LeViT a Vision Transformer in ConvNet's Clothing for Faster Inference - View it on GitHub