Open-source toolkit for optimizing and deploying AI models from multiple frameworks, converting vision, speech, and generative models to run efficiently on CPU, GPU, and NPU hardware.