how is your input photo?
how is your input photo?
Fhi. welcome
Fi dont suggest to generate samples during training
Fmoreover that extension is now sadly outdated
Fif you dont have gpu best is renting from runpod
Fif you cant afford runpod we have kaggle SDXL DreamBooth
Fand it says you have 6 GB VRAM
Fyou can't do training with 6 GB GPU
Fye ye sure 
to get scammed . bye
to get scammed . bye
N
NThis one is my input. Used the same in decorify
Fit is hugely changed
LWe could make NFTs
KWell... It would be a pain in the *** to change the windows redecorating 
BHey, i did not fix the error
BI have an AMD Radeon RX 6600 XT GPU
Fok
Fcan you open any desk?
Fi never tried on amd but lets give it a try
Bi get this when i try running webui
Fyou need to edit
Fweb ui user . bat file
Flet me send you
Fwebui-user.bat121B
Ftry this
Freplace with yours let me know
F@Bophedes
Bthis replaces the shell script right?
Bfile type
B
Fyes
Fin that folder you put
Bi am runnig the webui-user on its own right now
Bthat worked! it opened the stable diffusion on the web!
Fgreat
Bfollowing the How To Use SDXL in automatic1111 and not able to generate an image
Bi need the sdxl.tensor file i assume. but the tesnor download gives me error
P@Dr. Furkan Gözükara kohya dont train v1.5 models ? i tried realistic_vision_5.1 and got errors in load unet model.
FTensorRT wont work on AMD. it is NVIDIA feature
Fit supports
Fyou need to uncheck sdxl option
Fin first page
Pi uncheck this and checked v_prediction then booth unchecked?
PI'll try, another thing I noticed is that in sdxl with 200 repeat and 1 epoch I got 1280 steps. Without sdxl I got 6500 steps. My dataset is 16.
Fit makes 3200
Fwithout reg images
PEven if I provide regularization images?
Fif you provide
Fit becomes 6400
Pyes that's what happened, sdxl then generates fewer steps. So it must be right, I'll try to run it with SDXL unchecked.
Pi can specify vae-ft-mse-840000 for training ?
