Model cards

A model card is a summary of the capabilities, datasets, limitations, and training parameters for a given machine learning model. The RI Platform enables you to generate model cards that can be configured to show specific aspects of a given model.

Generating model cards

1. Sign in to an RI Platform instance.

   The Workspaces page appears.

2. Select a workspace.

   The workspace summary page appears, listing the projects in that workspace.

3. Select a project.

   The project summary page appears.

4. Select the Model Cards icon from the left bar.

   The model cards page appears.

5. Select a model from the drop-down and click Create Model Card.

   The model card configuration page appears.

6. Type information into the following data fields.

   Title and Subtitle Company and Creator Intended Use Case Why is this model important? Limitations Input, Output, and Architecture About Data Collection

7. Select toggles to define the content of the model card.

   Choose severity levels for test categories. The model card shows the selected levels.

8. Click Generate PDF.

   The RI Platform saves the selected changes and generates the PDF model card.

The model card is saved to disk.