Github user mpjlu commented on the issue: https://github.com/apache/spark/pull/17919 Hi @auskalia , you are right. repartition can improve the performance of recommendForAll. In my experiment for PR 17742, I have 120 cores, I use 20 partition for userFeatures, and itemFeatures. I also consider to provide interface to user to have a chance to do re-partition. Since you can set the partition number when train the model, I did not do that.
--- If your project is set up for it, you can reply to this email and have your reply appear on GitHub as well. If your project does not have this feature enabled and wishes so, or if the feature is enabled but not working, please contact infrastructure at infrastruct...@apache.org or file a JIRA ticket with INFRA. --- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org