std::string GBDT::ModelToIfElse()

in src/boosting/gbdt_model_text.cpp [124:284]
134 lines of code
12 McCabe index (conditional complexity)

std::string GBDT::ModelToIfElse(int num_iteration) const {
  std::stringstream str_buf;
  Common::C_stringstream(str_buf);

  str_buf << "#include \"gbdt.h\"" << '\n';
  str_buf << "#include <LightGBM/utils/common.h>" << '\n';
  str_buf << "#include <LightGBM/objective_function.h>" << '\n';
  str_buf << "#include <LightGBM/metric.h>" << '\n';
  str_buf << "#include <LightGBM/prediction_early_stop.h>" << '\n';
  str_buf << "#include <ctime>" << '\n';
  str_buf << "#include <sstream>" << '\n';
  str_buf << "#include <chrono>" << '\n';
  str_buf << "#include <string>" << '\n';
  str_buf << "#include <vector>" << '\n';
  str_buf << "#include <utility>" << '\n';
  str_buf << "namespace LightGBM {" << '\n';

  int num_used_model = static_cast<int>(models_.size());
  if (num_iteration > 0) {
    num_used_model = std::min(num_iteration * num_tree_per_iteration_, num_used_model);
  }

  // PredictRaw
  for (int i = 0; i < num_used_model; ++i) {
    str_buf << models_[i]->ToIfElse(i, false) << '\n';
  }

  str_buf << "double (*PredictTreePtr[])(const double*) = { ";
  for (int i = 0; i < num_used_model; ++i) {
    if (i > 0) {
      str_buf << " , ";
    }
    str_buf << "PredictTree" << i;
  }
  str_buf << " };" << '\n' << '\n';

  std::stringstream pred_str_buf;
  Common::C_stringstream(pred_str_buf);

  pred_str_buf << "\t" << "int early_stop_round_counter = 0;" << '\n';
  pred_str_buf << "\t" << "std::memset(output, 0, sizeof(double) * num_tree_per_iteration_);" << '\n';
  pred_str_buf << "\t" << "for (int i = 0; i < num_iteration_for_pred_; ++i) {" << '\n';
  pred_str_buf << "\t\t" << "for (int k = 0; k < num_tree_per_iteration_; ++k) {" << '\n';
  pred_str_buf << "\t\t\t" << "output[k] += (*PredictTreePtr[i * num_tree_per_iteration_ + k])(features);" << '\n';
  pred_str_buf << "\t\t" << "}" << '\n';
  pred_str_buf << "\t\t" << "++early_stop_round_counter;" << '\n';
  pred_str_buf << "\t\t" << "if (early_stop->round_period == early_stop_round_counter) {" << '\n';
  pred_str_buf << "\t\t\t" << "if (early_stop->callback_function(output, num_tree_per_iteration_))" << '\n';
  pred_str_buf << "\t\t\t\t" << "return;" << '\n';
  pred_str_buf << "\t\t\t" << "early_stop_round_counter = 0;" << '\n';
  pred_str_buf << "\t\t" << "}" << '\n';
  pred_str_buf << "\t" << "}" << '\n';

  str_buf << "void GBDT::PredictRaw(const double* features, double *output, const PredictionEarlyStopInstance* early_stop) const {" << '\n';
  str_buf << pred_str_buf.str();
  str_buf << "}" << '\n';
  str_buf << '\n';

  // PredictRawByMap
  str_buf << "double (*PredictTreeByMapPtr[])(const std::unordered_map<int, double>&) = { ";
  for (int i = 0; i < num_used_model; ++i) {
    if (i > 0) {
      str_buf << " , ";
    }
    str_buf << "PredictTree" << i << "ByMap";
  }
  str_buf << " };" << '\n' << '\n';

  std::stringstream pred_str_buf_map;
  Common::C_stringstream(pred_str_buf_map);

  pred_str_buf_map << "\t" << "int early_stop_round_counter = 0;" << '\n';
  pred_str_buf_map << "\t" << "std::memset(output, 0, sizeof(double) * num_tree_per_iteration_);" << '\n';
  pred_str_buf_map << "\t" << "for (int i = 0; i < num_iteration_for_pred_; ++i) {" << '\n';
  pred_str_buf_map << "\t\t" << "for (int k = 0; k < num_tree_per_iteration_; ++k) {" << '\n';
  pred_str_buf_map << "\t\t\t" << "output[k] += (*PredictTreeByMapPtr[i * num_tree_per_iteration_ + k])(features);" << '\n';
  pred_str_buf_map << "\t\t" << "}" << '\n';
  pred_str_buf_map << "\t\t" << "++early_stop_round_counter;" << '\n';
  pred_str_buf_map << "\t\t" << "if (early_stop->round_period == early_stop_round_counter) {" << '\n';
  pred_str_buf_map << "\t\t\t" << "if (early_stop->callback_function(output, num_tree_per_iteration_))" << '\n';
  pred_str_buf_map << "\t\t\t\t" << "return;" << '\n';
  pred_str_buf_map << "\t\t\t" << "early_stop_round_counter = 0;" << '\n';
  pred_str_buf_map << "\t\t" << "}" << '\n';
  pred_str_buf_map << "\t" << "}" << '\n';

  str_buf << "void GBDT::PredictRawByMap(const std::unordered_map<int, double>& features, double* output, const PredictionEarlyStopInstance* early_stop) const {" << '\n';
  str_buf << pred_str_buf_map.str();
  str_buf << "}" << '\n';
  str_buf << '\n';

  // Predict
  str_buf << "void GBDT::Predict(const double* features, double *output, const PredictionEarlyStopInstance* early_stop) const {" << '\n';
  str_buf << "\t" << "PredictRaw(features, output, early_stop);" << '\n';
  str_buf << "\t" << "if (average_output_) {" << '\n';
  str_buf << "\t\t" << "for (int k = 0; k < num_tree_per_iteration_; ++k) {" << '\n';
  str_buf << "\t\t\t" << "output[k] /= num_iteration_for_pred_;" << '\n';
  str_buf << "\t\t" << "}" << '\n';
  str_buf << "\t" << "}" << '\n';
  str_buf << "\t" << "if (objective_function_ != nullptr) {" << '\n';
  str_buf << "\t\t" << "objective_function_->ConvertOutput(output, output);" << '\n';
  str_buf << "\t" << "}" << '\n';
  str_buf << "}" << '\n';
  str_buf << '\n';

  // PredictByMap
  str_buf << "void GBDT::PredictByMap(const std::unordered_map<int, double>& features, double* output, const PredictionEarlyStopInstance* early_stop) const {" << '\n';
  str_buf << "\t" << "PredictRawByMap(features, output, early_stop);" << '\n';
  str_buf << "\t" << "if (average_output_) {" << '\n';
  str_buf << "\t\t" << "for (int k = 0; k < num_tree_per_iteration_; ++k) {" << '\n';
  str_buf << "\t\t\t" << "output[k] /= num_iteration_for_pred_;" << '\n';
  str_buf << "\t\t" << "}" << '\n';
  str_buf << "\t" << "}" << '\n';
  str_buf << "\t" << "if (objective_function_ != nullptr) {" << '\n';
  str_buf << "\t\t" << "objective_function_->ConvertOutput(output, output);" << '\n';
  str_buf << "\t" << "}" << '\n';
  str_buf << "}" << '\n';
  str_buf << '\n';


  // PredictLeafIndex
  for (int i = 0; i < num_used_model; ++i) {
    str_buf << models_[i]->ToIfElse(i, true) << '\n';
  }

  str_buf << "double (*PredictTreeLeafPtr[])(const double*) = { ";
  for (int i = 0; i < num_used_model; ++i) {
    if (i > 0) {
      str_buf << " , ";
    }
    str_buf << "PredictTree" << i << "Leaf";
  }
  str_buf << " };" << '\n' << '\n';

  str_buf << "void GBDT::PredictLeafIndex(const double* features, double *output) const {" << '\n';
  str_buf << "\t" << "int total_tree = num_iteration_for_pred_ * num_tree_per_iteration_;" << '\n';
  str_buf << "\t" << "for (int i = 0; i < total_tree; ++i) {" << '\n';
  str_buf << "\t\t" << "output[i] = (*PredictTreeLeafPtr[i])(features);" << '\n';
  str_buf << "\t" << "}" << '\n';
  str_buf << "}" << '\n';

  // PredictLeafIndexByMap
  str_buf << "double (*PredictTreeLeafByMapPtr[])(const std::unordered_map<int, double>&) = { ";
  for (int i = 0; i < num_used_model; ++i) {
    if (i > 0) {
      str_buf << " , ";
    }
    str_buf << "PredictTree" << i << "LeafByMap";
  }
  str_buf << " };" << '\n' << '\n';

  str_buf << "void GBDT::PredictLeafIndexByMap(const std::unordered_map<int, double>& features, double* output) const {" << '\n';
  str_buf << "\t" << "int total_tree = num_iteration_for_pred_ * num_tree_per_iteration_;" << '\n';
  str_buf << "\t" << "for (int i = 0; i < total_tree; ++i) {" << '\n';
  str_buf << "\t\t" << "output[i] = (*PredictTreeLeafByMapPtr[i])(features);" << '\n';
  str_buf << "\t" << "}" << '\n';
  str_buf << "}" << '\n';

  str_buf << "}  // namespace LightGBM" << '\n';

  return str_buf.str();
}