[None][test] Fix deepseek-r1 OOM issue for H100 perf test

[yufeiwu-nv]

Summary by CodeRabbit

• Tests
  • Updated internal test configuration for GPU compatibility verification.

---

Note: This release contains no user-facing changes. The updates are limited to testing infrastructure reorganization.

Weekly test found below deepseek fp8 models OOM on H100 (81*8GB), which is expected.
Therefore, modify test list.

Benchmark	input_output_len	reqs	con	maxbs	maxnt	kv_frac	ep/tp/gpus	Notes
deepseek_r1_0528_fp8-bench	1000,1000	20000	-	-	-	-	8/8/8	pytorch-float8
deepseek_r1_0528_fp8-bench	1000,2000	3000	-	-	-	-	8/8/8	pytorch-float8
deepseek_r1_fp8-bench	1000,2000	10	1	1	-	-	4/8/8	pytorch-float8
deepseek_r1_fp8-bench	1000,1000	3000	-	1024	4096	-	8/8/8	pytorch-float8
deepseek_r1_fp8-bench	1000,2000	5120	1024	128	1127	-	8/8/8	pytorch-float8
deepseek_r1_fp8-bench	2000,2000	200	-	256	1024	0.85	8/8/8	pytorch-float8
deepseek_r1_fp8-bench	128,128	-	-	32	-	-	8/8/8	pytorch-float8
deepseek_r1_fp8-bench	5000,500	200	-	512	2048	0.85	8/8/8	pytorch-float8
deepseek_r1_fp8-bench	128,128	-	-	32	-	-	8/8/8	pytorch-streaming-float8

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)

• 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.

[coderabbitai]

📝 Walkthrough

Walkthrough

Test condition blocks in the LLM performance YAML configuration are reorganized. Block #10 is redefined to target H20/H200/B200/B300 GPUs (removing H100), with RTX-6000 Server, RTX-6000D, and RTX6000-Server variants subsequently renumbered to blocks #11–#13. No test logic or case definitions are altered.

Changes

Cohort / File(s)	Summary
GPU Test Configuration Reorganization tests/integration/test_lists/qa/llm_perf_core.yml	Renumbered test condition blocks; removed H100 from block #10 (now H20/H200/B200/B300 only), shifted RTX-6000 variants to blocks #11–#13 with corresponding header updates.

Estimated code review effort

🎯 1 (Trivial) | ⏱️ ~3 minutes

🚥 Pre-merge checks | ✅ 2 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Description check	⚠️ Warning	PR description lacks key information required by the template. While a benchmark table and OOM context are provided, the formal description sections (Description, Test Coverage) are empty templates.	Complete the 'Description' section explaining why these specific deepseek models were removed. Fill in the 'Test Coverage' section listing affected test files and validation approach.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title clearly describes the main change: fixing a deepseek-r1 OOM issue for H100 performance tests, which aligns with the file modification in the test configuration.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment

Tip

Try Coding Plans. Let us write the prompt for your AI agent so you can ship faster (with fewer bugs).
Share your feedback on Discord.

---

_{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 the current code and only fix it if needed.

Inline comments:
In `@tests/integration/test_lists/qa/llm_perf_core.yml`:
- Around line 350-357: The header comment tokens (e.g., "H20, H200, B200, B300"
and "RTX-6000 Server") do not match the selectors in the condition.ranges
(system_gpu_count, compute_capability, gpu_memory) and may unintentionally
include other SKUs; either tighten the predicates in the condition (e.g., narrow
compute_capability and add explicit gpu_model or vendor tags, or raise
gpu_memory threshold) to exactly target the listed GPUs, or broaden the header
text to describe the selector intent (e.g., "any device with compute_capability
>=9.0 <12.0 and gpu_memory>90000") so scheduling intent is unambiguous; apply
the same change to the matching block that mentions "RTX-6000 Server" which
currently uses compute_capability lte 12.0.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: CHILL

Plan: Pro

Run ID: 01af4a24-7917-4d7f-ae27-98dfa7900e1c

📥 Commits

Reviewing files that changed from the base of the PR and between 2ee7dba and 66bf1ea.

📒 Files selected for processing (1)

• tests/integration/test_lists/qa/llm_perf_core.yml

[yufeiwu-nv]

/bot skip --comment "only test list modify"

[tensorrt-cicd]

PR_Github #37854 [ skip ] triggered by Bot. Commit: 66bf1ea Link to invocation

[tensorrt-cicd]

PR_Github #37854 [ skip ] completed with state SUCCESS. Commit: 66bf1ea
Skipping testing for commit 66bf1ea

Link to invocation

[yufeiwu-nv]

/bot skip --comment "only test list modify"

[tensorrt-cicd]

PR_Github #37949 [ skip ] triggered by Bot. Commit: cca58ce Link to invocation

[tensorrt-cicd]

PR_Github #37949 [ skip ] completed with state SUCCESS. Commit: cca58ce
Skipping testing for commit cca58ce

Link to invocation