Make weight and optimizer memory estimation take into account expert parallelism correctly

[YangFei1990]

What does this PR do ?

Fix memory estimation. Previously the function counted MoE routed expert parameters together with regular transformer parameters, so per-rank weight and optimizer memory did not account for expert tensor parallelism, expert parallelism, or expert data parallelism.

This change separates regular TP-sharded parameters, replicated non-expert parameters, and routed expert parameters. Routed experts now use ETP/EP for parameter ownership and EDP for distributed optimizer sharding, while shared experts remain on regular TP.

Issue tracking

Fixes #4050

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 the @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.

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.com or zijiey@nvidia.com.

[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.

[copy-pr-bot]

Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually.

Contributors can view more details about this message here.

[YangFei1990]

/claude review

[claude]

LGTM — the decomposition into TP-sharded, replicated, and routed-expert parameter categories is correct, and using separate data-parallel sizes for the distributed optimizer byte calculation is the right fix.

One suggestion: compute_weight_and_optimizer_memory has no unit tests today, and this PR rewrites most of its arithmetic. A parameterized unit test that constructs a mock args namespace and asserts known outputs for a few configurations (dense-only, MoE with ETP/EP > 1, distributed optimizer on/off, untied embeddings with PP=1) would protect against future regressions and make reviewing easier. The PR checklist also has the test boxes unchecked — worth adding before merging.

[YangFei1990]

/ok to test 60d595e

[deepakn94]

/ok to test 541a0d8

[svcnvidia-nemo-ci]

🔄 Merge queue validation started!

You can track the progress here: https://github.com/NVIDIA/Megatron-LM/actions/runs/25836565088