Registered Models & Versions

In the Registered Models section, you can manage and explore various iterations of machine learning models. Each Registered Model represents a unique model entity, and Model Versions are iterations of these models, each with its configuration and deployment settings. This structure allows users to track and manage the lifecycle of models from development through deployment.

Registred Models

Overview

• Create: Utilize the "Create" button in the toolbar to register a new model. Provide a brief description, and the model will seamlessly join the existing table of registered models.

• Search Functionality: The search field allows you to quickly explore and locate specific registered models within the table.

Activity Log

The Activity Log tab provides a chronological view of activity logs related to the registered model.

Registred Model Specific Page

By selecting a registered model, you'll be directed to a dedicated page with detailed information about that model. Here, you can:

• Search Versions by ID: Easily locate specific model versions using their unique identification.

• Display Activity Logs: Chronological view of logs related to the model version's creation.

• Create New Version: Click on the "Create" button to generate a new version. Choose the run name from the list, and the version will be seamlessly added to the table.

• Delete Registered Model: Access the settings tab to delete the registered model when needed.

Model Versions

Overview

Upon selecting a specific version, you'll navigate to the "Model Version Overview" page.

Activity Log

The Activity Log tab provides a chronological view of activity logs related to the model version.

Model Version Comments

To add helpful comments, navigate to the Comment tab.

Model Version Actions

By navigating to the Actions tab, you can apply the following actions:

• Change Stage: Modify the stage of the model version, choosing from options such as Production, Archived, or Staging.

• Copy configuration: Copy configuration from another Model Version.

• Save as model blueprint: Create a model blueprint using this model version configuration as a base.

• Delete Model Version: Easily remove a model version by clicking the "Delete Model Version" button.

Deploying the Model Version

To deploy a model version, ensure that you have first set up the necessary configurations. Proper configuration of the model version is essential to ensure that the deployment aligns with operational requirements and resource capabilities.

Setting Model Version Configuration

Before proceeding with the deployment, configure the model version according to your specific needs. For detailed guidance on configuring model versions, visit the Model Version Configuration page.

Deployment Process

Once the model version is appropriately configured, you can deploy it by following these steps:

1. Navigate to the Overview Tab: Under the Model Version Overview page, locate the Overview tab.
2. Deploy the Model: Click the Deploy button to start the deployment process. If the model is already deployed, you will see an Undeploy button, which you can use to stop the current deployment.

This process enables you to transition your model from a development or staging environment into staging/production, leveraging the scalability and robustness of the deployment platform.