fix: deterministic implementation of _float32_rms_norm_fwd#2971
Open
ashors1 wants to merge 3 commits into
Open
fix: deterministic implementation of _float32_rms_norm_fwd#2971ashors1 wants to merge 3 commits into
ashors1 wants to merge 3 commits into
Conversation
Signed-off-by: Anna Shors <ashors@nvidia.com>
Contributor
|
/claude review |
Contributor
|
/ok to test 4ced4b6 |
![claude[bot]](https://avatars.githubusercontent.com/in/1236702?s=60&v=4){kind=small}
Signed-off-by: Anna Shors <ashors@nvidia.com>
Contributor
|
/ok to test e3a3de0 |
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
3 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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 ?
This PR fixes an issue we're observing with DPO in nemo-rl where the initial loss does not match expectations. Please refer to NVIDIA-NeMo/RL#2662 for details.
Essentially, the current implementation of fp32 rmsnorm causes nondeterminism across grad/no_grad contexts, which is problematic for DPO where we expect the forward passes to be fully deterministic. This PR fixes that nondeterminism by ensuring that rmsnorm is treated as an atomic operation.
Changelog
Before your PR is "Ready for review"
Pre checks:
If you haven't finished some of the above items you can still open "Draft" PR.
Additional Information