onnxruntime

ONNX Runtime is a cross-platform machine-learning accelerator designed for both inference and training. Key capabilities include: - Inference Acceleration: Supports models from deep learning frameworks like PyTorch and TensorFlow/Keras, as well as classical ML libraries such as scikit-learn, LightGBM, and XGBoost. It optimizes performance through graph transformations and hardware accelerators across various operating systems and drivers. - Training Acceleration: Provides acceleration for transformer model training on multi-node NVIDIA GPUs, integrating with existing PyTorch training scripts. The project is compatible with multiple hardware configurations and provides a comprehensive set of documentation, tutorials, and sample repositories for both inferencing and training workflows.