Fine-tune ColBERT using teacher scores

AnswerDotAI / RAGatouille

Easily use and train state of the art late-interaction retrieval methods (ColBERT) in any RAG pipeline. Designed for modularity and ease-of-use, backed by research.

Apache License 2.0

3.07k stars 210 forks source link

Fine-tune ColBERT using teacher scores #251

Open jens-create opened 2 months ago

jens-create commented 2 months ago

Thanks for making this framework!

I'd like to know if it is possible to fine-tune e.g. the answerai colbert v1 small model using a teacher model and its scores' instead of positive and negative documents?

seplee commented 1 month ago

That would be a sweet feature.