Testing payload for new keras saving format. The mvas themselves are the same as the previous iteration (at least at the time of creation).

ECLChargedPIDMVA payload containing the trained BDTs with HadronFrac and the version of v8 of the CNN (with HadronFrac) that has been retrained with release 9 (new keras interface) 2024_07_02_rel-09_w_PSD. Replaced the HadronFrac_good custom variable with a conditionalVariable that uses the same cuts.

ECLChargedPIDMVA payload containing the trained BDTs and v10 of the CNN. Both are WITHOUT PSD information.

payload for ECLChargedPIDMVA using the CNN files v9 withhout PSD info and a matching BDT

Nominal XGB payload optimised for 20mb. Uses v5 of the CNN in the barrel for p < 1 GeV. With dropout layers for CNN for p < 0.6. Optimisied in both CNN regions. BDT and CNN trained with EP3 backgrounds.

Nominal XGB payload optimised for 20mb. Uses v3 of the CNN in the barrel for p < 1 GeV. BDT and CNN trained with EP3 backgrounds.

This payload includes CNN based particle ID in the ECL barrel. Improves particle ID performance.

Nominal XGB payload optimised for 20mb, includes crystal energy at low p, uses the CNN in the barrel below 1GeV

Nominal XGB payload optimised for 20mb, includes crystal energy at low p

Weights for the ECLChargedPIDMVA module using xgboost with parameters tuned such that the weightfile is approximately 20mb. Removed per crystal energy.

Weights for the ECLChargedPIDMVA module using xgboost with parameters tuned such that the weightfile is approximately 60mb. Fixed issue with rare tracks that are not able to be extrapolated to the ECL surface with helixExtTheta.

tag to add missing software trigger skim payloads to exp7-12. Intended for MC15rd and proc13 but can likely be used for later processings too. Based on the hlt_filters_exp24 GT.

testing payload for patching the HLT skims