PyTorch Conference North America schedule live for Oct 20-21 in San Jose

The PyTorch Conference North America has released its full schedule for October 20–21 in San Jose, bringing together over 3,000 developers, researchers, and practitioners. The program covers training and inference, compiler innovations, responsible AI, and the broader PyTorch ecosystem. Notable sessions include 'Observability Tooling for Cudagraph Workloads' by Natalia Gimelshein and Driss Guessous of Meta, 'Unlocking the Full Potential of TorchDynamo: Accelerating, Comparing, and Debugging ML Systems' from UC Berkeley and University of Washington, and 'Scaling Foundation Models: From Broken to Near-Linear Multi-Node Training' by Pinterest’s Sheng Huang. Early bird passes are $599 until July 31 (a $400 savings), and the poster CFP closes July 26 at 11:59 p.m. PDT.

Sponsorship opportunities are available for the conference, which draws technical leaders in AI infrastructure and applications. Diamond-level sponsors include AWS, Crusoe, Google, and Qualcomm, with platinum sponsors Arm, Baseten, Clockwork.io, Intel, Lightning AI, Red Hat, and Gold sponsors Citadel Securities, d-Matrix, DE Shaw & Co, Datadog, Jane Street, Lemurian Labs, and rebellions. The event provides a unique networking and learning opportunity for those working with PyTorch, from core development to production deployments.