Examples with JavaSparkContext org.apache.spark.api.java.JavaSparkContext used on opensource projects

Example 41 with JavaSparkContext

use of org.apache.spark.api.java.JavaSparkContext in project deeplearning4j by deeplearning4j.

the class TextPipelineTest method testWordFreqAccIdentifyingStopWords.

@Test
public void testWordFreqAccIdentifyingStopWords() throws Exception {
    JavaSparkContext sc = getContext();
    //  word2vec.setRemoveStop(false);
    JavaRDD<String> corpusRDD = getCorpusRDD(sc);
    Broadcast<Map<String, Object>> broadcastTokenizerVarMap = sc.broadcast(word2vec.getTokenizerVarMap());
    TextPipeline pipeline = new TextPipeline(corpusRDD, broadcastTokenizerVarMap);
    JavaRDD<List<String>> tokenizedRDD = pipeline.tokenize();
    pipeline.updateAndReturnAccumulatorVal(tokenizedRDD);
    Counter<String> wordFreqCounter = pipeline.getWordFreqAcc().value();
    assertEquals(wordFreqCounter.getCount("is"), 0, 0);
    assertEquals(wordFreqCounter.getCount("this"), 0, 0);
    assertEquals(wordFreqCounter.getCount("are"), 0, 0);
    assertEquals(wordFreqCounter.getCount("a"), 0, 0);
    assertEquals(wordFreqCounter.getCount("STOP"), 4, 0);
    assertEquals(wordFreqCounter.getCount("strange"), 2, 0);
    assertEquals(wordFreqCounter.getCount("flowers"), 1, 0);
    assertEquals(wordFreqCounter.getCount("world"), 1, 0);
    assertEquals(wordFreqCounter.getCount("red"), 1, 0);
    sc.stop();
}

Example 42 with JavaSparkContext

use of org.apache.spark.api.java.JavaSparkContext in project deeplearning4j by deeplearning4j.

the class TextPipelineTest method testBuildVocabCache.

@Test
public void testBuildVocabCache() throws Exception {
    JavaSparkContext sc = getContext();
    JavaRDD<String> corpusRDD = getCorpusRDD(sc);
    Broadcast<Map<String, Object>> broadcastTokenizerVarMap = sc.broadcast(word2vec.getTokenizerVarMap());
    TextPipeline pipeline = new TextPipeline(corpusRDD, broadcastTokenizerVarMap);
    pipeline.buildVocabCache();
    VocabCache<VocabWord> vocabCache = pipeline.getVocabCache();
    assertTrue(vocabCache != null);
    log.info("VocabWords: " + vocabCache.words());
    assertEquals(5, vocabCache.numWords());
    VocabWord redVocab = vocabCache.tokenFor("red");
    VocabWord flowerVocab = vocabCache.tokenFor("flowers");
    VocabWord worldVocab = vocabCache.tokenFor("world");
    VocabWord strangeVocab = vocabCache.tokenFor("strange");
    log.info("Red word: " + redVocab);
    log.info("Flower word: " + flowerVocab);
    log.info("World word: " + worldVocab);
    log.info("Strange word: " + strangeVocab);
    assertEquals(redVocab.getWord(), "red");
    assertEquals(redVocab.getElementFrequency(), 1, 0);
    assertEquals(flowerVocab.getWord(), "flowers");
    assertEquals(flowerVocab.getElementFrequency(), 1, 0);
    assertEquals(worldVocab.getWord(), "world");
    assertEquals(worldVocab.getElementFrequency(), 1, 0);
    assertEquals(strangeVocab.getWord(), "strange");
    assertEquals(strangeVocab.getElementFrequency(), 2, 0);
    sc.stop();
}

Example 43 with JavaSparkContext

use of org.apache.spark.api.java.JavaSparkContext in project deeplearning4j by deeplearning4j.

the class TestSparkComputationGraph method testBasic.

@Test
public void testBasic() throws Exception {
    JavaSparkContext sc = this.sc;
    RecordReader rr = new CSVRecordReader(0, ",");
    rr.initialize(new FileSplit(new ClassPathResource("iris.txt").getTempFileFromArchive()));
    MultiDataSetIterator iter = new RecordReaderMultiDataSetIterator.Builder(1).addReader("iris", rr).addInput("iris", 0, 3).addOutputOneHot("iris", 4, 3).build();
    List<MultiDataSet> list = new ArrayList<>(150);
    while (iter.hasNext()) list.add(iter.next());
    ComputationGraphConfiguration config = new NeuralNetConfiguration.Builder().optimizationAlgo(OptimizationAlgorithm.STOCHASTIC_GRADIENT_DESCENT).learningRate(0.1).graphBuilder().addInputs("in").addLayer("dense", new DenseLayer.Builder().nIn(4).nOut(2).build(), "in").addLayer("out", new OutputLayer.Builder(LossFunctions.LossFunction.MCXENT).nIn(2).nOut(3).build(), "dense").setOutputs("out").pretrain(false).backprop(true).build();
    ComputationGraph cg = new ComputationGraph(config);
    cg.init();
    TrainingMaster tm = new ParameterAveragingTrainingMaster(true, numExecutors(), 1, 10, 1, 0);
    SparkComputationGraph scg = new SparkComputationGraph(sc, cg, tm);
    scg.setListeners(Collections.singleton((IterationListener) new ScoreIterationListener(1)));
    JavaRDD<MultiDataSet> rdd = sc.parallelize(list);
    scg.fitMultiDataSet(rdd);
    //Try: fitting using DataSet
    DataSetIterator iris = new IrisDataSetIterator(1, 150);
    List<DataSet> list2 = new ArrayList<>();
    while (iris.hasNext()) list2.add(iris.next());
    JavaRDD<DataSet> rddDS = sc.parallelize(list2);
    scg.fit(rddDS);
}

Example 44 with JavaSparkContext

use of org.apache.spark.api.java.JavaSparkContext in project deeplearning4j by deeplearning4j.

the class TestCompareParameterAveragingSparkVsSingleMachine method testOneExecutor.

@Test
public void testOneExecutor() {
    //Idea: single worker/executor on Spark should give identical results to a single machine
    int miniBatchSize = 10;
    int nWorkers = 1;
    for (boolean saveUpdater : new boolean[] { true, false }) {
        JavaSparkContext sc = getContext(nWorkers);
        try {
            //Do training locally, for 3 minibatches
            int[] seeds = { 1, 2, 3 };
            MultiLayerNetwork net = new MultiLayerNetwork(getConf(12345, Updater.RMSPROP));
            net.init();
            INDArray initialParams = net.params().dup();
            for (int i = 0; i < seeds.length; i++) {
                DataSet ds = getOneDataSet(miniBatchSize, seeds[i]);
                if (!saveUpdater)
                    net.setUpdater(null);
                net.fit(ds);
            }
            INDArray finalParams = net.params().dup();
            //Do training on Spark with one executor, for 3 separate minibatches
            TrainingMaster tm = getTrainingMaster(1, miniBatchSize, saveUpdater);
            SparkDl4jMultiLayer sparkNet = new SparkDl4jMultiLayer(sc, getConf(12345, Updater.RMSPROP), tm);
            sparkNet.setCollectTrainingStats(true);
            INDArray initialSparkParams = sparkNet.getNetwork().params().dup();
            for (int i = 0; i < seeds.length; i++) {
                List<DataSet> list = getOneDataSetAsIndividalExamples(miniBatchSize, seeds[i]);
                JavaRDD<DataSet> rdd = sc.parallelize(list);
                sparkNet.fit(rdd);
            }
            INDArray finalSparkParams = sparkNet.getNetwork().params().dup();
            assertEquals(initialParams, initialSparkParams);
            assertNotEquals(initialParams, finalParams);
            assertEquals(finalParams, finalSparkParams);
        } finally {
            sc.stop();
        }
    }
}

Example 45 with JavaSparkContext

use of org.apache.spark.api.java.JavaSparkContext in project deeplearning4j by deeplearning4j.

the class TestCompareParameterAveragingSparkVsSingleMachine method testAverageEveryStepGraph.

@Test
public void testAverageEveryStepGraph() {
    //Idea: averaging every step with SGD (SGD updater + optimizer) is mathematically identical to doing the learning
    // on a single machine for synchronous distributed training
    //BUT: This is *ONLY* the case if all workers get an identical number of examples. This won't be the case if
    // we use RDD.randomSplit (which is what occurs if we use .fit(JavaRDD<DataSet> on a data set that needs splitting),
    // which might give a number of examples that isn't divisible by number of workers (like 39 examples on 4 executors)
    //This is also ONLY the case using SGD updater
    int miniBatchSizePerWorker = 10;
    int nWorkers = 4;
    for (boolean saveUpdater : new boolean[] { true, false }) {
        JavaSparkContext sc = getContext(nWorkers);
        try {
            //Do training locally, for 3 minibatches
            int[] seeds = { 1, 2, 3 };
            //                CudaGridExecutioner executioner = (CudaGridExecutioner) Nd4j.getExecutioner();
            ComputationGraph net = new ComputationGraph(getGraphConf(12345, Updater.SGD));
            net.init();
            INDArray initialParams = net.params().dup();
            for (int i = 0; i < seeds.length; i++) {
                DataSet ds = getOneDataSet(miniBatchSizePerWorker * nWorkers, seeds[i]);
                if (!saveUpdater)
                    net.setUpdater(null);
                net.fit(ds);
            }
            INDArray finalParams = net.params().dup();
            //                executioner.addToWatchdog(finalParams, "finalParams");
            //Do training on Spark with one executor, for 3 separate minibatches
            TrainingMaster tm = getTrainingMaster(1, miniBatchSizePerWorker, saveUpdater);
            SparkComputationGraph sparkNet = new SparkComputationGraph(sc, getGraphConf(12345, Updater.SGD), tm);
            sparkNet.setCollectTrainingStats(true);
            INDArray initialSparkParams = sparkNet.getNetwork().params().dup();
            for (int i = 0; i < seeds.length; i++) {
                List<DataSet> list = getOneDataSetAsIndividalExamples(miniBatchSizePerWorker * nWorkers, seeds[i]);
                JavaRDD<DataSet> rdd = sc.parallelize(list);
                sparkNet.fit(rdd);
            }
            System.out.println(sparkNet.getSparkTrainingStats().statsAsString());
            INDArray finalSparkParams = sparkNet.getNetwork().params().dup();
            //                executioner.addToWatchdog(finalSparkParams, "finalSparkParams");
            float[] fp = finalParams.data().asFloat();
            float[] fps = finalSparkParams.data().asFloat();
            System.out.println("Initial (Local) params:       " + Arrays.toString(initialParams.data().asFloat()));
            System.out.println("Initial (Spark) params:       " + Arrays.toString(initialSparkParams.data().asFloat()));
            System.out.println("Final (Local) params: " + Arrays.toString(fp));
            System.out.println("Final (Spark) params: " + Arrays.toString(fps));
            assertEquals(initialParams, initialSparkParams);
            assertNotEquals(initialParams, finalParams);
            assertArrayEquals(fp, fps, 1e-5f);
            double sparkScore = sparkNet.getScore();
            assertTrue(sparkScore > 0.0);
            assertEquals(net.score(), sparkScore, 1e-3);
        } finally {
            sc.stop();
        }
    }
}