Merge branch 'openvinotoolkit:develop' into develop

.github/workflows/api_changes_check.yml

@@ -61,3 +61,15 @@ jobs:
               repo: context.repo.repo,
               labels: ["API"]
             })
+      - name: Add release label
+        uses: actions/github-script@v6
+        if: ${{ contains(github.event.pull_request.base.ref, 'release_v') }}
+        with:
+          github-token: ${{ secrets.ADD_LABELS_WITH_REST_API }}
+          script: |
+            github.rest.issues.addLabels({
+              issue_number: context.issue.number,
+              owner: context.repo.owner,
+              repo: context.repo.repo,
+              labels: ["release_target"]
+            })

.github/workflows/post_pr_merge.yml

@@ -23,11 +23,15 @@ jobs:
     with:
       merge_commit_sha: ${{ github.event.pull_request.merge_commit_sha }}
       last_sha_in_pr: ${{ github.event.pull_request.head.sha }}
-      coverage_artifact_name_in_pr: coverage_xml_common
+      coverage_artifact_name_in_pr: coverage_common
+    secrets:
+      CODECOV_TOKEN: ${{ secrets.CODECOV_TOKEN }}
   upload-coverage-onnx:
     if: github.event.pull_request.merged == true
     uses: ./.github/workflows/upload_coverage_for_develop.yml
     with:
       merge_commit_sha: ${{ github.event.pull_request.merge_commit_sha }}
       last_sha_in_pr: ${{ github.event.pull_request.head.sha }}
-      coverage_artifact_name_in_pr: coverage_xml_onnx
+      coverage_artifact_name_in_pr: coverage_onnx
+    secrets:
+      CODECOV_TOKEN: ${{ secrets.CODECOV_TOKEN }}

.github/workflows/precommit.yml

@@ -21,16 +21,18 @@ jobs:
         run: make install-common-test
       - name: Run common precommit test scope
         run: make test-common
+        env:
+          NNCF_COVERAGE: 1
       - name: Upload coverage report as artifact
         uses: actions/upload-artifact@v3
         with:
-          name: coverage_xml_common
+          name: coverage_common
           path: ./coverage.xml
       - name: Upload coverage report to codecov
         uses: codecov/codecov-action@v3
         with:
           token: ${{ secrets.CODECOV_TOKEN }}
-          name: codecov-common
+          name: coverage_common
   onnx:
     runs-on: ubuntu-20.04
     steps:
@@ -44,14 +46,16 @@ jobs:
         run: make install-onnx-test
       - name: Run ONNX precommit test scope
         run: make test-onnx
+        env:
+          NNCF_COVERAGE: 1
       - name: Upload coverage report as artifact
         uses: actions/upload-artifact@v3
         with:
-          name: coverage_xml_onnx
+          name: coverage_onnx
           path: ./coverage.xml
       - name: Upload coverage report to codecov
         uses: codecov/codecov-action@v3
         with:
           token: ${{ secrets.CODECOV_TOKEN }}
-          name: codecov-onnx
+          name: coverage_onnx
 

.github/workflows/upload_coverage_for_develop.yml

@@ -12,6 +12,9 @@ on:
       coverage_artifact_name_in_pr:
         required: true
         type: string
+    secrets:
+      CODECOV_TOKEN:
+        required: true
 
 jobs:
   upload-coverage:
@@ -20,18 +23,18 @@ jobs:
     steps:
       - uses: actions/checkout@v3  # codecov uploader demands that the scanned files be present when uploading
         with:
-          ref: ${{ github.event.pull_request.merge_commit_sha }}
+          ref: ${{ inputs.merge_commit_sha }}
       - uses: dawidd6/action-download-artifact@v2
         with:
           workflow: precommit.yml
           check_artifacts: true
-          commit: ${{ github.event.pull_request.head.sha }}  # this is the latest commit in the PR
-          name: coverage_xml
+          commit: ${{ inputs.last_sha_in_pr }}  # this is the latest commit in the PR
+          name: ${{ inputs.coverage_artifact_name_in_pr }}
       - name: Upload coverage report to Codecov
         run: |
           curl -Os https://uploader.codecov.io/latest/linux/codecov
           chmod +x codecov
 
           # github.event.pull_request.merge_commit_sha is the fresh commit in the develop,
           # provided that github.event.pull_request.merged == true
-          ./codecov -f ./coverage.xml -t ${{ secrets.CODECOV_TOKEN }}  -C ${{ github.event.pull_request.merge_commit_sha }} -B develop -n "codecov-onnx"
+          ./codecov -f ./coverage.xml -t ${{ secrets.CODECOV_TOKEN }}  -C ${{ inputs.merge_commit_sha }} -B develop -n "${{ inputs.coverage_artifact_name_in_pr }}"

Makefile

@@ -1,10 +1,19 @@
 JUNITXML_PATH ?= nncf-tests.xml
-COVERAGE ?= --cov=./ --cov-report=xml
+
+ifdef NNCF_COVERAGE
+	COVERAGE_ARGS ?= --cov=./ --cov-report=xml
+else
+	COVERAGE_ARGS :=
+endif
 
 ifdef DATA
 	DATA_ARG := --data $(DATA)
 endif
 
+ifdef WEEKLY_MODELS
+	WEEKLY_MODELS_ARG := --weekly-models $(WEEKLY_MODELS)
+endif
+
 install-pre-commit:
 	pip install pre-commit==3.2.2
 
@@ -26,7 +35,7 @@ install-onnx-dev: install-onnx-test install-pre-commit install-pylint
 	pip install -r examples/post_training_quantization/onnx/mobilenet_v2/requirements.txt
 
 test-onnx:
-	pytest ${COVERAGE} tests/onnx $(DATA_ARG) --junitxml ${JUNITXML_PATH}
+	pytest ${COVERAGE_ARGS} tests/onnx $(DATA_ARG) --junitxml ${JUNITXML_PATH}
 
 pylint-onnx:
 	pylint --rcfile .pylintrc               \
@@ -60,8 +69,7 @@ install-openvino-dev: install-openvino-test install-pre-commit install-pylint
 	pip install -r examples/post_training_quantization/openvino/yolov8_quantize_with_accuracy_control/requirements.txt
 
 test-openvino:
-    # omitting ${COVERAGE} for internal runs since they seem to introduce a major slowdown
-	pytest tests/openvino $(DATA_ARG) --junitxml ${JUNITXML_PATH}
+	pytest ${COVERAGE_ARGS} tests/openvino $(DATA_ARG) --junitxml ${JUNITXML_PATH}
 
 pylint-openvino:
 	pylint --rcfile .pylintrc               \
@@ -91,8 +99,7 @@ install-tensorflow-dev: install-tensorflow-test install-pre-commit install-pylin
 	pip install -r examples/post_training_quantization/tensorflow/mobilenet_v2/requirements.txt
 
 test-tensorflow:
-    # omitting ${COVERAGE} for internal runs since they seem to introduce a major slowdown
-	pytest tests/common tests/tensorflow    \
+	pytest ${COVERAGE_ARGS} tests/common tests/tensorflow    \
 		--junitxml ${JUNITXML_PATH}         \
 		$(DATA_ARG)
 
@@ -121,8 +128,13 @@ install-torch-dev: install-torch-test install-pre-commit install-pylint
 	pip install -r examples/post_training_quantization/torch/ssd300_vgg16/requirements.txt
 
 test-torch:
-    # omitting ${COVERAGE} for internal runs since they seem to introduce a major slowdown
-	pytest tests/common tests/torch --junitxml ${JUNITXML_PATH} $(DATA_ARG)
+	pytest ${COVERAGE_ARGS} tests/common tests/torch -m "not weekly and not nightly" --junitxml ${JUNITXML_PATH} $(DATA_ARG)
+
+test-torch-nightly:
+	pytest ${COVERAGE_ARGS} tests/torch -m nightly --junitxml ${JUNITXML_PATH} $(DATA_ARG)
+
+test-torch-weekly:
+	pytest ${COVERAGE_ARGS} tests/torch -m weekly --junitxml ${JUNITXML_PATH} $(DATA_ARG) ${WEEKLY_MODELS_ARG}
 
 COMMON_PYFILES := $(shell python3 tools/collect_pylint_input_files_for_backend.py common)
 pylint-torch:
@@ -160,7 +172,7 @@ pylint-common:
 		$(COMMON_PYFILES)
 
 test-common:
-	pytest ${COVERAGE} tests/common $(DATA_ARG) --junitxml ${JUNITXML_PATH}
+	pytest ${COVERAGE_ARGS} tests/common $(DATA_ARG) --junitxml ${JUNITXML_PATH}
 
 test-examples:
 	pytest tests/cross_fw/examples -s --junitxml ${JUNITXML_PATH}

README.md

@@ -34,7 +34,7 @@ learning frameworks.
 | Compression algorithm                                                       |OpenVINO|PyTorch|   TensorFlow   |     ONNX       |
 |:----------------------------------------------------------------------------| :---: | :---: |:--------:|:------------------:|
 | [Post-Training Quantization](./docs/compression_algorithms/post_training/Quantization.md) | Supported | Supported |Supported| Supported |
-| [Weights Compression](./docs/compression_algorithms/CompressWeights.md) | Not supported | Supported |Not supported| Not supported |
+| [Weights Compression](./docs/compression_algorithms/CompressWeights.md) | Supported | Supported |Not supported| Not supported |
 
 ### Training-Time Compression Algorithms
 
@@ -280,13 +280,15 @@ A collection of ready-to-run Jupyter* notebooks are available to demonstrate how
 - [NNCF Post-Training Optimization of Segment Anything Model](https://github.com/openvinotoolkit/openvino_notebooks/tree/main/notebooks/237-segment-anything)
 - [NNCF Post-Training Optimization of CLIP Model](https://github.com/openvinotoolkit/openvino_notebooks/tree/main/notebooks/228-clip-zero-shot-image-classification)
 - [NNCF Post-Training Optimization of ImageBind Model](https://github.com/openvinotoolkit/openvino_notebooks/tree/main/notebooks/239-image-bind)
+- [NNCF Post-Training Optimization of Whisper Model](https://github.com/openvinotoolkit/openvino_notebooks/tree/main/notebooks/227-whisper-subtitles-generation)
 - [Quantize a Segmentation Model and Show Live Inference](https://github.com/openvinotoolkit/openvino_notebooks/blob/main/notebooks/110-ct-segmentation-quantize)
 - [Training to Deployment with TensorFlow and OpenVINO](https://github.com/openvinotoolkit/openvino_notebooks/blob/main/notebooks/301-tensorflow-training-openvino)
 - [Migrate quantization from POT API to NNCF API](https://github.com/openvinotoolkit/openvino_notebooks/blob/main/notebooks/111-yolov5-quantization-migration)
 - [Post-Training Quantization of Pytorch model with NNCF](https://github.com/openvinotoolkit/openvino_notebooks/tree/main/notebooks/112-pytorch-post-training-quantization-nncf)
 - [Optimizing PyTorch models with NNCF of OpenVINO by 8-bit quantization](https://github.com/openvinotoolkit/openvino_notebooks/tree/main/notebooks/302-pytorch-quantization-aware-training)
 - [Optimizing TensorFlow models with NNCF of OpenVINO by 8-bit quantization](https://github.com/openvinotoolkit/openvino_notebooks/tree/main/notebooks/305-tensorflow-quantization-aware-training)
 - [Accelerate Inference of Sparse Transformer Models with OpenVINO and 4th Gen Intel Xeon Scalable Processors](https://github.com/openvinotoolkit/openvino_notebooks/blob/main/notebooks/116-sparsity-optimization)
+- [Quantization with accuracy control using NNCF](https://github.com/openvinotoolkit/openvino_notebooks/tree/main/notebooks/122-quantizing-model-with-accuracy-control)
 
 ### Post-Training Quantization Samples
 
@@ -366,11 +368,11 @@ You may also use one of the Dockerfiles in the [docker](./docker) directory to b
 - Python\* 3.7 or later
 - Supported frameworks:
   - PyTorch\* >=1.13.0, <2.1
-  - TensorFlow\* >=2.4.0, <=2.12.0
+  - TensorFlow\* >=2.8.4, <=2.12.1
   - ONNX\* ~=1.13.1
   - OpenVINO\* >=2022.3.0
 
-This repository is tested on Python* 3.8.10, PyTorch* 2.0.1 (NVidia CUDA\* Toolkit 11.7) and TensorFlow* 2.12.0 (NVidia CUDA\* Toolkit 11.8).
+This repository is tested on Python* 3.8.10, PyTorch* 2.0.1 (NVidia CUDA\* Toolkit 11.7) and TensorFlow* 2.12.1 (NVidia CUDA\* Toolkit 11.8).
 
 ## NNCF Compressed Model Zoo
 

ReleaseNotes.md

@@ -1,5 +1,65 @@
 # Release Notes
 
+## New in Release 2.6.0
+
+Post-training Quantization:
+
+- Features:
+  - Added `CPU_SPR` device type support.
+  - Added quantizers scales unification.
+  - Added quantization scheme for ReduceSum operation.
+  - Added new types (ReduceL2, ReduceSum, Maximum) to the ignored scope for `ModelType.Transformer`.
+  - (OpenVINO) Added SmoothQuant algorithm.
+  - (OpenVINO) Added ChannelAlignment algorithm.
+  - (OpenVINO) Added HyperparameterTuner algorithm.
+  - (PyTorch) Added FastBiasCorrection algorithm support.
+  - (OpenVINO, ONNX) Added embedding weights quantization.
+  - (OpenVINO, PyTorch) Added new `compress_weights` method that provides data-free [INT8 weights compression](docs/compression_algorithms/CompressWeights.md).
+- Fixes:
+  - Fixed detection of decomposed post-processing in models.
+  - Multiple fixes (new patterns, bugfixes, etc.) to solve [#1936](https://github.com/openvinotoolkit/nncf/issues/1936) issue.
+  - Fixed model reshaping while quantization to keep original model shape.
+  - (OpenVINO) Added support for sequential models quanitzation.
+  - (OpenVINO) Fixed in-place statistics cast to support empty dimensions.
+  - (OpenVINO, ONNX) Fixed quantization of the MatMul operation with weights rank > 2.
+  - (OpenVINO, ONNX) Fixed BiasCorrection algorithm to enable [CLIP model quantization](https://github.com/openvinotoolkit/openvino_notebooks/tree/main/notebooks/228-clip-zero-shot-image-classification).
+- Improvements:
+  - Optimized `quantize(…)` pipeline (up to 4.3x speed up in total).
+  - Optimized `quantize_with_accuracy_control(…)` pipelilne (up to 8x speed up for [122-quantizing-model-with-accuracy-control](https://github.com/openvinotoolkit/openvino_notebooks/tree/main/notebooks/122-quantizing-model-with-accuracy-control) notebook).
+  - Optimized general statistics collection (up to 1.2x speed up for ONNX backend).
+  - Ignored patterns separated from Fused patterns scheme (with multiple patterns addition).
+- Tutorials:
+  - [Post-Training Optimization of Segment Anything Model](https://github.com/openvinotoolkit/openvino_notebooks/tree/main/notebooks/237-segment-anything).
+  - [Post-Training Optimization of CLIP Model](https://github.com/openvinotoolkit/openvino_notebooks/tree/main/notebooks/228-clip-zero-shot-image-classification).
+  - [Post-Training Optimization of ImageBind Model](https://github.com/openvinotoolkit/openvino_notebooks/tree/main/notebooks/239-image-bind).
+  - [Post-Training Optimization of Whisper Model](https://github.com/openvinotoolkit/openvino_notebooks/tree/main/notebooks/227-whisper-subtitles-generation).
+  - [Post-Training Optimization with accuracy control](https://github.com/openvinotoolkit/openvino_notebooks/tree/main/notebooks/122-quantizing-model-with-accuracy-control).
+
+Compression-aware training:
+
+- Features:
+  - Added shape pruning processor for BootstrapNAS algorithm.
+  - Added KD loss for BootstrapNAS algorithm.
+  - Added `validate_scopes` parameter for NNCF configuration.
+  - (PyTorch) Added PyTorch 2.0 support.
+  - (PyTorch) Added `.strip()` option to API.
+  - (PyTorch) Enabled bfloat data type for quantization kernels.
+  - (PyTorch) Quantized models can now be `torch.jit.trace`d without calling `.strip()`.
+  - (PyTorch) Added support for overridden `forward` instance attribute on model objects passed into `create_compressed_model`.
+  - (Tensorflow) Added Tensorflow 2.12 support.
+- Fixes:
+  - (PyTorch) Fixed padding adjustment issue in the elastic kernel to work with the different active kernel sizes.
+  - (PyTorch) Fixed the torch graph tracing in the case the tensors belonging to parallel edges are interleaved in the order of the tensor argument.
+  - (PyTorch) Fixed recurrent nodes matching (LSTM, GRU cells) condition with the strict rule to avoid adding not necessary nodes to the ignored scope.
+  - (PyTorch) Fixed `torch.jit.script` wrapper so that user-side handling exceptions during `torch.jit.script` invocation do not cause NNCF to be permanently disabled.
+  - (PyTorch, Tensorflow) Adjusted quantizer propagation algorithm to check if quantizer propagation will result in output quantization.
+  - (PyTorch) Added redefined `__class__` method for ProxyModule that avoids causing error while calling `.super()` in forward method.
+- Deprecations/Removals:
+  - (PyTorch) Removed deprecated `NNCFNetwork.__getattr__`, `NNCFNetwork.get_nncf_wrapped_model` methods.
+- Requirements:
+  - Updated PyTorch version (2.0.1).
+  - Updated Tensorflow version (2.12.0).
+
 ## New in Release 2.5.0
 
 Post-training Quantization:

codecov.yml

@@ -3,10 +3,11 @@ ignore:
   - "tests"
   - "tools"
 
-#codecov:
-#  notify:
-#     after_n_builds: 4
-#     wait_for_ci: yes
+codecov:
+  notify:
+     after_n_builds: 2
+     wait_for_ci: no
+  max_report_age: off
 
 coverage:
   status:
@@ -25,7 +26,8 @@ coverage:
         informational: true
         only_pulls: true
         paths:
-          - "nncf/onnx"  # extend this once we collect coverage reports for more than just onnx part of precommit
+          - "nncf/onnx"
+          - "nncf/common"  # extend this once we collect coverage reports for more than just onnx and common part of precommit
 
 comment:
   layout: "diff, flags, files"

docs/compression_algorithms/CompressWeights.md

@@ -1,8 +1,10 @@
 ### Weights Compression
 
+[OpenVINO](https://github.com/openvinotoolkit/openvino) is the preferred backend to run Weights Compression with, and PyTorch is also supported.
+
 #### The algorithm description
 
-The Weights Compression algorithm is aimed at compressing the weights of the models and can be used to optimize the model footprint and performance of large models where the size of weights is relatively larger than the size of activations, for example, Large Language Models (LLM). The algorithm compresses weights only for Linear and Embedding layers. It is also possible to keep the precision of the original weights and insert FakeQuantize operations by setting `use_fake_quantize` parameter to `True`.
+The Weights Compression algorithm is aimed at compressing the weights of the models and can be used to optimize the model footprint and performance of large models where the size of weights is relatively larger than the size of activations, for example, Large Language Models (LLM). The algorithm compresses weights only for Linear and Embedding layers.
 
 #### User guide
 
@@ -13,15 +15,8 @@ from nncf import compress_weights
 compressed_model = compress_weights(model)
 ```
 
-- Insert FakeQuantize layers for weights of linear layers and embeddings
-
-```python
-from nncf import compress_weights
-model_with_fake_quantize = compress_weights(model, use_fake_quantize=True)
-```
-
 ##### Limitations
 
-- The algorithm is supported for PyTorch only.
+- The algorithm is supported for OpenVINO and PyTorch models.
 - The compression applies in-place.
 - The compressed model is not trainable.

examples/post_training_quantization/onnx/mobilenet_v2/requirements.txt

@@ -4,4 +4,4 @@ scikit-learn
 fastdownload
 onnx~=1.13.1
 onnxruntime~=1.14.1
-openvino-dev==2023.0.1
+openvino-dev==2023.1

examples/post_training_quantization/openvino/anomaly_stfpm_quantize_with_accuracy_control/requirements.txt

@@ -1,2 +1,2 @@
 anomalib==0.6.0
-openvino-dev==2023.0.1
+openvino-dev==2023.1

examples/post_training_quantization/openvino/mobilenet_v2/requirements.txt

@@ -2,4 +2,4 @@ torchvision
 tqdm
 scikit-learn
 fastdownload
-openvino-dev==2023.0.1
+openvino-dev==2023.1