{
  "auto_map": {
    "AutoProcessor": "processor.MuLanProcessor"
  },
  "duration": 10,
  "fbank_kwargs": {
    "dither": 0.0,
    "frame_length": 25.0,
    "frame_shift": 10.0,
    "htk_compat": true,
    "num_mel_bins": 128,
    "sample_frequency": 16000,
    "use_energy": false,
    "window_type": "hanning"
  },
  "feature_extractor_type": "AudioSpectrogramTransformerFeatureExtractor",
  "freq_mask": {
    "mask_param": 48
  },
  "n_frames": 1000,
  "n_mels": 128,
  "processor_class": "MuLanProcessor",
  "sample_rate": 16000,
  "spec_mean": -4.2677393,
  "spec_norm": true,
  "spec_std": 4.5689974,
  "time_mask": {
    "mask_param": 192
  },
  "training": true
}