llama.cpp

b4024af6 - llama : skip main_gpu validation when no devices are available (#23405)