Beyond Pre-Training: Learning For Knowledge Updates In Language Models