Dynamic Reduction Network-based electron energy regression using the SONIC service #35839
Conversation
|
-code-checks Logs: https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-35839/26205
Code check has found code style and quality issues which could be resolved by applying following patch(s)
|
please cleanup the commit history to remove the intermediate files (especially the large ones) also, I'd rather have updates in .gitignore in a separate PR (if they are really needed). So, please remove it from this PR. |
cmsbuild
added
code-checks-pending
and removed
core-pending
code-checks-rejected
labels
|
-code-checks Logs: https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-35839/26222
Code check has found code style and quality issues which could be resolved by applying following patch(s)
|
| #ifndef Progression_EGM_DRN_SCEnergyCorrectorDRN_h | ||
| #define Progression_EGM_DRN_SCEnergyCorrectorDRN_h | ||
|
|
||
| #include "HeterogeneousCore/SonicTriton/interface/TritonEDProducer.h" |
| void setEventSetup(const edm::EventSetup& es); | ||
| void setEvent(const edm::Event& e); | ||
|
|
||
| //std::pair<double, double> getCorrections(const reco::SuperCluster& sc) const; |
There was a problem hiding this comment.
delete commented-out code
| void makeInput(const edm::Event& iEvent, TritonInputMap& iInput, const reco::SuperClusterCollection& inputSCs) const; | ||
| TritonOutput<float> getOutput(const TritonOutputMap& iOutput); | ||
|
|
||
| //void modifyObject(reco::SuperCluster& sc) const; |
There was a problem hiding this comment.
delete commented-out code
|
|
||
| #include <vdt/vdtMath.h> | ||
|
|
||
| #include <iostream> |
|
|
||
| void SCEnergyCorrectorDRN::makeInput(const edm::Event& iEvent, TritonInputMap& iInput, const reco::SuperClusterCollection& inputSCs ) const { | ||
|
|
||
| std::vector<unsigned> nHits = {}; |
There was a problem hiding this comment.
unnecessary default initialization
| timeout = cms.untracked.uint32(10), | ||
| modelName = cms.string('MustacheEE'), | ||
| modelVersion = cms.string(""), | ||
| modelConfigPath = cms.FileInPath("DRNData/models/MustacheEE/config.pbtxt"), |
RecoEcal/EgammaClusterProducers/python/SCEnergyCorrectorDRNProducer_cfi.py
Outdated
Show resolved
Hide resolved
| static float sigmoid(float x){ | ||
| return 1.0f/(1.0f + exp(-x)); | ||
| } | ||
| static float ln2 = log(2); |
There was a problem hiding this comment.
since this is only used inside the static function logcorrection below, just define it inside that function
| corrSCs->push_back(inputSC); | ||
| corrSCs->back().setEnergy(corrEn); | ||
| corrSCs->back().setCorrectedEnergy(corrEn); | ||
| //corrSCs->back().setCorrectedEnergyUncertainty(-1.0); |
There was a problem hiding this comment.
delete commented-out code
| @@ -0,0 +1,560 @@ | |||
| // -*- C++ -*- | |||
There was a problem hiding this comment.
Virtually every line of code in this file violates CMS coding rules: http://cms-sw.github.io/cms_coding_rules.html
If this really needs to be included in the official repository, please clean it up.
ssrothman
force-pushed
the
port_to_egm
branch
from
f53242d to
45b7741
Compare
|
-code-checks Logs: https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-35839/26224
Code check has found code style and quality issues which could be resolved by applying following patch(s)
|
ssrothman
force-pushed
the
port_to_egm
branch
from
45b7741 to
f53242d
Compare
|
-code-checks Logs: https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-35839/26227 ERROR: Build errors found during clang-tidy run. |
|
kind ping on this |
cmsbuild
added
code-checks-pending
tests-pending
and removed
tests-approved
code-checks-approved
labels
|
+code-checks Logs: https://cmssdt.cern.ch/SDT/code-checks/cms-sw-PR-35839/27992 |
|
Pull request #35839 was updated. @jpata, @cmsbuild, @clacaputo, @slava77 can you please check and sign again. |
|
@cmsbuild please test |
|
+1 Summary: https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-d7e2d0/22099/summary.html The following merge commits were also included on top of IB + this PR after doing git cms-merge-topic:
You can see more details here: Comparison SummarySummary:
|
|
+reconstruction
|
cmsbuild
added
fully-signed
reconstruction-approved
and removed
reconstruction-pending
pending-signatures
labels
|
This pull request is fully signed and it will be integrated in one of the next master IBs (tests are also fine). This pull request will now be reviewed by the release team before it's merged. @perrotta, @dpiparo, @qliphy (and backports should be raised in the release meeting by the corresponding L2) |
|
+1 |
PR description:
This PR commits the first tranche of code to apply the newly developed Dynamic Reduction Network (DRN)-based energy regression for electrons implemented using pytorch. This commit is intended for the ECAL calibration use case where the energy of electron Mustache superclusters are corrected. This does not require to be added to the full reconstruction workflow and can be applied on the fly when the necessary nTuples needed for ECAL calibration are produced using the ECALELF framework.
The DRN model is implemented in CMSSW (here and here) using the SONIC service, modifying the existing implementation of the BDT based regression and the corresponding producer.
The SONIC config file and torchscript regression weights needed are intended to be hosted on git cms-data. A PR will be created there committing these files referencing the current PR here.
A summary of this DRN model as well as the rationale of the CMSSW implementation may be found here.
This is a draft PR pending some careful checks of resource usage. In particular, prior to submitting the full PR we will
PR validation:
Validation has been performed using a custom cfg intended to be run using cmsRun that has been also committed in the test folder (here). Pending the cms-data PR, the config and weights files must be cloned from https://github.com/ssrothman/DRNData in order to run the validation config. A histogram of the predicted energy over the generated energy for both the DRN and the 2018 UL BDT (as well as the raw energy) is embedded here:
It has also been confirmed that this implementation yields identical results to the standalone python implementation used for training and validation of the DRN, such as in the slides linked above.
if this PR is a backport please specify the original PR and why you need to backport that PR:
As of now a backport is not envisaged, but might be re-visited later.
@Sam-Harper @lgray @kpedro88 @rchatter @lfinco @swagata87 @jainshilpi @SohamBhattacharya @simonepigazzini @violatingcp