Github user akopich commented on the issue: https://github.com/apache/spark/pull/19565 I'm saying they are not the same, but for larger datasets this should not matter. There is a change in logic. The hack with `val batchSize = (miniBatchFraction * corpusSize).ceil.toInt` is not used anymore. The function `updateLambda` uses the real number of non-empty docs.
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org