Due to the structure of our workflow (using cf.MLEvaluation to create categories in cf.ProduceColumns), it currently happens that cf.MLEvaluation runs once without and once with requiring this particular category Producer.
This issue might be solved by removing the MLModel from the command line when calling cf.CreateDatacards (which is resolved automatically at the moment). However, this also means that the ML output columns would not be loaded and that the MLTraining init might not be running.
Solutions:
add the produced columns from the MLTraining both in uses and produces of the category Producer
run the MLTraining init function in the category Producer init aswell
Due to the structure of our workflow (using cf.MLEvaluation to create categories in cf.ProduceColumns), it currently happens that cf.MLEvaluation runs once without and once with requiring this particular category Producer.
This issue might be solved by removing the MLModel from the command line when calling cf.CreateDatacards (which is resolved automatically at the moment). However, this also means that the ML output columns would not be loaded and that the MLTraining init might not be running. Solutions: