Synchronize the request counts for EP inference with strict matching

[santhnm2]

What does this PR do ?

Synchronizes the decode and prefill request counts for EP inference with strict matching. Otherwise the best graph config selected could vary on each rank (in particular, the token counts) and lead to hangs. Example observed when running ShareGPT trace:

[Rank 4] local_batch_dims=[20]: 1 P + 16 D, adjusted_batch_dim=[22]: 1 P + 16 D
[Rank 0] local_batch_dims=[22]: 1 P + 16 D, adjusted_batch_dim=[22]: 1 P + 16 D
[Rank 3] local_batch_dims=[20]: 1 P + 16 D, adjusted_batch_dim=[22]: 1 P + 16 D
[Rank 6] local_batch_dims=[21]: 1 P + 17 D, adjusted_batch_dim=[22]: 1 P + 17 D
[Rank 1] local_batch_dims=[22]: 1 P + 16 D, adjusted_batch_dim=[22]: 1 P + 16 D
[Rank 5] local_batch_dims=[20]: 1 P + 16 D, adjusted_batch_dim=[22]: 1 P + 16 D
[Rank 7] local_batch_dims=[21]: 1 P + 17 D, adjusted_batch_dim=[22]: 1 P + 17 D
[Rank 2] local_batch_dims=[20]: 1 P + 16 D, adjusted_batch_dim=[22]: 1 P + 16 D

[Rank 0] best_batch_dim=[32]: 16 P + 16 D
[Rank 3] best_batch_dim=[32]: 16 P + 16 D
[Rank 6] best_batch_dim=[64]: 16 P + 48 D
[Rank 5] best_batch_dim=[32]: 16 P + 16 D
[Rank 1] best_batch_dim=[32]: 16 P + 16 D
[Rank 7] best_batch_dim=[64]: 16 P + 48 D
[Rank 4] best_batch_dim=[32]: 16 P + 16 D
[Rank 2] best_batch_dim=[32]: 16 P + 16 D

⚠️ 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 the @mcore-oncall.

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

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

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.

[sidsingh-nvidia]

LGTM, thank you!

[santhnm2]

/ok to test ec27195

[sidsingh-nvidia]

Can you check whether this works with full_iteration cuda graphs?