This model is a TensorFlow port of ViT B-16 [1] trained with recipes from [2]. It was first pre-trained on ImageNet-21k and was then fine-tuned on the ImageNet-1k dataset. You can refer to
this notebook to know how the porting was done.
[1] An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale:
https://arxiv.org/abs/2010.11929[2] How to train your ViT? Data, Augmentation, and Regularization in Vision Transformers:
https://arxiv.org/abs/2106.10270
