<h2 id="hello-welcome!-xsqmwg1sn">Hello welcome!</h2>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)...<img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/fad83b19-fd0a-4834-9691-97901e36e1af/width=525/fad83b19-fd0a-4834-9691-97901e36e1af.jpeg" />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).

Files with SHA256: 64af500dbcda0f1ead9df5b856a4753ce255ee13bfa3185e61a3511bff12d930