Namespace: Ml

namespace ml

Functions

ROOT::RDF::RNode StandardTransformer(ROOT::RDF::RNode df, const std::string &inputname, const std::string &outputname, const std::string &param_file, const std::string &var_type)

Function to perform a standard transformation of input variables for NN evaluation.

Parameters:

  • df – The input dataframe

  • inputname – name of the variable which should be transformed

  • outputname – name of the output column

  • param_file – path to a json file with a dictionary of mean and std values

  • var_type – variable data type for correct processing e.g. “i” for integer or “f” for float

Returns:

a new dataframe containing the new column

ROOT::RDF::RNode StandardTransformer(ROOT::RDF::RNode df, correctionManager::CorrectionManager &correctionManager, const std::string &inputname, const std::string &outputname, const std::string &param_file, const std::string &var_type)

Function to perform a standard transformation of input variables for NN evaluation.

Parameters:

  • df – The input dataframe

  • correctionManager – correction manager

  • inputname – name of the variable which should be transformed

  • outputname – name of the output column

  • param_file – path to a json file with a dictionary of mean and std values

  • var_type – variable data type for correct processing e.g. “i” for integer or “f” for float

Returns:

a new dataframe containing the new column

template<std::size_t nParameters>
inline ROOT::RDF::RNode GenericOnnxEvaluator(ROOT::RDF::RNode df, OnnxSessionManager &onnxSessionManager, const std::string &outputname, const std::string &model_file_path, const std::vector<std::string> &input_vec)

Generic Function to evaluate an ONNX model using the ONNX Runtime Due to unknowns reasons, this function must be implemented inline in the header file, otherwise the linker will complain about undefined references. Moving the implementation to the source file will result in a linker error. Why, I don’t know… This generic implementation currenty supports only NNs with one input tensor and one output tensor

Parameters:

  • df – the dataframe to add the quantity to

  • onnxSessionManager – The OnnxSessionManager object to handle the runtime session. By default this is called onnxSessionManager and created in the main function

  • outputname – Name of the output column

  • model_file_path – Path to the ONNX model file

  • input_vec – Vector of input variable names, the order of the variables must match the order of the input nodes in the ONNX model

Returns:

a dataframe with the filter applied