[guyueh1]

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

What does this PR do?

Fix bug: in-place modification of the routing_map tensor cause autograd failure

⚠️ 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
• If this PR adds or changes a GPU kernel (Triton, jit_fuser/torch.compile, CUDA extension, TE or external-library dispatch, or a scatter/index accumulation), I have added or updated its bit-exact determinism test and registered it in tests/unit_tests/determinism/kernels/manifest.py (guide)
• 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.

[guyueh1]

What does this PR do?

Fix an autograd aliasing issue between fused MoE routing and routing-map padding.

Bug

pad_routing_map() transposed the input routing map and then padded it in place. Because the transpose is a view, this also modified the original routing map and incremented its tensor version.

When the routing map came from the fused router, autograd had saved that tensor for backward. The later in-place modification caused backward to fail with a tensor-version mismatch, typically:

RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation

Scope

The issue affects MoE training configurations using:

• the alltoall token dispatcher;
• --moe-router-fusion;
• --moe-router-padding-for-quantization;
• FP8 or FP4 expert computation; and
• the standard, non-fused pad_routing_map() path.

Inference and configurations without router padding or fused routing are unaffected.

Fix

Clone the routing map before transposing and padding it. Padding now operates on independent storage, preserving the original tensor and its autograd version counter.

Unit test

Adds test_pad_routing_map_does_not_modify_input, which verifies that:

• the input routing map is unchanged;
• its tensor version counter is unchanged; and
• the returned map is padded to the requested multiple.

Reproduction

On an affected revision, add the following options to a working FP8 MoE training command:

torchrun --nproc-per-node=8 pretrain_gpt.py <existing-model-and-data-options> \
  --num-experts 8 \
  --expert-model-parallel-size 8 \
  --moe-token-dispatcher-type alltoall \
  --moe-router-fusion \
  --moe-router-padding-for-quantization \
  --fp8-format hybrid

[ko3n1g]

/review

[svcnemo-autobot]

External PR at bed4c72 not vetted, review thus rejected

[ko3n1g]

/ok to test bed4c72

[ko3n1g]

/review

[svcnvidia-nemo-ci]

LGTM

Review generated by codex

[nemo-automation-bot]

🔄 Merge queue validation started!

You can track the progress here: https://github.com/NVIDIA/Megatron-LM/actions/runs/35947791505