Multi vector search in qdrant

Short Course Q&A Multi-vector Image Retrieval
1

Hi,

i was curious how is search being done in qdrant with this code

all_results = {}
    for vector_name in vector_names:
results = client.query_points(
            collection_name=collection_name,
            query=query_embedding,
            using=vector_name,
            search_params=qdrant_models.SearchParams(
                quantization=qdrant_models.QuantizationSearchParams(
                    rescore=False,
                )
            ),
            limit=limit,
        )
all_results[vector_name] = results.points

Is it performing brute search? Note this in in lesson 3 helper file and muvera is in the next lesson.

1 Like
2

Thank you @eng_ai for taking the course and posting this question. Let me check it with the team to come back with insights!

3

Looks to me like, if the indentation you have posted is correct, then “results = …” is outside of the for-loop, so it’s only going to run once.

And nothing inside the for-loop is saving the “results” value, and “vector_name” is always going to be the result of the last for-loop iteration.

So that’s rather confusing code.

4

Just got a replied and indeed, it is performing brute search.

Greetings!

1 Like
6

Is the for-loop redundant, or is it just badly-formatted in this topic?