This version has been trained off of roughly 1000 images, with enough variety of art styles that your generations without any artist LORAs should come out pretty plain looking-- which, if you want a good base to add LORAs to, is a good thing in my opinion.
That being said, considering the quality of hyperpreg images online, your faces might come out a bit more warped than usual on the first go around. Just be prepared to do a bit of inpainting on some artist LORAs.
Source images range in size from octomom to omegapreg, and additionally range in shape from smooth and circular to bumpy and oblong. Also occasionally generates images with stretch marks if you're into that.

