On the recordRead 14 Nov 202612 facts
Captum
Plans and prices
Captum
Free
Open-source library · Installation via conda or pip
Read 14 Nov 2026 · captum.ai
On the record
- Algorithms and metrics
- Beyond attribution methods, Captum includes infidelity and sensitivity metrics to estimate explanation trustworthiness.captum.ai · read 14 Nov 2026
- Attribution algorithms
- Its algorithms help identify features contributing to a model’s output, including Integrated Gradients.captum.ai · read 14 Nov 2026
- DeepLIFT limitation
- The documented DeepLIFT implementation supports the Rescale rule; RevealCancel is not supported in the described implementation.captum.ai · read 14 Nov 2026
- Distribution
- The site documents installation using either conda or pip.captum.ai · read 14 Nov 2026
- Maintainer
- The site footer identifies Meta Platforms, Inc. in its copyright notice.captum.ai · read 14 Nov 2026
- Model compatibility caveat
- Some hook-based methods do not work with functional nonlinearities; JIT models do not support hook-dependent methods.captum.ai · read 14 Nov 2026
- Parallel models
- The FAQ states that DataParallel and DistributedDataParallel models are supported with all model types.captum.ai · read 14 Nov 2026
- PyTorch models
- Captum supports most types of PyTorch models and is intended to require minimal changes to the original neural network.captum.ai · read 14 Nov 2026
- Research use
- Researchers can implement interpretability algorithms and benchmark them against methods in the library.captum.ai · read 14 Nov 2026
- Supported modalities
- The product site says Captum supports interpretability across vision, text, and other modalities.captum.ai · read 14 Nov 2026
- What it does
- Captum is an open-source library for interpreting machine-learning models built on PyTorch.captum.ai · read 14 Nov 2026
- Who it is for
- The stated primary audiences are model developers and interpretability researchers; application engineers are also identified as users.captum.ai · read 14 Nov 2026
