| { | |
| "model": { | |
| "type": "nemotron_speech", | |
| "vocab_size": 13088, | |
| "num_mels": 128, | |
| "fft_size": 512, | |
| "hop_length": 160, | |
| "win_length": 400, | |
| "preemph": 0.97, | |
| "log_eps": 5.96046448e-08, | |
| "subsampling_factor": 8, | |
| "left_context": 140, | |
| "conv_context": 8, | |
| "pre_encode_cache_size": 9, | |
| "sample_rate": 16000, | |
| "chunk_samples": 17920, | |
| "blank_id": 13087, | |
| "max_symbols_per_step": 10, | |
| "encoder": { | |
| "filename": "encoder.onnx", | |
| "hidden_size": 1024, | |
| "num_hidden_layers": 24, | |
| "inputs": { | |
| "audio_features": "audio_signal", | |
| "input_lengths": "length", | |
| "cache_last_channel": "cache_last_channel", | |
| "cache_last_time": "cache_last_time", | |
| "cache_last_channel_len": "cache_last_channel_len", | |
| "lang_id": "lang_id" | |
| }, | |
| "outputs": { | |
| "encoder_outputs": "outputs", | |
| "output_lengths": "encoded_lengths", | |
| "cache_last_channel_next": "cache_last_channel_next", | |
| "cache_last_time_next": "cache_last_time_next", | |
| "cache_last_channel_len_next": "cache_last_channel_len_next" | |
| } | |
| }, | |
| "decoder": { | |
| "filename": "decoder.onnx", | |
| "hidden_size": 640, | |
| "num_hidden_layers": 2, | |
| "inputs": { | |
| "targets": "targets", | |
| "lstm_hidden_state": "h_in", | |
| "lstm_cell_state": "c_in" | |
| }, | |
| "outputs": { | |
| "outputs": "decoder_output", | |
| "lstm_hidden_state": "h_out", | |
| "lstm_cell_state": "c_out" | |
| } | |
| }, | |
| "joiner": { | |
| "filename": "joint.onnx", | |
| "inputs": { | |
| "encoder_outputs": "encoder_output", | |
| "decoder_outputs": "decoder_output" | |
| }, | |
| "outputs": { | |
| "logits": "joint_output" | |
| } | |
| }, | |
| "vad": { | |
| "filename": "silero_vad.onnx", | |
| "threshold": 0.3, | |
| "silence_duration_ms": 3360, | |
| "prefix_padding_ms": 560 | |
| } | |
| } | |
| } |