need ggml-v3 to run. llama-cpp-python > 0.1.57 from https://github.com/ymcui/Chinese-LLaMA-Alpaca