Installation

scikit-fallback supports Python 3.9+ and depends on scikit-learn>=1.0, numpy, and scipy. An optional requirement is matplotlib (for visualization of metrics from skfb.metrics).

Install with pip

Usually these dependencies all pre-installed in any ML-powered environment; otherwise, scikit-fallback tries installing newer versions of the dependencies along with itself:

pip install -U scikit-fallback

Note

You might encounter warnings from scikit-fallback if you have scikit-learn<1.3. They are about several private features of scikit-learn and shouldn’t affect performance and API. Also, note that some older versions of scikit-learn don’t support numpy~=2.0.

Build from Source

To build scikit-fallback from source, clone the project, set up an environment, install the package and all the dependencies:

git clone https://github.com/sanjaradylov/scikit-fallback.git
cd scikit-fallback
# Your environment activation here, e.g., venv
python -m pip install -e ".[tests]"
pre-commit install
pre-commit run --all-files