[https://nvbugs/6414760][fix] Rewrite the test with the PyTorch LLM API (LLM + SamplingParams(n=4…

[trtllm-agent]

Summary

• Root cause: PR [TRTLLM-13781][chore] Remove legacy TensorRT examples #15763 (e5a05b2) removed examples/models/core/qwen/{requirements.txt,convert_checkpoint.py} as part of legacy TRT cleanup, but test_qwen_e2e_cpprunner_large_new_tokens still consumed the qwen_example_root fixture + convert_weights + trtllm-build + ModelRunnerCpp
• Fix: Rewrite the test with the PyTorch LLM API (LLM + SamplingParams(n=4, best_of=4)) to preserve the RCCA invariant (no zero-length among 4 returned sequences) without touching deleted legacy files; keep the same test id and parametrize
• 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/6414760

Summary by CodeRabbit

• Tests
  • Simplified an end-to-end generation test to use a more direct inference flow.
  • Improved coverage for multi-sample text generation by checking that each returned completion contains tokens.
  • Reduced test setup requirements, making the test easier to run and maintain.

[coderabbitai]

📝 Walkthrough

Walkthrough

The test test_qwen_e2e_cpprunner_large_new_tokens in test_e2e.py was rewritten to use tensorrt_llm.LLM and SamplingParams for generation instead of convert_weights, trtllm-build, and ModelRunnerCpp. The unused convert_weights import was removed and the test signature simplified.

Changes

Qwen e2e test rewrite

Layer / File(s)	Summary
Simplified generation flow tests/integration/defs/test_e2e.py	Removed convert_weights import; replaced the conversion/build/tokenizer/ModelRunnerCpp flow in test_qwen_e2e_cpprunner_large_new_tokens with direct tensorrt_llm.LLM/SamplingParams generation, asserting all completions have non-empty token_ids; test signature reduced to (model_name, model_path, llm_venv).

Estimated code review effort: 2 (Simple) | ~10 minutes

Suggested reviewers: yibinl-nvidia, nv-guomingz

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title is specific and matches the main change: rewriting the Qwen test to use the PyTorch LLM API for bug 6414760.
Description check	✅ Passed	The description clearly states the root cause, fix, test plan, and bug link, with only minor template-section differences.
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.}

[coderabbitai]

🧹 Nitpick comments (1)

tests/integration/defs/test_e2e.py (1)

75-92: 📐 Maintainability & Code Quality | 🔵 Trivial | 💤 Low value

llm_venv fixture is now unused.

The rewritten test runs generation in-process via LLM, so the llm_venv fixture is no longer exercised. Requesting it still triggers workspace creation and teardown (see llm_venv in conftest.py) for no purpose. Consider dropping it from the signature if parametrization/markers don't depend on it. Note: model_name should stay if it's supplied by a @pytest.mark.parametrize decorator (lines 31-74, not shown).

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@tests/integration/defs/test_e2e.py` around lines 75 - 92, The test
test_qwen_e2e_cpprunner_large_new_tokens no longer uses the llm_venv fixture
because generation now runs directly through LLM in-process. Remove llm_venv
from the test function signature so pytest does not create and tear down the
unused workspace, and keep model_name only if it is still needed by the existing
parametrization in this test module.

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Nitpick comments:
In `@tests/integration/defs/test_e2e.py`:
- Around line 75-92: The test test_qwen_e2e_cpprunner_large_new_tokens no longer
uses the llm_venv fixture because generation now runs directly through LLM
in-process. Remove llm_venv from the test function signature so pytest does not
create and tear down the unused workspace, and keep model_name only if it is
still needed by the existing parametrization in this test module.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: a2c9155d-d12d-4cb0-b09b-7190fe8eea68

📥 Commits

Reviewing files that changed from the base of the PR and between a0c406f and 5aaaa5a.

📒 Files selected for processing (1)

• tests/integration/defs/test_e2e.py

[Wanli-Jiang]

I will include the fix in #15907 to save CI resources.