Comment by sailingparrot

It's trained using the same architecture, and with a very similar dataset, so it should be very close.