GRPO loss fix #535
Merged
GRPO loss fix #535
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
|
Yes, masked_mean is indeed correct to use here. Also check my comment at #534 (comment) The current kl1 estimator will completely ignore the ref_logprobs during trainining. kl2 and kl3 are fine. |
|
hamishivi
approved these changes
Jan 31, 2025
| @@ -487,6 +487,34 @@ def remove_padding(sequences, pad_token_id): | |||
| return [[inneritem for inneritem in item if inneritem != pad_token_id] for item in sequences] | |||
|
|
|||
|
|
|||
| class MetricsTracker: | |||
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Outdated suggestions cannot be applied.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.
CC @gauravpandeyamu @hamishivi for a second look.
I think the
pg_loss = pg_loss + (args.beta * kl).mean()is incorrect because the mean operation happens after the scores are added with the corresponding KLThe correct impl should be
Let me know what you think.