org.deeplearning4j.arbiter.task.MultiLayerNetworkTaskCreator Maven / Gradle / Ivy

Go to download
/*
 *
 *  * Copyright 2016 Skymind,Inc.
 *  *
 *  *    Licensed under the Apache License, Version 2.0 (the "License");
 *  *    you may not use this file except in compliance with the License.
 *  *    You may obtain a copy of the License at
 *  *
 *  *        http://www.apache.org/licenses/LICENSE-2.0
 *  *
 *  *    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.
 *
 */
package org.deeplearning4j.arbiter.task;

import lombok.AllArgsConstructor;
import lombok.NoArgsConstructor;
import org.apache.commons.lang3.exception.ExceptionUtils;
import org.deeplearning4j.arbiter.DL4JConfiguration;
import org.deeplearning4j.arbiter.listener.BaseUIStatusReportingListener;
import org.deeplearning4j.arbiter.listener.UIStatusReportingListener;
import org.deeplearning4j.arbiter.optimize.api.Candidate;
import org.deeplearning4j.arbiter.optimize.api.OptimizationResult;
import org.deeplearning4j.arbiter.optimize.api.TaskCreator;
import org.deeplearning4j.arbiter.optimize.api.data.DataProvider;
import org.deeplearning4j.arbiter.optimize.api.evaluation.ModelEvaluator;
import org.deeplearning4j.arbiter.optimize.api.score.ScoreFunction;
import org.deeplearning4j.arbiter.optimize.runner.Status;
import org.deeplearning4j.arbiter.optimize.runner.listener.candidate.UICandidateStatusListener;
import org.deeplearning4j.datasets.iterator.DataSetIterator;
import org.deeplearning4j.earlystopping.EarlyStoppingConfiguration;
import org.deeplearning4j.earlystopping.EarlyStoppingResult;
import org.deeplearning4j.earlystopping.trainer.EarlyStoppingTrainer;
import org.deeplearning4j.nn.multilayer.MultiLayerNetwork;
import org.deeplearning4j.ui.components.text.ComponentText;

import java.util.concurrent.Callable;

@AllArgsConstructor @NoArgsConstructor
public class MultiLayerNetworkTaskCreator implements TaskCreator{

    private ModelEvaluator modelEvaluator;

    @Override
    public Callable> create(
            Candidate candidate, DataProvider dataProvider,
            ScoreFunction scoreFunction,
            UICandidateStatusListener statusListener) {

        return new DL4JLearningTask<>(candidate,dataProvider,scoreFunction,modelEvaluator,statusListener);

    }


    private static class DL4JLearningTask implements Callable> {

        private Candidate candidate;
        private DataProvider dataProvider;
        private ScoreFunction scoreFunction;
        private ModelEvaluator modelEvaluator;

        private BaseUIStatusReportingListener dl4jListener;

        public DL4JLearningTask(Candidate candidate, DataProvider dataProvider, ScoreFunction scoreFunction, ModelEvaluator modelEvaluator, UICandidateStatusListener listener) {
            this.candidate = candidate;
            this.dataProvider = dataProvider;
            this.scoreFunction = scoreFunction;
            this.modelEvaluator = modelEvaluator;

            dl4jListener = new UIStatusReportingListener(listener);
        }


        @Override
        public OptimizationResult call() throws Exception {
            //Create network
            MultiLayerNetwork net = new MultiLayerNetwork(candidate.getValue().getMultiLayerConfiguration());
            net.init();
            net.setListeners(dl4jListener);

            //Early stopping or fixed number of epochs:
            DataSetIterator dataSetIterator = dataProvider.testData(candidate.getDataParameters());


            EarlyStoppingConfiguration esConfig = candidate.getValue().getEarlyStoppingConfiguration();
            EarlyStoppingResult esResult = null;
            if(esConfig != null){
                EarlyStoppingTrainer trainer = new EarlyStoppingTrainer(esConfig,net,dataSetIterator,dl4jListener);
                try{
                    esResult = trainer.fit();
                    net = esResult.getBestModel();  //Can return null if failed OR if
                } catch(Exception e){
                    dl4jListener.postReport(Status.Failed, null,
                            new ComponentText("Unexpected exception during model training\n", null),
                            new ComponentText(ExceptionUtils.getStackTrace(e), null));
                    throw e;
                }

                switch(esResult.getTerminationReason()){
                    case Error:
                        dl4jListener.postReport(Status.Failed, esResult);
                        break;
                    case IterationTerminationCondition:
                    case EpochTerminationCondition:
                        dl4jListener.postReport(Status.Complete, esResult);
                        break;
                }

            } else {
                //Fixed number of epochs
                int nEpochs = candidate.getValue().getNumEpochs();
                for( int i=0; i(candidate, net, score, candidate.getIndex(), additionalEvaluation);
        }
    }
}