Question about usage of embed instead of query_embed in query method in QdrantFastembedMixin

[martinkozle]

qdrant-client/qdrant_client/qdrant_fastembed.py

Line 698 in e2ad066

sparse_vector = list(sparse_embedding_model_inst.embed(documents=query_text))[0]

Hello, I have a question.

For the Fastembed mixin, why is embed used for sparse embeddings instead of query_embed? From what I read, for bm25 specifically at least, is that it doesn't use term frequency and length, but instead sets a uniform 1.0 for each token. What is the effect of that difference here? Is 'query_embed' for sparse embeddings not meant to be used with Qdrant?