intfloat

/

multilingual-e5-large

like 803

You can use any dense retrieval training framework, just replace the model initialization with this one.

Personally, I would recommend Tevatron and SimLM codebase.

Hi,
What codebase can I refer to if I want to implement the pre-training stage in your e5 paper?

@Chuzhan The implementation for pre-training part is even simpler, you can use the same codebase by removing the hard negatives.

Is there a tutorial on how to fine tune this model on a Greek sentence similarity dataset?

I am not aware of any, but you can adapt existing codebase for your specific needs.