Dear ROOTers,
I am experiencing an issue with some input variable distributions when training some methods with the kFolding procedure using TMVA on lxplus EL9 and ROOT v6.30 (attached picture inputvariables_el9.png). The distributions of input variables are peaking on some specific anomalous values for some kFolds (not all). This strange behaviour is not observed on CENTOS7 machines and ROOT v6.24 (attached picture inputvariables_lxplus7.png).
During the training on EL9 machine, an error message similar to the following one appears in the log of the kFold presenting the strange distributions reported above:
Error in <TNetXNGFile::ReadBuffers>: [ERROR] Server responded with an error: [3027] readv past EOF
Error in <TBranch::GetBasket>: File: root://eoslhcb.cern.ch//eos/lhcb/user/t/tfulghes/output/Run2/MVA/B2KstTauE_TauTo3Pi_CombMVA_kFolding_0.root at byte:0, branch:ln_B_ENDVERTEX_CHI2, entry:7974, badread=1, nerrors=1, basketnumber=1
Error in <TBranch::GetBasket>: File: root://eoslhcb.cern.ch//eos/lhcb/user/t/tfulghes/output/Run2/MVA/B2KstTauE_TauTo3Pi_CombMVA_kFolding_0.root at byte:0, branch:ln_B_ENDVERTEX_CHI2, entry:326934, badread=0, nerrors=2, basketnumber=41
Error in <TBranch::GetBasket>: File: root://eoslhcb.cern.ch//eos/lhcb/user/t/tfulghes/output/Run2/MVA/B2KstTauE_TauTo3Pi_CombMVA_kFolding_0.root at byte:0, branch:ln_tau_ENDVERTEX_CHI2, entry:326934, badread=0, nerrors=3, basketnumber=41
Error in <TBranch::GetBasket>: File: root://eoslhcb.cern.ch//eos/lhcb/user/t/tfulghes/output/Run2/MVA/B2KstTauE_TauTo3Pi_CombMVA_kFolding_0.root at byte:0, branch:ln_Kst_ORIVX_FDCHI2, entry:326934, badread=0, nerrors=4, basketnumber=41
Error in <TBranch::GetBasket>: File: root://eoslhcb.cern.ch//eos/lhcb/user/t/tfulghes/output/Run2/MVA/B2KstTauE_TauTo3Pi_CombMVA_kFolding_0.root at byte:0, branch:ln_tau_ORIVX_FDCHI2, entry:326934, badread=0, nerrors=5, basketnumber=41
Error in <TBranch::GetBasket>: File: root://eoslhcb.cern.ch//eos/lhcb/user/t/tfulghes/output/Run2/MVA/B2KstTauE_TauTo3Pi_CombMVA_kFolding_0.root at byte:0, branch:B_FitPV_kstedtf_s2_M, entry:326934, badread=0, nerrors=6, basketnumber=41
Error in <TBranch::GetBasket>: File: root://eoslhcb.cern.ch//eos/lhcb/user/t/tfulghes/output/Run2/MVA/B2KstTauE_TauTo3Pi_CombMVA_kFolding_0.root at byte:0, branch:ln_B_ENDVERTEX_CHI2, entry:326935, badread=0, nerrors=7, basketnumber=41
Error in <TBranch::GetBasket>: File: root://eoslhcb.cern.ch//eos/lhcb/user/t/tfulghes/output/Run2/MVA/B2KstTauE_TauTo3Pi_CombMVA_kFolding_0.root at byte:0, branch:ln_tau_ENDVERTEX_CHI2, entry:326935, badread=0, nerrors=8, basketnumber=41
Error in <TBranch::GetBasket>: File: root://eoslhcb.cern.ch//eos/lhcb/user/t/tfulghes/output/Run2/MVA/B2KstTauE_TauTo3Pi_CombMVA_kFolding_0.root at byte:0, branch:ln_Kst_ORIVX_FDCHI2, entry:326935, badread=0, nerrors=9, basketnumber=41
Error in <TBranch::GetBasket>: File: root://eoslhcb.cern.ch//eos/lhcb/user/t/tfulghes/output/Run2/MVA/B2KstTauE_TauTo3Pi_CombMVA_kFolding_0.root at byte:0, branch:ln_tau_ORIVX_FDCHI2, entry:326935, badread=0, nerrors=10, basketnumber=41
Before the training, I checked that the input files used for the signal and background proxies were not corrupted, and no one seemed to have a problem.
Do you have any hint regarding this strange behaviour, or was it already observed?
Many thanks in advance!
inputvariables_el9.png:
inputvariables_lxplus7.png:
