optilab 21

Simple framework for optimization functions and metamodels

Optilab

Optilab is a lightweight and flexible python framework for testing black-box optimization.

Features

• ✅ 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.

How to install

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

Try the demos

Learn how to use optilab by using our demo notebook. See demo/tutorial.ipynb.

CLI tool

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.

usage: python -m optilab [-h] [--hide_plots] [--test_y] [--test_evals] [--entries ENTRIES [ENTRIES ...]] [--raw_values] [--hide_outliers] pickle_path

Optilab CLI utility.

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.
  --entries ENTRIES [ENTRIES ...]
                        Space separated list of indexes of entries to include in analysis.
  --raw_values          If specified, y values below tolerance are not substituted by tolerance value.
  --hide_outliers       If specified, outliers will not be shown in the box plot.

Docker

This project comes with a docker container. You can pull it from dockerhub:

docker pull mlojek/optilab

Or build it yourself:

make docker