Do I need to normalize embedded vector when encode the query/documents?

#1
by Phicks2105 - opened

I read some article said that I can normalize the embedding vector when using the encode method from sentence-transformers to increase accuracy when performing cosine-similarity. It also said that for some model, it already normalize that for me. So do I really need to do it with this model or I just don't need it.

Thank you

Sign up or log in to comment