imagenet1k Huggingface extraction

[AasherH]

Hi. I downloaded the imagenet1k file from huggingface, which has the train_images_0.tar.gz file. However, whenever I extract it (by running tar -xf train_images_0.tar.gz), I only get images, and not subfolders, which torchvision expects. Therefore, whenever I run main.py, I get a FileNotFoundError(f"Couldn't find any class folder in {directory}.").

Does anyone know how to successfully curate the data so that I can get the ijepa code working?

Thanks!

[mikerabbat]

Hi, try this https://github.com/williamFalcon/pytorch-imagenet-dataset/blob/master/README.md