processor_config.json

{
  "n_mels": 8,
  "n_fft": 512,
  "win_length": 400,
  "hop_length": 160,
  "audio_processor": {
    "feature_extractor_type": "GraniteSpeechFeatureExtractor",
    "melspec_kwargs": {
      "hop_length": 160,
      "n_fft": 512,
      "n_mels": 8,
      "sample_rate": 16000,
      "win_length": 400
    },
    "projector_downsample_rate": 5,
    "projector_window_size": 15,
    "sampling_rate": 16000
  },
  "audio_token": "<|audio|>",
  "processor_class": "GraniteSpeechProcessor"
}