You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
As I was configuring different combinations of pytorch and CUDA I realized if your device's current nvvc version and torch's compiled cuda version differ- which is totally ok for torch itself since you can run any code on the new cuda 11.5 machines while torch lags behind at 11.3 - you will receive an error for mismatched versions for installing extensions through distutils and cppextensions.
Of course I realize reverting the torch version back to 1.9 is totally fine and you can easily install the package. Any plans for a change in the structure of extensions(Sorry I am not familiar how exactly you could bypass this mismatch issue).
The text was updated successfully, but these errors were encountered:
Hi,
As I was configuring different combinations of
pytorch
andCUDA
I realized if your device's currentnvvc
version and torch's compiled cuda version differ- which is totally ok for torch itself since you can run any code on the newcuda 11.5
machines while torch lags behind at11.3
- you will receive an error for mismatched versions for installing extensions throughdistutils
andcppextensions
.Of course I realize reverting the
torch
version back to1.9
is totally fine and you can easily install the package. Any plans for a change in the structure of extensions(Sorry I am not familiar how exactly you could bypass this mismatch issue).The text was updated successfully, but these errors were encountered: