How to initialize the Transformers with ImageNet-1k weights and fine-tune them?

[abellnav]

Good afternoon,

First, I would like to thank you for the code provided, and I find the work very interesting. Please could you share the code you have used to initialize the Transformers with ImageNet-1k weights and to fine-tune them using another database given in the ‘build_dataset’ calls for Image classification? I tried to import pretrained Vision Transformers from HuggingFace using the argument ‘finetune’, but the corresponding code which considers ‘args.finetune’ gives me errors.

Thank you very much in advance.

Best