No scale factor applied when concat image related information

[liuliu]

Hi, this is a minor nit and asking to see if there are explicit motivations to do so. In the process of supporting this model in Draw Things, I noticed that unlike Inpainting models, where the encoded image multiplied the "scale factor": https://github.com/runwayml/stable-diffusion/blob/main/ldm/models/diffusion/ddpm.py#L550 while in instruct pix2pix, we don't: https://github.com/timothybrooks/instruct-pix2pix/blob/main/edit_cli.py#L103

Not a big deal to me as I figured out, modified a bit and it worked exactly like expected. But want to call it out and see if there are some considerations applied so I know where my modifications should be applied (whether treat the edit model as a special case, or just modify the first layer conv2d weights).