Resize control if there is a tensor size mismatch

[asagi4]

Fixes PatchModelAddDownscale when used with controlnets, and possibly some other nodes doing similar things as well.

Fixes #3022

I'm not sure if there's a situation where the operation can fail after this, so I didn't remove the try/except

[comfyanonymous]

Does this actually work and would it be better to use nearest instead of bicubic?

[asagi4]

I've used this locally for quite a while (never just got around to sending a PR) and it does at least let the controlnet do something, which is an improvement over things not working at all (Instead you just get a bunch of "warning control could not be applied torch.Size([2, 1280, 20, 16]) torch.Size([2, 1280, 40, 32])" ). You can also look at the linked issue where a couple people have reported using it.

I haven't tested different algorithms.

[mcmonkey4eva]

"it does something", or "it does what it's supposed to do"?

Is this an attempt at hacking SD1 and SDXL controlnets to cross over, or...?

[asagi4]

Iit allows the controlnet to work instead of failing and doing nothing when PatchModelAddDownscale does its downscaling, by scaling the control tensor to match; this has nothing to do with mixing models.

[mcmonkey4eva]

ooh the PatchModelAddDownscale (Kohya Deep Shrink) node, my bad, I skimmed past that the first look.