Request torch.cuda RNG states only if CUDA is available
#19234
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
awaelchli
added
bug
awaelchli
added
the
fun
Codecov Report
Additional details and impacted files@@ Coverage Diff @@
## master #19234 +/- ##
=========================================
- Coverage 65% 54% -11%
=========================================
Files 445 440 -5
Lines 37289 37191 -98
=========================================
- Hits 24285 20075 -4210
- Misses 13004 17116 +4112 |
Borda
approved these changes
Jan 10, 2024
carmocca
approved these changes
Jan 10, 2024
awaelchli
added a commit
that referenced
this pull request
Jan 30, 2024
lexierule
pushed a commit
that referenced
this pull request
Jan 31, 2024
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What does this PR do?
If the driver on your system is outdated for the version of PyTorch installed,
torch.cuda.is_available()will returnFalsewith a warning to update the driver. If the user then tries to init CUDA, a RuntimeError is raised:In Lightning, we don't select the CUDA accelerator unless
torch.cuda.is_available()isTrue, but if it returns False and the driver is broken, we still can get an initialization error when attempting to get the rng states here:pytorch-lightning/src/lightning/fabric/utilities/seed.py
Line 113 in f75f3bc
A user ran into this: https://lightning.ai/forums/t/cant-train-with-a-too-old-nvidia-driver-even-with-cpu-accelerator/4634
This PR makes it so that you can run on CPU even if your CUDA driver is broken.
cc @Borda @carmocca @justusschock @awaelchli