12.8 Publishing a Pipeline

If your satisfied with the results of the execution you can publish the pipeline. Publish the pipeline will backup the current model pipeline with non editable mode. To publish the pipeline, follow these steps:

  1. Click Launch Workspace This icon shows the launch workspace. next to corresponding Workspace to Launch Workspace and display the Dashboard window with application configuration and model creation menu.
  2. This icon shows the model pipeline menu.In the Mega menu, click Modeling and select Pipelines from the drop down to display the

    Model Pipeline window.

  3. Select the Objective from the list. The publish canvas is displayed.
  4. Click Publish. Publish Pipeline pop-up is displayed.
  5. Enter the field details. See the following table.

    Table 12-6 Publish Pipeline

    Field or Icon Description
    Model Name The field displays the name of the Model. Modify the name if required.
    Model Description The field displays the description for the Model. Enter or modify the description if required.
    Technique Enter the registered technique to use.
    Run Version Select a rub version.
    Variable Mapping The table displays the OFSAA variables and datasets used in the creation of the Training Model.
    Script

    The table displays the Paragraphs created in the Training Model. Select the Paragraphs that you want to use to create the Scoring Model.

    Track Output - Select this to track the output of the paragraph.
  6. Select the required configuration and click Publish to publish the pipeline or click Cancel to go back to previous page.

To view the published model follow the subsequent steps:

  1. Navigate to Model Pipeline page
  2. Click Models in-line with the Object Name. published models are displayed.

    Figure 12-33 Published Model


    Published Model

The published models are then deployed in production to be consumed by users. The iterations of comparison between various models lead to the elimination of undesired models and the filter of a few robust ones that can be considered for deployment in production. Modelers then use their better judgment to consolidate their choice and fix on one model - the champion model. The champion model is also called the scoring model or the actual model in this document.