Using vision.models with the CIFAR dataset?

[smth]

The pre-trained models are all based on Imagenet-12. Most of the model definitions also need 224x224 as the minimum input size. For CIFAR, you are better off with other repos.