View and manage registered models¶

In the Registry, model packages are grouped into registered models as versions, allowing you to categorize them based on the business problem they solve. Once you add registered models, you can search and filter them. You can also view model and version info, share your registered models (and the versions they contain) with other users, and download model packages. Registered models can contain DataRobot (Workbench and Classic), custom, and external models as registered model versions.

In the top-left corner of the Model directory page, you can search and filter the table of registered models:

Click Filter to apply or modify filters on the table of registered models. You can filter by Target, Target type, Created by, Last modified, and Global model (a premium feature), then click Apply filters:

To clear an active filter, in the Filters applied row, you can click x on the filter badge. You can also click Clear all to remove every filter applied.

Click Search and enter the registered model name, target name, registered model tag, or related item metadata (e.g., a model, Use Case, or training dataset, or the owner of any of those items) to locate a specific model in the table of registered models.

To open the list of registered model versions, click a registered model name or the right arrow (). Click the down arrow () to close the list of registered model versions:

Once you locate the registered model or model version you are looking for, you can access information about the registered model or version, along with a variety of management actions.

View registered model details¶

Click a registered model in the Model directory table to open a list of associated versions. To view more information about the registered model itself, click the Actions menu located in the last column for each registered model, then click View details:

From that panel, you can access the Target, the Global model tag (a premium feature), Created by information, Last modified by information, and the Registered Model ID:

Add registered model tags¶

On a registered model's Overview tab, in the Tags group box, click + Add to add key-value tags, view existing tags, or click + Add tag to add to existing tags:

In the Add key value(s) dialog box, configure the following settings and then click Add:

View registered model deployments¶

On a registered model's Deployments tab, view model deployments for all versions of a registered model, in addition to the associated deployment status information. Click a deployment row to open the deployment's overview page:

Share registered models¶

In the actions menu , located in the last column for each registered model on the Model directory page, use the Share option to grant permissions to a registered model:

When the registered model panel is open, in the actions group located in the upper-right corner of the panel, click the share icon :

In the Share dialog box, you can search for a user to Share with, select an access level for that user, and click Share:

Registered models are the model artifacts used for sharing, not model packages. When you share a registered model, you automatically share each model package contained in that registered model.

Delete registered models¶

In the actions menu , located in the last column for each registered model on the Model directory page, use the Delete option to delete a registered model and all associated versions:

When the registered model panel is open, in the actions group located in the upper-right corner of the panel, click the delete icon :

To confirm the deletion of a registered model and all associated versions, in the Delete registered model dialog box, click Delete.

Add a new version to a registered model¶

In the actions menu , located in the last column for each registered model on the Model directory page, use the + Add version option to add a version to the registered model:

When the registered model panel is open, in the actions group located in the upper-right corner of the panel, click Add version:

Then, you can register a new DataRobot, external, or custom model as a version of the registered model, with the relevant fields pre-filled from the registered model details or the model in Workbench.

View version details¶

To open a registered model version, click the row for the version you want to access. The first time you open a version, it opens to the Overview tab; however, if you switch tabs, that selection persists as you open other versions. The registered model version panel includes the following tabs:

Update registered model version stage¶

Each registered model version has two stage types: the System stage and the User stage:

DataRobot determines the System stage—it cannot be changed. The User stage is configurable from any tab on the registered model version details panel:

Changes in the registered model version stage generate system events. These events can be tracked with notification policies, through "single event" triggers for each possible user stage transition, and through two event groups: the Model Version Events group and the Model Version Stage Transition Events group.

Download a model package or Scoring Code¶

To download the model package associated with a registered model version, you can open the registered model version and, in the upper-right corner of the version panel, open the Actions menu . From the menu click either of the following:

• Use portable prediction server: Download the model package and copy a code snippet to run the portable prediction server (PPS). Use these assets to configure a remote DataRobot execution environment for DataRobot model packages (.mlpkg files) distributed as a self-contained Docker image. For more information, see the PPS documentation

• Download scoring code: Download a Scoring Code JAR file from DataRobot and copy the Java, Python, or CLI snippet used to make predictions. Scoring Code is portable and executable in any computing environment. This method is useful for low-latency applications that cannot fully support REST API performance or lack network access. For more information, see the Scoring Code documentation.