[Good First Issue][NNCF]: Suppress UserWarning: TORCH_CUDA_ARCH_LIST

[alexsu52]

Context

This issue relatives to improve user experience of NNCF PyTorch examples on GPU. The output of such examples has the following warning:

INFO:nncf:Compiling and loading torch extension: quantized_functions_cuda...
/nncf_torch/lib/python3.10/site-packages/torch/utils/cpp_extension.py:1964: UserWarning: TORCH_CUDA_ARCH_LIST is not set, all archs for visible cards are included for compilation. 
If this is not desired, please set os.environ['TORCH_CUDA_ARCH_LIST'].
  warnings.warn(

This warning was triggered by compilation custom quantization kernels in NNCF.

Requirement: GPU

How to reproduce:
python examples/post_training_quantization/torch/mobilenet_v2/main.py

What needs to be done?

Suppress UserWarning: TORCH_CUDA_ARCH_LIST by adding visible card architectures in TORCH_CUDA_ARCH_LIST before compilation in case if TORCH_CUDA_ARCH_LIST is not set by user. The list of CUDA architectures can be gotten by torch.cuda.get_arch_list()

Example Pull Requests

No response

Resources

• Contribution guide - start here!
• Intel DevHub Discord channel - engage in discussions, ask questions and talk to OpenVINO developers
• How to link your Pull Request to an issue

Contact points

@alexsu52
@AlexanderDokuchaev

Ticket

No response

[devesh-2002]

Hi. May I work upon this issue?

[alexsu52]

@devesh-2002, sure. you can use take to assign for yourself.

[devesh-2002]

.take

[github-actions]

Thank you for looking into this issue! Please let us know if you have any questions or require any help.