Kiding_Ryuu | LORA STYLE | IL

Hello welcome!

Here I leave a small introduction of the LoRA:

This style is based on pastel sketches, without much detail and with random backgrounds but it also contains images with more detailed faces although I added them as "sketches" and then trained it as a constant... I don't know how that works but the other parameters made the mix of styles more progressive and each epoch adds information instead of training all the data at once. (that's what I understood, I didn't really look closely because I also chose the last epoch hehe)...

Off Topic: So... yesterday I made a version for iIlus Possible and it came out totally incompatible with other versions for Illustrious, the thing is that I trained it again with IL V0.1 but this time I used old parameters decreasing the weight and I changed the "cosine with restarts" for "constant" (I honestly don't know if that changed the result of the lora but I really liked the results, maybe I'll try it in other training sessions).