31. CvT: Introducing Convolutions to Vision Transformer