How does the InferenceNet interact with the denoising unet

[phtphtpht]

I can't understand how does the InferenceNet interact with the denoising unet. In the code, the two modules seem completely seperate from each other, and there is no information exchange happen in the training stage.
Maybe someone kind enough can point out the interaction part of the code for me?
有大佬知道代码中InferenceNet和Denoising Unet是怎么交互的吗，没有看到inferencenet的信息被传递给Denoising Unet..

[CJ416]

在 src.models.mutualattention 里面的 referenceattentioncontrol类里面，有update这个方法

[CJ416]

By the way, it should be 'interact' but not 'iteract' (dog head manually)

[phtphtpht]

By the way, it should be 'interact' but not 'iteract' (dog head manually)

哈哈，谢谢提醒和解答！