Coremltools

Core ML Tools is Apple's open source Python package for converting trained machine learning models into the Core ML format so they can run inside apps on Apple devices. Developers train a model in a third party framework, convert it with coremltools into a Core ML model package, and then integrate it into an iOS, macOS or other Apple platform app through the Core ML framework. It is used by app developers and machine learning engineers shipping on device AI. The package converts neural networks from TensorFlow 1 and 2 and PyTorch, tree ensembles from XGBoost and scikit learn, generalized linear models and support vector machines from scikit learn and LIBSVM, and scikit learn preprocessing pipelines. Beyond conversion, it reads, writes and optimizes Core ML models to reduce storage, power consumption and inference latency, and on macOS it can verify a conversion by running predictions. Because Core ML runs models on the device using the CPU, GPU and Neural Engine, apps need no network connection for inference, which keeps user data private and responses fast. Core ML Tools is free. It is published by Apple under a permissive open source license, installs with pip, and has no paid tier. The listed readme.io address now redirects to the official Core ML Tools Guide hosted on GitHub Pages, which, alongside an API reference and examples, is the current home of the documentation.