Optilab is a lightweight and flexible python framework for testing black-box optimization.
- ✅ Intuitive interface to quickly prototype and run optimizers and metamodels.
- 📚 High quality documentation.
- 📈 Objective functions, optimizers, plotting and data handling.
- ⋙ CLI functionality to easily summarize results of previous experiments.
- 🚀 Multiprocessing for faster computation.
Optilab has been tested to work on python versions 3.11 and above. To install it from PyPI, run:
pip install optilab
You can also install from source by cloning this repo and running:
make install
Learn how to use optilab by using our demo notebook. See demo/tutorial.ipynb.
Optilab comes with a powerful CLI tool to easily summarize your experiments. It allows for plotting the results and performing statistical testing to check for statistical significance in optimization results.
Optilab CLI utility.
usage: python -m optilab [-h] [--hide_plots] [--test_y] [--test_evals] pickle_path
positional arguments:
pickle_path Path to pickle file or directory with optimization runs.
options:
-h, --help show this help message and exit
--hide_plots Hide plots when running the script.
--test_y Perform Mann-Whitney U test on y values.
--test_evals Perform Mann-Whitney U test on eval values.
This project comes with a docker container. You can pull it from dockerhub:
docker pull mlojek/optilab
Or build it yourself:
make docker