| { | |
| "model_name": "campplus", | |
| "model_id": "iic/speech_campplus_sv_zh_en_16k-common_advanced", | |
| "revision": "v1.0.0", | |
| "framework": "pytorch", | |
| "inputs": [ | |
| { | |
| "name": "feature", | |
| "shape": [1, 360, 80], | |
| "dtype": "float32", | |
| "layout": "NTC" | |
| } | |
| ], | |
| "outputs": [ | |
| { | |
| "name": "embedding", | |
| "shape": [1, 192], | |
| "dtype": "float32" | |
| } | |
| ], | |
| "feature": { | |
| "type": "kaldi_fbank", | |
| "num_mel_bins": 80, | |
| "sample_rate": 16000, | |
| "dither": 0, | |
| "mean_nor": true, | |
| "frames": 360, | |
| "min_wav_len": 57900, | |
| "chunk": { | |
| "dur": 1.5, | |
| "step": 0.75 | |
| } | |
| }, | |
| "quant": { | |
| "target_hardware": "AX650", | |
| "npu_mode": "NPU3", | |
| "data_type": "U16", | |
| "smooth_quant": true, | |
| "calibration": "30 real speech fbank chunks (5 wavs)", | |
| "embedding_cosine": 0.99569, | |
| "macs": 4064810496, | |
| "model_size_bytes": 10748172 | |
| } | |
| } | |