Hello, I need some advice since I am stuck on this.
I need to change the background arithmetic that happens for some reason. My goal is to create my own custom operators. so that everything in the backend will be using my own arithmetic implementation. It has worked for keras but it was simple. My logic is to create a wrapper class for the torch tensor and the torch layers. The reason I want to wrap the layers is because I want it to look the same for the users. Before I start hammering at this any advice??