{
    "model_type": "llama",
    "architectures": ["LLaMAForCausalLM"],
    "hidden_size": 4096,
    "num_hidden_layers": 32,
    "num_attention_heads": 32,
    "intermediate_size": 8192,
    "max_position_embeddings": 512,
    "vocab_size": 50265,
    "padding_idx": 0,
    "layer_norm_eps": 1e-5
}