Learning to Merge Tokens via Decoupled Embedding for Efficient Vision Transformers

Devs

Learning to Merge Tokens via Decoupled Embedding for Efficient Vision Transformers | Read Paper on Bytez