pupuvocoder_large/args.json

{
    "base_config": "egs/afgen/exp_config_afvocoder_base.json",
    "dataset": [
        "afgen_all_data",
    ],
    "exp_name": "afvocoder_large",
    "inference": {
        "batch_size": 1,
    },
    "log_dir": "/home/vocod/experiments/afgen",
    "model": {
        "afvocoder": {
            "resblock": "1",
            "resblock_dilation_sizes": [
                [
                    1,
                    3,
                    5,
                ],
                [
                    1,
                    3,
                    5,
                ],
                [
                    1,
                    3,
                    5,
                ],
            ],
            "resblock_kernel_sizes": [
                3,
                7,
                11,
            ],
            "upsample_initial_channel": 1536,
            "upsample_kernel_sizes": [
                16,
                16,
                4,
                4,
                4,
            ],
            "upsample_rates": [
                8,
                8,
                2,
                2,
                2,
            ],
        },
        "discriminators": [
            "msd",
            "mpd",
            "mssbstftd",
            "mssbcqtd",
        ],
        "generator": "afvocoder",
        "mpd": {
            "discriminator_channel_mult_factor": 1,
            "mpd_reshapes": [
                2,
                3,
                5,
                7,
                11,
                17,
                23,
                37,
            ],
            "use_spectral_norm": false,
        },
        "mssbcqtd": {
            "bins_per_octaves": [
                24,
                36,
                48,
            ],
            "dilations": [
                1,
                2,
                4,
            ],
            "filters": 32,
            "filters_scale": 1,
            "hop_lengths": [
                1024,
                512,
                512,
            ],
            "in_channels": 1,
            "max_filters": 1024,
            "n_octaves": [
                10,
                10,
                10,
            ],
            "out_channels": 1,
        },
        "msstftd": {
            "filters": 32,
        },
    },
    "model_type": "AFGenVocoder",
    "preprocess": {
        "align_mel_duration": false,
        "audio_dir": "audios",
        "bits": 8,
        "contentvec_dir": "contentvec",
        "cut_mel_frame": 128,
        "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": false,
        "extract_mcep": false,
        "extract_mel": false,
        "extract_mert_feature": false,
        "extract_one_hot": false,
        "extract_phone": false,
        "extract_pitch": true,
        "extract_uv": true,
        "extract_wenet_feature": false,
        "extract_whisper_feature": false,
        "f0_max": 1975.5,
        "f0_min": 32.7,
        "file_lst": "file.lst",
        "fmax": 22050,
        "fmin": 0,
        "hop_size": 512,
        "imaginary_dir": "imaginarys",
        "is_mu_law": false,
        "lab_dir": "labs",
        "label_dir": "labels",
        "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": 2048,
        "n_mel": 128,
        "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_bin": 256,
        "pitch_dir": "pitches",
        "pitch_extractor": "rmvpe",
        "pitch_max": 1100.0,
        "pitch_min": 50.0,
        "pitch_norm": false,
        "pitch_remove_outlier": false,
        "processed_dir": "data/",
        "raw_data": "raw_data",
        "real_dir": "reals",
        "ref_level_db": 20,
        "sample_rate": 44100,
        "spk2id": "singers.json",
        "symbols_dict": "symbols.dict",
        "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": true,
        "use_lab": false,
        "use_label": false,
        "use_linear": false,
        "use_log_scale_energy": false,
        "use_log_scale_pitch": false,
        "use_mel": false,
        "use_min_max_norm_mel": false,
        "use_one_hot": false,
        "use_phn_seq": false,
        "use_phone": false,
        "use_phone_duration": false,
        "use_phone_energy": false,
        "use_phone_pitch": false,
        "use_spkid": false,
        "use_text": false,
        "use_uv": true,
        "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": 2048,
    },
    "supported_model_type": [
        "GANVocoder",
        "Fastspeech2",
        "DiffSVC",
        "Transformer",
        "EDM",
        "Autotune",
        "CD",
    ],
    "task_type": "afgen",
    "train": {
        "adamw": {
            "adam_b1": 0.8,
            "adam_b2": 0.99,
            "lr": 0.0001,
        },
        "batch_size": 4,
        "criterions": [
            "feature",
            "discriminator",
            "generator",
            "multimel",
        ],
        "dataloader": {
            "num_worker": 32,
            "persistent_workers": true,
            "pin_memory": true,
            "prefetch_factor": 4,
        },
        "ddp": true,
        "exponential_lr": {
            "lr_decay": 0.9999996,
        },
        "freeze_step": 0,
        "gradient_accumulation_step": 1,
        "keep_checkpoint_max": 5,
        "keep_last": [
            3,
            -1,
        ],
        "max_epoch": 1000000,
        "max_steps": 1000000,
        "multi_speaker_training": false,
        "optimizer": "AdamW",
        "random_seed": 114514,
        "reducelronplateau": {
            "factor": 0.8,
            "min_lr": 0.0001,
            "patience": 10,
        },
        "run_eval": [
            true,
        ],
        "sampler": {
            "drop_last": true,
            "holistic_shuffle": true,
        },
        "save_checkpoint_stride": [
            1,
        ],
        "save_checkpoints_steps": 10000,
        "save_summary_steps": 500,
        "scheduler": "ReduceLROnPlateau",
        "total_training_steps": 50000,
        "tracker": [
            "tensorboard",
        ],
        "valid_interval": 10000,
    },
    "use_custom_dataset": [],
}