I am working on a new dataset and will train a v2 later this week but for now I am releasing an earlier epoch of v1. Should be much less over trained than the first vers and much easier to use :)