remove hardcoded torch.backends.cuda.matmul.allow_tf32 = True#18
Open
Parskatt wants to merge 4 commits intonaver:masterfrom
Open
remove hardcoded torch.backends.cuda.matmul.allow_tf32 = True#18Parskatt wants to merge 4 commits intonaver:masterfrom
Parskatt wants to merge 4 commits intonaver:masterfrom
Conversation
allow_tf32 can screw up stuff that relies on high precision, don't hardcode.
|
@yocabon Would be great if we could get this PR in - at least the allow_tf32 part I spent more than a day to find out why one (unrelated) model worked with cuda11.8, but not anymore with cuda>12 and it turned out it had to do with low precision matmuls which were enabled in croco.py |
This file contains hidden or 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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
allow_tf32 can screw up stuff that relies on high precision, don't hardcode.
In my case, it messed up torch.cdist so that a bunch of small values went to 0 (not good).