Cml

CML, short for Continuous Machine Learning, is an open source command line tool from Iterative, the team behind DVC, that brings continuous integration and delivery practices to machine learning projects. It runs inside existing CI pipelines on GitHub, GitLab or Bitbucket to automate machine provisioning, model training and evaluation, comparison of experiments across project history, and monitoring of changing datasets. It is aimed at data science and MLOps teams that want reproducible, reviewable model work without adopting a separate ML platform. The central feature is automatic reporting: after a pipeline trains a model, CML posts a report with metrics and plots as a comment on the pull request, so reviewers can judge a change to data or code by its results. Teams manage experiments with Git branching and track who trained a model or modified data and when, while data and models are codified with DVC rather than committed to the repository. Combined with DVC, a pipeline can pull data, reproduce the training run and publish a metrics diff and loss plot against the main branch. CML needs no databases or extra services; it relies on the Git host plus the team's own cloud on AWS, Azure, GCP or Kubernetes, and ships as an npm package, Docker images and a GitHub Action. CML is free and open source. No paid plan is published for it; compute costs come from whichever CI runners or cloud machines the team uses. The site currently promotes Iterative's newer open source project, DataChain, alongside CML.