FS#73668 - python-pytorch-cuda needs to be recompiled with cuda 11.6
Attached to Project:
Community Packages
Opened by Benoit Brummer (trougnouf) - Monday, 07 February 2022, 13:42 GMT
Last edited by Sven-Hendrik Haase (Svenstaro) - Sunday, 27 February 2022, 19:03 GMT
Opened by Benoit Brummer (trougnouf) - Monday, 07 February 2022, 13:42 GMT
Last edited by Sven-Hendrik Haase (Svenstaro) - Sunday, 27 February 2022, 19:03 GMT
|
Details
Description:
Cannot import pytorch because it links to previous cuda library. Additional info: * package version(s): python-pytorch-cuda 1.10.2-1, cuda 11.6.0-1 Steps to reproduce: [trougnouf@d]: ~>$ python Python 3.10.2 (main, Jan 15 2022, 19:56:27) [GCC 11.1.0] on linux Type "help", "copyright", "credits" or "license" for more information. >>> import torch Traceback (most recent call last): File "<stdin>", line 1, in <module> File "/usr/lib/python3.10/site-packages/torch/__init__.py", line 197, in <module> from torch._C import * # noqa: F403 ImportError: libcupti.so.11.5: cannot open shared object file: No such file or directory >>> |
This task depends upon
Closed by Sven-Hendrik Haase (Svenstaro)
Sunday, 27 February 2022, 19:03 GMT
Reason for closing: Fixed
Sunday, 27 February 2022, 19:03 GMT
Reason for closing: Fixed
the git version complains that the cuda compiler identification is unknown, I guess pytorch isn't compatible yet and (imo) cuda got updated too early.
downgrading cuda/cudnn restores functionality. (cuda 11.5.1-1, cudnn 8.3.0.98-1)
10error.log (14.4 KiB)
10error.log (168.8 KiB)