Simplify usage; integrate Sentence Transformers (+ LlamaIndex/LangChain, etc.)

#1
by tomaarsen HF staff - opened

Hello!

Preface

First of all, congratulations on your model release and MTEB placement. I agree completely with your paper's perspective that data is key, and data quality is the biggest make-or-break. I'm working on more convenient hard negative mining in Sentence Transformers to help common practitioners improve their data quality. Out of curiosity, are you planning on releasing your curated synthetic data?

Also, I like your work on lower precision evaluations.

Pull Request overview

  • Integrate model with Sentence Transformers

Details

I've integrated your model with Sentence Transformers, allowing more convenient usage. Sentence Transformers has also been integrated with many third party applications (LangChain, LlamaIndex, Haystack, SetFit, BERTopic, etc.), that would allow your model to be used more easily. I've included a usage snippet in the README.md in this PR. As you can see, the results are identical.

  • Tom Aarsen
tomaarsen changed pull request status to open
Junseong changed pull request status to merged
Linq org

Thank you for the PR to integrate the model to Sentence Transformers.

While we don't have immediate plans to release the data, we consider sharing more detailed information in the next version.
This would depend on whether we write a more comprehensive technical report to validate our hypothesis during the model's further development.

Thank you again.

Sign up or log in to comment