[jira] [Created] (SPARK-29811) Missing persist on oldDataset in ml.RandomForestRegressor.train()

["Dong Wang (Jira)" <ji...@apache.org>]

Dong Wang created SPARK-29811:
---------------------------------

             Summary: Missing persist on oldDataset in ml.RandomForestRegressor.train()
                 Key: SPARK-29811
                 URL: https://issues.apache.org/jira/browse/SPARK-29811
             Project: Spark
          Issue Type: Improvement
          Components: ML
    Affects Versions: 2.4.3
            Reporter: Dong Wang


The rdd oldDataset in ml.regression.RandomForestRegressor.train() needs to be persisted, because it used in two actions in RandomForest.run() and oldDataset.first().
{code:scala}
override protected def train(
      dataset: Dataset[_]): RandomForestRegressionModel = instrumented { instr =>
    val categoricalFeatures: Map[Int, Int] =
      MetadataUtils.getCategoricalFeatures(dataset.schema($(featuresCol)))
    val oldDataset: RDD[LabeledPoint] = extractLabeledPoints(dataset) // Needs to persist
    val strategy =
      super.getOldStrategy(categoricalFeatures, numClasses = 0, OldAlgo.Regression, getOldImpurity)

    instr.logPipelineStage(this)
    instr.logDataset(dataset)
    instr.logParams(this, labelCol, featuresCol, predictionCol, impurity, numTrees,
      featureSubsetStrategy, maxDepth, maxBins, maxMemoryInMB, minInfoGain,
      minInstancesPerNode, seed, subsamplingRate, cacheNodeIds, checkpointInterval)
   // First use oldDataset
    val trees = RandomForest
      .run(oldDataset, strategy, getNumTrees, getFeatureSubsetStrategy, getSeed, Some(instr))
      .map(_.asInstanceOf[DecisionTreeRegressionModel])
   // Second use oldDataset
    val numFeatures = oldDataset.first().features.size
    instr.logNamedValue(Instrumentation.loggerTags.numFeatures, numFeatures)
    new RandomForestRegressionModel(uid, trees, numFeatures)
  }
{code}

This issue is reported by our tool CacheCheck, which is used to dynamically detecting persist()/unpersist() api misuses.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org