What is a project in Iterative Studio

A project in Iterative Studio is an interactive representation of the ML datasets, hyperparameters, models and metrics defined in your Git repositories. These values are configured in your project's dvc.yaml file. Additionally, live metrics that you send to Iterative Studio using DVCLive are also included in the project.

Within a project, you can:

Explore all the details of the experiments that you have pushed to your Git repository.
Visualize the experiments using plots and trend charts.
Compare experiments.
Run new experiments by submitting new Git commits directly from Iterative Studio.
Track live metrics and plots by sending them to Iterative Studio by using DVCLive.

Prepare your repositories

To display your project's content in Iterative Studio, initialize DVC in your project's Git repository and create dvc.yaml. When running model training and evaluation, save metrics and plots in the files defined in dvc.yaml.

If you are working with a non-DVC repository, you can indicate which files contain metrics and hyperparameters that Iterative Studio should display in the project. However, we strongly recommend using DVC to avail of all the features of Iterative Studio.

To add model metadata to your repositories, you can use Iterative Studio Model Registry, or the underlying GTO or MLEM. Learn more

To run new experiments from Iterative Studio, integrate your repositories with a CI setup that includes a model training process. You can use the wizard provided by Iterative Studio to automatically generate the workflow configuration for the model training CI job. Learn more

To track live metrics and plots of running experiments, configure the STUDIO_ACCESS_TOKEN environment variable and use DVCLive in your training pipeline. You can also do this for experiments that you run from Iterative Studio if you configure the CI job accordingly. Learn more