chaeo workflows and recently created ilastik ones have a consistent first few arguments: file paths, list of models, etc.; and these are expected by batch running utility others do not and so they error in batch mode
ilastik workflows return a mix of manual dict and WorkflowRunRecord objects
Making progress on this in context of workflows involving the ZMask / RoiSet class on chaeo_zmask_class branch; this becomes more straightforward when workflows compose models with generic functionality e.g. semantic segmentation. Parameters then specific to model implementation e.g. ilastik are then contained in JSON-serializable pydantic objects.
Developing on dev_pipeline_refactor: design template involving inner pipeline call, parameter PyDantic model(s), and API wrapper(s) in a single module
added selected label
assigned to @rhodes
removed selected label
added staged label
New pipeline architecture standardizes this more than before, closed with: !58 (merged)
closed