[None][fix] Pass dtype to AllReduce ctor to enable MNNVL all-reduce fo… by nv-guomingz · Pull Request #15547 · NVIDIA/TensorRT-LLM

nv-guomingz · 2026-06-23T13:22:52Z

…r Qwen3.5

On NVL multi-node systems, AllReduce must be given dtype at construction so it can build the MNNVL all-reduce path (its Lamport workspace is sized by the dtype's element size). If dtype is omitted, mnnvl_allreduce is None, so the op falls back to the generic NCCL all-reduce across nodes — functionally correct but lower performance than the NVLink-fabric MNNVL path.

nv-guomingz · 2026-06-23T13:24:58Z

/bot run --disable-fail-fast

coderabbitai · 2026-06-23T13:26:58Z

📝 Walkthrough

Walkthrough

In modeling_qwen3_next.py, three AllReduce constructor calls are updated to pass dtype=config.torch_dtype explicitly. The affected constructors are in Qwen3NextSparseMoeBlock, Qwen3NextLinearDecoderLayer, and Qwen3NextFullAttentionDecoderLayer.

Changes

AllReduce dtype propagation in Qwen3Next

Layer / File(s)	Summary
Pass dtype to AllReduce in all three decoder components `tensorrt_llm/_torch/models/modeling_qwen3_next.py`	`AllReduce` construction in `Qwen3NextSparseMoeBlock` (line 128), `Qwen3NextLinearDecoderLayer` (line 350), and `Qwen3NextFullAttentionDecoderLayer` (line 515) now each supply `dtype=config.torch_dtype`; previously `dtype` was omitted from all three calls.

Estimated code review effort

🎯 1 (Trivial) | ⏱️ ~2 minutes

🚥 Pre-merge checks | ✅ 3 | ❌ 2

❌ Failed checks (1 warning, 1 inconclusive)

Check name	Status	Explanation	Resolution
Description check	⚠️ Warning	The pull request description provides sufficient context about the technical rationale for the changes but lacks required sections from the template.	Add missing template sections: provide a clear short description, explicitly list test coverage validation, and complete the PR checklist items required for this repository.
Title check	❓ Inconclusive	The title is truncated and appears incomplete ('fo…' suggests the text was cut off). While it references the main change (passing dtype to AllReduce), the truncation makes it unclear and prevents full assessment of clarity.	Complete the pull request title to fully convey the change. A complete title might be: '[fix] Pass dtype to AllReduce ctor to enable MNNVL all-reduce for Qwen3' or similar, ensuring all key information is visible.

✅ Passed checks (3 passed)

Check name	Status	Explanation
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.

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

✨ Finishing Touches

🧪 Generate unit tests (beta)

Create PR with unit tests

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

tensorrt-cicd · 2026-06-23T13:31:24Z

PR_Github #55247 [ run ] triggered by Bot. Commit: cc449b3 Link to invocation

tensorrt-cicd · 2026-06-23T21:11:18Z

PR_Github #55247 [ run ] completed with state SUCCESS. Commit: cc449b3
/LLM/main/L0_MergeRequest_PR pipeline #44204 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

nv-guomingz · 2026-06-24T02:51:28Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-06-24T02:58:07Z

PR_Github #55383 [ run ] triggered by Bot. Commit: cc449b3 Link to invocation

tensorrt-cicd · 2026-06-24T05:50:58Z

PR_Github #55383 [ run ] completed with state FAILURE. Commit: cc449b3
/LLM/main/L0_MergeRequest_PR pipeline #44328 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

nv-guomingz · 2026-06-24T11:39:49Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-06-24T11:45:57Z

PR_Github #55493 [ run ] triggered by Bot. Commit: cc449b3 Link to invocation

tensorrt-cicd · 2026-06-24T12:10:29Z

PR_Github #55493 [ run ] completed with state FAILURE. Commit: cc449b3
/LLM/main/L0_MergeRequest_PR pipeline #44418 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

nv-guomingz · 2026-06-24T16:16:33Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-06-24T16:23:26Z

PR_Github #55540 [ run ] triggered by Bot. Commit: cc449b3 Link to invocation

tensorrt-cicd · 2026-06-24T16:52:09Z

PR_Github #55540 [ run ] completed with state FAILURE. Commit: cc449b3
/LLM/main/L0_MergeRequest_PR pipeline #44461 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

tburt-nv · 2026-06-24T21:12:07Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-06-24T21:18:42Z

PR_Github #55593 [ run ] triggered by Bot. Commit: cc449b3 Link to invocation

tensorrt-cicd · 2026-06-24T22:42:25Z

PR_Github #55593 [ run ] completed with state FAILURE. Commit: cc449b3
/LLM/main/L0_MergeRequest_PR pipeline #44510 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

nv-guomingz · 2026-06-25T01:47:26Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-06-25T01:53:03Z

PR_Github #55649 [ run ] triggered by Bot. Commit: cc449b3 Link to invocation

tensorrt-cicd · 2026-06-25T02:21:36Z

PR_Github #55649 [ run ] completed with state FAILURE. Commit: cc449b3
/LLM/main/L0_MergeRequest_PR pipeline #44560 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

nv-guomingz · 2026-06-25T06:39:49Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-06-25T06:45:27Z

PR_Github #55726 [ run ] triggered by Bot. Commit: cc449b3 Link to invocation

tensorrt-cicd · 2026-06-25T07:52:29Z

PR_Github #55726 [ run ] completed with state FAILURE. Commit: cc449b3
/LLM/main/L0_MergeRequest_PR pipeline #44628 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

nv-guomingz · 2026-06-25T13:05:45Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-06-25T13:11:25Z

PR_Github #55784 [ run ] triggered by Bot. Commit: cc449b3 Link to invocation

tensorrt-cicd · 2026-06-25T13:46:35Z

PR_Github #55784 [ run ] completed with state FAILURE. Commit: cc449b3
/LLM/main/L0_MergeRequest_PR pipeline #44680 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

tensorrt-cicd · 2026-06-30T13:52:10Z

PR_Github #56614 [ run ] triggered by Bot. Commit: 56f30f1 Link to invocation

tensorrt-cicd · 2026-06-30T18:26:34Z

PR_Github #56614 [ run ] completed with state SUCCESS. Commit: 56f30f1
/LLM/main/L0_MergeRequest_PR pipeline #45441 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

nv-guomingz · 2026-07-01T03:24:44Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-07-01T03:31:21Z

PR_Github #56812 [ run ] triggered by Bot. Commit: 56f30f1 Link to invocation

tensorrt-cicd · 2026-07-01T04:19:32Z

PR_Github #56812 [ run ] completed with state SUCCESS. Commit: 56f30f1
/LLM/main/L0_MergeRequest_PR pipeline #45626 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

nv-guomingz · 2026-07-01T05:28:18Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-07-01T05:34:59Z

PR_Github #56845 [ run ] triggered by Bot. Commit: 7ccb7bf Link to invocation

tensorrt-cicd · 2026-07-01T12:06:07Z

PR_Github #56845 [ run ] completed with state SUCCESS. Commit: 7ccb7bf
/LLM/main/L0_MergeRequest_PR pipeline #45657 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

nv-guomingz · 2026-07-01T14:24:00Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-07-01T14:31:19Z

PR_Github #56949 [ run ] triggered by Bot. Commit: 7ccb7bf Link to invocation

tensorrt-cicd · 2026-07-01T14:42:52Z

PR_Github #56949 [ run ] completed with state FAILURE. Commit: 7ccb7bf
/LLM/main/L0_MergeRequest_PR pipeline #45754 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

nv-guomingz · 2026-07-02T01:01:10Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-07-02T01:07:39Z

PR_Github #57050 [ run ] triggered by Bot. Commit: 7ccb7bf Link to invocation

tensorrt-cicd · 2026-07-02T03:49:47Z

PR_Github #57050 [ run ] completed with state FAILURE. Commit: 7ccb7bf
/LLM/main/L0_MergeRequest_PR pipeline #45845 completed with status: 'ABORTED'

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

Link to invocation

nv-guomingz · 2026-07-02T05:53:18Z

/bot run --disable-fail-fast

…r Qwen3.5 Signed-off-by: nv-guomingz <137257613+nv-guomingz@users.noreply.github.com>

nv-guomingz · 2026-07-03T10:52:57Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-07-03T10:58:43Z

PR_Github #57458 [ run ] triggered by Bot. Commit: 207491e Link to invocation

tensorrt-cicd · 2026-07-03T19:54:17Z

PR_Github #57458 [ run ] completed with state SUCCESS. Commit: 207491e
/LLM/main/L0_MergeRequest_PR pipeline #46196 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

nv-guomingz · 2026-07-04T04:02:59Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-07-04T04:08:10Z

PR_Github #57532 [ run ] triggered by Bot. Commit: 207491e Link to invocation

tensorrt-cicd · 2026-07-04T08:03:28Z

PR_Github #57532 [ run ] completed with state SUCCESS. Commit: 207491e
/LLM/main/L0_MergeRequest_PR pipeline #46262 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

nv-guomingz · 2026-07-06T01:03:11Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-07-06T01:09:24Z

PR_Github #57643 [ run ] triggered by Bot. Commit: 207491e Link to invocation

tensorrt-cicd · 2026-07-06T02:01:51Z

PR_Github #57643 [ run ] completed with state SUCCESS. Commit: 207491e
/LLM/main/L0_MergeRequest_PR pipeline #46366 completed with status: 'SUCCESS'

CI Report

Link to invocation

…o… (NVIDIA#15547) Signed-off-by: nv-guomingz <137257613+nv-guomingz@users.noreply.github.com>

nv-guomingz requested a review from a team as a code owner June 23, 2026 13:22

nv-guomingz requested a review from Wanli-Jiang June 23, 2026 13:22

github-actions Bot assigned nv-guomingz Jun 23, 2026

nv-guomingz changed the title ~~None][fix] Pass dtype to AllReduce ctor to enable MNNVL all-reduce fo…~~ [None][fix] Pass dtype to AllReduce ctor to enable MNNVL all-reduce fo… Jun 23, 2026

Wanli-Jiang approved these changes Jun 25, 2026

View reviewed changes

nv-guomingz force-pushed the user/guomingz/fix-mnnvl-qwen3.5 branch from 56f30f1 to 7ccb7bf Compare July 1, 2026 05:28

None][fix] Pass dtype to AllReduce ctor to enable MNNVL all-reduce fo…

207491e

…r Qwen3.5 Signed-off-by: nv-guomingz <137257613+nv-guomingz@users.noreply.github.com>

nv-guomingz force-pushed the user/guomingz/fix-mnnvl-qwen3.5 branch from 7ccb7bf to 207491e Compare July 3, 2026 10:52

nv-guomingz disabled auto-merge July 4, 2026 04:02

nv-guomingz merged commit b420e72 into NVIDIA:main Jul 6, 2026
7 checks passed

BrianLi23 pushed a commit to BrianLi23/TensorRT-LLM that referenced this pull request Jul 9, 2026

[None][fix] Pass dtype to AllReduce ctor to enable MNNVL all-reduce f…

835ccf1

…o… (NVIDIA#15547) Signed-off-by: nv-guomingz <137257613+nv-guomingz@users.noreply.github.com>

Uh oh!

Conversation

nv-guomingz commented Jun 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

nv-guomingz commented Jun 23, 2026

Uh oh!

coderabbitai Bot commented Jun 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Walkthrough

Changes

Estimated code review effort

❌ Failed checks (1 warning, 1 inconclusive)

Uh oh!

tensorrt-cicd commented Jun 23, 2026

Uh oh!

tensorrt-cicd commented Jun 23, 2026

Uh oh!

nv-guomingz commented Jun 24, 2026

Uh oh!

tensorrt-cicd commented Jun 24, 2026

Uh oh!

tensorrt-cicd commented Jun 24, 2026

Uh oh!

nv-guomingz commented Jun 24, 2026

Uh oh!

tensorrt-cicd commented Jun 24, 2026

Uh oh!

tensorrt-cicd commented Jun 24, 2026

Uh oh!

nv-guomingz commented Jun 24, 2026

Uh oh!

tensorrt-cicd commented Jun 24, 2026

Uh oh!

tensorrt-cicd commented Jun 24, 2026

Uh oh!

tburt-nv commented Jun 24, 2026

Uh oh!

tensorrt-cicd commented Jun 24, 2026

Uh oh!

tensorrt-cicd commented Jun 24, 2026

Uh oh!

nv-guomingz commented Jun 25, 2026

Uh oh!

tensorrt-cicd commented Jun 25, 2026

Uh oh!

tensorrt-cicd commented Jun 25, 2026

Uh oh!

nv-guomingz commented Jun 25, 2026

Uh oh!

tensorrt-cicd commented Jun 25, 2026

Uh oh!

tensorrt-cicd commented Jun 25, 2026

Uh oh!

nv-guomingz commented Jun 25, 2026

Uh oh!

tensorrt-cicd commented Jun 25, 2026

Uh oh!

tensorrt-cicd commented Jun 25, 2026

Uh oh!

tensorrt-cicd commented Jun 30, 2026

Uh oh!

tensorrt-cicd commented Jun 30, 2026

Uh oh!

nv-guomingz commented Jul 1, 2026

Uh oh!

tensorrt-cicd commented Jul 1, 2026

Uh oh!

tensorrt-cicd commented Jul 1, 2026

Uh oh!

nv-guomingz commented Jul 1, 2026

Uh oh!

tensorrt-cicd commented Jul 1, 2026

Uh oh!

tensorrt-cicd commented Jul 1, 2026

Uh oh!

nv-guomingz commented Jul 1, 2026

Uh oh!

tensorrt-cicd commented Jul 1, 2026

Uh oh!

nv-guomingz commented Jun 23, 2026 •

edited

Loading

coderabbitai Bot commented Jun 23, 2026 •

edited

Loading