Repository: spark Updated Branches: refs/heads/branch-1.3 62c53be2a -> 81648a7b1
[SPARK-6120] [mllib] Warnings about memory in tree, ensemble model save Issue: When the Python DecisionTree example in the programming guide is run, it runs out of Java Heap Space when using the default memory settings for the spark shell. This prints a warning. CC: mengxr Author: Joseph K. Bradley <[email protected]> Closes #4864 from jkbradley/dt-save-heap and squashes the following commits: 02e8daf [Joseph K. Bradley] fixed based on code review 7ecb1ed [Joseph K. Bradley] Added warnings about memory when calling tree and ensemble model save with too small a Java heap size (cherry picked from commit c2fe3a6ff1a48a9da54d2c2c4d80ecd06cdeebca) Signed-off-by: Xiangrui Meng <[email protected]> Project: http://git-wip-us.apache.org/repos/asf/spark/repo Commit: http://git-wip-us.apache.org/repos/asf/spark/commit/81648a7b Tree: http://git-wip-us.apache.org/repos/asf/spark/tree/81648a7b Diff: http://git-wip-us.apache.org/repos/asf/spark/diff/81648a7b Branch: refs/heads/branch-1.3 Commit: 81648a7b12b5e6e4178d7e8191cca2ec45331580 Parents: 62c53be Author: Joseph K. Bradley <[email protected]> Authored: Mon Mar 2 22:33:51 2015 -0800 Committer: Xiangrui Meng <[email protected]> Committed: Mon Mar 2 22:35:22 2015 -0800 ---------------------------------------------------------------------- .../mllib/tree/model/DecisionTreeModel.scala | 27 ++++++++++++++++++-- .../mllib/tree/model/treeEnsembleModels.scala | 27 ++++++++++++++++++-- 2 files changed, 50 insertions(+), 4 deletions(-) ---------------------------------------------------------------------- http://git-wip-us.apache.org/repos/asf/spark/blob/81648a7b/mllib/src/main/scala/org/apache/spark/mllib/tree/model/DecisionTreeModel.scala ---------------------------------------------------------------------- diff --git a/mllib/src/main/scala/org/apache/spark/mllib/tree/model/DecisionTreeModel.scala b/mllib/src/main/scala/org/apache/spark/mllib/tree/model/DecisionTreeModel.scala index 060fd5b..8a57ebc 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/tree/model/DecisionTreeModel.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/tree/model/DecisionTreeModel.scala @@ -23,7 +23,7 @@ import org.json4s._ import org.json4s.JsonDSL._ import org.json4s.jackson.JsonMethods._ -import org.apache.spark.SparkContext +import org.apache.spark.{Logging, SparkContext} import org.apache.spark.annotation.Experimental import org.apache.spark.api.java.JavaRDD import org.apache.spark.mllib.linalg.Vector @@ -32,6 +32,7 @@ import org.apache.spark.mllib.tree.configuration.Algo._ import org.apache.spark.mllib.util.{Loader, Saveable} import org.apache.spark.rdd.RDD import org.apache.spark.sql.{DataFrame, Row, SQLContext} +import org.apache.spark.util.Utils /** * :: Experimental :: @@ -115,7 +116,7 @@ class DecisionTreeModel(val topNode: Node, val algo: Algo) extends Serializable override protected def formatVersion: String = "1.0" } -object DecisionTreeModel extends Loader[DecisionTreeModel] { +object DecisionTreeModel extends Loader[DecisionTreeModel] with Logging { private[tree] object SaveLoadV1_0 { @@ -187,6 +188,28 @@ object DecisionTreeModel extends Loader[DecisionTreeModel] { val sqlContext = new SQLContext(sc) import sqlContext.implicits._ + // SPARK-6120: We do a hacky check here so users understand why save() is failing + // when they run the ML guide example. + // TODO: Fix this issue for real. + val memThreshold = 768 + if (sc.isLocal) { + val driverMemory = sc.getConf.getOption("spark.driver.memory") + .orElse(Option(System.getenv("SPARK_DRIVER_MEMORY"))) + .map(Utils.memoryStringToMb) + .getOrElse(512) + if (driverMemory <= memThreshold) { + logWarning(s"$thisClassName.save() was called, but it may fail because of too little" + + s" driver memory (${driverMemory}m)." + + s" If failure occurs, try setting driver-memory ${memThreshold}m (or larger).") + } + } else { + if (sc.executorMemory <= memThreshold) { + logWarning(s"$thisClassName.save() was called, but it may fail because of too little" + + s" executor memory (${sc.executorMemory}m)." + + s" If failure occurs try setting executor-memory ${memThreshold}m (or larger).") + } + } + // Create JSON metadata. val metadata = compact(render( ("class" -> thisClassName) ~ ("version" -> thisFormatVersion) ~ http://git-wip-us.apache.org/repos/asf/spark/blob/81648a7b/mllib/src/main/scala/org/apache/spark/mllib/tree/model/treeEnsembleModels.scala ---------------------------------------------------------------------- diff --git a/mllib/src/main/scala/org/apache/spark/mllib/tree/model/treeEnsembleModels.scala b/mllib/src/main/scala/org/apache/spark/mllib/tree/model/treeEnsembleModels.scala index 4897906..30a8f7c 100644 --- a/mllib/src/main/scala/org/apache/spark/mllib/tree/model/treeEnsembleModels.scala +++ b/mllib/src/main/scala/org/apache/spark/mllib/tree/model/treeEnsembleModels.scala @@ -24,7 +24,7 @@ import org.json4s._ import org.json4s.JsonDSL._ import org.json4s.jackson.JsonMethods._ -import org.apache.spark.SparkContext +import org.apache.spark.{Logging, SparkContext} import org.apache.spark.annotation.Experimental import org.apache.spark.api.java.JavaRDD import org.apache.spark.mllib.linalg.Vector @@ -34,6 +34,7 @@ import org.apache.spark.mllib.tree.configuration.EnsembleCombiningStrategy._ import org.apache.spark.mllib.util.{Loader, Saveable} import org.apache.spark.rdd.RDD import org.apache.spark.sql.SQLContext +import org.apache.spark.util.Utils /** * :: Experimental :: @@ -250,7 +251,7 @@ private[tree] sealed class TreeEnsembleModel( def totalNumNodes: Int = trees.map(_.numNodes).sum } -private[tree] object TreeEnsembleModel { +private[tree] object TreeEnsembleModel extends Logging { object SaveLoadV1_0 { @@ -277,6 +278,28 @@ private[tree] object TreeEnsembleModel { val sqlContext = new SQLContext(sc) import sqlContext.implicits._ + // SPARK-6120: We do a hacky check here so users understand why save() is failing + // when they run the ML guide example. + // TODO: Fix this issue for real. + val memThreshold = 768 + if (sc.isLocal) { + val driverMemory = sc.getConf.getOption("spark.driver.memory") + .orElse(Option(System.getenv("SPARK_DRIVER_MEMORY"))) + .map(Utils.memoryStringToMb) + .getOrElse(512) + if (driverMemory <= memThreshold) { + logWarning(s"$className.save() was called, but it may fail because of too little" + + s" driver memory (${driverMemory}m)." + + s" If failure occurs, try setting driver-memory ${memThreshold}m (or larger).") + } + } else { + if (sc.executorMemory <= memThreshold) { + logWarning(s"$className.save() was called, but it may fail because of too little" + + s" executor memory (${sc.executorMemory}m)." + + s" If failure occurs try setting executor-memory ${memThreshold}m (or larger).") + } + } + // Create JSON metadata. implicit val format = DefaultFormats val ensembleMetadata = Metadata(model.algo.toString, model.trees(0).algo.toString, --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
