Brutex
/
cavis
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Fixing tests 

Signed-off-by: brian <brian@brutex.de>
enhance-build-infrastructure
Brian Rosenberger 2023-05-15 14:24:01 +02:00
parent 8d73a7a410
commit 0bed17c97f
20 changed files with 482 additions and 607 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
cavis-dnn/cavis-dnn-core/src/test/java/org/deeplearning4j/earlystopping/TestEarlyStopping.java
View File

@ -45,6 +45,7 @@ import org.deeplearning4j.nn.conf.NeuralNetConfiguration;
import org.deeplearning4j.nn.conf.layers.*;
import org.deeplearning4j.nn.conf.layers.variational.BernoulliReconstructionDistribution;
import org.deeplearning4j.nn.conf.layers.variational.VariationalAutoencoder;
import org.deeplearning4j.nn.conf.serde.CavisMapper;
import org.deeplearning4j.nn.multilayer.MultiLayerNetwork;
import org.deeplearning4j.nn.weights.WeightInit;
import org.deeplearning4j.optimize.api.BaseTrainingListener;
@ -924,8 +925,8 @@ public class TestEarlyStopping extends BaseDL4JTest {
};
for(EpochTerminationCondition e : etc ){
String s = NeuralNetConfiguration.mapper().writeValueAsString(e);
EpochTerminationCondition c = NeuralNetConfiguration.mapper().readValue(s, EpochTerminationCondition.class);
String s = CavisMapper.getMapper(CavisMapper.Type.JSON).writeValueAsString(e);
EpochTerminationCondition c = CavisMapper.getMapper(CavisMapper.Type.JSON).readValue(s, EpochTerminationCondition.class);
assertEquals(e, c);
}
@ -936,8 +937,8 @@ public class TestEarlyStopping extends BaseDL4JTest {
};
for(IterationTerminationCondition i : itc ){
String s = NeuralNetConfiguration.mapper().writeValueAsString(i);
IterationTerminationCondition c = NeuralNetConfiguration.mapper().readValue(s, IterationTerminationCondition.class);
String s = CavisMapper.getMapper(CavisMapper.Type.JSON).writeValueAsString(i);
IterationTerminationCondition c = CavisMapper.getMapper(CavisMapper.Type.JSON).readValue(s, IterationTerminationCondition.class);
assertEquals(i, c);
}
}

2
cavis-dnn/cavis-dnn-core/src/test/java/org/deeplearning4j/gradientcheck/CNN3DGradientCheckTest.java
View File

@ -112,7 +112,7 @@ public class CNN3DGradientCheckTest extends BaseDL4JTest {
NeuralNetConfiguration conf = NeuralNetConfiguration.builder()
.dataType(DataType.DOUBLE)
.updater(new NoOp()).weightInit(WeightInit.LECUN_NORMAL)
.updater(new NoOp())
.dist(new NormalDistribution(0, 1))
.layer(0, Convolution3D.builder().activation(afn).kernelSize(kernel)
.stride(stride).nIn(convNIn).nOut(convNOut1).hasBias(false)

3
cavis-dnn/cavis-dnn-core/src/test/java/org/deeplearning4j/gradientcheck/LossFunctionGradientCheck.java
View File

@ -32,6 +32,7 @@ import org.deeplearning4j.nn.conf.distribution.UniformDistribution;
import org.deeplearning4j.nn.conf.layers.DenseLayer;
import org.deeplearning4j.nn.conf.layers.LossLayer;
import org.deeplearning4j.nn.conf.layers.OutputLayer;
import org.deeplearning4j.nn.conf.serde.CavisMapper;
import org.deeplearning4j.nn.multilayer.MultiLayerNetwork;
import org.junit.jupiter.api.Test;
import org.nd4j.linalg.activations.Activation;
@ -336,7 +337,7 @@ public class LossFunctionGradientCheck extends BaseDL4JTest {
// to ensure that we carry the parameters through
// the serializer.
try{
ObjectMapper m = NeuralNetConfiguration.mapper();
ObjectMapper m = CavisMapper.getMapper(CavisMapper.Type.JSON);
String s = m.writeValueAsString(lossFunctions[i]);
ILossFunction lf2 = m.readValue(s, lossFunctions[i].getClass());
lossFunctions[i] = lf2;

5
cavis-dnn/cavis-dnn-core/src/test/java/org/deeplearning4j/regressiontest/TestDistributionDeserializer.java
View File

@ -23,6 +23,7 @@ package org.deeplearning4j.regressiontest;
import org.deeplearning4j.BaseDL4JTest;
import org.deeplearning4j.nn.conf.NeuralNetConfiguration;
import org.deeplearning4j.nn.conf.distribution.*;
import org.deeplearning4j.nn.conf.serde.CavisMapper;
import org.junit.jupiter.api.Test;
import com.fasterxml.jackson.databind.ObjectMapper;
@ -38,7 +39,7 @@ public class TestDistributionDeserializer extends BaseDL4JTest {
new Distribution[] {new NormalDistribution(3, 0.5), new UniformDistribution(-2, 1),
new GaussianDistribution(2, 1.0), new BinomialDistribution(10, 0.3)};
ObjectMapper om = NeuralNetConfiguration.mapper();
ObjectMapper om = CavisMapper.getMapper(CavisMapper.Type.JSON);
for (Distribution d : distributions) {
String json = om.writeValueAsString(d);
@ -50,7 +51,7 @@ public class TestDistributionDeserializer extends BaseDL4JTest {
@Test
public void testDistributionDeserializerLegacyFormat() throws Exception {
ObjectMapper om = NeuralNetConfiguration.mapper();
ObjectMapper om = CavisMapper.getMapper(CavisMapper.Type.JSON);
String normalJson = "{\n" + " \"normal\" : {\n" + " \"mean\" : 0.1,\n"
+ " \"std\" : 1.2\n" + " }\n" + " }";

4
cavis-dnn/cavis-dnn-modelimport/src/test/java/org/deeplearning4j/nn/modelimport/keras/configurations/JsonTest.java
View File

@ -41,8 +41,8 @@ public class JsonTest extends BaseDL4JTest {
};
for(InputPreProcessor p : pp ){
String s = NeuralNetConfiguration.mapper().writeValueAsString(p);
InputPreProcessor p2 = NeuralNetConfiguration.mapper().readValue(s, InputPreProcessor.class);
String s = CavisMapper.getMapper(CavisMapper.Type.JSON).writeValueAsString(p);
InputPreProcessor p2 = CavisMapper.getMapper(CavisMapper.Type.JSON).readValue(s, InputPreProcessor.class);
assertEquals(p, p2);
}

9
cavis-dnn/cavis-dnn-nn/src/main/java/org/deeplearning4j/nn/conf/ComputationGraphConfiguration.java
View File

@ -34,6 +34,7 @@ import org.deeplearning4j.nn.conf.layers.recurrent.LastTimeStep;
import org.deeplearning4j.nn.conf.layers.samediff.SameDiffVertex;
import org.deeplearning4j.nn.conf.memory.MemoryReport;
import org.deeplearning4j.nn.conf.memory.NetworkMemoryReport;
import org.deeplearning4j.nn.conf.serde.CavisMapper;
import org.deeplearning4j.nn.conf.serde.JsonMappers;
import org.deeplearning4j.nn.weights.IWeightInit;
import org.deeplearning4j.nn.weights.WeightInit;
@ -110,7 +111,7 @@ public class ComputationGraphConfiguration implements Serializable, Cloneable {
* @return YAML representation of configuration
*/
public String toYaml() {
ObjectMapper mapper = NeuralNetConfiguration.mapperYaml();
ObjectMapper mapper = CavisMapper.getMapper(CavisMapper.Type.YAML);
synchronized (mapper) {
try {
return mapper.writeValueAsString(this);
@ -127,7 +128,7 @@ public class ComputationGraphConfiguration implements Serializable, Cloneable {
* @return {@link ComputationGraphConfiguration}
*/
public static ComputationGraphConfiguration fromYaml(String json) {
ObjectMapper mapper = NeuralNetConfiguration.mapperYaml();
ObjectMapper mapper = CavisMapper.getMapper(CavisMapper.Type.YAML);
try {
return mapper.readValue(json, ComputationGraphConfiguration.class);
} catch (IOException e) {
@ -140,7 +141,7 @@ public class ComputationGraphConfiguration implements Serializable, Cloneable {
*/
public String toJson() {
//As per NeuralNetConfiguration.toJson()
ObjectMapper mapper = NeuralNetConfiguration.mapper();
ObjectMapper mapper =CavisMapper.getMapper(CavisMapper.Type.JSON);
synchronized (mapper) {
//JSON mappers are supposed to be thread safe: however, in practice they seem to miss fields occasionally
//when writeValueAsString is used by multiple threads. This results in invalid JSON. See issue #3243
@ -160,7 +161,7 @@ public class ComputationGraphConfiguration implements Serializable, Cloneable {
*/
public static ComputationGraphConfiguration fromJson(String json) {
//As per NeuralNetConfiguration.fromJson()
ObjectMapper mapper = NeuralNetConfiguration.mapper();
ObjectMapper mapper =CavisMapper.getMapper(CavisMapper.Type.JSON);
ComputationGraphConfiguration conf;
try {
conf = mapper.readValue(json, ComputationGraphConfiguration.class);

123
cavis-dnn/cavis-dnn-nn/src/main/java/org/deeplearning4j/nn/conf/NeuralNetBaseBuilderConfiguration.java
View File

@ -21,6 +21,7 @@
package org.deeplearning4j.nn.conf;
import com.fasterxml.jackson.annotation.JsonIgnore;
import com.fasterxml.jackson.annotation.JsonProperty;
import com.fasterxml.jackson.annotation.JsonTypeInfo;
import com.fasterxml.jackson.databind.JsonNode;
import com.fasterxml.jackson.databind.ObjectMapper;
@ -39,6 +40,7 @@ import org.deeplearning4j.nn.conf.inputs.InputType;
import org.deeplearning4j.nn.conf.layers.BaseLayerConfiguration;
import org.deeplearning4j.nn.conf.layers.ConvolutionLayer;
import org.deeplearning4j.nn.conf.layers.LayerConfiguration;
import org.deeplearning4j.nn.conf.serde.CavisMapper;
import org.deeplearning4j.nn.conf.serde.JsonMappers;
import org.deeplearning4j.nn.conf.stepfunctions.StepFunction;
import org.deeplearning4j.nn.conf.weightnoise.IWeightNoise;
@ -417,113 +419,6 @@ public abstract class NeuralNetBaseBuilderConfiguration implements INeuralNetwor
@Getter @Setter @lombok.Builder.Default private double biasInit = 0.0;
@Getter @Setter @lombok.Builder.Default private double gainInit = 1.0;
/**
* Handle {@link WeightInit} and {@link Distribution} from legacy configs in Json format. Copied
* from handling of {@link Activation} above.
*
* @return True if all is well and layer iteration shall continue. False else-wise.
*/
private static boolean handleLegacyWeightInitFromJson(
String json, LayerConfiguration l, ObjectMapper mapper, JsonNode confs, int layerCount) {
if ((l instanceof BaseLayerConfiguration)
&& ((BaseLayerConfiguration) l).getWeightInit() == null) {
try {
JsonNode jsonNode = mapper.readTree(json);
if (confs == null) {
confs = jsonNode.get("confs");
}
if (confs instanceof ArrayNode) {
ArrayNode layerConfs = (ArrayNode) confs;
JsonNode outputLayerNNCNode = layerConfs.get(layerCount);
if (outputLayerNNCNode == null) {
return false; // Should never happen...
}
JsonNode layerWrapperNode = outputLayerNNCNode.get("layer");
if (layerWrapperNode == null || layerWrapperNode.size() != 1) {
return true;
}
JsonNode layerNode = layerWrapperNode.elements().next();
JsonNode weightInit =
layerNode.get("weightInit"); // Should only have 1 element: "dense", "output", etc
JsonNode distribution = layerNode.get("dist");
Distribution dist = null;
if (distribution != null) {
dist = mapper.treeToValue(distribution, Distribution.class);
}
if (weightInit != null) {
final IWeightInit wi =
WeightInit.valueOf(weightInit.asText()).getWeightInitFunction(dist);
((BaseLayerConfiguration) l).setWeightInit(wi);
}
}
} catch (IOException e) {
log.warn(
"ILayer with null WeightInit detected: " + l.getName() + ", could not parse JSON",
e);
}
}
return true;
}
/**
* Object mapper for serialization of configurations
*
* @return
*/
public static ObjectMapper mapperYaml() {
return JsonMappers.getMapperYaml();
}
/**
* Object mapper for serialization of configurations
*
* @return
*/
public static ObjectMapper mapper() {
return JsonMappers.getMapper();
}
public static NeuralNetBaseBuilderConfiguration fromYaml(String input) {
throw new RuntimeException("Needs fixing - not supported."); // TODO
}
/**
* @return JSON representation of NN configuration
*/
public String toYaml() {
ObjectMapper mapper = NeuralNetBaseBuilderConfiguration.mapperYaml();
synchronized (mapper) {
try {
return mapper.writeValueAsString(this);
} catch (com.fasterxml.jackson.core.JsonProcessingException e) {
throw new RuntimeException(e);
}
}
}
/**
* @return JSON representation of NN configuration
*/
public String toJson() {
ObjectMapper mapper = NeuralNetBaseBuilderConfiguration.mapper();
synchronized (mapper) {
// JSON mappers are supposed to be thread safe: however, in practice they seem to miss fields
// occasionally
// when writeValueAsString is used by multiple threads. This results in invalid JSON. See
// issue #3243
try {
return mapper.writeValueAsString(this);
} catch (com.fasterxml.jackson.core.JsonProcessingException e) {
throw new RuntimeException(e);
}
}
}
@Override
public NeuralNetBaseBuilderConfiguration clone() {
NeuralNetBaseBuilderConfiguration clone;
@ -562,15 +457,7 @@ public abstract class NeuralNetBaseBuilderConfiguration implements INeuralNetwor
List<Object> innerConfigurations$value = new ArrayList<>(); // initialize with an empty list
public B activation(Activation activation) {
this.activation = activation;
return self();
}
@JsonIgnore
public B activation(IActivation activation) {
this.activation = activation;
return self();
}
/**
* Set constraints to be applied to all layers. Default: no constraints.<br>
* Constraints can be used to enforce certain conditions (non-negativity of parameters, max-norm
@ -897,6 +784,7 @@ public abstract class NeuralNetBaseBuilderConfiguration implements INeuralNetwor
*
* @param distribution Distribution to use for weight initialization
*/
@JsonIgnore @Deprecated
public B weightInit(Distribution distribution) {
this.weightInit$value = new WeightInitDistribution(distribution);
this.weightInit$set = true;
@ -909,6 +797,7 @@ public abstract class NeuralNetBaseBuilderConfiguration implements INeuralNetwor
return self();
}
@JsonProperty("weightInit") //this is needed for Jackson < 2.4, otherwise JsonIgnore on the other setters will ignore this also
public B weightInit(IWeightInit iWeightInit) {
this.weightInit$value = iWeightInit;
this.weightInit$set = true;
@ -921,6 +810,7 @@ public abstract class NeuralNetBaseBuilderConfiguration implements INeuralNetwor
* @param distribution
* @return
*/
@JsonIgnore
public B dist(@NonNull Distribution distribution) {
return weightInit(distribution);
}
@ -951,5 +841,6 @@ public abstract class NeuralNetBaseBuilderConfiguration implements INeuralNetwor
innerConfigurations$set = true;
return self();
}
}
}

638
cavis-dnn/cavis-dnn-nn/src/main/java/org/deeplearning4j/nn/conf/NeuralNetConfiguration.java
View File

@ -24,39 +24,24 @@ import com.fasterxml.jackson.annotation.JsonIgnore;
import com.fasterxml.jackson.annotation.JsonIgnoreProperties;
import com.fasterxml.jackson.core.JsonProcessingException;
import com.fasterxml.jackson.databind.*;
import com.fasterxml.jackson.databind.annotation.JsonSerialize;
import com.fasterxml.jackson.databind.exc.InvalidTypeIdException;
import com.fasterxml.jackson.databind.json.JsonMapper;
import com.fasterxml.jackson.databind.node.ArrayNode;
import java.util.*;
import java.util.stream.Collectors;
import lombok.*;
import lombok.experimental.SuperBuilder;
import lombok.extern.jackson.Jacksonized;
import lombok.extern.slf4j.Slf4j;
import net.brutex.ai.dnn.api.IModel;
import org.deeplearning4j.nn.conf.distribution.Distribution;
import org.deeplearning4j.nn.conf.inputs.InputType;
import org.deeplearning4j.nn.conf.layers.*;
import org.deeplearning4j.nn.conf.layers.recurrent.LastTimeStep;
import org.deeplearning4j.nn.conf.memory.LayerMemoryReport;
import org.deeplearning4j.nn.conf.memory.MemoryReport;
import org.deeplearning4j.nn.conf.memory.NetworkMemoryReport;
import org.deeplearning4j.nn.conf.serde.JsonMappers;
import org.deeplearning4j.nn.weights.IWeightInit;
import org.deeplearning4j.nn.weights.WeightInit;
import org.deeplearning4j.nn.conf.serde.CavisMapper;
import org.deeplearning4j.util.OutputLayerUtil;
import org.nd4j.linalg.activations.Activation;
import org.nd4j.linalg.factory.Nd4j;
import org.nd4j.linalg.learning.config.IUpdater;
import org.nd4j.linalg.learning.config.Sgd;
import org.nd4j.linalg.lossfunctions.LossFunctions;
import org.nd4j.linalg.lossfunctions.impl.LossBinaryXENT;
import org.nd4j.linalg.lossfunctions.impl.LossMCXENT;
import org.nd4j.linalg.lossfunctions.impl.LossMSE;
import org.nd4j.linalg.lossfunctions.impl.LossNegativeLogLikelihood;
import java.io.IOException;
import java.util.*;
import java.util.stream.Collectors;
/**
* Deeplearning4j is a domain-specific language to configure deep neural networks, which are made of
@ -64,71 +49,50 @@ import java.util.stream.Collectors;
* and their hyperparameters. Hyperparameters are variables that determine how a neural network
* learns. They include how many times to update the weights of the model, how to initialize those
* weights, which activation function to attach to the nodes, which optimization algorithm to use,
* and how fast the model should learn. This is what one configuration would look like:
* <br/><br/>
*
* NeuralNetConfiguration conf = NeuralNetConfiguration.builder()<br/>
* .weightInit(WeightInit.XAVIER) .activation(Activation.RELU)<br/>
* .optimizationAlgo(OptimizationAlgorithm.STOCHASTIC_GRADIENT_DESCENT)<br/>
* .updater(new Sgd(0.05)) //... other hyperparameters <br/>
* .backprop(true)<br/>
* .build();<br/><br/>
*
* With Deeplearning4j, you add a layer
* by calling layer on the NeuralNetConfiguration.NeuralNetConfigurationBuilder(), specifying its place in the order of
* and how fast the model should learn. This is what one configuration would look like: <br>
* <br>
* NeuralNetConfiguration conf = NeuralNetConfiguration.builder()<br>
* .weightInit(WeightInit.XAVIER) .activation(Activation.RELU)<br>
* .optimizationAlgo(OptimizationAlgorithm.STOCHASTIC_GRADIENT_DESCENT)<br>
* .updater(new Sgd(0.05)) //... other hyperparameters <br>
* .backprop(true)<br>
* .build();<br>
* <br>
* With Deeplearning4j, you add a layer by calling layer on the
* NeuralNetConfiguration.NeuralNetConfigurationBuilder(), specifying its place in the order of
* layers (the zero-indexed layer below is the input layer), the number of input and output nodes,
* nIn and nOut, as well as the type: DenseLayer.<br/><br/>
*
* .layer(0, DenseLayer.builder().nIn(784).nOut(250)<br/>
* .build())<br/><br/>
*
* Once you've configured your net, you train the
* model with model.fit.
* nIn and nOut, as well as the type: DenseLayer.<br>
* <br>
* .layer(0, DenseLayer.builder().nIn(784).nOut(250)<br>
* .build())<br>
* <br>
* Once you've configured your net, you train the model with model.fit.
*/
@Data
@Slf4j
@Jacksonized
@JsonIgnoreProperties(value={"net"}, ignoreUnknown = true)
@EqualsAndHashCode(exclude = {"net"}, callSuper = true)
//@JsonIdentityInfo(generator= ObjectIdGenerators.IntSequenceGenerator.class, property="@id")
@JsonIgnoreProperties(value = {"net"})
@EqualsAndHashCode(callSuper = true)
// @JsonIdentityInfo(generator= ObjectIdGenerators.IntSequenceGenerator.class, property="@id")
//The inner builder, that we can then extend ...
@SuperBuilder //TODO fix access
// The inner builder, that we can then extend ...
@Jacksonized
@SuperBuilder // TODO fix access
public class NeuralNetConfiguration extends NeuralNetBaseBuilderConfiguration {
private IModel net;
private static final int DEFAULT_TBPTT_LENGTH = 20;
private boolean initCalled = false;
@Getter
@Setter
@NonNull
@lombok.Builder.Default
@Deprecated
@Getter @Setter @NonNull @lombok.Builder.Default @Deprecated
protected WorkspaceMode trainingWorkspaceMode = WorkspaceMode.ENABLED;
@Getter
@Setter
@NonNull
@lombok.Builder.Default
@Deprecated
@Getter @Setter @NonNull @lombok.Builder.Default @Deprecated
protected WorkspaceMode inferenceWorkspaceMode = WorkspaceMode.ENABLED;
@Getter
@Setter
@lombok.Builder.Default
protected int iterationCount = 0;
//Counter for the number of epochs completed so far. Used for per-epoch schedules
@Getter
@Setter
@lombok.Builder.Default
protected int epochCount = 0;
@lombok.Builder.Default
protected double dampingFactor = 100;
@Getter @Setter @lombok.Builder.Default protected int iterationCount = 0;
// Counter for the number of epochs completed so far. Used for per-epoch schedules
@Getter @Setter @lombok.Builder.Default protected int epochCount = 0;
@lombok.Builder.Default protected double dampingFactor = 100;
@EqualsAndHashCode.Exclude private IModel net;
private boolean initCalled = false;
// gradient keys used for ensuring order when getting and setting the gradient
@lombok.Builder.Default private LinkedHashSet<String> netWideVariables = new LinkedHashSet<>();
@ -143,22 +107,19 @@ public class NeuralNetConfiguration extends NeuralNetBaseBuilderConfiguration {
*/
@Getter @Setter @Builder.Default private IUpdater updater = new Sgd();
/**
* Sets the cuDNN algo mode for convolutional layers, which impacts performance and memory usage of cuDNN.
* See {@link ConvolutionLayer.AlgoMode} for details. Defaults to "PREFER_FASTEST", but "NO_WORKSPACE" uses less memory.
* <br>
* Note: values set by this method will be applied to all applicable layers in the network, unless a different
* value is explicitly set on a given layer. In other words: values set via this method are used as the default
* value, and can be overridden on a per-layer basis.
* Sets the cuDNN algo mode for convolutional layers, which impacts performance and memory usage
* of cuDNN. See {@link ConvolutionLayer.AlgoMode} for details. Defaults to "PREFER_FASTEST", but
* "NO_WORKSPACE" uses less memory. <br>
* Note: values set by this method will be applied to all applicable layers in the network, unless
* a different value is explicitly set on a given layer. In other words: values set via this
* method are used as the default value, and can be overridden on a per-layer basis.
*
* @param cudnnAlgoMode cuDNN algo mode to use
*/
@Getter
@Setter
@lombok.Builder.Default
@Getter @Setter @lombok.Builder.Default
private ConvolutionLayer.AlgoMode cudnnAlgoMode = ConvolutionLayer.AlgoMode.PREFER_FASTEST;
/**
* Create a neural net configuration from json
*
@ -166,270 +127,23 @@ public class NeuralNetConfiguration extends NeuralNetBaseBuilderConfiguration {
* @return {@link NeuralNetConfiguration}
*/
public static NeuralNetConfiguration fromJson(String json) {
//ObjectMapper mapper = NeuralNetConfiguration.mapper();
JsonMapper mapper = JsonMapper.builder()
.enable(DeserializationFeature.FAIL_ON_UNKNOWN_PROPERTIES)
.build();
ObjectMapper mapper = CavisMapper.getMapper(CavisMapper.Type.JSON);
try {
return mapper.readValue(json, NeuralNetConfiguration.class);
} catch (JsonProcessingException e) {
throw new RuntimeException(e);
}
/*
try {
conf = mapper.readValue(json, NeuralNetConfiguration.class);
} catch (InvalidTypeIdException e) {
if (e.getMessage().contains("@class")) {
try {
//JSON may be legacy (1.0.0-alpha or earlier), attempt to load it using old format
return JsonMappers.getLegacyMapper().readValue(json, NeuralNetConfiguration.class);
} catch (InvalidTypeIdException e2) {
//Check for legacy custom layers: "Could not resolve type id 'CustomLayer' as a subtype of [simple type, class org.deeplearning4j.nn.conf.layers.ILayer]: known type ids = [Bidirectional, CenterLossOutputLayer, CnnLossLayer, ..."
//1.0.0-beta5: dropping support for custom layers defined in pre-1.0.0-beta format. Built-in layers from these formats still work
String msg = e2.getMessage();
if (msg != null && msg.contains("Could not resolve type id")) {
throw new RuntimeException(
"Error deserializing NeuralNetConfiguration - configuration may have a custom " +
"layer, vertex or preprocessor, in pre version 1.0.0-beta JSON format.\nModels in legacy format with custom"
+
" layers should be loaded in 1.0.0-beta to 1.0.0-beta4 and saved again, before loading in the current version of DL4J",
e);
}
throw new RuntimeException(e2);
} catch (IOException e2) {
throw new RuntimeException(e2);
}
}
throw new RuntimeException(e);
} catch (IOException e) {
//Check if this exception came from legacy deserializer...
String msg = e.getMessage();
if (msg != null && msg.contains("legacy")) {
throw new RuntimeException(
"Error deserializing NeuralNetConfiguration - configuration may have a custom " +
"layer, vertex or preprocessor, in pre version 1.0.0-alpha JSON format. These layers can be "
+
"deserialized by first registering them with NeuralNetConfiguration.registerLegacyCustomClassesForJSON(Class...)",
e);
}
throw new RuntimeException(e);
}
//To maintain backward compatibility after loss function refactoring (configs generated with v0.5.0 or earlier)
// Previously: enumeration used for loss functions. Now: use classes
// IN the past, could have only been an OutputLayer or RnnOutputLayer using these enums
int layerCount = 0;
JsonNode confs = null;
for (LayerConfiguration nnc : conf.getFlattenedLayerConfigurations()) {
LayerConfiguration l = nnc;
if (l instanceof BaseOutputLayer && ((BaseOutputLayer) l).getLossFunction() == null) {
//lossFn field null -> may be an old config format, with lossFunction field being for the enum
//if so, try walking the JSON graph to extract out the appropriate enum value
BaseOutputLayer ol = (BaseOutputLayer) l;
try {
JsonNode jsonNode = mapper.readTree(json);
if (confs == null) {
confs = jsonNode.get("confs");
}
if (confs instanceof ArrayNode) {
ArrayNode layerConfs = (ArrayNode) confs;
JsonNode outputLayerNNCNode = layerConfs.get(layerCount);
if (outputLayerNNCNode == null) {
throw new RuntimeException(
"should never happen"); //return conf; //Should never happen...
}
JsonNode outputLayerNode = outputLayerNNCNode.get("layer");
JsonNode lossFunctionNode = null;
if (outputLayerNode.has("output")) {
lossFunctionNode = outputLayerNode.get("output").get("lossFunction");
} else if (outputLayerNode.has("rnnoutput")) {
lossFunctionNode = outputLayerNode.get("rnnoutput").get("lossFunction");
}
if (lossFunctionNode != null) {
String lossFunctionEnumStr = lossFunctionNode.asText();
LossFunctions.LossFunction lossFunction = null;
try {
lossFunction = LossFunctions.LossFunction.valueOf(lossFunctionEnumStr);
} catch (Exception e) {
log.warn(
"OutputLayer with null LossFunction or pre-0.6.0 loss function configuration detected: could not parse JSON",
e);
}
if (lossFunction != null) {
switch (lossFunction) {
case MSE:
ol.setLossFunction(new LossMSE());
break;
case XENT:
ol.setLossFunction(new LossBinaryXENT());
break;
case NEGATIVELOGLIKELIHOOD:
ol.setLossFunction(new LossNegativeLogLikelihood());
break;
case MCXENT:
ol.setLossFunction(new LossMCXENT());
break;
//Remaining: TODO
case SQUARED_LOSS:
case RECONSTRUCTION_CROSSENTROPY:
default:
log.warn(
"OutputLayer with null LossFunction or pre-0.6.0 loss function configuration detected: could not set loss function for {}",
lossFunction);
break;
}
}
}
} else {
log.warn(
"OutputLayer with null LossFunction or pre-0.6.0 loss function configuration detected: could not parse JSON: layer 'confs' field is not an ArrayNode (is: {})",
(confs != null ? confs.getClass() : null));
}
} catch (IOException e) {
log.warn(
"OutputLayer with null LossFunction or pre-0.6.0 loss function configuration detected: could not parse JSON",
e);
break;
}
}
//Also, pre 0.7.2: activation functions were Strings ("activationFunction" field), not classes ("activationFn")
//Try to load the old format if necessary, and create the appropriate IActivation instance
if ((l instanceof BaseLayerConfiguration) && ((BaseLayerConfiguration) l).getActivationFn() == null) {
try {
JsonNode jsonNode = mapper.readTree(json);
if (confs == null) {
confs = jsonNode.get("confs");
}
if (confs instanceof ArrayNode) {
ArrayNode layerConfs = (ArrayNode) confs;
JsonNode outputLayerNNCNode = layerConfs.get(layerCount);
if (outputLayerNNCNode == null) {
throw new RuntimeException(
"Should never happen"); //return conf; //Should never happen...
}
JsonNode layerWrapperNode = outputLayerNNCNode.get("layer");
if (layerWrapperNode == null || layerWrapperNode.size() != 1) {
continue;
}
JsonNode layerNode = layerWrapperNode.elements().next();
JsonNode activationFunction = layerNode.get(
"activationFunction"); //Should only have 1 element: "dense", "output", etc
if (activationFunction != null) {
Activation ia = Activation.fromString(activationFunction.asText());
((BaseLayerConfiguration) l).setActivation(ia.getActivationFunction());
}
}
} catch (IOException e) {
log.warn(
"ILayer with null ActivationFn field or pre-0.7.2 activation function detected: could not parse JSON",
e);
}
}
if (!handleLegacyWeightInitFromJson(json, l, mapper, confs, layerCount)) {
return conf;
}
layerCount++;
}
return conf;
*/
}
/**
* Handle {@link WeightInit} and {@link Distribution} from legacy configs in Json format. Copied
* from handling of {@link Activation} above.
*
* @return True if all is well and layer iteration shall continue. False else-wise.
*/
private static boolean handleLegacyWeightInitFromJson(String json, LayerConfiguration l,
ObjectMapper mapper,
JsonNode confs, int layerCount) {
if ((l instanceof BaseLayerConfiguration) ) { //&& ((BaseLayerConfiguration) l).getWeightInit() == null) {
try {
JsonNode jsonNode = mapper.readTree(json);
if (confs == null) {
confs = jsonNode.get("confs");
}
if (confs instanceof ArrayNode) {
ArrayNode layerConfs = (ArrayNode) confs;
JsonNode outputLayerNNCNode = layerConfs.get(layerCount);
if (outputLayerNNCNode == null) {
return false; //Should never happen...
}
JsonNode layerWrapperNode = outputLayerNNCNode.get("layer");
if (layerWrapperNode == null || layerWrapperNode.size() != 1) {
return true;
}
JsonNode layerNode = layerWrapperNode.elements().next();
JsonNode weightInit = layerNode.get(
"weightInit"); //Should only have 1 element: "dense", "output", etc
JsonNode distribution = layerNode.get("dist");
Distribution dist = null;
if (distribution != null) {
dist = mapper.treeToValue(distribution, Distribution.class);
}
if (weightInit != null) {
final IWeightInit wi = WeightInit.valueOf(weightInit.asText())
.getWeightInitFunction(dist);
((BaseLayerConfiguration) l).setWeightInit(wi);
}
}
} catch (IOException e) {
log.warn(
"ILayer with null WeightInit detected: " + l.getName() + ", could not parse JSON",
e);
}
}
return true;
}
/**
* Object mapper for serialization of configurations
*
* @return
*/
public static ObjectMapper mapperYaml() {
return JsonMappers.getMapperYaml();
}
/**
* Object mapper for serialization of configurations
*
* @return
*/
public static ObjectMapper mapper() {
return JsonMappers.getMapper();
}
public static NeuralNetConfiguration fromYaml(String input) {
throw new RuntimeException("Needs fixing - not supported."); //TODO
throw new RuntimeException("Needs fixing - not supported."); // TODO
}
/**
* @return JSON representation of NN configuration
*/
public String toYaml() {
ObjectMapper mapper = NeuralNetConfiguration.mapperYaml();
ObjectMapper mapper = CavisMapper.getMapper(CavisMapper.Type.YAML);
synchronized (mapper) {
try {
return mapper.writeValueAsString(this);
@ -443,14 +157,12 @@ public class NeuralNetConfiguration extends NeuralNetBaseBuilderConfiguration {
* @return JSON representation of NN configuration
*/
public String toJson() {
JsonMapper mapper = JsonMapper.builder()
.enable(SerializationFeature.INDENT_OUTPUT)
.enable(MapperFeature.SORT_PROPERTIES_ALPHABETICALLY)
.build();
//ObjectMapper mapper = NeuralNetConfiguration.mapper();
ObjectMapper mapper = CavisMapper.getMapper(CavisMapper.Type.JSON);
synchronized (mapper) {
//JSON mappers are supposed to be thread safe: however, in practice they seem to miss fields occasionally
//when writeValueAsString is used by multiple threads. This results in invalid JSON. See issue #3243
// JSON mappers are supposed to be thread safe: however, in practice they seem to miss fields
// occasionally
// when writeValueAsString is used by multiple threads. This results in invalid JSON. See
// issue #3243
try {
return mapper.writeValueAsString(this);
} catch (com.fasterxml.jackson.core.JsonProcessingException e) {
@ -469,7 +181,9 @@ public class NeuralNetConfiguration extends NeuralNetBaseBuilderConfiguration {
public NeuralNetConfiguration clone() {
NeuralNetConfiguration clone;
clone = (NeuralNetConfiguration) super.clone();
if(getStepFunction() != null) { clone.setStepFunction(getStepFunction().clone()); }
if (getStepFunction() != null) {
clone.setStepFunction(getStepFunction().clone());
}
clone.netWideVariables = new LinkedHashSet<>(netWideVariables);
clone.setInnerConfigurations(new ArrayList<>(innerConfigurations));
@ -489,98 +203,106 @@ public class NeuralNetConfiguration extends NeuralNetBaseBuilderConfiguration {
clone.setDataType(this.getDataType());
return clone;
}
/**
*
*/
/** */
@Override
public void init() {
if(initCalled) return;
initCalled=true;
if (initCalled) return;
initCalled = true;
/**
* Run init() for each layer
*/
for( NeuralNetConfiguration nconf : getNetConfigurations() ) {
/** Run init() for each layer */
for (NeuralNetConfiguration nconf : getNetConfigurations()) {
nconf.init();
}
//getNetConfigurations().stream().forEach( conf -> {
// conf.init(); //do not call on self
//}); //call init on all embedded net configurations
// getNetConfigurations().stream().forEach( conf -> {
// conf.init(); //do not call on self
// }); //call init on all embedded net configurations
//TODO do not put inside self to avoid serialization issues
// TODO do not put inside self to avoid serialization issues
// innerConfigurations.add(0, this); //put this configuration at first place
/**
* Inherit network wide configuration setting to those layer configurations
* that do not have an individual setting (nor a default)
* Inherit network wide configuration setting to those layer configurations that do not have an
* individual setting (nor a default)
*/
for(LayerConfiguration lconf : this.getFlattenedLayerConfigurations()) {
for (LayerConfiguration lconf : this.getFlattenedLayerConfigurations()) {
lconf.runInheritance();
}
getLayerConfigurations().stream().forEach( lconf -> lconf.setNetConfiguration(this)); //set this as net config for all layers (defined in here, not stacked
getLayerConfigurations().stream()
.forEach(
lconf ->
lconf.setNetConfiguration(
this)); // set this as net config for all layers (defined in here, not stacked
//Validate BackpropType setting
// Validate BackpropType setting
if ((tbpttBackLength != DEFAULT_TBPTT_LENGTH || tbpttFwdLength != DEFAULT_TBPTT_LENGTH)
&& backpropType != BackpropType.TruncatedBPTT) {
log.warn("Truncated backpropagation through time lengths have been configured with values "
+ tbpttFwdLength
+ " and " + tbpttBackLength + " but backprop type is set to " + backpropType
+ ". TBPTT configuration" +
" settings will only take effect if backprop type is set to BackpropType.TruncatedBPTT");
log.warn(
"Truncated backpropagation through time lengths have been configured with values "
+ tbpttFwdLength
+ " and "
+ tbpttBackLength
+ " but backprop type is set to "
+ backpropType
+ ". TBPTT configuration"
+ " settings will only take effect if backprop type is set to BackpropType.TruncatedBPTT");
}
if (backpropType == BackpropType.TruncatedBPTT && isValidateTbpttConfig()) {
//Check for invalid combination - tbptt plus LastTimeStepLayer or
// Check for invalid combination - tbptt plus LastTimeStepLayer or
for (int i = 0; i < getFlattenedLayerConfigurations().size(); i++) {
LayerConfiguration l = getFlattenedLayerConfigurations().get(i);
if (l instanceof LastTimeStep || l instanceof GlobalPoolingLayer) {
throw new IllegalStateException(
"Invalid network configuration detected: Truncated backpropagation through time (TBPTT)"
+
" cannot be used with layer " + i + " of type " + l.getClass().getName()
+ ": TBPTT is incompatible with this layer type (which is designed " +
"to process entire sequences at once, and does support the type of sequence segments that TPBTT uses).\n"
+
"This check can be disabled using validateTbpttConfig(false) but this is not recommended.");
+ " cannot be used with layer "
+ i
+ " of type "
+ l.getClass().getName()
+ ": TBPTT is incompatible with this layer type (which is designed "
+ "to process entire sequences at once, and does support the type of sequence segments that TPBTT uses).\n"
+ "This check can be disabled using validateTbpttConfig(false) but this is not recommended.");
}
}
}
if (getInputType() == null && inputPreProcessors.get(0) == null) {
//User hasn't set the InputType. Sometimes we can infer it...
// For example, Dense/RNN layers, where preprocessor isn't set -> user is *probably* going to feed in
// User hasn't set the InputType. Sometimes we can infer it...
// For example, Dense/RNN layers, where preprocessor isn't set -> user is *probably* going to
// feed in
// standard feedforward or RNN data
//This isn't the most elegant implementation, but should avoid breaking backward compatibility here
//Can't infer InputType for CNN layers, however (don't know image dimensions/depth)
// This isn't the most elegant implementation, but should avoid breaking backward
// compatibility here
// Can't infer InputType for CNN layers, however (don't know image dimensions/depth)
LayerConfiguration firstLayer = getFlattenedLayerConfigurations().get(0);
if (firstLayer instanceof BaseRecurrentLayer) {
BaseRecurrentLayer brl = (BaseRecurrentLayer) firstLayer;
val nIn = brl.getNIn();
if (nIn > 0) {
setInputType( InputType.recurrent(nIn, brl.getDataFormat()));
setInputType(InputType.recurrent(nIn, brl.getDataFormat()));
}
} else if (firstLayer instanceof DenseLayer || firstLayer instanceof EmbeddingLayer
} else if (firstLayer instanceof DenseLayer
|| firstLayer instanceof EmbeddingLayer
|| firstLayer instanceof OutputLayer) {
//Can't just use "instanceof FeedForwardLayer" here. ConvolutionLayer is also a FeedForwardLayer
// Can't just use "instanceof FeedForwardLayer" here. ConvolutionLayer is also a
// FeedForwardLayer
FeedForwardLayer ffl = (FeedForwardLayer) firstLayer;
val nIn = ffl.getNIn();
if (nIn > 0) {
setInputType( InputType.feedForward(nIn));
setInputType(InputType.feedForward(nIn));
}
}
}
//Add preprocessors and set nIns, if InputType has been set
// Add preprocessors and set nIns, if InputType has been set
// Builder.inputType field can be set in 1 of 4 ways:
// 1. User calls setInputType directly
// 2. Via ConvolutionLayerSetup -> internally calls setInputType(InputType.convolutional(...))
// 3. Via the above code: i.e., assume input is as expected by the RNN or dense layer -> sets the inputType field
if(inputPreProcessors == null) {
// 3. Via the above code: i.e., assume input is as expected by the RNN or dense layer -> sets
// the inputType field
if (inputPreProcessors == null) {
inputPreProcessors = new HashMap<>();
}
if (getInputType() != null) {
@ -588,7 +310,7 @@ public class NeuralNetConfiguration extends NeuralNetBaseBuilderConfiguration {
for (int i = 0; i < getFlattenedLayerConfigurations().size(); i++) {
LayerConfiguration l = getFlattenedLayerConfigurations().get(i);
if (inputPreProcessors.get(i) == null) {
//Don't override preprocessor setting, but set preprocessor if required...
// Don't override preprocessor setting, but set preprocessor if required...
@NonNull
InputPreProcessor inputPreProcessor = l.getPreProcessorForInputType(currentInputType);
if (inputPreProcessor != null) {
@ -602,41 +324,47 @@ public class NeuralNetConfiguration extends NeuralNetBaseBuilderConfiguration {
}
if (i > 0) {
LayerConfiguration layer = getFlattenedLayerConfigurations().get(i - 1);
//convolution 1d is an edge case where it has rnn input type but the filters
//should be the output
// convolution 1d is an edge case where it has rnn input type but the filters
// should be the output
if (layer instanceof Convolution1DLayer) {
if (l instanceof DenseLayer && getInputType() instanceof InputType.InputTypeRecurrent) {
FeedForwardLayer feedForwardLayer = (FeedForwardLayer) l;
if (getInputType() instanceof InputType.InputTypeRecurrent) {
InputType.InputTypeRecurrent recurrent = (InputType.InputTypeRecurrent) getInputType();
InputType.InputTypeRecurrent recurrent =
(InputType.InputTypeRecurrent) getInputType();
feedForwardLayer.setNIn(recurrent.getTimeSeriesLength());
}
} else {
l.setNIn(currentInputType,
isOverrideNinUponBuild()); //Don't override the nIn setting, if it's manually set by the user
l.setNIn(
currentInputType,
isOverrideNinUponBuild()); // Don't override the nIn setting, if it's manually set
// by the user
}
} else {
l.setNIn(currentInputType,
isOverrideNinUponBuild()); //Don't override the nIn setting, if it's manually set by the user
l.setNIn(
currentInputType,
isOverrideNinUponBuild()); // Don't override the nIn setting, if it's manually set
// by the user
}
} else {
l.setNIn(currentInputType,
isOverrideNinUponBuild()); //Don't override the nIn setting, if it's manually set by the user
l.setNIn(
currentInputType,
isOverrideNinUponBuild()); // Don't override the nIn setting, if it's manually set by
// the user
}
currentInputType = l.getOutputType(i, currentInputType);
}
}
Nd4j.getRandom().setSeed(getSeed());
//Validate output layer configuration
// Validate output layer configuration
if (isValidateOutputLayerConfig()) {
//Validate output layer configurations...
// Validate output layer configurations...
for (LayerConfiguration n : getFlattenedLayerConfigurations()) {
OutputLayerUtil.validateOutputLayer(n.getName(), n); //No-op for non output/loss layers
OutputLayerUtil.validateOutputLayer(n.getName(), n); // No-op for non output/loss layers
}
}
}
@ -662,26 +390,28 @@ public class NeuralNetConfiguration extends NeuralNetBaseBuilderConfiguration {
layerName = String.valueOf(i);
}
//Pass input type through preprocessor, if necessary
// Pass input type through preprocessor, if necessary
InputPreProcessor preproc = getInputPreProcess(i);
//TODO memory requirements for preprocessor
// TODO memory requirements for preprocessor
if (preproc != null) {
inputType = preproc.getOutputType(inputType);
}
LayerMemoryReport report = getFlattenedLayerConfigurations().get(i).getMemoryReport(inputType);
LayerMemoryReport report =
getFlattenedLayerConfigurations().get(i).getMemoryReport(inputType);
memoryReportMap.put(layerName, report);
inputType = getFlattenedLayerConfigurations().get(i).getOutputType(i, inputType);
}
return new NetworkMemoryReport(memoryReportMap, NeuralNetConfiguration.class,
"MultiLayerNetwork", inputType);
return new NetworkMemoryReport(
memoryReportMap, NeuralNetConfiguration.class, "MultiLayerNetwork", inputType);
}
/**
* For the given input shape/type for the network, return a list of activation sizes for each
* layer in the network.<br> i.e., list.get(i) is the output activation sizes for layer i
* layer in the network.<br>
* i.e., list.get(i) is the output activation sizes for layer i
*
* @param inputType Input type for the network
* @return A lits of activation types for the network, indexed by layer number
@ -715,38 +445,47 @@ public class NeuralNetConfiguration extends NeuralNetBaseBuilderConfiguration {
public void addNetWideVariable(String variable) {
if (!netWideVariables.contains(variable)) {
netWideVariables.add(variable);
log.trace("Adding neural network wide variable '{}' to the list of variables. New length is {}.", variable, netWideVariables.size());
log.trace(
"Adding neural network wide variable '{}' to the list of variables. New length is {}.",
variable,
netWideVariables.size());
}
log.trace("Skipped adding neural network wide variable '{}' to the list of variables. It was already present. Length remains {}.", variable, netWideVariables.size());
log.trace(
"Skipped adding neural network wide variable '{}' to the list of variables. It was already present. Length remains {}.",
variable,
netWideVariables.size());
}
public void clearNetWideVariable() {
netWideVariables.clear();
log.trace("Adding neural network wide variables have been cleared. New length is {}.", netWideVariables.size());
log.trace(
"Adding neural network wide variables have been cleared. New length is {}.",
netWideVariables.size());
}
/**
* From the list of layers and neural net configurations, only return the Layer Configurations that
* are defined in this neural network (it does not include embedded neural network configuration
* layers)
* From the list of layers and neural net configurations, only return the Layer Configurations
* that are defined in this neural network (it does not include embedded neural network
* configuration layers)
*
* @return list with layer configurations
*/
@JsonIgnore
public List<LayerConfiguration> getLayerConfigurations() {
return innerConfigurations.stream()
.filter(obj -> (obj instanceof LayerConfiguration))
.map( obj -> (LayerConfiguration)obj )
.collect( Collectors.toList());
.map(obj -> (LayerConfiguration) obj)
.collect(Collectors.toList());
}
/**
* From the list of layers and neural net configurations, only return the neural net configurations
* From the list of layers and neural net configurations, only return the neural net
* configurations
*
* @return list with neural net configurations
*/
//@Synchronized("innerConfigurationsLock")
// @Synchronized("innerConfigurationsLock")
@JsonIgnore
public List<NeuralNetConfiguration> getNetConfigurations() {
List<NeuralNetConfiguration> list;
@ -767,35 +506,42 @@ public class NeuralNetConfiguration extends NeuralNetBaseBuilderConfiguration {
* @return list of layer configurations
*/
public List<LayerConfiguration> getFlattenedLayerConfigurations(NeuralNetConfiguration conf) {
List<LayerConfiguration> ret = new ArrayList<>(); //create the final return list
//When properly initialized, _this_ configuration is set first in the list, however we
//can find cases where this is not true, thus the first configuration is another net or layer configuration
//and should not be skipped. In essence, skip first configuration if that is "this".
//TODO: skipping not needed anymore as we removed _this_ from innerConfigurations
List<LayerConfiguration> ret = new ArrayList<>(); // create the final return list
// When properly initialized, _this_ configuration is set first in the list, however we
// can find cases where this is not true, thus the first configuration is another net or layer
// configuration
// and should not be skipped. In essence, skip first configuration if that is "this".
// TODO: skipping not needed anymore as we removed _this_ from innerConfigurations
int iSkip = 0;
if(conf.getInnerConfigurations().size()>0 && conf.getInnerConfigurations().get(0).equals(this)) { iSkip=1;}
conf.getInnerConfigurations().stream().skip(iSkip)
.forEach(obj -> {
//if Layer Config, include in list and inherit parameters from this conf
//else if neural net configuration, call self recursively to resolve layer configurations
if (conf.getInnerConfigurations().size() > 0
&& conf.getInnerConfigurations().get(0).equals(this)) {
iSkip = 1;
}
conf.getInnerConfigurations().stream()
.skip(iSkip)
.forEach(
obj -> {
// if Layer Config, include in list and inherit parameters from this conf
// else if neural net configuration, call self recursively to resolve layer
// configurations
if (obj instanceof LayerConfiguration) {
((LayerConfiguration) obj).setNetConfiguration(conf);
ret.add((LayerConfiguration) obj);
} else if (obj instanceof NeuralNetConfiguration)
ret.addAll(getFlattenedLayerConfigurations(
(NeuralNetConfiguration) obj));
} else if (obj instanceof NeuralNetConfiguration)
ret.addAll(getFlattenedLayerConfigurations((NeuralNetConfiguration) obj));
else {
log.error(
"The list of layers and neural network configurations does contain an object of {}. Element will be ignored.",
obj.getClass().getSimpleName());
"The list of layers and neural network configurations does contain an object of {}. Element will be ignored.",
obj.getClass().getSimpleName());
}
});
return ret;
}
/**
* Sames as {@link #getFlattenedLayerConfigurations(NeuralNetConfiguration)}, but uses this configurations
* list of configurations
* Sames as {@link #getFlattenedLayerConfigurations(NeuralNetConfiguration)}, but uses this
* configurations list of configurations
*
* @return list of layer configurations
*/
@JsonIgnore
@ -805,6 +551,7 @@ public class NeuralNetConfiguration extends NeuralNetBaseBuilderConfiguration {
/**
* Add a new layer to the first position
*
* @param layer configuration
*/
public void setLayer(@NonNull LayerConfiguration layer) {
@ -817,26 +564,28 @@ public class NeuralNetConfiguration extends NeuralNetBaseBuilderConfiguration {
}
/**
* Deprecated, do not use. Workaround for old tests
* and getFlattenedLayerConfigurations().get(0);
* Deprecated, do not use. Workaround for old tests and getFlattenedLayerConfigurations().get(0);
*
* @return
*/
@Deprecated @JsonIgnore
@Deprecated
@JsonIgnore
public LayerConfiguration getFirstLayer() {
log.warn("This getFirstLayer method is an ugly workaround and will be removed.");
return getFlattenedLayerConfigurations().get(0);
}
/*
protected boolean canEqual(final Object other) {
return other instanceof NeuralNetConfiguration;
}
*/
protected boolean canEqual(final Object other) {
return other instanceof NeuralNetConfiguration;
}
public static abstract class NeuralNetConfigurationBuilder<C extends NeuralNetConfiguration,
B extends NeuralNetConfiguration.NeuralNetConfigurationBuilder<C, B>> extends
NeuralNetBaseBuilderConfigurationBuilder<C, B> {
public abstract static class NeuralNetConfigurationBuilder<
C extends NeuralNetConfiguration,
B extends NeuralNetConfiguration.NeuralNetConfigurationBuilder<C, B>>
extends NeuralNetBaseBuilderConfigurationBuilder<C, B> {
public ComputationGraphConfiguration.GraphBuilder graphBuilder() {
return new ComputationGraphConfiguration.GraphBuilder(this);
@ -845,10 +594,9 @@ public class NeuralNetConfiguration extends NeuralNetBaseBuilderConfiguration {
public NeuralNetConfigurationBuilder clone() {
try {
return (NeuralNetConfigurationBuilder) super.clone();
} catch(CloneNotSupportedException ex) {
} catch (CloneNotSupportedException ex) {
throw new RuntimeException(ex);
}
}
}
}

4
cavis-dnn/cavis-dnn-nn/src/main/java/org/deeplearning4j/nn/conf/distribution/Distribution.java
View File

@ -25,8 +25,8 @@ import com.fasterxml.jackson.annotation.JsonTypeInfo;
import java.io.Serializable;
@JsonTypeInfo(use = JsonTypeInfo.Id.CLASS, include = JsonTypeInfo.As.PROPERTY, property = "type",
defaultImpl = LegacyDistributionHelper.class)
@JsonTypeInfo(use = JsonTypeInfo.Id.CLASS, property = "@class")
public abstract class Distribution implements Serializable, Cloneable {
private static final long serialVersionUID = 5401741214954998498L;

15
cavis-dnn/cavis-dnn-nn/src/main/java/org/deeplearning4j/nn/conf/distribution/NormalDistribution.java
View File

@ -20,6 +20,7 @@
package org.deeplearning4j.nn.conf.distribution;
import com.fasterxml.jackson.annotation.JsonTypeInfo;
import lombok.Data;
import lombok.EqualsAndHashCode;
import com.fasterxml.jackson.annotation.JsonCreator;
@ -48,21 +49,7 @@ public class NormalDistribution extends Distribution {
this.std = std;
}
public double getMean() {
return mean;
}
public void setMean(double mean) {
this.mean = mean;
}
public double getStd() {
return std;
}
public void setStd(double std) {
this.std = std;
}
@Override
public int hashCode() {

9
cavis-dnn/cavis-dnn-nn/src/main/java/org/deeplearning4j/nn/conf/memory/MemoryReport.java
View File

@ -24,6 +24,7 @@ import lombok.EqualsAndHashCode;
import lombok.NonNull;
import org.deeplearning4j.nn.conf.CacheMode;
import org.deeplearning4j.nn.conf.NeuralNetConfiguration;
import org.deeplearning4j.nn.conf.serde.CavisMapper;
import org.nd4j.linalg.api.buffer.DataBuffer;
import org.nd4j.linalg.api.buffer.DataType;
import org.nd4j.linalg.api.buffer.util.DataTypeUtil;
@ -151,7 +152,7 @@ public abstract class MemoryReport {
public String toJson() {
try {
return NeuralNetConfiguration.mapper().writeValueAsString(this);
return CavisMapper.getMapper(CavisMapper.Type.JSON).writeValueAsString(this);
} catch (JsonProcessingException e) {
throw new RuntimeException(e);
}
@ -159,7 +160,7 @@ public abstract class MemoryReport {
public String toYaml() {
try {
return NeuralNetConfiguration.mapperYaml().writeValueAsString(this);
return CavisMapper.getMapper(CavisMapper.Type.YAML).writeValueAsString(this);
} catch (JsonProcessingException e) {
throw new RuntimeException(e);
}
@ -167,7 +168,7 @@ public abstract class MemoryReport {
public static MemoryReport fromJson(String json) {
try {
return NeuralNetConfiguration.mapper().readValue(json, MemoryReport.class);
return CavisMapper.getMapper(CavisMapper.Type.JSON).readValue(json, MemoryReport.class);
} catch (IOException e) {
throw new RuntimeException(e);
}
@ -175,7 +176,7 @@ public abstract class MemoryReport {
public static MemoryReport fromYaml(String yaml) {
try {
return NeuralNetConfiguration.mapperYaml().readValue(yaml, MemoryReport.class);
return CavisMapper.getMapper(CavisMapper.Type.JSON).readValue(yaml, MemoryReport.class);
} catch (IOException e) {
throw new RuntimeException(e);
}

2
cavis-dnn/cavis-dnn-nn/src/main/java/org/deeplearning4j/nn/conf/serde/BaseNetConfigDeserializer.java
View File

@ -251,7 +251,7 @@ public abstract class BaseNetConfigDeserializer<T> extends StdDeserializer<T> im
Distribution d = null;
if(w == WeightInit.DISTRIBUTION && on.has("dist")){
String dist = on.get("dist").toString();
d = NeuralNetConfiguration.mapper().readValue(dist, Distribution.class);
d = CavisMapper.getMapper(CavisMapper.Type.JSON).readValue(dist, Distribution.class);
}
IWeightInit iwi = w.getWeightInitFunction(d);
baseLayerConfiguration.setWeightInit(iwi);

60
cavis-dnn/cavis-dnn-nn/src/main/java/org/deeplearning4j/nn/conf/serde/CavisMapper.java Normal file
View File

@ -0,0 +1,60 @@
/*
*
* ******************************************************************************
* *
* * This program and the accompanying materials are made available under the
* * terms of the Apache License, Version 2.0 which is available at
* * https://www.apache.org/licenses/LICENSE-2.0.
* *
* * See the NOTICE file distributed with this work for additional
* * information regarding copyright ownership.
* * Unless required by applicable law or agreed to in writing, software
* * distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
* * WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
* * License for the specific language governing permissions and limitations
* * under the License.
* *
* * SPDX-License-Identifier: Apache-2.0
* *****************************************************************************
*
*/
package org.deeplearning4j.nn.conf.serde;
import com.fasterxml.jackson.databind.DeserializationFeature;
import com.fasterxml.jackson.databind.MapperFeature;
import com.fasterxml.jackson.databind.ObjectMapper;
import com.fasterxml.jackson.databind.SerializationFeature;
import com.fasterxml.jackson.databind.json.JsonMapper;
import com.fasterxml.jackson.dataformat.yaml.YAMLMapper;
import lombok.NonNull;
public class CavisMapper {
public static ObjectMapper getMapper(@NonNull Type type) {
ObjectMapper mapper;
switch (type) {
case JSON:
mapper = JsonMapper.builder()
.enable(SerializationFeature.INDENT_OUTPUT)
.enable(MapperFeature.SORT_PROPERTIES_ALPHABETICALLY)
.enable(DeserializationFeature.FAIL_ON_UNKNOWN_PROPERTIES)
.enable(DeserializationFeature.FAIL_ON_INVALID_SUBTYPE)
.enable(DeserializationFeature.FAIL_ON_IGNORED_PROPERTIES)
.build();
break;
case YAML:
mapper = YAMLMapper.builder().build();
break;
default:
throw new RuntimeException("Mapper type not recognised.");
}
return mapper;
}
public static enum Type {
JSON,
YAML
}
}

2
cavis-dnn/cavis-dnn-nn/src/main/java/org/deeplearning4j/nn/conf/serde/ComputationGraphConfigurationDeserializer.java
View File

@ -96,7 +96,7 @@ public class ComputationGraphConfigurationDeserializer
}
jsonSubString = s.substring((int) charOffsetStart - 1, charOffsetEnd.intValue());
ObjectMapper om = NeuralNetConfiguration.mapper();
ObjectMapper om = CavisMapper.getMapper(CavisMapper.Type.JSON);
JsonNode rootNode = om.readTree(jsonSubString);
ObjectNode verticesNode = (ObjectNode) rootNode.get("vertices");

2
cavis-dnn/cavis-dnn-nn/src/main/java/org/deeplearning4j/nn/conf/serde/NeuralNetConfigurationDeserializer.java
View File

@ -78,7 +78,7 @@ public class NeuralNetConfigurationDeserializer extends BaseNetConfigDeserialize
}
String jsonSubString = s.substring((int) charOffsetStart - 1, (int) charOffsetEnd);
ObjectMapper om = NeuralNetConfiguration.mapper();
ObjectMapper om = CavisMapper.getMapper(CavisMapper.Type.JSON);
JsonNode rootNode = om.readTree(jsonSubString);
ArrayNode confsNode = (ArrayNode)rootNode.get("confs");

9
cavis-dnn/cavis-dnn-nn/src/main/java/org/deeplearning4j/nn/transferlearning/FineTuneConfiguration.java
View File

@ -47,6 +47,7 @@ import org.deeplearning4j.nn.conf.layers.ConvolutionLayer;
import org.deeplearning4j.nn.conf.layers.LayerConfiguration;
import org.deeplearning4j.nn.conf.layers.LayerValidation;
import org.deeplearning4j.nn.conf.layers.SubsamplingLayer;
import org.deeplearning4j.nn.conf.serde.CavisMapper;
import org.deeplearning4j.nn.conf.stepfunctions.StepFunction;
import org.deeplearning4j.nn.conf.weightnoise.IWeightNoise;
import org.deeplearning4j.nn.weights.IWeightInit;
@ -118,7 +119,7 @@ public class FineTuneConfiguration {
public static FineTuneConfiguration fromJson(String json) {
try {
return NeuralNetConfiguration.mapper().readValue(json, FineTuneConfiguration.class);
return CavisMapper.getMapper(CavisMapper.Type.JSON).readValue(json, FineTuneConfiguration.class);
} catch (IOException e) {
throw new RuntimeException(e);
}
@ -126,7 +127,7 @@ public class FineTuneConfiguration {
public static FineTuneConfiguration fromYaml(String yaml) {
try {
return NeuralNetConfiguration.mapperYaml().readValue(yaml, FineTuneConfiguration.class);
return CavisMapper.getMapper(CavisMapper.Type.YAML).readValue(yaml, FineTuneConfiguration.class);
} catch (IOException e) {
throw new RuntimeException(e);
}
@ -322,7 +323,7 @@ public class FineTuneConfiguration {
public String toJson() {
try {
return NeuralNetConfiguration.mapper().writeValueAsString(this);
return CavisMapper.getMapper(CavisMapper.Type.JSON).writeValueAsString(this);
} catch (JsonProcessingException e) {
throw new RuntimeException(e);
}
@ -330,7 +331,7 @@ public class FineTuneConfiguration {
public String toYaml() {
try {
return NeuralNetConfiguration.mapperYaml().writeValueAsString(this);
return CavisMapper.getMapper(CavisMapper.Type.YAML).writeValueAsString(this);
} catch (JsonProcessingException e) {
throw new RuntimeException(e);
}

2
cavis-dnn/cavis-dnn-nn/src/main/java/org/deeplearning4j/nn/weights/IWeightInit.java
View File

@ -26,7 +26,7 @@ import com.fasterxml.jackson.annotation.JsonTypeInfo;
import java.io.Serializable;
@JsonTypeInfo(use = JsonTypeInfo.Id.CLASS, include = JsonTypeInfo.As.PROPERTY, property = "@class")
@JsonTypeInfo(use = JsonTypeInfo.Id.CLASS, property = "@class")
@JsonAutoDetect(fieldVisibility = JsonAutoDetect.Visibility.ANY, getterVisibility = JsonAutoDetect.Visibility.NONE,
setterVisibility = JsonAutoDetect.Visibility.NONE)
public interface IWeightInit extends Serializable {

12
cavis-dnn/cavis-dnn-nn/src/main/java/org/deeplearning4j/nn/weights/WeightInitDistribution.java
View File

@ -20,19 +20,24 @@
package org.deeplearning4j.nn.weights;
import lombok.EqualsAndHashCode;
import com.fasterxml.jackson.annotation.JsonCreator;
import lombok.*;
import org.deeplearning4j.nn.conf.distribution.Distribution;
import org.deeplearning4j.nn.conf.distribution.Distributions;
import org.deeplearning4j.nn.conf.distribution.NormalDistribution;
import org.nd4j.linalg.api.ndarray.INDArray;
import org.nd4j.linalg.api.rng.distribution.impl.OrthogonalDistribution;
import com.fasterxml.jackson.annotation.JsonProperty;
@EqualsAndHashCode
@NoArgsConstructor
public class WeightInitDistribution implements IWeightInit {
private final Distribution distribution;
@Getter @Setter
private Distribution distribution;
public WeightInitDistribution(@JsonProperty("distribution") Distribution distribution) {
public WeightInitDistribution(@NonNull Distribution distribution) {
if(distribution == null) {
// Would fail later below otherwise
throw new IllegalArgumentException("Must set distribution!");
@ -40,6 +45,7 @@ public class WeightInitDistribution implements IWeightInit {
this.distribution = distribution;
}
@Override
public INDArray init(double fanIn, double fanOut, long[] shape, char order, INDArray paramView) {
//org.nd4j.linalg.api.rng.distribution.Distribution not serializable

176
cavis-dnn/cavis-dnn-nn/src/test/java/net/brutex/ai/dnn/serde/NeuralNetConfigurationSerdeTest.java Normal file
View File

@ -0,0 +1,176 @@
/*
*
* ******************************************************************************
* *
* * This program and the accompanying materials are made available under the
* * terms of the Apache License, Version 2.0 which is available at
* * https://www.apache.org/licenses/LICENSE-2.0.
* *
* * See the NOTICE file distributed with this work for additional
* * information regarding copyright ownership.
* * Unless required by applicable law or agreed to in writing, software
* * distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
* * WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
* * License for the specific language governing permissions and limitations
* * under the License.
* *
* * SPDX-License-Identifier: Apache-2.0
* *****************************************************************************
*
*/
package net.brutex.ai.dnn.serde;
import org.deeplearning4j.nn.conf.NeuralNetConfiguration;
import org.deeplearning4j.nn.conf.distribution.NormalDistribution;
import org.deeplearning4j.nn.conf.layers.DenseLayer;
import org.deeplearning4j.nn.conf.layers.OutputLayer;
import org.deeplearning4j.nn.weights.WeightInit;
import org.junit.jupiter.api.Test;
import org.nd4j.linalg.lossfunctions.LossFunctions;
import static org.junit.jupiter.api.Assertions.*;
class NeuralNetConfigurationSerdeTest {
@Test
void toYaml() {
}
@Test
void toJson() {
final var conf = NeuralNetConfiguration.builder()
.weightInit(new NormalDistribution(3, 2))
.layer(DenseLayer.builder().nIn(100).nOut(30).build())
.layer(OutputLayer.builder().lossFunction(LossFunctions.LossFunction.SQUARED_LOSS).build())
.build();
assertEquals(conf.toJson(), NeuralNetConfiguration.fromJson(conf.toJson()).toJson());
}
@Test
void toJson2() {
final var conf = NeuralNetConfiguration.builder()
.weightInit(WeightInit.IDENTITY)
.layer(DenseLayer.builder().nIn(100).nOut(30).build())
.layer(OutputLayer.builder().lossFunction(LossFunctions.LossFunction.SQUARED_LOSS).build())
.build();
assertEquals(conf.toJson(), NeuralNetConfiguration.fromJson(conf.toJson()).toJson());
}
@Test
void fromJson() {
final String json = "{\n" +
" \"activation\" : null,\n" +
" \"seed\" : 12345, \n" +
" \"allParamConstraints\" : [ ],\n" +
" \"backpropType\" : \"Standard\",\n" +
" \"biasConstraints\" : [ ],\n" +
" \"biasInit\" : 0.0,\n" +
" \"biasUpdater\" : null,\n" +
" \"cacheMode\" : \"NONE\",\n" +
" \"constrainWeights\" : [ ],\n" +
" \"convolutionMode\" : \"Truncate\",\n" +
" \"cudnnAlgoMode\" : \"PREFER_FASTEST\",\n" +
" \"dampingFactor\" : 100.0,\n" +
" \"dataType\" : \"FLOAT\",\n" +
" \"epochCount\" : 0,\n" +
" \"gainInit\" : 1.0,\n" +
" \"gradientNormalization\" : \"None\",\n" +
" \"gradientNormalizationThreshold\" : 0.0,\n" +
" \"idropOut\" : null,\n" +
" \"inferenceWorkspaceMode\" : \"ENABLED\",\n" +
" \"initCalled\" : false,\n" +
" \"innerConfigurations\" : [ {\n" +
" \"org.deeplearning4j.nn.conf.layers.DenseLayer\" : {\n" +
" \"activation\" : [ \"org.nd4j.linalg.activations.Activation\", \"IDENTITY\" ],\n" +
" \"allParamConstraints\" : null,\n" +
" \"biasConstraints\" : null,\n" +
" \"biasInit\" : 0.0,\n" +
" \"biasUpdater\" : null,\n" +
" \"constrainWeights\" : [ ],\n" +
" \"constraints\" : null,\n" +
" \"dataType\" : null,\n" +
" \"dropOut\" : null,\n" +
" \"gainInit\" : 0.0,\n" +
" \"gradientNormalization\" : \"None\",\n" +
" \"gradientNormalizationThreshold\" : 1.0,\n" +
" \"hasBias\" : true,\n" +
" \"hasLayerNorm\" : false,\n" +
" \"name\" : null,\n" +
" \"nout\" : 30,\n" +
" \"regularization\" : [ ],\n" +
" \"regularizationBias\" : [ ],\n" +
" \"type\" : \"UNKNOWN\",\n" +
" \"variables\" : [ ],\n" +
" \"weightConstraints\" : null,\n" +
" \"weightInit\" : null,\n" +
" \"weightNoise\" : null\n" +
" }\n" +
" }, {\n" +
" \"org.deeplearning4j.nn.conf.layers.OutputLayer\" : {\n" +
" \"activation\" : [ \"org.nd4j.linalg.activations.Activation\", \"IDENTITY\" ],\n" +
" \"allParamConstraints\" : null,\n" +
" \"biasConstraints\" : null,\n" +
" \"biasInit\" : 0.0,\n" +
" \"biasUpdater\" : null,\n" +
" \"constrainWeights\" : [ ],\n" +
" \"constraints\" : null,\n" +
" \"dataType\" : null,\n" +
" \"dropOut\" : null,\n" +
" \"gainInit\" : 0.0,\n" +
" \"gradientNormalization\" : \"None\",\n" +
" \"gradientNormalizationThreshold\" : 1.0,\n" +
" \"hasBias\" : true,\n" +
" \"lossFunction\" : {\n" +
" \"@class\" : \"org.nd4j.linalg.lossfunctions.impl.LossMSE\"\n" +
" },\n" +
" \"name\" : null,\n" +
" \"nout\" : 0,\n" +
" \"regularization\" : [ ],\n" +
" \"regularizationBias\" : [ ],\n" +
" \"type\" : \"UNKNOWN\",\n" +
" \"variables\" : [ ],\n" +
" \"weightConstraints\" : null,\n" +
" \"weightInit\" : null,\n" +
" \"weightNoise\" : null\n" +
" }\n" +
" } ],\n" +
" \"inputPreProcessors\" : { },\n" +
" \"inputType\" : null,\n" +
" \"iterationCount\" : 0,\n" +
" \"maxNumLineSearchIterations\" : 5,\n" +
" \"miniBatch\" : false,\n" +
" \"minimize\" : true,\n" +
" \"name\" : \"Anonymous INeuralNetworkConfiguration\",\n" +
" \"netWideVariables\" : [ ],\n" +
" \"optimizationAlgo\" : \"STOCHASTIC_GRADIENT_DESCENT\",\n" +
" \"overrideNinUponBuild\" : true,\n" +
" \"regularization\" : [ ],\n" +
" \"regularizationBias\" : [ ],\n" +
" \"stepFunction\" : null,\n" +
" \"tbpttBackLength\" : 20,\n" +
" \"tbpttFwdLength\" : 20,\n" +
" \"trainingWorkspaceMode\" : \"ENABLED\",\n" +
" \"updater\" : {\n" +
" \"@class\" : \"org.nd4j.linalg.learning.config.Sgd\",\n" +
" \"learningRate\" : 0.001\n" +
" },\n" +
" \"validateOutputLayerConfig\" : true,\n" +
" \"validateTbpttConfig\" : true,\n" +
" \"weightInit\" : {\n" +
" \"@class\" : \"org.deeplearning4j.nn.weights.WeightInitIdentity\",\n" +
" \"scale\" : null\n" +
" },\n" +
" \"weightNoise\" : null\n" +
"}";
final var conf = NeuralNetConfiguration.builder()
.weightInit(WeightInit.IDENTITY)
.layer(DenseLayer.builder().nIn(100).nOut(30).build())
.layer(OutputLayer.builder().lossFunction(LossFunctions.LossFunction.SQUARED_LOSS).build())
.seed(12345)
.build();
NeuralNetConfiguration conf2 = NeuralNetConfiguration.fromJson(json);
assertEquals(conf.toJson(),conf2.toJson());
assertEquals(conf, conf2);
}
}

3
cavis-ui/cavis-ui-vertx/src/main/java/org/deeplearning4j/ui/module/train/TrainModule.java
View File

@ -42,6 +42,7 @@ import org.deeplearning4j.nn.conf.NeuralNetConfiguration;
import org.deeplearning4j.nn.conf.graph.GraphVertex;
import org.deeplearning4j.nn.conf.graph.LayerVertex;
import org.deeplearning4j.nn.conf.layers.*;
import org.deeplearning4j.nn.conf.serde.CavisMapper;
import org.deeplearning4j.nn.conf.serde.JsonMappers;
import org.deeplearning4j.ui.VertxUIServer;
import org.deeplearning4j.ui.api.HttpMethod;
@ -909,7 +910,7 @@ public class TrainModule implements UIModule {
} else {
try {
NeuralNetConfiguration layer =
NeuralNetConfiguration.mapper().readValue(config, NeuralNetConfiguration.class);
CavisMapper.getMapper(CavisMapper.Type.JSON).readValue(config, NeuralNetConfiguration.class);
return new Triple<>(null, null, layer);
} catch (Exception e) {
log.error("",e);