Updated to Version 1.1
Text encoder was not trained in this version but U-Net has much better training
Description
Greatly Improved for Version 1.1
FAQ
Comments (15)
can this generate some blown graphics cards?
Set the steps high enough and the resolution at max and it just might. (LORA not responsible for burning graphics cards)
I love that Fermi example image
Thanks
Using a GPU to generate a GPU is so meta
imagine those generated GPUs could do bimining for ya 😂
Generating NVIDIA GPU on my AMD GPU is even more funny
self portrait
Radeon GPU loRA when??? :0
Maybe if I convert my models to Onyx and Direct ML....maybe
@Felldude oooooooooooooo gotcha gotcha
@CappyAdams Training at 512 is pushing my 3050 to the brink, I'd be curious if the microsoft/Nvidia partnership would increase the speed as much as they claim. Also I am in the middle of training an XL Lora assuming I don't break training with my web browsing shenanigans I have 3 hours to go
@Felldude ah I see