case class TrainingContext(batchIndex: Int, sharedStateSingleton: SharedSingleton[SharedState], schema: StructType, numCols: Int, numInitScoreClasses: Int, trainingParams: BaseTrainParams, networkParams: NetworkParams, columnParams: ColumnParams, datasetParams: String, featureNames: Option[Array[String]], numTasksPerExecutor: Int, validationData: Option[Broadcast[Array[Row]]], serializedReferenceDataset: Option[Array[Byte]], partitionCounts: Option[Array[Long]]) extends Serializable with Product

Object to encapsulate all information about a training session that does not change during execution and can be created on the driver. There is also a reference to the shared state in an executor, which can change over time.

Linear Supertypes
Product, Equals, Serializable, Serializable, AnyRef, Any
Ordering
  1. Alphabetic
  2. By Inheritance
Inherited
  1. TrainingContext
  2. Product
  3. Equals
  4. Serializable
  5. Serializable
  6. AnyRef
  7. Any
  1. Hide All
  2. Show All
Visibility
  1. Public
  2. All

Instance Constructors

  1. new TrainingContext(batchIndex: Int, sharedStateSingleton: SharedSingleton[SharedState], schema: StructType, numCols: Int, numInitScoreClasses: Int, trainingParams: BaseTrainParams, networkParams: NetworkParams, columnParams: ColumnParams, datasetParams: String, featureNames: Option[Array[String]], numTasksPerExecutor: Int, validationData: Option[Broadcast[Array[Row]]], serializedReferenceDataset: Option[Array[Byte]], partitionCounts: Option[Array[Long]])

Value Members

  1. val batchIndex: Int
  2. val columnParams: ColumnParams
  3. val datasetParams: String
  4. val earlyStoppingRound: Int
  5. val featureNames: Option[Array[String]]
  6. val hasGroups: Boolean
  7. val hasGroupsAsInt: Int
  8. val hasInitialScores: Boolean
  9. val hasInitialScoresAsInt: Int
  10. val hasValidationData: Boolean
  11. val hasWeights: Boolean
  12. val hasWeightsAsInt: Int
  13. val improvementTolerance: Double
  14. def incrementArrayProcessedSignal(log: Logger): Int
  15. def incrementDataPrepDoneSignal(log: Logger): Unit
  16. val isBulk: Boolean
  17. val isClassification: Boolean
  18. val isProvideTrainingMetric: Boolean
  19. val isStreaming: Boolean
  20. val microBatchSize: Int
  21. val networkParams: NetworkParams
  22. val numCols: Int
  23. val numInitScoreClasses: Int
  24. val numTasksPerExecutor: Int
  25. val partitionCounts: Option[Array[Long]]
  26. val schema: StructType
  27. val serializedReferenceDataset: Option[Array[Byte]]
  28. def sharedState(): SharedState
  29. val sharedStateSingleton: SharedSingleton[SharedState]
  30. def shouldCreateValidationDataset(): Boolean

    Determines if the current task should calculate the validation Dataset.

    Determines if the current task should calculate the validation Dataset. Only 1 task per executor needs to do it, and first one to call this gets the assignment.

    returns

    True if the current task should create, false otherwise.

  31. implicit def toInt(b: Boolean): Int
  32. val trainingParams: BaseTrainParams
  33. val useSingleDatasetMode: Boolean
  34. val validationData: Option[Broadcast[Array[Row]]]