[TorchAO] [OVQuantizer] Update Torch AO Dependency and Refactor Executorch Tests

[anzr299]

Changes

Include tests for executorch, OVQuantizer
Remove OVQuantizer from NNCF
Move the tests in appropriate directory

Reason for changes

Migrate from torch.ao -> torchao

Notable differences in torch.ao and torchao implementation which affect test references:

1. Constant folding in torch.ao produced the new constant with the name _frozen_param{index}. In the new torchao implementation it keeps the same name as the node being replaced with.
2. The old deprecated XNNPACK Quantizer was being used before which is now changed to point to the new XNNPACKQuantizer implementation in https://github.com/pytorch/executorch/tree/main/backends/xnnpack/quantizer which now explicitly quantizes the conv_transpose2D operator and its weights.

Reference graphs changed due to above:

1. tests/executorch/data/ao_export_quantization_OpenVINOQuantizer/mobilenet_v3_small.dot
2. tests/executorch/data/ao_export_quantization_OpenVINOQuantizer/resnet18.dot
3. tests/executorch/data/ao_export_quantization_OpenVINOQuantizer/swin_v2_t.dot
4. tests/executorch/data/ao_export_quantization_OpenVINOQuantizer/synthetic_transformer.dot
5. tests/executorch/data/ao_export_quantization_OpenVINOQuantizer/unet.dot
6. tests/executorch/data/ao_export_quantization_OpenVINOQuantizer/vit_b_16.dot
7. tests/executorch/data/ao_export_quantization_OpenVINOQuantizer/yolo11n_sdpa_block.dot
8. tests/executorch/data/XNNPACKQuantizer/unet.dot (Due to difference Number 2)
9. tests/executorch/data/XNNPACKQuantizer/unet_ref_qconfig.json ((Due to difference Number 2))

Related tickets

CVS-176783

Tests

PTQ Conformance - PTQ-789 - Pass
Examples - https://github.com/openvinotoolkit/nncf/actions/runs/21204888584 - Pass
WC conformance - https://github.com/openvinotoolkit/nncf/actions/runs/21204893707 - Pass

pytest tests/executorch results:

All tests in tests/executorch pass

[MaximProshin]

Agreed to merge it after the CF to mitigate any risks with a new dependency. @anzr299 @AlexanderDokuchaev

[daniil-lyakhov]

Aamir, as we discussed it with Alexander Suslov, the main idea was to get rid of torch.ao and do not use any external dependencies (like torchao)
The task is to remove torch.ao imports and replace them with code in the nncf codebase, no external dependencies

[Copilot]

Pull request overview

This PR migrates the codebase from using torch.ao to torchao as a dependency, updates the OpenVINOQuantizer to use the executorch implementation, and refactors the test structure for executorch tests. The migration addresses breaking changes in PyTorch's quantization API and aligns with the newer torchao package architecture.

Changes:

• Migration of all torch.ao.quantization imports to torchao.quantization.pt2e
• Removal of NNCF's internal OpenVINOQuantizer implementation in favor of executorch's version
• Introduction of quantizer_params alongside compression_params for better parameter separation
• Update of test reference files to reflect behavior changes in torchao vs torch.ao
• Addition of torchao and executorch dependencies across requirements files

Reviewed changes

Copilot reviewed 58 out of 100 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
tests/torch/requirements.txt	Added torchao dependency
tests/torch/fx/test_model_transformer.py	Updated imports from torch.ao to torchao
tests/post_training/test_quantize_conformance.py	Added quantizer_params support and printing
tests/post_training/requirements.txt	Added torchao and executorch, updated scikit-learn version
tests/post_training/pipelines/image_classification_base.py	Updated imports and refactored parameter handling
tests/post_training/pipelines/base.py	Added quantizer_params parameter throughout pipeline
tests/post_training/model_scope.py	Added add_quantizer_compression_params function
tests/executorch/test_quantizer_compression.py	Updated test data directory paths
tests/executorch/test_ptq.py	Updated imports and test references
tests/executorch/requirements.txt	Updated dependency versions
tests/executorch/data/*	Multiple new reference files for test data
tests/cross_fw/shared/helpers.py	Added torchao to backend packages
src/nncf/torch/quantization/strip.py	Updated observer imports
src/nncf/quantization/algorithms/min_max/torch_fx_backend.py	Updated observer imports
src/nncf/experimental/torch/fx/transformations.py	Updated imports and comments
src/nncf/experimental/torch/fx/quantization/quantizer/torch_ao_adapter.py	Updated imports and documentation
src/nncf/experimental/torch/fx/quantization/quantizer/openvino_quantizer.py	File removed
src/nncf/experimental/torch/fx/quantization/quantizer/openvino_adapter.py	Updated to use executorch's OpenVINOQuantizer
src/nncf/experimental/torch/fx/quantization/quantize_pt2e.py	Updated imports and removed internal quantizer reference
src/nncf/experimental/torch/fx/quantization/quantize_model.py	Updated imports
src/nncf/experimental/torch/fx/init.py	Removed OpenVINOQuantizer export
src/nncf/common/tensor_statistics/collectors.py	Updated comment reference
examples/*/requirements.txt	Added torchao dependency
constraints.txt	Added torchao==0.14.0 constraint

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.