Open AndriusVaitkus97 opened 9 months ago
Can you check whether it's the initial or final hists that are slowing things down? Tagging @IvanOleksiyuk in case he has any ideas
Can you check whether it's the initial or final hists that are slowing things down? Tagging @IvanOleksiyuk in case he has any ideas
I don't how to exactly check that, but I looked at the timestamps of when the plots are created. I have this:
-rw-rw-r-- 1 avaitkus avaitkus 233K Feb 9 22:09 initial_abs_eta.png
-rw-rw-r-- 1 avaitkus avaitkus 241K Feb 9 23:41 initial_mass.png
-rw-rw-r-- 1 avaitkus avaitkus 214K Feb 9 20:32 initial_ptlow.png
-rw-rw-r-- 1 avaitkus avaitkus 230K Feb 9 19:16 initial_pt.png
-rw-rw-r-- 1 avaitkus avaitkus 113K Feb 9 23:41 train_abs_eta.png
-rw-rw-r-- 1 avaitkus avaitkus 115K Feb 9 23:41 train_mass.png
-rw-rw-r-- 1 avaitkus avaitkus 98K Feb 9 23:41 train_ptlow.png
-rw-rw-r-- 1 avaitkus avaitkus 106K Feb 9 23:41 train_pt.png
It seems to be that the initial plots are the ones taking ages, given that the start time was 16:17.
UPD: Actually the initial plots also are really weird, why are there so many categories?
Built-in plotting with UPP runs really slow for me. I ran
preprocess --config configs/xbb-avaitkus.yaml --plots
on this commit: https://github.com/umami-hep/umami-preprocessing/commit/7099a2980baade4f1cc7a98ebc2538f3d6ebd9c0, on the train file with 60M jets. Below is the log file.As you can see it took me over 7 hours. @zcapjdb has a similar issue