FastDeploy: Full-Scene, High-Performance AI Deployment Tool (Presented by Baidu Online Network Technology (Beijing) Co., Ltd.)

FastDeploy is a full-scene, extremely efficient, easy-to-use and flexible AI deployment toolkit for cloud, mobile, and edge. It unifies Paddle and the ecological AI Deployment Engine API including Paddle Inference, Paddle Lite, TensorRT, ONNX Runtime, Poros, and other inference engines to help developers flexibly switch multiple inference engine backends with a single command. It also integrates Triton Inference Server to help developers rapidly deploy to cloud, mobile, and edge in one toolkit. Integrating AI acceleration libraries such as CV-CUDA, FastTokenier, FlyCV, and PaddleSlim automatic compression tool achieves end-to-end performance optimization of AI models. FastDeploy designs a unified deployment API for different languages, and you only need three lines of core code to achieve high-performance AI deployment. You can complete the industrial AI deployment with the 160-plus state-of-the-art models demo.