Repository navigation
Reduce memory consumption in batched_forward_pass - #234
Merged
Merged
Conversation
atsumoto
changed the title
|
The documentation is not available anymore as the PR was closed or merged. |
younesbelkada
approved these changes
Mar 21, 2023
younesbelkada
left a comment
younesbelkada
left a comment
Contributor
There was a problem hiding this comment.
Thanks a lot for fixing and for taking care of the memory consumption
This looks very good to me!
Would love to hear @lvwerra 's thoughts here
yxliu-TAMU
pushed a commit
to mincheolseong/ECEN743-GRPO-Project-Proposal
that referenced
this pull request
Apr 20, 2025
* Reduce memory consumption by not storing logits in forward_pass * Add docstring of return_logits
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
4 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
This PR reduces memory consumption in
batched_forward_passof PPOTrainer, by avoiding the storage of logits when they are not necessary.Before this PR,
batched_forward_passstored all of the model'slogitsall the time like other tensors such asvaluesandlogprobs. Here,logitstensors have a much larger size (batch_size * tokens * vocabulary_size) compared tologprobsandvaluestensors (batch_size × tokens), consuming a significant amount of cuda memory.I have modified
batched_forward_passto avoid unnecessary storage oflogits, which is only required when calculating entropy in thelossmethod.