[None][test] Add support for nemotron_3_ultra_550b_nvfp4 model in performance tests and configurations

[yufeiwu-nv]

• Updated model paths to include nemotron_3_ultra_550b_nvfp4 in HF_MODEL_PATH.
• Added configuration settings for nemotron_3_ultra_550b_nvfp4 in pytorch_model_config.py.
• Included new performance test cases for nemotron_3_ultra_550b_nvfp4 in test_perf.py and updated llm_perf_core.yml.
• Cleaned up legacy model name handling in test_perf.py.

Signed-off-by: yufeiwu-nv 230315618+yufeiwu-nv@users.noreply.github.com

Summary by CodeRabbit

Release Notes

• Tests
  • Added performance test support for Nemotron 3 Ultra 550B model with NVFP4 quantization.
  • Updated performance testing infrastructure from TensorRT to PyTorch backend.
  • Refined GPU performance test matrix for improved coverage and test configurations.

Description

Test Coverage

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)

• If PR introduces API changes, an appropriate PR label is added - either api-compatible or api-breaking. For api-breaking, include BREAKING in the PR title.

• 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.

[yufeiwu-nv]

/bot skip --comment "only test modify"

[tensorrt-cicd]

PR_Github #53064 [ skip ] triggered by Bot. Commit: ea9ef23 Link to invocation

[tensorrt-cicd]

PR_Github #53064 [ skip ] completed with state SUCCESS. Commit: ea9ef23
Skipping testing for commit ea9ef23

Link to invocation

[coderabbitai]

📝 Walkthrough

Walkthrough

This PR adds support for the NVIDIA Nemotron-3-Ultra-550B model to TensorRT-LLM performance testing infrastructure, refactors model directory and benchmark command generation logic, and reorganizes the GPU performance test matrix across multiple hardware and backend configurations.

Changes

Performance test infrastructure and Nemotron model support

Layer / File(s)	Summary
Model path and configuration setup tests/integration/defs/perf/_model_paths.py, tests/integration/defs/perf/test_perf.py, tests/integration/defs/perf/pytorch_model_config.py	Registers nemotron_3_ultra_550b_nvfp4 in HF_MODEL_PATH pointing to nvidia/NVIDIA-Nemotron-3-Ultra-550B-A55B-NVFP4, adds it to NEMOTRON_SUPER_MODELS and TRUST_REMOTE_CODE_MODELS sets, and defines PyTorch-specific runtime configuration including attention DP, MoE backend (CUTEDSL), CUDA graph padding, and KV-cache settings.
Model directory resolution and bench command refactoring tests/integration/defs/perf/test_perf.py	get_model_dir() now directly checks HF_MODEL_PATH first, falls back to MODEL_PATH_DICT with llm_models_root(), and returns empty string otherwise, removing legacy _hf suffix stripping. get_trtllm_bench_build_command() always appends _hf suffix to model names passed to trtllm-bench. get_trtllm_bench_command() simplifies engine directory path to use only TP/PP subdirectories, removing intermediate hf_model_name component.
Performance test matrix reorganization tests/integration/test_lists/qa/llm_perf_core.yml, tests/integration/test_lists/test-db/l0_perf.yml	In qa/llm_perf_core.yml: condition #2 narrowed to single qwen3.5_27b min-latency test; condition #3 expanded with nemotron_nano_12b_v2, qwen3.5_27b, and llama_v3.3_nemotron_super_49b throughput variants; added nemotron_3_ultra_550b_nvfp4-serve test set. In l0_perf.yml: migrated H100 l0_perf from TensorRT to PyTorch backend for llama_v3.1_8b_instruct; added L40S PyTorch condition for bielik_11b_v2.2_instruct bfloat16.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~25 minutes

Possibly related PRs

• NVIDIA/TensorRT-LLM#14952: Modifies tests/integration/test_lists/qa/llm_perf_core.yml to update perf test cases and Nemotron-related model entries with overlapping test matrix changes.

Suggested reviewers

• niukuo
• StanleySun639
• 2ez4bz
• tburt-nv

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 inconclusive)

Check name	Status	Explanation	Resolution
Description check	❓ Inconclusive	The description provides a bullet-point list of changes but lacks details on the motivation/rationale, test coverage specifics, and whether API changes apply. The template's Description and Test Coverage sections are not adequately filled.	Add context explaining why this model is being added, specify which test files validate the new configuration, and clarify whether checklist items regarding API changes and dependencies apply.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly describes the main change: adding support for the nemotron_3_ultra_550b_nvfp4 model in performance tests and configurations, which aligns with the changeset additions across multiple test files.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.
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.}

[coderabbitai]

Actionable comments posted: 1

🤖 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.

Inline comments:
In `@tests/integration/defs/perf/test_perf.py`:
- Around line 1042-1049: The build command list uses an unnecessary f-string for
a static literal; update the build_cmd construction (where self._build_script,
"--log_level=info", f"--workspace={engine_dir}", f"--model={model_name}",
f"--model_path={model_dir}", "build") by removing the leading f from the
"--log_level=info" item so it becomes a plain string; keep the other f-strings
for workspace, model and model_path that interpolate engine_dir, model_name and
model_dir respectively.

🪄 Autofix (Beta)

Fix all unresolved CodeRabbit comments on this PR:

• Push a commit to this branch (recommended)
• Create a new PR with the fixes

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Enterprise

Run ID: 49b63a26-7cc0-400f-8dec-10d15bd63e4a

📥 Commits

Reviewing files that changed from the base of the PR and between ba6ba1f and f782448.

📒 Files selected for processing (5)

• tests/integration/defs/perf/_model_paths.py
• tests/integration/defs/perf/pytorch_model_config.py
• tests/integration/defs/perf/test_perf.py
• tests/integration/test_lists/qa/llm_perf_core.yml
• tests/integration/test_lists/test-db/l0_perf.yml

💤 Files with no reviewable changes (1)

• tests/integration/test_lists/test-db/l0_perf.yml

[yufeiwu-nv]

/bot skip --comment "only test modify"

[yufeiwu-nv]

/bot skip --comment "only test modify"

[tensorrt-cicd]

PR_Github #53250 [ skip ] triggered by Bot. Commit: 5c94d9b Link to invocation

[tensorrt-cicd]

PR_Github #53250 [ skip ] completed with state SUCCESS. Commit: 5c94d9b
Skipping testing for commit 5c94d9b

Link to invocation