Repository navigation
Re-add fused GDN TransformerEngine integration - #7583
Merged
Merged
Conversation
Re-add the GDN TransformerEngine integration from upstream PR NVIDIA#6645 (54c62df), which was reverted in PR NVIDIA#7326 (1e137dd) as part of a combined revert for unrelated CI failures. Adds a TransformerEngine GatedDeltaNetAttention kernel backend option (gdn_kernel_backend: transformer_engine|fla|torch) alongside the existing FLA/torch paths, resolved against the current gdn.py two-stage forward split introduced by the ShortcutMoE change (NVIDIA#6959) that landed after the revert. Fixes a golden-config mismatch in the original commit: the gdn_kernel_backend default is "fla", but tests/unit_tests/models/test_hybrid_moe_model.py's GOLDEN_CONFIG had recorded "transformer_engine", which would have failed test_constructor's config-drift check. Co-authored-by: gautham-kollu <gkollu@nvidia.com> Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Signed-off-by: Kirthi Shankar Sivamani <ksivamani@nvidia.com>
svcnvidia-nemo-ci
marked this pull request as draft
September 22, 2026 15:13
Contributor
|
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. |
ksivaman
marked this pull request as ready for review
September 22, 2026 15:16
zhongbozhu
approved these changes
Sep 22, 2026
Contributor
|
/ok to test 5a3f5f3 |
Phlip79
approved these changes
Sep 22, 2026
jaredcasper
approved these changes
Sep 22, 2026
svcnvidia-nemo-ci
added
Approved
Contributor
|
Deterministic mode is not supported with TE backend. Hence adding "exempt" label |
Contributor
|
/claude fix |
Contributor
|
❌ Claude fix stopped because a workflow step failed. Inspect the run. |
Member
Author
|
/ok to test 4fbd64f |
Signed-off-by: Kirthi Shankar Sivamani <ksivamani@nvidia.com>
Member
Author
|
/ok to test bf3c4fb |
Contributor
|
/ok to test 6f31b04 |
Signed-off-by: Kirthi Shankar Sivamani <ksivamani@nvidia.com>
Member
Author
|
/ok to test 6078a33 |
|
🔄 Merge queue validation started! You can track the progress here: https://github.com/NVIDIA/Megatron-LM/actions/runs/38062626973 |
|
🔄 Merge queue validation started! You can track the progress here: https://github.com/NVIDIA/Megatron-LM/actions/runs/38063492099 |
5 tasks
3 of 5 tasks
This branch was previously deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Re-add the GDN TransformerEngine integration from upstream #6645 which was reverted in #7326 as part of a combined revert for unrelated CI failures.
Pre-checks
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 intests/unit_tests/determinism/kernels/manifest.py(guide)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"
.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.