PyTorch 2.1 - New Features and Accelerating Generative AI Models

Explore the latest developments in PyTorch 2.1 through this informative conference talk by Supriya Rao, Engineering Manager at Meta. Dive into new features in compile, distributed, inference, export, and edge technologies, as well as optimization techniques like quantization and pruning. Discover how these advancements can significantly accelerate popular generative AI models, offering substantial performance improvements. Gain practical insights into enhancing AI model efficiency using PyTorch 2.1, making this session valuable for data engineers, AI researchers, and machine learning enthusiasts. Follow along as Rao covers the background and metrics, detailed PyTorch 2.1 features, methods for accelerating generative AI models, and presents the impressive results of these innovations.

- Introduction
- Agenda
- Background and Metrics
- PyTorch 2.1 Features
- Accelerating Generative AI Models
- Results