[
https://issues.apache.org/jira/browse/SPARK-21476?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16447545#comment-16447545
]
Tamilselvan Veeramani commented on SPARK-21476:
-----------------------------------------------
I am facing the same issue - high task deserialization with RD model when
calling model.transform(dataset).select("probability"); since only
predict(features.asInstanceOf[Vector]) implemented in transformImpl method of
RandomForestClassificationModel and still other methods
predictRaw(features.asInstanceOf[FeaturesType]) &
predictProbability(features.asInstanceOf[FeaturesType]) are not implemented and
hence Spark Structured Streaming with RF model continue to be big challenge
with high task deserialization time. any plan to implement these method near
soon ? thanks
> RandomForest classification model not using broadcast in transform
> ------------------------------------------------------------------
>
> Key: SPARK-21476
> URL: https://issues.apache.org/jira/browse/SPARK-21476
> Project: Spark
> Issue Type: Improvement
> Components: ML
> Affects Versions: 2.2.0
> Reporter: Saurabh Agrawal
> Priority: Minor
>
> I notice significant task deserialization latency while running prediction
> with pipelines using RandomForestClassificationModel. While digging into the
> source, found that the transform method in RandomForestClassificationModel
> binds to its parent ProbabilisticClassificationModel and the only concrete
> definition that RandomForestClassificationModel provides and which is
> actually used in transform is that of predictRaw. Broadcasting is not being
> used in predictRaw.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
