unitaryai/detoxify
View on GitHub ↗ ★ 1,314 Python
Trained models & code to predict toxic comments on all 3 Jigsaw Toxic Comment Challenges. Built using ⚡ Pytorch Lightning and 🤗 Transformers.
3 mentions on Hacker News by 3 people: 2 stories, 1 comment · first 2020-11-12 · latest 2021-11-10
20082026
-
…sadly still can happen :-/ We have fully analyzed the training dataset (1128 GB) using Detoxify ( https://github.com/unitaryai/detoxify ) to filter out problematic content. But of course detecting toxicity is a tough challenge in itself,…
comment by MasterScrat · 2021-11-10 · in Show HN: Cedille, the largest French language model, released in open source