CML - CI/CD for Machine Learning - Use GitHub Actions & GitLab CI for Machine Learning

Hi, Dmitry here! I’m a co-creator of Continuous Machine Learning, or CML for short: https://github.com/iterative/cml CI/CD is a well-established practice in software development. However, it is not easy to use CI/CD in with ML. That’s why CML addresses three hurdles for making ML work with CI: 1. In ML, pass/fail tests aren't enough. CML helps to bring metrics and data visualizations to CI/CD reports. You can even get a Tensorboard in your report! 2. Data plays a similar role as code. CML helps to transfer multi-GB datasets to CI runners using DVC. 3. CPU and GPU are needed. CML can automatically provision and deploy cloud compute instances (AWS, Azure, GCP, Ali) for model training using Docker Machine. Our philosophy is that ML projects, and MLOps practices, should be built on top of traditional engineering tools like GitHub Actions and GitLab CI/CD and not as a separate AI platform. The goal is to extend DevOps’ wins from software development to ML.

Very exciting! I used this team's previous product (DVC) at a research lab at Caltech. This looks like a very useful tool.

What a great product done by an exceptional team. Great job Ivan and the rest of the team!