We're using PDG very successfully on 1,000's of jobs in Deadline and it's running very smoothly.
One of the tasks is a Python script that runs on a worker "by itself" (i.e. nothing else runs on that worker or even that physical machine) but currently it seems like the Deadline job runs Hython which then runs the script. This seems.... needlessly expensive? It's just a Python script
Am I missing something vis a vis having just the script run without Hython? I'm using a Python Script TOP, but I wonder if I need to use the Python Processor? The latter I don't really understand at all currently
The Python Script generates one work item for each input work item, and configures that work item to run the script code you specified on the node. The Python Processor node on the other hand allows you to actually define how work items are generated, e.g. you can generate 3 work items from each input, or a random number of work items for each input work item, or some custom logic based on parameters. It's useful if you want to e.g. fan out the number of work items you have, as opposed to the Python Script which is always just a one-to-one relationship with the input.
As for running with hython, the Python Script Python Executable section that you can use to customize which Python executable is used to run the script, if the node is running work items out of process. By default it uses hython, but you can tell it use plain Python or a custom path instead.
