org.tribuo.interop.tensorflow

Class TensorflowUtil

java.lang.Object

org.tribuo.interop.tensorflow.TensorflowUtil

public class TensorflowUtil
extends Object

Helper functions for working with Tensorflow.

Field Summary

Fields

Modifier and Type	Field and Description
static String	ASSIGN_OP
static String	ASSIGN_PLACEHOLDER
static String	DTYPE
static String	PLACEHOLDER
static String	VARIABLE_V2

Constructor Summary

Constructors

Constructor and Description
TensorflowUtil()

Method Summary

Modifier and Type	Method and Description
static void	annotateGraph(org.tensorflow.Graph graph, org.tensorflow.Session session) Annotates a graph with an extra placeholder and assign operation for each VariableV2.
static void	closeTensorList(List<org.tensorflow.Tensor<?>> tensorList) Closes a list of Tensors.
static Object	convertTensorToArray(org.tensorflow.Tensor<?> tensor) Extracts the appropriate type of primitive array from a Tensor.
static Object	convertTensorToScalar(org.tensorflow.Tensor<?> tensor) Converts a Tensor into a scalar object, boxing the primitive types.
static void	deserialise(org.tensorflow.Session session, Map<String,Object> tensorMap) Writes a map containing the name of each Tensorflow VariableV2 and the associated parameter array into the supplied session.
static String	generatePlaceholderName(String variableName) Creates a name for a placeholder based on the supplied variable name.
static Object	newBooleanArray(long[] shape) Creates a new primitive boolean array of up to 8 dimensions, using the supplied shape.
static Object	newByteArray(long[] shape) Creates a new primitive byte array of up to 8 dimensions, using the supplied shape.
static Object	newDoubleArray(long[] shape) Creates a new primitive double array of up to 8 dimensions, using the supplied shape.
static Object	newFloatArray(long[] shape) Creates a new primitive float array of up to 8 dimensions, using the supplied shape.
static Object	newIntArray(long[] shape) Creates a new primitive int array of up to 8 dimensions, using the supplied shape.
static Object	newLongArray(long[] shape) Creates a new primitive long array of up to 8 dimensions, using the supplied shape.
static Map<String,Object>	serialise(org.tensorflow.Graph graph, org.tensorflow.Session session) Extracts a Map containing the name of each Tensorflow VariableV2 and the associated parameter array.

Methods inherited from class java.lang.Object

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

Field Detail

VARIABLE_V2

public static final String VARIABLE_V2

See Also:

Constant Field Values

ASSIGN_OP

public static final String ASSIGN_OP

See Also:

Constant Field Values

ASSIGN_PLACEHOLDER

public static final String ASSIGN_PLACEHOLDER

See Also:

Constant Field Values

PLACEHOLDER

public static final String PLACEHOLDER

See Also:

Constant Field Values

DTYPE

public static final String DTYPE

See Also:

Constant Field Values

Constructor Detail

TensorflowUtil

public TensorflowUtil()

Method Detail

newBooleanArray

public static Object newBooleanArray(long[] shape)

Creates a new primitive boolean array of up to 8 dimensions, using the supplied shape.

Does not check the shape to see if all it's elements are positive.

Parameters:

shape - The shape of array to create.

Returns:

A boolean array.

newByteArray

public static Object newByteArray(long[] shape)

Creates a new primitive byte array of up to 8 dimensions, using the supplied shape.

Does not check the shape to see if all it's elements are positive.

Parameters:

shape - The shape of array to create.

Returns:

A byte array.

newIntArray

public static Object newIntArray(long[] shape)

Creates a new primitive int array of up to 8 dimensions, using the supplied shape.

Does not check the shape to see if all it's elements are positive.

Parameters:

shape - The shape of array to create.

Returns:

A int array.

newLongArray

public static Object newLongArray(long[] shape)

Creates a new primitive long array of up to 8 dimensions, using the supplied shape.

Does not check the shape to see if all it's elements are positive.

Parameters:

shape - The shape of array to create.

Returns:

A long array.

newFloatArray

public static Object newFloatArray(long[] shape)

Creates a new primitive float array of up to 8 dimensions, using the supplied shape.

Does not check the shape to see if all it's elements are positive.

Parameters:

shape - The shape of array to create.

Returns:

A float array.

newDoubleArray

public static Object newDoubleArray(long[] shape)

Creates a new primitive double array of up to 8 dimensions, using the supplied shape.

Does not check the shape to see if all it's elements are positive.

Parameters:

shape - The shape of array to create.

Returns:

A double array.

closeTensorList

public static void closeTensorList(List<org.tensorflow.Tensor<?>> tensorList)

Closes a list of Tensors.

Parameters:

tensorList - The list of tensors to close.

convertTensorToArray

public static Object convertTensorToArray(org.tensorflow.Tensor<?> tensor)

Extracts the appropriate type of primitive array from a Tensor.

Returns an object as the user doesn't know what type is in the Tensor.

Parameters:

tensor - The tensor to read.

Returns:

A primitive array.

convertTensorToScalar

public static Object convertTensorToScalar(org.tensorflow.Tensor<?> tensor)

Converts a Tensor into a scalar object, boxing the primitive types.

Does not close the Tensor.

Parameters:

tensor - The tensor to convert.

Returns:

A boxed scalar.

annotateGraph

public static void annotateGraph(org.tensorflow.Graph graph,
                                 org.tensorflow.Session session)

Annotates a graph with an extra placeholder and assign operation for each VariableV2. This allows the graph to be deserialised using deserialise(Session, Map).

This operation can either be done each time the Graph is loaded before deserialise is called, or once, and the updated graphDef persisted with the Map produced by serialise.

Requires a session to correctly get the output type of a VariableV2. This isn't strictly necessary, but there aren't typed ways to get outputs in the TF version we use.

Parameters:

graph - The graph to annotate.

session - The session to use.

generatePlaceholderName

public static String generatePlaceholderName(String variableName)

Creates a name for a placeholder based on the supplied variable name.

Parameters:

variableName - The variable name to use as a base.

Returns:

A name for the placeholder.

serialise

public static Map<String,Object> serialise(org.tensorflow.Graph graph,
                                           org.tensorflow.Session session)

Extracts a Map containing the name of each Tensorflow VariableV2 and the associated parameter array. This map can then be serialised to disk.

Parameters:

graph - The graph to read operations from.

session - The session to read from.

Returns:

A map containing all variable names and parameter arrays.

deserialise

public static void deserialise(org.tensorflow.Session session,
                               Map<String,Object> tensorMap)

Writes a map containing the name of each Tensorflow VariableV2 and the associated parameter array into the supplied session.

Parameters:

session - The session to write to.

tensorMap - The parameter map to write.