Continuous integration means you automatically run your machine learning (ML) pipelines when you push your Git commits and get quick results on your ML model in a ReadMe file that is available to all team members in your Github workspace and email in a user-friendly format.

Imagine creating a Dockerfile every time you make changes to your ML model's Python file, and then sending the results with graphs and related information directly from the Python script. Github Actions can do this and save a lot of time while you improve your model. This is a simple CI-CD pipeline.

The basic idea comes from DevOps and is simple. Every time you push your commits to your Github branch, a new Docker container with dependencies is created. Then the model training is started and finally your results are delivered. All workstream processes are defined in a special Continuous Machine Learning (CML) YAML file.
In this example, I'm using a simple Tensorflow regression model, a simple graph created with the Python matplotlib library, and a standard Github account with a public repository.

For this purpose we use:
– GitHub Actions: https://github.com/features/actions
– CML: https://github.com/iterative/cml
The example project is available here on Github: https://github.com/vb100/ml-ops-ci/tree/experiment

The content of the tutorial:
0:00 – Introduction
1:15 – What the model is about.
1:30 – Create a new Github branch with the cml.yaml file for the ML workstream.
2:46 – Execute Git action with print statement in terminal
8:13 – Perform Git action with text information in the readme file
13:18 – Perform Git action with graphical information in the readme file
19:54 – Improve the model and run GitAction again.
23:09 – Summary and thank you

This is the first video in a series dedicated to MLOps (Machine Learning Operations). Subscribe to the channel to learn more!

#mlops #gitactions #docker

Please take the opportunity to connect with your friends and family and share this video with them if you find it useful.