Examples with GridExecutioner org.nd4j.linalg.api.ops.executioner.GridExecutioner used on opensource projects

Example 31 with GridExecutioner

use of org.nd4j.linalg.api.ops.executioner.GridExecutioner in project deeplearning4j by deeplearning4j.

the class ScoreExamplesWithKeyFunctionAdapter method call.

@Override
public Iterable<Tuple2<K, Double>> call(Iterator<Tuple2<K, MultiDataSet>> iterator) throws Exception {
    if (!iterator.hasNext()) {
        return Collections.emptyList();
    }
    ComputationGraph network = new ComputationGraph(ComputationGraphConfiguration.fromJson(jsonConfig.getValue()));
    network.init();
    INDArray val = params.value().unsafeDuplication();
    if (val.length() != network.numParams(false))
        throw new IllegalStateException("Network did not have same number of parameters as the broadcast set parameters");
    network.setParams(val);
    List<Tuple2<K, Double>> ret = new ArrayList<>();
    List<MultiDataSet> collect = new ArrayList<>(batchSize);
    List<K> collectKey = new ArrayList<>(batchSize);
    int totalCount = 0;
    while (iterator.hasNext()) {
        collect.clear();
        collectKey.clear();
        int nExamples = 0;
        while (iterator.hasNext() && nExamples < batchSize) {
            Tuple2<K, MultiDataSet> t2 = iterator.next();
            MultiDataSet ds = t2._2();
            int n = ds.getFeatures(0).size(0);
            if (n != 1)
                throw new IllegalStateException("Cannot score examples with one key per data set if " + "data set contains more than 1 example (numExamples: " + n + ")");
            collect.add(ds);
            collectKey.add(t2._1());
            nExamples += n;
        }
        totalCount += nExamples;
        MultiDataSet data = org.nd4j.linalg.dataset.MultiDataSet.merge(collect);
        INDArray scores = network.scoreExamples(data, addRegularization);
        double[] doubleScores = scores.data().asDouble();
        for (int i = 0; i < doubleScores.length; i++) {
            ret.add(new Tuple2<>(collectKey.get(i), doubleScores[i]));
        }
    }
    if (Nd4j.getExecutioner() instanceof GridExecutioner)
        ((GridExecutioner) Nd4j.getExecutioner()).flushQueueBlocking();
    if (log.isDebugEnabled()) {
        log.debug("Scored {} examples ", totalCount);
    }
    return ret;
}

Example 32 with GridExecutioner

use of org.nd4j.linalg.api.ops.executioner.GridExecutioner in project deeplearning4j by deeplearning4j.

the class ParameterAveragingTrainingWorker method getInitialModelGraph.

@Override
public ComputationGraph getInitialModelGraph() {
    if (configuration.isCollectTrainingStats())
        stats = new ParameterAveragingTrainingWorkerStats.ParameterAveragingTrainingWorkerStatsHelper();
    if (configuration.isCollectTrainingStats())
        stats.logBroadcastGetValueStart();
    NetBroadcastTuple tuple = broadcast.getValue();
    if (configuration.isCollectTrainingStats())
        stats.logBroadcastGetValueEnd();
    //Don't want to have shared configuration object: each may update its iteration count (for LR schedule etc) individually
    ComputationGraph net = new ComputationGraph(tuple.getGraphConfiguration().clone());
    //Can't have shared parameter array across executors for parameter averaging, hence the 'true' for clone parameters array arg
    net.init(tuple.getParameters().unsafeDuplication(), false);
    if (tuple.getUpdaterState() != null) {
        //Again: can't have shared updater state
        net.setUpdater(new ComputationGraphUpdater(net, tuple.getUpdaterState().unsafeDuplication()));
    }
    if (Nd4j.getExecutioner() instanceof GridExecutioner)
        ((GridExecutioner) Nd4j.getExecutioner()).flushQueueBlocking();
    configureListeners(net, tuple.getCounter().getAndIncrement());
    if (configuration.isCollectTrainingStats())
        stats.logInitEnd();
    return net;
}

Example 33 with GridExecutioner

use of org.nd4j.linalg.api.ops.executioner.GridExecutioner in project deeplearning4j by deeplearning4j.

the class ParameterAveragingTrainingWorker method getFinalResult.

@Override
public ParameterAveragingTrainingResult getFinalResult(ComputationGraph network) {
    INDArray updaterState = null;
    if (saveUpdater) {
        ComputationGraphUpdater u = network.getUpdater();
        if (u != null)
            updaterState = u.getStateViewArray();
    }
    if (Nd4j.getExecutioner() instanceof GridExecutioner)
        ((GridExecutioner) Nd4j.getExecutioner()).flushQueueBlocking();
    Collection<StorageMetaData> storageMetaData = null;
    Collection<Persistable> listenerStaticInfo = null;
    Collection<Persistable> listenerUpdates = null;
    if (listenerRouterProvider != null) {
        StatsStorageRouter r = listenerRouterProvider.getRouter();
        if (r instanceof VanillaStatsStorageRouter) {
            //TODO this is ugly... need to find a better solution
            VanillaStatsStorageRouter ssr = (VanillaStatsStorageRouter) r;
            storageMetaData = ssr.getStorageMetaData();
            listenerStaticInfo = ssr.getStaticInfo();
            listenerUpdates = ssr.getUpdates();
        }
    }
    return new ParameterAveragingTrainingResult(network.params(), updaterState, network.score(), storageMetaData, listenerStaticInfo, listenerUpdates);
}

Example 34 with GridExecutioner

use of org.nd4j.linalg.api.ops.executioner.GridExecutioner in project deeplearning4j by deeplearning4j.

the class ParameterAveragingTrainingWorker method getFinalResult.

@Override
public ParameterAveragingTrainingResult getFinalResult(MultiLayerNetwork network) {
    INDArray updaterState = null;
    if (saveUpdater) {
        Updater u = network.getUpdater();
        if (u != null)
            updaterState = u.getStateViewArray();
    }
    if (Nd4j.getExecutioner() instanceof GridExecutioner)
        ((GridExecutioner) Nd4j.getExecutioner()).flushQueueBlocking();
    Collection<StorageMetaData> storageMetaData = null;
    Collection<Persistable> listenerStaticInfo = null;
    Collection<Persistable> listenerUpdates = null;
    if (listenerRouterProvider != null) {
        StatsStorageRouter r = listenerRouterProvider.getRouter();
        if (r instanceof VanillaStatsStorageRouter) {
            //TODO this is ugly... need to find a better solution
            VanillaStatsStorageRouter ssr = (VanillaStatsStorageRouter) r;
            storageMetaData = ssr.getStorageMetaData();
            listenerStaticInfo = ssr.getStaticInfo();
            listenerUpdates = ssr.getUpdates();
        }
    }
    return new ParameterAveragingTrainingResult(network.params(), updaterState, network.score(), storageMetaData, listenerStaticInfo, listenerUpdates);
}

Example 35 with GridExecutioner

use of org.nd4j.linalg.api.ops.executioner.GridExecutioner in project deeplearning4j by deeplearning4j.

the class ParameterAveragingElementAddFunction method call.

@Override
public ParameterAveragingAggregationTuple call(ParameterAveragingAggregationTuple tuple, ParameterAveragingTrainingResult result) throws Exception {
    if (tuple == null) {
        return ParameterAveragingAggregationTuple.builder().parametersSum(result.getParameters()).updaterStateSum(result.getUpdaterState()).scoreSum(result.getScore()).aggregationsCount(1).sparkTrainingStats(result.getSparkTrainingStats()).listenerMetaData(result.getListenerMetaData()).listenerStaticInfo(result.getListenerStaticInfo()).listenerUpdates(result.getListenerUpdates()).build();
    }
    INDArray params = tuple.getParametersSum().addi(result.getParameters());
    INDArray updaterStateSum;
    if (tuple.getUpdaterStateSum() == null) {
        updaterStateSum = result.getUpdaterState();
    } else {
        updaterStateSum = tuple.getUpdaterStateSum();
        if (result.getUpdaterState() != null)
            updaterStateSum.addi(result.getUpdaterState());
    }
    double scoreSum = tuple.getScoreSum() + result.getScore();
    SparkTrainingStats stats = tuple.getSparkTrainingStats();
    if (result.getSparkTrainingStats() != null) {
        if (stats == null)
            stats = result.getSparkTrainingStats();
        else
            stats.addOtherTrainingStats(result.getSparkTrainingStats());
    }
    if (Nd4j.getExecutioner() instanceof GridExecutioner)
        ((GridExecutioner) Nd4j.getExecutioner()).flushQueueBlocking();
    Collection<StorageMetaData> listenerMetaData = tuple.getListenerMetaData();
    if (listenerMetaData == null)
        listenerMetaData = result.getListenerMetaData();
    else {
        Collection<StorageMetaData> newMeta = result.getListenerMetaData();
        if (newMeta != null)
            listenerMetaData.addAll(newMeta);
    }
    Collection<Persistable> listenerStaticInfo = tuple.getListenerStaticInfo();
    if (listenerStaticInfo == null)
        listenerStaticInfo = result.getListenerStaticInfo();
    else {
        Collection<Persistable> newStatic = tuple.getListenerStaticInfo();
        if (newStatic != null)
            listenerStaticInfo.addAll(newStatic);
    }
    Collection<Persistable> listenerUpdates = tuple.getListenerUpdates();
    if (listenerUpdates == null)
        listenerUpdates = result.getListenerUpdates();
    else {
        Collection<Persistable> newUpdates = result.getListenerUpdates();
        if (newUpdates != null)
            listenerUpdates.addAll(newUpdates);
    }
    return new ParameterAveragingAggregationTuple(params, updaterStateSum, scoreSum, tuple.getAggregationsCount() + 1, stats, listenerMetaData, listenerStaticInfo, listenerUpdates);
}