Hello,
Thank you for providing this excellent codebase! It has been very helpful for my work. I was wondering if it would be possible to extend the current implementation to support the DeiT model, specifically the facebook/deit-base-distilled-patch16-224 variant.
Currently, your code supports ViT and other models, but I am interested in using DeiT for similar visual tasks. Could you provide any guidance or code modifications to make it compatible with the DeiT architecture?
Any insights or plans for future support would be greatly appreciated!