Cannot train a Dreambooth model with WD 1.4
Greetings,
I am training a Dreambooth model on Waifu Diffusion 1.4 and there is something weird going on. I would like to see if anyone knows what it could be
Here:
The model is trained just fine, it works when generating samples. The image above is from the current checkpoint, nothing was changed.
But then I have to generate the .ckpt file to use on the AUTOMATIC1111 GUI.
However, if I try to generate the images with the "compiled" ckpt...
Only garbage comes out with the "ckpt", but the regular Diffusers (non ckpt) model works fine
Do you have any idea of what could be going on?
What is the process to generate ckpts with Waifu Diffusion 1.4?
Thank you
Do you have a textual inversion that you made just in case with the same token clcsgm ? the UI will prioritize it first
I had a similar issue where i prepared in advance a textual inversion with my token but never trained it