| { | |
| "base_config": "egs/vocoder/diffusion/exp_config_base.json", | |
| "preprocess": { | |
| // acoustic features | |
| "extract_mel": true, | |
| "extract_audio": true, | |
| // Features used for model training | |
| "use_mel": true, | |
| "use_audio": true, | |
| }, | |
| "model": { | |
| "generator": "diffwave", | |
| "diffwave": { | |
| "residual_channels": 64, | |
| "residual_layers": 30, | |
| "dilation_cycle_length": 10, | |
| "noise_schedule_factors": [1.0e-4, 0.05, 50], | |
| "inference_noise_schedule": [0.0001, 0.001, 0.01, 0.05, 0.2, 0.5], | |
| "upsample_factors": [16, 16], | |
| } | |
| }, | |
| "inference": { | |
| "batch_size": 1, | |
| } | |
| } | |