I’m using the fastai.gan.gan_critic() and pre-training it as in the course (but with a very different application).
Is it normal for the train loss to be a lot smaller than validation loss after just one epoch, or am I doing something very wrong?
train loss: 0.026537
valid_loss: 0.425462
accuracy_thresh_expand: 0.862143
Edit: Sorry, I had messed up my transforms , I fixed it.