mikemccand commented on issue #14007: URL: https://github.com/apache/lucene/issues/14007#issuecomment-2494478945
> I'd rather like it to be done on top of the codec API. E.g. computing a good scalar quantization for a given model offline, and then using it in the way in to index vectors directly as byte[]. Lucene's HNSW KNN already supports this -- users can provide either `byte[]` or `float[]` input vectors. So a user could pre-quantize their `float[]` vectors, with full knowledge of the model that produced those vectors and a "good" quantization approach (instead of Lucene's per-segment quantile based estimation), turn those into `byte[]` to send to Lucene. -- 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
