Making Vision Transformers Efficient from A Token Sparsification View

Open in new window