Open skunkwerk opened 1 year ago
Thanks for this work. What changes would be required to use SDXL instead of the 1.4 checkpoint?
Did you modify it successfully? After I migrated to stable diffusion xl, the training results didn't give me the desired identity. Both text encoders use different embedding managers, looking forward to someone implementing celeb-basis+SD XL together!
We are trying to make our overall pipeline more automated. At the same time, we will adapt our method to other diffusion models like SDXL.
Thanks for this work. What changes would be required to use SDXL instead of the 1.4 checkpoint?