According to their docs[1], three different ways of quantizing your model are supported - one of which is quantization-aware training (3).

[1] - https://pytorch.org/docs/master/quantization.html#quantizati...