{ "architectures": [ "LlamaForCausalLM" ], "model_type": "llama", "vocab_size": 128256 }