What is Captum?
Captum is a model interpretability library for PyTorch. It provides a range of tools and algorithms to help users understand and interpret the behavior of their PyTorch models. The library supports multiple modalities, including vision and text, enabling the analysis of diverse model types.
As an open-source and extensible library, Captum facilitates research in interpretability. Users can readily implement and benchmark new interpretability algorithms, contributing to the advancement of the field. Captum integrates smoothly with existing PyTorch models, requiring only minimal adjustments.
Features
- Multi-Modal: Supports interpretability of models across modalities including vision, text, and more.
- Built on PyTorch: Supports most types of PyTorch models and can be used with minimal modification to the original neural network.
- Extensible: Open source, generic library for interpretability research. Easily implement and benchmark new algorithms.
Use Cases
- Understanding the decision-making process of PyTorch models.
- Debugging and improving the performance of PyTorch models.
- Implementing and benchmarking new interpretability algorithms.
- Analyzing models across various modalities, including vision and text.
