Add bge-m3 dense, sparse, multi-vector

embeddings-benchmark / mteb

MTEB: Massive Text Embedding Benchmark

https://arxiv.org/abs/2210.07316

Apache License 2.0

2k stars 277 forks source link

Add bge-m3 dense, sparse, multi-vector #1529

Open yjoonjang opened 5 days ago

yjoonjang commented 5 days ago

https://huggingface.co/BAAI/bge-m3

bge-m3 model has 3 versions: dense, sparse, multi_vector. However, sentence-transformers only allow dense versions, so I think it would be great if mteb makes it available to use other versions.

Can I work on this ?

Samoed commented 4 days ago

Of course! That would be great!

KennethEnevoldsen commented 2 days ago

Probably related to #1211. We currently have no good way to log variations on a model (though I think it is something we should figure out how to do)