NVIDIA Triton Inference

NVIDIA open-source inference server — serve any model on any hardware at scale.