Release Pretrained Models

Release of Models pretrained with self-supervised approach on 5 small datasets (CIFAR10/100, CINIC10, SVHN, Tiny-Imagenet) across 3 vision transformer architectures: ViT, Swin and CaiT.