about inpainting model
how to use the "Realistic_Vision_V2.0-inpainting.ckpt" in Diffusers
same question, I guess you switch to this checkpoint model during using inpainting mode?
Same question, can I convert ckpt to diffusers by "diffusers/scripts/convert_original_stable_diffusion_to_diffusers.py" directly? Raising warning "Some weights of the model checkpoint at openai/clip-vit-large-patch14 were not used when initializing CLIPTextModel" when I convert it.
Hi! In my experience, I realized that the inpainting model cannot be converted to diffusers. You can't train an inpainting version of a model, instead you have to train a regular version of the model, then turn it into an inpainting version using a model merge. Use the "Add Difference" method.
A + (B - C) * M
Model A = sd1.5-inpainting (3.97 GB)
Model B = your trained model
Model C = sd1.5-pruned (7.17 GB)
M = 1