85 lines
3.1 KiB
JSON
85 lines
3.1 KiB
JSON
{
|
||
"default_model": "Qwen/Qwen3-ASR-0.6B",
|
||
"models": {
|
||
"Qwen/Qwen3-ASR-1.7B": {
|
||
"alias": "1.7b",
|
||
"directory": "Qwen/Qwen3-ASR-1.7B",
|
||
"description": "Qwen3-ASR 1.7B，GPU 可选大模型"
|
||
},
|
||
"Qwen/Qwen3-ASR-0.6B": {
|
||
"alias": "0.6b",
|
||
"directory": "Qwen/Qwen3-ASR-0.6B",
|
||
"description": "Qwen3-ASR 0.6B，GPU 默认轻量模型"
|
||
}
|
||
},
|
||
"auxiliary_models": {
|
||
"damo/speech_fsmn_vad_zh-cn-16k-common-pytorch": {
|
||
"alias": "vad",
|
||
"directory": "damo/speech_fsmn_vad_zh-cn-16k-common-pytorch",
|
||
"kind": "vad",
|
||
"description": "FunASR FSMN VAD",
|
||
"revision": "v2.0.2",
|
||
"required_files": ["configuration.json", "config.yaml", "model.pb"],
|
||
"min_total_size_bytes": 1000000
|
||
},
|
||
"iic/speech_campplus_speaker-diarization_common": {
|
||
"alias": "diarization",
|
||
"directory": "iic/speech_campplus_speaker-diarization_common",
|
||
"kind": "diarization",
|
||
"description": "CAM++ speaker diarization",
|
||
"required_files": [
|
||
"configuration.json",
|
||
"config.yaml",
|
||
"onnx/asd.onnx",
|
||
"onnx/face_recog_ir101.onnx",
|
||
"onnx/fqa.onnx",
|
||
"onnx/version-RFB-320.onnx"
|
||
],
|
||
"min_total_size_bytes": 50000000
|
||
},
|
||
"iic/speech_campplus_sv_zh-cn_16k-common": {
|
||
"alias": "speaker",
|
||
"directory": "iic/speech_campplus_sv_zh-cn_16k-common",
|
||
"kind": "speaker_verification",
|
||
"description": "Configured CAM++ speaker verification",
|
||
"revision": "v2.0.2",
|
||
"required_files": ["configuration.json", "config.yaml", "campplus_cn_common.bin"],
|
||
"min_total_size_bytes": 10000000
|
||
},
|
||
"iic/speech_eres2netv2_sv_zh-cn_16k-common": {
|
||
"alias": "realtime-speaker",
|
||
"directory": "iic/speech_eres2netv2_sv_zh-cn_16k-common",
|
||
"kind": "realtime_speaker_verification",
|
||
"description": "Realtime speaker verification",
|
||
"required_files": ["configuration.json"],
|
||
"any_files": ["*"],
|
||
"min_total_size_bytes": 10000000
|
||
},
|
||
"damo/speech_campplus_sv_zh-cn_16k-common": {
|
||
"alias": "speaker-damo",
|
||
"directory": "damo/speech_campplus_sv_zh-cn_16k-common",
|
||
"kind": "speaker_verification",
|
||
"description": "CAM++ speaker verification dependency",
|
||
"required_files": ["configuration.json", "config.yaml", "campplus_cn_common.bin"],
|
||
"min_total_size_bytes": 10000000
|
||
},
|
||
"damo/speech_campplus-transformer_scl_zh-cn_16k-common": {
|
||
"alias": "speaker-transformer",
|
||
"directory": "damo/speech_campplus-transformer_scl_zh-cn_16k-common",
|
||
"kind": "speaker_transformer",
|
||
"description": "CAM++ Transformer dependency",
|
||
"required_files": ["configuration.json", "campplus_cn_encoder.pt", "transformer_backend.pt"],
|
||
"min_total_size_bytes": 10000000
|
||
},
|
||
"Qwen/Qwen3-ForcedAligner-0.6B": {
|
||
"alias": "forced-aligner",
|
||
"directory": "Qwen/Qwen3-ForcedAligner-0.6B",
|
||
"kind": "forced_aligner",
|
||
"description": "Qwen3 word-level forced aligner",
|
||
"required_files": ["config.json"],
|
||
"any_files": ["*.safetensors", "*.bin"],
|
||
"min_total_size_bytes": 500000000
|
||
}
|
||
}
|
||
}
|