[TRTLLM-14609][chore] Remove legacy MoE path in CuteDslFusedMoE by xxi-nv · Pull Request #16863 · NVIDIA/TensorRT-LLM

xxi-nv · 2026-07-25T01:41:00Z

Description

For the sake of MoE code health, this PR removes the legacy (pre-ConfigurableMoE) forward path from CuteDslFusedMoE.

ConfigurableMoE has been the default MoE path for months (ENABLE_CONFIGURABLE_MOE defaults to "1"), so the backend-internal forward pipeline and the legacy all-to-all method selection are no longer reachable in the default flow.

What is removed (`tensorrt_llm/_torch/modules/fused_moe/fused_moe_cute_dsl.py`)

select_alltoall_method_type()
forward_chunk()

Net effect: 43 deletions in a single file.

What is deliberately preserved

The live contract called by ConfigurableMoE / MoEScheduler is untouched: run_moe() and supports_moe_output_in_alltoall_workspace().

Test Coverage

Verified on OCI GB200 (1 node, 4 GPUs) with the container pinned by jenkins/current_image_tags.properties, running tests/unittest/_torch/modules/moe/test_moe_module.py filtered to this backend (excluding the MEGAMOE_CUTEDSL variant):

Scope	Result
`test_configurable_moe_single_gpu`	9 passed
`test_configurable_moe_multi_gpu`	20 passed
`test_configurable_moe_multi_gpu_eplb`	2 passed

31 passed, 0 failed.

PR Checklist

PR title follows [JIRA/NVBUG][type] description
Commit is signed off (DCO)
Tests pass for the affected backend

Dev Engineer Review

Removed the legacy pre-ConfigurableMoE forward path from CuteDslFusedMoE, including select_alltoall_method_type() and forward_chunk().
Removed now-unused allgather and AlltoallMethodType imports.
The active ConfigurableMoE/MoEScheduler contract remains unchanged, including run_moe() and supports_moe_output_in_alltoall_workspace().
No config or test-list changes.
OCI GB200 validation passed all 31 affected-backend tests.

QA Engineer Review

No test changes.

ConfigurableMoE has been the default MoE path for months (ENABLE_CONFIGURABLE_MOE defaults to 1) and drives compute through backend.run_moe while the wrapper owns communication and chunk scheduling. The backend-level legacy MoE forward path is dead code on that path. Remove it; run_moe and the live ConfigurableMoE contract (quantize_input, get_workspaces, supports_moe_output_in_alltoall_workspace, forward_fake, etc.) are unchanged. Signed-off-by: xxi <xxi@nvidia.com>

xxi-nv · 2026-07-26T23:58:55Z

/bot run

coderabbitai · 2026-07-26T23:59:51Z

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 9a940c7a-3e13-446b-af6a-e38deb0ef20b

📥 Commits

Reviewing files that changed from the base of the PR and between 4b7d719 and 347661a.

📒 Files selected for processing (1)

tensorrt_llm/_torch/modules/fused_moe/fused_moe_cute_dsl.py

💤 Files with no reviewable changes (1)

tensorrt_llm/_torch/modules/fused_moe/fused_moe_cute_dsl.py

Walkthrough

The Cute DSL fused MoE module removes unused imports, the select_alltoall_method_type() override, and the legacy forward_chunk() implementation.

Changes

Cute DSL MoE cleanup

Layer / File(s)	Summary
Remove obsolete Cute DSL paths `tensorrt_llm/_torch/modules/fused_moe/fused_moe_cute_dsl.py`	Removes unused distributed and interface imports, the all-to-all method selection override, and the legacy routing and chunk-forwarding implementation.

Estimated code review effort: 2 (Simple) | ~10 minutes

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title follows the required ticket/type format and clearly states the main change.
Description check	✅ Passed	The description includes the required sections and provides a clear summary, tests, and checklist.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
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.

✨ Finishing Touches

🧪 Generate unit tests (beta)

Create PR with unit tests

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

tensorrt-cicd · 2026-07-27T00:06:21Z

PR_Github #61759 [ run ] triggered by Bot. Commit: 347661a Link to invocation

tensorrt-cicd · 2026-07-27T03:25:37Z

PR_Github #61759 [ run ] completed with state FAILURE. Commit: 347661a
/LLM/main/L0_MergeRequest_PR pipeline #49961 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

xxi-nv · 2026-07-27T05:32:41Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-07-27T05:39:45Z

PR_Github #61821 [ run ] triggered by Bot. Commit: 347661a Link to invocation

tensorrt-cicd · 2026-07-27T09:04:39Z

PR_Github #61821 [ run ] completed with state SUCCESS. Commit: 347661a
/LLM/main/L0_MergeRequest_PR pipeline #50014 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

xxi-nv · 2026-07-27T09:09:02Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-07-27T09:15:39Z

PR_Github #61876 [ run ] triggered by Bot. Commit: 347661a Link to invocation

tensorrt-cicd · 2026-07-27T11:05:11Z

PR_Github #61876 [ run ] completed with state SUCCESS. Commit: 347661a
/LLM/main/L0_MergeRequest_PR pipeline #50067 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

xxi-nv · 2026-07-27T11:09:45Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-07-27T11:16:08Z

PR_Github #61896 [ run ] triggered by Bot. Commit: 347661a Link to invocation

xxi-nv · 2026-07-27T12:00:09Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-07-27T12:05:59Z

PR_Github #61906 [ run ] triggered by Bot. Commit: 7a8d83a Link to invocation

tensorrt-cicd · 2026-07-27T12:09:42Z

PR_Github #61896 [ run ] completed with state ABORTED. Commit: 347661a

Link to invocation

tensorrt-cicd · 2026-07-27T17:12:31Z

PR_Github #61906 [ run ] completed with state FAILURE. Commit: 7a8d83a
/LLM/main/L0_MergeRequest_PR pipeline #50097 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

xxi-nv · 2026-07-27T17:17:05Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-07-27T17:24:53Z

PR_Github #61952 [ run ] triggered by Bot. Commit: 7a8d83a Link to invocation

tensorrt-cicd · 2026-07-27T22:01:25Z

PR_Github #61952 [ run ] completed with state FAILURE. Commit: 7a8d83a
/LLM/main/L0_MergeRequest_PR pipeline #50146 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

xxi-nv · 2026-07-27T22:03:52Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-07-27T22:16:31Z

PR_Github #62005 [ run ] triggered by Bot. Commit: 7a8d83a Link to invocation

tensorrt-cicd · 2026-07-28T01:24:12Z

PR_Github #62005 [ run ] completed with state SUCCESS. Commit: 7a8d83a
/LLM/main/L0_MergeRequest_PR pipeline #50195 completed with status: 'SUCCESS'

CI Report

Link to invocation

github-actions Bot assigned xxi-nv Jul 25, 2026

xxi-nv marked this pull request as ready for review July 26, 2026 23:58

xxi-nv requested a review from a team as a code owner July 26, 2026 23:58

xxi-nv requested review from leslie-fang25 and rosong11 July 26, 2026 23:58

Merge branch 'main' into xxi/moe-rm-legacy-cutedsl

7a8d83a

leslie-fang25 approved these changes Jul 28, 2026

View reviewed changes

xxi-nv merged commit ede2cad into NVIDIA:main Jul 28, 2026
7 checks passed

xxi-nv deleted the xxi/moe-rm-legacy-cutedsl branch July 28, 2026 05:12

coderabbitai Bot mentioned this pull request Jul 29, 2026

[TRTLLM-14609][chore] Remove ENABLE_CONFIGURABLE_MOE escape hatch and remaining MoE legacy relics #17009

Merged

4 tasks

Uh oh!

Conversation

xxi-nv commented Jul 25, 2026 • edited by coderabbitai Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

What is removed (tensorrt_llm/_torch/modules/fused_moe/fused_moe_cute_dsl.py)

What is deliberately preserved

Test Coverage

PR Checklist

Uh oh!

xxi-nv commented Jul 26, 2026

Uh oh!

coderabbitai Bot commented Jul 26, 2026

Walkthrough

Changes

Uh oh!

tensorrt-cicd commented Jul 27, 2026

Uh oh!

tensorrt-cicd commented Jul 27, 2026

Uh oh!

xxi-nv commented Jul 27, 2026

Uh oh!

tensorrt-cicd commented Jul 27, 2026

Uh oh!

tensorrt-cicd commented Jul 27, 2026

Uh oh!

xxi-nv commented Jul 27, 2026

Uh oh!

tensorrt-cicd commented Jul 27, 2026

Uh oh!

tensorrt-cicd commented Jul 27, 2026

Uh oh!

xxi-nv commented Jul 27, 2026

Uh oh!

tensorrt-cicd commented Jul 27, 2026

Uh oh!

xxi-nv commented Jul 27, 2026

Uh oh!

tensorrt-cicd commented Jul 27, 2026

Uh oh!

tensorrt-cicd commented Jul 27, 2026

Uh oh!

tensorrt-cicd commented Jul 27, 2026

Uh oh!

xxi-nv commented Jul 27, 2026

Uh oh!

tensorrt-cicd commented Jul 27, 2026

Uh oh!

tensorrt-cicd commented Jul 27, 2026

Uh oh!

xxi-nv commented Jul 27, 2026

Uh oh!

tensorrt-cicd commented Jul 27, 2026

Uh oh!

tensorrt-cicd commented Jul 28, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

xxi-nv commented Jul 25, 2026 •

edited by coderabbitai Bot

Loading

What is removed (`tensorrt_llm/_torch/modules/fused_moe/fused_moe_cute_dsl.py`)