Use vector bulk scoring in entry-point and filter hnsw search #15500
+109
−53
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
benwtrent
changed the title
| int size = Lucene99HnswVectorsReader.EXHAUSTIVE_BULK_SCORE_ORDS; | ||
| // visit limit is a "best effort" limit given our bulk scoring logic; assert that we are | ||
| // within | ||
| // reasonable bounds |
Contributor
There was a problem hiding this comment.
nit: clean up spotless multi-line comment
| results.collect(eps[0], score); | ||
| } | ||
| } else { | ||
| scorer.bulkScore(eps, scores, eps.length); |
Contributor
There was a problem hiding this comment.
Saying this outloud in case my assumption is wrong. I assume the reason we don't need benchmarking here is that we know from prior work at the leaf level that there's definitely a benefit to bulk scoring eps here instead of doing an early termination check for each entry point.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
We currently use the bulk scoring interface on the lowest HNSW level, but there isn't any technical reason why we cannot use bulk scoring on higher levels to find the entry point. For highly connected graphs with many levels, this could provide a small speed improvement.
Additionally, I noticed that we were not doing bulk scoring when using the Filtered hnsw searcher. Again, we should be using bulk scoring even when using the filter optimized HNSW scorer.