Explore essential techniques for troubleshooting and testing machine learning codebases and deep neural networks in this 43-minute lecture from the Full Stack Deep Learning 2022 course. Learn about software testing fundamentals, including tools like pytest, doctests, and codecov, as well as clean code practices using black, flake8, and shellcheck. Discover automation strategies with GitHub Actions, and delve into ML-specific testing approaches for data, training processes, and models. Understand the importance of testing in production and the ML Test Score concept. Gain insights into troubleshooting models and performance issues to enhance your ML development skills.
Overview
Syllabus
Overview
Testing software
Testing tools: pytest, doctests, codecov
Clean code tools: black, flake8, shellcheck
Automation
GitHub Actions for automation
Testing ML systems
Testing data
Testing training
Testing models
Test in production
The ML Test Score
Troubleshooting models
Troubleshooting performance
Outro
Taught by
The Full Stack
Related Courses
-
Building a Reproducible Model Workflow
-
CI/CD for Machine Learning
-
Python Fundamentals for MLOps
-
Computer Vision Fundamentals with Google Cloud
-
Extinguishing the Garbage Fire of ML Testing - Improving Reliability and Quality in Machine Learning
-
ML Testing & Explainability - Full Stack Deep Learning - Spring 2021
