old-school inpainting support

[anotherjesse]

It would be nice if you could mask/inpaint with your dreambooths.

while stable diffusion 1.5 / 2.x has moved to having specialized models/weights for inpainting, as @andreasjansson says:

yes, since SD 1.5 inpainting has been separate models, trained by extending the u-net with another two channels (mask and masked image). the legacy inpainting pipeline is still better than nothing, and that should work on dreambooth models too