A high-performance neural network inference framework optimized for mobile, embedded, and desktop model deployment.