facebook / Ax

Ax is an accessible, general-purpose platform for understanding, managing, deploying, and automating adaptive experiments. Adaptive experimentation is the machine-learning guided process of iteratively exploring a (possibly infinite) parameter space in order to identify optimal configurations in a resource-efficient manner. Ax currently supports Bayesian optimization and bandit optimization as exploration strategies. Bayesian optimization in Ax is powered by BoTorch, a modern library for Bayesian optimization research built on PyTorch. For full documentation and tutorials, see the Ax website Why Ax? • **Expressive API**: Ax has an expressive API that can address many real-world optimization tasks. It handles complex search spaces, multiple objectives, constraints on both parameters and outcomes, and noisy observations. It supports suggesting multiple designs to evaluate in parallel (both synchronously and asynchronously) and the ability to early-stop evaluations. • **Strong performance out of the box**: Ax abstracts away optimization details that are important but obscure, providing sensible defaults and enabling practitioners to leverage advanced techniques otherwise only accessible to optimization experts. • **State-of-the-art methods**: Ax leverages state-of-the-art Bayesian optimization algorithms implemented in BoTorch, to deliver strong performance across a variety of problem classes. • **Flexible:** Ax is highly configurable, allowing researchers to plug in novel optimization algorithms, models, and experimentation flows. • **Production ready:** Ax offers automation and orchestration features as well as robust error handling for real-world deployment at scale. Getting Started To run a simple optimization loop in Ax (using the Booth response surface as the artificial evaluation function): Installation Ax requires Python 3.11 or newer. A full list of Ax's direct dependencies can be found in pyproject.toml. We recommend installing Ax via pip, even if using Conda environment: Installation will use Python wheels from PyPI, available for OSX, Linux, and Windows. _Note_: Make sure the being used to install is actually the one from the newly created Conda environment. If you're using a Unix-based OS, you can use to check. Installing with Extras Ax can be installed with additional dependencies, which are not included in the default installation. For example, in order to use Ax within a Jupyter notebook, install Ax with the extra: Extras for using Ax with MySQL storage ( ), for running Ax's tutorial's locally ( ), and for installing all dependencies necessary for developing Ax ( ) are also available. Install Ax from source You can install the latest (bleeding edge) version from GitHub using . The bleeding edge for Ax depends on bleeding edge versions of BoTorch and GPyTorch. We therefore recommend installing those from Github as well. Join the Ax Community Getting help Please open an issue on our issues page with any questions, feature requests or bug reports! If posting a bug report, please include a minimal reproducible example (as a code snippet) that we can use to reproduce and debug the problem you encountered. Contributing See the CONTRIBUTING file for how to help out. When contributing to Ax, we recommend cloning the repository and installing all optional dependencies: See recommendation for installing PyTorch for MacOS users above. The above example limits the cloned directory size via the argument to . If you require the entire commit history you may remove this argument. Citing Ax If you use Ax, please cite the following paper: > M. Olson, E. Santorella, L. C. Tiao, S. Cakmak, D. Eriksson, M. Garrard, S. Daulton, M. Balandat, E. Bakshy, E. Kashtelyan, Z. J. Lin, S. Ament, B. Beckerman, E. Onofrey, P. Igusti, C. Lara, B. Letham, C. Cardoso, S. S. Shen, A. C. Lin, and M. Grange. Ax: A platform for Adaptive Experimentation. In AutoML 2025 ABCD Track, 2025. License Ax is licensed under the MIT license.