TensorRT

NVIDIA SDK for high-performance deep learning inference on GPUs. Optimizes trained models for deployment with plugins, an ONNX parser, and example applications for developers.