FM / opt

/usr/local/lib64/python3.6/site-packages/torch/include/caffe2/opt/

/usr/local/lib64/python3.6/site-packages/torch/include/caffe2/opt

Name	Size	Mode	Actions
annotations.h	2141	0644	edit dl rm
backend_cutting.h	586	0644	edit dl rm
backend_transformer_base.h	2773	0644	edit dl rm
bound_shape_inferencer.h	5884	0644	edit dl rm
converter.h	3013	0644	edit dl rm
device.h	442	0644	edit dl rm
distributed.h	1126	0644	edit dl rm
fakefp16_transform.h	584	0644	edit dl rm
fusion.h	4141	0644	edit dl rm
glow_net_transform.h	1625	0644	edit dl rm
mobile.h	388	0644	edit dl rm
onnxifi_op.h	19497	0644	edit dl rm
onnxifi_transformer.h	6859	0644	edit dl rm
onnx_convert.h	1279	0644	edit dl rm
optimizer.h	402	0644	edit dl rm
optimize_ideep.h	351	0644	edit dl rm
passes.h	2642	0644	edit dl rm
shape_info.h	4784	0644	edit dl rm
tvm_transformer.h	2992	0644	edit dl rm

Edit: /usr/local/lib64/python3.6/site-packages/torch/include/caffe2/opt/converter.h (3013B)

#ifndef CAFFE2_OPT_CONVERTER_H
#define CAFFE2_OPT_CONVERTER_H

#include "caffe2/core/common.h"
#include "caffe2/core/logging.h"
#include "caffe2/opt/annotations.h"
#include "caffe2/proto/caffe2_pb.h"
#include "nomnigraph/Graph/Graph.h"
#include "nomnigraph/Representations/ControlFlow.h"
#include "nomnigraph/Representations/NeuralNet.h"

#include <unordered_map>

namespace caffe2 {

TORCH_API void injectDataEdgeIndicators(caffe2::NetDef* net);
TORCH_API void removeDataEdgeIndicators(caffe2::NetDef* net);

// Default conversion to a NNModule
// Optionally strict -- which checks for various input and output conditions.
// Optionally this function will update a vector that maps operators in the
// netdef positionally to NodeRefs in the resultant NNModule.
TORCH_API nom::repr::NNModule convertToNNModule(
    const caffe2::NetDef& net,
    bool strict = false,
    std::vector<nom::repr::NNGraph::NodeRef>* = nullptr);
TORCH_API caffe2::NetDef convertToCaffe2Proto(nom::repr::NNModule&);

// Pass in an oldNet to copy all the attributes of that network.
// Be warned that transformations that modify the graph's inputs or outputs
// are not reflected in changes to external_input or external_output.
TORCH_API caffe2::NetDef convertToCaffe2Proto(
    nom::repr::NNModule&,
    const caffe2::NetDef& oldNet);

// Use these functions instead of the registry directly.
TORCH_API std::unique_ptr<nom::repr::NeuralNetOperator>
convertToNeuralNetOperator(const caffe2::OperatorDef& op);

TORCH_API caffe2::OperatorDef convertToOperatorDef(
    const nom::repr::NNGraph::NodeRef& instrNode);

// If the annotation doesn't exist, attempt to add it
TORCH_API Caffe2Annotation* getOrAddCaffe2Annotation(
    nom::repr::NNGraph::NodeRef& instrNode);

class TORCH_API Converter {
 public:
  explicit Converter() = default;
  virtual std::unique_ptr<nom::repr::NeuralNetOperator>
  convertToNeuralNetOperator(const OperatorDef&) = 0;
  virtual OperatorDef convertToOperatorDef(const nom::repr::NeuralNetOperator*);
  static std::map<std::string, caffe2::Argument> getArgumentsFromOperator(
      caffe2::OperatorDef op);

virtual ~Converter() {}

protected:
  caffe2::DeviceOption getDeviceOption(
      const nom::repr::NeuralNetOperator* nnOp) const;
};

C10_DECLARE_REGISTRY(ConverterRegistry, Converter);
#define REGISTER_CONVERTER(name, cls) \
  C10_REGISTER_CLASS(ConverterRegistry, name, cls)

#define TRIVIAL_CONVERTER(opName)                                             \
  class opName##Converter : public Converter {                                \
    std::unique_ptr<nom::repr::NeuralNetOperator> convertToNeuralNetOperator( \
        const OperatorDef& op) override {                                     \
      return std::make_unique<nom::repr::opName>();                     \
    }                                                                         \
    virtual ~opName##Converter() {}                                           \
  };

} // namespace caffe2

#endif // CAFFE2_OPT_CONVERTER_H

#ifndef CAFFE2_OPT_CONVERTER_H #define CAFFE2_OPT_CONVERTER_H #include "caffe2/core/common.h" #include "caffe2/core/logging.h" #include "caffe2/opt/annotations.h" #include "caffe2/proto/caffe2_pb.h" #include "nomnigraph/Graph/Graph.h" #include "nomnigraph/Representations/ControlFlow.h" #include "nomnigraph/Representations/NeuralNet.h" #include namespace caffe2 { TORCH_API void injectDataEdgeIndicators(caffe2::NetDef* net); TORCH_API void removeDataEdgeIndicators(caffe2::NetDef* net); // Default conversion to a NNModule // Optionally strict -- which checks for various input and output conditions. // Optionally this function will update a vector that maps operators in the // netdef positionally to NodeRefs in the resultant NNModule. TORCH_API nom::repr::NNModule convertToNNModule( const caffe2::NetDef& net, bool strict = false, std::vector* = nullptr); TORCH_API caffe2::NetDef convertToCaffe2Proto(nom::repr::NNModule&); // Pass in an oldNet to copy all the attributes of that network. // Be warned that transformations that modify the graph's inputs or outputs // are not reflected in changes to external_input or external_output. TORCH_API caffe2::NetDef convertToCaffe2Proto( nom::repr::NNModule&, const caffe2::NetDef& oldNet); // Use these functions instead of the registry directly. TORCH_API std::unique_ptr convertToNeuralNetOperator(const caffe2::OperatorDef& op); TORCH_API caffe2::OperatorDef convertToOperatorDef( const nom::repr::NNGraph::NodeRef& instrNode); // If the annotation doesn't exist, attempt to add it TORCH_API Caffe2Annotation* getOrAddCaffe2Annotation( nom::repr::NNGraph::NodeRef& instrNode); class TORCH_API Converter { public: explicit Converter() = default; virtual std::unique_ptr convertToNeuralNetOperator(const OperatorDef&) = 0; virtual OperatorDef convertToOperatorDef(const nom::repr::NeuralNetOperator*); static std::map getArgumentsFromOperator( caffe2::OperatorDef op); virtual ~Converter() {} protected: caffe2::DeviceOption getDeviceOption( const nom::repr::NeuralNetOperator* nnOp) const; }; C10_DECLARE_REGISTRY(ConverterRegistry, Converter); #define REGISTER_CONVERTER(name, cls) \ C10_REGISTER_CLASS(ConverterRegistry, name, cls) #define TRIVIAL_CONVERTER(opName) \ class opName##Converter : public Converter { \ std::unique_ptr convertToNeuralNetOperator( \ const OperatorDef& op) override { \ return std::make_unique(); \ } \ virtual ~opName##Converter() {} \ }; } // namespace caffe2 #endif // CAFFE2_OPT_CONVERTER_H