| { | |
| "model_type": "silero_vad_v5", | |
| "sample_rate": 16000, | |
| "chunk_size": 512, | |
| "context_size": 64, | |
| "filter_length": 256, | |
| "hop_length": 128, | |
| "encoder_channels": [ | |
| 129, | |
| 128, | |
| 64, | |
| 64, | |
| 128 | |
| ], | |
| "encoder_kernel_sizes": [ | |
| 3, | |
| 3, | |
| 3, | |
| 3 | |
| ], | |
| "encoder_strides": [ | |
| 1, | |
| 2, | |
| 2, | |
| 1 | |
| ], | |
| "lstm_hidden_size": 128, | |
| "lstm_num_layers": 1 | |
| } |