jbellis commented on issue #12615: URL: https://github.com/apache/lucene/issues/12615#issuecomment-1772704049
Responding top to bottom, > I wonder how much the speed difference is due to (1) Vectors being out of memory (and if they used PQ for diskann, if they did, we should test PQ with HNSW). (2) Not merging to a single segment and searching multiple segments serially. (1) 90% of it is the PQ, yes. I assume that storing the vector inline w/ the graph also helps some but I did not test that separately. You could definitely get a big speed up just using PQ on top of HNSW. (2) Single segment in both cases. (JVector leaves segment management as an exercise for the user.) -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org
