root/TMVA/ROperator_Identity.hxx

0001 #ifndef TMVA_SOFIE_ROPERATOR_IDENTITY
0002 #define TMVA_SOFIE_ROPERATOR_IDENTITY
0003
0004 #include "TMVA/SOFIE_common.hxx"
0005 #include "TMVA/ROperator.hxx"
0006 #include "TMVA/RModel.hxx"
0007
0008 #include <sstream>
0009
0010 namespace TMVA{
0011 namespace Experimental{
0012 namespace SOFIE{
0013
0014 template <typename T>
0015 class ROperator_Identity final : public ROperator
0016 {
0017
0018 private:
0019
0020    bool fIsInputInitialized = false;
0021    std::string fNX;
0022    std::string fNY;
0023    std::vector<size_t> fShape;
0024
0025 public:
0026    ROperator_Identity(){}
0027    ROperator_Identity(std::string nameX, std::string nameY):
0028       fNX(UTILITY::Clean_name(nameX)), fNY(UTILITY::Clean_name(nameY)){
0029          fInputTensorNames = { fNX };
0030          fOutputTensorNames = { fNY };
0031       }
0032
0033    std::vector<ETensorType> TypeInference(std::vector<ETensorType> input) override {
0034       return input;
0035    }
0036
0037    std::vector<std::vector<size_t>> ShapeInference(std::vector<std::vector<size_t>> input) override {
0038       auto ret = input; //suggest copy to compiler
0039       return ret;
0040    }
0041
0042    void Initialize(RModel& model) override {
0043        //input must be a graph input, or already initialized intermediate tensor
0044       if (model.CheckIfTensorAlreadyExist(fNX) == false){
0045         throw std::runtime_error("TMVA SOFIE Identity Op Input Tensor is not found in model");
0046       }
0047       fShape = model.GetTensorShape(fNX);
0048       if (model.IsInitializedTensor(fNX)) {
0049          // we need to check if is a weight (initialized) or a constant tensor
0050          // in the first case we need to create a constant tensor with the output, in teh second we
0051          // need to generate the identy code in the GenerateInitCode
0052          if (model.IsConstantTensor(fNX)) {
0053             auto inputData = static_cast<T*>(model.GetInitializedTensorData(fNX).get());
0054             model.AddConstantTensor<T>(fNY, fShape, inputData);
0055             fIsOutputConstant = true;
0056          } else {
0057             fIsInputInitialized = true;
0058             // need to create a dummy intermediate tensor for the declaration
0059             // this could probably be improved to save memory
0060             model.AddIntermediateTensor(fNY, model.GetTensorType(fNX), fShape);
0061          }
0062       } else
0063          model.AddIntermediateTensor(fNY, model.GetTensorType(fNX), fShape);
0064    }
0065
0066    std::string GenerateInitCode() override {
0067       // generate init code for identity operator
0068       if (!fIsInputInitialized) return "";
0069       std::stringstream out;
0070       out << "\n//------ IDENTITY\n";
0071       // just copy the tensor pointers
0072       out << SP << SP << "tensor_" << fNY << " = tensor_" << fNX << ";\n";
0073       return out.str();
0074    }
0075
0076
0077    std::string Generate(std::string OpName) override {
0078       if (fIsOutputConstant || fIsInputInitialized) return "";
0079       OpName = "op_" + OpName;
0080       if (fShape.empty()) {
0081          throw std::runtime_error("TMVA SOFIE Operator Identity called to Generate without being initialized first");
0082       }
0083       std::stringstream out;
0084       out << "\n//------ IDENTITY\n";
0085       // just copy the tensor pointers
0086       out << SP << SP << "tensor_" << fNY << " = tensor_" << fNX << ";\n";
0087       return out.str();
0088    }
0089
0090 };
0091
0092 }//SOFIE
0093 }//Experimental
0094 }//TMVA
0095
0096
0097 #endif //TMVA_SOFIE_ROPERATOR_IDENTITY