--dataset torch/CIFAR10

[TalRub104]

Hi,
After downloading the CIFAR-10 dataset using the following command:
python3 -c "from torchvision.datasets import CIFAR10; CIFAR10(CIFAR10, download=True)"

I attempted to run the following command to fine-tune on CIFAR-10:

python -m torch.distributed.launch --nproc_per_node=8 train.py CIFAR10 --dataset CIFAR10 --experiment vssm_tiny_v2_TRADE_none_reg --output output/TRADE/vssm_tiny_v2/none --model vssm_tiny_v2 --config configs/xcit-adv-training-cifar10.yaml --mean 0.4914 0.4822 0.4465 --std 0.2471 0.2435 0.2616 --normalize-model

However, I encountered the following error:
RuntimeError: Found 0 images in subfolders of CIFAR10. Supported image extensions are .png, .jpg, .jpeg

When I use the --dataset torch/CIFAR10 option instead, the script runs without any issues. Could you please clarify the following:

How can I make the --dataset CIFAR10 option work as expected?
If not, could you update the "XCiT-S12, XCiT-M12, and XCiT-L12 fine-tuning on CIFAR-10 and CIFAR-100" command?

Thanks!