[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]
2023-11: Warosu is now out of extended maintenance.

/vt/ - Virtual Youtubers

Search:


View post   

>> No.70986910 [View]
File: 2.20 MB, 1248x1824, 00049-1743869136.png [View same] [iqdb] [saucenao] [google]
70986910

>>70983739
>ccc anon denying someone a 1.5 version of a lora kek.
There's an explanation for that, mainly that all of my 1.5 stuff (nai model, settings, testing set-up with 1.5 models and wildcards) resides on another server that is currently down until I reinstall stuff on it. Plus there was already a 1.5 lora for that chuuba iirc so I'd rather update the ones that don't have XL versions than improving 1.5 loras.
I can make the training work on my current set-up and bake something quick but I will defer any intense 1.5 lora testing to others if they would like, as I'm getting too many crashes as it is.

>>70986556
Good luck! Did you add the captions as well? Maybe I will make a Kohya GUI tutorial at one point on how to tag a dataset, use dataset preparation and train basic character loras but I expect a bunch of "acshually you should increase the learning rate because pony does this and remove the text from images and don't use shuffle and lmao using GUI" and other crap like that

Navigation
View posts[+24][+48][+96]