Hi All,
This MR updates the TauGNN to use the SaltModel EDM loading infrastructure.
Detail of the motivation and discussions can be found in https://indico.cern.ch/event/1595490/
The following tests has been performed on 9490 events in data25_13p6TeV.00507349.physics_Main.merge.AOD.f1633_m2272/data25_13p6TeV.00507349.physics_Main.merge.AOD.f1633_m2272._lb0152._0002.1, running AOD->DAOD_PHYS production
-
CPU overhead estimation with this MR. -
Larger scale validation beyond CI.
The total CPU time for TauJets_TauIDDecorKernel went from 154.004s to 153.184s, which is a 0.5% reduction.
| TauJets_TauIDDecorKernel | main | This MR |
|---|---|---|
| Count | 9489 | 9489 |
| cpuTime | 154004 | 153184 |
| malloc | 2223635 | 2223635 |
| wallTime | 154936 | 155166 |
diff-root between the two resulting DAOD_PHYS files shows no difference:
acmd diff-root run-main/DAOD_PHYS.myOutput.pool.root run-MR/DAOD_PHYS.myOutput.pool.root --nan-equal
...
Py:diff-root INFO comparing over [9490] entries...
Py:diff-root INFO comparing [1068] leaves over entries...
Py:diff-root INFO Found [35636139] identical leaves
Py:diff-root INFO Found [0] different leaves
Py:diff-root INFO all good.Undraft to invite comments.
Edited by Qichen Dong