config.json

{   
    "diffusion_config":{
        "T": 200,
        "beta_0": 0.0001,
        "beta_T": 0.02
    },
    "wavenet_config": {
        "in_channels": 1, 
        "res_channels": 256, 
        "skip_channels": 256, 
        "out_channels": 1, 
        "num_res_layers": 36, 
        "dilation_cycle": 12, 
        "diffusion_step_embed_dim_in": 128,
        "diffusion_step_embed_dim_mid": 512,
        "diffusion_step_embed_dim_out": 512
    },
    "train_config": {
        "output_directory": "logs/checkpoint",
        "tensorboard_directory": "logs/tensorboard",
        "ckpt_iter": "max",
        "iters_per_ckpt": 10000,
        "iters_per_logging": 100,
        "n_iters": 1000001,
        "learning_rate": 2e-4,
        "batch_size_per_gpu": 2
    },
    "trainset_config":{
        "data_path": "/tmp2/SC09/train",
        "segment_length": 16000,
        "sampling_rate": 16000
    },
    "gen_config":{
        "tensorboard_directory": "logs/tensorboard",
        "output_directory": "speeches",
        "ckpt_path": "logs/checkpoint"
    },
    "dist_config": {
        "dist_backend": "nccl",
        "dist_url": "tcp://localhost:54321"
    }
}