| { | |
| "ssl_in_16k": 89600, | |
| "sample_rate": 44100, | |
| "ssl_sample_rate": 16000, | |
| "wavlm_hop": 320, | |
| "ssl_dim": 768, | |
| "token_hz": 25, | |
| "token_samples": 1764, | |
| "downsample_factor": 2, | |
| "n_fft": 392, | |
| "hop_length": 98, | |
| "win_length": 392, | |
| "n_bins": 197, | |
| "istft_pad": 147, | |
| "wave_upsample_factor": 2, | |
| "upsampler_total": 9, | |
| "frames_per_tok": 18, | |
| "wave_interpolation_mode": "linear", | |
| "normalize_ssl_features": true, | |
| "global_dim": 128, | |
| "chunk": 6, | |
| "enc_left": 128, | |
| "enc_right": 6, | |
| "dec_left": 128, | |
| "dec_right": 6, | |
| "enc_tokens": 140, | |
| "dec_tokens": 140, | |
| "enc_ssl_frames": 279, | |
| "dec_stft_frames": 2520 | |
| } |