Skip to content

[plugin][OOT Benchmark] Refine OOT benchmark(manual trigger) to cover key models #409

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
zejunchen-zejun wants to merge 11 commits into
base: main
Choose a base branch
from
+447 −160
Open

[plugin][OOT Benchmark] Refine OOT benchmark(manual trigger) to cover key models #409

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
11 commits
Select commit Hold shift + click to select a range
934cae4
[plugin][oot benchmark] refine the OOT benchmark workflow
zejunchen-zejun Mar 24, 2026
27fd782
add model qwen3.5
zejunchen-zejun Mar 25, 2026
5f9cf75
set 4 GPU machine for Kimi-K2 TP4
zejunchen-zejun Mar 25, 2026
f236d2c
if the model has not been chosen, the gpu runner
zejunchen-zejun Mar 25, 2026
62437ef
change the config
zejunchen-zejun Mar 25, 2026
f8c77fb
remove redundant env flag for gptoss
zejunchen-zejun Mar 25, 2026
5f5e1d4
Merge branch 'main' into zejun/add_benchmark_3.24
zejunchen-zejun Mar 26, 2026
7a11c93
add specific branch trigger OOT benchmark
zejunchen-zejun Mar 26, 2026
8c8833c
change the oot benchmark behavior
zejunchen-zejun Mar 26, 2026
80bff24
refine the docker remove logic and rebuild logic
zejunchen-zejun Mar 26, 2026
f325fd6
Merge branch 'main' into zejun/add_benchmark_3.24
wuhuikx Mar 26, 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
28 changes: 19 additions & 9 deletions .github/benchmark/oot_benchmark_models.json
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,33 +1,33 @@
[
{
"display": "DeepSeek-R1 FP8",
"display": "DeepSeek-R1 FP8 TP8",
"source_path": "deepseek-ai/DeepSeek-R1-0528",
"path": "/models/DeepSeek-R1-0528",
"prefix": "deepseek-r1-fp8",
"extra_args": "--trust-remote-code --tensor-parallel-size 8 --max-num-batched-tokens 16384 --max-model-len 16384",
"bench_args": "",
"runner": "atom-mi355-8gpu.predownload",
"env_vars": "AITER_QUICK_REDUCE_QUANTIZATION=INT4\nATOM_ENABLE_QK_NORM_ROPE_CACHE_QUANT_FUSION=1"
"env_vars": "AITER_QUICK_REDUCE_QUANTIZATION=INT4"
},
{
"display": "DeepSeek-R1 MXFP4",
"display": "DeepSeek-R1 MXFP4 TP8",
"source_path": "amd/DeepSeek-R1-0528-MXFP4",
"path": "/models/DeepSeek-R1-0528-MXFP4",
"prefix": "deepseek-r1-mxfp4",
"extra_args": "--trust-remote-code --tensor-parallel-size 8 --max-num-batched-tokens 16384 --max-model-len 16384",
"bench_args": "",
"runner": "atom-mi355-8gpu.predownload",
"env_vars": "AITER_QUICK_REDUCE_QUANTIZATION=INT4\nATOM_ENABLE_QK_NORM_ROPE_CACHE_QUANT_FUSION=1"
"env_vars": "AITER_QUICK_REDUCE_QUANTIZATION=INT4"
},
{
"display": "gpt-oss-120b",
"display": "gpt-oss-120b TP1",
"source_path": "amd/gpt-oss-120b-w-mxfp4-a-fp8",
"path": "/models/gpt-oss-120b",
"prefix": "gpt-oss-120b",
"extra_args": "--trust-remote-code --tensor-parallel-size 1 --max-num-batched-tokens 16384 --max-model-len 16384",
"bench_args": "",
"runner": "linux-atom-mi355-1",
"env_vars": "AITER_QUICK_REDUCE_QUANTIZATION=INT4\nATOM_ENABLE_QK_NORM_ROPE_CACHE_QUANT_FUSION=1"
"runner": "atom-mi355-8gpu.predownload",
"env_vars": "ATOM_ENABLE_QK_NORM_ROPE_CACHE_QUANT_FUSION=1"
},
{
"display": "Kimi-K2-Thinking-MXFP4 TP4",
Expand All @@ -38,7 +38,7 @@
"extra_args": "--trust-remote-code --tensor-parallel-size 4 --max-num-batched-tokens 16384 --max-model-len 16384",
"bench_args": "",
"runner": "atom-mi355-8gpu.predownload",
"env_vars": "AITER_QUICK_REDUCE_QUANTIZATION=INT4\nATOM_ENABLE_QK_NORM_ROPE_CACHE_QUANT_FUSION=1"
"env_vars": "AITER_QUICK_REDUCE_QUANTIZATION=INT4"
},
{
"display": "Kimi-K2-Thinking-MXFP4 TP8",
Expand All @@ -49,6 +49,16 @@
"extra_args": "--trust-remote-code --tensor-parallel-size 8 --max-num-batched-tokens 16384 --max-model-len 16384",
"bench_args": "",
"runner": "atom-mi355-8gpu.predownload",
"env_vars": "AITER_QUICK_REDUCE_QUANTIZATION=INT4\nATOM_ENABLE_QK_NORM_ROPE_CACHE_QUANT_FUSION=1"
"env_vars": "AITER_QUICK_REDUCE_QUANTIZATION=INT4"
},
{
"display": "Qwen3.5-397B-A17B-FP8 TP8",
"source_path": "Qwen/Qwen3.5-397B-A17B-FP8",
"path": "/models/Qwen3.5-397B-A17B-FP8",
"prefix": "qwen3-5-397b-a17b-fp8",
"extra_args": "--trust-remote-code --tensor-parallel-size 8 --max-num-batched-tokens 16384 --max-model-len 16384",
"bench_args": "",
"runner": "atom-mi355-8gpu.predownload",
"env_vars": "AITER_QUICK_REDUCE_QUANTIZATION=INT4\nATOM_DISABLE_VLLM_PLUGIN_ATTENTION=1\nATOM_USE_CUSTOM_ALL_GATHER=0"
}
]
12 changes: 12 additions & 0 deletions .github/scripts/oot_benchmark_to_dashboard.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -50,6 +50,15 @@ def append_metric(
entries.append(entry)


def is_dashboard_publish_allowed(payload: dict) -> bool:
publish_flag = payload.get("dashboard_publish_allowed")
if publish_flag is None:
return True
if isinstance(publish_flag, bool):
return publish_flag
return str(publish_flag).strip().lower() not in {"0", "false", "no"}


def build_entries(result_dir: Path, run_url: str | None) -> list[dict]:
entries: list[dict] = []

Expand All @@ -62,6 +71,9 @@ def build_entries(result_dir: Path, run_url: str | None) -> list[dict]:
except (OSError, UnicodeDecodeError, json.JSONDecodeError):
continue

if not is_dashboard_publish_allowed(payload):
continue

if "output_throughput" not in payload:
continue

Expand Down
Loading
Loading