package tensorflow.serving

Get desktop application:
View/edit binary Protocol Buffers messages

An asset file or set of sharded files with the same name that will be bound to a tensor at init / session_bundle load time.

• optional TensorBinding tensor_binding = 1

  The tensor to bind the asset filename to.

• string filename = 2

  The filename within the assets directory. Note: does not include the base path or asset directory prefix. Base paths can and will change when models are deployed for serving.

ClassificationSignature specifies a graph that takes an input and returns classes and their scores. WARNING(break-tutorial-inline-code): The following code snippet is in-lined in tutorials, please update tutorial documents accordingly whenever code changes.

Used in: Signature

• optional TensorBinding input = 1

• optional TensorBinding classes = 2

• optional TensorBinding scores = 3

GenericSignature specifies a map from logical name to Tensor name. Typical application of GenericSignature is to use a single GenericSignature that includes all of the Tensor nodes and target names that may be useful at serving, analysis or debugging time. The recommended name for this signature in the ModelManifest is "generic_bindings".

Used in: Signature

• map<string, TensorBinding> map = 1

RegressionSignature specifies a graph that takes an input and returns an output.

Used in: Signature

• optional TensorBinding input = 1

• optional TensorBinding output = 2

A Signature specifies the inputs and outputs of commonly used graphs.

Used in: Signatures

• oneof type

  • RegressionSignature regression_signature = 1

  • ClassificationSignature classification_signature = 2

  • GenericSignature generic_signature = 3

Signatures of model export.

• optional Signature default_signature = 1

  Default signature of the graph. WARNING(break-tutorial-inline-code): The following code snippet is in-lined in tutorials, please update tutorial documents accordingly whenever code changes.

• map<string, Signature> named_signatures = 2

  Named signatures of the graph.

A binding to a tensor including the name and, possibly in the future, type or other metadata. For example, this may specify whether a tensor supports batch vs single inference.

Used in: AssetFile, ClassificationSignature, GenericSignature, RegressionSignature

• string tensor_name = 1

  The name of the tensor to bind to.