Standalone evaluator

Allows to integrate the CatBoost code into Android projects and simplifies the integration of CatBoost in the CERN experiments.

This method of using a trained model is not recommended due to several limitations:

  • Only models with float features are supported.
  • Dependency from the FlatBuffers library. The flatc toolkit must either be built manually or integrated into your build system.

Refer to the CMake project and an example in the CatBoost repository for more details.

A code snippet:

NCatboostStandalone::TOwningEvaluator evaluator("model.cbm");
auto modelFloatFeatureCount = (size_t)evaluator.GetFloatFeatureCount();
std::cout << "Model uses: " << modelFloatFeatureCount << " float features" << std::endl;
std::vector<float> features(modelFloatFeatureCount);
std::cout << evaluator.Apply(features, NCatboostStandalone::EPredictionType::RawValue) << std::endl;