llama.cpp

bf9d14cd - test: remember to set n_thread for cpu backend