I have created embeddings using local

aasanchez75

I have created embeddings using "local:BAAI/bge-small-en-v1.5" and stored in Elasticsearch. I am using OpenAI for the query engine. However, I got this error

Plain Text

BadRequestError: BadRequestError(400, 'search_phase_execution_exception', 'failed to create query: the query vector has a different dimension [1536] than the index vectors [384]')

. My question is whether I should look for a Huggingface model with vector dimension = 1536 or there is a way to query an VectorStore whose vector dimension is different to the vector dimension used in the query engine?

10 comments

LLogan M

It sounds like you aren't using bge for the queries 👀

aasanchez75

yes, I want to use OpenAI.

aasanchez75

like in this example from the docs https://docs.llamaindex.ai/en/stable/examples/vector_stores/Elasticsearch_demo.html

aasanchez75

where instead to use SimpleDirectoryReader, I read data from a Elasticsearch index

aasanchez75

but the query is done using OpenAI as far I can understand

LLogan M

You need to use the same embeddings model for queries as you did to create the index

But remember, theres two models -- The LLM and the embedding model

The LLM can change at any time. But the embedding model has to be constant

I'm guessing you want to use openai as the LLM, which is fine, but you need to continue using bge for your embeddings

LLogan M

If you can share your code, I can probably fix it

aasanchez75

Yes, you were right, it is working now!

LLogan M

nice!

aasanchez75

thanks!

Add a reply

Find answers from the community

I have created embeddings using local