mikemccand commented on issue #13519:
URL: https://github.com/apache/lucene/issues/13519#issuecomment-2194673340
OK I managed to run `knnPerfTest.py` from `luceneutil`, using `mpnet`
vectors (768 dims) and I think I am also seeing horrific performance for `int8`
but OK for `int4` and `int7`:
```
quantizedBits recall latency nDoc fanout maxConn beamWidth
visited index ms
32 0.983 2.55 250000 20 64 250 9294 66542 1.00
post-filter
4 0.645 2.13 250000 20 64 250 13010 79605 1.00
post-filter
7 0.943 1.87 250000 20 64 250 11775 78806 1.00
post-filter
8 0.002 2.81 250000 20 64 250 23879 79177 1.00
post-filter
```
NOTE: this is my first time successfully running `knnPerfTest.py` so it's
entirely possible I messed something up! But given that I'm seeing decent
recall for unquantized (32 bit) and 7 bit, I think the 8 bit result is
believable and horrible.
--
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
