{
    "llm_model": "llm.mnn",
    "llm_weight": "llm.mnn.weight",
    "backend_type": "opencl",
    "precision": "low",
    "memory": "low",
    "sampler_type": "greedy"
}