Github user jkbradley commented on the pull request:
https://github.com/apache/spark/pull/5330#issuecomment-91082537
@MechCoder Apologies for the delay; I've been traveling. Thanks for the
updates! The only issue I see now is with the use of broadcast variables; it's
an issue with understanding how they work exactly. Please read this:
[https://spark.apache.org/docs/latest/programming-guide.html#broadcast-variables]
(Actually, it's worth reading the whole programming guide.)
They should be treated as read-only, so you should:
* maintain baseLearners (and weights) as regular Arrays, not broadcast
variables
* create a new broadcast variable on each iteration for the new tree
* never assign a value to a broadcast variable
Let me know if you have questions about it. (I can also push an update to
the PR if you prefer.)
---
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
