| { | |
| "base_config": "config/vits.json", | |
| "dataset": [ | |
| "hifitts", | |
| ], | |
| "model": { | |
| "filter_channels": 768, | |
| "gin_channels": 256, | |
| "hidden_channels": 192, | |
| "inter_channels": 192, | |
| "kernel_size": 3, | |
| "n_heads": 2, | |
| "n_layers": 6, | |
| "n_layers_q": 3, | |
| "n_speakers": 10, | |
| "p_dropout": 0.1, | |
| "resblock": "1", | |
| "resblock_dilation_sizes": [ | |
| [ | |
| 1, | |
| 3, | |
| 5, | |
| ], | |
| [ | |
| 1, | |
| 3, | |
| 5, | |
| ], | |
| [ | |
| 1, | |
| 3, | |
| 5, | |
| ], | |
| ], | |
| "resblock_kernel_sizes": [ | |
| 3, | |
| 7, | |
| 11, | |
| ], | |
| "text_token_num": 512, | |
| "upsample_initial_channel": 512, | |
| "upsample_kernel_sizes": [ | |
| 16, | |
| 16, | |
| 4, | |
| 4, | |
| ], | |
| "upsample_rates": [ | |
| 8, | |
| 8, | |
| 2, | |
| 2, | |
| ], | |
| "use_sdp": true, | |
| "use_spectral_norm": false, | |
| }, | |
| "model_type": "VITS", | |
| "preprocess": { | |
| "add_blank": true, | |
| "align_mel_duration": false, | |
| "audio_dir": "audios", | |
| "bits": 8, | |
| "contentvec_dir": "contentvec", | |
| "data_augment": false, | |
| "dur_dir": "durs", | |
| "duration_dir": "duration", | |
| "emo2id": "emo2id.json", | |
| "energy_dir": "energys", | |
| "energy_extract_mode": "from_mel", | |
| "energy_norm": false, | |
| "energy_remove_outlier": false, | |
| "extract_acoustic_token": false, | |
| "extract_amplitude_phase": false, | |
| "extract_audio": true, | |
| "extract_contentvec_feature": false, | |
| "extract_duration": false, | |
| "extract_energy": false, | |
| "extract_label": false, | |
| "extract_linear_spec": true, | |
| "extract_mcep": false, | |
| "extract_mel": true, | |
| "extract_mert_feature": false, | |
| "extract_phone": true, | |
| "extract_pitch": false, | |
| "extract_uv": false, | |
| "extract_wenet_feature": false, | |
| "extract_whisper_feature": false, | |
| "file_lst": "file.lst", | |
| "fmax": null, | |
| "fmin": 0, | |
| "hop_size": 256, | |
| "imaginary_dir": "imaginarys", | |
| "lab_dir": "labs", | |
| "label_dir": "labels", | |
| "language": "en-us", | |
| "lexicon_path": "./text/lexicon/librispeech-lexicon.txt", | |
| "linear_dir": "linears", | |
| "log_amplitude_dir": "log_amplitudes", | |
| "mcep_dir": "mcep", | |
| "mel_dir": "mels", | |
| "mel_extract_mode": "", | |
| "mel_min_max_norm": false, | |
| "min_level_db": -115, | |
| "n_fft": 1024, | |
| "n_mel": 80, | |
| "num_silent_frames": 8, | |
| "phase_dir": "phases", | |
| "phone_dir": "phones", | |
| "phone_energy_dir": "phone_energys", | |
| "phone_extractor": "espeak", | |
| "phone_pitch_dir": "phone_pitches", | |
| "phone_seq_file": "phone_seq_file", | |
| "pitch_dir": "pitches", | |
| "pitch_extractor": "parselmouth", | |
| "pitch_norm": false, | |
| "pitch_remove_outlier": false, | |
| "raw_data": "raw_data", | |
| "real_dir": "reals", | |
| "ref_level_db": 20, | |
| "sample_rate": 24000, | |
| "segment_size": 8192, | |
| "spk2id": "spk2id.json", | |
| "symbols_dict": "symbols.dict", | |
| "text_cleaners": [ | |
| "english_cleaners", | |
| ], | |
| "train_file": "train.json", | |
| "trim_fft_size": 512, | |
| "trim_hop_size": 128, | |
| "trim_silence": false, | |
| "trim_top_db": 30, | |
| "trimmed_wav_dir": "trimmed_wavs", | |
| "use_amplitude_phase": false, | |
| "use_audio": true, | |
| "use_dur": false, | |
| "use_emoid": false, | |
| "use_frame_duration": false, | |
| "use_frame_energy": false, | |
| "use_frame_pitch": false, | |
| "use_lab": false, | |
| "use_label": false, | |
| "use_linear": true, | |
| "use_log_scale_energy": false, | |
| "use_log_scale_pitch": false, | |
| "use_mel": true, | |
| "use_min_max_norm_mel": false, | |
| "use_one_hot": false, | |
| "use_phn_seq": false, | |
| "use_phone": true, | |
| "use_phone_duration": false, | |
| "use_phone_energy": false, | |
| "use_phone_pitch": false, | |
| "use_spkid": true, | |
| "use_text": false, | |
| "use_uv": false, | |
| "use_wav": false, | |
| "use_wenet": false, | |
| "utt2emo": "utt2emo", | |
| "utt2spk": "utt2spk", | |
| "uv_dir": "uvs", | |
| "valid_file": "valid.json", | |
| "wav_dir": "wavs", | |
| "wenet_dir": "wenet", | |
| "win_size": 1024, | |
| }, | |
| "supported_model_type": [ | |
| "Fastspeech2", | |
| "VITS", | |
| "VALLE", | |
| ], | |
| "task_type": "tts", | |
| "train": { | |
| "AdamW": { | |
| "betas": [ | |
| 0.8, | |
| 0.99, | |
| ], | |
| "eps": 1e-09, | |
| }, | |
| "adamw": { | |
| "lr": 0.0004, | |
| }, | |
| "batch_size": 16, | |
| "betas": [ | |
| 0.8, | |
| 0.99, | |
| ], | |
| "c_kl": 1.0, | |
| "c_mel": 45, | |
| "dataloader": { | |
| "num_worker": 32, | |
| "pin_memory": true, | |
| }, | |
| "ddp": true, | |
| "eps": 1e-09, | |
| "fp16_run": true, | |
| "gradient_accumulation_step": 1, | |
| "init_lr_ratio": 1, | |
| "keep_checkpoint_max": 5, | |
| "keep_last": [ | |
| 3, | |
| -1, | |
| ], | |
| "learning_rate": 0.0002, | |
| "lr_decay": 0.999875, | |
| "max_epoch": -1, | |
| "max_steps": 1000000, | |
| "multi_speaker_training": true, | |
| "optimizer": "AdamW", | |
| "random_seed": 10086, | |
| "reducelronplateau": { | |
| "factor": 0.8, | |
| "min_lr": 0.0001, | |
| "patience": 10, | |
| }, | |
| "run_eval": [ | |
| false, | |
| true, | |
| ], | |
| "sampler": { | |
| "drop_last": true, | |
| "holistic_shuffle": true, | |
| }, | |
| "save_checkpoint_stride": [ | |
| 5, | |
| 20, | |
| ], | |
| "save_checkpoints_steps": 10000, | |
| "save_summary_steps": 500, | |
| "scheduler": "ReduceLROnPlateau", | |
| "total_training_steps": 50000, | |
| "tracker": [ | |
| "tensorboard", | |
| ], | |
| "valid_interval": 10000, | |
| "warmup_epochs": 0, | |
| }, | |
| "use_custom_dataset": false, | |
| } |