Token Compensator: Altering Inference Cost of Vision Transformer without Re-Tuning

VLM
Re-assign community