EfficientViT: Memory Efficient Vision Transformer with Cascaded Group Attention | IEEE Conference Publication | IEEE Xplore