head/config.json

{
    "medusa_num_heads": 3,
    "medusa_num_layers": 1,
    "transformers_version": "4.34.1",
    "base_model_name_or_path": "vicuna-13b-v1.5",
    "vocab_size": 32000
}