Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[Feature] DRY repetition penalty #1350

Closed
2 tasks done
ghost opened this issue Sep 8, 2024 · 3 comments
Closed
2 tasks done

[Feature] DRY repetition penalty #1350

ghost opened this issue Sep 8, 2024 · 3 comments

Comments

@ghost
Copy link

ghost commented Sep 8, 2024

Checklist

Motivation

DRY is a modern repetition penalty which ramps up penalties on N-grams to avoid looping behavior.

The penalty is commonly used on Llama 3.1 8B by practitioners to avoid repetition spirals.

Related resources

oobabooga/text-generation-webui#5677

https://www.reddit.com/r/SillyTavernAI/comments/1eg2pq5/good_info_on_dry_to_get_you_started_has/

@zhyncs
Copy link
Member

zhyncs commented Sep 8, 2024

@vnkc1 ref #1187

@ghost
Copy link
Author

ghost commented Sep 8, 2024

Thanks

@ghost
Copy link
Author

ghost commented Sep 8, 2024

Closing this issue

@ghost ghost closed this as completed Sep 8, 2024
This issue was closed.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant