0

possibly a very simple error but wanted to ask as I am unable to understand what the issue is or find an appropriate solution.

I am working on Ubuntu 22.04.2 LTS and while checking nvcc -V, I see: Build cuda_11.2.r11.2/compiler.29618528_0

With nvidia-smi, I see: NVIDIA-SMI 470.182.03 Driver Version: 470.182.03 CUDA Version: 11.4

If I am reading the NVIDIA docs correctly, everything should be compatible.

https://docs.nvidia.com/deploy/cuda-compatibility/

What am I missing? Why is this error popping up?

...torch/utils/cpp_extension.py", line 991, in _get_cuda_arch_flags raise ValueError("Unknown CUDA arch ({}) or GPU not supported".format(arch)) ValueError: Unknown CUDA arch (8.0) or GPU not supported

Any help and especially explanation would be appreciated as I am none the wiser from the other posts.

Short description above

talonmies
  • 70,661
  • 34
  • 192
  • 269
Vinay Pradhan
  • 1
  • 1
    from [here](https://stackoverflow.com/questions/66968382/valueerror-unknown-cuda-arch-8-6-or-gpu-not-supported): "Assuming this is related to pytorch keep in mind that pytorch does not use your system's cuda. It installs it's own version along with the pytorch binaries. See this answer and this answer for more details on how to get pytorch working with CUDA." You haven't indicated what pytorch version you are trying to use, but it is evidently pulling in a version of CUDA that is 10.x or older, somewhere in your environment. – Robert Crovella Aug 22 '23 at 15:12
  • It means the version of PyTorch you are using lacks support for your GPU – talonmies Aug 22 '23 at 15:12
  • Thanks for the link @RobertCrovella, I am reading through now. – Vinay Pradhan Aug 22 '23 at 15:13
  • Thanks @talonmies. I am trying to find out how can i find a compatibility table for further clarification. – Vinay Pradhan Aug 22 '23 at 15:14
  • Related: https://stackoverflow.com/a/76882398/11790637 – ihdv Aug 23 '23 at 15:07

0 Answers0