Installation

BOSS is written in Python 3 and distributed as a PyPI package.

Virtual environments (recommended)

We recommend installing BOSS inside a virtual enviroment (venv, conda…).

To create a virtual environment ‘boss’ with venv:

python3 -m venv boss
source boss/bin/activate

or with conda:

conda create -n boss
conda activate boss

The latest stable BOSS version can then be installed using pip:

python3 -m pip install aalto-boss

User installation

If you are not using virtual enviroments, we recommend doing a user installation with pip:

python3 -m pip install --user aalto-boss

On Linux, this will place the python module under $HOME/.local/lib/python3.X/site-packages where X should be substituted for the minor version of your Python distribution. During the installation an executable called boss, which is used to run BOSS, is built and placed under $HOME/.local/bin. In order for your BOSS installation to function properly, these paths must (if not already present) be appended to the correct enviroment variables:

export PATH=$PATH:$HOME/.local/bin
# Do not forget to substitute X for your Python minor version
export PYTHONPATH=$PYTHONPATH:$HOME/.local/lib/python3.X/site-packages

To check that the installation was successful, calling the boss executable without any arguments should result in usage instruction being printed. During the installation you may see a large number of dependencies being installed by pip. Notably, BOSS uses the Gaussian process framework GPy as a backend.

Development version

To install the development version, clone the source code from GitLab and install with pip:

git clone https://gitlab.com/cest-group/boss.git
cd boss
python3 -m pip install --user .

Experimental features

After acquiring the source code, experimental features can be installed from their own development branches:

git checkout [branch_name]
cd boss
python3 -m pip install --user .

WARNING: Experimental features that are in development may produce inaccurate results. Use at your own risk!

Known issues

Rarely, on HPC systems, warnings related to the use of cython in GPy have been observed. This can typically be resolved by using a more fully-featured Python distribution that ships with cython, such as Anaconda.