In data_fetcher.py, currently all of prompt reco and re-reco files for a particular run, for all versions of re-reco are taken for training.
A partial fix is included in https://github.com/AutoDQM/AutoDQM_ML/pull/34 which ignores prompt reco and only selects the first re-reco file version found for a given run (version DQM_V000X for all X are grabbed, but reco DDMMMYYYY-vXZ for only Z=1 taken).
Likely affects the AutoDQM testbed too, but at least is not handled correctly for AutoDQM-ML
In data_fetcher.py, currently all of prompt reco and re-reco files for a particular run, for all versions of re-reco are taken for training.
A partial fix is included in https://github.com/AutoDQM/AutoDQM_ML/pull/34 which ignores prompt reco and only selects the first re-reco file version found for a given run (version DQM_V000X for all X are grabbed, but reco DDMMMYYYY-vXZ for only Z=1 taken).
Likely affects the AutoDQM testbed too, but at least is not handled correctly for AutoDQM-ML