What is the point of AzureML modules?

Question

Hello!

I created an Azure ML pipeline in Python and used multiple PythonScriptSteps for each of my tasks. For example, I have three training steps running in parallel, so I create three PythonScriptSteps in a for loop with my train.py script and different data. Later, I came across the ModuleStep, which seems to do exactly this, but with an extra layer of (seemingly pointless) abstraction. What does the ModuleStep add to a PythonScriptStep?

Also, I imagined the ModuleStep might make it possible to use a custom PythonScriptStep in the pipeline designer (by creating a new drag and drop module), however this doesn't seem to be the case. Is there any way of doing this?

Answer 1

Just to close this question, I have since discovered that the ModuleStep does create a custom drag-and-drop module in the designer. I don't know if I'd missed this (I imagine so) or if this is a new feature. Either way, that's the answer. @YutongTie-MSFT can you confirm if this was recently added?