Github user viirya commented on the issue: https://github.com/apache/spark/pull/23236 > Compared to the error values from the test failures above, they match up until the 10th batch but then these continue until the 16th where it has a timeout I suspect that might because as the resource usage is heavy, `StreamingLogisticRegressionWithSGD`'s training speed on input batch stream can't always catch up predict batch stream. So the model doesn't reach expected improvement in error yet.
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org