DynamicViT: Efficient Vision Transformers with Dynamic Token Sparsification 109.5K viewsedited 03:47