NVIDIA Open-Sources TensorRT Library Components
TensorRT is NVIDIA's flagship platform for deep learning inference and focused for doing so on NVIDIA GPU hardware. TensorRT is built atop CUDA and provides a wealth of optimizations and other features.
Included via NVIDIA/TensorRT on GitHub are indeed sources to this C++ library though limited to the plug-ins and Caffe/ONNX parsers and sample code. Building the open-source TensorRT code still depends upon the proprietary CUDA as well as other common build dependencies. But nice at least seeing the TensorRT code more open now than previously.