Search in sources :

Example 16 with MultiDataSet

use of org.nd4j.linalg.dataset.api.MultiDataSet in project deeplearning4j by deeplearning4j.

the class RecordReaderMultiDataSetIteratorTest method testSplittingCSVSequenceMeta.

@Test
public void testSplittingCSVSequenceMeta() throws Exception {
    //need to manually extract
    for (int i = 0; i < 3; i++) {
        new ClassPathResource(String.format("csvsequence_%d.txt", i)).getTempFileFromArchive();
        new ClassPathResource(String.format("csvsequencelabels_%d.txt", i)).getTempFileFromArchive();
        new ClassPathResource(String.format("csvsequencelabelsShort_%d.txt", i)).getTempFileFromArchive();
    }
    ClassPathResource resource = new ClassPathResource("csvsequence_0.txt");
    String featuresPath = resource.getTempFileFromArchive().getAbsolutePath().replaceAll("0", "%d");
    resource = new ClassPathResource("csvsequencelabels_0.txt");
    String labelsPath = resource.getTempFileFromArchive().getAbsolutePath().replaceAll("0", "%d");
    SequenceRecordReader featureReader = new CSVSequenceRecordReader(1, ",");
    SequenceRecordReader labelReader = new CSVSequenceRecordReader(1, ",");
    featureReader.initialize(new NumberedFileInputSplit(featuresPath, 0, 2));
    labelReader.initialize(new NumberedFileInputSplit(labelsPath, 0, 2));
    SequenceRecordReader featureReader2 = new CSVSequenceRecordReader(1, ",");
    SequenceRecordReader labelReader2 = new CSVSequenceRecordReader(1, ",");
    featureReader2.initialize(new NumberedFileInputSplit(featuresPath, 0, 2));
    labelReader2.initialize(new NumberedFileInputSplit(labelsPath, 0, 2));
    RecordReaderMultiDataSetIterator srrmdsi = new RecordReaderMultiDataSetIterator.Builder(1).addSequenceReader("seq1", featureReader2).addSequenceReader("seq2", labelReader2).addInput("seq1", 0, 1).addInput("seq1", 2, 2).addOutputOneHot("seq2", 0, 4).build();
    srrmdsi.setCollectMetaData(true);
    int count = 0;
    while (srrmdsi.hasNext()) {
        MultiDataSet mds = srrmdsi.next();
        MultiDataSet fromMeta = srrmdsi.loadFromMetaData(mds.getExampleMetaData(RecordMetaData.class));
        assertEquals(mds, fromMeta);
        count++;
    }
    assertEquals(3, count);
}
Also used : RecordMetaData(org.datavec.api.records.metadata.RecordMetaData) CSVSequenceRecordReader(org.datavec.api.records.reader.impl.csv.CSVSequenceRecordReader) SequenceRecordReader(org.datavec.api.records.reader.SequenceRecordReader) CSVSequenceRecordReader(org.datavec.api.records.reader.impl.csv.CSVSequenceRecordReader) MultiDataSet(org.nd4j.linalg.dataset.api.MultiDataSet) ClassPathResource(org.nd4j.linalg.io.ClassPathResource) NumberedFileInputSplit(org.datavec.api.split.NumberedFileInputSplit) Test(org.junit.Test)

Example 17 with MultiDataSet

use of org.nd4j.linalg.dataset.api.MultiDataSet in project deeplearning4j by deeplearning4j.

the class RecordReaderMultiDataSetIteratorTest method testsBasicMeta.

@Test
public void testsBasicMeta() throws Exception {
    //As per testBasic - but also loading metadata
    RecordReader rr2 = new CSVRecordReader(0, ",");
    rr2.initialize(new FileSplit(new ClassPathResource("iris.txt").getTempFileFromArchive()));
    RecordReaderMultiDataSetIterator rrmdsi = new RecordReaderMultiDataSetIterator.Builder(10).addReader("reader", rr2).addInput("reader", 0, 3).addOutputOneHot("reader", 4, 3).build();
    rrmdsi.setCollectMetaData(true);
    int count = 0;
    while (rrmdsi.hasNext()) {
        MultiDataSet mds = rrmdsi.next();
        MultiDataSet fromMeta = rrmdsi.loadFromMetaData(mds.getExampleMetaData(RecordMetaData.class));
        assertEquals(mds, fromMeta);
        count++;
    }
    assertEquals(150 / 10, count);
}
Also used : RecordMetaData(org.datavec.api.records.metadata.RecordMetaData) MultiDataSet(org.nd4j.linalg.dataset.api.MultiDataSet) RecordReader(org.datavec.api.records.reader.RecordReader) ImageRecordReader(org.datavec.image.recordreader.ImageRecordReader) CSVSequenceRecordReader(org.datavec.api.records.reader.impl.csv.CSVSequenceRecordReader) CSVRecordReader(org.datavec.api.records.reader.impl.csv.CSVRecordReader) SequenceRecordReader(org.datavec.api.records.reader.SequenceRecordReader) CSVRecordReader(org.datavec.api.records.reader.impl.csv.CSVRecordReader) FileSplit(org.datavec.api.split.FileSplit) ClassPathResource(org.nd4j.linalg.io.ClassPathResource) Test(org.junit.Test)

Example 18 with MultiDataSet

use of org.nd4j.linalg.dataset.api.MultiDataSet in project deeplearning4j by deeplearning4j.

the class RecordReaderMultiDataSetIteratorTest method testVariableLengthTSMeta.

@Test
public void testVariableLengthTSMeta() throws Exception {
    //need to manually extract
    for (int i = 0; i < 3; i++) {
        new ClassPathResource(String.format("csvsequence_%d.txt", i)).getTempFileFromArchive();
        new ClassPathResource(String.format("csvsequencelabels_%d.txt", i)).getTempFileFromArchive();
        new ClassPathResource(String.format("csvsequencelabelsShort_%d.txt", i)).getTempFileFromArchive();
    }
    //Set up SequenceRecordReaderDataSetIterators for comparison
    ClassPathResource resource = new ClassPathResource("csvsequence_0.txt");
    String featuresPath = resource.getTempFileFromArchive().getAbsolutePath().replaceAll("0", "%d");
    resource = new ClassPathResource("csvsequencelabelsShort_0.txt");
    String labelsPath = resource.getTempFileFromArchive().getAbsolutePath().replaceAll("0", "%d");
    //Set up
    SequenceRecordReader featureReader3 = new CSVSequenceRecordReader(1, ",");
    SequenceRecordReader labelReader3 = new CSVSequenceRecordReader(1, ",");
    featureReader3.initialize(new NumberedFileInputSplit(featuresPath, 0, 2));
    labelReader3.initialize(new NumberedFileInputSplit(labelsPath, 0, 2));
    SequenceRecordReader featureReader4 = new CSVSequenceRecordReader(1, ",");
    SequenceRecordReader labelReader4 = new CSVSequenceRecordReader(1, ",");
    featureReader4.initialize(new NumberedFileInputSplit(featuresPath, 0, 2));
    labelReader4.initialize(new NumberedFileInputSplit(labelsPath, 0, 2));
    RecordReaderMultiDataSetIterator rrmdsiStart = new RecordReaderMultiDataSetIterator.Builder(1).addSequenceReader("in", featureReader3).addSequenceReader("out", labelReader3).addInput("in").addOutputOneHot("out", 0, 4).sequenceAlignmentMode(RecordReaderMultiDataSetIterator.AlignmentMode.ALIGN_START).build();
    RecordReaderMultiDataSetIterator rrmdsiEnd = new RecordReaderMultiDataSetIterator.Builder(1).addSequenceReader("in", featureReader4).addSequenceReader("out", labelReader4).addInput("in").addOutputOneHot("out", 0, 4).sequenceAlignmentMode(RecordReaderMultiDataSetIterator.AlignmentMode.ALIGN_END).build();
    rrmdsiStart.setCollectMetaData(true);
    rrmdsiEnd.setCollectMetaData(true);
    int count = 0;
    while (rrmdsiStart.hasNext()) {
        MultiDataSet mdsStart = rrmdsiStart.next();
        MultiDataSet mdsEnd = rrmdsiEnd.next();
        MultiDataSet mdsStartFromMeta = rrmdsiStart.loadFromMetaData(mdsStart.getExampleMetaData(RecordMetaData.class));
        MultiDataSet mdsEndFromMeta = rrmdsiEnd.loadFromMetaData(mdsEnd.getExampleMetaData(RecordMetaData.class));
        assertEquals(mdsStart, mdsStartFromMeta);
        assertEquals(mdsEnd, mdsEndFromMeta);
        count++;
    }
    assertFalse(rrmdsiStart.hasNext());
    assertFalse(rrmdsiEnd.hasNext());
    assertEquals(3, count);
}
Also used : RecordMetaData(org.datavec.api.records.metadata.RecordMetaData) CSVSequenceRecordReader(org.datavec.api.records.reader.impl.csv.CSVSequenceRecordReader) SequenceRecordReader(org.datavec.api.records.reader.SequenceRecordReader) CSVSequenceRecordReader(org.datavec.api.records.reader.impl.csv.CSVSequenceRecordReader) MultiDataSet(org.nd4j.linalg.dataset.api.MultiDataSet) ClassPathResource(org.nd4j.linalg.io.ClassPathResource) NumberedFileInputSplit(org.datavec.api.split.NumberedFileInputSplit) Test(org.junit.Test)

Example 19 with MultiDataSet

use of org.nd4j.linalg.dataset.api.MultiDataSet in project deeplearning4j by deeplearning4j.

the class RecordReaderMultiDataSetIteratorTest method testsBasic.

@Test
public void testsBasic() throws Exception {
    //Load details from CSV files; single input/output -> compare to RecordReaderDataSetIterator
    RecordReader rr = new CSVRecordReader(0, ",");
    rr.initialize(new FileSplit(new ClassPathResource("iris.txt").getTempFileFromArchive()));
    RecordReaderDataSetIterator rrdsi = new RecordReaderDataSetIterator(rr, 10, 4, 3);
    RecordReader rr2 = new CSVRecordReader(0, ",");
    rr2.initialize(new FileSplit(new ClassPathResource("iris.txt").getTempFileFromArchive()));
    MultiDataSetIterator rrmdsi = new RecordReaderMultiDataSetIterator.Builder(10).addReader("reader", rr2).addInput("reader", 0, 3).addOutputOneHot("reader", 4, 3).build();
    while (rrdsi.hasNext()) {
        DataSet ds = rrdsi.next();
        INDArray fds = ds.getFeatureMatrix();
        INDArray lds = ds.getLabels();
        MultiDataSet mds = rrmdsi.next();
        assertEquals(1, mds.getFeatures().length);
        assertEquals(1, mds.getLabels().length);
        assertNull(mds.getFeaturesMaskArrays());
        assertNull(mds.getLabelsMaskArrays());
        INDArray fmds = mds.getFeatures(0);
        INDArray lmds = mds.getLabels(0);
        assertNotNull(fmds);
        assertNotNull(lmds);
        assertEquals(fds, fmds);
        assertEquals(lds, lmds);
    }
    assertFalse(rrmdsi.hasNext());
    //need to manually extract
    for (int i = 0; i < 3; i++) {
        new ClassPathResource(String.format("csvsequence_%d.txt", i)).getTempFileFromArchive();
        new ClassPathResource(String.format("csvsequencelabels_%d.txt", i)).getTempFileFromArchive();
        new ClassPathResource(String.format("csvsequencelabelsShort_%d.txt", i)).getTempFileFromArchive();
    }
    //Load time series from CSV sequence files; compare to SequenceRecordReaderDataSetIterator
    ClassPathResource resource = new ClassPathResource("csvsequence_0.txt");
    String featuresPath = resource.getTempFileFromArchive().getAbsolutePath().replaceAll("0", "%d");
    resource = new ClassPathResource("csvsequencelabels_0.txt");
    String labelsPath = resource.getTempFileFromArchive().getAbsolutePath().replaceAll("0", "%d");
    SequenceRecordReader featureReader = new CSVSequenceRecordReader(1, ",");
    SequenceRecordReader labelReader = new CSVSequenceRecordReader(1, ",");
    featureReader.initialize(new NumberedFileInputSplit(featuresPath, 0, 2));
    labelReader.initialize(new NumberedFileInputSplit(labelsPath, 0, 2));
    SequenceRecordReaderDataSetIterator iter = new SequenceRecordReaderDataSetIterator(featureReader, labelReader, 1, 4, false);
    SequenceRecordReader featureReader2 = new CSVSequenceRecordReader(1, ",");
    SequenceRecordReader labelReader2 = new CSVSequenceRecordReader(1, ",");
    featureReader2.initialize(new NumberedFileInputSplit(featuresPath, 0, 2));
    labelReader2.initialize(new NumberedFileInputSplit(labelsPath, 0, 2));
    MultiDataSetIterator srrmdsi = new RecordReaderMultiDataSetIterator.Builder(1).addSequenceReader("in", featureReader2).addSequenceReader("out", labelReader2).addInput("in").addOutputOneHot("out", 0, 4).build();
    while (iter.hasNext()) {
        DataSet ds = iter.next();
        INDArray fds = ds.getFeatureMatrix();
        INDArray lds = ds.getLabels();
        MultiDataSet mds = srrmdsi.next();
        assertEquals(1, mds.getFeatures().length);
        assertEquals(1, mds.getLabels().length);
        assertNull(mds.getFeaturesMaskArrays());
        assertNull(mds.getLabelsMaskArrays());
        INDArray fmds = mds.getFeatures(0);
        INDArray lmds = mds.getLabels(0);
        assertNotNull(fmds);
        assertNotNull(lmds);
        assertEquals(fds, fmds);
        assertEquals(lds, lmds);
    }
    assertFalse(srrmdsi.hasNext());
}
Also used : CSVSequenceRecordReader(org.datavec.api.records.reader.impl.csv.CSVSequenceRecordReader) SequenceRecordReader(org.datavec.api.records.reader.SequenceRecordReader) DataSet(org.nd4j.linalg.dataset.DataSet) MultiDataSet(org.nd4j.linalg.dataset.api.MultiDataSet) RecordReader(org.datavec.api.records.reader.RecordReader) ImageRecordReader(org.datavec.image.recordreader.ImageRecordReader) CSVSequenceRecordReader(org.datavec.api.records.reader.impl.csv.CSVSequenceRecordReader) CSVRecordReader(org.datavec.api.records.reader.impl.csv.CSVRecordReader) SequenceRecordReader(org.datavec.api.records.reader.SequenceRecordReader) FileSplit(org.datavec.api.split.FileSplit) ClassPathResource(org.nd4j.linalg.io.ClassPathResource) NumberedFileInputSplit(org.datavec.api.split.NumberedFileInputSplit) MultiDataSetIterator(org.nd4j.linalg.dataset.api.iterator.MultiDataSetIterator) INDArray(org.nd4j.linalg.api.ndarray.INDArray) MultiDataSet(org.nd4j.linalg.dataset.api.MultiDataSet) CSVSequenceRecordReader(org.datavec.api.records.reader.impl.csv.CSVSequenceRecordReader) CSVRecordReader(org.datavec.api.records.reader.impl.csv.CSVRecordReader) Test(org.junit.Test)

Example 20 with MultiDataSet

use of org.nd4j.linalg.dataset.api.MultiDataSet in project deeplearning4j by deeplearning4j.

the class RecordReaderMultiDataSetIteratorTest method testVariableLengthTS.

@Test
public void testVariableLengthTS() throws Exception {
    //need to manually extract
    for (int i = 0; i < 3; i++) {
        new ClassPathResource(String.format("csvsequence_%d.txt", i)).getTempFileFromArchive();
        new ClassPathResource(String.format("csvsequencelabels_%d.txt", i)).getTempFileFromArchive();
        new ClassPathResource(String.format("csvsequencelabelsShort_%d.txt", i)).getTempFileFromArchive();
    }
    //Set up SequenceRecordReaderDataSetIterators for comparison
    ClassPathResource resource = new ClassPathResource("csvsequence_0.txt");
    String featuresPath = resource.getTempFileFromArchive().getAbsolutePath().replaceAll("0", "%d");
    resource = new ClassPathResource("csvsequencelabelsShort_0.txt");
    String labelsPath = resource.getTempFileFromArchive().getAbsolutePath().replaceAll("0", "%d");
    SequenceRecordReader featureReader = new CSVSequenceRecordReader(1, ",");
    SequenceRecordReader labelReader = new CSVSequenceRecordReader(1, ",");
    featureReader.initialize(new NumberedFileInputSplit(featuresPath, 0, 2));
    labelReader.initialize(new NumberedFileInputSplit(labelsPath, 0, 2));
    SequenceRecordReader featureReader2 = new CSVSequenceRecordReader(1, ",");
    SequenceRecordReader labelReader2 = new CSVSequenceRecordReader(1, ",");
    featureReader2.initialize(new NumberedFileInputSplit(featuresPath, 0, 2));
    labelReader2.initialize(new NumberedFileInputSplit(labelsPath, 0, 2));
    SequenceRecordReaderDataSetIterator iterAlignStart = new SequenceRecordReaderDataSetIterator(featureReader, labelReader, 1, 4, false, SequenceRecordReaderDataSetIterator.AlignmentMode.ALIGN_START);
    SequenceRecordReaderDataSetIterator iterAlignEnd = new SequenceRecordReaderDataSetIterator(featureReader2, labelReader2, 1, 4, false, SequenceRecordReaderDataSetIterator.AlignmentMode.ALIGN_END);
    //Set up
    SequenceRecordReader featureReader3 = new CSVSequenceRecordReader(1, ",");
    SequenceRecordReader labelReader3 = new CSVSequenceRecordReader(1, ",");
    featureReader3.initialize(new NumberedFileInputSplit(featuresPath, 0, 2));
    labelReader3.initialize(new NumberedFileInputSplit(labelsPath, 0, 2));
    SequenceRecordReader featureReader4 = new CSVSequenceRecordReader(1, ",");
    SequenceRecordReader labelReader4 = new CSVSequenceRecordReader(1, ",");
    featureReader4.initialize(new NumberedFileInputSplit(featuresPath, 0, 2));
    labelReader4.initialize(new NumberedFileInputSplit(labelsPath, 0, 2));
    RecordReaderMultiDataSetIterator rrmdsiStart = new RecordReaderMultiDataSetIterator.Builder(1).addSequenceReader("in", featureReader3).addSequenceReader("out", labelReader3).addInput("in").addOutputOneHot("out", 0, 4).sequenceAlignmentMode(RecordReaderMultiDataSetIterator.AlignmentMode.ALIGN_START).build();
    RecordReaderMultiDataSetIterator rrmdsiEnd = new RecordReaderMultiDataSetIterator.Builder(1).addSequenceReader("in", featureReader4).addSequenceReader("out", labelReader4).addInput("in").addOutputOneHot("out", 0, 4).sequenceAlignmentMode(RecordReaderMultiDataSetIterator.AlignmentMode.ALIGN_END).build();
    while (iterAlignStart.hasNext()) {
        DataSet dsStart = iterAlignStart.next();
        DataSet dsEnd = iterAlignEnd.next();
        MultiDataSet mdsStart = rrmdsiStart.next();
        MultiDataSet mdsEnd = rrmdsiEnd.next();
        assertEquals(1, mdsStart.getFeatures().length);
        assertEquals(1, mdsStart.getLabels().length);
        //assertEquals(1, mdsStart.getFeaturesMaskArrays().length); //Features data is always longer -> don't need mask arrays for it
        assertEquals(1, mdsStart.getLabelsMaskArrays().length);
        assertEquals(1, mdsEnd.getFeatures().length);
        assertEquals(1, mdsEnd.getLabels().length);
        //assertEquals(1, mdsEnd.getFeaturesMaskArrays().length);
        assertEquals(1, mdsEnd.getLabelsMaskArrays().length);
        assertEquals(dsStart.getFeatureMatrix(), mdsStart.getFeatures(0));
        assertEquals(dsStart.getLabels(), mdsStart.getLabels(0));
        assertEquals(dsStart.getLabelsMaskArray(), mdsStart.getLabelsMaskArray(0));
        assertEquals(dsEnd.getFeatureMatrix(), mdsEnd.getFeatures(0));
        assertEquals(dsEnd.getLabels(), mdsEnd.getLabels(0));
        assertEquals(dsEnd.getLabelsMaskArray(), mdsEnd.getLabelsMaskArray(0));
    }
    assertFalse(rrmdsiStart.hasNext());
    assertFalse(rrmdsiEnd.hasNext());
}
Also used : CSVSequenceRecordReader(org.datavec.api.records.reader.impl.csv.CSVSequenceRecordReader) SequenceRecordReader(org.datavec.api.records.reader.SequenceRecordReader) CSVSequenceRecordReader(org.datavec.api.records.reader.impl.csv.CSVSequenceRecordReader) MultiDataSet(org.nd4j.linalg.dataset.api.MultiDataSet) DataSet(org.nd4j.linalg.dataset.DataSet) MultiDataSet(org.nd4j.linalg.dataset.api.MultiDataSet) ClassPathResource(org.nd4j.linalg.io.ClassPathResource) NumberedFileInputSplit(org.datavec.api.split.NumberedFileInputSplit) Test(org.junit.Test)

Aggregations

MultiDataSet (org.nd4j.linalg.dataset.api.MultiDataSet)28 Test (org.junit.Test)12 ClassPathResource (org.nd4j.linalg.io.ClassPathResource)11 INDArray (org.nd4j.linalg.api.ndarray.INDArray)10 MultiDataSetIterator (org.nd4j.linalg.dataset.api.iterator.MultiDataSetIterator)10 SequenceRecordReader (org.datavec.api.records.reader.SequenceRecordReader)8 CSVSequenceRecordReader (org.datavec.api.records.reader.impl.csv.CSVSequenceRecordReader)8 DataSet (org.nd4j.linalg.dataset.DataSet)8 FileSplit (org.datavec.api.split.FileSplit)7 ImageRecordReader (org.datavec.image.recordreader.ImageRecordReader)6 ComputationGraph (org.deeplearning4j.nn.graph.ComputationGraph)6 RecordReader (org.datavec.api.records.reader.RecordReader)5 CSVRecordReader (org.datavec.api.records.reader.impl.csv.CSVRecordReader)5 NumberedFileInputSplit (org.datavec.api.split.NumberedFileInputSplit)5 ArrayList (java.util.ArrayList)4 RecordMetaData (org.datavec.api.records.metadata.RecordMetaData)4 GridExecutioner (org.nd4j.linalg.api.ops.executioner.GridExecutioner)4 AsyncMultiDataSetIterator (org.deeplearning4j.datasets.iterator.AsyncMultiDataSetIterator)3 DataSetIterator (org.nd4j.linalg.dataset.api.iterator.DataSetIterator)3 Random (java.util.Random)2