The ONNX-mlpack converter can take any ONNX graph and convert it to an mlpack neural network. This is done by matching individual mlpack layers to subgraphs of the ONNX network using a rule matching engine, and then selecting the overall best match of the network.

1. Dependencies: make sure that you have, available on your system,

• mlpack and its dependencies ensmallen, cereal, and Armadillo](https://arma.sourceforge.net); if not installed, these will be autodownloaded during the CMake comnfiguration
• ONNX (on Debian, libonnx-dev is sufficient)
• Protobuf (on Debian, libprotobuf-dev is sufficient)

2. Configuration: create a build directory and use CMake to configure:

mkdir build/
cd build/
cmake ../

3. Build the converter:

make onnx_mlpack_converter

4. Run the converter:

src/onnx_mlpack_converter <input_network.onnx> <output_mlpack_network.bin>

Once the converter has been run, you can load the network into mlpack as a DAGNetwork:

#define MLPACK_ENABLE_ANN_SERIALIZATION
#include <mlpack.hpp>

using namespace mlpack;

int main()
{
  DAGNetwork<> network;
  Load("output_mlpack_network.bin", network);

  // now you can use network.Predict(), network.Train(), etc.
}

This repository is under active development! At this particular moment, support is pretty primitive, but it is actively being expanded. If you have a network that does not match properly, or encounter other problems, please feel free to open a bug report and we will look into the issue.