refer https://github.com/ggml-org/llama.cpp/discussions/15396#discussioncomment-14154375
· Sign up or log in to comment