Plot NAS search progression non-interactively (#2025)

### Changes
As stated in the title

### Reason for changes
Windows pre-commits fail because they cannot find tk during the sanity
tests. The plotting in sanity tests doesn't actually need interactivity,
so agg can be used instead of tk as a backend which does not pose
cross-platform problems.

### Related tickets
N/A

### Tests
windows/precommit_torch_cpu

nncf/common/accuracy_aware_training/runner.py

@@ -20,6 +20,7 @@
 from nncf.api.compression import CompressionAlgorithmController
 from nncf.api.compression import CompressionStage
 from nncf.common.logging import nncf_logger
+from nncf.common.plotting import noninteractive_plotting
 from nncf.common.utils.helpers import configure_accuracy_aware_paths
 from nncf.common.utils.tensorboard import prepare_for_tensorboard
 from nncf.config.schemata.defaults import AA_COMPRESSION_RATE_STEP_REDUCTION_FACTOR
@@ -486,20 +487,17 @@ def update_training_history(self, compression_rate, metric_value):
         self._compressed_training_history.append((compression_rate, accuracy_budget))
 
         if IMG_PACKAGES_AVAILABLE:
-            backend = plt.get_backend()
-            plt.switch_backend("agg")
-            plt.ioff()
-            fig = plt.figure()
-            plt.plot(self.compressed_training_history.keys(), self.compressed_training_history.values())
-            buf = io.BytesIO()
-            plt.savefig(buf, format="jpeg")
-            buf.seek(0)
-            image = PIL.Image.open(buf)
-            self.add_tensorboard_image(
-                "compression/accuracy_aware/acc_budget_vs_comp_rate", image, len(self.compressed_training_history)
-            )
-            plt.close(fig)
-            plt.switch_backend(backend)
+            with noninteractive_plotting():
+                fig = plt.figure()
+                plt.plot(self.compressed_training_history.keys(), self.compressed_training_history.values())
+                buf = io.BytesIO()
+                plt.savefig(buf, format="jpeg")
+                buf.seek(0)
+                image = PIL.Image.open(buf)
+                self.add_tensorboard_image(
+                    "compression/accuracy_aware/acc_budget_vs_comp_rate", image, len(self.compressed_training_history)
+                )
+                plt.close(fig)
 
     @property
     def compressed_training_history(self):

nncf/common/plotting.py

@@ -0,0 +1,22 @@
+# Copyright (c) 2023 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 contextlib import contextmanager
+
+from matplotlib import pyplot as plt
+
+
+@contextmanager
+def noninteractive_plotting():
+    backend = plt.get_backend()
+    plt.switch_backend("agg")
+    plt.ioff()
+    yield
+    plt.switch_backend(backend)

nncf/experimental/torch/nas/bootstrapNAS/search/search.py

@@ -29,6 +29,7 @@
 from nncf import NNCFConfig
 from nncf.common.initialization.batchnorm_adaptation import BatchnormAdaptationAlgorithm
 from nncf.common.logging import nncf_logger
+from nncf.common.plotting import noninteractive_plotting
 from nncf.common.utils.decorators import skip_if_dependency_unavailable
 from nncf.common.utils.os import safe_open
 from nncf.config.extractors import get_bn_adapt_algo_kwargs
@@ -399,45 +400,46 @@ def visualize_search_progression(self, filename="search_progression") -> NoRetur
         """
         import matplotlib.pyplot as plt
 
-        plt.figure()
-        colormap = plt.cm.get_cmap("viridis")
-        col = range(int(self.search_params.num_evals / self.search_params.population))
-        for i in range(0, len(self.search_records), self.search_params.population):
-            c = [col[int(i / self.search_params.population)]] * len(
-                self.search_records[i : i + self.search_params.population]
-            )
-            plt.scatter(
-                [abs(row[2]) for row in self.search_records][i : i + self.search_params.population],
-                [abs(row[4]) for row in self.search_records][i : i + self.search_params.population],
-                s=9,
-                c=c,
-                alpha=0.5,
-                marker="D",
-                cmap=colormap,
-            )
-        plt.scatter(
-            *tuple(abs(ev.input_model_value) for ev in self.evaluator_handlers),
-            marker="s",
-            s=120,
-            color="blue",
-            label="Input Model",
-            edgecolors="black",
-        )
-        if None not in self.best_vals:
+        with noninteractive_plotting():
+            plt.figure()
+            colormap = plt.cm.get_cmap("viridis")
+            col = range(int(self.search_params.num_evals / self.search_params.population))
+            for i in range(0, len(self.search_records), self.search_params.population):
+                c = [col[int(i / self.search_params.population)]] * len(
+                    self.search_records[i : i + self.search_params.population]
+                )
+                plt.scatter(
+                    [abs(row[2]) for row in self.search_records][i : i + self.search_params.population],
+                    [abs(row[4]) for row in self.search_records][i : i + self.search_params.population],
+                    s=9,
+                    c=c,
+                    alpha=0.5,
+                    marker="D",
+                    cmap=colormap,
+                )
             plt.scatter(
-                *tuple(abs(val) for val in self.best_vals),
-                marker="o",
+                *tuple(abs(ev.input_model_value) for ev in self.evaluator_handlers),
+                marker="s",
                 s=120,
-                color="yellow",
-                label="BootstrapNAS A",
+                color="blue",
+                label="Input Model",
                 edgecolors="black",
-                linewidth=2.5,
             )
-        plt.legend()
-        plt.title("Search Progression")
-        plt.xlabel(self.efficiency_evaluator_handler.name)
-        plt.ylabel(self.accuracy_evaluator_handler.name)
-        plt.savefig(f"{self._log_dir}/{filename}.png")
+            if None not in self.best_vals:
+                plt.scatter(
+                    *tuple(abs(val) for val in self.best_vals),
+                    marker="o",
+                    s=120,
+                    color="yellow",
+                    label="BootstrapNAS A",
+                    edgecolors="black",
+                    linewidth=2.5,
+                )
+            plt.legend()
+            plt.title("Search Progression")
+            plt.xlabel(self.efficiency_evaluator_handler.name)
+            plt.ylabel(self.accuracy_evaluator_handler.name)
+            plt.savefig(f"{self._log_dir}/{filename}.png")
 
     def save_evaluators_state(self) -> NoReturn:
         """