intel-openvino
How to export PyTorch models with unsupported layers to ONNX and then to Intel OpenVINO