Data augmentation on the validation set

[chinmay5]

Thank you so much for the well-written and clear code. However, I had a small doubt. I noted that you have used random transform for the validation dataset as well. I thought, we should stick to deterministic transforms for the validation sets and test sets.

Although I get an idea why the test set used 10 crops but I am not sure why we are using random transforms in the case of our validation dataset. Any help from you would be highly appreciated.