testing out training loras with less images! i also didn't compress these ones, so the model got a handle on the way i like to do lineart :)