Multi-criteria Token Fusion with One-step-ahead Attention for Efficient Vision Transformers

Devs

Multi-criteria Token Fusion with One-step-ahead Attention for Efficient Vision Transformers | Read Paper on Bytez