Running, Deploying Fastai successes: Links, Tutorials, Stories

For running inference on your GPU from C++ applications I would suggest using OpenCV once you have exported your ONNX model from Pytorch (OpenCV 4.2.0 has CUDA DNN backend).