Someone is probably going to port it to llama.cpp soon.