I'm training ControlNet on my dataset, and setting sd_locked = False turns out to improve model performance significantly.
I learned that unlocking sd will also train some parts of the diffusion model, so the base model also changed, right? But one of the problems is that I can not use it on automatical1111 sd webui, which I can use to combine other techniques like LoRA with ControlNet. Does it mean if I unlocked SD when training ControlNet, I have to use the official code base to take full use of the trained model, i.e. running in prediction mode?
Is it possible to save the fine-tuned SD along with ControlNet so I can use them in other places together? Appreciate any advice.
Thanks for the great work!
I'm training ControlNet on my dataset, and setting
sd_locked = Falseturns out to improve model performance significantly. I learned that unlocking sd will also train some parts of the diffusion model, so the base model also changed, right? But one of the problems is that I can not use it on automatical1111 sd webui, which I can use to combine other techniques like LoRA with ControlNet. Does it mean if I unlocked SD when training ControlNet, I have to use the official code base to take full use of the trained model, i.e. running in prediction mode? Is it possible to save the fine-tuned SD along with ControlNet so I can use them in other places together? Appreciate any advice.