PyTorch's top priorities have always been flexibility and hackability, while continuing to deliver good performance. PyTorch 2.0 introduces new compiler technologies that improves PyTorch's performance, while maintaining the eager-mode development and user experience you know and love. In this workshop I’ll introduce new PyTorch 2.0 provide an overview of the technology stack that powers the new torch.compile() API: TorchDynamo, AITAutograd, PrimTorch and TorchInductor. The new compiler stack reduces training times across a wide range of workloads while being fully backwards compatible. Bring your laptops, or connect to remote GPU powered systems to run the examples.

Workshop format: presentation and hands-on

Target audience: This talk is intended for a those looking to start using PyTorch and experienced PyTorch users.

Knowledge prerequisites: Some experience with Python would be beneficial

Hardware/software prerequisites: TBD

Learning objectives: Attendees will leave with an understanding of how to take full advantage of PyTorch 2.0.