Fix quantile parameter not being used for MEAN_NO_OUTLIERS aggragator

openvinotoolkit · Apr 18, 2024 · 92e491c · 92e491c
1 parent 573b0c3
commit 92e491c
Show file tree

Hide file tree

Showing 3 changed files with 27 additions and 17 deletions.
diff --git a/nncf/quantization/algorithms/min_max/onnx_backend.py b/nncf/quantization/algorithms/min_max/onnx_backend.py
@@ -42,7 +42,7 @@
 from nncf.quantization.algorithms.min_max.backend import MinMaxAlgoBackend
 from nncf.quantization.fake_quantize import FakeConvertParameters
 from nncf.quantization.fake_quantize import FakeQuantizeParameters
-from nncf.quantization.range_estimator import RangeEstimatorParameters
+from nncf.quantization.range_estimator import RangeEstimatorParameters, AggregatorType
 
 
 class ONNXMinMaxAlgoBackend(MinMaxAlgoBackend):
@@ -211,11 +211,14 @@ def get_statistic_collector(
                 statistic_type = StatisticsType.ABS_MAX
             reducer = ONNX_REDUCERS_MAP[statistic_type](**kwargs)
 
-            aggregator = AGGREGATORS_MAP[params.aggregator_type](
-                num_samples=num_samples,
-                aggregation_axes=aggregation_axes,
-                tensor_processor=ONNXNNCFCollectorTensorProcessor,
-            )
+            kwargs = {
+                "num_samples": num_samples,
+                "aggregation_axes": aggregation_axes,
+                "tensor_processor": ONNXNNCFCollectorTensorProcessor
+            }
+            if params.aggregator_type == AggregatorType.MEAN_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

diff --git a/nncf/quantization/algorithms/min_max/openvino_backend.py b/nncf/quantization/algorithms/min_max/openvino_backend.py
@@ -42,6 +42,7 @@
 from nncf.quantization.algorithms.min_max.backend import MinMaxAlgoBackend
 from nncf.quantization.fake_quantize import FakeConvertParameters
 from nncf.quantization.fake_quantize import FakeQuantizeParameters
+from nncf.quantization.range_estimator import AggregatorType
 
 
 class OVMinMaxAlgoBackend(MinMaxAlgoBackend):
@@ -195,11 +196,14 @@ def get_statistic_collector(
                 statistic_type = StatisticsType.ABS_MAX
             reducer = OV_REDUCERS_MAP[statistic_type](**kwargs)
 
-            aggregator = AGGREGATORS_MAP[params.aggregator_type](
-                num_samples=num_samples,
-                aggregation_axes=aggregation_axes,
-                tensor_processor=OVNNCFCollectorTensorProcessor,
-            )
+            kwargs = {
+                "num_samples": num_samples,
+                "aggregation_axes": aggregation_axes,
+                "tensor_processor": OVNNCFCollectorTensorProcessor
+            }
+            if params.aggregator_type == AggregatorType.MEAN_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

diff --git a/nncf/quantization/algorithms/min_max/torch_backend.py b/nncf/quantization/algorithms/min_max/torch_backend.py
@@ -33,7 +33,7 @@
 from nncf.quantization.algorithms.min_max.backend import MinMaxAlgoBackend
 from nncf.quantization.fake_quantize import FakeConvertParameters
 from nncf.quantization.fake_quantize import FakeQuantizeParameters
-from nncf.quantization.range_estimator import RangeEstimatorParameters
+from nncf.quantization.range_estimator import RangeEstimatorParameters, AggregatorType
 from nncf.torch.graph.graph import PTNNCFGraph
 from nncf.torch.graph.graph import PTTargetPoint
 from nncf.torch.graph.transformations.command_creation import create_quantizer_insertion_command
@@ -195,11 +195,14 @@ def get_statistic_collector(
                     statistic_type = StatisticsType.ABS_MAX
                 reducer = PT_REDUCERS_MAP[statistic_type](reduction_axes=reduction_axes)
 
-            aggregator = AGGREGATORS_MAP[params.aggregator_type](
-                aggregation_axes=aggregation_axes,
-                num_samples=num_samples,
-                tensor_processor=PTNNCFCollectorTensorProcessor,
-            )
+            kwargs = {
+                "num_samples": num_samples,
+                "aggregation_axes": aggregation_axes,
+                "tensor_processor": PTNNCFCollectorTensorProcessor
+            }
+            if params.aggregator_type == AggregatorType.MEAN_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