Version 0.3 released - now usable directly in notebook environments

60 views

Skip to first unread message

James Wexler

unread,

Apr 7, 2021, 8:17:02 AM4/7/21

to lit-annoucements

Version 0.3 of The Language Interpretability Tool (LIT) has been released. It contains UI improvements, bug fixes, and a number of new features, including:

Notebook mode - Use LIT directly inside of Jupyter or colab notebooks (demo).
New annotated text visualization module added.
Allow saving/loading of generated datapoints, and dynamic adding of new datasets by path in the UI.
Added synchronized scrolling between duplicated modules when comparing datapoints or models.
Added a focus service for visually linking focus (i.e. hover) states between components.
Allow layouts to be specified on LIT creation in python.

We hope you enjoy the updates and we look forward to adding tons of new features and continually improving the UI. For feedback, such as feature requests and bug reports, please use GitHub issues.