[https://nvbugs/6528742][test] Disable V2 in legacy Mamba test

[VALLIS-NERIA]

What changed

• Explicitly set KvCacheConfig.use_kv_cache_manager_v2=False in the parameterized Nemotron V3 Super legacy Mamba cache manager test.

Why

The model defaults can otherwise select MambaHybridCacheManagerV2, even when the test sets TRTLLM_USE_PY_MAMBA=1. This made the python_mamba_cache case stop exercising the intended legacy MixedMambaHybridCacheManager.

Pinning V2 off keeps both legacy Python and C++ manager parameterizations on their intended paths.

Validation

• pre-commit run --files tests/integration/defs/accuracy/test_llm_api_pytorch.py
• pytest tests/unittest/_torch/executor/test_mamba_cache_manager.py -k hybrid_cache_manager_factory_honors_v2_setting — 4 passed
• B200, 4 GPUs:
  TestNemotronV3Super::test_fp8_4gpus[attention_dp_on-python_mamba_cache] — passed
  • MMLU: 85.526 (threshold 83.883)
  • GSM8K: 92.456 (threshold 88.897)
  • Runtime log confirmed MixedMambaHybridCacheManager

Dev Engineer Review

• Updated test_fp8_4gpus to explicitly set KvCacheConfig.use_kv_cache_manager_v2=False, ensuring the test exercises the intended legacy Mamba cache manager regardless of model defaults or TRTLLM_USE_PY_MAMBA.
• The change is minimal, uses a valid existing API field, and does not alter production behavior or public interfaces.
• Validation reported: pre-commit checks, four passing unit tests, and a successful B200 4-GPU run with MMLU and GSM8K scores above thresholds.

QA Engineer Review

• Modified test: test_fp8_4gpus in tests/integration/defs/accuracy/test_llm_api_pytorch.py.
• No test-list changes were identified; coverage should remain associated with its existing test-list entry.
• Verdict: sufficient.

[VALLIS-NERIA]

/bot run --disable-fail-fast

[tensorrt-cicd]

PR_Github #62722 [ run ] triggered by Bot. Commit: adfb17f Link to invocation

[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: 0c507eaa-9a14-4dd0-bc67-96a46ce81685

📥 Commits

Reviewing files that changed from the base of the PR and between 78ab4e7 and adfb17f.

📒 Files selected for processing (1)

• tests/integration/defs/accuracy/test_llm_api_pytorch.py

---

Walkthrough

The FP8 four-GPU Nemotron test now explicitly disables KV-cache-manager-v2 in its KvCacheConfig.

Changes

KV Cache Configuration

Layer / File(s)	Summary
Explicit KV cache manager setting tests/integration/defs/accuracy/test_llm_api_pytorch.py	The FP8 four-GPU test sets use_kv_cache_manager_v2=False in KvCacheConfig.

Estimated code review effort: 1 (Trivial) | ~2 minutes

Suggested reviewers: bowenfu

🚥 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 is concise, specific, and matches the test-only change to disable KV-cache-manager V2 in the legacy Mamba test.
Description check	✅ Passed	The description clearly explains what changed, why, and how it was validated; only the checklist section is omitted.
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.}

[tensorrt-cicd]

PR_Github #62722 [ run ] completed with state SUCCESS. Commit: adfb17f
/LLM/main/L0_MergeRequest_PR pipeline #50857 completed with status: 'SUCCESS'

CI Report

Link to invocation