search

stanford-crfm / helm

Holistic Evaluation of Language Models (HELM), a framework to increase the transparency of language models (https://arxiv.org/abs/2211.09110). This framework is also used to evaluate text-to-image models in Holistic Evaluation of Text-to-Image Models (HEIM) (https://arxiv.org/abs/2311.04287).
https://crfm.stanford.edu/helm
Apache License 2.0
1.89k stars 244 forks source link

color code tokens for language modeling #962

Open percyliang opened 2 years ago

percyliang commented 2 years ago

When displaying language modeling results, color code each token based on its probability.

rishibommasani commented 2 years ago

Feel free to revert, but I think this is a P3 for now.

yifanmai commented 1 year ago

Noting that @dlwh implemented this for Levanter. We could consider turning it into a reusable React component.

dlwh commented 1 year ago

mine's a crappy pure html page (and needs to be), but the code is here https://github.com/stanford-crfm/levanter/blob/main/src/levanter/visualization.py#L12

would probably be better to quantize or use client-side html to do the interpolation (css won't :( )