Group nodes
Group nodes are a way to convert a set of nodes into a single node, thus making the algorithm easier to re-use and share. Groups can be also setup to repeat their contents per input (loop) for memory efficiency.
Creating a group
You can create groups manually by adding the Group node into the pipeline:
Alternatively, you can also select a set of nodes and group them via a context menu action.
To edit the contents of a group, 🖱 double-click the node or open the node parameters and click the Edit graph button.
This will open a dedicated pipeline editor that by default has only two nodes Group input and Group output.
Use the + buttons to add the desired inputs and outputs to the interface nodes. The changes will be reflected in the group node:
Creating a loop
Groups can be configured to loop over their inputs, meaning if for example a group receives 10 image files, you can configure that the group's content is executed 10 times, once per image file.
To enable the loop feature, set the Graph iteration mode to Loop (single data per slot) or Loop (multiple data per slot).
Continue on failure
Similar to graph partitions, group nodes support the "Continue on failure" feature that would ignore failed loop iterations and instead create a notification.