[GRPO] Fix: Processing ref logprobs in batches #3740
Merged
Conversation
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
Until commit d6a969f ref_per_token_logps used to be calculated inside _compute_loss or compute_liger_loss which used to get input of size per_device_batch_size. Since it moved to _generate_and_score_completions it now need to be passed a batch_size. Without it, memory consumption is not independent of the number of gradient accumulations.
|
The docs for this PR live here. All of your documentation changes will be reflected on that endpoint. The docs are available until 30 days after the last update. |
marcandrelarochelle
pushed a commit
to marcandrelarochelle/trl
that referenced
this pull request
Jul 29, 2025
Co-authored-by: LeonEricsson <70749762+LeonEricsson@users.noreply.github.com>
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Outdated suggestions cannot be applied.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.
Until commit d6a969f, ref_per_token_logps was calculated inside _compute_loss or compute_liger_loss, which gets input of size per_device_batch_size.
Since it moved to _generate_and_score_completions it now need to be passed a batch_size. Without it, memory consumption is not independent of the number of gradient accumulations and can spike when beta != 0