Megatron-FSDP: Hybrid FSDP over replicated weights, single-rank reduction and bucket padding fixes

[rapatel]

• I, the PR author, have personally reviewed every line of this PR.

Note: This PR is stacked on #6594. This branch includes that PR's commit as its base, so it should be reviewed and merged after it; the three commits here are the only new changes.

What does this PR do?

Allows Hybrid FSDP to shard optimizer state over DP-Outer on top of a DP-Inner strategy that leaves the model weights replicated, so ZeRO-1 on DP-Outer can sit above ZeRO-1/ZeRO-2 inner sharding rather than requiring ZeRO-3. Includes two independent fixes: gradients are now scaled correctly when the reduction group holds a single rank, which is wrong in main today whenever expert parallelism spans the data parallel domain, and parameter groups are split by chunk size factor instead of widening it to the least common multiple of incompatible row sizes. On DeepSeek-V3 671B the resulting ZeRO-3(experts) + ZeRO-2(dense) + ZeRO-1(DP-Outer) configuration measures 1450 TFLOP/s/GPU against 1195 for uniform ZeRO-3, matching its loss curve.

⚠️ For major changes (either in lines of code or in its impact), please make sure to first share a design doc with the team. If you're unsure what's the best way to do so, contact @NVIDIA/mcore-oncall.

Issue tracking

For PRs from open-source community contributors:

• New features: a linked issue is required. Please open a feature request and reference it here before submitting the PR.
• Small updates (bug fixes, minor improvements): a linked issue is recommended and will accelerate the PR review process.

Linked issue:

Contribution process

Pre-checks

• I have added relevant unit tests
• I have added relevant functional tests
• I have added proper typing to my code Typing guidelines
• I have added relevant documentation
• I have run the autoformatter.sh on my PR

Code review

Feel free to message or comment @NVIDIA/mcore-oncall to help accelerate your merge into main. The less complex your PR is, the faster it will be approved and merged!

All PRs start as draft. If you open a non-draft PR, it will be automatically converted to draft.

Step 1: Mark PR as "Ready for Review"

1. When your PR is ready, click Ready for Review.
2. An oncall reviewer is auto-assigned and expert reviewers are notified based on your changes.
   • Some PRs may jump straight to step 2. This is determined by .github/CODEOWNERS.

⚠️ Only mark as ready once merge-conflicts are resolved and the CI is passing.
Final Review might get declined if these requirements are not fulfilled.

Step 2: Final Review

For PRs that change megatron/core, once all expert reviewers have approved, the Final Review label is applied automatically and final reviewers are assigned.

For PRs outside megatron/core, this step is skipped.

Step 3: Approved

Once all required reviewers have approved, the Approved label is applied automatically.

Merge

Any member of mcore-engineers will be able to merge your PR.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[github-actions]

This PR has been automatically converted to draft because all PRs must start as drafts.

When you are ready for review, click Ready for Review to begin the review process. This will:

1. Add the oncall reviewer (optional reviewer)
2. Add required review teams based on your changes

See the contribution guide for more details.

[wujingyue]

Is it a stack? Can you diffbase according to skills/mcore-split-pr/SKILL.md?