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
flatctoolkit must either be built manually or integrated into your build system.
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;