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

Adds support for Stella_en_v5 embedding model - 1.5B variant #2551

Merged
merged 7 commits into from
Oct 13, 2024

Conversation

AnubhabB
Copy link
Contributor

@AnubhabB AnubhabB commented Oct 8, 2024

Stella_en_1.5B_v5 is a top ranking model (top 5) in Retrieval and Reranking tasks as of 8th October 2024.

Model Card

This PR adds support for the model along with some examples.

Tasks:

  • Check license: Model is licensed MIT
  • Check numerical accuracy - Verified against source pytorch implementation. At f32 candle implementation yields mathematically equivalent up to a 3-decimal places rounded .. beyond which we observe some variations. Code used to verify numeric equivalence can be found here.
  • Authors example from the model card added and reproduced.

Note:

Stella_400M seems to follow a significantly different implementation. Have not implemented it in this PR.

Closes #2525

@LaurentMazare LaurentMazare merged commit f553ab5 into huggingface:main Oct 13, 2024
@LaurentMazare
Copy link
Collaborator

Thanks!

EricLBuehler pushed a commit to EricLBuehler/candle that referenced this pull request Nov 26, 2024
…face#2551)

* Stella_en_1.5B_v5

* Separated  creation. This is a critical step for numerical accuracy and would be documented in the readme

* EmbedDim would require clone and copy

* WIP: example

* Examples added

* a litte more in README
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.

[QUESTION] Protocol of adding a new model (Stella_en_<*>_v5 family) implementation with Candle
2 participants