Yes, llama.cpp does that.