org.deeplearning4j.nn.modelimport.keras.utils

Class KerasModelUtils

java.lang.Object

org.deeplearning4j.nn.modelimport.keras.utils.KerasModelUtils

public class KerasModelUtils
extends Object

Constructor Summary

Constructors

Constructor and Description
KerasModelUtils()

Method Summary

Modifier and Type	Method and Description
static Model	copyWeightsToModel(Model model, Map<String,KerasLayer> kerasLayers) Helper function to import weights from nested Map into existing model.
static String	determineKerasBackend(Map<String,Object> modelConfig, KerasModelConfiguration config) Determine Keras backend
static int	determineKerasMajorVersion(Map<String,Object> modelConfig, KerasModelConfiguration config) Determine Keras major version
static void	importWeights(Hdf5Archive weightsArchive, String weightsRoot, Map<String,KerasLayer> layers, int kerasVersion, String backend) Store weights to import with each associated Keras layer.
static Map<String,Object>	parseJsonString(String json) Convenience function for parsing JSON strings.
static Map<String,Object>	parseModelConfig(String modelJson, String modelYaml) Parse Keras model configuration from JSON or YAML string representation
static Map<String,Object>	parseYamlString(String yaml) Convenience function for parsing YAML strings.

Methods inherited from class java.lang.Object

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

Constructor Detail

KerasModelUtils

public KerasModelUtils()

Method Detail

copyWeightsToModel

public static Model copyWeightsToModel(Model model,
                                       Map<String,KerasLayer> kerasLayers)
                                throws InvalidKerasConfigurationException

Helper function to import weights from nested Map into existing model. Depends critically on matched layer and parameter names. In general this seems to be straightforward for most Keras models and layersOrdered, but there may be edge cases.

Parameters:

model - DL4J Model interface

Returns:

DL4J Model interface

Throws:

InvalidKerasConfigurationException - Invalid Keras config

determineKerasMajorVersion

public static int determineKerasMajorVersion(Map<String,Object> modelConfig,
                                             KerasModelConfiguration config)
                                      throws InvalidKerasConfigurationException

Determine Keras major version

Parameters:

modelConfig - parsed model configuration for keras model

config - basic model configuration (KerasModelConfiguration)

Returns:

Major Keras version (1 or 2)

Throws:

InvalidKerasConfigurationException - Invalid Keras config

determineKerasBackend

public static String determineKerasBackend(Map<String,Object> modelConfig,
                                           KerasModelConfiguration config)

Determine Keras backend

Parameters:

modelConfig - parsed model configuration for keras model

config - basic model configuration (KerasModelConfiguration)

Returns:

Keras backend string

Throws:

InvalidKerasConfigurationException - Invalid Keras config

importWeights

public static void importWeights(Hdf5Archive weightsArchive,
                                 String weightsRoot,
                                 Map<String,KerasLayer> layers,
                                 int kerasVersion,
                                 String backend)
                          throws InvalidKerasConfigurationException,
                                 UnsupportedKerasConfigurationException

Store weights to import with each associated Keras layer.

Parameters:

weightsArchive - Hdf5Archive

weightsRoot - root of weights in HDF5 archive

Throws:

InvalidKerasConfigurationException - Invalid Keras configuration

UnsupportedKerasConfigurationException

parseModelConfig

public static Map<String,Object> parseModelConfig(String modelJson,
                                                  String modelYaml)
                                           throws IOException,
                                                  InvalidKerasConfigurationException

Parse Keras model configuration from JSON or YAML string representation

Parameters:

modelJson - JSON string representing model (potentially null)

modelYaml - YAML string representing model (potentially null)

Returns:

Model configuration as Map

Throws:

IOException - IO exception

InvalidKerasConfigurationException - Invalid Keras config

parseJsonString

public static Map<String,Object> parseJsonString(String json)
                                          throws IOException

Convenience function for parsing JSON strings.

Parameters:

json - String containing valid JSON

Returns:

Nested (key,value) map of arbitrary depth

Throws:

IOException - IO exception

parseYamlString

public static Map<String,Object> parseYamlString(String yaml)
                                          throws IOException

Convenience function for parsing YAML strings.

Parameters:

yaml - String containing valid YAML

Returns:

Nested (key,value) map of arbitrary depth

Throws:

IOException - IO exception