[GitHub] [lucene] benwtrent commented on issue #11963: Improve vector quantization API

2022-12-14 Thread GitBox
benwtrent commented on issue #11963: URL: https://github.com/apache/lucene/issues/11963#issuecomment-1351900658 OK, step one of this is done with the `KnnByteVectorQuery`. I am next approaching a new `KnnByteVectorField` and this will cause some refactoring to `LeafReader` or

[GitHub] [lucene] benwtrent commented on issue #11963: Improve vector quantization API

2022-12-02 Thread GitBox
benwtrent commented on issue #11963: URL: https://github.com/apache/lucene/issues/11963#issuecomment-1335239951 @rmuir thank you! I currently have a major part of the refactor already written locally. But, it will be big and will need its own rounds of review to make sure its the way we

[GitHub] [lucene] benwtrent commented on issue #11963: Improve vector quantization API

2022-12-01 Thread GitBox
benwtrent commented on issue #11963: URL: https://github.com/apache/lucene/issues/11963#issuecomment-1334127223 @jpountz > which I don't like much for only two types that need to be supported? I do think we will add more types in the future. Specifically `binary` which can

[GitHub] [lucene] benwtrent commented on issue #11963: Improve vector quantization API

2022-11-22 Thread GitBox
benwtrent commented on issue #11963: URL: https://github.com/apache/lucene/issues/11963#issuecomment-1323680557 This is a naive question, but is there a design principle I am missing that makes `VectorValues` out of line? I don't know how often folks will read the vector values