Merge UCSD efforts into main by aashayarora · Pull Request #121 · cmstas/run3-vbsvvh

aashayarora · 2026-09-02T21:30:23Z

Mainly just QCD resampling, script to generate the PDFs and the relevant reader in utils.
Update the BDT weights after retraining on nominal JECs applied jet collection.

kmohrman · 2026-09-03T14:12:33Z

Thanks for preparing this PR @aashayarora!

Before getting into the details, I wanted to understand why the configs in the old configs directory are needed, is there something in particular these channels need that are not available on the main configs on the master branch?

mmfsz

Thank you Aashay! I have two comments:

On this branch the 2025 era is keyed as 2025Prompt in the C++ (corrections.cpp, weights.h, utils.cpp), but etc/dataset_names_ref.py and LUMI_DICT in etc/make_sample_jsons.py still tag the 2025 data as 2025. So anyone generating input jsons with make_sample_jsons.py on this branch gets year = "2025" in the metadata, which matches none of the 2025Prompt table entries. For JEC/JER that throws at runtime; for pileup, golden JSON, b-tagging, and the is2025 trigger define it silently falls through. Could you either switch the C++ here to 2025, or update dataset_names_ref.py, LUMI_DICT, and etc/build_trigger_logical_or.py to 2025Prompt?
Can you clarify for which MC samples and with which preselection the resampling was derived for? I left two inline comments to suggest changing the .root file names to something more descriptive. These could include also something indicating at which point the resampling is done, i.e. can one use these also in other channels, or are they specific to one channel?

kmohrman · 2026-09-09T13:40:52Z

Thanks for preparing this PR @aashayarora!

Before getting into the details, I wanted to understand why the configs in the old configs directory are needed, is there something in particular these channels need that are not available on the main configs on the master branch?

Regarding this, just for reference: As discussed on the Mattermost, we modified the way the paths and prefixes are handed (#122) so that this PR can now hopefully move to using the central common jsons.

…ing PDFs

…annels

kmohrman · 2026-09-12T21:05:24Z

-const std::string trigger_logic_string_0lep1FJ = "(is2016 && (( ((shortname==\"JetHT\") || !isData)  && ((HLT_AK8DiPFJet280_200_TrimMass30_BTagCSV_p20 == true) || (HLT_AK8PFHT700_TrimR0p1PT0p03Mass50 == true) || (HLT_AK8PFJet360_TrimMass30 == true) || (HLT_PFHT650_WideJetMJJ900DEtaJJ1p5 == true) || (HLT_PFHT650_WideJetMJJ950DEtaJJ1p5 == true) || (HLT_PFHT800 == true) || (HLT_PFHT900 == true) || (HLT_PFJet450 == true)) ))) || (is2017 && (( ((shortname==\"JetHT\") || !isData)  && ((HLT_AK8PFHT800_TrimMass50 == true) || (HLT_AK8PFJet330_BTagCSV_p17 == true) || (HLT_AK8PFJet400_TrimMass30 == true) || (HLT_AK8PFJet420_TrimMass30 == true) || (HLT_AK8PFJet500 == true) || (HLT_PFHT1050 == true) || (HLT_PFJet500 == true)) ))) || (is2018 && (( ((shortname==\"JetHT\") || !isData)  && ((HLT_AK8PFHT800_TrimMass50 == true) || (HLT_AK8PFJet330_TrimMass30_PFAK8BoostedDoubleB_np2 == true) || (HLT_AK8PFJet400_TrimMass30 == true) || (HLT_AK8PFJet420_TrimMass30 == true) || (HLT_AK8PFJet500 == true) || (HLT_PFHT1050 == true) || (HLT_PFJet500 == true)) ))) || (is2022 && (( ((shortname==\"JetMET\") || !isData)  && ((HLT_AK8PFJet250_SoftDropMass40_PFAK8ParticleNetBB0p35 == true) || (HLT_AK8PFJet425_SoftDropMass40 == true)) ))) || (is2023 && (( ((shortname==\"JetMET\") || !isData)  && ((HLT_AK8PFJet250_SoftDropMass40_PNetBB0p06 == true) || (HLT_AK8PFJet425_SoftDropMass40 == true)) ))) || (is2024 && (( ((shortname==\"JetMET\") || !isData)  && ((HLT_AK8PFJet250_SoftDropMass40_PNetBB0p06 == true) || (HLT_AK8PFJet425_SoftDropMass30 == true)) ))) || (is2025 && (( ((shortname==\"JetMET\") || !isData)  && ((HLT_AK8PFJet250_SoftDropMass40_PNetBB0p06 == true) || (HLT_AK8PFJet425_SoftDropMass30 == true)) )))";
-
+const std::string trigger_logic_string_singlelep = "(is2016 && ((((isData && (shortname == \"SingleMuon\")) || !isData) && (HLT_IsoMu24 == true || HLT_IsoTkMu24 == true)) || (((isData && (shortname == \"SingleElectron\")) || !isData) && (HLT_Ele27_eta2p1_WPTight_Gsf == true) && !(HLT_IsoMu24 == true || HLT_IsoTkMu24 == true)))) ||(is2017 && ((((isData && (shortname == \"SingleMuon\")) || !isData) && (HLT_IsoMu27 == true)) || (((isData && (shortname == \"SingleElectron\")) || !isData) && (HLT_Ele32_WPTight_Gsf_L1DoubleEG == true) && !(HLT_IsoMu27 == true)))) ||(is2018 && ((((isData && (shortname == \"SingleMuon\")) || !isData) && (HLT_IsoMu24 == true)) || (((isData && (shortname == \"EGamma\")) || !isData) && (HLT_Ele32_WPTight_Gsf == true) && !(HLT_IsoMu24 == true)))) ||((is2022 || is2023 || is2024 || is2025) && ((((isData && (shortname == \"Muon\" || shortname == \"Muon0\" || shortname == \"Muon1\")) || !isData) && (HLT_IsoMu24 == true)) || (((isData && (shortname == \"EGamma\" || shortname == \"EGamma0\" || shortname == \"EGamma1\" || shortname == \"EGamma2\" || shortname == \"EGamma3\")) || !isData) && (HLT_Ele30_WPTight_Gsf == true) && !(HLT_IsoMu24 == true))))";
+const std::string trigger_logic_string_ht  = "(is2016 && (HLT_PFHT800 || HLT_PFHT900)) || (!is2016 && HLT_PFHT1050)";


I am not sure if this trigger_logic_string_ht is exactly what we want. I think the trigger_logic_string_ht is used by 0lep_3FJ and 0lep_2FJ, and I think these channels include JetHT, MET, and SingleMuon datasets (for Run 2). I'm not sure if this trigger string actually removes the overlap among these datasets.
Could you confirm how the overlap is being handled for these channels?

for the 0lep3FJ channel, in our custom configs we were only using JetHT and JetMET. Why do we need the MET and SingleMuon datasets?

I agree those don't seem to be needed. But I think they are skimmed for that channel and the jsons exist (maybe it was for some study?) so unless there is some guard against it that I am missing, I think in the current workflow they would be run (and overlap among them would not be handled).

kmohrman · 2026-09-12T21:57:14Z

+    // calls HvsQCD / VvsQCD -- identical formulas, Xbb/(Xbb+QCD) and
+    // (Xqq/3+Xcs)/(Xqq/3+Xcs+QCD), so the templates apply unchanged. See
+    // AK8JetsSelection in selections.cpp.
+    df = df.Redefine("fatjet_Hbb", [](const RVec<RVec<float>>& hv) { return hv[0]; }, {"_qcd_resampled_hv"});


Are the fatjet_Hbb and the fatjet_Wqq the objects we want to use here? Just not sure if here it should be the capital or lower case ones in this context and how that interfaces with how the corrections are applied (maybe @mmfsz could comment)

I believe only FatJet_Hbb and FatJet_Wqq are defined in selections.cpp. I think this should fail at runtime?

df = applyObjectMaskNewAffix(df, "_good_ak8jets", "FatJet", "fatjet");
This redefines the collection with the new mask, but regardless it shouldn't matter.

Ah sorry, that is true, so it should not fail at runtime. The only issue is that it's not ideal that the fatjet collection (which is nominal JECs) ends up resampled, while the FatJet[nominal_JEC_mask] collection remains not-resampled. At the same time, you probably don't want to resample FatJet, so I think it's fine to leave it as is.

kmohrman · 2026-09-12T22:00:19Z

+
+def spec_path(run):
+    """The production spec JSON the preselection itself is run on for this CR's skim."""
+    return os.path.join(_PRESEL, "etc", "old_config", "0Lep3FJ", f"0Lep3FJ_{run}-data.json")


I think this "old_config" area is outdated now right?

This script never needs to be run again in principle, but it should be easy to configure it in case someone does need to. But yes, we have moved to using the central configs.

Ahh, ok. Could you add some comment to the top (unless it's already in some readme or comment?) that describes this, and mentions it would need to be re-configured if it needs to be run again?

kmohrman · 2026-09-12T22:11:54Z

+_HERE = os.path.dirname(os.path.abspath(__file__))      # preselection/data
+_PRESEL = os.path.dirname(_HERE)                        # preselection
+_SUFFIX = "" if RUN == "run3" else f"_{RUN}"
+OUT_ROOT = os.path.join(_HERE, f"resampling_pdfs{_SUFFIX}.root")


Is the applyQCDScoreResampling in utils expecting something like "resampling_pdfs_QCD_run2.root"? If so maybe the QCD needs to get added here (or removed there), unless the string is getting modified somewhere else I guess

kmohrman · 2026-09-12T22:22:14Z

Thanks for the PR @aashayarora! In general I think it looks pretty good. I've just left a few comments above.

mmfsz · 2026-09-15T19:27:51Z

Hi Aashay, I believe my previous comments have all been addressed. There are a few outstanding comments from Kelci and then I just wanted to ask if you could add two README.md files under the data and the bdt folders to explain what the scripts are and give instructions of how to reproduce your results? Other than that, I am happy for this to be merged.

aashayarora added 3 commits September 1, 2026 16:18

add qcd resampling

56a8e3f

minor bugfix

ac24b48

update resampling PDFs after JECs

34d0c43

aashayarora requested review from kmohrman and mmfsz September 2, 2026 21:30

Delete misc script

2e3d872

Merge remote-tracking branch 'origin' into ucsd-presel

43883e2

mmfsz reviewed Sep 9, 2026

View reviewed changes

Comment thread preselection/data/resampling_pdfs_QCD_run3.root

mmfsz reviewed Sep 9, 2026

View reviewed changes

Comment thread preselection/data/resampling_pdfs_QCD_run2.root

mmfsz reviewed Sep 9, 2026

View reviewed changes

aashayarora and others added 7 commits September 10, 2026 09:47

update BDT training for new nominal JECs applied jets, rename resampl…

4283c92

…ing PDFs

Merge remote-tracking branch 'origin/ucsd-presel' into ucsd-presel

7604761

update all 2025Prompts to 2025 to match scheme in main

ac9980b

Merge branch 'main' into ucsd-presel

fc13d1c

update weights again, add bdt training script

aab3e7c

remove old_config, switch to using same config generation as other ch…

608a000

…annels

fix dangling artifacts

02903b4

kmohrman reviewed Sep 12, 2026

View reviewed changes

Comment thread preselection/src/selections.cpp

Uh oh!

Conversation

aashayarora commented Sep 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

kmohrman commented Sep 3, 2026

Uh oh!

Uh oh!

Uh oh!

mmfsz left a comment

Choose a reason for hiding this comment

Uh oh!

kmohrman commented Sep 9, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

kmohrman commented Sep 12, 2026

Uh oh!

mmfsz commented Sep 15, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

aashayarora commented Sep 2, 2026 •

edited

Loading