To start updating models in your system with Machine Learning Operations (MLOps), configure the appropriate access rights for data scientist operators, update Prediction Studio settings, and ensure that your Business Change pipeline meets the requirements.
Data scientist operators
Configure the following settings for the operator IDs of the data scientists who update models:
Ensure that the operator's default access context is the same as the application context in which the predictions are updated.
For example, if the prediction is updated in the CDHSample application, then the default access group for the operator must be CDHSample.
For more information, see Defining operator contact information and application access.
- Optional: If you want to enable email notifications for the model
- Add the notification email to the operator ID.
Configure the operator ID with a default access group that includes the PegaRULES:PegaAPI access role.
For example, if the default access group is CDHSample:Administrators, then that access group must include the PegaRULES:PegaAPI access role.
For more information, see Defining work routing settings for an operator.
Prediction Studio settings
Enable the model update feature by configuring the following artifacts, and then adding them to the Prediction Studio settings:
- An analytics repository for storing machine learning models.
- Optional: A work queue for data scientists if you want to use a different work queue than the default DataScientistWorkqueue.
- Optional: An email account if you want to enable email notifications about model changes.
For more information, see Enabling Machine Learning Operations.
Business Change pipeline
Ensure that the Business Change pipeline meets the following requirements:
- All instances (Development, Sandbox, BOE, Production) are on the same versions of Pega Platform and Pega Customer Decision Hub.
- Pega 1:1 Operations Manager and Prediction Studio are on the Business Operations Environment (BOE) instance of your application.
- The models have the same names in all instances.
- The predictors for the models are configured in all instances and are in sync.
Previous topic Updating active models in predictions with MLOps Next topic Replacing models in predictions with MLOps