Token Recycling for Efficient Sequential Inference with Vision Transformers

Open in new window