EnCodec/1024/base/config.json

{
    "n_filters": 64,
    "strides": [
        8,
        5,
        4,
        2
    ],
    "dimension": 512,
    "semantic_dimension": 768,
    "bidirectional": true,
    "dilation_base": 2,
    "residual_kernel_size": 3,
    "n_residual_layers": 1,
    "lstm_layers": 2,
    "activation": "ELU",
    "codebook_size": 1024,
    "n_q": 8,
    "num_mels": 80,
    "num_freq": 1025,
    "n_fft": 1024,
    "hop_size": 240,
    "win_size": 1024,
    "fmin": 0,
    "fmax": 8000,
    "fmax_for_loss": null,
    "mel_loss_lambdas": [
        45,
        1,
        1,
        1
    ],
    "recon_loss_lambda": 500,
    "commitment_loss_lambda": 10,
    "distill_loss_lambda": 120,
    "distill_type": "d_axis",
    "semantic_model_path": "/remote-home/share/models/hubert/ls_960/hubert-base-ls960",
    "semantic_model_layer": "avg",
    "train_files": "train_file_list.txt",
    "valid_files": "dev_file_list.txt",
    "results_folder": "Log/spt_base",
    "sample_rate": 16000,
    "batch_size": 8,
    "epochs": 1000,
    "learning_rate": 0.0001,
    "intial_learning_rate": 0.0001,
    "num_warmup_steps": 0,
    "betas": [
        0.9,
        0.99
    ],
    "adam_b2": 0.9,
    "lr_decay": 0.98,
    "seed": 1234,
    "segment_size": 48000,
    "wd": 0,
    "num_workers": 8,
    "log_steps": 100,
    "stdout_steps": 10,
    "save_model_steps": 1000,
    "num_ckpt_keep": 6,
    "showpiece_num": 8
}