[None][chore] Remove the duplicate Rubin fused FC12 op registration by leslie-fang25 · Pull Request #19476 · NVIDIA/TensorRT-LLM

leslie-fang25 · 2026-09-21T06:23:20Z

#19184 synced a second copy of Fc12FusedInputsHelper, the FC12 runner and trtllm::cute_dsl_nvfp4_fc12_fused_rubin from an internal branch whose kernel takes activation_type / situ_beta / situ_linear_beta kwargs. The kernel on main (#18357) does not, so that copy would raise TypeError if it were ever reached. It never is: the #18357 registration later in the module re-registers the same op name and rebinds the same class names, so the second copy is what actually runs. Keep only the working copy.

Dev Engineer Review

The change removes the duplicate Rubin NVFP4 fused FC12 registration and retains the working implementation. This removes unsupported activation_type, situ_beta, and situ_linear_beta arguments from the retained registration path and prevents duplicate operation and class registration.

The source change removes 855 lines from tensorrt_llm/_torch/custom_ops/cute_dsl_custom_ops.py. Verify that all callers use the retained FC12 implementation and that no caller depends on the removed helper or operator. Test results are unavailable.

QA Engineer Review

No test changes.

Per-File QA Perspective

tensorrt_llm/_torch/custom_ops/cute_dsl_custom_ops.py: QA should verify Rubin NVFP4 fused FC12 execution through the retained registration, including operator discovery, supported activation behavior, and SM107 execution. QA should also confirm that no runtime path references the removed cute_dsl_nvfp4_fc12_fused_rubin implementation or _run_nvfp4_fc12_fused_rubin helper.

Description

Test Coverage

PR Checklist

Please review the following before submitting your PR:

PR description clearly explains what and why. If using CodeRabbit's summary, please make sure it makes sense.
PR Follows TRT-LLM CODING GUIDELINES to the best of your knowledge.
Test cases are provided for new code paths (see test instructions)
If PR introduces API changes, an appropriate PR label is added - either api-compatible or api-breaking. For api-breaking, include BREAKING in the PR title.
Any new dependencies have been scanned for license and vulnerabilities
CODEOWNERS updated if ownership changes
Documentation updated as needed
Update tava architecture diagram if there is a significant design change in PR.
The reviewers assigned automatically/manually are appropriate for the PR.
Please check this after reviewing the above items as appropriate for this PR.

GitHub Bot Help

To see a list of available CI bot commands, please comment /bot help.

NVIDIA#19184 synced a second copy of Fc12FusedInputsHelper, the FC12 runner and trtllm::cute_dsl_nvfp4_fc12_fused_rubin from an internal branch whose kernel takes activation_type / situ_beta / situ_linear_beta kwargs. The kernel on main (NVIDIA#18357) does not, so that copy would raise TypeError if it were ever reached. It never is: the NVIDIA#18357 registration later in the module re-registers the same op name and rebinds the same class names, so the second copy is what actually runs. Keep only the working copy. Signed-off-by: Leslie Fang <leslief@nvidia.com>

coderabbitai · 2026-09-21T06:24:53Z

Understand this PR’s impact

Explore downstream dependencies and potential security impact with Blast Radius.

View blast radius →

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Repository: NVIDIA/TensorRT-LLM/.coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 15eb7314-960e-44c8-8323-99d812bd9abb

📥 Commits

Reviewing files that changed from the base of the PR and between eca9e52 and 57dcc31.

📒 Files selected for processing (1)

tensorrt_llm/_torch/custom_ops/cute_dsl_custom_ops.py

💤 Files with no reviewable changes (1)

tensorrt_llm/_torch/custom_ops/cute_dsl_custom_ops.py

Included review availability: Your plan provides up to 12 included reviews per hour; 11 remain after this review.

Walkthrough

The PR removes Rubin NVFP4 fused FC1+FC2 CuteDSL support from the custom operations module, including kernel integration, execution helpers, runtime initialization, operator registration, and fake implementation.

Changes

Rubin NVFP4 fused FC12 removal

Layer / File(s)	Summary
Remove fused FC12 implementation and operator `tensorrt_llm/_torch/custom_ops/cute_dsl_custom_ops.py`	Removes Rubin-specific kernel imports and compatibility handling, autotuning and validation logic, memory initialization, execution helpers, the fused FC1+FC2 wrapper, and the `trtllm::cute_dsl_nvfp4_fc12_fused_rubin` operator with its fake implementation.

Priority: ⬇️ Low

Estimated code review effort: 2 (Simple) | ~15 minutes

Change: Bug fix

Suggested reviewers: bowenfu

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 0…
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Title check	✅ Passed	The title clearly identifies the removal of the duplicate Rubin fused FC12 operation registration and follows the required [None][chore] format.
Description check	✅ Passed	The description clearly explains the duplicate implementation, the unsupported keyword arguments, the registration conflict, and the intended fix. The test coverage section is not completed, but the d…

✨ Finishing Touches 💡 1

🛠️ Fix failing CI checks 💡

Commit to this branch
Create a new PR

🧪 Generate unit tests (beta)

Create a new PR

_{Comment @coderabbitai help to get the list of available commands.}

leslie-fang25 · 2026-09-21T06:30:46Z

/bot run --disable-fail-fast

github-actions · 2026-09-21T06:32:06Z

Automatically added "ci: full pre-merge approved" because this PR has satisfied the required GitHub review approvals. Unresolved review conversations and other required checks remain independent merge requirements.

leslie-fang25 · 2026-09-21T06:37:12Z

@xguannv for vis

tensorrt-cicd · 2026-09-21T06:37:56Z

PR_Github #74748 [ run ] triggered by Bot. Commit: 57dcc31 Link to invocation

tensorrt-cicd · 2026-09-21T10:02:16Z

PR_Github #74748 [ run ] completed with state SUCCESS. Commit: 57dcc31
/LLM/main/L0_MergeRequest_PR pipeline #61528 completed with status: 'FAILURE'

CI Report

⚠️ Action Required:

Please check the failed tests and fix your PR
If you cannot view the failures, ask the CI triggerer to share details
Once fixed, request an NVIDIA team member to trigger CI again

CI Agent Failure Analysis

Link to invocation

leslie-fang25 · 2026-09-21T10:06:26Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-09-21T10:12:28Z

PR_Github #74798 [ run ] triggered by Bot. Commit: 57dcc31 Link to invocation

tensorrt-cicd · 2026-09-21T11:01:41Z

PR_Github #74798 [ run ] completed with state SUCCESS. Commit: 57dcc31
/LLM/main/L0_MergeRequest_PR pipeline #61574 completed with status: 'SUCCESS'

CI Report

Link to invocation

leslie-fang25 requested a review from a team as a code owner September 21, 2026 06:23

leslie-fang25 requested review from asfiyab-nvidia and tongyuantongyu September 21, 2026 06:23

github-actions Bot assigned leslie-fang25 Sep 21, 2026

leslie-fang25 requested a review from reasonsolo September 21, 2026 06:24

reasonsolo approved these changes Sep 21, 2026

View reviewed changes

tongyuantongyu approved these changes Sep 21, 2026

View reviewed changes

trtllm-agent added the ci: full pre-merge approved label Sep 21, 2026

leslie-fang25 merged commit 2480b57 into NVIDIA:main Sep 21, 2026
19 of 21 checks passed

Navigation Menu

[None][chore] Remove the duplicate Rubin fused FC12 op registration - #19476

[None][chore] Remove the duplicate Rubin fused FC12 op registration#19476
leslie-fang25 merged 1 commit into
NVIDIA:mainfrom
leslie-fang25:leslie/remove_redundant_fc12_op_register

leslie-fang25 commented Sep 21, 2026 •

edited

Loading

Uh oh!

coderabbitai Bot commented Sep 21, 2026 •

edited

Loading

Uh oh!

leslie-fang25 commented Sep 21, 2026

Uh oh!

github-actions Bot commented Sep 21, 2026

Uh oh!

leslie-fang25 commented Sep 21, 2026

Uh oh!

tensorrt-cicd commented Sep 21, 2026

Uh oh!

tensorrt-cicd commented Sep 21, 2026

Uh oh!

leslie-fang25 commented Sep 21, 2026

Uh oh!

tensorrt-cicd commented Sep 21, 2026

Uh oh!

tensorrt-cicd commented Sep 21, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Conversation

leslie-fang25 commented Sep 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Dev Engineer Review

QA Engineer Review

Per-File QA Perspective

Description

Test Coverage

PR Checklist

GitHub Bot Help

Uh oh!

coderabbitai Bot commented Sep 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Walkthrough

Changes

Uh oh!

leslie-fang25 commented Sep 21, 2026

Uh oh!

github-actions Bot commented Sep 21, 2026

Uh oh!

leslie-fang25 commented Sep 21, 2026

Uh oh!

tensorrt-cicd commented Sep 21, 2026

Uh oh!

tensorrt-cicd commented Sep 21, 2026

Uh oh!

leslie-fang25 commented Sep 21, 2026

Uh oh!

tensorrt-cicd commented Sep 21, 2026

Uh oh!

tensorrt-cicd commented Sep 21, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

leslie-fang25 commented Sep 21, 2026 •

edited

Loading

coderabbitai Bot commented Sep 21, 2026 •

edited

Loading