I am trying to train a U-Net on my data of shape (3,32,32,32) but my loss function has weird spikes in it that keep on decreasing. I do not understand what inference should I ma
