hnswlib
hnswlib copied to clipboard
Header-only C++/python library for fast approximate nearest neighbors
I used to develop everything on my local machine, but is moving to gcs now. I am seeing errors loading model index from gcs. Wondering is there support index loading...
Hi! Thanks for this wonderful library! Every empty Index with max_elements of 100 is taking 2.5 MB of RAM, which is quite expensive overhead for index object without data. Initially...
is it possible to unmark_deleted?
@yurymalkov : I have multiple indexers producing many sharded hnsw indexes , but servers perhaps can handle multiple such shards . I read in the paper that a distributed implementation...
Collect_metrics@searchBaseLayer : We should not increment metric_distance_computation with the size of neighbors, since there may be some neighbors already visited.
Hi, I'm using the library to perform k-NN search. However, I have to regularly update the associated embeddings. Is this possible? What's the optimal way to do it? I'm using...
Hi, I tried to create an index for around 60k embeddings of size 128 in a highend machine with 36 cores (with parameters ef_construction=300, M=32, ef=10). The index file generated...
I have an index containing 8 million items. Retrieval works fine for 1 million query with k=60. However, after I deleted 18k items using mark_deleted(). There were 8 queries giving...
Proposed signature is `tableint replacePoint(const void *data_point, labeltype oldLabel, labeltype newLabel)`. This will allow to change the label without deleting one node and creating another one.