Calling Both Multivector Modes
#6
by
kamilersz
- opened
Is it possible to get both dense and multivector embedding in a single call?
I believe this will optimize performance given query and reranker will need both types of vectors
Hi kamilersz, you can generate multivector embedding ? Can you share me a example to do it ?
If you mean just only multivector embedding, just set the return multivector params to True. It's already provided in the API
How about use VLLM directly?