config.json

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