Skip to content

[None][test] fix perf test cases issue of incorrect match #12096

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

Open
ruodil wants to merge 2 commits into
base: main
Choose a base branch
from
+15 −2
Open

[None][test] fix perf test cases issue of incorrect match #12096

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
c4905ec
fix perf test cases issue
ruodil Mar 11, 2026
d3be4bb
Update tests/integration/test_lists/qa/llm_perf_core.yml
ruodil Mar 12, 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
17 changes: 15 additions & 2 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 @@ -5,7 +5,8 @@ llm_perf_core:
# ===============================================================================
# 1: All GPUs common tests(L20, L40S, H100, H20, H200, GB200, B200, B300, GB300, RTX-6000D, RTX-6000-Server test cases)
# 2: L20, L40S, H100, H20, H200
# 3: L40S, H100, H20, H200
# 3: L40S, H100, H20, H200 (4 GPUs)
# 3b: L40S, H100, H20, H200 (8 GPUs)
# 4: H100, H20, H200 test cases
# 5: H100, H20, H200, GB200, B200, B300, GB300, RTX6000-D, RTX6000-Server test cases
# 6: GB200, B200, B300, GB300, RTX6000-Server test cases
Expand Down Expand Up @@ -136,6 +137,17 @@ llm_perf_core:
- perf/test_perf.py::test_perf[llama_v3.3_70b_instruct_fp8-bench-pytorch-float8-input_output_len:2000,500-gpus:4]
- perf/test_perf.py::test_perf[llama_v3.3_70b_instruct_fp8-bench-pytorch-float8-input_output_len:128,128-gpus:4]
- perf/test_perf.py::test_perf[llama_v3.3_70b_instruct_fp8-bench-pytorch-float8-input_output_len:512,32-gpus:4] #llama_v3.3_70b_instruct_fp8


# 3b: H100, H20, H200 (8 GPUs)
- condition:
ranges:
system_gpu_count:
gte: 8
compute_capability:
gt: 8.0
Copy link
Collaborator

@yufeiwu-nv yufeiwu-nv Mar 11, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Suggested change
gt: 8.0
gte: 9.0

Copy link
Collaborator

@yufeiwu-nv yufeiwu-nv Mar 11, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

maybe remove 8XL40S because it's low-priority and 8XGPU is SYS topo which lead to unstable

Copy link
Collaborator

@yufeiwu-nv yufeiwu-nv Mar 11, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

if remove 8XL40S, maybe you can merge 3b into 4

lte: 9.0
tests:
- perf/test_perf.py::test_perf[llama_v3.3_70b_instruct_fp8-bench-pytorch-streaming-float8-input_output_len:512,32-gpus:8]
- perf/test_perf.py::test_perf[llama_v3.3_70b_instruct_fp8-bench-pytorch-float8-input_output_len:512,32-gpus:8]
- perf/test_perf.py::test_perf[llama_v3.3_70b_instruct_fp8-bench-pytorch-float8-input_output_len:2000,200-gpus:8]
Expand Down Expand Up @@ -185,7 +197,6 @@ llm_perf_core:
- perf/test_perf.py::test_perf[phi_4_multimodal_instruct_audio-bench-pytorch-bfloat16-input_output_len:1000,1000-loras:1-con:250]
- perf/test_perf.py::test_perf[phi_4_multimodal_instruct_image-bench-pytorch-bfloat16-input_output_len:1000,1000-loras:1-con:250]
- perf/test_perf.py::test_perf[deepseek_v3_lite_fp8-bench-pytorch-float8-input_output_len:128,128]
- perf/test_perf.py::test_perf[qwen3_235b_a22b_fp8-bench-pytorch-float8-input_output_len:1000,2000-con:256-ep:8-gpus:8]


# 6: GB200, B200, B300, GB300, RTX6000-Server test cases
Expand Down Expand Up @@ -381,6 +392,8 @@ llm_perf_core:
lte: 12.0
tests:
- perf/test_perf.py::test_perf[llama_v4_maverick_17b_128e_instruct_fp8-bench-pytorch-float8-input_output_len:128,128-ep:8-tp:8-gpus:8]
# qwen3_235b_a22b_fp8
- perf/test_perf.py::test_perf[qwen3_235b_a22b_fp8-bench-pytorch-float8-input_output_len:1000,2000-con:256-ep:8-gpus:8]


# 12: RTX-6000D, RTX-6000 Server test cases
Expand Down
Loading