{
    "architectures": ["LLaMAForCausalLM"],
    "hidden_size": 4096,
    "num_attention_heads": 32,
    "num_hidden_layers": 24,
    "model_max_length": 2048
}