Transparency

The transparency module configures the explainers and visualisers that generate attributions for your model predictions.

Each transparency entry defines one named explainer, its algorithm, and the visualisers that should render its outputs. The current implementation supports Captum- and SHAP-based explainers behind the same config surface.

Transparency module documentation

• Configuration
• Supported libraries
• Output