Model Cataloging

Model Cataloging allows users to save the model related information in such a way that it can be reused using the supported functions from the Vantage Advanced SQL and Machine Learning Engines via SQL, Teradata Python Package (teradataml) or Teradata R client Package (tdplyr) client analytic libraries

teradataml offers APIs to use Model Cataloging, allowing users to:

• Save a model and related information to the Advanced SQL database;
• List the saved models;
• Describe a model;
• Retrieve a model;
• Publish a model;
• Delete a model.

In order for any user to use the model cataloging feature in teradataml, setup must be performed by the administrator on the Vantage system. The required scripts along with the instructions to set up the Vantage system are part of this bundle vantage_model_cataloging_1.0.0.tar.gz.

The sandbox environment is based on a SLES12 SP3 docker image that contains a Python distribution (interpreter and add-on libraries) based on the latest Teradata In-nodes Python release for SLES12-SP3:

• Python interpreter (Version 3.7.7 or Version 3.6.7)
• Add-on libraries for Python

User can choose to setup the docker environment and test Python scripts by running them inside the docker container, or user can directly execute scripts on Vantage.

The sandbox environment docker image  name is "stosandbox:1.0".

The docker image size is about 4 GB. Due to the large size, Teradata recommends downloading it beforehand, and saving it into a local folder.