Package org.tribuo.interop.tensorflow

Interface OutputConverter<T extends Output<T>>

Type Parameters:
T - The output type.
All Superinterfaces:
com.oracle.labs.mlrg.olcut.config.Configurable, ProtoSerializable<org.tribuo.interop.tensorflow.protos.OutputConverterProto>, com.oracle.labs.mlrg.olcut.provenance.Provenancable<com.oracle.labs.mlrg.olcut.provenance.ConfiguredObjectProvenance>, Serializable
All Known Implementing Classes:
LabelConverter, MultiLabelConverter, RegressorConverter
public interface OutputConverter<T extends Output<T>> extends com.oracle.labs.mlrg.olcut.config.Configurable, ProtoSerializable<org.tribuo.interop.tensorflow.protos.OutputConverterProto>, com.oracle.labs.mlrg.olcut.provenance.Provenancable<com.oracle.labs.mlrg.olcut.provenance.ConfiguredObjectProvenance>, Serializable
Converts the Output into a Tensor and vice versa.

Also provides the loss function for this output type, along with the function which converts the TF graph output into a well formed output float (e.g., a softmax for classification, a sigmoid for multi-label, or the identity function for regression).

N.B. TensorFlow support is experimental and may change without a major version bump.

  • Method Details

    • loss

      BiFunction<org.tensorflow.op.Ops,com.oracle.labs.mlrg.olcut.util.Pair<org.tensorflow.op.core.Placeholder<? extends org.tensorflow.types.family.TNumber>,org.tensorflow.Operand<org.tensorflow.types.family.TNumber>>,org.tensorflow.Operand<org.tensorflow.types.family.TNumber>> loss()
      The loss function associated with this prediction type.
      Returns:
      The TF loss function.

    • outputTransformFunction

      <U extends org.tensorflow.types.family.TNumber> BiFunction<org.tensorflow.op.Ops,org.tensorflow.Operand<U>,org.tensorflow.op.Op> outputTransformFunction()
      Produces an output transformation function that applies the operation to the graph from the supplied Ops, taking a graph output operation.

      For example this function will apply a softmax for classification, a sigmoid for multi-label, or the identity function for regression.

      Type Parameters:
      U - The type of the graph output.
      Returns:
      A function which applies the appropriate transformation function.

    • convertToPrediction

      Prediction<T> convertToPrediction(org.tensorflow.Tensor tensor, ImmutableOutputInfo<T> outputIDInfo, int numValidFeatures, Example<T> example)
      Converts a Tensor into a Prediction.
      Parameters:
      tensor - The tensor to convert.
      outputIDInfo - The output info to use to identify the outputs.
      numValidFeatures - The number of valid features used by the prediction.
      example - The example to insert into the prediction.
      Returns:
      A prediction object.

    • convertToOutput

      T convertToOutput(org.tensorflow.Tensor tensor, ImmutableOutputInfo<T> outputIDInfo)
      Converts a Tensor into the specified output type.
      Parameters:
      tensor - The tensor to convert.
      outputIDInfo - The output info to use to identify the outputs.
      Returns:
      A output.

    • convertToBatchPrediction

      List<Prediction<T>> convertToBatchPrediction(org.tensorflow.Tensor tensor, ImmutableOutputInfo<T> outputIDInfo, int[] numValidFeatures, List<Example<T>> examples)
      Converts a Tensor containing multiple outputs into a list of Predictions.
      Parameters:
      tensor - The tensor to convert.
      outputIDInfo - The output info to use to identify the outputs.
      numValidFeatures - The number of valid features used by the prediction.
      examples - The example to insert into the prediction.
      Returns:
      A list of predictions.

    • convertToBatchOutput

      List<T> convertToBatchOutput(org.tensorflow.Tensor tensor, ImmutableOutputInfo<T> outputIDInfo)
      Converts a Tensor containing multiple outputs into a list of Outputs.
      Parameters:
      tensor - The tensor to convert.
      outputIDInfo - The output info to use to identify the outputs.
      Returns:
      A list of outputs.

    • convertToTensor

      org.tensorflow.Tensor convertToTensor(T output, ImmutableOutputInfo<T> outputIDInfo)
      Converts an Output into a Tensor representing it's output.
      Parameters:
      output - The output to convert.
      outputIDInfo - The output info to use to identify the outputs.
      Returns:
      A Tensor representing this output.

    • convertToTensor

      org.tensorflow.Tensor convertToTensor(List<Example<T>> examples, ImmutableOutputInfo<T> outputIDInfo)
      Converts a list of Example into a Tensor representing all the outputs in the list. It accepts a list of Example rather than a list of Output for efficiency reasons.
      Parameters:
      examples - The examples to convert.
      outputIDInfo - The output info to use to identify the outputs.
      Returns:
      A Tensor representing all the supplied Outputs.

    • generatesProbabilities

      boolean generatesProbabilities()
      Does this OutputConverter generate probabilities.
      Returns:
      True if it produces a probability distribution in the Prediction.

    • getTypeWitness

      default Class<T> getTypeWitness()
      The type witness used when deserializing the TensorFlow model from a protobuf.

      The default implementation throws UnsupportedOperationException for compatibility with implementations which don't use protobuf serialization. This implementation will be removed in the next major version of Tribuo.

      Returns:
      The output class this object produces.