How does the ivfflat index use search_preassigned

[lambda7xx]

Summary

Platform

OS:

Faiss version:

Installed from:

Faiss compilation options:

Running on:

• CPU
• GPU

Interface:

• C++
• Python

Reproduction instructions

[lambda7xx]

my faiss-cpu version is 1.8. my code is below

index.search_preassigned(emb.shape[0], emb, nnn, candidate_cluster, None, topk_distance, topk_indices)

I try to use the search_preassigned API for ivfflat. I want to search the top-nnn vector in the candidate cluster and set the centroid_dis as None(https://faiss.ai/cpp_api/struct/structfaiss_1_1IndexIVFFlat.html). Base on the document(https://faiss.ai/cpp_api/struct/structfaiss_1_1IndexIVFFlat.html), it has the search_preassigned.

but my error is

TypeError: handle_Index.<locals>.replacement_search_preassigned() takes 5 positional arguments but 8 were given
type(self.index):<class 'faiss.swigfaiss_avx2.IndexIVFFlat'>

[ndtreviv]

I recently upgraded FAISS and had the same/similar error:

in ivf_search_preassigned
index_ivf.search_preassigned(
handle_Index.<locals>.replacement_search_preassigned() takes 5 positional arguments but 9 were given

[lambda7xx]

I recently upgraded FAISS and had the same/similar error:

in ivf_search_preassigned
index_ivf.search_preassigned(
handle_Index.<locals>.replacement_search_preassigned() takes 5 positional arguments but 9 were given

same error. They do not support for flat index. You can try Ivf-pq index