Hi all ,
here is my 1st Lora , it didnt result at all in what i was meaning, but since it work for nearly 3hours for 4680step and about 30 epoch on 86 pictures and the text who should have gone with ^^
you jsut have to put lora in selection or in the prompt ,
normaly it should have respond to jtree , jbuilding , or jcastle , but no it would seem from my try
Here what i did notice about this lora when you put it on
around a weight of 0.6 the face mostly is affected and it made it a lot unprefect / ugly , the rest keep about the same going worst as you go toward 1 , on some prompt i did notice some change of composition at around 1.35 weight
When i tried on my ink drawing prompt it made less drawing but more toward realist and not ugly
On anime prompt didnt change much, less details some
So depending what you do it might serve or not
And as for the embedding i did before there been some who still download it and a few who even put in favorite
Maybe it will be off some use to some of you , if you do end using it , dont hesitate to post all or some of it on here im curious about how people would apply it
well have fun with it , if i can i will try to put the dataset it try to train on , if any got suggestion about things