Getting started with the Orchestration Pipelines editor
The Pipelines editor is a graphical canvas where you can drag and drop nodes that you connect together into a pipeline for automating machine model operations.
You can open the Pipelines editor by creating a new Pipelines asset or editing an existing Pipelines asset. To create a new asset in your project from the Assets tab, click New asset > Automate model lifecycle. To edit an existing asset, click the pipeline asset name on the Assets tab.
The canvas opens with a set of annotated tools for you to use to create a pipeline. The canvas includes the following components:
- The node palette provides nodes that represent various actions for manipulating assets and altering the flow of control in a pipeline. For example, you can add nodes to create assets such as data files, AutoAI experiments, or deployment spaces. You can configure node actions based on conditions if files import successfully, such as feeding data into a notebook. You can also use nodes to run and update assets. As you build your pipeline, you connect the nodes, then configure operations on the nodes to create the pipeline. To organize nodes visually, you can also customize the size of nodes by dragging node corners with your mouse with the exception of some, such as loop nodes. These pipelines create a dynamic flow that addresses specific stages of the machine learning lifecycle.
- The toolbar includes shortcuts to options related to running, editing, and viewing the pipeline.
- The parameters pane provides context-sensitive options for configuring the elements of your pipeline.
The toolbar
Use the Pipeline editor toolbar to:
- Run the pipeline as a trial run or a scheduled job
- View the history of pipeline runs
- Cut, copy, or paste canvas objects
- View the edit history of your pipeline and undo or redo actions
- Save the pipeline
- When you save, you can see all the users who is viewing the flow
- Delete a selected node
- Drop a comment onto the canvas. The comments accept basic styling of:
- Basic HTML/CSS syntax, such as
<div>,<span>,<p>,<font>,<size>,<font-style>,<font-clour>,<text-align>,<background-color>. - Basic Markdown syntax, such as
**(bold),_(italic),#(header).
- Basic HTML/CSS syntax, such as
- Drop an annotation on the canvas. The annotation supports HTML with its built-in styling toolbar and does not accept HTML coding.
The styling of any comments are preserved when migrating pipelines, for example when migrating a flow to a new project or recreating it in a DataStage environment.
- Configure global objects, such as pipeline parameters or user variables
- Manage default settings
- Arrange nodes vertically
- View last saved timestamp
- Zoom in or out
- Fit the pipeline to the view
- Show or hide global messages
Hover over an icon on the toolbar to view the shortcut text.
The node palette
The node palette provides the objects that you need to create an end-to-end pipeline. Click a top-level node in the palette to see the related nodes.
| Node category | Description | Node type |
|---|---|---|
| Copy | Use nodes to copy an asset or file, import assets, or export assets | Copy assets Export assets Import assets |
| Create | Create assets or containers for assets | Create AutoAI experiment Create AutoAI time series experiment Create batch deployment Create data asset Create deployment space Create online deployment |
| Wait | Specify node-level conditions for advancing the pipeline run | Wait for all results Wait for any result Wait for file |
| Control | Specify error handling | Loop in parallel Loop in sequence Set user variables Terminate pipeline |
| Update | Update the configuration settings for a space, asset, or job. | Update AutoAI experiment Update batch deployment Update deployment space Update online deployment |
| Delete | Remove a specified asset, job, or space. | Delete AutoAI experiment Delete batch deployment Delete deployment space Delete online deployment |
| Run | Run an existing or ad hoc job. | Run AutoAI experiment Run Bash script Run batch deployment Run Data Refinery job Run DataStage job Run notebook job Run pipeline job Run Pipelines component job Run SPSS Modeler job |
The parameters pane
Double-click a node to edit its configuration options. Depending on the type, a node can define various input and output options or even allow the user to add inputs or outputs dynamically. You can define the source of values in various ways. For example, you can specify that the source of value for "ML asset" input for a batch deployment must be the output from a run notebook node.
For more information on parameters, see Configuring pipeline components.
Next steps
Parent topic: Pipelines