{ "wandb": { "mode": "disabled", "checkpoint": "checkpoint_name", "fork_checkpoint": false, "project": "project_name", "wandb_prefix": "wandb_prefix", "dir": "wandb_dir" }, "local": { "checkpoint_model": null, "checkpoint_discriminator": null }, "checkpoint_dir": "path/to/checkpoints", "seed": 123, "learning_rate": 0.0001, "dataset": "Jamendo, LibriTTS", "train_datafile": "data/train.txt", "validation_datafile": "data/val_short.txt", "n_epochs": 500000, "step_checkpoint": 10000, "step_media_log": 2000, "batch_grad_log": 2000, "batch_size": 32, "sample_rate": 44100, "exp_gamma": 0.9995, "adam_b1": 0.8, "adam_b2": 0.99, "segment_size": 16384, "segment_size_val": 262144, "n_cache_reuse": 30, "num_workers": 4, "prefetch_factor": 2, "use_discriminator": true, "unfreeze": { "steps": 100000, "loss_multiplier": { "loss_z": 0, "loss_ms_mel": 15, "loss_ms_stft": 1, "loss_adv_gen": 1, "loss_adv_feat": 2, "loss_waveform": 1 } }, "model": { "latent_dim": 1024, "n_codebooks": 9, "codebook_dim": 8, "codebook_size": 1024, "n_resblocks": 6, "initial_out_channels": 1024, "intermediate_dim": 1152, "resblock_type": "AMP", "resblock_kernel_sizes": [3, 3, 7, 7, 11, 11], "resblock_dilations": [[1, 3, 5], [1, 3, 5], [1, 3, 5], [1, 3, 5], [1, 3, 5], [1, 3, 5]], "predict_type": "z", "activation": "snake" }, "disc": { "resolutions": [[1024, 120, 600], [2048, 240, 1200], [512, 50, 240]], "periods": [2, 3, 5, 7, 11], "discriminator_channel_mult": 1, "use_spectral_norm": false }, "mel": { "n_fft": 1024, "win_length": 1024, "hop_length": 256, "f_min": 0, "f_max": null, "n_mels": 128 }, "loss_multiplier": { "loss_z": 15, "loss_ms_mel": 15, "loss_ms_stft": 1, "loss_adv_gen": 1, "loss_adv_feat": 2, "loss_waveform": 1 }, "backend": { "master_port": 12359 } }