complete classification by error
Classification via ProcessClassifier
and EmbeddingsInspector
is okay.
But to analyze failed workchains, should also have a customized sequence which classifies by exit_status
hierarchies, identifies specific error causes and counts their number of occurrences. The goal is to have a complete hierarchy of errors which occurred. The idea is not to identify all possible errors, only the ones which are present in the thesis data_generation dataset.