llama.cpp

b5af7ad8 - llama : refactor quantization to avoid <mutex> header