ort provides a Rust interface to Microsoft’s ONNX Runtime library, enabling high-performance inference and training on models exported in the Open Neural Network Exchange (ONNX) format. It supports a wide range of hardware accelerators including CUDA, Core ML, DirectML, OpenVINO, and others for device or datacenter deployment. The library targets developers who need to deploy PyTorch, TensorFlow, Keras, scikit-learn, or PaddlePaddle models from Rust applications.