config.json

{
  "model_type": "redimnet2",
  "model_name": "b6",
  "embed_dim": 192,
  "sample_rate": 16000,
  "n_mels": 64,
  "hop_length": 160,
  "n_fft": 512,
  "max_duration_sec": 60,
  "architecture": {
    "encoder": "ReDimNet2Encoder",
    "initial_channels": 64,
    "out_channels": 224,
    "pooling": "ASTP",
    "return_2d_output": false
  },
  "training": {
    "train_type": "lm",
    "dataset": "vox2",
    "loss": "sphereface2"
  },
  "format": {
    "file_format": "pytorch_state_dict",
    "key_prefix": "_orig_mod.",
    "strip_prefix": true
  }
}