53" -> "798 dequantize_per_tensor_default_53" [label=None, style=solid]; +"798 dequantize_per_tensor_default_53" -> "799 permute_11" [label=None, style=solid]; +"799 permute_11" -> "800 view_86" [label="(197, 1, 12, 64)", style=solid]; +"800 view_86" -> "801 linear_41" [label="(197, 768)", style=solid]; +"801 linear_41" -> "802 view_87" [label="(197, 768)", style=solid]; +"802 view_87" -> "803 transpose_65" [label="(197, 1, 768)", style=solid]; +"803 transpose_65" -> "804 dropout_31" [label="(1, 197, 768)", style=solid]; +"804 dropout_31" -> "805 add_21" [label="(1, 197, 768)", style=solid]; +"805 add_21" -> "806 layer_norm_21" [label="(1, 197, 768)", style=solid]; +"805 add_21" -> "816 add_22" [label="(1, 197, 768)", style=solid]; +"806 layer_norm_21" -> "807 quantize_per_tensor_default_54" [label="(1, 197, 768)", style=solid]; +"807 quantize_per_tensor_default_54" -> "808 dequantize_per_tensor_default_54" [label=None, style=solid]; +"808 dequantize_per_tensor_default_54" -> "809 linear_42" [label=None, style=solid]; +"809 linear_42" -> "810 gelu_10" [label="(1, 197, 3072)", style=solid]; +"810 gelu_10" -> "811 quantize_per_tensor_default_55" [label="(1, 197, 3072)", style=solid]; +"811 quantize_per_tensor_default_55" -> "812 dequantize_per_tensor_default_55" [label=None, style=solid]; +"812 dequantize_per_tensor_default_55" -> "813 dropout_32" [label=None, style=solid]; +"813 dropout_32" -> "814 linear_43" [label="(1, 197, 3072)", style=solid]; +"814 linear_43" -> "815 dropout_33" [label="(1, 197, 768)", style=solid]; +"815 dropout_33" -> "816 add_22" [label="(1, 197, 768)", style=solid]; +"816 add_22" -> "817 layer_norm_22" [label="(1, 197, 768)", style=solid]; +"816 add_22" -> "851 add_23" [label="(1, 197, 768)", style=solid]; +"817 layer_norm_22" -> "818 quantize_per_tensor_default_56" [label="(1, 197, 768)", style=solid]; +"818 quantize_per_tensor_default_56" -> "819 dequantize_per_tensor_default_56" [label=None, style=solid]; +"819 dequantize_per_tensor_default_56" -> "820 transpose_66" [label=None, style=solid]; +"820 transpose_66" -> "821 linear_44" [label="(197, 1, 768)", style=solid]; +"821 linear_44" -> "822 unflatten_11" [label="(197, 1, 2304)", style=solid]; +"822 unflatten_11" -> "823 unsqueeze_11" [label="(197, 1, 3, 768)", style=solid]; +"823 unsqueeze_11" -> "824 transpose_67" [label="(1, 197, 1, 3, 768)", style=solid]; +"824 transpose_67" -> "825 squeeze_11" [label="(3, 197, 1, 1, 768)", style=solid]; +"825 squeeze_11" -> "826 contiguous_11" [label="(3, 197, 1, 768)", style=solid]; +"826 contiguous_11" -> "827 quantize_per_tensor_default_57" [label="(3, 197, 1, 768)", style=solid]; +"826 contiguous_11" -> "832 select_35" [label="(3, 197, 1, 768)", style=solid]; +"827 quantize_per_tensor_default_57" -> "828 dequantize_per_tensor_default_85" [label=None, style=solid]; +"827 quantize_per_tensor_default_57" -> "829 dequantize_per_tensor_default_84" [label=None, style=solid]; +"828 dequantize_per_tensor_default_85" -> "831 select_34" [label=None, style=solid]; +"829 dequantize_per_tensor_default_84" -> "830 select_33" [label=None, style=solid]; +"830 select_33" -> "833 view_88" [label="(197, 1, 768)", style=solid]; +"831 select_34" -> "835 view_89" [label="(197, 1, 768)", style=solid]; +"832 select_35" -> "837 view_90" [label="(197, 1, 768)", style=solid]; +"833 view_88" -> "834 transpose_68" [label="(197, 12, 64)", style=solid]; +"834 transpose_68" -> "839 view_91" [label="(12, 197, 64)", style=solid]; +"835 view_89" -> "836 transpose_69" [label="(197, 12, 64)", style=solid]; +"836 transpose_69" -> "840 view_92" [label="(12, 197, 64)", style=solid]; +"837 view_90" -> "838 transpose_70" [label="(197, 12, 64)", style=solid]; +"838 transpose_70" -> "841 view_93" [label="(12, 197, 64)", style=solid]; +"839 view_91" -> "842 scaled_dot_product_attention_11" [label="(1, 12, 197, 64)", style=solid]; +"840 view_92" -> "842 scaled_dot_product_attention_11" [label="(1, 12, 197, 64)", style=solid]; +"841 view_93" -> "842 scaled_dot_product_attention_11" [label="(1, 12, 197, 64)", style=solid]; +"842 scaled_dot_product_attention_11" -> "843 quantize_per_tensor_default_58" [label="(1, 12, 197, 64)", style=solid]; +"843 quantize_per_tensor_default_58" -> "844 dequantize_per_tensor_default_58" [label=None, style=solid]; +"844 dequantize_per_tensor_default_58" -> "845 permute_12" [label=None, style=solid]; +"845 permute_12" -> "846 view_94" [label="(197, 1, 12, 64)", style=solid]; +"846 view_94" -> "847 linear_45" [label="(197, 768)", style=solid]; +"847 linear_45" -> "848 view_95" [label="(197, 768)", style=solid]; +"848 view_95" -> "849 transpose_71" [label="(197, 1, 768)", style=solid]; +"849 transpose_71" -> "850 dropout_34" [label="(1, 197, 768)", style=solid]; +"850 dropout_34" -> "851 add_23" [label="(1, 197, 768)", style=solid]; +"851 add_23" -> "852 layer_norm_23" [label="(1, 197, 768)", style=solid]; +"851 add_23" -> "862 add_24" [label="(1, 197, 768)", style=solid]; +"852 layer_norm_23" -> "853 quantize_per_tensor_default_59" [label="(1, 197, 768)", style=solid]; +"853 quantize_per_tensor_default_59" -> "854 dequantize_per_tensor_default_59" [label=None, style=solid]; +"854 dequantize_per_tensor_default_59" -> "855 linear_46" [label=None, style=solid]; +"855 linear_46" -> "856 gelu_11" [label="(1, 197, 3072)", style=solid]; +"856 gelu_11" -> "857 quantize_per_tensor_default_60" [label="(1, 197, 3072)", style=solid]; +"857 quantize_per_tensor_default_60" -> "858 dequantize_per_tensor_default_60" [label=None, style=solid]; +"858 dequantize_per_tensor_default_60" -> "859 dropout_35" [label=None, style=solid]; +"859 dropout_35" -> "860 linear_47" [label="(1, 197, 3072)", style=solid]; +"860 linear_47" -> "861 dropout_36" [label="(1, 197, 768)", style=solid]; +"861 dropout_36" -> "862 add_24" [label="(1, 197, 768)", style=solid]; +"862 add_24" -> "863 layer_norm_24" [label="(1, 197, 768)", style=solid]; +"863 layer_norm_24" -> "864 quantize_per_tensor_default_61" [label="(1, 197, 768)", style=solid]; +"864 quantize_per_tensor_default_61" -> "865 dequantize_per_tensor_default_61" [label=None, style=solid]; +"865 dequantize_per_tensor_default_61" -> "866 slice_1" [label=None, style=solid]; +"866 slice_1" -> "867 select_36" [label="(1, 197, 768)", style=solid]; +"867 select_36" -> "868 linear_48" [label="(1, 768)", style=solid]; +"868 linear_48" -> "869 output_1" [label="(1, 1000)", style=solid]; } From 01dfd6d8c2f8f94d9d6fa67ac6e438d4654a9277 Mon Sep 17 00:00:00 2001 From: dlyakhov Date: Mon, 2 Dec 2024 19:30:08 +0100 Subject: [PATCH 11/15] Conformance: nncf.quantize_pt2e and OpenVINOQuantize support --- tests/post_training/conftest.py | 5 + .../data/ptq_reference_data.yaml | 32 +++++ tests/post_training/model_scope.py | 9 +- tests/post_training/pipelines/base.py | 17 ++- .../pipelines/image_classification_base.py | 132 ++++++++++++++++-- .../image_classification_torchvision.py | 7 +- .../test_quantize_conformance.py | 7 + 7 files changed, 190 insertions(+), 19 deletions(-) diff --git a/tests/post_training/conftest.py b/tests/post_training/conftest.py index a05d20d0d5c..6bdae739ff6 100644 --- a/tests/post_training/conftest.py +++ b/tests/post_training/conftest.py @@ -19,6 +19,11 @@ def pytest_addoption(parser): parser.addoption("--fp32", action="store_true", help="Test original model") parser.addoption("--cuda", action="store_true", help="Enable CUDA_TORCH backend") parser.addoption("--benchmark", action="store_true", help="Run benchmark_app") + parser.addoption( + "--validate-in-backend", + action="store_true", + help="Validate quantized model in native backend, not in openvino.", + ) parser.addoption( "--extra-columns", action="store_true", diff --git a/tests/post_training/data/ptq_reference_data.yaml b/tests/post_training/data/ptq_reference_data.yaml index 94f70f3a931..1a7b2daa2ef 100644 --- a/tests/post_training/data/ptq_reference_data.yaml +++ b/tests/post_training/data/ptq_reference_data.yaml @@ -38,6 +38,14 @@ torchvision/resnet18_backend_CUDA_TORCH: metric_value: 0.69152 torchvision/resnet18_backend_FX_TORCH: metric_value: 0.6946 +torchvision/resnet18_backend_OV_QUANTIZER_NNCF: + metric_value: 0.6946 +torchvision/resnet18_backend_OV_QUANTIZER_AO: + metric_value: 0.6946 +torchvision/resnet18_backend_X86_QUANTIZER_NNCF: + metric_value: 0.6946 +torchvision/resnet18_backend_X86_QUANTIZER_AO: + metric_value: 0.6946 torchvision/mobilenet_v3_small_BC_backend_FP32: metric_value: 0.6766 torchvision/mobilenet_v3_small_BC_backend_OV: @@ -46,18 +54,42 @@ torchvision/mobilenet_v3_small_BC_backend_ONNX: metric_value: 0.6679 torchvision/mobilenet_v3_small_BC_backend_FX_TORCH: metric_value: 0.6679 +torchvision/mobilenet_v3_small_BC_backend_OV_QUANTIZER_NNCF: + metric_value: 0.6679 +torchvision/mobilenet_v3_small_BC_backend_OV_QUANTIZER_AO: + metric_value: 0.6679 +torchvision/mobilenet_v3_small_BC_backend_X86_QUANTIZER_NNCF: + metric_value: 0.6679 +torchvision/mobilenet_v3_small_BC_backend_X86_QUANTIZER_AO: + metric_value: 0.6679 torchvision/vit_b_16_backend_FP32: metric_value: 0.8107 torchvision/vit_b_16_backend_OV: metric_value: 0.80948 torchvision/vit_b_16_backend_FX_TORCH: metric_value: 0.80922 +torchvision/vit_b_16_backend_OV_QUANTIZER_NNCF: + metric_value: 0.80922 +torchvision/vit_b_16_backend_OV_QUANTIZER_AO: + metric_value: 0.80922 +torchvision/vit_b_16_backend_X86_QUANTIZER_NNCF: + metric_value: 0.80922 +torchvision/vit_b_16_backend_X86_QUANTIZER_AO: + metric_value: 0.80922 torchvision/swin_v2_s_backend_FP32: metric_value: 0.83712 torchvision/swin_v2_s_backend_OV: metric_value: 0.83638 torchvision/swin_v2_s_backend_FX_TORCH: metric_value: 0.8360 +torchvision/swin_v2_s_backend_OV_QUANTIZER_NNCF: + metric_value: 0.8360 +torchvision/swin_v2_s_backend_OV_QUANTIZER_AO: + metric_value: 0.8360 +torchvision/swin_v2_s_backend_X86_QUANTIZER_NNCF: + metric_value: 0.8360 +torchvision/swin_v2_s_backend_X86_QUANTIZER_AO: + metric_value: 0.8360 timm/crossvit_9_240_backend_CUDA_TORCH: metric_value: 0.7275 timm/crossvit_9_240_backend_FP32: diff --git a/tests/post_training/model_scope.py b/tests/post_training/model_scope.py index 54b49d63a21..42b9b16610e 100644 --- a/tests/post_training/model_scope.py +++ b/tests/post_training/model_scope.py @@ -24,6 +24,7 @@ from nncf.quantization.advanced_parameters import AdvancedScaleEstimationParameters from nncf.quantization.advanced_parameters import AdvancedSmoothQuantParameters from tests.post_training.pipelines.base import ALL_PTQ_BACKENDS +from tests.post_training.pipelines.base import FX_BACKENDS from tests.post_training.pipelines.base import NNCF_PTQ_BACKENDS from tests.post_training.pipelines.base import BackendType from tests.post_training.pipelines.causal_language_model import CausalLMHF @@ -87,7 +88,7 @@ "model_id": "resnet18", "pipeline_cls": ImageClassificationTorchvision, "compression_params": {}, - "backends": [BackendType.FX_TORCH, BackendType.TORCH, BackendType.CUDA_TORCH, BackendType.OV, BackendType.ONNX], + "backends": FX_BACKENDS + [BackendType.TORCH, BackendType.CUDA_TORCH, BackendType.OV, BackendType.ONNX], "batch_size": 128, }, { @@ -98,7 +99,7 @@ "fast_bias_correction": False, "preset": QuantizationPreset.MIXED, }, - "backends": [BackendType.FX_TORCH, BackendType.OV, BackendType.ONNX], + "backends": FX_BACKENDS + [BackendType.OV, BackendType.ONNX], "batch_size": 128, }, { @@ -109,7 +110,7 @@ "model_type": ModelType.TRANSFORMER, "advanced_parameters": AdvancedQuantizationParameters(smooth_quant_alpha=0.15), }, - "backends": [BackendType.FX_TORCH, BackendType.OV], + "backends": FX_BACKENDS + [BackendType.OV], "batch_size": 1, }, { @@ -120,7 +121,7 @@ "model_type": ModelType.TRANSFORMER, "advanced_parameters": AdvancedQuantizationParameters(smooth_quant_alpha=0.5), }, - "backends": [BackendType.FX_TORCH, BackendType.OV], + "backends": FX_BACKENDS + [BackendType.OV], "batch_size": 1, }, # Timm models diff --git a/tests/post_training/pipelines/base.py b/tests/post_training/pipelines/base.py index 41e9bcf17e4..a949387a497 100644 --- a/tests/post_training/pipelines/base.py +++ b/tests/post_training/pipelines/base.py @@ -44,6 +44,10 @@ class BackendType(Enum): TORCH = "TORCH" CUDA_TORCH = "CUDA_TORCH" FX_TORCH = "FX_TORCH" + OV_QUANTIZER_NNCF = "OV_QUANTIZER_NNCF" + OV_QUANTIZER_AO = "OV_QUANTIZER_AO" + X86_QUANTIZER_NNCF = "X86_QUANTIZER_NNCF" + X86_QUANTIZER_AO = "X86_QUANTIZER_AO" ONNX = "ONNX" OV = "OV" OPTIMUM = "OPTIMUM" @@ -52,6 +56,13 @@ class BackendType(Enum): NNCF_PTQ_BACKENDS = [BackendType.TORCH, BackendType.CUDA_TORCH, BackendType.ONNX, BackendType.OV] ALL_PTQ_BACKENDS = NNCF_PTQ_BACKENDS PT_BACKENDS = [BackendType.TORCH, BackendType.CUDA_TORCH] +FX_BACKENDS = [ + BackendType.FX_TORCH, + BackendType.OV_QUANTIZER_NNCF, + BackendType.OV_QUANTIZER_AO, + BackendType.X86_QUANTIZER_NNCF, + BackendType.X86_QUANTIZER_AO, +] OV_BACKENDS = [BackendType.OV, BackendType.OPTIMUM] LIMIT_LENGTH_OF_STATUS = 120 @@ -211,6 +222,7 @@ def __init__( reference_data: dict, no_eval: bool, run_benchmark_app: bool, + validate_in_backend: bool = False, params: dict = None, batch_size: int = 1, memory_monitor: bool = False, @@ -227,6 +239,7 @@ def __init__( self.memory_monitor = memory_monitor self.no_eval = no_eval self.run_benchmark_app = run_benchmark_app + self.validate_in_backend = validate_in_backend self.output_model_dir: Path = self.output_dir / self.reported_name / self.backend.value self.output_model_dir.mkdir(parents=True, exist_ok=True) self.model_name = f"{self.reported_name}_{self.backend.value}" @@ -405,8 +418,8 @@ def save_compressed_model(self) -> None: ) self.path_compressed_ir = self.output_model_dir / "model.xml" ov.serialize(ov_model, self.path_compressed_ir) - elif self.backend == BackendType.FX_TORCH: - exported_model = torch.export.export(self.compressed_model, (self.dummy_tensor,)) + elif self.backend in FX_BACKENDS: + exported_model = torch.export.export(self.model, (self.dummy_tensor,)) ov_model = ov.convert_model(exported_model, example_input=self.dummy_tensor.cpu(), input=self.input_size) self.path_compressed_ir = self.output_model_dir / "model.xml" ov.serialize(ov_model, self.path_compressed_ir) diff --git a/tests/post_training/pipelines/image_classification_base.py b/tests/post_training/pipelines/image_classification_base.py index 22e60a5ae3b..c28c2762474 100644 --- a/tests/post_training/pipelines/image_classification_base.py +++ b/tests/post_training/pipelines/image_classification_base.py @@ -12,15 +12,27 @@ import copy import os +os.environ["TORCHINDUCTOR_FREEZING"] = "1" + +from itertools import islice + import numpy as np import openvino as ov import torch from sklearn.metrics import accuracy_score +from torch.ao.quantization.quantize_pt2e import convert_pt2e +from torch.ao.quantization.quantize_pt2e import prepare_pt2e +from torch.ao.quantization.quantizer.x86_inductor_quantizer import X86InductorQuantizer +from torch.ao.quantization.quantizer.x86_inductor_quantizer import get_default_x86_inductor_quantization_config from torchvision import datasets import nncf from nncf.common.logging.track_progress import track +from nncf.experimental.common.quantization.algorithms.quantizer.openvino_quantizer import OpenVINOQuantizer +from nncf.experimental.torch.fx.quantization.quantize_pt2e import quantize_pt2e from tests.post_training.pipelines.base import DEFAULT_VAL_THREADS +from tests.post_training.pipelines.base import FX_BACKENDS +from tests.post_training.pipelines.base import BackendType from tests.post_training.pipelines.base import PTQTestPipeline @@ -33,18 +45,15 @@ def prepare_calibration_dataset(self): self.calibration_dataset = nncf.Dataset(loader, self.get_transform_calibration_fn()) - def _validate(self): - val_dataset = datasets.ImageFolder(root=self.data_dir / "imagenet" / "val", transform=self.transform) - val_loader = torch.utils.data.DataLoader(val_dataset, batch_size=1, num_workers=2, shuffle=False) - - dataset_size = len(val_loader) - - # Initialize result tensors for async inference support. - predictions = np.zeros((dataset_size)) - references = -1 * np.ones((dataset_size)) + def _validate_ov( + self, + val_loader: torch.utils.data.DataLoader, + predictions: np.ndarray, + references: np.ndarray, + dataset_size: int, + ): core = ov.Core() - if os.environ.get("INFERENCE_NUM_THREADS"): # Set CPU_THREADS_NUM for OpenVINO inference inference_num_threads = os.environ.get("INFERENCE_NUM_THREADS") @@ -73,8 +82,111 @@ def process_result(request, userdata): references[i] = target infer_queue.wait_all() + return predictions, references + + def _validate_torch_compile( + self, val_loader: torch.utils.data.DataLoader, predictions: np.ndarray, references: np.ndarray + ): + # compiled_model = torch.compile(self.compressed_model, backend="openvino") + q_num = 0 + for node in self.compressed_model.graph.nodes: + if ".quantize_per" in str(node.target): + q_num += 1 + + print(f"Qunatize ops num: {q_num}") + + if self.backend in [BackendType.X86_QUANTIZER_AO, BackendType.X86_QUANTIZER_NNCF]: + compiled_model = torch.compile(self.compressed_model) + else: + compiled_model = torch.compile(self.compressed_model, backend="openvino") + + for i, (images, target) in enumerate(val_loader): + # W/A for memory leaks when using torch DataLoader and OpenVINO + pred = compiled_model(images) + pred = torch.argmax(pred, dim=1) + predictions[i] = pred.numpy() + references[i] = target.numpy() + return predictions, references + + def _validate(self): + val_dataset = datasets.ImageFolder(root=self.data_dir / "imagenet" / "val", transform=self.transform) + val_loader = torch.utils.data.DataLoader(val_dataset, batch_size=1, num_workers=2, shuffle=False) + + dataset_size = len(val_loader) + + # Initialize result tensors for async inference support. + predictions = np.zeros((dataset_size)) + references = -1 * np.ones((dataset_size)) + + if self.backend in FX_BACKENDS: + predictions, references = self._validate_torch_compile(val_loader, predictions, references) + else: + predictions, references = self._validate_ov(val_loader, predictions, references, dataset_size) acc_top1 = accuracy_score(predictions, references) self.run_info.metric_name = "Acc@1" self.run_info.metric_value = acc_top1 + + def _compress_torch_ao(self, quantizer): + + prepared_model = prepare_pt2e(self.model, quantizer) + subset_size = self.compression_params.get("subset_size", 300) + for data in islice(self.calibration_dataset.get_inference_data(), subset_size): + prepared_model(data) + self.compressed_model = convert_pt2e(prepared_model) + + def _compress_nncf_pt2e(self, quantizer): + pt2e_kwargs = {} + for key in ( + "subset_size", + "fast_bias_correction", + "smooth_quant", + "bias_correction_params", + "smooth_quant_params", + "activations_range_estimator_params", + "weights_range_estimator_params", + ): + if key in self.compression_params: + pt2e_kwargs[key] = self.compression_params[key] + smooth_quant = False + if self.compression_params.get("model_type", False): + smooth_quant = self.compression_params["model_type"] == nncf.ModelType.TRANSFORMER + self.compressed_model = quantize_pt2e( + self.model, quantizer, self.calibration_dataset, smooth_quant=smooth_quant, fold_quantize=False + ) + + def _compress(self): + """ + Quantize self.model + """ + if self.backend not in FX_BACKENDS or self.backend == BackendType.FX_TORCH: + super()._compress() + return + + if self.backend in [BackendType.OV_QUANTIZER_AO, BackendType.OV_QUANTIZER_NNCF]: + quantizer_kwargs = {} + for key in ( + "mode", + "preset", + "target_device", + "model_type", + "ignored_scope", + "overflow_fix", + "quantize_outputs", + "activations_quantization_params", + "weights_quantization_params", + "quantizer_propagation_rule", + ): + if key in self.compression_params: + quantizer_kwargs[key] = self.compression_params[key] + quantizer = OpenVINOQuantizer(**quantizer_kwargs) + else: + + quantizer = X86InductorQuantizer() + quantizer.set_global(get_default_x86_inductor_quantization_config()) + + if self.backend in [BackendType.OV_QUANTIZER_NNCF, BackendType.X86_QUANTIZER_NNCF]: + self._compress_nncf_pt2e(quantizer) + else: + self._compress_torch_ao(quantizer) diff --git a/tests/post_training/pipelines/image_classification_torchvision.py b/tests/post_training/pipelines/image_classification_torchvision.py index eb2ffeb96a0..2c07752fce5 100644 --- a/tests/post_training/pipelines/image_classification_torchvision.py +++ b/tests/post_training/pipelines/image_classification_torchvision.py @@ -19,6 +19,7 @@ from torchvision import models from nncf.torch import disable_patching +from tests.post_training.pipelines.base import FX_BACKENDS from tests.post_training.pipelines.base import PT_BACKENDS from tests.post_training.pipelines.base import BackendType from tests.post_training.pipelines.image_classification_base import ImageClassificationBase @@ -74,7 +75,7 @@ def prepare_model(self) -> None: if self.batch_size > 1: # Dynamic batch_size shape export self.input_size[0] = -1 - if self.backend == BackendType.FX_TORCH: + if self.backend in FX_BACKENDS: with torch.no_grad(): with disable_patching(): self.model = self.model_params.export_fn(model, (self.dummy_tensor,)) @@ -120,7 +121,7 @@ def _dump_model_fp32(self) -> None: ) ov.serialize(ov_model, self.fp32_model_dir / "model_fp32.xml") - if self.backend == BackendType.FX_TORCH: + if self.backend in FX_BACKENDS: exported_model = torch.export.export(self.model, (self.dummy_tensor,)) ov_model = ov.convert_model(exported_model, example_input=self.dummy_tensor, input=self.input_size) ov.serialize(ov_model, self.fp32_model_dir / "fx_model_fp32.xml") @@ -132,7 +133,7 @@ def prepare_preprocessor(self) -> None: self.transform = self.model_params.weights.transforms() def get_transform_calibration_fn(self): - if self.backend in [BackendType.FX_TORCH] + PT_BACKENDS: + if self.backend in FX_BACKENDS + PT_BACKENDS: device = torch.device("cuda" if self.backend == BackendType.CUDA_TORCH else "cpu") def transform_fn(data_item): diff --git a/tests/post_training/test_quantize_conformance.py b/tests/post_training/test_quantize_conformance.py index 20504a8b086..075b838e6d7 100644 --- a/tests/post_training/test_quantize_conformance.py +++ b/tests/post_training/test_quantize_conformance.py @@ -75,6 +75,11 @@ def fixture_run_benchmark_app(pytestconfig): return pytestconfig.getoption("benchmark") +@pytest.fixture(scope="session", name="validate_in_backend") +def fixture_validate_in_backend(pytestconfig): + return pytestconfig.getoption("validate_in_backend") + + @pytest.fixture(scope="session", name="extra_columns") def fixture_extra_columns(pytestconfig): return pytestconfig.getoption("extra_columns") @@ -266,6 +271,7 @@ def test_ptq_quantization( run_torch_cuda_backend: bool, subset_size: Optional[int], run_benchmark_app: bool, + validate_in_backend: bool, capsys: pytest.CaptureFixture, extra_columns: bool, memory_monitor: bool, @@ -293,6 +299,7 @@ def test_ptq_quantization( "data_dir": data_dir, "no_eval": no_eval, "run_benchmark_app": run_benchmark_app, + "validate_in_backend": validate_in_backend, "batch_size": batch_size, "memory_monitor": memory_monitor, } From 117cb9324f2b789e7ad8d7b09d0cc412a0c157e2 Mon Sep 17 00:00:00 2001 From: dlyakhov Date: Tue, 3 Dec 2024 15:19:07 +0100 Subject: [PATCH 12/15] No grad during the TorchFX model validation --- .../pipelines/image_classification_base.py | 55 ++++++++++++------- 1 file changed, 34 insertions(+), 21 deletions(-) diff --git a/tests/post_training/pipelines/image_classification_base.py b/tests/post_training/pipelines/image_classification_base.py index c28c2762474..6c649862b8c 100644 --- a/tests/post_training/pipelines/image_classification_base.py +++ b/tests/post_training/pipelines/image_classification_base.py @@ -30,6 +30,7 @@ from nncf.common.logging.track_progress import track from nncf.experimental.common.quantization.algorithms.quantizer.openvino_quantizer import OpenVINOQuantizer from nncf.experimental.torch.fx.quantization.quantize_pt2e import quantize_pt2e +from nncf.torch import disable_patching from tests.post_training.pipelines.base import DEFAULT_VAL_THREADS from tests.post_training.pipelines.base import FX_BACKENDS from tests.post_training.pipelines.base import BackendType @@ -95,18 +96,20 @@ def _validate_torch_compile( print(f"Qunatize ops num: {q_num}") - if self.backend in [BackendType.X86_QUANTIZER_AO, BackendType.X86_QUANTIZER_NNCF]: - compiled_model = torch.compile(self.compressed_model) - else: - compiled_model = torch.compile(self.compressed_model, backend="openvino") - - for i, (images, target) in enumerate(val_loader): - # W/A for memory leaks when using torch DataLoader and OpenVINO - pred = compiled_model(images) - pred = torch.argmax(pred, dim=1) - predictions[i] = pred.numpy() - references[i] = target.numpy() - return predictions, references + with disable_patching(): + with torch.no_grad(): + if self.backend in [BackendType.X86_QUANTIZER_AO, BackendType.X86_QUANTIZER_NNCF]: + compiled_model = torch.compile(self.compressed_model) + else: + compiled_model = torch.compile(self.compressed_model, backend="openvino") + + for i, (images, target) in enumerate(val_loader): + # W/A for memory leaks when using torch DataLoader and OpenVINO + pred = compiled_model(images) + pred = torch.argmax(pred, dim=1) + predictions[i] = pred.numpy() + references[i] = target.numpy() + return predictions, references def _validate(self): val_dataset = datasets.ImageFolder(root=self.data_dir / "imagenet" / "val", transform=self.transform) @@ -130,11 +133,13 @@ def _validate(self): def _compress_torch_ao(self, quantizer): - prepared_model = prepare_pt2e(self.model, quantizer) - subset_size = self.compression_params.get("subset_size", 300) - for data in islice(self.calibration_dataset.get_inference_data(), subset_size): - prepared_model(data) - self.compressed_model = convert_pt2e(prepared_model) + with disable_patching(): + with torch.no_grad(): + prepared_model = prepare_pt2e(self.model, quantizer) + subset_size = self.compression_params.get("subset_size", 300) + for data in islice(self.calibration_dataset.get_inference_data(), subset_size): + prepared_model(data) + self.compressed_model = convert_pt2e(prepared_model) def _compress_nncf_pt2e(self, quantizer): pt2e_kwargs = {} @@ -152,17 +157,25 @@ def _compress_nncf_pt2e(self, quantizer): smooth_quant = False if self.compression_params.get("model_type", False): smooth_quant = self.compression_params["model_type"] == nncf.ModelType.TRANSFORMER - self.compressed_model = quantize_pt2e( - self.model, quantizer, self.calibration_dataset, smooth_quant=smooth_quant, fold_quantize=False - ) + with disable_patching(): + with torch.no_grad(): + self.compressed_model = quantize_pt2e( + self.model, quantizer, self.calibration_dataset, smooth_quant=smooth_quant, fold_quantize=False + ) def _compress(self): """ Quantize self.model """ - if self.backend not in FX_BACKENDS or self.backend == BackendType.FX_TORCH: + if self.backend not in FX_BACKENDS: super()._compress() + return + if self.backend == BackendType.FX_TORCH: + with disable_patching(): + with torch.no_grad(): + super()._compress() + return if self.backend in [BackendType.OV_QUANTIZER_AO, BackendType.OV_QUANTIZER_NNCF]: quantizer_kwargs = {} From 44ebb6cdb78a52c7862a6a9e27daad45c343de3a Mon Sep 17 00:00:00 2001 From: dlyakhov Date: Wed, 4 Dec 2024 13:21:14 +0100 Subject: [PATCH 13/15] quantization params are being forwarded to quantize_pt2e/OpenVINOQuantizer --- .../pipelines/image_classification_base.py | 46 ++++++++++++++----- 1 file changed, 35 insertions(+), 11 deletions(-) diff --git a/tests/post_training/pipelines/image_classification_base.py b/tests/post_training/pipelines/image_classification_base.py index 6c649862b8c..0bbbd893ac1 100644 --- a/tests/post_training/pipelines/image_classification_base.py +++ b/tests/post_training/pipelines/image_classification_base.py @@ -27,6 +27,7 @@ from torchvision import datasets import nncf +from nncf import AdvancedQuantizationParameters from nncf.common.logging.track_progress import track from nncf.experimental.common.quantization.algorithms.quantizer.openvino_quantizer import OpenVINOQuantizer from nncf.experimental.torch.fx.quantization.quantize_pt2e import quantize_pt2e @@ -146,21 +147,40 @@ def _compress_nncf_pt2e(self, quantizer): for key in ( "subset_size", "fast_bias_correction", - "smooth_quant", - "bias_correction_params", - "smooth_quant_params", - "activations_range_estimator_params", - "weights_range_estimator_params", ): if key in self.compression_params: pt2e_kwargs[key] = self.compression_params[key] + + advanced_parameters: AdvancedQuantizationParameters = self.compression_params.get( + "advanced_parameters", AdvancedQuantizationParameters() + ) + + sq_params = advanced_parameters.smooth_quant_alphas + sq_alpha = advanced_parameters.smooth_quant_alpha + if sq_alpha is not None: + if sq_alpha < 0: + sq_params.convolution = -1 + sq_params.matmul = -1 + else: + sq_params.matmul = sq_alpha + pt2e_kwargs["smooth_quant_params"] = sq_params + pt2e_kwargs["bias_correction_params"] = advanced_parameters.bias_correction_params + pt2e_kwargs["activations_range_estimator_params"] = advanced_parameters.activations_range_estimator_params + pt2e_kwargs["weights_range_estimator_params"] = advanced_parameters.weights_range_estimator_params + smooth_quant = False if self.compression_params.get("model_type", False): smooth_quant = self.compression_params["model_type"] == nncf.ModelType.TRANSFORMER + with disable_patching(): with torch.no_grad(): self.compressed_model = quantize_pt2e( - self.model, quantizer, self.calibration_dataset, smooth_quant=smooth_quant, fold_quantize=False + self.model, + quantizer, + self.calibration_dataset, + smooth_quant=smooth_quant, + fold_quantize=False, + **pt2e_kwargs, ) def _compress(self): @@ -185,14 +205,18 @@ def _compress(self): "target_device", "model_type", "ignored_scope", - "overflow_fix", - "quantize_outputs", - "activations_quantization_params", - "weights_quantization_params", - "quantizer_propagation_rule", ): if key in self.compression_params: quantizer_kwargs[key] = self.compression_params[key] + advanced_parameters: AdvancedQuantizationParameters = self.compression_params.get( + "advanced_parameters", AdvancedQuantizationParameters() + ) + quantizer_kwargs["overflow_fix"] = advanced_parameters.overflow_fix + quantizer_kwargs["quantize_outputs"] = advanced_parameters.quantize_outputs + quantizer_kwargs["activations_quantization_params"] = advanced_parameters.activations_quantization_params + quantizer_kwargs["weights_quantization_params"] = advanced_parameters.weights_quantization_params + quantizer_kwargs["quantizer_propagation_rule"] = advanced_parameters.quantizer_propagation_rule + quantizer = OpenVINOQuantizer(**quantizer_kwargs) else: From 78ee5dfba8ac0371432d6bb0cd048ed79fb240d1 Mon Sep 17 00:00:00 2001 From: dlyakhov Date: Tue, 26 Nov 2024 16:39:03 +0100 Subject: [PATCH 14/15] [Conformance] Ultralytics yolov8n and yolo11n --- .../data/ptq_reference_data.yaml | 12 ++ tests/post_training/model_scope.py | 80 +++++++++++ .../pipelines/ultralytics_detection.py | 128 ++++++++++++++++++ tests/post_training/requirements.txt | 1 + 4 files changed, 221 insertions(+) create mode 100644 tests/post_training/pipelines/ultralytics_detection.py diff --git a/tests/post_training/data/ptq_reference_data.yaml b/tests/post_training/data/ptq_reference_data.yaml index 1a7b2daa2ef..bbd16c9c7da 100644 --- a/tests/post_training/data/ptq_reference_data.yaml +++ b/tests/post_training/data/ptq_reference_data.yaml @@ -90,6 +90,18 @@ torchvision/swin_v2_s_backend_X86_QUANTIZER_NNCF: metric_value: 0.8360 torchvision/swin_v2_s_backend_X86_QUANTIZER_AO: metric_value: 0.8360 +ultralytics/yolov8n_backend_FP32: + metric_value: 0.6056 +ultralytics/yolov8n_backend_FX_TORCH: + metric_value: 0.61417 +ultralytics/yolov8n_backend_OV: + metric_value: 0.6188 +ultralytics/yolo11n_backend_FP32: + metric_value: 0.6770 +ultralytics/yolo11n_backend_FX_TORCH: + metric_value: 0.6735 +ultralytics/yolo11n_backend_OV: + metric_value: 0.6752 timm/crossvit_9_240_backend_CUDA_TORCH: metric_value: 0.7275 timm/crossvit_9_240_backend_FP32: diff --git a/tests/post_training/model_scope.py b/tests/post_training/model_scope.py index 42b9b16610e..bf6218cb18a 100644 --- a/tests/post_training/model_scope.py +++ b/tests/post_training/model_scope.py @@ -33,6 +33,7 @@ from tests.post_training.pipelines.image_classification_torchvision import ImageClassificationTorchvision from tests.post_training.pipelines.lm_weight_compression import LMWeightCompression from tests.post_training.pipelines.masked_language_modeling import MaskedLanguageModelingHF +from tests.post_training.pipelines.ultralytics_detection import UltralyticsDetection QUANTIZATION_MODELS = [ # HF models @@ -124,6 +125,85 @@ "backends": FX_BACKENDS + [BackendType.OV], "batch_size": 1, }, + # Ultralytics models + { + "reported_name": "ultralytics/yolov8n", + "model_id": "yolov8n", + "pipeline_cls": UltralyticsDetection, + "compression_params": { + "preset": nncf.QuantizationPreset.MIXED, + "ignored_scope": nncf.IgnoredScope( + types=["mul", "sub", "sigmoid", "__getitem__"], + subgraphs=[ + nncf.Subgraph( + inputs=["cat_13", "cat_14", "cat_15"], + outputs=["output"], + ) + ], + ), + }, + "backends": [BackendType.FX_TORCH], + "batch_size": 1, + }, + { + "reported_name": "ultralytics/yolov8n", + "model_id": "yolov8n", + "pipeline_cls": UltralyticsDetection, + "compression_params": { + "preset": QuantizationPreset.MIXED, + "ignored_scope": nncf.IgnoredScope( + types=["Multiply", "Subtract", "Sigmoid"], + subgraphs=[ + nncf.Subgraph( + inputs=["/model.22/Concat", "/model.22/Concat_1", "/model.22/Concat_2"], + outputs=["output0/sink_port_0"], + ) + ], + ), + }, + "backends": [BackendType.OV], + "batch_size": 1, + }, + { + "reported_name": "ultralytics/yolo11n", + "model_id": "yolo11n", + "pipeline_cls": UltralyticsDetection, + "compression_params": { + "model_type": nncf.ModelType.TRANSFORMER, + "preset": QuantizationPreset.MIXED, + "ignored_scope": nncf.IgnoredScope( + types=["mul", "sub", "sigmoid", "__getitem__"], + subgraphs=[ + nncf.Subgraph( + inputs=["cat_13", "cat_14", "cat_15"], + outputs=["output"], + ) + ], + ), + }, + "backends": [BackendType.FX_TORCH], + "batch_size": 1, + }, + { + "reported_name": "ultralytics/yolo11n", + "model_id": "yolo11n", + "pipeline_cls": UltralyticsDetection, + "compression_params": { + "model_type": nncf.ModelType.TRANSFORMER, + "preset": QuantizationPreset.MIXED, + "ignored_scope": nncf.IgnoredScope( + types=["Multiply", "Subtract", "Sigmoid"], + subgraphs=[ + nncf.Subgraph( + inputs=["/model.23/Concat", "/model.23/Concat_1", "/model.23/Concat_2"], + outputs=["output0/sink_port_0"], + ) + ], + ), + }, + "backends": [BackendType.OV], + "batch_size": 1, + }, # Timm models { "reported_name": "timm/crossvit_9_240", diff --git a/tests/post_training/pipelines/ultralytics_detection.py b/tests/post_training/pipelines/ultralytics_detection.py new file mode 100644 index 00000000000..3b4766f30e6 --- /dev/null +++ b/tests/post_training/pipelines/ultralytics_detection.py @@ -0,0 +1,128 @@ +# Copyright (c) 2024 Intel Corporation +# Licensed under the Apache License, Version 2.0 (the "License"); +# you may not use this file except in compliance with the License. +# You may obtain a copy of the License at +# http://www.apache.org/licenses/LICENSE-2.0 +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +from pathlib import Path +from typing import Dict, Tuple + +import openvino as ov +import torch +from ultralytics import YOLO +from ultralytics.data.utils import check_det_dataset +from ultralytics.engine.validator import BaseValidator as Validator +from ultralytics.utils.torch_utils import de_parallel + +import nncf +from nncf.torch import disable_patching +from tests.post_training.pipelines.base import OV_BACKENDS +from tests.post_training.pipelines.base import BackendType +from tests.post_training.pipelines.base import PTQTestPipeline + + +class UltralyticsDetection(PTQTestPipeline): + """Pipeline for Yolo detection models from the Ultralytics repository""" + + def prepare_model(self) -> None: + if self.batch_size != 1: + raise RuntimeError("Batch size > 1 is not supported") + + model_path = f"{self.fp32_model_dir}/{self.model_id}" + yolo = YOLO(f"{model_path}.pt") + self.validator, self.data_loader = self._prepare_validation(yolo, "coco128.yaml") + self.dummy_tensor = torch.ones((1, 3, 640, 640)) + + if self.backend in OV_BACKENDS + [BackendType.FP32]: + onnx_model_path = Path(f"{model_path}.onnx") + ir_model_path = self.fp32_model_dir / "model_fp32.xml" + yolo.export(format="onnx", dynamic=True, half=False) + ov.save_model(ov.convert_model(onnx_model_path), ir_model_path) + self.model = ov.Core().read_model(ir_model_path) + + if self.backend == BackendType.FX_TORCH: + pt_model = yolo.model + # Run mode one time to initialize all + # internal variables + pt_model(self.dummy_tensor) + + with torch.no_grad(): + with disable_patching(): + self.model = torch.export.export(pt_model, args=(self.dummy_tensor,), strict=False).module() + + def prepare_preprocessor(self) -> None: + pass + + @staticmethod + def _validate_fx( + model: ov.Model, data_loader: torch.utils.data.DataLoader, validator: Validator, num_samples: int = None + ) -> Tuple[Dict, int, int]: + compiled_model = torch.compile(model, backend="openvino") + for batch_i, batch in enumerate(data_loader): + if num_samples is not None and batch_i == num_samples: + break + batch = validator.preprocess(batch) + preds = compiled_model(batch["img"]) + preds = validator.postprocess(preds) + validator.update_metrics(preds, batch) + stats = validator.get_stats() + return stats, validator.seen, validator.nt_per_class.sum() + + @staticmethod + def _validate_ov( + model: ov.Model, data_loader: torch.utils.data.DataLoader, validator: Validator, num_samples: int = None + ) -> Tuple[Dict, int, int]: + model.reshape({0: [1, 3, -1, -1]}) + compiled_model = ov.compile_model(model) + output_layer = compiled_model.output(0) + for batch_i, batch in enumerate(data_loader): + if num_samples is not None and batch_i == num_samples: + break + batch = validator.preprocess(batch) + preds = torch.from_numpy(compiled_model(batch["img"])[output_layer]) + preds = validator.postprocess(preds) + validator.update_metrics(preds, batch) + stats = validator.get_stats() + return stats, validator.seen, validator.nt_per_class.sum() + + def get_transform_calibration_fn(self): + def transform_func(batch): + return self.validator.preprocess(batch)["img"] + + return transform_func + + def prepare_calibration_dataset(self): + self.calibration_dataset = nncf.Dataset(self.data_loader, self.get_transform_calibration_fn()) + + @staticmethod + def _prepare_validation(model: YOLO, data: str) -> Tuple[Validator, torch.utils.data.DataLoader]: + custom = {"rect": False, "batch": 1} # method defaults + args = {**model.overrides, **custom, "mode": "val"} # highest priority args on the right + + validator = model._smart_load("validator")(args=args, _callbacks=model.callbacks) + stride = 32 # default stride + validator.stride = stride # used in get_dataloader() for padding + validator.data = check_det_dataset(data) + validator.init_metrics(de_parallel(model)) + + data_loader = validator.get_dataloader(validator.data.get(validator.args.split), validator.args.batch) + + return validator, data_loader + + def _validate(self): + if self.backend == BackendType.FP32: + stats, _, _ = self._validate_ov(self.model, self.data_loader, self.validator) + elif self.backend in OV_BACKENDS: + stats, _, _ = self._validate_ov(self.compressed_model, self.data_loader, self.validator) + elif self.backend == BackendType.FX_TORCH: + stats, _, _ = self._validate_fx(self.compressed_model, self.data_loader, self.validator) + else: + raise RuntimeError(f"Backend {self.backend} is not supported in UltralyticsDetection") + + self.run_info.metric_name = "mAP50(B)" + self.run_info.metric_value = stats["metrics/mAP50(B)"] diff --git a/tests/post_training/requirements.txt b/tests/post_training/requirements.txt index 93103066aab..e2de7cef9d0 100644 --- a/tests/post_training/requirements.txt +++ b/tests/post_training/requirements.txt @@ -19,3 +19,4 @@ timm==0.9.2 transformers==4.38.2 whowhatbench @ git+https://github.com/andreyanufr/who_what_benchmark@456d3584ce628f6c8605f37cd9a3ab2db1ebf933 datasets==2.21.0 +ultralytics==8.3.27 From 7bda046676a803a0de2e0ef6269d34ed9d55d6d9 Mon Sep 17 00:00:00 2001 From: dlyakhov Date: Fri, 6 Dec 2024 15:18:13 +0100 Subject: [PATCH 15/15] Fixes / ref commit yolo validation --- .../data/ptq_reference_data.yaml | 16 +++ tests/post_training/model_scope.py | 9 +- tests/post_training/pipelines/base.py | 112 +++++++++++++++++- .../pipelines/ultralytics_detection.py | 21 +++- .../test_quantize_conformance.py | 7 -- 5 files changed, 150 insertions(+), 15 deletions(-) diff --git a/tests/post_training/data/ptq_reference_data.yaml b/tests/post_training/data/ptq_reference_data.yaml index bbd16c9c7da..5f5421647ee 100644 --- a/tests/post_training/data/ptq_reference_data.yaml +++ b/tests/post_training/data/ptq_reference_data.yaml @@ -94,12 +94,28 @@ ultralytics/yolov8n_backend_FP32: metric_value: 0.6056 ultralytics/yolov8n_backend_FX_TORCH: metric_value: 0.61417 +ultralytics/yolov8n_backend_OV_QUANTIZER_NNCF: + metric_value: 0.61417 +ultralytics/yolov8n_backend_OV_QUANTIZER_AO: + metric_value: 0.61417 +ultralytics/yolov8n_backend_X86_QUANTIZER_NNCF: + metric_value: 0.61417 +ultralytics/yolov8n_backend_X86_QUANTIZER_AO: + metric_value: 0.61417 ultralytics/yolov8n_backend_OV: metric_value: 0.6188 ultralytics/yolo11n_backend_FP32: metric_value: 0.6770 ultralytics/yolo11n_backend_FX_TORCH: metric_value: 0.6735 +ultralytics/yolo11n_backend_OV_QUANTIZER_NNCF: + metric_value: 0.6735 +ultralytics/yolo11n_backend_OV_QUANTIZER_AO: + metric_value: 0.6735 +ultralytics/yolo11n_backend_X86_QUANTIZER_NNCF: + metric_value: 0.6735 +ultralytics/yolo11n_backend_X86_QUANTIZER_AO: + metric_value: 0.6735 ultralytics/yolo11n_backend_OV: metric_value: 0.6752 timm/crossvit_9_240_backend_CUDA_TORCH: diff --git a/tests/post_training/model_scope.py b/tests/post_training/model_scope.py index bf6218cb18a..1e98872ca5b 100644 --- a/tests/post_training/model_scope.py +++ b/tests/post_training/model_scope.py @@ -15,6 +15,7 @@ import nncf from nncf import ModelType from nncf import QuantizationPreset +from nncf.experimental.torch.fx.quantization.backend_parameters import FXBackendParameters from nncf.parameters import BackupMode from nncf.parameters import CompressWeightsMode from nncf.parameters import SensitivityMetric @@ -142,7 +143,7 @@ ], ), }, - "backends": [BackendType.FX_TORCH], + "backends": FX_BACKENDS, "batch_size": 1, }, { @@ -180,8 +181,12 @@ ) ], ), + "advanced_parameters": AdvancedQuantizationParameters( + # Weight compression decreases accuracu on 0.2 + backend_params={FXBackendParameters.COMPRESS_WEIGHTS: False} + ), }, - "backends": [BackendType.FX_TORCH], + "backends": FX_BACKENDS, "batch_size": 1, }, { diff --git a/tests/post_training/pipelines/base.py b/tests/post_training/pipelines/base.py index a949387a497..5687b81795d 100644 --- a/tests/post_training/pipelines/base.py +++ b/tests/post_training/pipelines/base.py @@ -8,6 +8,11 @@ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. + +import os + +os.environ["TORCHINDUCTOR_FREEZING"] = "1" + import datetime as dt import gc import os @@ -15,9 +20,11 @@ import time from abc import ABC from abc import abstractmethod +from copy import deepcopy from dataclasses import dataclass from datetime import timedelta from enum import Enum +from itertools import islice from pathlib import Path from typing import Dict, Optional @@ -27,9 +34,17 @@ import torch from memory_profiler import memory_usage from optimum.intel import OVQuantizer +from torch.ao.quantization.quantize_pt2e import convert_pt2e +from torch.ao.quantization.quantize_pt2e import prepare_pt2e +from torch.ao.quantization.quantizer.x86_inductor_quantizer import X86InductorQuantizer +from torch.ao.quantization.quantizer.x86_inductor_quantizer import get_default_x86_inductor_quantization_config import nncf +from nncf import AdvancedQuantizationParameters from nncf import TargetDevice +from nncf.experimental.common.quantization.algorithms.quantizer.openvino_quantizer import OpenVINOQuantizer +from nncf.experimental.torch.fx.quantization.quantize_pt2e import quantize_pt2e +from nncf.torch import disable_patching from tests.cross_fw.shared.command import Command from tools.memory_monitor import MemoryType from tools.memory_monitor import MemoryUnit @@ -358,7 +373,7 @@ class PTQTestPipeline(BaseTestPipeline): Base class to test post training quantization. """ - def _compress(self): + def _compress_nncf_quantize(self): """ Quantize self.model """ @@ -495,3 +510,98 @@ def collect_data_from_stdout(self, stdout: str): stats = PTQTimeStats() stats.fill(stdout) self.run_info.stats_from_output = stats + + def _compress_torch_ao(self, quantizer): + with disable_patching(): + with torch.no_grad(): + prepared_model = prepare_pt2e(deepcopy(self.model), quantizer) + subset_size = self.compression_params.get("subset_size", 300) + for data in islice(self.calibration_dataset.get_inference_data(), subset_size): + prepared_model(data) + self.compressed_model = convert_pt2e(prepared_model) + + def _compress_nncf_pt2e(self, quantizer): + pt2e_kwargs = {} + for key in ( + "subset_size", + "fast_bias_correction", + ): + if key in self.compression_params: + pt2e_kwargs[key] = self.compression_params[key] + + advanced_parameters: AdvancedQuantizationParameters = self.compression_params.get( + "advanced_parameters", AdvancedQuantizationParameters() + ) + + sq_params = advanced_parameters.smooth_quant_alphas + sq_alpha = advanced_parameters.smooth_quant_alpha + if sq_alpha is not None: + if sq_alpha < 0: + sq_params.convolution = -1 + sq_params.matmul = -1 + else: + sq_params.matmul = sq_alpha + pt2e_kwargs["smooth_quant_params"] = sq_params + pt2e_kwargs["bias_correction_params"] = advanced_parameters.bias_correction_params + pt2e_kwargs["activations_range_estimator_params"] = advanced_parameters.activations_range_estimator_params + pt2e_kwargs["weights_range_estimator_params"] = advanced_parameters.weights_range_estimator_params + + smooth_quant = False + if self.compression_params.get("model_type", False): + smooth_quant = self.compression_params["model_type"] == nncf.ModelType.TRANSFORMER + + with disable_patching(): + with torch.no_grad(): + self.compressed_model = quantize_pt2e( + self.model, + quantizer, + self.calibration_dataset, + smooth_quant=smooth_quant, + fold_quantize=False, + **pt2e_kwargs, + ) + + def _compress(self): + """ + Quantize self.model + """ + if self.backend not in FX_BACKENDS: + self._compress_nncf_quantize() + + return + if self.backend == BackendType.FX_TORCH: + with disable_patching(): + with torch.no_grad(): + self._compress_nncf_quantize() + return + + if self.backend in [BackendType.OV_QUANTIZER_AO, BackendType.OV_QUANTIZER_NNCF]: + quantizer_kwargs = {} + for key in ( + "mode", + "preset", + "target_device", + "model_type", + "ignored_scope", + ): + if key in self.compression_params: + quantizer_kwargs[key] = self.compression_params[key] + advanced_parameters: AdvancedQuantizationParameters = self.compression_params.get( + "advanced_parameters", AdvancedQuantizationParameters() + ) + quantizer_kwargs["overflow_fix"] = advanced_parameters.overflow_fix + quantizer_kwargs["quantize_outputs"] = advanced_parameters.quantize_outputs + quantizer_kwargs["activations_quantization_params"] = advanced_parameters.activations_quantization_params + quantizer_kwargs["weights_quantization_params"] = advanced_parameters.weights_quantization_params + quantizer_kwargs["quantizer_propagation_rule"] = advanced_parameters.quantizer_propagation_rule + + quantizer = OpenVINOQuantizer(**quantizer_kwargs) + else: + + quantizer = X86InductorQuantizer() + quantizer.set_global(get_default_x86_inductor_quantization_config()) + + if self.backend in [BackendType.OV_QUANTIZER_NNCF, BackendType.X86_QUANTIZER_NNCF]: + self._compress_nncf_pt2e(quantizer) + else: + self._compress_torch_ao(quantizer) diff --git a/tests/post_training/pipelines/ultralytics_detection.py b/tests/post_training/pipelines/ultralytics_detection.py index 3b4766f30e6..41c6ba1cbd4 100644 --- a/tests/post_training/pipelines/ultralytics_detection.py +++ b/tests/post_training/pipelines/ultralytics_detection.py @@ -21,6 +21,7 @@ import nncf from nncf.torch import disable_patching +from tests.post_training.pipelines.base import FX_BACKENDS from tests.post_training.pipelines.base import OV_BACKENDS from tests.post_training.pipelines.base import BackendType from tests.post_training.pipelines.base import PTQTestPipeline @@ -45,7 +46,7 @@ def prepare_model(self) -> None: ov.save_model(ov.convert_model(onnx_model_path), ir_model_path) self.model = ov.Core().read_model(ir_model_path) - if self.backend == BackendType.FX_TORCH: + if self.backend in FX_BACKENDS: pt_model = yolo.model # Run mode one time to initialize all # internal variables @@ -60,9 +61,17 @@ def prepare_preprocessor(self) -> None: @staticmethod def _validate_fx( - model: ov.Model, data_loader: torch.utils.data.DataLoader, validator: Validator, num_samples: int = None + model: ov.Model, + data_loader: torch.utils.data.DataLoader, + validator: Validator, + backend: BackendType, + num_samples: int = None, ) -> Tuple[Dict, int, int]: - compiled_model = torch.compile(model, backend="openvino") + if backend in [BackendType.FX_TORCH, BackendType.OV_QUANTIZER_AO, BackendType.OV_QUANTIZER_NNCF]: + compiled_model = torch.compile(model, backend="openvino") + else: + compiled_model = torch.compile(model) + for batch_i, batch in enumerate(data_loader): if num_samples is not None and batch_i == num_samples: break @@ -119,8 +128,10 @@ def _validate(self): stats, _, _ = self._validate_ov(self.model, self.data_loader, self.validator) elif self.backend in OV_BACKENDS: stats, _, _ = self._validate_ov(self.compressed_model, self.data_loader, self.validator) - elif self.backend == BackendType.FX_TORCH: - stats, _, _ = self._validate_fx(self.compressed_model, self.data_loader, self.validator) + elif self.backend in FX_BACKENDS: + stats, _, _ = self._validate_fx( + self.compressed_model, self.data_loader, self.validator, backend=self.backend + ) else: raise RuntimeError(f"Backend {self.backend} is not supported in UltralyticsDetection") diff --git a/tests/post_training/test_quantize_conformance.py b/tests/post_training/test_quantize_conformance.py index 075b838e6d7..20504a8b086 100644 --- a/tests/post_training/test_quantize_conformance.py +++ b/tests/post_training/test_quantize_conformance.py @@ -75,11 +75,6 @@ def fixture_run_benchmark_app(pytestconfig): return pytestconfig.getoption("benchmark") -@pytest.fixture(scope="session", name="validate_in_backend") -def fixture_validate_in_backend(pytestconfig): - return pytestconfig.getoption("validate_in_backend") - - @pytest.fixture(scope="session", name="extra_columns") def fixture_extra_columns(pytestconfig): return pytestconfig.getoption("extra_columns") @@ -271,7 +266,6 @@ def test_ptq_quantization( run_torch_cuda_backend: bool, subset_size: Optional[int], run_benchmark_app: bool, - validate_in_backend: bool, capsys: pytest.CaptureFixture, extra_columns: bool, memory_monitor: bool, @@ -299,7 +293,6 @@ def test_ptq_quantization( "data_dir": data_dir, "no_eval": no_eval, "run_benchmark_app": run_benchmark_app, - "validate_in_backend": validate_in_backend, "batch_size": batch_size, "memory_monitor": memory_monitor, }