Created by: JiayiFeng

Currently, our scale_op only support float and double. That's not enough, int and int64 should also be supported.