is the text encoder bug in kohya fixed for dreambooth?
is the text encoder bug in kohya fixed for dreambooth?
Fi did testing for both sd 1.5 and sdxl
Fworks great
Mi tried training with the latest json but got horrible results.
the model did not output the face at all until i just used "ohwx man" and nothing else
the model did not output the face at all until i just used "ohwx man" and nothing else
Fyou must have mistake somewhere
Fyou did training sd 1.5 or sdxl?
Msdxl
Mdoes the json need to be updated post fix or does it still work fine?
Vwhich is the best preprocessor and model for depth ? any tips
Fyou need to use our latest json
Mi was using that one
Fsadly dont know i always try all
Fshow me your training dataset and output images
Fyou can DM me
Vokk doc
Mok
Fok load your config
Fwhich has folders
Fall of your images are corrupted
FHi could you take a look at the naming convention for batch outputs, it seems to rename them into an odd format that doesnt follow say, aftereffects protocol FILENAME_####.png etc
Fit names them based on the time of output not the frame number
Fye i made that way
Ffile name is harder
Flol i have been working since 18 hours now 
Fcan it just retain the original names?
Foh god well there's no rush on this man, get some rest
Fmaybe we can append unique variable to original names
Fgood idea
Ftomorrow will check hopefully
Fyeah like add prefix something like that
VOMG ..look at this..Anyone knows how to create this MECHA CYBORG look?
Fis it possible to select a different model to use with supir?
Fn/mind found it
L
A@Dr. Furkan Gözükara When U try to run Llava it says "LLaVA is not available. Please add text manually." I installed with v10.. what should I do ?
AI don't find that pt 1 pt2 pt3 files that u have in the video
R@Dr. Furkan Gözükara Hey did you notice Runpod is totally empty on GPU's? All you can find are A5000's and only 4 regions
Ris the same for you? I used to be able to easily get a 4090 and A6000
Ri can see more options on Vast.ai
RHave you tried it? How you dealing with the GPU scarcity on Runpod?
Fye they need more gpus
Fhavnet tried yet
Rthere's no network volume in vast though, hence why i suppose runpod is doing so well
Ryou can only setup the persistent volume like a pod per instance. and transfer out if it gets rented to someone else.
Ri guess the disadvantage with vast is that it is a market, and doesnt rent out persistent space
R@Dr. Furkan Gözükara will this template work?
Rit is not 2.2.0
Rnevermind, found it
Rbtw @Dr. Furkan Gözükara is RTX 4000 Ada ok for Supir, it is half the cost of 4090?
Rit is 20GB vram