Captum

A PyTorch library designed for understanding model predictions across different modalities.

Back to Search Results
The AI REPORT pick
Dev Tools
Research & Insights
Open Source
Overview
ABOUT

Captum serves as a PyTorch library dedicated to enhancing model interpretability. It offers a range of tools and methods that facilitate the understanding and attribution of predictions made by PyTorch models in diverse domains, such as visual data and textual content. Compatible with the majority of PyTorch models, Captum simplifies the integration and evaluation of new interpretability algorithms, making it a valuable resource for researchers and developers alike.

USE CASE

Research & Insights

KEY FEATURES
  • Supports multi-modal interpretability
  • Built on the PyTorch framework
  • Open source and easily extensible
Pricing
Open Source
Free only