fftnet_config.json

{
  "name": "fftnet-vocoder",
  "n_gpu": 1,
  "arch": {
    "type": "FFTNet",
    "args": {
      "n_blocks": 1,
      "n_layers": 11,
      "classes": 256,
      "radix": 2,
      "descending": true,
      "aux_channels": 80,
      "fft_channels": 128
    }
  },
  "data_loader": {
    "type": "RandomWaveFileLoader",
    "args": {
      "data_dir": "data/",
      "batch_size": 5,
      "num_workers": 2,
      "segment": 5000,
      "quantization_channels": 256,
      "injected_noise": true
    }
  },
  "optimizer": {
    "type": "Adam",
    "args": {
      "lr": 1e-3
    }
  },
  "loss": "cross_entropy",
  "lr_scheduler": {
    "type": "StepLR",
    "args": {
      "step_size": 100000,
      "gamma": 0.1
    }
  },
  "trainer": {
    "steps": 10000,
    "save_dir": "saved/",
    "save_freq": 1000,
    "verbosity": 2
  },
  "visualization": {
    "tensorboardX": false,
    "log_dir": "saved/runs"
  },
  "feature": {
    "type": "get_logmel",
    "args": {
      "sr": 16000,
      "n_fft": 1024,
      "hop_length": 256,
      "n_mels":80,
      "fmax": 8000
    }
  }
}