Skip to content
This repository has been archived by the owner on Feb 4, 2022. It is now read-only.

An experimental version of the GPT-J-6B Colab notebook that supports repetition_penalty from Hugging Face's transformers.

Notifications You must be signed in to change notification settings

VE-FORBRYDERNE/gpt-j-6b-filter-test

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

4 Commits
 
 
 
 

Repository files navigation

GPT-J-6B is a GPT-like language model from EleutherAI for text generation.

This is a version of their inference Colab notebook with an additional repetition_penalty setting that makes tokens that have already appeared less likely to reappear in the hopes that it improves the quality of the output text. The technique was originally described in section 4.1 of this paper.

You can try out this new Colab notebook here.

About

An experimental version of the GPT-J-6B Colab notebook that supports repetition_penalty from Hugging Face's transformers.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published