Model deployments are a managed resource within the Oracle Cloud Infrastructure (OCI) Data Science service. They allow you to deploy machine learning models as web applications (HTTP endpoints). They provide real-time predictions and enables you to quickly productionalize your models.

The ads.model.deployment module allows you to deploy models using the Data Science service. This module is built on top of the oci Python SDK. It is designed to simplify data science workflows.

A model artifact is a ZIP archive of the files necessary to deploy your model. The model artifact contains the score.py file. This file has the Python code that is used to load the model and perform predictions. The model artifact also contains the runtime.yaml file. This file is used to define the conda environment used by the model deployment.

ADS supports deploying a model artifact from the Data Science model catalog, or the URI of a directory that can be in the local block storage or in Object Storage.

You can integrate model deployments with the OCI Logging service. The system allows you to store access and prediction logs ADS provides APIs to simplify the interaction with the Logging service, see ADS Logging.

The ads.model.deployment module provides the following classes, which are used to deploy and manage the model.

  • ModelDeployer: It creates a new deployment. It is also used to delete, list, and update existing deployments.

  • ModelDeployment: Encapsulates the information and actions for an existing deployment.

  • ModelDeploymentProperties: Stores the properties used to deploy a model.