Skip to content

GitLab Integration

Introduced in GitLab 15.11, and generally available in GitLab 16.1, "Machine learning model experiments" allows users to track experiments directly in their GitLab project (under Project (sidebar) > Analyze > Model Experiments), as close as possible to your model code. This feature is relatively new, and is still receiving important feature requests for improvement. On another note, that also means that right now some tooling may be missing when using it.

Architecture
GitLab Experiments

The important feature is the MLflow compatibility. That means you can use MLflow to push to GitLab as a tracking URI. You can learn more in the MLflow client compatibility documentation.

Note

An important thing to note is that your SharingHub may be configured with GitLab integration, and the code generator in the MLflow helper will use GitLab as a tracking URI in the snippets generated. Be sure to check this helper!