ONNX Runtime
Cross-platform engine for running and training ONNX machine learning models
What it is
ONNX Runtime executes models exported to the ONNX format from PyTorch, TensorFlow, scikit-learn, XGBoost and similar libraries, applying graph optimizations and dispatching work to hardware accelerators through pluggable execution providers. It runs on desktop, server, mobile and web targets, and also offers a training add-on that speeds up transformer training on multi-GPU NVIDIA setups. Developers embed it in applications and services to serve trained models with lower latency and cost across different hardware.