PyTorch implementation and pretrained models for DINO. For details, see Emerging Properties in Self-Supervised Vision Transformers. Run DINO with ViT-small network on a single node with 8 GPUs for 100 ...