[https://nvbugs/6422315][fix] Decorate PythonMambaCacheManager.add_dummy_requests with…

[trtllm-agent]

Summary

• Root cause: add_dummy_requests writes into inference-tensor fields (prev_num_accepted_tokens/cache_buf_idx) without a torch.inference_mode() guard, unlike sibling helpers _prepare_mamba_cache_blocks and _refresh_dummy_request_mask.
• Fix: Decorate PythonMambaCacheManager.add_dummy_requests with @torch.inference_mode() and remove the corresponding waiver line.
• 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/6422315

Summary by CodeRabbit

• Bug Fixes

  • Improved cache setup behavior during inference-only execution, reducing unnecessary runtime tracking overhead.
  • Removed an outdated test exception entry so that the related accuracy check now runs as part of the integration suite.

• Tests

  • Updated integration test expectations to better reflect current behavior.

[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: 3961846e-1661-4201-86c7-ebcc615de53c

📥 Commits

Reviewing files that changed from the base of the PR and between ce67288 and 2c60957.

📒 Files selected for processing (2)

• tensorrt_llm/_torch/pyexecutor/mamba_cache_manager.py
• tests/integration/test_lists/waives.txt

💤 Files with no reviewable changes (1)

• tests/integration/test_lists/waives.txt

---

📝 Walkthrough

Walkthrough

This PR makes two unrelated small changes: it adds a @torch.inference_mode() decorator to PythonMambaCacheManager.add_dummy_requests, and removes a SKIP waiver entry for TestNemotronSuperV3::test_mtp[bf16_ws4_180gb-trtllm] from the integration test waives list.

Changes

Small unrelated fixes

Layer / File(s)	Summary
Inference mode decorator tensorrt_llm/_torch/pyexecutor/mamba_cache_manager.py	Adds @torch.inference_mode() decorator to PythonMambaCacheManager.add_dummy_requests.
Waive list update tests/integration/test_lists/waives.txt	Removes the SKIP entry for TestNemotronSuperV3::test_mtp[bf16_ws4_180gb-trtllm] (nvbugs 6422315).

Estimated code review effort: 1 (Trivial) | ~3 minutes

Possibly related PRs

• NVIDIA/TensorRT-LLM#15960: Both PRs modify TestNemotronSuperV3 MTP-related SKIP waivers in waives.txt.
• NVIDIA/TensorRT-LLM#16027: Both PRs touch the same waive entry for test_mtp under the trtllm configuration.
• NVIDIA/TensorRT-LLM#15959: Both PRs remove SKIP waiver entries from waives.txt.

Suggested reviewers: xinhe-nv, jieli-matrix

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title is specific and matches the main change: fixing add_dummy_requests under inference mode.
Description check	✅ Passed	It covers the issue, fix, test plan, and bug link, though the template's Description and PR Checklist sections are omitted.
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.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

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

[hnover-nv]

Duplicate with 16000