Model Cataloging allows users to save model-related information in a way that it can be reused by the supported functions of Machine Learning Engine or Advanced SQL Engine via SQL, Teradata Package for Python (teradataml) or Teradata Package for R (tdplyr).
For example, an ML Engine DecisionForest model saved by using SQL can be retrieved to use with teradataml for scoring with the DecisionForestPredict function from ML Engine or Advanced SQL Engine. Similarly, an ML Engine or Advanced SQL Engine model saved by using tdplyr can be described and retrieved by teradataml. See save_model() for more information on what information is saved.
- Save information and related objects to the catalog for the model created using the teradataml analytic functions, using save_model();
- List the saved models, using list_models();
- Describe a model, using describe_model();
- Retrieve a model for reuse, using retrieve_model();
- Publish a model to set the access level and status of saved model, using publish_model();
- Delete a model, using delete_model().
To use the teradataml offering related to model cataloging, the Model Catalog must be set up on Vantage system.
The scripts along with the instructions to set up the Model Catalog can be downloaded from the Teradata Package for Python - teradataml page on https://downloads.teradata.com/.