Skip to content

Update HNSW indexes to Lucene 10#3097

Merged
lintool merged 22 commits intolucene10from
lucene10-hnsw
Jan 19, 2026
Merged

Update HNSW indexes to Lucene 10#3097
lintool merged 22 commits intolucene10from
lucene10-hnsw

Conversation

@lintool
Copy link
Member

@lintool lintool commented Jan 18, 2026

Note here that I'd like to merge into the lucene10 branch, not master.

This patch updates hnsw and hnsw-sqv x onnx and cached for both BEIR and MS MARCO: code updated, parameters re-tuned.

@zdann15 can you please take a look also?

@lintool lintool requested review from lilyjge and sahel-sh January 18, 2026 16:31
@lintool
Copy link
Member Author

lintool commented Jan 19, 2026

BTW, int8 was renamed to sqv because ScalarQuantizedVectors is the name of the Lucene class that implements the quantization.

Specifically: https://lucene.apache.org/core/10_3_2/core/org/apache/lucene/codecs/lucene99/Lucene99HnswScalarQuantizedVectorsFormat.html

And according to the doc: "Constructs a format using default graph construction parameters with 7 bit quantization", so int8 is actually a misnomer.

@lilyjge
Copy link
Member

lilyjge commented Jan 19, 2026

LGTM

@lintool lintool merged commit 0055e12 into lucene10 Jan 19, 2026
@lintool lintool deleted the lucene10-hnsw branch January 19, 2026 19:12
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants