Had an old unused LoRA lying about, it was literally trained on only 2 very similar images of a certain cat girl who had no name; I'm calling her Kaede Momoka. The first version I trained locally with sub-optimal parameters, so it didn't accurately get the character down right. The 2nd version was done on this site's trainer, with 100 repeats and on 10 epochs--it was massively overtrained by the 3rd epoch I think, so I'm settling with the 2nd epoch; it has more than enough training to know who Kaede is, but not trained to the point that she's always stuck in the same position regardless of prompts lol.

Incidentally, I do have the 10th epoch, which is very much extreme, so if anyone's crazy enough to request it, I won't mind uploading that version, too.