Get Started with Iterative Studio
Sign in to your Iterative Studio dashboard using
your GitHub.com, GitLab.com, or Bitbucket.org account, or with your email
address. Upon sign-in, the Projects dashboard opens. It includes some Demo
projects for you to explore.
When you sign up, you're on the Free plan. To switch to the Basic plan create a team first, then go to the Team settings to change the plan. To sign up for the Enterprise plan, please schedule a call (see pricing details).
Run and track experiments
- Iterative Studio works with Git repositories. So first, prepare your Git repositories and make sure that the connection to your Git server has been set up.
-
Click on
Add a Project
and follow the on-screen instructions to search and connect to the desired Git repository. You canSkip and Continue
theProject settings
.Once the project is created, remember to configure project settings if needed.
-
Iterative Studio parses your project's
dvc.yaml
file to identify data, metrics, plots and hyperparameters.If you are not using DVC, you can separately indicate which files contain metrics and hyperparameters. However, we strongly recommend using DVC to avail of all the features of Iterative Studio.
-
Each project on the Projects dashboard displays some of the metrics from your Git repository (such as
avg_prec
androc_auc
in the following project). -
Click on the project name to open the project table and explore all your ML experiments.
-
You can submit new experiments by changing hyperparameters and datasets. This triggers model training if your repository has appropriate CI/CD actions set up.
-
To track live metrics and plots of running experiments, set the
STUDIO_ACCESS_TOKEN
environment variable and use DVCLive in your training pipeline.
Manage models
-
Click on the
Models
tab to open the central Models dashboard. Iterative Studio uses your project'sartifacts.yaml
file to identify ML models and specially formatted Git tags to identify model versions and stage assignments. -
Click on the model name to open its details page.
-
You can perform the following actions to manage the life cycle of models:
- Register new models from your Git repositories and remote (cloud) storages.
- Register model versions
- Assign stages (e.g., development, staging, production)
- Unassign stages, deregister versions or deprecate (remove) models
Collaborate
-
You can create a team and invite collaborators. Each team will have its own projects dashboard. To create teams with more than 2 team members, sign up for the Basic or Enterprise plan.
-
You can also make your projects public.