The Teradata Package for Python (teradataml), which is Teradata's Python package for client-side scripts, is extended with support for Google Vertex AI. This support is added to the tdapiclient library.
The tdapiclient library allow Vantage users to use Vantage tables and queries with Vertex AI platform for model building as well as model scoring.
- Both model building and model scoring can happen on Vertex AI cloud platform.
- Model scoring has the option of using UDF (API_Request) for in-database scoring.
- This Python library also has the option of moving models built on Vertex AI to in-database through BYOM.
Google provides Package aiplatform, a Python SDK to interact with Vertex AI.
The following APIs introduced in the Vertex AI SDK Class Overview are supported through tdapiclient when interacting with Vertex AI.
| Google Vertex AI Interface | Classes supported through tdapiclient |
|---|---|
| Data | TabularDataset, internally created when user calls fit method. |
| Training | The following classes, internally created in the tdapiclient Vertex object wrapper class.
|
| Model | Supported |
| Prediction | Endpoint, online endpoint can be created by calling the deploy method. |
| Tracking | Experiment, internally created in create_tdapi_context method. |
The Google Vertex AI teradataml extension library (tdapiclient) includes the following functions and interfaces.