Skip to content

[None][test] Adjust timeout cases in QA perf test #16894

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
yufeiwu-nv merged 2 commits into from
Jul 27, 2026
Merged

[None][test] Adjust timeout cases in QA perf test #16894

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
29286b3
Update LLM performance test configurations in llm_perf_core.yml to ad…
yufeiwu-nv Jul 27, 2026
f14b548
Merge branch 'main' into fp4
yufeiwu-nv Jul 27, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 5 additions & 3 deletions tests/integration/test_lists/qa/llm_perf_core.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -37,7 +37,8 @@ llm_perf_core:
# test overlap scheduler
# con:1 paired with a small model is an intentional design choice—it amplifies host-side overhead and simplifies execution timelines to the maximum extent.
- perf/test_perf.py::test_perf[qwen3_0.6b-bench-pytorch-bfloat16-maxnt:2048-input_output_len:8000,1000-reqs:256-con:1-pp:4-gpus:4]
- perf/test_perf.py::test_perf[qwen3_30b_a3b-bench-pytorch-bfloat16-maxnt:2048-input_output_len:8000,1000-reqs:256-con:1-pp:4-gpus:4]
# 30B MoE uses fewer reqs to fit the 1h timeout; con:1 host-overhead signal is unaffected
- perf/test_perf.py::test_perf[qwen3_30b_a3b-bench-pytorch-bfloat16-maxnt:2048-input_output_len:8000,1000-reqs:64-con:1-pp:4-gpus:4]


# 2: L40S, H100, H20
Expand All @@ -60,9 +61,7 @@ llm_perf_core:
- perf/test_perf.py::test_perf[llama_v3.3_nemotron_super_49b-bench-pytorch-bfloat16-input_output_len:128,128-tp:4-gpus:4]
- perf/test_perf.py::test_perf[llama_v3.3_nemotron_super_49b-bench-pytorch-bfloat16-input_output_len:500,2000-tp:4-gpus:4]
- perf/test_perf.py::test_perf[llama_v3.3_nemotron_super_49b-bench-pytorch-bfloat16-input_output_len:2000,500-tp:4-gpus:4]
- perf/test_perf.py::test_perf[llama_v3.3_nemotron_super_49b-bench-pytorch-bfloat16-input_output_len:8000,1000-tp:4-gpus:4]
- perf/test_perf.py::test_perf[llama_v3.3_nemotron_super_49b-bench-pytorch-bfloat16-input_output_len:1000,2000-tp:4-gpus:4]
- perf/test_perf.py::test_perf[llama_v3.3_nemotron_super_49b-bench-pytorch-bfloat16-input_output_len:8000,1000-con:250-tp:4-gpus:4] #max_throughput


# 3: H100, H20 test cases
Expand Down Expand Up @@ -98,6 +97,9 @@ llm_perf_core:
- perf/test_perf.py::test_perf[llama_v3.3_nemotron_super_49b-bench-pytorch-bfloat16-input_output_len:1000,2000-tp:2-gpus:2]
- perf/test_perf.py::test_perf[llama_v3.3_nemotron_super_49b-bench-pytorch-bfloat16-maxbs:1-input_output_len:8000,1000-reqs:10-con:1-tp:2-gpus:2] #min_latency
- perf/test_perf.py::test_perf[llama_v3.3_nemotron_super_49b-bench-pytorch-bfloat16-input_output_len:8000,1000-con:250-tp:2-gpus:2] #max_throughput
#llama_v3.3_nemotron_super_49b (nemotron-nas BF16 94G, 4-GPU): too slow on L40S PCIe-only interconnect
- perf/test_perf.py::test_perf[llama_v3.3_nemotron_super_49b-bench-pytorch-bfloat16-input_output_len:8000,1000-tp:4-gpus:4]
- perf/test_perf.py::test_perf[llama_v3.3_nemotron_super_49b-bench-pytorch-bfloat16-input_output_len:8000,1000-con:250-tp:4-gpus:4] #max_throughput


# 4: H100, H20, GB200, B200, B300, GB300, RTX6000-Server test cases
Expand Down
Loading