Support for custom diffusion fine tuning

[Snimm]

https://github.com/adobe-research/custom-diffusion

Searching for custom diffusion in this repo gives nothing, so I am asking:

Adobe created this fine-tuning technique which is arguably better than dreambooth and textual inversion. It is already implemented in the diffusers library, so it should not be too much work. It would be nice if we can load custom diffusion checkpoints and also train custom diffusion models. Here are some already trained custom diffusion models.

https://www.cs.cmu.edu/~custom-diffusion/assets/models/

[TingTingin]

you can already load custom diffusion bins they work like regular inversions no training yet though