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 user's guide-101 for the simplest of demonstrations!
As simple and as complex you want
- Simple and natural, but also explicit and verbose, search domain definitions
- Minimal and non-intrusive client interface for reporting target function values
- Database logging (currently powered by MongoDB)
- Flexible configuration
- Explicit experiment termination conditions
- Algorithms algorithms algorithms: Skopt's bayesian optimizers are at hand without writing. Random search is the default. only a single line of code.
- More algorithms: Implementing and distributing algorithms is as easy as possible! Check developer's guide-101. Expect algorithm plugins to pop out quickly!
- Came up with an idea? Your intuition is still at play: Help your optima hunter now by a command line interface.
- And other many more already there or coming soon!
Install Oríon (beta) by running:
pip install git+https://github.com/mila-udem/orion.git@master
For more information read the full installation docs. .. TODO Once website is published use a full hyperlink to the installation page
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:
- Issue Tracker: github.com/mila-udem/orion/issues
- Source Code: github.com/mila-udem/orion
Start by starring and forking our Github repo!
Thanks for the support!
The project is licensed under the BSD license.