In ordinary software, tests check that code behaves the same way every time. In machine learning, a system can be broken even when the code is perfect, because the data may have drifted or a new model may be quietly worse than the last one. That means testing ML covers three things: the code, the data, and the model itself.
You'll start with what makes testing ML systems different, then learn the basics of CI/CD: continuous integration to build and test every change, and continuous delivery to ship validated changes automatically. From there you'll build real workflows with GitHub Actions, and see how CML (Continuous Machine Learning) adapts those pipelines for ML by running training and reporting results directly in your pull requests.
Automated, tested changes are what let you ship models with confidence. The next module, Model Serving & Deployment, covers where those models end up.
What testing ML is and how it differs from testing deterministic code, and why data and models need tests as well as code.
A quick introduction to DevOps, continuous integration and delivery, and how a CI/CD pipeline can improve productivity and code quality.
A complete beginner tutorial on creating and using GitHub Actions workflows to automate CI/CD.
An introduction to CML, an open-source library for running machine learning projects through CI/CD with GitHub Actions or GitLab CI.