[https://nvbugs/5848377][fix] fix deepeplowlatency with trtllm moe backend running fp8 DS_R1 by leslie-fang25 · Pull Request #11266 · NVIDIA/TensorRT-LLM

leslie-fang25 · 2026-02-04T07:38:06Z

Summary by CodeRabbit

Bug Fixes
- Fixed data type consistency in distributed expert selection operations
- Improved token count handling across multiple ranks during expert aggregation
- Enhanced dynamic expert routing to respect input token specifications

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)
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

/bot [-h] ['run', 'kill', 'skip', 'reuse-pipeline'] ...

Provide a user friendly way for developers to interact with a Jenkins server.

Run /bot [-h|--help] to print this help message.

See details below for each supported subcommand.

Details

run [--reuse-test (optional)pipeline-id --disable-fail-fast --skip-test --stage-list "A10-PyTorch-1, xxx" --gpu-type "A30, H100_PCIe" --test-backend "pytorch, cpp" --add-multi-gpu-test --only-multi-gpu-test --disable-multi-gpu-test --post-merge --extra-stage "H100_PCIe-TensorRT-Post-Merge-1, xxx" --detailed-log --debug(experimental)]

Launch build/test pipelines. All previously running jobs will be killed.

--reuse-test (optional)pipeline-id (OPTIONAL) : Allow the new pipeline to reuse build artifacts and skip successful test stages from a specified pipeline or the last pipeline if no pipeline-id is indicated. If the Git commit ID has changed, this option will be always ignored. The DEFAULT behavior of the bot is to reuse build artifacts and successful test results from the last pipeline.

--disable-reuse-test (OPTIONAL) : Explicitly prevent the pipeline from reusing build artifacts and skipping successful test stages from a previous pipeline. Ensure that all builds and tests are run regardless of previous successes.

--disable-fail-fast (OPTIONAL) : Disable fail fast on build/tests/infra failures.

--skip-test (OPTIONAL) : Skip all test stages, but still run build stages, package stages and sanity check stages. Note: Does NOT update GitHub check status.

--stage-list "A10-PyTorch-1, xxx" (OPTIONAL) : Only run the specified test stages. Examples: "A10-PyTorch-1, xxx". Note: Does NOT update GitHub check status.

--gpu-type "A30, H100_PCIe" (OPTIONAL) : Only run the test stages on the specified GPU types. Examples: "A30, H100_PCIe". Note: Does NOT update GitHub check status.

--test-backend "pytorch, cpp" (OPTIONAL) : Skip test stages which don't match the specified backends. Only support [pytorch, cpp, tensorrt, triton]. Examples: "pytorch, cpp" (does not run test stages with tensorrt or triton backend). Note: Does NOT update GitHub pipeline status.

--only-multi-gpu-test (OPTIONAL) : Only run the multi-GPU tests. Note: Does NOT update GitHub check status.

--disable-multi-gpu-test (OPTIONAL) : Disable the multi-GPU tests. Note: Does NOT update GitHub check status.

--add-multi-gpu-test (OPTIONAL) : Force run the multi-GPU tests in addition to running L0 pre-merge pipeline.

--post-merge (OPTIONAL) : Run the L0 post-merge pipeline instead of the ordinary L0 pre-merge pipeline.

--extra-stage "H100_PCIe-TensorRT-Post-Merge-1, xxx" (OPTIONAL) : Run the ordinary L0 pre-merge pipeline and specified test stages. Examples: --extra-stage "H100_PCIe-TensorRT-Post-Merge-1, xxx".

--detailed-log (OPTIONAL) : Enable flushing out all logs to the Jenkins console. This will significantly increase the log volume and may slow down the job.

--debug (OPTIONAL) : Experimental feature. Enable access to the CI container for debugging purpose. Note: Specify exactly one stage in the stage-list parameter to access the appropriate container environment. Note: Does NOT update GitHub check status.

For guidance on mapping tests to stage names, see docs/source/reference/ci-overview.md
and the scripts/test_to_stage_mapping.py helper.

kill

kill

Kill all running builds associated with pull request.

skip

skip --comment COMMENT

Skip testing for latest commit on pull request. --comment "Reason for skipping build/test" is required. IMPORTANT NOTE: This is dangerous since lack of user care and validation can cause top of tree to break.

reuse-pipeline

reuse-pipeline

Reuse a previous pipeline to validate current commit. This action will also kill all currently running builds associated with the pull request. IMPORTANT NOTE: This is dangerous since lack of user care and validation can cause top of tree to break.

…ckend running fp8 DS_R1 Signed-off-by: leslie-fang25 <leslief@nvidia.com>

coderabbitai · 2026-02-04T07:41:55Z

📝 Walkthrough

Walkthrough

These changes enhance the MOE (Mixture of Experts) module by adding precision handling for weights during low-latency communication, passing maximum token counts to combine operations, and dynamically synchronizing top-k selection based on token-specific expert routing data.

Changes

Cohort / File(s)	Summary
MOE Communication Layer `tensorrt_llm/_torch/modules/fused_moe/communication/deep_ep_low_latency.py`	Added type guard in `DeepEPLowLatency.combine` to cast `deep_ep_topk_weights` to float32 before proceeding with low-precision combine paths, ensuring data type consistency.
MOE Configuration & Routing `tensorrt_llm/_torch/modules/fused_moe/configurable_moe.py`, `tensorrt_llm/_torch/modules/fused_moe/fused_moe_trtllm_gen.py`	Enhanced `configurable_moe._forward_chunk_impl` to compute and pass `all_rank_max_num_tokens` to the combine operation. Updated `fused_moe_trtllm_gen.run_moe` to dynamically synchronize `top_k` with actual top-k from `token_selected_experts` when available.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

🚥 Pre-merge checks | ✅ 2 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Description check	⚠️ Warning	The PR description is mostly empty with only template boilerplate present. Critical sections like Description and Test Coverage are not filled out, leaving no explanation of the issue, solution, or test coverage.	Fill in the Description section explaining the bug and the fix, and the Test Coverage section listing relevant tests that validate the changes.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly identifies the bug fix (NVBugs 5848377) and describes the specific issue being fixed: deepeplowlatency with trtllm MOE backend running FP8 DS_R1.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing touches

📝 Generate docstrings

🧪 Generate unit tests (beta)

Create PR with unit tests
Post copyable unit tests in a comment

Important

Action Needed: IP Allowlist Update

If your organization protects your Git platform with IP whitelisting, please add the new CodeRabbit IP address to your allowlist:

✨ 136.113.208.247/32 (new)
34.170.211.100/32
35.222.179.152/32

Reviews will stop working after February 8, 2026 if the new IP is not added to your allowlist.

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

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

Signed-off-by: leslie-fang25 <leslief@nvidia.com>

leslie-fang25 · 2026-02-04T07:55:44Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-02-04T08:02:37Z

PR_Github #34719 [ run ] triggered by Bot. Commit: 3f0b89f

tensorrt-cicd · 2026-02-04T17:30:39Z

PR_Github #34830 [ run ] triggered by Bot. Commit: 3f0b89f

tensorrt-cicd · 2026-02-05T00:21:14Z

PR_Github #34830 [ run ] completed with state SUCCESS. Commit: 3f0b89f
/LLM/main/L0_MergeRequest_PR pipeline #26869 completed with status: 'FAILURE'

⚠️ 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

leslie-fang25 · 2026-02-05T02:11:56Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-02-05T02:18:38Z

PR_Github #34877 [ run ] triggered by Bot. Commit: 3f0b89f

tensorrt_llm/_torch/modules/fused_moe/fused_moe_trtllm_gen.py

tensorrt-cicd · 2026-02-05T06:48:40Z

PR_Github #34877 [ run ] completed with state SUCCESS. Commit: 3f0b89f
/LLM/main/L0_MergeRequest_PR pipeline #26905 completed with status: 'FAILURE'

⚠️ 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

leslie-fang25 · 2026-02-05T06:59:48Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-02-05T07:05:54Z

PR_Github #34931 [ run ] triggered by Bot. Commit: 3f0b89f

tensorrt-cicd · 2026-02-05T08:49:43Z

PR_Github #34931 [ run ] completed with state SUCCESS. Commit: 3f0b89f
/LLM/main/L0_MergeRequest_PR pipeline #26948 completed with status: 'FAILURE'

⚠️ 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

leslie-fang25 · 2026-02-05T09:08:56Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-02-05T09:15:58Z

PR_Github #34951 [ run ] triggered by Bot. Commit: 3f0b89f

tensorrt-cicd · 2026-02-05T11:36:12Z

PR_Github #34951 [ run ] completed with state SUCCESS. Commit: 3f0b89f
/LLM/main/L0_MergeRequest_PR pipeline #26965 completed with status: 'SUCCESS'

Co-authored-by: Tailing Yuan <yuantailing@gmail.com> Signed-off-by: Leslie Fang <leslief@nvidia.com>

leslie-fang25 · 2026-02-06T02:11:54Z

Hi @yuantailing @ChristinaZ, the comments have been addressed. Please kindly take a look again.

leslie-fang25 · 2026-02-06T02:12:24Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-02-06T02:18:35Z

PR_Github #35042 [ run ] triggered by Bot. Commit: 182a127

tensorrt-cicd · 2026-02-06T05:55:20Z

PR_Github #35042 [ run ] completed with state SUCCESS. Commit: 182a127
/LLM/main/L0_MergeRequest_PR pipeline #27044 completed with status: 'FAILURE'

⚠️ 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

tensorrt-cicd · 2026-02-09T03:53:14Z

PR_Github #35255 [ run ] completed with state SUCCESS. Commit: b421a36
/LLM/main/L0_MergeRequest_PR pipeline #27220 completed with status: 'SUCCESS'

Signed-off-by: leslie-fang25 <leslief@nvidia.com>

leslie-fang25 · 2026-02-09T05:08:37Z

/bot run --disable-fail-fast

Signed-off-by: leslie-fang25 <leslief@nvidia.com>

tensorrt-cicd · 2026-02-09T05:14:34Z

PR_Github #35275 [ run ] triggered by Bot. Commit: 8a59f96

ChristinaZ

LGTM, thanks.

leslie-fang25 · 2026-02-09T07:39:29Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-02-09T07:45:49Z

PR_Github #35300 [ run ] triggered by Bot. Commit: 8a59f96

leslie-fang25 · 2026-02-09T09:05:08Z

/bot run --disable-fail-fast --disable-reuse-test

tensorrt-cicd · 2026-02-09T09:11:07Z

PR_Github #35311 [ run ] triggered by Bot. Commit: 8a59f96

tensorrt-cicd · 2026-02-09T13:16:28Z

PR_Github #35311 [ run ] completed with state SUCCESS. Commit: 8a59f96
/LLM/main/L0_MergeRequest_PR pipeline #27269 completed with status: 'FAILURE'

⚠️ 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

xxi-nv · 2026-02-09T13:55:17Z

/bot run --disable-fail-fast

xxi-nv · 2026-02-09T13:56:15Z

/bot kill

tensorrt-cicd · 2026-02-09T14:01:09Z

PR_Github #35339 [ ] completed with state FAILURE. Commit: ``

tensorrt-cicd · 2026-02-09T14:02:02Z

PR_Github #35340 [ kill ] triggered by Bot. Commit: 8a59f96

tensorrt-cicd · 2026-02-09T14:02:04Z

PR_Github #35340 [ kill ] completed with state SUCCESS. Commit: 8a59f96
Successfully killed previous jobs for commit 8a59f96

leslie-fang25 · 2026-02-09T23:40:19Z

/bot run --disable-fail-fast

leslie-fang25 · 2026-02-10T01:35:19Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-02-10T01:41:33Z

PR_Github #35390 [ run ] triggered by Bot. Commit: 8a59f96

…ckend running fp8 DS_R1 Cherry-picked from PR NVIDIA#11266. Signed-off-by: Xi Xu <xxi@nvidia.com>

tensorrt-cicd · 2026-02-10T05:24:26Z

PR_Github #35390 [ run ] completed with state SUCCESS. Commit: 8a59f96
/LLM/main/L0_MergeRequest_PR pipeline #27336 completed with status: 'FAILURE'

⚠️ 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

leslie-fang25 · 2026-02-10T05:26:19Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-02-10T05:32:46Z

PR_Github #35421 [ run ] triggered by Bot. Commit: 8a59f96

tensorrt-cicd · 2026-02-10T07:27:46Z

PR_Github #35421 [ run ] completed with state SUCCESS. Commit: 8a59f96
/LLM/main/L0_MergeRequest_PR pipeline #27361 completed with status: 'FAILURE'

⚠️ 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

leslie-fang25 · 2026-02-10T07:32:17Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-02-10T07:37:56Z

PR_Github #35459 [ run ] triggered by Bot. Commit: 8a59f96

tensorrt-cicd · 2026-02-10T10:51:25Z

PR_Github #35459 [ run ] completed with state SUCCESS. Commit: 8a59f96
/LLM/main/L0_MergeRequest_PR pipeline #27394 completed with status: 'SUCCESS'

[https://nvbugs/5848377][fix] fix deepeplowlatency with trtllm gen ba…

8337494

…ckend running fp8 DS_R1 Signed-off-by: leslie-fang25 <leslief@nvidia.com>

leslie-fang25 requested a review from a team as a code owner February 4, 2026 07:38

leslie-fang25 requested a review from xxi-nv February 4, 2026 07:38

format

3f0b89f

Signed-off-by: leslie-fang25 <leslief@nvidia.com>

leslie-fang25 requested review from ChristinaZ, rosenrodt and yuantailing February 5, 2026 02:12

xxi-nv approved these changes Feb 5, 2026

View reviewed changes

yuantailing reviewed Feb 5, 2026

View reviewed changes

tensorrt_llm/_torch/modules/fused_moe/fused_moe_trtllm_gen.py Outdated Show resolved Hide resolved

leslie-fang25 and others added 2 commits February 6, 2026 10:09

Update tensorrt_llm/_torch/modules/fused_moe/fused_moe_trtllm_gen.py

8fd919e

Co-authored-by: Tailing Yuan <yuantailing@gmail.com> Signed-off-by: Leslie Fang <leslief@nvidia.com>

Merge branch 'main' into leslie/fix_5848377

182a127

leslie-fang25 requested a review from yuantailing February 6, 2026 02:10

update the check conditions

1c366da

Signed-off-by: leslie-fang25 <leslief@nvidia.com>

leslie-fang25 requested a review from yuantailing February 9, 2026 05:08

yuantailing approved these changes Feb 9, 2026

View reviewed changes

update the dim

8a59f96

Signed-off-by: leslie-fang25 <leslief@nvidia.com>

ChristinaZ approved these changes Feb 9, 2026

View reviewed changes

xxi-nv added a commit to xxi-nv/TensorRT-LLM that referenced this pull request Feb 10, 2026

[https://nvbugs/5848377][fix] fix deepeplowlatency with trtllm moe ba…

429e0d0

…ckend running fp8 DS_R1 Cherry-picked from PR NVIDIA#11266. Signed-off-by: Xi Xu <xxi@nvidia.com>

leslie-fang25 merged commit d6e4954 into NVIDIA:main Feb 10, 2026
5 checks passed

Comments

Conversation

leslie-fang25 commented Feb 4, 2026 • edited by coderabbitai bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary by CodeRabbit

Description

Test Coverage

PR Checklist

GitHub Bot Help

kill

skip

reuse-pipeline

Uh oh!

coderabbitai bot commented Feb 4, 2026

Walkthrough

Changes

Estimated code review effort

Uh oh!

leslie-fang25 commented Feb 4, 2026

Uh oh!

tensorrt-cicd commented Feb 4, 2026

Uh oh!

tensorrt-cicd commented Feb 4, 2026

Uh oh!

tensorrt-cicd commented Feb 5, 2026

Uh oh!

leslie-fang25 commented Feb 5, 2026

Uh oh!

tensorrt-cicd commented Feb 5, 2026

Uh oh!

Uh oh!

tensorrt-cicd commented Feb 5, 2026

Uh oh!

leslie-fang25 commented Feb 5, 2026

Uh oh!

tensorrt-cicd commented Feb 5, 2026

Uh oh!

tensorrt-cicd commented Feb 5, 2026

Uh oh!

leslie-fang25 commented Feb 5, 2026

Uh oh!

tensorrt-cicd commented Feb 5, 2026

Uh oh!

tensorrt-cicd commented Feb 5, 2026

Uh oh!

leslie-fang25 commented Feb 6, 2026

Uh oh!

leslie-fang25 commented Feb 6, 2026

Uh oh!

tensorrt-cicd commented Feb 6, 2026

Uh oh!

tensorrt-cicd commented Feb 6, 2026

Uh oh!

tensorrt-cicd commented Feb 9, 2026

Uh oh!

leslie-fang25 commented Feb 9, 2026

Uh oh!

tensorrt-cicd commented Feb 9, 2026

Uh oh!

ChristinaZ left a comment

Choose a reason for hiding this comment

Uh oh!

leslie-fang25 commented Feb 9, 2026

Uh oh!

tensorrt-cicd commented Feb 9, 2026

Uh oh!

leslie-fang25 commented Feb 9, 2026

Uh oh!

tensorrt-cicd commented Feb 9, 2026

Uh oh!

tensorrt-cicd commented Feb 9, 2026

Uh oh!

xxi-nv commented Feb 9, 2026

Uh oh!

xxi-nv commented Feb 9, 2026

Uh oh!

tensorrt-cicd commented Feb 9, 2026

Uh oh!

tensorrt-cicd commented Feb 9, 2026

Uh oh!

leslie-fang25 commented Feb 4, 2026 •

edited by coderabbitai bot

Loading