Learning Interpretability Tool (LIT)

Welcome to 🔥LIT, the Learning Interpretability Tool!

If you want to jump in and start playing with the LIT UI, check out the hosted demos at https://pair-code.github.io/lit/demos/.

• Main Site
• Getting Started
  • Installation
  • Hosted demos
  • LIT with your model
  • Using LIT in notebooks
  • Stand-alone components
  • Run an existing example
• Examples
  • Classification
  • Regression / Scoring
  • Sequence-to-Sequence
  • Language Modeling
  • Structured Prediction
  • Multimodal
• UI Guide
  • General Layout
  • Datapoint Selections
  • Toolbars
  • Comparing Models
  • Comparing Datapoints
  • Slices
  • Module Details
  • User Journeys
• Components & Features
  • Framework and Model Support
  • Input and Output Types
  • Token-based Salience
  • Pixel-based Salience
  • Attention
  • Embedding Projector
  • Aggregate Analysis
  • TCAV
  • Counterfactual Analysis
• Python API
  • Design Overview
  • Adding Models and Data
  • Datasets
  • Models
  • Interpretation Components
  • Type System
  • Server Configuration
  • Accessing the LIT UI in Notebooks
  • Using LIT components outside of LIT
• Frontend Development
  • High Level Overview
  • Application Architecture
  • Modules (LitModule)
  • Style Guide
  • Development Tips (open-source)
  • Custom Client / Modules
• Running in Docker
  • Basic Usage
  • Integrating Custom LIT Instances with the Default Docker Image
  • Building Your Own Image
• Glossary
• FAQ
  • Model and Data Types
  • Languages
  • Scale
  • Privacy and Security
  • Workflow and Integrations
• GitHub