org.apache.ignite.ml.inference.parser

Class TensorFlowBaseModelParser<I,O>

java.lang.Object

org.apache.ignite.ml.inference.parser.TensorFlowBaseModelParser<I,O>

Type Parameters:

I - Type of model input.

O - Type of model output.

All Implemented Interfaces:

Serializable, ModelParser<I,O,Model<I,O>>

Direct Known Subclasses:

TensorFlowGraphModelParser, TensorFlowSavedModelModelParser

public abstract class TensorFlowBaseModelParser<I,O>
extends Object
implements ModelParser<I,O,Model<I,O>>

Base class for TensorFlow model parsers. Contains the logic that is common for models saved as "SavedModel" and as a simple graph.

See Also:

Serialized Form

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static interface	TensorFlowBaseModelParser.InputTransformer<I> Input transformer that accepts input and transforms it into tensor.
static interface	TensorFlowBaseModelParser.OutputTransformer<O> Output transformer that accepts tensors and transforms them into output.

Constructor Summary

Constructors

Constructor and Description
TensorFlowBaseModelParser()

Method Summary

Modifier and Type	Method and Description
Model<I,O>	parse(byte[] mdl) Accepts serialized model represented by byte array, parses it and returns Model.
abstract org.tensorflow.Session	parseModel(byte[] mdl) Parses model specified in serialized form as byte array.
TensorFlowBaseModelParser<I,O>	withInput(String name, TensorFlowBaseModelParser.InputTransformer<I> transformer) Setter that allows to specify additional input graph node and correspondent transformer that allows to transform input into tensor.
TensorFlowBaseModelParser<I,O>	withOutput(List<String> names, TensorFlowBaseModelParser.OutputTransformer<O> transformer) Setter that allows to specify output graph nodes and correspondent transformer that allow to transform tensors into output.

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

TensorFlowBaseModelParser

public TensorFlowBaseModelParser()

Method Detail

parse

public Model<I,O> parse(byte[] mdl)

Accepts serialized model represented by byte array, parses it and returns Model.

Specified by:

parse in interface ModelParser<I,O,Model<I,O>>

Parameters:

mdl - Serialized model represented by byte array.

Returns:

Inference model.

parseModel

public abstract org.tensorflow.Session parseModel(byte[] mdl)

Parses model specified in serialized form as byte array.

Parameters:

mdl - Inference model in serialized form as byte array.

Returns:

TensorFlow session that encapsulates the TensorFlow graph parsed from serialized model.

withInput

public TensorFlowBaseModelParser<I,O> withInput(String name,
                                                TensorFlowBaseModelParser.InputTransformer<I> transformer)

Setter that allows to specify additional input graph node and correspondent transformer that allows to transform input into tensor.

Parameters:

name - Name of the input graph node.

transformer - Transformer that allows to transform input into tensor.

Returns:

This instance.

withOutput

public TensorFlowBaseModelParser<I,O> withOutput(List<String> names,
                                                 TensorFlowBaseModelParser.OutputTransformer<O> transformer)

Setter that allows to specify output graph nodes and correspondent transformer that allow to transform tensors into output.

Parameters:

names - List of output graph node names.

transformer - Transformer that allow to transform tensors into output.

Returns:

This instance.