ClearScape Analytics ModelOps product automates and governs the model deployment, schedule, and monitoring process in an industrialized and repeatable fashion. Our focus is in the following areas:
- Model Deployment: Once the model code is developed using end-user tool and language of choice, ModelOps provides a simple and intuitive way to operationalize the model with Vantage.
- By deploying in Vantage, we can leverage in-database scale with no extra technology or infrastructure investment.
- Deploying in Edge container might be useful for some use cases where model execution is required to be done in specific environment or device.
- Model Lifecycle: With ModelOps, the model implementation process is tracked from training, evaluation, approval to deployment and retirement. We permit the auditability of the model getting access to the end-to-end metadata of the process.
- Model Governance: We organize the information of the models in projects, allowing the different Data Science departments control access to their models.
- Model Monitoring: When models are deployed, ModelOps provides best-in-class capabilities to understand model metrics and datasets/features drift. Data Science teams and Machine Learning operations can get alerts and act early on poorly performing and decaying models.
ModelOps can work with both Bring your Own Models (BYOM)/third-party models and Git models using the Teradata modelops (tmo) command-line-tool published in https://pypi.org/project/teradatamodelops/:
- BYOM/third-party models
In ModelOps, data scientist can operationalize models that have been trained externally. Data scientists convert these models to one of the supported BYOM formats (PMML, ONNX, H2O, SAS, and so on) and operationalize with ModelOps, leveraging all the capabilities of ModelOps, with a different lifecycle ( ).
You can find the supported BYOM formats in the Software and Compatibility section of the Teradata Vantage™ - Bring Your Own Model User Guide, B700-1111 - Git models
Git models refer to the models with their training, evaluation, and scoring files stored in git repository and for which we also manage the training stage of their lifecycle (
). This is different from the lifecycle of BYOM models.