| { | |
| "train": { | |
| "log_interval": 200, | |
| "eval_interval": 2000, | |
| "seed": 1234, | |
| "epochs": 10000, | |
| "learning_rate": 0.0001, | |
| "betas": [ | |
| 0.8, | |
| 0.99 | |
| ], | |
| "eps": 1e-09, | |
| "batch_size": 8, | |
| "fp16_run": false, | |
| "lr_decay": 0.999875, | |
| "segment_size": 20480, | |
| "init_lr_ratio": 1, | |
| "warmup_epochs": 0, | |
| "c_mel": 45, | |
| "c_kl": 1.0 | |
| }, | |
| "data": { | |
| "training_files": "dump/train_files_gs.list", | |
| "validation_files": "dump/val_files_gs.list", | |
| "max_wav_value": 32768.0, | |
| "sampling_rate": 32000, | |
| "filter_length": 2048, | |
| "hop_length": 640, | |
| "win_length": 2048, | |
| "n_mel_channels": 128, | |
| "mel_fmin": 0.0, | |
| "mel_fmax": null, | |
| "add_blank": true, | |
| "n_speakers": 5500, | |
| "cleaned_text": true | |
| }, | |
| "model": { | |
| "inter_channels": 192, | |
| "hidden_channels": 192, | |
| "filter_channels": 768, | |
| "n_heads": 2, | |
| "n_layers": 6, | |
| "kernel_size": 3, | |
| "p_dropout": 0.1, | |
| "resblock": "1", | |
| "resblock_kernel_sizes": [ | |
| 3, | |
| 7, | |
| 11 | |
| ], | |
| "resblock_dilation_sizes": [ | |
| [ | |
| 1, | |
| 3, | |
| 5 | |
| ], | |
| [ | |
| 1, | |
| 3, | |
| 5 | |
| ], | |
| [ | |
| 1, | |
| 3, | |
| 5 | |
| ] | |
| ], | |
| "upsample_rates": [ | |
| 10, | |
| 8, | |
| 2, | |
| 2, | |
| 2 | |
| ], | |
| "upsample_initial_channel": 512, | |
| "upsample_kernel_sizes": [ | |
| 16, | |
| 16, | |
| 8, | |
| 2, | |
| 2 | |
| ], | |
| "n_layers_q": 3, | |
| "use_spectral_norm": false, | |
| "gin_channels": 192, | |
| "semantic_frame_rate": "16hz", | |
| "freeze_quantizer": false, | |
| "use_reference_enc": true, | |
| "prosody_layers": 12 | |
| }, | |
| "s2_ckpt_dir": "logs/fsqhifigan", | |
| "content_module": "whisper" | |
| } |