Explore the challenges and opportunities of building a mobile-optimized machine learning platform in this 27-minute conference talk from F8 2019. Discover Facebook's approach to moving AI/ML technology from servers to edge devices, covering topics such as Natural Language Processing, video processing, speech recognition, and content understanding on mobile. Learn about the benefits of mobile AI, the "Yard Effect," and key techniques like quantization and machine learning compilers. Gain insights into the mobile hardware landscape, kernel libraries, and benchmarking strategies used to optimize AI performance on mobile devices.
Overview
Syllabus
Intro
Agenda
Natural Language Processing NLP
Accessibility
Why Mobile
Challenges
The Yard Effect
Peak Performance
Mobile Ecosystem
ML Pipeline
Mobile AI Platform
Challenge
Flexibility
Core Techniques
Mobile Hardware Landscape
Kernel Libraries
QM Pack
Machine Learning Compiler
Quantization
Results
Quantization Flow
Benchmarking
Outro
Taught by
Meta Developers
