Installation
ModelDoctor requires Python 3.9+.
Standard Installation
To install the core diagnostic engine:
Optional Extras
ModelDoctor provides optional dependencies for extended functionality:
# Install dashboard dependencies (Jinja2, HTML rendering)
pip install "modeldoctor[dashboard]"
# Install explainability dependencies (SHAP)
pip install "modeldoctor[shap]"
# Install all optional dependencies
pip install "modeldoctor[all]"
Development Installation
To install ModelDoctor from source for development or contribution:
git clone https://github.com/modeldoctor/modeldoctor.git
cd modeldoctor
# Install in editable mode with dev dependencies (pytest, black, mkdocs)
pip install -e ".[dev]"
Dependency Notes
ModelDoctor relies on the following core dependencies:
- scikit-learn >= 1.0
- numpy >= 1.20
- pandas >= 1.3
- pydantic >= 2.0
SHAP Support
If you want ModelDoctor to automatically compute and render SHAP feature importance charts, you must install the [shap] extra. Without it, ModelDoctor will fallback to native permutation importance.
Dashboard Support
The HTML dashboard relies on Jinja2 for rendering templates. If you plan to call report.dashboard() or export HTML reports, ensure the [dashboard] extra is installed.
Common Installation Errors
ModuleNotFoundError: No module named 'shap'
You are trying to run the ExplainabilityEngine without SHAP installed. Run pip install "modeldoctor[shap]".
TemplateNotFound Error
You are attempting to render the HTML dashboard but the Jinja2 templates are missing or the library was installed without dashboard support. Run pip install "modeldoctor[dashboard]".
Pydantic Validation Errors on Import
ModelDoctor uses Pydantic v2. If you have Pydantic v1 installed, you may see compatibility errors. Upgrade using pip install -U pydantic.