KEMBAR78
Refactoring tying mechanism + small fixes by NouamaneTazi · Pull Request #62 · huggingface/nanotron · GitHub
Skip to content

Conversation

NouamaneTazi
Copy link
Member

@NouamaneTazi NouamaneTazi commented Feb 12, 2024

  • fix tied embeddings in llama (we didn't tie before)
  • refactor KV weights tying

This PR is a preparation for a next PR to remove module_id_to_prefix from nanotron

✅ tied weights tests pass

Copy link
Member

@3outeille 3outeille left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

lgtm !

@NouamaneTazi NouamaneTazi merged commit 939522e into main Feb 12, 2024
Copy link
Member

@thomwolf thomwolf left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Very nice!

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

Successfully merging this pull request may close these issues.

3 participants