[MoE]Enable Casting-Free FP8-Flow-MoE Blockwise FP8 Dataflow

[xiaoxi-wangfj]

What does this PR do ?

This PR introduces a set of FP8-centric MoE dataflow optimizations that improve FP8 communication performance and reduce CPU overhead through operator fusion, while maintaining numerical stability in existing FP8 training workflows.

This PR is designed to be used in conjunction with PR NVIDIA/TransformerEngine#2544

Further optimizations are introduced via two additional PRs:

• Fuse permute+pad and unpermute+unpad ops for FP8/FP4 training  #2763
• [PyTorch] Fuse permute+pad and unpermute+unpad ops for FP8 optimization TransformerEngine#1921

Background / Motivation

Detailed optimization points are described in the joint PR NVIDIA/TransformerEngine#2544

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]

Loading

Pre-checks

• I want this PR in a versioned release and have added the appropriate Milestone (e.g., Core 0.8)
• 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

The following process is enforced via the CODEOWNERS file for changes into megatron/core. For changes outside of megatron/core, it is up to the PR author whether or not to tag the Final Reviewer team.

For MRs into `main` branch

(Step 1): Add PR label Expert Review

(Step 2): Collect the expert reviewers reviews

1. Attach the Expert Review label when your PR is ready for review.
2. GitHub auto-assigns expert reviewers based on your changes. They will get notified and pick up your PR soon.

⚠️ Only proceed to the next step once all reviewers have approved, merge-conflict are resolved and the CI is passing.
Final Review might get declined if these requirements are not fulfilled.

(Step 3): Final Review

1. Add Final Review label
2. GitHub auto-assigns final reviewers based on your changes. They will get notified and pick up your PR soon.

(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, select Cherry-pick to 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.com or zijiey@nvidia.com.

Merging your PR

Any member of core-adlr and core-nemo 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.

[Phlip79]

Can you please add tests for the new features?

[Phlip79]

Please add tests

[xiaoxi-wangfj]

@Phlip79 tests are added.

[Phlip79]

/claude review

[claude]

Two small issues flagged inline:

1. Bug in FusedCombine.forward (fused_a2a.py:201-202): ctx.fp8_recipe = config.fp8_recipe will crash when config is None — the preceding line correctly guards against this but this one doesn't.

2. Typo in error message (transformer_config.py:1436): "moe_expert recompute_modules is supported fp8 blockwise recipe" is missing "only" and "with".

[Phlip79]

/ok to test 2dfcf5c

[Phlip79]

Please fix Claude's comments