Quantization preferred way

What is the preferred/suggested way to perform quantization in Pytorch? Just want to ensure that I pick up the path that is going to become the standard for PyTorch in future.