OK - First of all I love the auto-rescaling when no argument is given. 👍

My only nitpick is this doesn't sort the problem when ylim is set by the user as (say) 10 ** -172. All the ticks will be blank. But I think _get_yticks() can be tweaked to get this -

def _get_yticks(ylim_oom):
    order_of_magnitude = int(np.floor(np.log10(ylim_oom) / _OOM_SIZE)) * -1
    ylim_oom = 10 ** - ((order_of_magnitude -1) * _OOM_SIZE)
    return np.geomspace(start=1, stop=ylim_oom, num=order_of_magnitude + 1)

This will then correctly provide the ticks spaced out by 10 ** -50, but the ylim (set elsewhere) will add some extra y-axis on top if needed to get to the users requested value ylim. This extra space will not have any minor ticks (so there will be no tick with 10 ** -172 for instance), but I don't think there needs to be - this is the expected behaviour, in my mind at least.

My only nitpick is this doesn't sort the problem when ylim is set by the user as (say) 10 ** -172. All the ticks will be blank.

Ooh, not a nitpick, a good catch! Thanks for the fix. Happy for me to merge it in after?

Yes please!