Description
CML, short for Continuous Machine Learning, is a tool that brings CI and CD practices to machine learning projects. It runs in GitLab, GitHub or Bitbucket pipelines and generates reports with metrics and plots that appear in each pull or merge request, so the team can compare experiments on the same screen as the code. The site describes a Git-based workflow for data science, where DVC tracks datasets and models outside the Git repository. A minimal setup adds a pipeline file that trains a model, writes the metrics and a chart to a report file, and posts that report with the cml comment command. CML is installed as an npm package, @dvcorg/cml, and a Docker image is provided for pipelines. The site says it needs no extra database or service: you can build your own ML platform from GitHub or GitLab and a cloud provider such as AWS, Azure, GCP or Kubernetes. The site also promotes DataChain, a related open-source tool for working with unstructured data in Python.…
Highlights
- Posts metric and plot reports to pull requests
- Works with GitHub, GitLab and Bitbucket
- Pairs with DVC for data and model versioning
- Installs as an npm package
- No extra service required
Ratings & Reviews
No ratings yet
Be the first to rate CML.
People also visit

ml-pipeline
Designs and implements production-grade ML pipeline infrastructure: configures
CML FAQ
What is CML?
A CI and CD tool for machine learning projects that generates reports from training runs and posts them to pull or merge requests.
Which platforms does it run on?
GitHub, GitLab and Bitbucket pipelines, according to the site's examples.
How do I install it?
Install it globally with npm i -g @dvcorg/cml, or use the iterativeai/cml Docker image in a pipeline.
Does it need extra services?
No. The site says you can build an ML platform with just GitHub or GitLab and your cloud provider, with no database or complex setup.
How does it fit with DVC?
The site says you can codify data and models with DVC so that they are not pushed to the Git repository.
What is DataChain?
A related open-source release from the same team for working with unstructured data in Python, using AI helpers at scale.
What are the best CML alternatives?
Popular alternatives to CML include MLRun.org, Metaflow and Nannyml. See all CML alternatives





