[Perf] Optimize Higgs Audio v3 serving (#4204)

Signed-off-by: Sy03 <1370724210@qq.com>
Signed-off-by: Yueqian Lin <linyueqian@outlook.com>
Co-authored-by: Yueqian Lin <linyueqian@outlook.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>
This commit is contained in:
authored and GitHub committed 2026-06-11 13:29:31 -07:00
1 parent eb964dcec5
commit e27fd46ab9
21 files changed
+1976 -734

No files matched your search

@@ -0,0 +1,74 @@
[
{
"test_name": "test_higgs_audio_v3",
"server_params": {
"model": "bosonai/higgs-audio-v3-tts-4b",
"stage_overrides": {
"0": {
"gpu_memory_utilization": 0.3
},
"1": {
"gpu_memory_utilization": 0.2
}
},
"extra_cli_args": [
"--deploy-config",
"vllm_omni/deploy/higgs_multimodal_qwen3.yaml",
"--trust-remote-code"
]
},
"benchmark_params": [
{
"task": "default_voice",
"eval_phase": "latency",
"dataset_name": "seed-tts-text",
"backend": "openai-audio-speech",
"endpoint": "/v1/audio/speech",
"dataset_path": "benchmarks/build_dataset/seed_tts_smoke",
"num_prompts": [
20
],
"max_concurrency": [
1
],
"seed_tts_locale": "en",
"percentile-metrics": "ttft,e2el,audio_rtf,audio_ttfp,audio_duration",
"baseline": {
"median_audio_ttfp_ms": [
3000
],
"median_audio_rtf": [
0.5
]
}
},
{
"task": "default_voice",
"eval_phase": "throughput",
"dataset_name": "seed-tts-text",
"backend": "openai-audio-speech",
"endpoint": "/v1/audio/speech",
"dataset_path": "benchmarks/build_dataset/seed_tts_smoke",
"num_prompts": [
500
],
"max_concurrency": [
16
],
"seed_tts_locale": "en",
"percentile-metrics": "ttft,e2el,audio_rtf,audio_ttfp,audio_duration",
"baseline": {
"median_audio_ttfp_ms": [
5000
],
"median_audio_rtf": [
1.2
],
"audio_throughput": [
30
]
}
}
]
}
]