Captum.Ai

Open Site

3.50

Visit website

Captum is a model interpretability library designed for use with PyTorch, facilitating the analysis of machine learning models across various modalities, including vision and text. It offers multi-modal support, enabling users to explore and understand model predictions without extensive modifications to the original architecture.

The library is open-source and allows researchers to benchmark new interpretability algorithms efficiently. With its simple installation process via Conda or Pip, users can quickly set up the environment to start implementing interpretability techniques.