ONNX Runtime is a cross-platform machine-learning accelerator for both inference and training. It enables faster model execution by leveraging hardware accelerators, graph optimizations, and transforms across different hardware, drivers, and operating systems. The inference component supports models from deep learning frameworks like PyTorch and TensorFlow/Keras as well as classical ML libraries including scikit-learn, LightGBM, and XGBoost. The training component accelerates multi-node GPU training for transformer models through a one-line integration with existing PyTorch scripts.