I have to thank _mammoh, who is helping me with the training and the dataset

I had to make an almost exact copy of c0c's LoRa, since he did the same thing(I'm referring to this LoRa, to generating Coco in this specific style.),although it doesn't have metadata for some reason, I know there's an option in Kohya.

But it's not the same; you have to use "shortstack" to make it more faithful to the original style and c0c`s LoRa, apart from the fact that the weight is 0.9 and not 1, in my opinion it looks better.

I want to do the rest of the LoRas he had about Coco, it's going to take a while of course.