Deploying a Decision Optimization model by using the user interface

You can save a model for deployment in the Decision Optimization experiment UI and promote it to your Watson Machine Learning deployment space.

Before you begin

You must have a deployment space associated with your project.

About this task

When you're satisfied with its results, reliability, and performance, you can deploy a model inside Decision Optimization experiment UI with Watson Machine Learning.

The main stages for deployment are as follows:

From the Decision Optimization experiment UI, save your model scenario as a Watson Machine Learning model in your Project.
Promote your Watson Machine Learning model to your deployment space.
From your deployment space, create a new deployment.
You can then create and run jobs to your deployed model.

These stages are detailed in the following procedure.

Procedure

To save your model for deployment:

In the Decision Optimization experiment UI, either from the Scenario or from the Overview pane, click the menu icon for the scenario that you want to deploy, and select Save for deployment
Specify a name for your model and add a description, if needed, then click Next.
1. Review the Input and Output schema and select the tables you want to include in the schema.
2. Review the Run parameters and add, modify or delete any parameters as necessary.
3. Review the Environment and Model files that are listed in the Review and save window.
4. Click Save.
The model is then available in the Models section of your project.

To promote your model to your deployment space:

View your model in the Models section of your project.
You can see a summary with input and output schema. Click Promote to deployment space.
In the Promote to space window that opens, check that the Target space field displays the name of your deployment space and click Promote.
Click the link deployment space in the message that you receive that confirms successful promotion.
Your promoted model is displayed in the Assets tab of your Deployment space. The information pane shows you the Type, Software specification, description and any defined tags such as the Python version used.

To create a new deployment:

From the Assets tab of your deployment space, open your model and click New Deployment.
In the Create a deployment window that opens, specify a name for your deployment and select a Hardware specification.
Click Create to create the deployment. Your deployment window opens from which you can later create jobs.

Results

You can access information about your deployment on the Deployments tab of your model in your deployment space.

Creating and running Decision Optimization jobs

You can create and run jobs to your deployed model.

Procedure

Return to your deployment space by using the navigation path and (if the data pane isn't already open) click the data icon to open the data pane. Upload your input data tables, and solution and kpi output tables here. (You must have output tables defined in your model to be able to see the solution and kpi values.)
Open your deployment model, by selecting it in the Deployments tab of your deployment space and click New job.
Define the details of your job by entering a name, and an optional description for your job and click Next.
Configure your job by selecting a hardware specification and Next.
You can choose to schedule your job here, or leave the default schedule option off and click Next. You can also optionally choose to turn on notifications or click Next.
Choose the data that you want to use in your job by clicking Select the source for each of your input and output tables. Click Next.
You can now review and create your model by clicking Create.
When you receive a successful job creation message, you can then view it by opening it from your deployment space. There you can see the run status of your job.
Open the run for your job.
Your job log opens and you can also view and copy the payload information.

Results

You can create and monitor jobs, and get solutions by using the watson.ai Python client. See the RunDeployedModel notebook in the DO-samples. Select the relevant product and version subfolder.