Captum: a Unified and Generic Model Interpretability Library for PyTorch

International Conference on Learning Representations (ICLR)


In this paper we introduce a novel, unified, open-source model interpretability library for PyTorch (Paszke et al., 2019). The library contains generic implementations of a number of gradient and perturbation-based attribution algorithms, also known as feature, neuron and layer importance algorithms, as well as a set of evaluation metrics for these algorithms. It can be used for both classification and non-classification models including graph-structured models built on Neural Networks (NN). In this paper we give a high-level overview of supported attribution algorithms and show how to perform memory-efficient and scalable computations. We emphasize that the three main characteristics of the library are multimodality, extensibility and ease of use. Multimodality supports different modality of inputs such as image, text, audio or video. Extensibility allows adding new algorithms and features. The library is also designed for easy understanding and use, and is a valuable tool for building Responsible AI frameworks.

Featured Publications