srowen commented on issue #20793: [WIP][SPARK-23643] Shrinking the buffer in hashSeed up to size of the seed parameter URL: https://github.com/apache/spark/pull/20793#issuecomment-464757057 @MaxGekk this is looking directionally good; if you have time, would be good to get it in as it unearthed some issues in the tests too. For the many ML tests that fail because a value is not quite in tolerances, there are two solutions: - Just change the expected value - Increase the tolerance Really increasing the tolerance and maybe running tests with more seeds is the principled thing to do. However because we fix a seed, the tolerance is not really about variance due to the random seed state, but the implementation. I can see the argument for just changing the expected value to the new value that is returned; it's also simpler.
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
