benwtrent commented on PR #15140: URL: https://github.com/apache/lucene/pull/15140#issuecomment-3249748627
With oversampling (3x - 5x) and quantized scoring (so the dominating cost of floating point ops goes away), I have seen other administrivia of HNSW searching be more and more the cause of latency. It would be interesting to see the latency for `k==100 & oversampling=3x` with binary quantization. Like @msokolov I wouldn't expect anything on regular float32 (though, you never know until you benchmark it!) -- 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: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
