I want to use torchvision.datasets.FakeData and have a training and testing set. What is the best way to achieve this? If I create two of these, how do I ensure they are generating the same classes?

FakeData uses the argument num_classes to return the dummy targets.

I’m not sure, why you would need a training and test dataset, as your model won’t learn anything useful and FakeData is usually used for debugging/testing.

1 Like