jtibshirani commented on pull request #416: URL: https://github.com/apache/lucene/pull/416#issuecomment-953198759
> As can be seen from the comparison, there is very slight change that the hierarchy brings: a small increase in recall by at the expense of lower QPSs It looks like QPS is sometimes worse, but often better (like in all the sift-128-euclidean runs, num_cands >=50). I wonder if the first runs are affected by a lack of warm-up? My original set-up you linked to didn't include a warm-up, and in LUCENE-9937 we found that this can have a big impact on the first runs. -- 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
