The Language Interpretability Tool (LIT) now has a
website with documentation, tutorials, and hosted
demos. With the demos, you can explore LIT on a variety of different tasks and model types (for example: classification for sentiment analysis using BERT, text generation with T5) without needing to install LIT yourself.
Also we've released version 0.2 of LIT today. This version has an improved UI, bug fixes, and a number of new features, including customizable layouts, and salience maps through integrated gradients. Install the new version either through pip (`pip install lit-nlp`), or by cloning the GitHub repo and building it locally. More details on setup can be found on the website's
setup guide.
Please join this lit-annoucements Google Group to stay up to date on LIT as we continue to build it out. For feedback, such as feature requests and bug reports, please use
GitHub issues.
- The LIT team