{
    "llm_model": "deepseek-7b-int4.mnn",
    "llm_weight": "deepseek-7b-int4.mnn.weight",

    "backend_type": "cpu",
    "thread_num": 4,
    "precision": "low",
    "memory": "low"
}