Y_range sigmoid vs clamp

In regression problems, we can use the y_range parameter to force the outputs to be in a certain range, let’s say [0,1]. In one of its lesson (can’t remember which one) Jeremy mention that we should probably add a bit of padding to make sure the model can predict the edge values like 0 and 1. Because values would have to be pretty large to actually predict 1 after the sigmoid. So maybe change y_range=[-0.1, 1.1].

Would it be a good idea to then add clamp(0,1) to that to force the values to be between 0 and 1? I was wondering why fastai only used the sigmoid and no clamp.

Thanks!