Debugging Machine Learning on the Edge with MLExray

Explore the challenges and solutions for debugging machine learning models deployed on edge devices in this insightful conference talk by Michelle Nquyen from Stanford. Gain an understanding of the traditional model deployment process, its benefits, and potential issues that can arise. Learn about MLExray, a powerful Python API designed to address common problems in edge ML deployment. Discover how to use reference pipelines, handle preprocessing errors, implement MLExray assertions, and manage quantization issues. Examine hardware requirements and witness a live demonstration of MLExray in action. Discuss the tool's current limitations and explore additional resources like Pixi for further learning. Equip yourself with the knowledge to effectively debug and optimize machine learning models for edge computing environments.

Introduction
Michelles background
Why debugging ML on the edge
Traditional model
Benefits
Deployment
What can go wrong
MLExray
Python API
Reference Pipelines
Issues
Preprocessing errors
MLExray assertions
Quantization
Reference Pipeline
Hardware Requirements
MLExray Demo
Limitations
Pixi
Resources
Questions