Error of jetmet_tools for Run3 JEC files in coffea 0.7.22

[ParticleChef]

Dear experts,
I am trying to apply JEC, JER corrections using txt files from twiki.
I can make the corrections for jecs like L1, L2, L3 but I cannot load the Uncertainty file. I found that the format of name is different.
Names must be formatted as ''. is in coffea/jetmet_tools/JetCorrectionUncertainty.py file but name of jec for run3 has one more '_'.
And also I tried changing name of .txt file but it has same error like this.

Traceback (most recent call last): File "check_jecs2.py", line 111, in <module> print("function jet_factory: ", jet_factory("2022pre","AK4PFPuppi", False)) File "check_jecs2.py", line 82, in jet_factory return jet_factory_factory(jec_files(year, jec_algo, dojer), year) File "check_jecs2.py", line 78, in jet_factory_factory jec_stack = JECStack(ext.make_evaluator()) File "/home/jhong/.local/lib/python3.8/site-packages/coffea/jetmet_tools/JECStack.py", line 58, in __init__ self._junc = JetCorrectionUncertainty( File "/home/jhong/.local/lib/python3.8/site-packages/coffea/jetmet_tools/JetCorrectionUncertainty.py", line 84, in __init__ raise Exception( Exception: Summer22_22Sep2023_V2_MC_Uncertainty_AK4PFPuppi is a <class 'coffea.lookup_tools.jme_standard_function.jme_standard_function'> and not a jec_uncertainty_lookup!

I use the jetfactory code like this:
`
def jet_factory_factory(files, year):
ext = extractor()
directory='data/JetMETCorr/'+year
for filename in files:
ext.add_weight_sets([f"* * {directory+'/'+filename}"])
ext.finalize()
jec_stack = JECStack(ext.make_evaluator())
return CorrectedJetsFactory(jec_name_map, jec_stack)

def jet_factory(year, jec_algo, dojer):
return jet_factory_factory(jec_files(year, jec_algo, dojer), year)

`

I attached one uncertainty file I used

Summer22_22Sep2023_V2_MC_Uncertainty_AK4PFPuppi.txt

Desktop (please complete the following information):

• OS: linux
• Version Centos7

[nsmith-]

In #1134 there is a WIP implementation of correctionlib-supported jetmet tools in v0.7.x. Seeing this through is more likely to be long-term useful than patching the text file parser. On the other hand, also in CalVer we have no correctionlib-based jetmet tools in coffea proper. There are examples in https://gitlab.cern.ch/cms-analysis-corrections/JME/examples/-/blob/latest/jercExample.py?ref_type=heads of doing it by hand.

@anpicci @NJManganelli @garvitaa @ikrommyd and others, is there an existing run 3 JME recipe?

[nsmith-]

Also, this is a duplicate of #1111

[ikrommyd]

Here's what we currently do in HiggsDNA (CMS Hgg analyses): https://gitlab.cern.ch/cms-analysis/general/HiggsDNA/-/blob/master/higgs_dna/systematics/jet_systematics_json.py?ref_type=heads

[nsmith-]

I tracked down also the PocketCoffea implementation of correctionlib-based JEC: https://gitlab.cern.ch/cms-analysis/general/PocketCoffea/-/blob/main/pocket_coffea/lib/jets.py?ref_type=heads#L439
Note: this is based on coffea 0.7.x so likely you could use PocketCoffea or parts of it to do your JECs quite easily without moving to CalVer.
cc @valsdav

[ikrommyd]

I tracked down also the PocketCoffea implementation of correctionlib-based JEC: https://gitlab.cern.ch/cms-analysis/general/PocketCoffea/-/blob/main/pocket_coffea/lib/jets.py?ref_type=heads#L439 Note: this is based on coffea 0.7.x so likely you could use PocketCoffea or parts of it to do your JECs quite easily without moving to CalVer. cc @valsdav

0.7.x is applicable very easily to calver virtual mode. (mostly zero changes)

[valsdav]

I tracked down also the PocketCoffea implementation of correctionlib-based JEC: https://gitlab.cern.ch/cms-analysis/general/PocketCoffea/-/blob/main/pocket_coffea/lib/jets.py?ref_type=heads#L439 Note: this is based on coffea 0.7.x so likely you could use PocketCoffea or parts of it to do your JECs quite easily without moving to CalVer. cc @valsdav

Ehi, I think that's in sync with the higgDNA implementation. It is currently tested against all the other frameworks here https://gitlab.cern.ch/cms-analysis/general/analysis_recipes/

[anpicci]

Hi all, I'm using #1134 for a Run3 analysis, and I don't see crashes.

Btw, sorry for not following up on the original PR for so long. If there is a general interest in closing #1134 for coffea 0.7, let me know, and I will allocate some time to do so.

Thanks,
Andrea

[nsmith-]

It would be good to wrap up #1134
For CalVer we might consider pointing to https://cms-analysis.docs.cern.ch/guidelines/snippets/corrections/#jet-energy-scale-corrections when it is eventually populated