| { | |
| "format_version": 1, | |
| "type": "s2", | |
| "model": { | |
| "model_params": { | |
| "style_encoder": { | |
| "dim": 256 | |
| }, | |
| "DiT": { | |
| "hidden_dim": 512, | |
| "num_heads": 8, | |
| "depth": 13, | |
| "class_dropout_prob": 0.1, | |
| "in_channels": 32, | |
| "content_dim": 32 | |
| }, | |
| "wavenet": { | |
| "hidden_dim": 512, | |
| "num_layers": 8, | |
| "kernel_size": 5, | |
| "dilation_rate": 1, | |
| "p_dropout": 0.2 | |
| } | |
| }, | |
| "conditioner": { | |
| "n_vocab": 178, | |
| "output_channels": 32, | |
| "ssl_dim": 768, | |
| "hidden_channels": 192, | |
| "prosody_dim": 16, | |
| "duration_bins": 512, | |
| "filter_channels": 768, | |
| "n_heads": 2, | |
| "n_layers": 6, | |
| "kernel_size": 3, | |
| "p_dropout": 0.1, | |
| "prenet": true, | |
| "duration_loss_weight": 1.0, | |
| "max_token_duration": 511, | |
| "max_infer_frames": 1000 | |
| }, | |
| "prompt": { | |
| "min_seconds": 1.0, | |
| "max_seconds": 5.0, | |
| "max_fraction": 0.5, | |
| "dropout": 0.1 | |
| } | |
| } | |
| } | |