No, i think llama.cpp only using testing model