Deploy and manage models and functions

Using Watson Machine Learning, you can deploy models and solutions so that you can put them into productive use, then monitor the deployed assets for fairness and explainability. You can also automate the AI lifecycle to keep your machine learning assets current. Deployment is the final stage of the lifecycle of a model or script, where you run your models and code. Watson Machine Learning provides the tools that you need to deploy an asset, such as a machine learning model or function, or a Decision Optimization solution.

Depending on what is installed and configured for your deployment, you can:

• Build, train, and deploy models from notebooks by using the Watson Machine Learning Python client library or the Watson Machine Learning API.
• Create AutoAI experiments. AutoAI automatically preprocesses your structured data, selects the best estimator for the data, and then generates model candidate pipelines for you to review and compare. Deploy the best performing pipeline as a machine learning model.
• Run experiments to train complex models in Experiment builder.
• Deploy your models so that you can score the models and generate predictions.

Deployment spaces

Deployment spaces contain deployable assets, deployments, deployment jobs, associated input and output data, and the associated environments. You can use spaces to deploy models and manage your deployments. The deployments dashboard provides an aggregate view of deployment activity available to you, across spaces.

For more information on deployment spaces, see Deployment spaces (Watson Machine Learning).

Adding assets to a space

The way to add an asset to a space depends on the asset type. In the case of some assets, you can add them directly to a space. Other asset types originate in a Watson Studio project and must be transferred from a project to a space. Third class are asset types that you can only add to a space as a dependency of another asset (these asset types are not directly exposed in the Assets tab in the UI). To get more information on adding assets to a space, see Adding assets to a deployment space.

Deploying assets

For proper deployment, you must set up a deployment space and then select and configure a specific deployment type. Refer to Deploying assets for information on:

• Creating an online deployment
• Creating a batch deployment
• Deploying Python functions
• Deploying NLP models
• Creating a deployment job
• Getting the deployment endpoint URL

Managing deployments

After you deploy assets, you can manage and update them to make sure they are performing well and to monitor their accuracy. Some ways that you can manage or update a deployment include:

• Managing deployment jobs. After you create one or more jobs, you can view and manage them from the Jobs tab of your deployment space.
• Updating a deployment. You can replace a model with a better-performing version without having to create a new deployment.
• Scaling a deployment to increase availability and throughput by creating replicas of the deployment.
• Evaluating a deployment for bias by using Watson OpenScale.
• Deleting a deployment to remove a deployment and free up resources.

Summary

This section provided a brief overview of deploying and managing models with IBM Cloud Pak for Data. For more information on this topic, see Deploying and managing models.