[None][fix] Add GlmMoeDsaForCausalLM to EPLB supported model list by qiaoxj07 · Pull Request #12607 · NVIDIA/TensorRT-LLM

qiaoxj07 · 2026-03-31T03:55:26Z

Summary

GLM-5 (GlmMoeDsaForCausalLM) uses the DeepSeekV3 MoE architecture but was missing from moe_model_arch_list in moe_load_balancer.py.
When moe_config.load_balancer.num_slots is set, maybe_create_moe_load_balancer() skips setup() because the arch is not in the list, but interface.py still accesses num_local_slots (which requires setup()), causing ValueError: Cannot calculate num_local_slots.
Fix: add GlmMoeDsaForCausalLM to the supported architecture list.

Test plan

Verify GLM-5 with moe_config.load_balancer.num_slots=256 no longer crashes during model init
Verify existing EPLB models (DeepSeek V3, Qwen3 MoE) are unaffected

🤖 Generated with Claude Code

Summary by CodeRabbit

New Features
- Added support for GlmMoeDsaForCausalLM model architecture in the MOE load balancer.

GlmMoeDsaForCausalLM (GLM-5) uses the DeepSeekV3 MoE architecture but was missing from moe_model_arch_list. This caused setup() to never be called on the load balancer config, so accessing num_local_slots during model init raised ValueError. Adding it to the list enables EPLB for GLM-5. Signed-off-by: Xianjie <5410381+qiaoxj07@users.noreply.github.com>

qiaoxj07 · 2026-03-31T03:56:40Z

/bot run --disable-fail-fast

coderabbitai · 2026-03-31T03:57:57Z

📝 Walkthrough

Walkthrough

This change extends the moe_model_arch_list to include support for a new model architecture, 'GlmMoeDsaForCausalLM', in the MOE load balancer module. No control flow or logic modifications were made.

Changes

Cohort / File(s)	Summary
MOE Model Architecture Support `tensorrt_llm/_torch/modules/fused_moe/moe_load_balancer.py`	Added `'GlmMoeDsaForCausalLM'` to the supported model architectures list in `moe_model_arch_list`.

Estimated code review effort

🎯 1 (Trivial) | ⏱️ ~2 minutes

🚥 Pre-merge checks | ✅ 3

✅ Passed checks (3 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly and concisely describes the main change: adding GlmMoeDsaForCausalLM to the supported model list. It follows the required format with [None][fix] prefix and directly relates to the changeset.
Description check	✅ Passed	The description provides a clear summary of the issue, explains the root cause, describes the fix, and includes a comprehensive test plan. All critical sections are addressed, making the PR's intent and changes understandable.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

_{✏️ 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 and usage tips.}

coderabbitai

Caution

Some comments are outside the diff and can’t be posted inline due to platform limitations.

⚠️ Outside diff range comments (1)

tensorrt_llm/_torch/modules/fused_moe/moe_load_balancer.py (1)
1-1: ⚠️ Potential issue | 🟠 Major

Add NVIDIA copyright header in this modified Python file.

This file is modified but the provided content has no NVIDIA OSS copyright header at the top.

As per coding guidelines, "All TensorRT-LLM Open Source Software code should contain an NVIDIA copyright header that includes the year of its latest meaningful modification."
🤖 Prompt for AI Agents
Verify each finding against the current code and only fix it if needed.

In `@tensorrt_llm/_torch/modules/fused_moe/moe_load_balancer.py` at line 1, This
file is missing the required NVIDIA OSS copyright header; add the standard
NVIDIA copyright header (with the year of the latest meaningful modification) at
the very top of the file before the first statement (before the existing "import
ctypes") so the module
tensorrt_llm/_torch/modules/fused_moe/moe_load_balancer.py contains the correct
license header.

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.

Outside diff comments:
In `@tensorrt_llm/_torch/modules/fused_moe/moe_load_balancer.py`:
- Line 1: This file is missing the required NVIDIA OSS copyright header; add the
standard NVIDIA copyright header (with the year of the latest meaningful
modification) at the very top of the file before the first statement (before the
existing "import ctypes") so the module
tensorrt_llm/_torch/modules/fused_moe/moe_load_balancer.py contains the correct
license header.

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 61380d38-9552-495d-ace6-636249243410

📥 Commits

Reviewing files that changed from the base of the PR and between f6db7e3 and 8a114de.

📒 Files selected for processing (1)

tensorrt_llm/_torch/modules/fused_moe/moe_load_balancer.py

tensorrt-cicd · 2026-03-31T04:02:25Z

PR_Github #40849 [ run ] triggered by Bot. Commit: 8a114de Link to invocation

dc3671

LGTM

tensorrt-cicd · 2026-03-31T12:43:03Z

PR_Github #40849 [ run ] completed with state SUCCESS. Commit: 8a114de
/LLM/main/L0_MergeRequest_PR pipeline #31858 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

Link to invocation

qiaoxj07 · 2026-04-01T01:08:55Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-04-01T01:15:55Z

PR_Github #41051 [ run ] triggered by Bot. Commit: 8a114de Link to invocation

tensorrt-cicd · 2026-04-01T05:38:19Z

PR_Github #41051 [ run ] completed with state FAILURE. Commit: 8a114de
/LLM/main/L0_MergeRequest_PR pipeline #32027 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

Link to invocation

qiaoxj07 · 2026-04-01T08:11:20Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-04-01T08:18:00Z

PR_Github #41158 [ run ] triggered by Bot. Commit: 6c52a3f Link to invocation

tensorrt-cicd · 2026-04-01T16:30:48Z

PR_Github #41158 [ run ] completed with state SUCCESS. Commit: 6c52a3f
/LLM/main/L0_MergeRequest_PR pipeline #32127 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

Link to invocation

qiaoxj07 · 2026-04-01T23:57:40Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-04-02T00:04:25Z

PR_Github #41283 [ run ] triggered by Bot. Commit: 6c52a3f Link to invocation

tensorrt-cicd · 2026-04-02T02:02:05Z

PR_Github #41283 [ run ] completed with state FAILURE. Commit: 6c52a3f
/LLM/main/L0_MergeRequest_PR pipeline #32241 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

Link to invocation

qiaoxj07 · 2026-04-02T05:39:42Z

/bot run --disable-fail-fast

tensorrt-cicd · 2026-04-02T05:45:30Z

PR_Github #41348 [ run ] triggered by Bot. Commit: 6c52a3f Link to invocation

tensorrt-cicd · 2026-04-02T09:38:54Z

PR_Github #41348 [ run ] completed with state SUCCESS. Commit: 6c52a3f
/LLM/main/L0_MergeRequest_PR pipeline #32296 completed with status: 'SUCCESS'

CI Report

Link to invocation

…IDIA#12607) Signed-off-by: Xianjie <5410381+qiaoxj07@users.noreply.github.com>

qiaoxj07 requested a review from a team as a code owner March 31, 2026 03:55

qiaoxj07 requested a review from yuxianq March 31, 2026 03:55

github-actions bot assigned qiaoxj07 Mar 31, 2026

coderabbitai bot reviewed Mar 31, 2026

View reviewed changes

qiaoxj07 requested a review from dc3671 March 31, 2026 03:59

dc3671 approved these changes Mar 31, 2026

View reviewed changes

dc3671 requested a review from xxi-nv March 31, 2026 04:17

yuxianq approved these changes Mar 31, 2026

View reviewed changes

Merge branch 'main' into fix/eplb-glm5-support

6c52a3f

qiaoxj07 merged commit 5c1c1e2 into NVIDIA:main Apr 2, 2026
5 checks passed

karen-sy pushed a commit to karen-sy/TensorRT-LLM that referenced this pull request Apr 7, 2026

[None][fix] Add GlmMoeDsaForCausalLM to EPLB supported model list (NV…

7d16047

…IDIA#12607) Signed-off-by: Xianjie <5410381+qiaoxj07@users.noreply.github.com>

Conversation

qiaoxj07 commented Mar 31, 2026 • edited by coderabbitai bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Test plan

Summary by CodeRabbit

Uh oh!

qiaoxj07 commented Mar 31, 2026

Uh oh!

coderabbitai bot commented Mar 31, 2026

Walkthrough

Changes

Estimated code review effort

Uh oh!

coderabbitai bot left a comment

Choose a reason for hiding this comment

Uh oh!

tensorrt-cicd commented Mar 31, 2026

Uh oh!

dc3671 left a comment

Choose a reason for hiding this comment

Uh oh!

tensorrt-cicd commented Mar 31, 2026

Uh oh!

qiaoxj07 commented Apr 1, 2026

Uh oh!

tensorrt-cicd commented Apr 1, 2026

Uh oh!

tensorrt-cicd commented Apr 1, 2026

Uh oh!

qiaoxj07 commented Apr 1, 2026

Uh oh!

tensorrt-cicd commented Apr 1, 2026

Uh oh!

tensorrt-cicd commented Apr 1, 2026

Uh oh!

qiaoxj07 commented Apr 1, 2026

Uh oh!

tensorrt-cicd commented Apr 2, 2026

Uh oh!

tensorrt-cicd commented Apr 2, 2026

Uh oh!

qiaoxj07 commented Apr 2, 2026

Uh oh!

tensorrt-cicd commented Apr 2, 2026

Uh oh!

tensorrt-cicd commented Apr 2, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

qiaoxj07 commented Mar 31, 2026 •

edited by coderabbitai bot

Loading