using different batchsize causes a error if cg already runs with different batch size. Error is at score.

example.java

package com.tactico.tm.asyncRL;

import org.deeplearning4j.nn.api.OptimizationAlgorithm;
import org.deeplearning4j.nn.conf.ComputationGraphConfiguration;
import org.deeplearning4j.nn.conf.NeuralNetConfiguration;
import org.deeplearning4j.nn.conf.Updater;
import org.deeplearning4j.nn.conf.inputs.InputType;
import org.deeplearning4j.nn.conf.layers.GravesLSTM;
import org.deeplearning4j.nn.graph.ComputationGraph;
import org.deeplearning4j.nn.weights.WeightInit;
import org.deeplearning4j.optimize.listeners.ScoreIterationListener;
import org.nd4j.linalg.api.ndarray.INDArray;
import org.nd4j.linalg.dataset.MultiDataSet;
import org.nd4j.linalg.factory.Nd4j;


public class Bug {
	public static void main (String[] args){
		ComputationGraphConfiguration.GraphBuilder confBuilder = new NeuralNetConfiguration.Builder()
                .iterations(1)
                .optimizationAlgo(OptimizationAlgorithm.STOCHASTIC_GRADIENT_DESCENT)
                .learningRate(0.1)
                .updater(Updater.RMSPROP)
 
                .weightInit(WeightInit.NORMALIZED)
                .graphBuilder()
                .setInputTypes(InputType.recurrent(2))
                .addInputs("input")
                .addLayer("layer", new GravesLSTM.Builder()
                .nIn(2)
                .nOut(5)
                .activation("tanh")
                .build(), "input");
		confBuilder.setOutputs("layer");
		ComputationGraphConfiguration cgconf = confBuilder.pretrain(false).backprop(true).build();
		ComputationGraph cg = new ComputationGraph(cgconf);
		cg.init();
		cg.setListeners(new ScoreIterationListener(Constants.NEURAL_NET_ITERATION_LISTENER));
		
		INDArray input = Nd4j.rand(1,2,1);
		cg.rnnTimeStep(input);
		INDArray input_train = Nd4j.rand(4,2,8);
		INDArray labels = Nd4j.rand(4,5,8);
		INDArray input_mask = Nd4j.rand(4,8);
		INDArray labels_mask = Nd4j.rand(4,8);
		MultiDataSet dataSet = new MultiDataSet(new INDArray[]{input_train}, new INDArray[]{labels}, new INDArray[]{input_mask}, new INDArray[]{labels_mask});
		double score = cg.score(dataSet,true);

		
	}

}