| { | |
| "format": "wakewords-fastconformer-v1", | |
| "features": { | |
| "sample_rate": 16000, | |
| "window_seconds": 2.0, | |
| "window_samples": 32000, | |
| "channels": "mean_to_mono", | |
| "resampler": "soxr_HQ", | |
| "n_mels": 128, | |
| "mel_frames": 200, | |
| "n_fft": 512, | |
| "window_samples_stft": 400, | |
| "hop_samples": 160, | |
| "window": "periodic_hann", | |
| "center": true, | |
| "pad_mode": "constant", | |
| "mel_scale": "slaney", | |
| "f_min": 0, | |
| "f_max": 8000, | |
| "power": 2, | |
| "log_floor": 1e-10, | |
| "normalization": "per_feature_over_full_two_second_window", | |
| "normalization_epsilon": 1e-05, | |
| "subsampling_factor": 8 | |
| }, | |
| "model": { | |
| "wakeword": "hey native", | |
| "threshold": 0.5, | |
| "backbone": "repcnn", | |
| "frontend": { | |
| "dim": 64, | |
| "depth": 3, | |
| "heads": 4, | |
| "feedforward_dim": 128, | |
| "subsampling_channels": 16, | |
| "conv_kernel": 9, | |
| "dropout": 0.1 | |
| }, | |
| "repcnn": { | |
| "channels": [ | |
| 48, | |
| 64, | |
| 96 | |
| ], | |
| "dim": 96, | |
| "dropout": 0.1 | |
| } | |
| } | |
| } | |