Having trouble adding tensor core support to pytorch

So I’ve been using this repo for some fps interpolation which works pretty well. I recently got a new GPU (2060super) to speed it up, which it did, however I recently found it that the tensor cores aren’t being used and that you have to activate them somehow?

I went online and found out that you need to use some sort of Nvidia/Apex extension for pytorch to be able to use those tensor cores, however even though on their repo they say that you only have to add 3 lines of code, I am completely confused on where to begin. Could anyone please help me?

Here’s the repo I am talking about: https://github.com/MortenHannemose/pytorch-vfi-cft
Here’s the nvidia/apex repo: https://github.com/NVIDIA/apex