| backbone: | |
| class_path: vocos.models.VocosBackbone | |
| init_args: | |
| adanorm_num_embeddings: null | |
| dim: 384 | |
| input_channels: 1024 | |
| intermediate_dim: 1152 | |
| layer_scale_init_value: null | |
| num_layers: 8 | |
| decay_mel_coeff: false | |
| evaluate_periodicty: true | |
| evaluate_pesq: true | |
| evaluate_utmos: true | |
| feature_extractor: | |
| class_path: vocos.feature_extractors.DACFeatures | |
| head: | |
| class_path: vocos.heads.ISTFTHead | |
| init_args: | |
| dim: 384 | |
| hop_length: 320 | |
| n_fft: 1280 | |
| padding: same | |
| initial_learning_rate: 0.0005 | |
| mel_loss_coeff: 45.0 | |
| mrd_loss_coeff: 1.0 | |
| num_warmup_steps: 0 | |
| pretrain_mel_steps: 0 | |
| sample_rate: 16000 | |