benwtrent commented on issue #13921: URL: https://github.com/apache/lucene/issues/13921#issuecomment-2416554406
@hanqiushi you don't need to do anything special. Since its a format, you only have to query the index. So, once you have indexed a bunch of [KnnFloatVectorField](https://lucene.apache.org/core/9_12_0/core/org/apache/lucene/document/KnnFloatVectorField.html), you can just search them with a [KnnFloatVectorQuery](https://lucene.apache.org/core/9_12_0/core/org/apache/lucene/search/KnnFloatVectorQuery.html). Example: ``` try (IndexReader reader = DirectoryReader.open(dir)) { IndexSearcher searcher = new IndexSearcher(reader); TopDocs topDocs = searcher.search(new KnnFloatVectorQuery( KNN_VECTOR_FIELD, // the field name. queryVector, // the float query vector, dims must be the same as the indexed fields k), // the `k` to search for, could be considered efSearch k); // how many docs to return ``` Some other tips: - Use `confidence_interval=0` in the format, this greatly improves `int4` quality. - Pass `compress=true` as well or you will not gain the memory advantages. SEE: https://lucene.apache.org/core/9_12_0/core/org/apache/lucene/codecs/lucene99/Lucene99HnswScalarQuantizedVectorsFormat.html#%3Cinit%3E(int,int,int,int,boolean,java.lang.Float,java.util.concurrent.ExecutorService) -- 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
