Installation and dependencies
Stable installation
Section titled “Stable installation”python -m pip install model-auditor==0.1.16The distribution name contains a hyphen. Python imports use an underscore:
from model_auditor import Auditor, InferenceConfigAuditor and InferenceConfig are exported from the package root. Import metrics
and the remaining public dataclasses from their modules.
Python requirement
Section titled “Python requirement”Package metadata requires:
Python >= 3.10The stable classifiers name Python 3.10, 3.11, and 3.12. There is no declared upper bound.
Core dependencies
Section titled “Core dependencies”Installing the package resolves:
| Dependency | Minimum version | Use |
|---|---|---|
| pandas | 2.2 | DataFrame storage, grouping, display, and resampling |
| NumPy | 2.1 | numerical arrays, thresholds, percentiles, and histogram bins |
| scikit-learn | 1.5 | ROC construction, AUROC, and average precision |
| SciPy | 1.14 | exact intervals and calibration fitting |
| tqdm | 4.0 | feature-level evaluation progress display |
The package is pure Python and declares no compiled extension of its own.
Optional dependencies
Section titled “Optional dependencies”Install the declared plotting extra for Matplotlib and Plotly:
python -m pip install 'model-auditor[plotting]'Install the styling extra for Jinja2-backed pandas tables:
python -m pip install 'model-auditor[styling]'Core evaluation imports neither extra.
Notebook environment
Section titled “Notebook environment”Jupyter is not required by the library, but style_dataframe() is designed for notebook display:
python -m pip install jupyterlab 'model-auditor[plotting,styling]'Verify versions
Section titled “Verify versions”import importlib.metadata
print(importlib.metadata.version("model-auditor"))Expected stable version:
0.1.16Source installation
Section titled “Source installation”git clone https://github.com/beatrice-b-m/model-auditor.gitcd model-auditorgit checkout v0.1.16
python -m venv .venvsource .venv/bin/activate
python -m pip install --upgrade pippython -m pip install -e '.[dev]'The project uses setuptools with setuptools-scm, so a Git checkout and tags determine the dynamic version.
The development extra includes pytest, Ruff, build, plotting, and styling tools:
python -m ruff check .python -m ruff format --check .python -m pytest -qpython -m build