Brad Heintz, Facebook AI Research
gtc-dc 2019
We’ll discuss the latest release of PyTorch and how it relates to the field of machine learning, which continues to grow in multiple directions and prompt questions related to hardware performance, scaling, and the privacy implications of human-centered AI tools. We’ll explain how this version of PyTorch answers these questions by enabling and accelerating machine learning on new platforms, offering distributed training on GPUs, and updating performance-focused tools like the just-in-time compiler. We’ll include examples from Facebook and other corporate and research institutions to illustrate how PyTorch is being applied to a variety of use cases.