[
https://issues.apache.org/jira/browse/SPARK-11898?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Sean Owen resolved SPARK-11898.
-------------------------------
Resolution: Fixed
Fix Version/s: 1.7.0
Issue resolved by pull request 9878
[https://github.com/apache/spark/pull/9878]
> Use broadcast for the global tables in Word2Vec
> -----------------------------------------------
>
> Key: SPARK-11898
> URL: https://issues.apache.org/jira/browse/SPARK-11898
> Project: Spark
> Issue Type: Improvement
> Components: MLlib
> Affects Versions: 1.5.2
> Reporter: yuhao yang
> Fix For: 1.7.0
>
>
> syn0Global and sync1Global in word2vec are quite large objects with size
> (vocab * vectorSize * 8), yet they are passed to worker using basic task
> serialization.
> Use broadcast can greatly improve the performance. My benchmark shows that,
> for 1M vocabulary and default vectorSize 100, changing to broadcast can help,
> 1. decrease the worker memory consumption by 45%.
> 2. decrease running time by 40%.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
