C++ inference: No longer requires Python for inference

Wednesday, August 23, 2017


Posted by Yukio Oobuchi

With version 0.9.4, Neural Network Libraries supports an inference in C++.

Even on the environments where it is difficult to run Python due to various restrictions, you can do inference with a trained model which is generated by save function, or exported by Neural Network Console *1.

Please see MNIST runtime example for more details.

  • *1 With version 0.9.4, it doesn’t support C++ inference on a network which contains RepeatStart, RepeatEnd, ReccurentInput, RecurrentOutput, Delay, and LSTM.