Iterative Studio works with the data, metrics,
hyperparameters and model metadata that you add to your ML project Git
repositories. It works very closely with your Git ecosystem.
This video illustrates how Iterative Studio works closely with your Git
ecosystem.
Note that we have renamed DVC Studio to Iterative Studio and Views to
Projects.
Using DVC and Git, you will push all your ML experiments to your GitHub,
GitLab or Bitbucket repositories as Git commits.
Using Iterative Studio, or using the command line interface (CLI) of GTO,
and possibly MLEM, you will push all your ML model details to the Git
repositories as Git commits and Git tags.
How Iterative Studio extracts your project data
When you connect to these Git repositories from Iterative Studio, the
project's dvc.yaml is used to identify all the data, metrics and
hyperparameters in your experiments.
You can set up your CI/CD actions (e.g. GitHub Actions) to run model training
upon the creation of Git commits, tags or pull requests. You can use CML in
your CI/CD actions for continuous machine learning.
When you
register new versions
of your ML models or
assign stages to them,
Iterative Studio creates annotated Git tags representing the actions.