[https://nvbugs/6248827][fix] Wrap both accesses in getattr(self.config, 'num_nextn_predict_layers', 0)…

[tensorrt-cicd]

Summary

• Root cause: DeepseekV3WeightLoader.load_weights unconditionally reads self.config.num_nextn_predict_layers, but _MistralPretrainedConfig (non-MTP, routed here since commit 651853b made MistralLarge3 delegate to super().load_weights) has no such attribute, raising AttributeError before weight loading.
• Fix: Wrap both accesses in getattr(self.config, 'num_nextn_predict_layers', 0) (mirroring the existing precedent in modeling_exaone_moe.py:649-650), keeping the normal weight-load path running; also removed the now-fixed waiver (waives.txt:126).
• Automated fix generated by repair-bot

Test plan

• Verify fix on the same GPU type as the original failure
• Check for regressions in related tests

Links

• Bug: https://nvbugs/6248827

Summary by CodeRabbit

• Bug Fixes

  • Improved robustness of model weight loading for DeepseekV3, with better handling of missing configuration attributes.

• Tests

  • Updated test configuration for accuracy validation.

[coderabbitai]

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: 9197c505-a541-4a0e-bc24-4e897e4f812c

📥 Commits

Reviewing files that changed from the base of the PR and between 846d0f4 and d121154.

📒 Files selected for processing (2)

• tensorrt_llm/_torch/models/modeling_deepseekv3.py
• tests/integration/test_lists/waives.txt

💤 Files with no reviewable changes (1)

• tests/integration/test_lists/waives.txt

---

📝 Walkthrough

Walkthrough

This PR refactors MTP weight detection in DeepseekV3 to handle missing config attributes gracefully, replacing a helper function with direct safe-default computation. It also updates test waivers for Mistral test parameterization.

Changes

DeepseekV3 weight loading and test waiver updates

Layer / File(s)	Summary
DeepseekV3 shared MTP checkpoint detection refactor tensorrt_llm/_torch/models/modeling_deepseekv3.py	Simplifies DeepseekV3WeightLoader.load_weights by removing the detect_shared_mtp_weights() helper and directly computing has_shared_mtp_weights using getattr defaults for num_nextn_predict_layers and _ckpt_num_nextn_predict_layers, making the logic more robust to missing config attributes.
Test waiver parameter update tests/integration/test_lists/waives.txt	Updates the waiver entry for TestMistralLarge3_675B::test_nvfp4_4gpus, replacing parameterization latency_moe_trtllm with latency_moe_trtllm_eagle and its associated NV bug link.

Estimated code review effort

🎯 1 (Trivial) | ⏱️ ~5 minutes

Possibly related PRs

• NVIDIA/TensorRT-LLM#14789: Both PRs modify the waiver for accuracy/test_llm_api_pytorch.py::TestMistralLarge3_675B::test_nvfp4_4gpus[latency_moe_trtllm*] parameterizations.
• NVIDIA/TensorRT-LLM#14457: Both PRs refactor shared-MTP weight detection in DeepseekV3 weight loading, using derived model and checkpoint layer counts.
• NVIDIA/TensorRT-LLM#14792: Both PRs adjust test waivers for accuracy/test_llm_api_pytorch.py::TestMistralLarge3_675B::test_nvfp4_4gpus parameterizations.

Suggested reviewers

• hlu1
• danielafrimi
• jieli-matrix
• joyang-nv

🚥 Pre-merge checks | ✅ 4 | ❌ 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 (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly identifies the main change: wrapping accesses to getattr(self.config, 'num_nextn_predict_layers', 0) to fix an AttributeError in DeepseekV3WeightLoader.
Description check	✅ Passed	The description includes root cause, solution, test plan verification, and bug link. All required template sections are adequately addressed.
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 and usage tips.}

[2ez4bz]

/bot run

[tensorrt-cicd]

PR_Github #57788 [ run ] triggered by Bot. Commit: c753b08 Link to invocation

[tensorrt-cicd]

PR_Github #57788 [ run ] completed with state SUCCESS. Commit: c753b08
/LLM/main/L0_MergeRequest_PR pipeline #46487 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

[MartinMarciniszyn]

Replaced by !15123