Skip to content

Ichristo/vllm results validation fix #76

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
cijohnson wants to merge 3 commits into
base: main
Choose a base branch
from
Open

Ichristo/vllm results validation fix #76

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
ba3ae71
Implement config-specific result validation
cijohnson Feb 2, 2026
277697a
vllm env vars can change per model
cijohnson Feb 2, 2026
c461731
Add safe iterator wrapper to handle
cijohnson Feb 3, 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 cvs/input/config_file/inference/inferencemax/mi300x_singlenode_inferencemax.json
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -45,9 +45,11 @@
"server_script": "gptoss_fp4_mi300x_docker.sh",
"bench_serv_script": "benchmark_serving.py",
"result_dict": {
"output_throughput_per_sec": "4200",
"mean_ttft_ms": "500",
"mean_tpot_ms": "15"
"ISL=8192,OSL=1024,TP=8,CONC=64": {
"total_throughput_per_sec": "4200",
"mean_ttft_ms": "500",
"mean_tpot_ms": "15"
}
}
}
}
Expand Down
205 changes: 192 additions & 13 deletions cvs/input/config_file/inference/vllm/mi355x_singlenode_vllm.json
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -69,10 +69,57 @@
"metric_percentiles": "99",
"server_script": "gpt-oss-120b_fp4_mi355x_vllm_docker.sh",
"bench_serv_script": "benchmark_serving.py",
"vllm_env_vars": {
"VLLM_USE_AITER_UNIFIED_ATTENTION": "1",
"VLLM_ROCM_USE_AITER_MHA": "0",
"VLLM_ROCM_USE_AITER_FUSED_MOE_A16W4": "1"
},
"result_dict": {
"output_throughput_per_sec": "10600",
"mean_ttft_ms": "200",
"mean_tpot_ms": "10"
"ISL=1024,OSL=1024,TP=1,CONC=16": {
"total_throughput_per_sec": "4651",
"mean_ttft_ms": "70",
"mean_tpot_ms": "8"
},
"ISL=1024,OSL=1024,TP=1,CONC=32": {
"total_throughput_per_sec": "7043",
"mean_ttft_ms": "180",
"mean_tpot_ms": "9"
},
"ISL=1024,OSL=1024,TP=1,CONC=64": {
"total_throughput_per_sec": "10677",
"mean_ttft_ms": "76",
"mean_tpot_ms": "13"
},
"ISL=1024,OSL=8192,TP=1,CONC=16": {
"total_throughput_per_sec": "2735",
"mean_ttft_ms": "57",
"mean_tpot_ms": "7"
},
"ISL=1024,OSL=8192,TP=1,CONC=32": {
"total_throughput_per_sec": "4038",
"mean_ttft_ms": "67",
"mean_tpot_ms": "10"
},
"ISL=1024,OSL=8192,TP=1,CONC=64": {
"total_throughput_per_sec": "6140",
"mean_ttft_ms": "93",
"mean_tpot_ms": "13"
},
"ISL=8192,OSL=1024,TP=1,CONC=16": {
"total_throughput_per_sec": "16509",
"mean_ttft_ms": "335",
"mean_tpot_ms": "24"
},
"ISL=8192,OSL=1024,TP=1,CONC=32": {
"total_throughput_per_sec": "22072",
"mean_ttft_ms": "320",
"mean_tpot_ms": "19"
},
"ISL=8192,OSL=1024,TP=1,CONC=64": {
"total_throughput_per_sec": "28863",
"mean_ttft_ms": "280",
"mean_tpot_ms": "22"
}
}
},
"qwen3-235b": {
Expand Down Expand Up @@ -117,10 +164,59 @@
"metric_percentiles": "99",
"server_script": "qwen3-235b-bf16_mi355x_vllm_docker.sh",
"bench_serv_script": "benchmark_serving.py",
"vllm_env_vars": {
"VLLM_V1_USE_PREFILL_DECODE_ATTENTION": "1",
"VLLM_ROCM_USE_AITER": "1",
"VLLM_ROCM_USE_AITER_MHA": "1",
"VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4",
"SAFETENSORS_FAST_GPU": "1"
},
"result_dict": {
"output_throughput_per_sec": "8000",
"mean_ttft_ms": "300",
"mean_tpot_ms": "12"
"ISL=1024,OSL=1024,TP=8,CONC=16": {
"total_throughput_per_sec": "2000",
"mean_ttft_ms": "850",
"mean_tpot_ms": "18"
},
"ISL=1024,OSL=1024,TP=8,CONC=32": {
"total_throughput_per_sec": "3435",
"mean_ttft_ms": "80",
"mean_tpot_ms": "10"
},
"ISL=1024,OSL=1024,TP=8,CONC=64": {
"total_throughput_per_sec": "5840",
"mean_ttft_ms": "260",
"mean_tpot_ms": "10"
},
"ISL=1024,OSL=8192,TP=8,CONC=16": {
"total_throughput_per_sec": "1119",
"mean_ttft_ms": "415",
"mean_tpot_ms": "25"
},
"ISL=1024,OSL=8192,TP=8,CONC=32": {
"total_throughput_per_sec": "1876",
"mean_ttft_ms": "70",
"mean_tpot_ms": "10"
},
"ISL=1024,OSL=8192,TP=8,CONC=64": {
"total_throughput_per_sec": "3139",
"mean_ttft_ms": "310",
"mean_tpot_ms": "14"
},
"ISL=8192,OSL=1024,TP=8,CONC=16": {
"total_throughput_per_sec": "7476",
"mean_ttft_ms": "300",
"mean_tpot_ms": "21"
},
"ISL=8192,OSL=1024,TP=8,CONC=32": {
"total_throughput_per_sec": "11312",
"mean_ttft_ms": "355",
"mean_tpot_ms": "27"
},
"ISL=8192,OSL=1024,TP=8,CONC=64": {
"total_throughput_per_sec": "16082",
"mean_ttft_ms": "450",
"mean_tpot_ms": "39"
}
}
},
"qwen3-80b": {
Expand Down Expand Up @@ -165,10 +261,59 @@
"metric_percentiles": "99",
"server_script": "qwen3-80b-bf16_mi355x_vllm_docker.sh",
"bench_serv_script": "benchmark_serving.py",
"vllm_env_vars": {
"AITER_ONLINE_TUNE": "0",
"VLLM_ROCM_USE_AITER": "1",
"VLLM_ROCM_USE_AITER_MOE": "1",
"VLLM_ROCM_USE_AITER_UNIFIED_ATTENTION": "0",
"VLLM_ROCM_USE_AITER_MHA": "0"
},
"result_dict": {
"output_throughput_per_sec": "12000",
"mean_ttft_ms": "150",
"mean_tpot_ms": "8"
"ISL=1024,OSL=1024,TP=1,CONC=16": {
"total_throughput_per_sec": "2003",
"mean_ttft_ms": "180",
"mean_tpot_ms": "9"
},
"ISL=1024,OSL=1024,TP=1,CONC=32": {
"total_throughput_per_sec": "3155",
"mean_ttft_ms": "160",
"mean_tpot_ms": "8"
},
"ISL=1024,OSL=1024,TP=1,CONC=64": {
"total_throughput_per_sec": "4570",
"mean_ttft_ms": "150",
"mean_tpot_ms": "8"
},
"ISL=1024,OSL=8192,TP=1,CONC=16": {
"total_throughput_per_sec": "1200",
"mean_ttft_ms": "220",
"mean_tpot_ms": "14"
},
"ISL=1024,OSL=8192,TP=1,CONC=32": {
"total_throughput_per_sec": "1800",
"mean_ttft_ms": "200",
"mean_tpot_ms": "12"
},
"ISL=1024,OSL=8192,TP=1,CONC=64": {
"total_throughput_per_sec": "2600",
"mean_ttft_ms": "180",
"mean_tpot_ms": "10"
},
"ISL=8192,OSL=1024,TP=1,CONC=16": {
"total_throughput_per_sec": "7500",
"mean_ttft_ms": "300",
"mean_tpot_ms": "18"
},
"ISL=8192,OSL=1024,TP=1,CONC=32": {
"total_throughput_per_sec": "11300",
"mean_ttft_ms": "280",
"mean_tpot_ms": "17"
},
"ISL=8192,OSL=1024,TP=1,CONC=64": {
"total_throughput_per_sec": "16000",
"mean_ttft_ms": "260",
"mean_tpot_ms": "16"
}
}
},
"deepseek-v31": {
Expand All @@ -182,7 +327,7 @@
32,
64
],
"model": "/models/dsv31-fp8",
"model": "deepseek-ai/DeepSeek-V3.1",
"num_prompts": "3200",
"sequence_combinations": [
{
Expand All @@ -208,10 +353,44 @@
"metric_percentiles": "99",
"server_script": "dsr1_fp8_mi355x_vllm_docker.sh",
"bench_serv_script": "benchmark_serving.py",
"vllm_env_vars": {
"VLLM_ROCM_USE_AITER": "1",
"VLLM_USE_AITER_TRITON_ROPE": "1",
"VLLM_ROCM_USE_AITER_RMSNORM": "1",
"VLLM_ROCM_USE_AITER_TRITON_LINEAR": "1",
"VLLM_ROCM_QUICK_REDUCE_QUANTIZATION": "INT4"
},
"result_dict": {
"output_throughput_per_sec": "9000",
"mean_ttft_ms": "250",
"mean_tpot_ms": "11"
"ISL=1024,OSL=1024,TP=8,CONC=16": {
"total_throughput_per_sec": "1944",
"mean_ttft_ms": "280",
"mean_tpot_ms": "12"
},
"ISL=1024,OSL=1024,TP=8,CONC=32": {
"total_throughput_per_sec": "2939",
"mean_ttft_ms": "260",
"mean_tpot_ms": "11"
},
"ISL=1024,OSL=1024,TP=8,CONC=64": {
"total_throughput_per_sec": "4834",
"mean_ttft_ms": "250",
"mean_tpot_ms": "11"
},
"ISL=1024,OSL=8192,TP=8,CONC=16": {
"total_throughput_per_sec": "1109",
"mean_ttft_ms": "320",
"mean_tpot_ms": "16"
},
"ISL=1024,OSL=8192,TP=8,CONC=32": {
"total_throughput_per_sec": "1676",
"mean_ttft_ms": "300",
"mean_tpot_ms": "14"
},
"ISL=1024,OSL=8192,TP=8,CONC=64": {
"total_throughput_per_sec": "2716",
"mean_ttft_ms": "280",
"mean_tpot_ms": "13"
}
}
}
}
Expand Down
Loading
Loading