Class TrainTest.TensorflowOptions

java.lang.Object
org.tribuo.interop.tensorflow.TrainTest.TensorflowOptions
All Implemented Interfaces:
com.oracle.labs.mlrg.olcut.config.Options
Enclosing class:
TrainTest

public static class TrainTest.TensorflowOptions extends Object implements com.oracle.labs.mlrg.olcut.config.Options
Options for training a model in TensorFlow.
  • Field Details

    • outputPath

      @Option(charName='f', longName="model-output-path", usage="Path to serialize model to.") public Path outputPath
      Path to serialize model to.
    • saveToProto

      @Option(longName="model-save-to-proto", usage="Save the Tribuo model out as a protobuf.") public boolean saveToProto
      Save the Tribuo model out as a protobuf.
    • trainingPath

      @Option(charName='u', longName="training-file", usage="Path to the libsvm format training file.") public Path trainingPath
      Path to the libsvm format training file.
    • testingPath

      @Option(charName='v', longName="testing-file", usage="Path to the libsvm format testing file.") public Path testingPath
      Path to the libsvm format testing file.
    • outputName

      @Option(charName='l', longName="output-name", usage="Name of the output operation.") public String outputName
      Name of the output operation.
    • gradientParamNames

      @Option(longName="optimizer-param-names", usage="Gradient optimizer param names, see org.tribuo.interop.tensorflow.GradientOptimiser.") public List<String> gradientParamNames
      Gradient optimizer param names, see org.tribuo.interop.tensorflow.GradientOptimiser.
    • gradientParamValues

      @Option(longName="optimizer-param-values", usage="Gradient optimizer param values, see org.tribuo.interop.tensorflow.GradientOptimiser.") public List<Float> gradientParamValues
      Gradient optimizer param values, see org.tribuo.interop.tensorflow.GradientOptimiser.
    • optimiser

      @Option(charName='g', longName="gradient-optimizer", usage="The gradient optimizer to use.") public GradientOptimiser optimiser
      The gradient optimizer to use.
    • testBatchSize

      @Option(longName="test-batch-size", usage="Test time minibatch size.") public int testBatchSize
      Test time minibatch size.
    • batchSize

      @Option(charName='b', longName="batch-size", usage="Minibatch size.") public int batchSize
      Minibatch size.
    • epochs

      @Option(charName='e', longName="num-epochs", usage="Number of gradient descent epochs.") public int epochs
      Number of gradient descent epochs.
    • loggingInterval

      @Option(longName="logging-interval", usage="Interval between logging the loss.") public int loggingInterval
      Interval between logging the loss.
    • inputName

      @Option(charName='n', longName="input-name", usage="Name of the input placeholder.") public String inputName
      Name of the input placeholder.
    • imageFormat

      @Option(longName="image-format", usage="Image format, in [W,H,C]. Defaults to MNIST.") public String imageFormat
      Image format, in [W,H,C]. Defaults to MNIST.
    • inputType

      @Option(charName='t', longName="input-type", usage="Input type.") public TrainTest.InputType inputType
      Input type.
    • protobufPath

      @Option(charName='m', longName="model-protobuf", usage="Path to the protobuf containing the network description.") public Path protobufPath
      Path to the protobuf containing the network description.
    • checkpointPath

      @Option(charName='p', longName="checkpoint-dir", usage="Path to the checkpoint base directory.") public Path checkpointPath
      Path to the checkpoint base directory.
  • Constructor Details

    • TensorflowOptions

      public TensorflowOptions()
  • Method Details

    • getOptionsDescription

      public String getOptionsDescription()
      Specified by:
      getOptionsDescription in interface com.oracle.labs.mlrg.olcut.config.Options
    • getGradientParams

      public Map<String,Float> getGradientParams()
      Zips the gradient parameter names and values.
      Returns:
      The gradient parameter map.