config.json

{
  "architectures": [
    "LlamaForCausalLM"
  ],
  "model_type": "llama",
  "torch_dtype": "float16",
  "use_cache": true,
  "vocab_size": 128000
}