metadata

license: mit
datasets:
  - s-nlp/paradetox
language:
  - en
library_name: transformers
tags:
  - text2text-generation
widget:
  - text: 'Toxic version: who the fuck are you?'
    example_title: toxic text 1
  - text: 'Toxic version: time to pull this rubbish .'
    example_title: toxic text 2
  - text: 'Toxic version: how many niggers does it take to change a light bulb ?'
    example_title: toxic text 3

This model is a finetuned T5 base model, for detoxifying texts.

How to Prompt ?

Input text : "Toxic version: <some toxic text>"
Then the output will be like : "Non-toxic version: <the non-toxic version of input text>"