Executorch: Enabling and Accelerating PyTorch Models on Edge Devices

Explore the challenges of on-device machine learning and discover how ExecuTorch, a new stack built on PyTorch 2, enables efficient and portable execution of PyTorch models on edge devices. In this 33-minute conference talk, Chen Lai from Meta introduces ExecuTorch, which was released as a preview version during the PyTorch conference 2023. Learn about the diverse landscape of edge devices, ranging from smartphones to embedded systems, and understand how ExecuTorch addresses the complexities of deploying ML models on these platforms. Gain insights into the innovative approach that maintains both user and developer experience while optimizing performance for on-device machine learning applications.