llama.cpp

daf2dd78 - quantize : skip tensor override when in fallback mode (#14995)