Ever wondered how/why to train model in low precision? What is fp32, bfloat16? Need for quantization? Role in training/inference & tuning? Look no further, this article covers all of this and more.