| { | |
| "sample_rate": 16000, | |
| "window_seconds": 15.0, | |
| "n_samples": 240000, | |
| "hop_length": 160, | |
| "n_rows": 1503, | |
| "n_mels": 128, | |
| "valid_frames": 1500, | |
| "enc_frames": 188, | |
| "pred_hidden": 640, | |
| "pred_layers": 2, | |
| "joint_hidden": 640, | |
| "n_logits": 8198, | |
| "vocab_size": 8192, | |
| "blank_idx": 8192, | |
| "num_durations": 5, | |
| "durations": [ | |
| 0, | |
| 1, | |
| 2, | |
| 3, | |
| 4 | |
| ], | |
| "precision": "float16", | |
| "io_precision": "float16", | |
| "deployment_target": "iOS18", | |
| "encoder_chunks": 2, | |
| "decode_width": 8, | |
| "n_fft": 512, | |
| "preemph": 0.97, | |
| "n_padded_samples": 240480 | |
| } |