Testing out Krea-2 model training. This epoch gives some solid results. End results are varied and I can't stress enough how prompt sensetive this LoRa is. Adding things such source_anime or taking it off can create two varying styles.