Installation¶

Requirements¶

For the differentiable PyTorch implementation, PyTorch 2.10+ is also required.

pip install fractional-information-gain

With PyTorch support (for differentiable loss functions):

pip install "fractional-information-gain[torch]"

To install the latest development version directly from GitHub instead:

pip install git+https://github.com/RenaissancePhilanthropy/fractional-information-gain.git

The package is installed as fractional-information-gain but imported as fig:

import fig
# or
from fig import fractional_information_gain_validation

The PyTorch module is not re-exported from the top-level package. Import it directly:

from fig.fig_torch import fractional_information_gain_validation_torch

The project uses uv for dependency management:

git clone https://github.com/RenaissancePhilanthropy/fractional-information-gain.git
cd fractional-information-gain
uv sync --dev --extra torch

This installs all development dependencies (pytest, ruff, pre-commit) and the PyTorch extra.

To run the test suite:

uv run python -m pytest tests/

The R implementation (figmetric) lives in the r/ subdirectory. See the R Package page for installation instructions.