When deploying large-scale deep learning applications, C++ may be a better choice than Python to meet application demands or to optimize model performance. Therefore, I specifically document my recent ...

While Facebook is working towards unifying PyTorch and Caffe2, developers can rely on ONNX for exchanging the models between the frameworks.

Microsoft is open-sourcing an optimized version of Google's BERT that uses ONNX Runtime and CPUs or GPUs to speed language model performance.