Fetching the paper…

Tokens-to-Token ViT: Training Vision Transformers from Scratch on ImageNet · Around