Bump torch to 2.11.0

[AlexanderDokuchaev]

Changes

• torch==2.11
• onnx-ir==0.2.1
• onnxscript==0.7.0
• Update cuda to 13
• Use torch.export.export instead of removed torch.export.export_for_training
• Temporary disables some testcase for linalg.cholesky_inverse (it falls only on t4)
• Disable test of gptqmodel example
• -removed disable_tf32_precision

Reason for changes

Related tickets

Tests

Test examples - success
Weight compression - success
PTQ-850

[Copilot]

Pull request overview

Updates NNCF’s dependency stack and test/export utilities to be compatible with PyTorch 2.11 (and associated ONNX tooling), including CI updates for CUDA 13.

Changes:

• Bump PyTorch/TorchVision and ONNX-related pinned versions across constraints and example requirements.
• Replace torch.export.export_for_training usages with torch.export.export across tests/examples.
• Update CI workflows to install CUDA 13.2.1 and adjust/disable specific tests/examples impacted by the upgrade.

Reviewed changes

Copilot reviewed 37 out of 37 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
tests/torch/test_tensor.py	Overrides CUDA cholesky_inverse test cases and adds a TODO note for T4-related failures.
tests/torch/fx/test_model_transformer.py	Updates export API usage and adjusts a comment near SWIN capture reproduction.
tests/torch/fx/helpers.py	Switches FX capture helper from export_for_training to export.
tests/torch/function_hook/test_wrapper.py	Adds explicit ids for dynamo/no-dynamo parametrization.
tests/torch/conftest.py	Updates global fp32 precision backend settings for tests.
tests/post_training/pipelines/image_classification_torchvision.py	Switches torchvision pipeline export function to torch.export.export.
tests/post_training/pipelines/fx_modelling.py	Switches HuggingFace export path to torch.export.export.
tests/cross_fw/examples/example_scope.json	Removes GPTQModel converter example entry (disables that example test).
src/nncf/torch/function_hook/handle_inner_functions.py	Updates internal MHA forward handling to match newer torch internals and replaces assert statements.
examples/quantization_aware_training/torch/resnet18/requirements.txt	Bumps torch/torchvision pins to 2.11.0/0.26.0.
examples/quantization_aware_training/torch/anomalib/requirements.txt	Bumps torch pin to 2.11.0.
examples/pruning/torch/resnet18/requirements.txt	Bumps torch/torchvision pins to 2.11.0/0.26.0.
examples/post_training_quantization/torch_fx/resnet18/requirements.txt	Bumps torch/torchvision pins to 2.11.0/0.26.0.
examples/post_training_quantization/torch/ssd300_vgg16/requirements.txt	Bumps torch/torchvision pins to 2.11.0/0.26.0.
examples/post_training_quantization/torch/mobilenet_v2/requirements.txt	Bumps torch/torchvision pins to 2.11.0/0.26.0.
examples/post_training_quantization/openvino/yolov8_quantize_with_accuracy_control/requirements.txt	Bumps torch pin to 2.11.0.
examples/post_training_quantization/openvino/yolo26/requirements.txt	Bumps torch pin to 2.11.0.
examples/post_training_quantization/onnx/yolov8_quantize_with_accuracy_control/requirements.txt	Bumps torch pin to 2.11.0.
examples/post_training_quantization/onnx/mobilenet_v2/requirements.txt	Bumps torch/torchvision pins to 2.11.0/0.26.0.
examples/llm_compression/torch_fx/tiny_llama/requirements.txt	Bumps torch/torchvision pins to 2.11.0/0.26.0.
examples/llm_compression/torch_fx/tiny_llama/modelling.py	Switches export path to torch.export.export.
examples/llm_compression/torch/downstream_qat_with_nls/requirements.txt	Bumps torch pin to 2.11.0.
examples/llm_compression/torch/distillation_qat_with_lora/requirements.txt	Bumps torch pin to 2.11.0.
examples/llm_compression/openvino/tiny_llama_synthetic_data/requirements.txt	Bumps torch pin to 2.11.0.
examples/llm_compression/openvino/tiny_llama_find_hyperparams/requirements.txt	Bumps torch pin to 2.11.0.
examples/llm_compression/openvino/tiny_llama/requirements.txt	Bumps torch pin to 2.11.0.
examples/llm_compression/openvino/smollm2_360m_fp8/requirements.txt	Bumps torch pin to 2.11.0.
examples/llm_compression/openvino/smollm2_360m_codebook/requirements.txt	Bumps torch/torchvision pins to 2.11.0/0.26.0.
examples/llm_compression/openvino/smollm2_360m_adaptive_codebook/requirements.txt	Bumps torch/torchvision pins to 2.11.0/0.26.0.
examples/llm_compression/onnx/tiny_llama_scale_estimation/requirements.txt	Bumps torch pin to 2.11.0.
examples/llm_compression/onnx/tiny_llama/requirements.txt	Bumps torch pin to 2.11.0.
docs/Installation.md	Updates documented PyTorch/CUDA versions for develop.
constraints.txt	Updates pinned torch/torchvision/onnx-ir/onnxscript versions.
README.md	Updates TorchFX export snippet to use torch.export.export.
.github/workflows/install.yml	Updates CI CUDA installer from 12.6.3 to 13.2.1.
.github/workflows/examples.yml	Updates examples CI CUDA installer from 12.6.3 to 13.2.1.
.github/workflows/call_precommit.yml	Updates pre-commit CI CUDA installer from 12.6.3 to 13.2.1.

Comments suppressed due to low confidence (1)

tests/post_training/pipelines/image_classification_torchvision.py:33

• There are two identical _torch_export function definitions back-to-back. This duplication is easy to miss and makes future edits error-prone; please remove one definition (and ensure only one canonical export helper is used in this module).