Is CSVSequenceRecordReader creating compatible dataset for training LSTM network? - deeplearning4j

Question

I want to train a simple LSTM network but I got the exception
java.lang.IllegalStateException: C (result) array is not F order or is a view. Nd4j.gemm requires the result array to be F order and not a view. C (result) array: [Rank: 2,Offset: 0 Order: f Shape: [10,1], stride: [1,10]]
I'm training a simple NN with a single LSTM cell and a single output cell for regression.
I created a training dataset of just 10 samples with variable sequence length (from 5 to 10) in csv files, each sample consists of just one value for the input and one value for the output.
I created a SequenceRecordReaderDataSetIterator from a CSVSequenceRecordReader.
When I train my network the code throws the exception.
I tried generating random dataset coding the dataset iterator directly with 'f shape' INDarray and the code runs without error.
So the problem seems to be the shape of tensors created by CSVSequenceRecordReader.
Does anyone have this problems?
SingleFileTimeSeriesDataReader.java
package org.mmarini.lstmtest;
import java.io.IOException;
import org.datavec.api.records.reader.SequenceRecordReader;
import org.datavec.api.records.reader.impl.csv.CSVSequenceRecordReader;
import org.datavec.api.split.NumberedFileInputSplit;
import org.deeplearning4j.datasets.datavec.SequenceRecordReaderDataSetIterator;
import org.nd4j.linalg.dataset.api.iterator.DataSetIterator;
/**
*
*/
public class SingleFileTimeSeriesDataReader {
private final int miniBatchSize;
private final int numPossibleLabels;
private final boolean regression;
private final String filePattern;
private final int maxFileIdx;
private final int minFileIdx;
private final int numInputs;
/**
*
* #param filePattern
* #param minFileIdx
* #param maxFileIdx
* #param numInputs
* #param numPossibleLabels
* #param miniBatchSize
* #param regression
*/
public SingleFileTimeSeriesDataReader(final String filePattern, final int minFileIdx, final int maxFileIdx,
final int numInputs, final int numPossibleLabels, final int miniBatchSize, final boolean regression) {
this.miniBatchSize = miniBatchSize;
this.numPossibleLabels = numPossibleLabels;
this.regression = regression;
this.filePattern = filePattern;
this.maxFileIdx = maxFileIdx;
this.minFileIdx = minFileIdx;
this.numInputs = numInputs;
}
/**
*
* #return
* #throws IOException
* #throws InterruptedException
*/
public DataSetIterator apply() throws IOException, InterruptedException {
final SequenceRecordReader reader = new CSVSequenceRecordReader(0, ",");
reader.initialize(new NumberedFileInputSplit(filePattern, minFileIdx, maxFileIdx));
final DataSetIterator iter = new SequenceRecordReaderDataSetIterator(reader, miniBatchSize, numPossibleLabels,
numInputs, regression);
return iter;
}
}
TestConfBuilder.java
/**
*
*/
package org.mmarini.lstmtest;
import org.deeplearning4j.nn.api.OptimizationAlgorithm;
import org.deeplearning4j.nn.conf.MultiLayerConfiguration;
import org.deeplearning4j.nn.conf.NeuralNetConfiguration;
import org.deeplearning4j.nn.conf.layers.LSTM;
import org.deeplearning4j.nn.conf.layers.RnnOutputLayer;
import org.deeplearning4j.nn.weights.WeightInit;
import org.nd4j.linalg.activations.Activation;
import org.nd4j.linalg.lossfunctions.LossFunctions.LossFunction;
/**
* #author mmarini
*
*/
public class TestConfBuilder {
private final int noInputUnits;
private final int noOutputUnits;
private final int noLstmUnits;
/**
*
* #param noInputUnits
* #param noOutputUnits
* #param noLstmUnits
*/
public TestConfBuilder(final int noInputUnits, final int noOutputUnits, final int noLstmUnits) {
super();
this.noInputUnits = noInputUnits;
this.noOutputUnits = noOutputUnits;
this.noLstmUnits = noLstmUnits;
}
/**
*
* #return
*/
public MultiLayerConfiguration build() {
final NeuralNetConfiguration.Builder builder = new NeuralNetConfiguration.Builder()
.weightInit(WeightInit.XAVIER).optimizationAlgo(OptimizationAlgorithm.STOCHASTIC_GRADIENT_DESCENT);
final LSTM lstmLayer = new LSTM.Builder().units(noLstmUnits).nIn(noInputUnits).activation(Activation.TANH)
.build();
final RnnOutputLayer outLayer = new RnnOutputLayer.Builder(LossFunction.MEAN_SQUARED_LOGARITHMIC_ERROR)
.activation(Activation.IDENTITY).nOut(noOutputUnits).nIn(noLstmUnits).build();
final MultiLayerConfiguration conf = builder.list(lstmLayer, outLayer).build();
return conf;
}
}
TestTrainingTest .java
package org.mmarini.lstmtest;
import static org.hamcrest.CoreMatchers.equalTo;
import static org.hamcrest.MatcherAssert.assertThat;
import static org.junit.jupiter.api.Assertions.assertNotNull;
import java.io.File;
import java.io.IOException;
import java.util.Arrays;
import org.deeplearning4j.datasets.iterator.INDArrayDataSetIterator;
import org.deeplearning4j.nn.conf.MultiLayerConfiguration;
import org.deeplearning4j.nn.multilayer.MultiLayerNetwork;
import org.junit.jupiter.api.Test;
import org.nd4j.linalg.api.ndarray.INDArray;
import org.nd4j.linalg.dataset.api.iterator.DataSetIterator;
import org.nd4j.linalg.factory.Nd4j;
import org.nd4j.linalg.primitives.Pair;
import org.nd4j.linalg.util.ArrayUtil;
class TestTrainingTest {
private static final int MINI_BATCH_SIZE = 10;
private static final int NUM_LABELS = 1;
private static final boolean REGRESSION = true;
private static final String SAMPLES_FILE = "src/test/resources/datatest/sample_%d.csv";
private static final int MIN_INPUTS_FILE_IDX = 0;
private static final int MAX_INPUTS_FILE_IDX = 9;
private static final int NUM_INPUTS_COLUMN = 1;
private static final int NUM_HIDDEN_UNITS = 1;
DataSetIterator createData() {
final double[][][] featuresAry = new double[][][] { { { 0.5, 0.2, 0.5 } }, { { 0.5, 1.0, 0.0 } } };
final double[] featuresData = ArrayUtil.flattenDoubleArray(featuresAry);
final int[] featuresShape = new int[] { 2, 1, 3 };
final INDArray features = Nd4j.create(featuresData, featuresShape, 'c');
final double[][][] labelsAry = new double[][][] { { { 1.0, -1.0, 1.0 }, { 1.0, -1.0, -1.0 } } };
final double[] labelsData = ArrayUtil.flattenDoubleArray(labelsAry);
final int[] labelsShape = new int[] { 2, 1, 3 };
final INDArray labels = Nd4j.create(labelsData, labelsShape, 'c');
final INDArrayDataSetIterator iter = new INDArrayDataSetIterator(
Arrays.asList(new Pair<INDArray, INDArray>(features, labels)), 2);
System.out.println(iter.inputColumns());
return iter;
}
private String file(String template) {
return new File(".", template).getAbsolutePath();
}
#Test
void testBuild() throws IOException, InterruptedException {
final SingleFileTimeSeriesDataReader reader = new SingleFileTimeSeriesDataReader(file(SAMPLES_FILE),
MIN_INPUTS_FILE_IDX, MAX_INPUTS_FILE_IDX, NUM_INPUTS_COLUMN, NUM_LABELS, MINI_BATCH_SIZE, REGRESSION);
final DataSetIterator data = reader.apply();
assertThat(data.inputColumns(), equalTo(NUM_INPUTS_COLUMN));
assertThat(data.totalOutcomes(), equalTo(NUM_LABELS));
final TestConfBuilder builder = new TestConfBuilder(NUM_INPUTS_COLUMN, NUM_LABELS, NUM_HIDDEN_UNITS);
final MultiLayerConfiguration conf = builder.build();
final MultiLayerNetwork net = new MultiLayerNetwork(conf);
assertNotNull(net);
net.init();
net.fit(data);
}
}
I expect not to throw any exception but I got the following exception:
java.lang.IllegalStateException: C (result) array is not F order or is a view. Nd4j.gemm requires the result array to be F order and not a view. C (result) array: [Rank: 2,Offset: 0 Order: f Shape: [10,1], stride: [1,10]]
at org.nd4j.base.Preconditions.throwStateEx(Preconditions.java:641)
at org.nd4j.base.Preconditions.checkState(Preconditions.java:304)
at org.nd4j.linalg.factory.Nd4j.gemm(Nd4j.java:980)
at org.deeplearning4j.nn.layers.recurrent.LSTMHelpers.backpropGradientHelper(LSTMHelpers.java:696)
at org.deeplearning4j.nn.layers.recurrent.LSTM.backpropGradientHelper(LSTM.java:122)
at org.deeplearning4j.nn.layers.recurrent.LSTM.backpropGradient(LSTM.java:93)
at org.deeplearning4j.nn.multilayer.MultiLayerNetwork.calcBackpropGradients(MultiLayerNetwork.java:1826)
at org.deeplearning4j.nn.multilayer.MultiLayerNetwork.computeGradientAndScore(MultiLayerNetwork.java:2644)
at org.deeplearning4j.nn.multilayer.MultiLayerNetwork.computeGradientAndScore(MultiLayerNetwork.java:2587)
at org.deeplearning4j.optimize.solvers.BaseOptimizer.gradientAndScore(BaseOptimizer.java:160)
at org.deeplearning4j.optimize.solvers.StochasticGradientDescent.optimize(StochasticGradientDescent.java:63)
at org.deeplearning4j.optimize.Solver.optimize(Solver.java:52)
at org.deeplearning4j.nn.multilayer.MultiLayerNetwork.fitHelper(MultiLayerNetwork.java:1602)
at org.deeplearning4j.nn.multilayer.MultiLayerNetwork.fit(MultiLayerNetwork.java:1521)
at org.mmarini.lstmtest.TestTrainingTest.testBuild(TestTrainingTest.java:77)
at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
at java.lang.reflect.Method.invoke(Method.java:498)
at org.junit.platform.commons.util.ReflectionUtils.invokeMethod(ReflectionUtils.java:532)
at org.junit.jupiter.engine.execution.ExecutableInvoker.invoke(ExecutableInvoker.java:115)
at org.junit.jupiter.engine.descriptor.TestMethodTestDescriptor.lambda$invokeTestMethod$6(TestMethodTestDescriptor.java:171)
at org.junit.platform.engine.support.hierarchical.ThrowableCollector.execute(ThrowableCollector.java:72)
at org.junit.jupiter.engine.descriptor.TestMethodTestDescriptor.invokeTestMethod(TestMethodTestDescriptor.java:167)
at org.junit.jupiter.engine.descriptor.TestMethodTestDescriptor.execute(TestMethodTestDescriptor.java:114)
at org.junit.jupiter.engine.descriptor.TestMethodTestDescriptor.execute(TestMethodTestDescriptor.java:59)
at org.junit.platform.engine.support.hierarchical.NodeTestTask.lambda$executeRecursively$4(NodeTestTask.java:108)
at org.junit.platform.engine.support.hierarchical.ThrowableCollector.execute(ThrowableCollector.java:72)
at org.junit.platform.engine.support.hierarchical.NodeTestTask.executeRecursively(NodeTestTask.java:98)
at org.junit.platform.engine.support.hierarchical.NodeTestTask.execute(NodeTestTask.java:74)
at java.util.ArrayList.forEach(ArrayList.java:1257)
at org.junit.platform.engine.support.hierarchical.SameThreadHierarchicalTestExecutorService.invokeAll(SameThreadHierarchicalTestExecutorService.java:38)
at org.junit.platform.engine.support.hierarchical.NodeTestTask.lambda$executeRecursively$4(NodeTestTask.java:112)
at org.junit.platform.engine.support.hierarchical.ThrowableCollector.execute(ThrowableCollector.java:72)
at org.junit.platform.engine.support.hierarchical.NodeTestTask.executeRecursively(NodeTestTask.java:98)
at org.junit.platform.engine.support.hierarchical.NodeTestTask.execute(NodeTestTask.java:74)
at java.util.ArrayList.forEach(ArrayList.java:1257)
at org.junit.platform.engine.support.hierarchical.SameThreadHierarchicalTestExecutorService.invokeAll(SameThreadHierarchicalTestExecutorService.java:38)
at org.junit.platform.engine.support.hierarchical.NodeTestTask.lambda$executeRecursively$4(NodeTestTask.java:112)
at org.junit.platform.engine.support.hierarchical.ThrowableCollector.execute(ThrowableCollector.java:72)
at org.junit.platform.engine.support.hierarchical.NodeTestTask.executeRecursively(NodeTestTask.java:98)
at org.junit.platform.engine.support.hierarchical.NodeTestTask.execute(NodeTestTask.java:74)
at org.junit.platform.engine.support.hierarchical.SameThreadHierarchicalTestExecutorService.submit(SameThreadHierarchicalTestExecutorService.java:32)
at org.junit.platform.engine.support.hierarchical.HierarchicalTestExecutor.execute(HierarchicalTestExecutor.java:57)
at org.junit.platform.engine.support.hierarchical.HierarchicalTestEngine.execute(HierarchicalTestEngine.java:51)
at org.junit.platform.launcher.core.DefaultLauncher.execute(DefaultLauncher.java:220)
at org.junit.platform.launcher.core.DefaultLauncher.lambda$execute$6(DefaultLauncher.java:188)
at org.junit.platform.launcher.core.DefaultLauncher.withInterceptedStreams(DefaultLauncher.java:202)
at org.junit.platform.launcher.core.DefaultLauncher.execute(DefaultLauncher.java:181)
at org.junit.platform.launcher.core.DefaultLauncher.execute(DefaultLauncher.java:128)
at org.eclipse.jdt.internal.junit5.runner.JUnit5TestReference.run(JUnit5TestReference.java:89)
at org.eclipse.jdt.internal.junit.runner.TestExecution.run(TestExecution.java:41)
at org.eclipse.jdt.internal.junit.runner.RemoteTestRunner.runTests(RemoteTestRunner.java:541)
at org.eclipse.jdt.internal.junit.runner.RemoteTestRunner.runTests(RemoteTestRunner.java:763)
at org.eclipse.jdt.internal.junit.runner.RemoteTestRunner.run(RemoteTestRunner.java:463)
at org.eclipse.jdt.internal.junit.runner.RemoteTestRunner.main(RemoteTestRunner.java:209)

Answer 1

Please see the DL4J Gitter community: https://gitter.im/deeplearning4j/deeplearning4j