Fix split_state_dict function for MoE models#3667
Open
eternally-z wants to merge 3 commits intoNVIDIA:devfrom
Open
Fix split_state_dict function for MoE models#3667eternally-z wants to merge 3 commits intoNVIDIA:devfrom
eternally-z wants to merge 3 commits intoNVIDIA:devfrom
Conversation
kunlunl
reviewed
Mar 3, 2026
kunlunl
approved these changes
Mar 3, 2026
Contributor
|
/ok to test c65fd5c |
Contributor
|
/ok to test e867137 |
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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 ?
main PR: #3676
This PR fixes an issue in _split_state_dict, which splits the state dict into sub-state dicts for chained optimizers.
The previous offset-based implementation failed in MoE scenarios. Since MoE parts often have a separate optimizer, the offset-based splitting logic caused chunk mis-assignment. As a result, the MoE optimizer was unable to obtain the correct state_dicts.
@AniZpZ
Contribution process
flowchart LR A[Pre-checks] --> B[PR Tests] subgraph Code Review/Approval C1[Expert Review] --> C2[Final Review] end B --> C1 C2 --> D[Merge]Pre-checks
Core 0.8)Code review
The following process is enforced via the CODEOWNERS file for changes into
megatron/core. For changes outside ofmegatron/core, it is up to the PR author whether or not to tag the Final Reviewer team.For MRs into `main` branch
Feel free to message or comment the @mcore-oncall to help accelerate your merge into main. The less complex your PR is, the faster it will be approved and merged!
(Step 1): Add PR label
Expert Review(Step 2): Collect the expert reviewers reviews
Expert Reviewlabel when your PR is ready for review.Final Review might get declined if these requirements are not fulfilled.
(Step 3): Final Review
Final Reviewlabel(Optional Step 4): Cherry-pick into release branch
If this PR also needs to be merged into
core_r*release branches, after this PR has been merged, selectCherry-pickto open a new PR into the release branch.For MRs into `dev` branch
The proposed review process for `dev` branch is under active discussion.MRs are mergable after one approval by either
eharper@nvidia.comorzijiey@nvidia.com.Merging your PR
Any member of core-adlr and
core-nemowill be able to merge your PR.