PyTorch on NVLink

I am planning install NVLink on my RTX 8000s. If I connect 2x GPUs with NVLink, do I have to reprogram my PyTorch program for NVLink? NVLink is under CUDA’s control so I do not have to reprogram for NVLink but I just want to make sure my tensorflow or pytorch programs runs after an installation of NVLinks.

Does anyone know about it?