[
https://issues.apache.org/jira/browse/SPARK-21972?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16171342#comment-16171342
]
zhengruifeng edited comment on SPARK-21972 at 9/19/17 8:54 AM:
---------------------------------------------------------------
Since persistence handling is very algorithm-dependent, I perfer this solution:
1, add method {{preprocess(dataset: Dataset[_]): DataFrame}} in {{Predictor}},
and call it before {{train()}}, we do casting, columns selection and data
persistence (if necessary) in it.
2, add method {{postprocess}}, and call it after {{train()}}, we unpersist the
intermediate dataframe if needed.
3, for specific purpose, we can override it, like in current PR,
{{GeneralizedLinearRegression}} override {{preprocess}} to add a column
selection and casting for {{offsetCol}}.
I personally think that, to handle persistence, we can use more information:
for examples, for algs containing param {{maxIter}}, if {{maxIter}} is set 0 or
1, the input dataset may be unnecsssary to cache.
was (Author: podongfeng):
Since persistence handling is very algorithm-dependent, I perfer this solution:
1, add method {preprocess(dataset: Dataset[_]): DataFrame} in {Predictor}, and
call it before {train()), we do casting, columns selection and data persistence
(if necessary) in it.
2, add method {postprocess}, and call it after {train()), we unpersist the
intermediate dataframe if needed.
3, for specific purpose, we can override it, like in current PR,
{GeneralizedLinearRegression} override {preprocess} to add a column selection
and casting for {offsetCol}.
I personally think that, to handle persistence, we can use more information:
for examples, for algs containing param {maxIter}, if {maxIter} is set 0 or 1,
the input dataset may be unnecsssary to cache.
> Allow users to control input data persistence in ML Estimators via a
> handlePersistence ml.Param
> -----------------------------------------------------------------------------------------------
>
> Key: SPARK-21972
> URL: https://issues.apache.org/jira/browse/SPARK-21972
> Project: Spark
> Issue Type: Improvement
> Components: ML, MLlib
> Affects Versions: 2.2.0
> Reporter: Siddharth Murching
>
> Several Spark ML algorithms (LogisticRegression, LinearRegression, KMeans,
> etc) call {{cache()}} on uncached input datasets to improve performance.
> Unfortunately, these algorithms a) check input persistence inaccurately
> ([SPARK-18608|https://issues.apache.org/jira/browse/SPARK-18608]) and b)
> check the persistence level of the input dataset but not any of its parents.
> These issues can result in unwanted double-caching of input data & degraded
> performance (see
> [SPARK-21799|https://issues.apache.org/jira/browse/SPARK-21799]).
> This ticket proposes adding a boolean {{handlePersistence}} param
> (org.apache.spark.ml.param) so that users can specify whether an ML algorithm
> should try to cache un-cached input data. {{handlePersistence}} will be
> {{true}} by default, corresponding to existing behavior (always persisting
> uncached input), but users can achieve finer-grained control over input
> persistence by setting {{handlePersistence}} to {{false}}.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
