[None][fix] Enforce minimum NVSHMEM_QP_DEPTH of 128 for DeepEP low latency

[Tabrizian]

Summary by CodeRabbit

Release Notes

• Bug Fixes
  • Improved stability for deep expert parallelism configurations with low-latency communication.

Description

Enforce a minimum NVSHMEM_QP_DEPTH of 128 for DeepEP low latency communication to avoid NVSHMEM errors when the token limit is small. Previously the depth was computed as 2 * (max_tokens + 1), which could be too low.

Test Coverage

• Existing DeepEP low latency MoE tests cover this code path.

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

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

[Tabrizian]

/bot run --disable-fail-fast

[coderabbitai]

📝 Walkthrough

Walkthrough

The DeepEPLowLatency class now enforces a minimum threshold of 128 for NVSHMEM_QP_DEPTH by computing it as the maximum of 128 and the previous calculation. The copyright year in the header has also been updated.

Changes

Cohort / File(s)	Summary
NVSHMEM Queue Depth Constraint tensorrt_llm/_torch/modules/fused_moe/communication/deep_ep_low_latency.py	Modified NVSHMEM_QP_DEPTH initialization to enforce a lower bound of 128 using max(128, 2 * (deep_ep_max_num_tokens + 1)). Updated copyright year header from 2025 to 2025-2026.

Estimated code review effort

🎯 1 (Trivial) | ⏱️ ~3 minutes

🚥 Pre-merge checks | ✅ 2 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly describes the main change: enforcing a minimum NVSHMEM_QP_DEPTH of 128 for DeepEP low latency, which matches the core modification in the changeset.
Description check	✅ Passed	The PR description clearly explains the issue (NVSHMEM errors with small token limits) and the solution (enforcing a minimum NVSHMEM_QP_DEPTH of 128). It includes test coverage information and a completed checklist.

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

✨ Finishing Touches

🧪 Generate unit tests (beta)

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

---

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

[tensorrt-cicd]

PR_Github #38548 [ run ] triggered by Bot. Commit: d14389d Link to invocation

[tensorrt-cicd]

PR_Github #38548 [ run ] completed with state SUCCESS. Commit: d14389d
/LLM/main/L0_MergeRequest_PR pipeline #29892 completed with status: 'SUCCESS'

CI Report

Link to invocation