Inference Shape Error: 48m_canny and sdxl

[Hammour-steak]

I encountered the above error when using 48m_canny and sdxl, would like to ask what is the reason? The dim of mask attention seems to be consistent with the batch?

[FriedFeid]

Hi Hammour-steak,

I have just run everything, but it works as intended for me.
You may have accidentally changed some parameters in the cdxl_encD_canny_48m.yaml.

Without the code itself, I can not state more.

If the error continues, could you share a minimal example?

Best,
FriedFeid

[Hammour-steak]

These are some of my Settings and test code is the same as README.md.

[Hammour-steak]

I found the problem, the input image must be 1024*1024. Thank you for your priceless work.