How does the Conv2DWeightNorm layer function?

[ArmanMazdaee]

Hi I was searching on how to implement the weight norm in the Tensorflow when I came across your repository. I have read the Conv2DWeightNorm implementation but I can't quite understand how does it function. Why there is no need to override the call method implementation? How does the kernel get normalize after each update? I would be glad if you could help me to grasp the mechanism.