EsViT: Efficient self-supervised Vision Transformers - View it on GitHub