{
  "architectures": ["ChatterboxModel"],
  "model_type": "chatterbox",
  "text_config": {
    "model_type": "gpt2",
    "n_layer": 24,
    "n_ctx": 8196,
    "n_embd": 1024,
    "hidden_size": 1024,
    "n_head": 16,
    "layer_norm_epsilon": 1e-05,
    "vocab_size": 50276,
    "speech_vocab_size": 6563
  },
  "audio_config": {
    "num_hidden_layers": 6
  }
}