The detected CUDA version (12.0) mismatches the version that was used to compile PyTorch (11.7)

1

Hi!
I get this runtimeerror when trying to run my code:

File “C:\Users\isakf\anaconda3\envs\fastreg\lib\site-packages\torch\utils\cpp_extension.py”, line 386, in _check_cuda_version
raise RuntimeError(CUDA_MISMATCH_MESSAGE.format(cuda_str_version, torch.version.cuda))
RuntimeError:
The detected CUDA version (12.0) mismatches the version that was used to compile

This message is from the torch.utils.cpp_extension module.

Im struggling to see where I can download the torch version with CUDA 12 enabled.

Do you have any tips?

2

The PyTorch binaries do not support CUDA 12 yet and you would have to build from source using a locally installed CUDA 12.0 toolkit.

2 Likes
3

Hello, just curious if there is any update on pip torch supporting xformers with cuda 12. Thank you!

4

We are already building nightly PyTorch binaries with CUDA 12.1 but I don’t know which CUDA versions are supported in xformers.

1 Like
5

same problem!!, thanks. The problem with xformers, is only in the installation.