3

I'm trying to extract data like input layers, output layers and their shapes from an onnx model. I know there is python interface to do this. I want to do something similar to this code but in c++. I have also pasted the code from the link. I have tried it in python and it has worked for me. I want to know if there are c++ API's to do the same.

import onnx

model = onnx.load(r"model.onnx")

# The model is represented as a protobuf structure and it can be accessed
# using the standard python-for-protobuf methods

# iterate through inputs of the graph
for input in model.graph.input:
    print (input.name, end=": ")
    # get type of input tensor
    tensor_type = input.type.tensor_type
    # check if it has a shape:
    if (tensor_type.HasField("shape")):
        # iterate through dimensions of the shape:
        for d in tensor_type.shape.dim:
            # the dimension may have a definite (integer) value or a symbolic identifier or neither:
            if (d.HasField("dim_value")):
                print (d.dim_value, end=", ")  # known dimension
            elif (d.HasField("dim_param")):
                print (d.dim_param, end=", ")  # unknown dimension with symbolic name
            else:
                print ("?", end=", ")  # unknown dimension with no name
    else:
        print ("unknown rank", end="")
    print()

Also I'm new to c++, please help me with the same.

Hitesh Kumar
  • 193
  • 3
  • 11

1 Answers1

7

ONNX format is essentially a protobuf, so it can be opened in any language protoc compiler supports.

In case of C++

  1. Take onnx proto file (onnx repo)
  2. Compile it with protoc --cpp_out=. onnx.proto3 command. It will generate onnx.proto3.pb.cc and onnx.proto3.pb.h files
  3. Link protobuf library (maybe protobuf-lite), generated cpp file and following code:
#include <fstream>
#include <cassert>

#include "onnx.proto3.pb.h"

void print_dim(const ::onnx::TensorShapeProto_Dimension &dim)
{
  switch (dim.value_case())
  {
  case onnx::TensorShapeProto_Dimension::ValueCase::kDimParam:
    std::cout << dim.dim_param();
    break;
  case onnx::TensorShapeProto_Dimension::ValueCase::kDimValue:
    std::cout << dim.dim_value();
    break;
  default:
    assert(false && "should never happen");
  }
}

void print_io_info(const ::google::protobuf::RepeatedPtrField< ::onnx::ValueInfoProto > &info)
{
  for (auto input_data: info)
  {
    auto shape = input_data.type().tensor_type().shape();
    std::cout << "  " << input_data.name() << ":";
    std::cout << "[";
    if (shape.dim_size() != 0)
    {
      int size = shape.dim_size();
      for (int i = 0; i < size - 1; ++i)
      {
        print_dim(shape.dim(i));
        std::cout << ", ";
      }
      print_dim(shape.dim(size - 1));
    }
    std::cout << "]\n";
  }
}

int main(int argc, char **argv)
{
  std::ifstream input("mobilenet.onnx", std::ios::ate | std::ios::binary); // open file and move current position in file to the end

  std::streamsize size = input.tellg(); // get current position in file
  input.seekg(0, std::ios::beg); // move to start of file

  std::vector<char> buffer(size);
  input.read(buffer.data(), size); // read raw data

  onnx::ModelProto model;
  model.ParseFromArray(buffer.data(), size); // parse protobuf

  auto graph = model.graph();

  std::cout << "graph inputs:\n";
  print_io_info(graph.input());

  std::cout << "graph outputs:\n";
  print_io_info(graph.output());
  return 0;
}
binarman
  • 121
  • 4