use_current_task
is False by default, let me see if this helps
well, it just shoved the dataset files with cryptic data_1/data_2 names among the artifacts.
I think you can simply save the creating task ID as a configuration parameter (I'm just thinking out loud)