I’m trying to visualize the activations stats of my model by calling the ActivationStats callback when defining my learner as defined in the convolutions chapter.
But by doing so the training process is getting extremely slow? Is this normal and that storing these maps reduces training speed drastically or am I doing something wrong?