Pytorch Model Inference optimzation

Hi Everyone!
Is there any technique/code to optimize the pytorch model’s inference. I can try the nightly build too if anything new has been added/supported in newer versions ?