Fetching the paper…

Rethinking and Improving Relative Position Encoding for Vision Transformer · Around