TFCSEnergyAndHitGAN.cxx

Go to the documentation of this file.

/*
  Copyright (C) 2002-2026 CERN for the benefit of the ATLAS collaboration
*/
 
#include "ISF_FastCaloSimEvent/TFCSEnergyAndHitGAN.h"
#include "ISF_FastCaloSimEvent/TFCSLateralShapeParametrizationHitBase.h"
#include "ISF_FastCaloSimEvent/TFCSSimulationState.h"
#include "ISF_FastCaloSimEvent/TFCSTruthState.h"
#include "ISF_FastCaloSimEvent/TFCSExtrapolationState.h"
#include "ISF_FastCaloSimEvent/TFCSCenterPositionCalculation.h"
#include "CxxUtils/as_const_ptr.h"
 
#include "TFile.h"
 
#include "HepPDT/ParticleData.hh"
#include "HepPDT/ParticleDataTable.hh"
 
#include "CLHEP/Random/RandGauss.h"
#include "CLHEP/Random/RandFlat.h"
 
#if defined(__FastCaloSimStandAlone__)
#include "CLHEP/Random/TRandomEngine.h"
#else
#include <CLHEP/Random/RanluxEngine.h>
#endif
 
#include <iostream>
#include <fstream>
#include <limits>
#include <cassert>
 
// LWTNN
#include "lwtnn/LightweightGraph.hh"
#include "lwtnn/parse_json.hh"
 
#include "XMLCoreParser/XMLCoreParser.h"
#include "XMLCoreParser/XMLCoreNode.h"
 
 
//=============================================
//======= TFCSEnergyAndHitGAN =========
//=============================================
 
TFCSEnergyAndHitGAN::TFCSEnergyAndHitGAN(const char *name, const char *title)
    : TFCSParametrizationBinnedChain(name, title) {
  set_GANfreemem();
}
 
TFCSEnergyAndHitGAN::~TFCSEnergyAndHitGAN() {
  if (m_input != nullptr) {
    delete m_input;
  }
  if (m_graph != nullptr) {
    delete m_graph;
  }
}
 
bool TFCSEnergyAndHitGAN::is_match_calosample(int calosample) const {
  if (get_Binning().find(calosample) == get_Binning().cend())
    return false;
  if (get_Binning().at(calosample).GetNbinsX() == 1)
    return false;
  return true;
}
 
unsigned int TFCSEnergyAndHitGAN::get_nr_of_init(unsigned int bin) const {
  if (bin >= m_bin_ninit.size())
    return 0;
  return m_bin_ninit[bin];
}
 
void TFCSEnergyAndHitGAN::set_nr_of_init(unsigned int bin, unsigned int ninit) {
  if (bin >= m_bin_ninit.size()) {
    m_bin_ninit.resize(bin + 1, 0);
    m_bin_ninit.shrink_to_fit();
  }
  m_bin_ninit[bin] = ninit;
}
 
// initialize lwtnn network
bool TFCSEnergyAndHitGAN::initializeNetwork(
    int pid, int etaMin, const std::string &FastCaloGANInputFolderName) {
 
  // initialize all necessary constants
  // FIXME eventually all these could be stored in the .json file
 
  ATH_MSG_INFO(
      "Using FastCaloGANInputFolderName: " << FastCaloGANInputFolderName);
  // get neural net JSON file as an std::istream object
  int etaMax = etaMin + 5;
 
  reset_match_all_pdgid();
  set_pdgid(pid);
  if (pid == 11)
    add_pdgid(-pid);
  if (pid == 211)
    add_pdgid(-pid);
  set_eta_min(etaMin / 100.0);
  set_eta_max(etaMax / 100.0);
  set_eta_nominal((etaMin + etaMax) / 200.0);
 
  std::string inputFile = std::format("{}/neural_net_{}_eta_{}_{}.json", 
                                     FastCaloGANInputFolderName, pid, etaMin, etaMax);
  ATH_MSG_INFO("For pid: " << pid << " and eta " << etaMin << "-" << etaMax
                           << ", loading json file " << inputFile);
  std::ifstream input(inputFile);
  if (!input.is_open()) { // check: verify the file actually exists and opened
    ATH_MSG_ERROR(std::format("Could not open json file: {}", inputFile));
    return false;
  }
  std::stringstream sin;
  sin << input.rdbuf();
  input.close();
  // build the graph
  auto config = lwt::parse_json_graph(sin);
  m_graph = new lwt::LightweightGraph(config);
  assert(m_graph!=nullptr);
  if (m_input != nullptr) {
    delete m_input;
  }
  m_input = new std::string(sin.str());
 
  m_GANLatentSize = 50;
 
  // Get all Binning histograms to store in memory
  GetBinning(pid, (etaMin + etaMax) / 2, FastCaloGANInputFolderName);
 
  if (m_GANLatentSize == 0) {
    ATH_MSG_ERROR("m_GANLatentSize uninitialized!");
    return false;
  }
 
  return true;
}
 
void TFCSEnergyAndHitGAN::GetBinning(
    int pid, int etaMid, const std::string &FastCaloGANInputFolderName) {
  const std::string xmlFullFileName = std::format("{}/binning.xml", FastCaloGANInputFolderName);
  ATH_MSG_DEBUG("Opening XML file in " << xmlFullFileName);
 
  std::vector<Binning> AllBinning;
  std::vector<int> EtaMaxList;
 
  XMLCoreParser p;
  std::unique_ptr<XMLCoreNode> doc = p.parse (xmlFullFileName);
  for (const XMLCoreNode* bin : doc->get_children ("Bins/Bin")) {
    int nodePid = bin->get_int_attrib ("pid");
    int nodeEtaMax = bin->get_int_attrib ("etaMax");
 
    Binning binsInLayer;
    bool correctentry = true;
    if (nodePid != pid)
      correctentry = false;
 
    for (const XMLCoreNode* nodeLayer : bin->get_children ("Layer")) {
      std::vector<double> edges;
      std::string s = nodeLayer->get_attrib ("r_edges");
      std::istringstream ss(s);
      std::string token;
 
      while (std::getline(ss, token, ',')) {
        edges.push_back(atof(token.c_str()));
      }
 
      int binsInAlpha = nodeLayer->get_int_attrib ("n_bin_alpha");
      int layer = nodeLayer->get_int_attrib ("id");
 
      if (correctentry)
        ATH_MSG_DEBUG("nodepid=" << nodePid << " nodeEtaMax="
                      << nodeEtaMax << " Layer: " << layer
                      << " binsInAlpha: " << binsInAlpha
                      << " edges: " << s);
 
      std::string name = "hist_pid_" + std::to_string(nodePid) +
        "_etaSliceNumber_" +
        std::to_string(EtaMaxList.size()) + "_layer_" +
        std::to_string(layer);
      int xBins = edges.size() - 1;
      if (xBins == 0) {
        xBins = 1; // remove warning
        edges.push_back(0);
        edges.push_back(1);
      }
      binsInLayer[layer] =
        TH2D(name.c_str(), name.c_str(), xBins, &edges[0],
             binsInAlpha, -TMath::Pi(), TMath::Pi());
      binsInLayer[layer].SetDirectory(nullptr);
    }
 
    if (!correctentry)
      continue;
    AllBinning.push_back(std::move(binsInLayer));
    EtaMaxList.push_back(nodeEtaMax);
  }
 
  int index = 0;
  for (int etaMax : EtaMaxList) {
    if (etaMid < etaMax) {
      m_Binning = AllBinning[index];
      break;
    }
    index++;
  }
  ATH_MSG_DEBUG("Done XML file");
}
 
const std::string
TFCSEnergyAndHitGAN::get_variable_text(TFCSSimulationState &simulstate,
                                       const TFCSTruthState *,
                                       const TFCSExtrapolationState *) const {
  return std::string(
      Form("layer=%d", simulstate.getAuxInfo<int>("GANlayer"_FCShash)));
}
 
bool TFCSEnergyAndHitGAN::fillFastCaloGanNetworkInputs(
    TFCSSimulationState &simulstate, const TFCSTruthState *truth,
    NetworkInputs &inputs, double &trueEnergy) const {
  // fill randomize latent space
  // FIXME: should this really be momentum
  trueEnergy = truth->P();
  double randUniformZ = 0.;
  // FIXME: make dependency on input particle eta, pdgid and energy
  for (int i = 0; i < m_GANLatentSize; i++) {
    randUniformZ = CLHEP::RandFlat::shoot(simulstate.randomEngine(), -1., 1.);
    inputs["node_0"].insert(
        std::pair<std::string, double>(std::to_string(i), randUniformZ));
  }
 
  // ATH_MSG_INFO( "Check label: " <<trueEnergy <<" "<<std::pow(2,22)<<"
  // "<<trueEnergy/std::pow(2,22));
  inputs["node_1"].insert(
      std::pair<std::string, double>("0", trueEnergy / (std::pow(2, 22))));
 
  return true;
}
 
bool TFCSEnergyAndHitGAN::fillEnergy(TFCSSimulationState &simulstate,
                                     const TFCSTruthState *truth,
                                     const TFCSExtrapolationState *extrapol,
                                     NetworkInputs inputs) const {
  const int pdgId = truth->pdgid();
  const float charge = HepPDT::ParticleID(pdgId).charge();
 
  float Einit;
  const float Ekin = truth->Ekin();
  if (OnlyScaleEnergy())
    Einit = simulstate.E();
  else
    Einit = Ekin;
 
  ATH_MSG_VERBOSE("Momentum " << truth->P() << " pdgId " << truth->pdgid());
 
  // compute the network output values
  // ATH_MSG_VERBOSE("neural network input = "<<inputs);
  ATH_MSG_VERBOSE("input size " << inputs["node_0"].size());
  NetworkOutputs outputs = m_graph->compute(inputs);
  // ATH_MSG_VERBOSE("neural network output = "<<outputs);
 
  ATH_MSG_VERBOSE("neural network output size = " << outputs.size());
 
  const Binning &binsInLayers = m_Binning;
  ATH_MSG_VERBOSE("Get binning");
 
  simulstate.set_E(0);
 
  int vox = 0;
  for (const auto& element : binsInLayers) {
    int layer = element.first;
 
    const TH2D *h = &element.second;
    int xBinNum = h->GetNbinsX();
    // If only one bin in r means layer is empty, no value should be added
    if (xBinNum == 1) {
      ATH_MSG_VERBOSE(" Layer "
                      << layer
                      << " has only one bin in r, this means is it not used, "
                         "skipping (this is needed to keep correct "
                         "syncronisation of voxel and layers)");
      // delete h;
      continue;
    }
 
    ATH_MSG_VERBOSE(" Getting energy for Layer " << layer);
 
    int yBinNum = h->GetNbinsY();
 
    // First fill energies
    for (int iy = 1; iy <= yBinNum; ++iy) {
      for (int ix = 1; ix <= xBinNum; ++ix) {
        double energyInVoxel = outputs["out_" + std::to_string(vox)];
        ATH_MSG_VERBOSE(" Vox " << vox << " energy " << energyInVoxel
                                << " binx " << ix << " biny " << iy);
 
        if (energyInVoxel == 0) {
          vox++;
          continue;
        }
 
        simulstate.add_E(layer, Einit * energyInVoxel);
        vox++;
      }
    }
  }
 
  for (unsigned int ichain = m_bin_start.back(); ichain < size(); ++ichain) {
    ATH_MSG_DEBUG("now run for all bins: " << chain()[ichain]->GetName());
    if (simulate_and_retry(chain()[ichain], simulstate, truth, extrapol) !=
        FCSSuccess) {
      return FCSFatal;
    }
  }
 
  vox = 0;
  for (const auto& element : binsInLayers) {
    int layer = element.first;
    simulstate.setAuxInfo<int>("GANlayer"_FCShash, layer);
    TFCSLateralShapeParametrizationHitBase::Hit hit;
 
    const TH2D *h = &element.second;
    int xBinNum = h->GetNbinsX();
    // If only one bin in r means layer is empty, no value should be added
    if (xBinNum == 1) {
      ATH_MSG_VERBOSE(" Layer "
                      << layer
                      << " has only one bin in r, this means is it not used, "
                         "skipping (this is needed to keep correct "
                         "syncronisation of voxel and layers)");
      // delete h;
      continue;
    }
 
    if (get_number_of_bins() > 0) {
      const int bin = get_bin(simulstate, truth, extrapol);
      if (bin >= 0 && bin < (int)get_number_of_bins()) {
        for (unsigned int ichain = m_bin_start[bin];
             ichain < TMath::Min(m_bin_start[bin] + get_nr_of_init(bin),
                                 m_bin_start[bin + 1]);
             ++ichain) {
          ATH_MSG_DEBUG("for " << get_variable_text(simulstate, truth, extrapol)
                               << " run init " << get_bin_text(bin) << ": "
                               << chain()[ichain]->GetName());
          if (chain()[ichain]->InheritsFrom(
                  TFCSLateralShapeParametrizationHitBase::Class())) {
            TFCSLateralShapeParametrizationHitBase *sim =
                static_cast<TFCSLateralShapeParametrizationHitBase *>(chain()[ichain]);
            if (sim->simulate_hit(hit, simulstate, truth, extrapol) !=
                FCSSuccess) {
              ATH_MSG_ERROR("error for "
                            << get_variable_text(simulstate, truth, extrapol)
                            << " run init " << get_bin_text(bin) << ": "
                            << chain()[ichain]->GetName());
              return false;
            }
          } else {
            ATH_MSG_ERROR("for "
                          << get_variable_text(simulstate, truth, extrapol)
                          << " run init " << get_bin_text(bin) << ": "
                          << chain()[ichain]->GetName()
                          << " does not inherit from "
                             "TFCSLateralShapeParametrizationHitBase");
            return false;
          }
        }
      } else {
        ATH_MSG_WARNING("nothing to init for "
                        << get_variable_text(simulstate, truth, extrapol)
                        << ": " << get_bin_text(bin));
      }
    }
 
    int binResolution = 5;
    if (layer == 1 || layer == 5) {
      binResolution = 1;
    }
 
    const double center_eta = hit.center_eta();
    const double center_phi = hit.center_phi();
    const double center_r = hit.center_r();
    const double center_z = hit.center_z();
 
    ATH_MSG_VERBOSE(" Layer " << layer << " Extrap eta " << center_eta
                              << " phi " << center_phi << " R " << center_r);
 
    const float dist000 =
        TMath::Sqrt(center_r * center_r + center_z * center_z);
    const float eta_jakobi = TMath::Abs(2.0 * TMath::Exp(-center_eta) /
                                        (1.0 + TMath::Exp(-2 * center_eta)));
 
    int nHitsAlpha;
    int nHitsR;
 
    int yBinNum = h->GetNbinsY();
 
    // Now create hits
    for (int iy = 1; iy <= yBinNum; ++iy) {
      for (int ix = 1; ix <= xBinNum; ++ix) {
 
        double energyInVoxel = outputs["out_" + std::to_string(vox)];
        ATH_MSG_VERBOSE(" Vox " << vox << " energy " << energyInVoxel
                                << " binx " << ix << " biny " << iy);
 
        if (energyInVoxel == 0) {
          vox++;
          continue;
        }
 
        const TAxis *x = h->GetXaxis();
        nHitsR = x->GetBinUpEdge(ix) - x->GetBinLowEdge(ix);
        if (yBinNum == 1) {
          // nbins in alpha depend on circumference lenght
          double r = x->GetBinUpEdge(ix);
          nHitsAlpha = ceil(2 * TMath::Pi() * r / binResolution);
        } else {
          // d = 2*r*sin (a/2r) this distance at the upper r must be 1mm for
          // layer 1 or 5, 5mm otherwise.
          const TAxis *y = h->GetYaxis();
          double angle = y->GetBinUpEdge(iy) - y->GetBinLowEdge(iy);
          double r = x->GetBinUpEdge(ix);
          double d = 2 * r * sin(angle / 2 * r);
          nHitsAlpha = ceil(d / binResolution);
        }
 
        nHitsAlpha = std::min(10, std::max(1, nHitsAlpha));
        nHitsR = std::min(10, std::max(1, nHitsR));
 
        for (int ir = 0; ir < nHitsR; ++ir) {
          const TAxis *x = h->GetXaxis();
          double r =
              x->GetBinLowEdge(ix) + x->GetBinWidth(ix) / (nHitsR + 1) * ir;
 
          for (int ialpha = 0; ialpha < nHitsAlpha; ++ialpha) {
            double alpha;
            if (yBinNum == 1) {
              alpha = CLHEP::RandFlat::shoot(simulstate.randomEngine(), -M_PI,
                                             M_PI);
            } else {
              const TAxis *y = h->GetYaxis();
              alpha = y->GetBinLowEdge(iy) +
                      y->GetBinWidth(iy) / (nHitsAlpha + 1) * ialpha;
            }
 
            hit.reset();
            hit.E() = Einit * energyInVoxel / (nHitsAlpha * nHitsR);
 
            if (layer <= 20) {
              float delta_eta_mm = r * cos(alpha);
              float delta_phi_mm = r * sin(alpha);
 
              ATH_MSG_VERBOSE("delta_eta_mm " << delta_eta_mm
                                              << " delta_phi_mm "
                                              << delta_phi_mm);
 
              // Particles with negative eta are expected to have the same shape
              // as those with positive eta after transformation: delta_eta -->
              // -delta_eta
              if (center_eta < 0.)
                delta_eta_mm = -delta_eta_mm;
              // We derive the shower shapes for electrons and positively charged hadrons.
              // Particle with the opposite charge are expected to have the same shower shape
              // after the transformation: delta_phi --> -delta_phi
              if ((charge < 0. && pdgId!=11) || pdgId==-11)
                delta_phi_mm = -delta_phi_mm;
 
              const float delta_eta = delta_eta_mm / eta_jakobi / dist000;
              const float delta_phi = delta_phi_mm / center_r;
 
              hit.eta() = center_eta + delta_eta;
              hit.phi() = center_phi + delta_phi;
 
              ATH_MSG_VERBOSE(" Hit eta " << hit.eta() << " phi " << hit.phi()
                                          << " layer " << layer);
            } else { // FCAL is in (x,y,z)
              const float hit_r = r * cos(alpha) + center_r;
              float delta_phi = r * sin(alpha) / center_r;
 
              // We derive the shower shapes for electrons and positively charged hadrons.
              // Particle with the opposite charge are expected to have the same shower shape
              // after the transformation: delta_phi --> -delta_phi
              if ((charge < 0. && pdgId!=11) || pdgId==-11)
                delta_phi = -delta_phi;
              const float hit_phi = delta_phi + center_phi;
              hit.x() = hit_r * cos(hit_phi);
              hit.y() = hit_r * sin(hit_phi);
              hit.z() = center_z;
              ATH_MSG_VERBOSE(" Hit x " << hit.x() << " y " << hit.y()
                                        << " layer " << layer);
            }
 
            if (get_number_of_bins() > 0) {
              const int bin = get_bin(simulstate, truth, extrapol);
              if (bin >= 0 && bin < (int)get_number_of_bins()) {
                for (unsigned int ichain =
                         m_bin_start[bin] + get_nr_of_init(bin);
                     ichain < m_bin_start[bin + 1]; ++ichain) {
                  ATH_MSG_DEBUG(
                      "for " << get_variable_text(simulstate, truth, extrapol)
                             << " run " << get_bin_text(bin) << ": "
                             << chain()[ichain]->GetName());
                  if (chain()[ichain]->InheritsFrom(
                          TFCSLateralShapeParametrizationHitBase::Class())) {
                    TFCSLateralShapeParametrizationHitBase *sim =
                        static_cast<TFCSLateralShapeParametrizationHitBase
                             *>(chain()[ichain]);
                    if (sim->simulate_hit(hit, simulstate, truth, extrapol) !=
                        FCSSuccess) {
                      ATH_MSG_ERROR(
                          "error for "
                          << get_variable_text(simulstate, truth, extrapol)
                          << " run init " << get_bin_text(bin) << ": "
                          << chain()[ichain]->GetName());
                      return false;
                    }
                  } else {
                    ATH_MSG_ERROR(
                        "for " << get_variable_text(simulstate, truth, extrapol)
                               << " run init " << get_bin_text(bin) << ": "
                               << chain()[ichain]->GetName()
                               << " does not inherit from "
                                  "TFCSLateralShapeParametrizationHitBase");
                    return false;
                  }
                }
              } else {
                ATH_MSG_WARNING(
                    "nothing to do for "
                    << get_variable_text(simulstate, truth, extrapol) << ": "
                    << get_bin_text(bin));
              }
            } else {
              ATH_MSG_WARNING("no bins defined, is this intended?");
            }
          }
        }
        vox++;
      }
    }
 
    ATH_MSG_VERBOSE("Number of voxels " << vox);
 
    // delete h;
    ATH_MSG_VERBOSE("Done layer " << layer);
  }
  if (simulstate.E() > std::numeric_limits<double>::epsilon()) {
    for (int ilayer = 0; ilayer < CaloCell_ID_FCS::MaxSample; ++ilayer) {
      simulstate.set_Efrac(ilayer, simulstate.E(ilayer) / simulstate.E());
    }
  }
 
  ATH_MSG_VERBOSE("Done particle");
 
  return true;
}
 
FCSReturnCode
TFCSEnergyAndHitGAN::simulate(TFCSSimulationState &simulstate,
                              const TFCSTruthState *truth,
                              const TFCSExtrapolationState *extrapol) const {
  for (unsigned int ichain = 0; ichain < m_bin_start[0]; ++ichain) {
    ATH_MSG_DEBUG("now run for all bins: " << chain()[ichain]->GetName());
    if (simulate_and_retry(chain()[ichain], simulstate, truth, extrapol) !=
        FCSSuccess) {
      return FCSFatal;
    }
  }
 
  // Compute all inputs to the network
  NetworkInputs inputs;
  double trueEnergy;
  ATH_MSG_VERBOSE("Get Inputs");
  if (!fillFastCaloGanNetworkInputs(simulstate, truth, inputs, trueEnergy)) {
    ATH_MSG_WARNING("Could not initialize network ");
    // bail out but do not stop the job
    return FCSSuccess;
  }
 
  ATH_MSG_VERBOSE("Fill Energies");
  if (!fillEnergy(simulstate, truth, extrapol, std::move(inputs))) {
    ATH_MSG_WARNING("Could not fill energies ");
    // bail out but do not stop the job
    return FCSSuccess;
  }
 
  return FCSSuccess;
}
 
void TFCSEnergyAndHitGAN::Print(Option_t *option) const {
  TFCSParametrization::Print(option);
  TString opt(option);
  bool shortprint = opt.Index("short") >= 0;
  bool longprint = msgLvl(MSG::DEBUG) || (msgLvl(MSG::INFO) && !shortprint);
  TString optprint = opt;
  optprint.ReplaceAll("short", "");
 
  if (longprint) {
    ATH_MSG_INFO(optprint << "  "
                          << "Graph=" << CxxUtils::as_const_ptr(m_graph)
                          << "; json input=" << CxxUtils::as_const_ptr(m_input)
                          << "; free mem=" << GANfreemem()
                          << "; latent space=" << m_GANLatentSize
                          << "; Binning size=" << m_Binning.size());
    for (const auto &l : m_Binning)
      if (is_match_calosample(l.first)) {
        ATH_MSG_INFO(optprint << "    "
                              << "layer=" << l.first
                              << " nR=" << l.second.GetNbinsX()
                              << " nalpha=" << l.second.GetNbinsY());
      }
  }
 
  TString prefix = "- ";
  for (unsigned int ichain = 0; ichain < size(); ++ichain) {
    if (ichain == 0 && ichain != m_bin_start.front()) {
      prefix = "> ";
      if (longprint)
        ATH_MSG_INFO(optprint << prefix << "Run for all bins");
    }
    for (unsigned int ibin = 0; ibin < get_number_of_bins(); ++ibin) {
      if (ichain == m_bin_start[ibin]) {
        if (ibin < get_number_of_bins() - 1)
          if (ichain == m_bin_start[ibin + 1])
            continue;
        prefix = Form("%-2d", ibin);
        if (longprint)
          ATH_MSG_INFO(optprint << prefix << "Run for " << get_bin_text(ibin));
      }
    }
    if (ichain == m_bin_start.back()) {
      prefix = "< ";
      if (longprint)
        ATH_MSG_INFO(optprint << prefix << "Run for all bins");
    }
    chain()[ichain]->Print(opt + prefix);
  }
}
 
void TFCSEnergyAndHitGAN::Streamer(TBuffer &R__b) {
  // Stream an object of class TFCSEnergyAndHitGAN
 
  if (R__b.IsReading()) {
    R__b.ReadClassBuffer(TFCSEnergyAndHitGAN::Class(), this);
    if (m_graph != nullptr) {
      delete m_graph;
      m_graph = nullptr;
    }
    if (m_input) {
      std::stringstream sin;
      sin.str(*m_input);
      auto config = lwt::parse_json_graph(sin);
      m_graph = new lwt::LightweightGraph(config);
    }
#ifndef __FastCaloSimStandAlone__
    // When running inside Athena, delete config to free the memory
    if (GANfreemem()) {
      delete m_input;
      m_input = nullptr;
    }
#endif
  } else {
    R__b.WriteClassBuffer(TFCSEnergyAndHitGAN::Class(), this);
  }
}
 
void TFCSEnergyAndHitGAN::unit_test(TFCSSimulationState *simulstate,
                                    const TFCSTruthState *truth,
                                    const TFCSExtrapolationState *extrapol) {
  if (!simulstate) {
    simulstate = new TFCSSimulationState();
#if defined(__FastCaloSimStandAlone__)
    simulstate->setRandomEngine(new CLHEP::TRandomEngine());
#else
    simulstate->setRandomEngine(new CLHEP::RanluxEngine());
#endif
  }
  if (!truth) {
    TFCSTruthState *t = new TFCSTruthState();
    t->SetPtEtaPhiM(20000, 0.225, 0, 130);
    t->set_pdgid(211);
    truth = t;
  }
  if (!extrapol) {
    TFCSExtrapolationState *e = new TFCSExtrapolationState();
    e->set_IDCaloBoundary_eta(truth->Eta());
    for (int i = 0; i < 24; ++i) {
      e->set_eta(i, TFCSExtrapolationState::SUBPOS_ENT, truth->Eta());
      e->set_eta(i, TFCSExtrapolationState::SUBPOS_EXT, truth->Eta());
      e->set_eta(i, TFCSExtrapolationState::SUBPOS_MID, truth->Eta());
      e->set_phi(i, TFCSExtrapolationState::SUBPOS_ENT, 0);
      e->set_phi(i, TFCSExtrapolationState::SUBPOS_EXT, 0);
      e->set_phi(i, TFCSExtrapolationState::SUBPOS_MID, 0);
      e->set_r(i, TFCSExtrapolationState::SUBPOS_ENT, 1500 + i * 10);
      e->set_r(i, TFCSExtrapolationState::SUBPOS_EXT, 1510 + i * 10);
      e->set_r(i, TFCSExtrapolationState::SUBPOS_MID, 1505 + i * 10);
      e->set_z(i, TFCSExtrapolationState::SUBPOS_ENT, 3500 + i * 10);
      e->set_z(i, TFCSExtrapolationState::SUBPOS_EXT, 3510 + i * 10);
      e->set_z(i, TFCSExtrapolationState::SUBPOS_MID, 3505 + i * 10);
    }
    extrapol = e;
  }
 
  TFCSEnergyAndHitGAN GAN("GAN", "GAN");
  GAN.setLevel(MSG::VERBOSE);
  static constexpr int pid = 211;
  int etaMin = 20;
  int etaMax = etaMin + 5;
  GAN.initializeNetwork(
      pid, etaMin,
      "/eos/atlas/atlascerngroupdisk/proj-simul/VoxalisationOutputs/nominal/"
      "GAN_michele_normE_MaxE/input_for_service_new");
  for (int i = 0; i < 24; ++i)
    if (GAN.is_match_calosample(i)) {
      TFCSCenterPositionCalculation *c = new TFCSCenterPositionCalculation(
          Form("center%d", i), Form("center layer %d", i));
      c->set_calosample(i);
      c->setExtrapWeight(0.5);
      c->setLevel(MSG::VERBOSE);
      c->set_pdgid(pid);
      if (pid == 11)
        //pid was set to 211
        //coverity[DEADCODE]
        c->add_pdgid(-pid);
      if (pid == 211)
        c->add_pdgid(-pid);
      c->set_eta_min(etaMin / 100.0);
      c->set_eta_max(etaMax / 100.0);
      c->set_eta_nominal((etaMin + etaMax) / 200.0);
 
      GAN.push_back_in_bin(c, i);
      GAN.set_nr_of_init(i, 1);
    }
 
  GAN.Print();
 
  TFile *fGAN = TFile::Open("FCSGANtest.root", "recreate");
  GAN.Write();
  fGAN->ls();
  fGAN->Close();
  delete fGAN;
  fGAN = TFile::Open("FCSGANtest.root");
  TFCSEnergyAndHitGAN *GAN2 = (TFCSEnergyAndHitGAN *)(fGAN->Get("GAN"));
  GAN2->Print();
 
  GAN2->setLevel(MSG::DEBUG);
  GAN2->simulate(*simulstate, truth, extrapol);
  simulstate->Print();
  delete fGAN;
}