I’ve noticed that most of the examples and most of my models result in train_loss > valid_loss. What are the best ways to correct that?
I should add that this still happens after trying many variations on # of epochs and lr
18 Likes