[SYSTEMDS-869] Error converting Matrix to Spark DataFrame with MLContext After Subsequent Executions

• Type: Bug
• Status: Closed
• Priority: Blocker
• Resolution: Fixed
• Fix Version/s: SystemML 0.11
• Component/s: APIs
• Assignee: Matthias Boehm
• Reporter: Mike Dusenberry
• Created: 25/Aug/16 21:07
• Updated: 20/Dec/16 20:32
• Resolved: 29/Sep/16 18:56

Description

Running the LeNet deep learning example notebook with the new MLContext API in Python results in the below error when converting the resulting Matrix to a Spark DataFrame via the toDF() call. This only occurs with the large LeNet example, and not for the similar "Softmax Classifier" example that has a smaller model.

Py4JJavaError: An error occurred while calling o34.asDataFrame.
: org.apache.hadoop.mapred.InvalidInputException: Input path does not exist: file:/Users/mwdusenb/Documents/Code/systemML/deep_learning/examples/scratch_space/_p85157_9.31.116.142/_t0/temp816_133
    at org.apache.hadoop.mapred.FileInputFormat.listStatus(FileInputFormat.java:251)
    at org.apache.hadoop.mapred.SequenceFileInputFormat.listStatus(SequenceFileInputFormat.java:45)
    at org.apache.hadoop.mapred.FileInputFormat.getSplits(FileInputFormat.java:270)
    at org.apache.spark.rdd.HadoopRDD.getPartitions(HadoopRDD.scala:199)
    at org.apache.spark.rdd.RDD$$anonfun$partitions$2.apply(RDD.scala:239)
    at org.apache.spark.rdd.RDD$$anonfun$partitions$2.apply(RDD.scala:237)
    at scala.Option.getOrElse(Option.scala:120)
    at org.apache.spark.rdd.RDD.partitions(RDD.scala:237)
    at org.apache.spark.rdd.MapPartitionsRDD.getPartitions(MapPartitionsRDD.scala:35)
    at org.apache.spark.rdd.RDD$$anonfun$partitions$2.apply(RDD.scala:239)
    at org.apache.spark.rdd.RDD$$anonfun$partitions$2.apply(RDD.scala:237)
    at scala.Option.getOrElse(Option.scala:120)
    at org.apache.spark.rdd.RDD.partitions(RDD.scala:237)
    at org.apache.spark.rdd.MapPartitionsRDD.getPartitions(MapPartitionsRDD.scala:35)
    at org.apache.spark.rdd.RDD$$anonfun$partitions$2.apply(RDD.scala:239)
    at org.apache.spark.rdd.RDD$$anonfun$partitions$2.apply(RDD.scala:237)
    at scala.Option.getOrElse(Option.scala:120)
    at org.apache.spark.rdd.RDD.partitions(RDD.scala:237)
    at org.apache.spark.Partitioner$.defaultPartitioner(Partitioner.scala:65)
    at org.apache.spark.rdd.PairRDDFunctions$$anonfun$groupByKey$3.apply(PairRDDFunctions.scala:642)
    at org.apache.spark.rdd.PairRDDFunctions$$anonfun$groupByKey$3.apply(PairRDDFunctions.scala:642)
    at org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:150)
    at org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:111)
    at org.apache.spark.rdd.RDD.withScope(RDD.scala:316)
    at org.apache.spark.rdd.PairRDDFunctions.groupByKey(PairRDDFunctions.scala:641)
    at org.apache.spark.api.java.JavaPairRDD.groupByKey(JavaPairRDD.scala:538)
    at org.apache.sysml.runtime.instructions.spark.utils.RDDConverterUtilsExt.binaryBlockToDataFrame(RDDConverterUtilsExt.java:502)
    at org.apache.sysml.api.mlcontext.MLContextConversionUtil.matrixObjectToDataFrame(MLContextConversionUtil.java:762)
    at org.apache.sysml.api.mlcontext.Matrix.asDataFrame(Matrix.java:111)
    at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
    at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
    at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
    at java.lang.reflect.Method.invoke(Method.java:497)
    at py4j.reflection.MethodInvoker.invoke(MethodInvoker.java:231)
    at py4j.reflection.ReflectionEngine.invoke(ReflectionEngine.java:381)
    at py4j.Gateway.invoke(Gateway.java:259)
    at py4j.commands.AbstractCommand.invokeMethod(AbstractCommand.java:133)
    at py4j.commands.CallCommand.execute(CallCommand.java:79)
    at py4j.GatewayConnection.run(GatewayConnection.java:209)
    at java.lang.Thread.run(Thread.java:745)

To setup, I used the instructions here , running the Example - MNIST LeNet.ipynb notebook. Additionally, to speed up the actual training time, I modified line 84 & 85 of mnist_lenet.dml to set the epochs = 1, and iters = 1.

Issue Links

• relates to: SYSTEMDS-914 MLContext Performance Improvements