mikemccand commented on PR #12255:
URL: https://github.com/apache/lucene/pull/12255#issuecomment-1540672865
Thanks @jbellis -- I'm not sure this change caused any difference.
Something in the past couple days tweaked the KNN results and this one jumped
out at me as a possibility.
This is the only detail the nightly benchmark produced:
```
Traceback (most recent call last):
File "/l/util.nightly/src/python/nightlyBench.py", line 1818, in <module>
run()
File "/l/util.nightly/src/python/nightlyBench.py", line 701, in run
raise RuntimeError('search result differences: %s' % str(errors))
RuntimeError: search result differences:
["query=KnnFloatVectorQuery:vector[0.028473025,...][100] filter=None sort=None
groupField=None hitCount=100: hit 15 has wrong field/score value ([17135768],
'0.9621086') vs ([26065483], '0.9620853')",
"query=KnnFloatVectorQuery:vector[-0.047548626,...][100] filter=None sort=None
groupField=None hitCount=100: hit 0 has wrong field/score value ([20712471],
'0.8335463') vs ([15605918], '0.8440397')",
"query=KnnFloatVectorQuery:vector[0.02625591,...][100] filter=None sort=None
groupField=None hitCount=100: hit 7 has wrong field/score value ([23761647],
'0.8285247') vs ([25459412], '0.8309758')"]
```
Unfortunately it is not so simple to reproduce these nightly benchmarks.
Note that they only check for exact hit/score differences and not any
precision/recall tradeoff.
--
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]
