Oríon is an asynchronous framework for black-box function optimization.
Its purpose is to serve as a meta-optimizer for machine learning models and training, as well as a flexible experimentation platform for large scale asynchronous optimization procedures.
Core design value is the minimum disruption of a researcher's workflow. It allows fast and efficient tuning, providing minimum simple non-intrusive (not even necessary!) helper client interface for a user's script.
So if ./run.py --mini-batch=50 looks like what you execute normally,
now what you have to do looks like this:
orion -n experiment_name ./run.py --mini-batch~'randint(32, 256)'
Check out our getting started guide or this presentation for an overview, or our scikit-learn example for a more hands-on experience. Finally we encourage you to browse our documentation.
Effortless to adopt, deeply customizable
- Adopt it with a single line of code
- Natively asynchronous, thus resilient and easy to parallelize
- Offers the latest established hyperparameter algorithms
- Elegant and rich search-space definitions
- Comprehensive configuration system with smart defaults
- Transparent persistence in local or remote database
- Integrate seamlessly your own hyper-optimization algorithms
- Language and configuration file agnostic
Install Oríon by running $ pip install orion. For more information consult the installation
guide.
- 2021-07-14 - SciPy 2021 (Video) (Slides)
- 2021-05-19 - Dask Summit 2021 (Video) (Slides)
- 2021-03-16 - AICamp (Video) (Slides)
- 2019-11-28 - Tech-talk @ Mila (Video) (Slides)
Do you have a question or issues? Do you want to report a bug or suggest a feature? Name it! Please contact us by opening an issue in our repository below and checkout our contribution guidelines:
- Issue Tracker: https://github.com/epistimio/orion/issues
- Source Code: https://github.com/epistimio/orion
Start by starring and forking our Github repo!
Thanks for the support!
If you use Oríon for published work, please cite our work using the following bibtex entry.
@software{xavier_bouthillier_2022_0_2_6,
author = {Xavier Bouthillier and
Christos Tsirigotis and
François Corneau-Tremblay and
Thomas Schweizer and
Lin Dong and
Pierre Delaunay and
Fabrice Normandin and
Mirko Bronzi and
Dendi Suhubdy and
Reyhane Askari and
Michael Noukhovitch and
Chao Xue and
Satya Ortiz-Gagné and
Olivier Breuleux and
Arnaud Bergeron and
Olexa Bilaniuk and
Steven Bocco and
Hadrien Bertrand and
Guillaume Alain and
Dmitriy Serdyuk and
Peter Henderson and
Pascal Lamblin and
Christopher Beckham},
title = {{Epistimio/orion: Asynchronous Distributed Hyperparameter Optimization}},
month = august,
year = 2022,
publisher = {Zenodo},
version = {v0.2.6,
doi = {10.5281/zenodo.3478592},
url = {https://doi.org/10.5281/zenodo.3478592}
}See ROADMAP.md.
The project is licensed under the BSD license.