org.deeplearning4j.nn.modelimport.keras.layers.recurrent

Class KerasRnnUtils

java.lang.Object

org.deeplearning4j.nn.modelimport.keras.layers.recurrent.KerasRnnUtils

public class KerasRnnUtils
extends Object

Constructor Summary

Constructors

Constructor and Description
KerasRnnUtils()

Method Summary

Modifier and Type	Method and Description
static double	getRecurrentDropout(KerasLayerConfiguration conf, Map<String,Object> layerConfig) Get recurrent weight dropout from Keras layer configuration.
static boolean	getUnrollRecurrentLayer(KerasLayerConfiguration conf, Map<String,Object> layerConfig) Get unroll parameter to decide whether to unroll RNN with BPTT or not.
static boolean	isRnnLayer(KerasLayer kerasLayer) Returns true if the given layer is an KerasLSTM, KerasSimpleRnn, KerasBidirectional

Methods inherited from class java.lang.Object

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

Constructor Detail

KerasRnnUtils

public KerasRnnUtils()

Method Detail

isRnnLayer

public static boolean isRnnLayer(KerasLayer kerasLayer)

Returns true if the given layer is an KerasLSTM, KerasSimpleRnn, KerasBidirectional

Parameters:

kerasLayer - the input layer

Returns:

getUnrollRecurrentLayer

public static boolean getUnrollRecurrentLayer(KerasLayerConfiguration conf,
                                              Map<String,Object> layerConfig)
                                       throws InvalidKerasConfigurationException

Get unroll parameter to decide whether to unroll RNN with BPTT or not.

Parameters:

conf - KerasLayerConfiguration

layerConfig - dictionary containing Keras layer properties

Returns:

boolean unroll parameter

Throws:

InvalidKerasConfigurationException - Invalid Keras configuration

getRecurrentDropout

public static double getRecurrentDropout(KerasLayerConfiguration conf,
                                         Map<String,Object> layerConfig)
                                  throws UnsupportedKerasConfigurationException,
                                         InvalidKerasConfigurationException

Get recurrent weight dropout from Keras layer configuration. Non-zero dropout rates are currently not supported.

Parameters:

conf - KerasLayerConfiguration

layerConfig - dictionary containing Keras layer properties

Returns:

recurrent dropout rate

Throws:

InvalidKerasConfigurationException - Invalid Keras configuration

UnsupportedKerasConfigurationException