Divvy is a Python library for parallelized optimisation of multiple experiments - if you can execute your experiment with commandline you can Divvy it!. It allows for efficient testing by allowing smarter optimisation of experimental parameters in a distribution fashion. Divvy supports a generic interface to a variety of standard optimisers with a focus on Bayesian Optimisation for data efficient parameter learning.
- python-yaml
Global installation:
sudo python setup.py install
User installation:
python setup.py install --user
divvy <path/to/config/file>
An example of configuration file file is given in examples/config.yaml.