Inference bug-fixes: Re-enable EP syncs for the legacy A2A dispatcher and re-simplify ep_sync accidentally reverted by #4306 - #4587
Conversation
|
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:
See the contribution guide for more details. |
|
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. |
|
/claude review |
Co-authored-by: claude[bot] <209825114+claude[bot]@users.noreply.github.com>
|
🔄 Merge queue validation started! You can track the progress here: https://github.com/NVIDIA/Megatron-LM/actions/runs/25334581956 |
|
🔄 Merge queue validation started! You can track the progress here: https://github.com/NVIDIA/Megatron-LM/actions/runs/25340409008 |
|
🔄 Merge queue validation started! You can track the progress here: https://github.com/NVIDIA/Megatron-LM/actions/runs/25370163787 |
… and re-simplify ep_sync accidentally reverted by NVIDIA#4306 (NVIDIA#4587) Co-authored-by: claude[bot] <209825114+claude[bot]@users.noreply.github.com> Signed-off-by: yhgalaxy <yhgalaxy@outlook.com>
… and re-simplify ep_sync accidentally reverted by NVIDIA#4306 (NVIDIA#4587) Co-authored-by: claude[bot] <209825114+claude[bot]@users.noreply.github.com> Signed-off-by: Jon Barker <jbarker@aws-cmh-slurm-1-vscode-02.cm.cluster>
… and re-simplify ep_sync accidentally reverted by NVIDIA#4306 (NVIDIA#4587) Co-authored-by: claude[bot] <209825114+claude[bot]@users.noreply.github.com>
… and re-simplify ep_sync accidentally reverted by NVIDIA#4306 (NVIDIA#4587) Co-authored-by: claude[bot] <209825114+claude[bot]@users.noreply.github.com> Signed-off-by: Dmytro Pykhtar <dpykhtar@nvidia.com>
What does this PR do ?
Enables EP syncs for the legacy A2A dispatcher. Plus disables non-decode cudagraphs for it, just like we did for the ncclallgather dispatcher. This fixes a broken gitlab functional test.
also fixes an incorrect merge from #4306, that reintroduced deprecated ep-sync logic inside
batch_dimension_utils.pyIssue tracking
For PRs from open-source community contributors:
Linked issue:
Contribution process
Pre-checks
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"
.github/CODEOWNERS.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, theFinal Reviewlabel 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
Approvedlabel 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.comorzijiey@nvidia.com.