fix(ppo): restore offload_after_refit semantics - #2782
Merged
yuki-97 merged 2 commits intoJun 11, 2026
Merged
Conversation
Signed-off-by: bg51717 <biguo@nvidia.com>
…er-refit-onload-regression
Member
Author
|
/ok to test 76dd30f |
yuki-97
reviewed
Jun 11, 2026
yuki-97
added
CI:docs
yuki-97
approved these changes
Jun 11, 2026
yuki-97
enabled auto-merge (squash)
June 11, 2026 17:11
Contributor
|
https://github.com/NVIDIA-NeMo/RL/actions/runs/27351602878/attempts/3?pr=2782
|
pengdurice
pushed a commit
to pengdurice/RL
that referenced
this pull request
Jun 12, 2026
Signed-off-by: bg51717 <biguo@nvidia.com>
ashors1
pushed a commit
that referenced
this pull request
Jun 27, 2026
Signed-off-by: bg51717 <biguo@nvidia.com> Signed-off-by: Anna Shors <ashors@nvidia.com>
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
2 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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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 ?
Restores the offload_after_refit() contract that PR #2530 accidentally broke, so the Megatron PPO policy worker actually offloads params to CPU after a colocated refit instead of leaving them on GPU. Aligns the policy lifecycle with GRPO/SFT/DPO (no finish_training on the policy side), and adds an explicit prepare_for_lp_inference() before refit_policy_generation so params are guaranteed on GPU when entering refit (covers the ckpt-only-step path that previously relied on the removed auto-onload).
Issues
close #2775