Github user hhbyyh commented on the issue: https://github.com/apache/spark/pull/19565 @akopich I'm actually leaning towards "filter after sample". 1. so we don't need to change `miniBatchFraction` in ` docs.sample(withReplacement = sampleWithReplacement, miniBatchFraction, randomGenerator.nextLong())`. 2. Minor: I'm not sure corpusSize = nonEmptyDoc is always good. I'd prefer to keep docs and corpusSize referring to the entire dataset.
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org