{
  "architectures": [
    "GazelleForConditionalGeneration"
  ],
  "audio_config": {
    "model_type": "wav2vec2"
  },
  "audio_model_id": "facebook/wav2vec2-base-960h",
  "audio_token_index": 156939,
  "hidden_size": 4096,
  "ignore_index": -100,
  "model_type": "gazelle",
  "projector_type": "mlp",
  "stack_factor": 8,
  "text_config": {
    "model_type": "llama",
    "vocab_size": 156940
  },
  "text_model_id": "meta-llama/Llama-3.1-8B-Instruct",
  "torch_dtype": "bfloat16",
  "transformers_version": "4.47.0",
  "vocab_size": 156940
}