Expanded and recaptioned the dataset, the thing should theoretically be a little more stable now (I don't know to be honest, it still fails in some cases).
A trial run of a LoRA for vaginal mushrooms. A work in progress.
