When you frequently look at some visualization, you will get a feeling when things are normal, or also when things are unexpected. If it is the first time you

When you frequently look at some visualization, you will get a feeling when things are normal, or also when things are unexpected. If it is the first time you look at this, or first time for this specific visualization, or first time for a new type of model, it will not tell you much, and you cannot really tell when things are wrong. You really need to compare it. E.g. in the readme, it shows the example of overfitting, where you see that activations became too large. But to tell whether this is unexpectedly large or not, you need to compare it to some other plots where everything is fine.