# LSTM Sec2Sec forecasting

**URL:** <https://community.konduit.ai/t/lstm-sec2sec-forecasting/1156>\
**Category:** DL4J\
**Created:** [February 5, 2021, 7:50pm UTC](https://community.konduit.ai/t/lstm-sec2sec-forecasting/1156 "2021-02-05T19:50:06Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![vurosevic](https://yyz1.discourse-cdn.com/flex035/user_avatar/community.konduit.ai/vurosevic/32/449_2.png) [@vurosevic](https://community.konduit.ai/u/vurosevic)\
**Post date:** [February 5, 2021, 7:50pm UTC](https://community.konduit.ai/t/lstm-sec2sec-forecasting/1156/1 "2021-02-05T19:50:06Z")

</div>

Hi,  
I try to make forecasting for one TS input.  
I need to have one sequence (fixed length) as input and one sequence as output (different fixed length as input). Can I do that with this configuration, or I should be use encoder/decoder system?

```
public MultiLayerConfiguration getLstmNetworkConfiguration() {
    NeuralNetConfiguration.Builder builder = new NeuralNetConfiguration.Builder();
    builder.seed(123);
    builder.biasInit(0);
    builder.miniBatch(true);
    builder.updater(updater);
    builder.weightInit(WeightInit.XAVIER);
    builder.optimizationAlgo(OptimizationAlgorithm.STOCHASTIC_GRADIENT_DESCENT);
    builder.l2(0.0001);

    NeuralNetConfiguration.ListBuilder listBuilder = builder.list();

    for (int i = 0; i < HIDDEN_LAYER_CONT; i++) {
        LSTM.Builder hiddenLayerBuilder = new LSTM.Builder();
        hiddenLayerBuilder.nIn(i == 0 ? experimentDto.getNumInputs() : HIDDEN_LAYER_WIDTH);
        hiddenLayerBuilder.nOut(HIDDEN_LAYER_WIDTH);
        hiddenLayerBuilder.activation(Activation.TANH);
        hiddenLayerBuilder.weightInit(WeightInit.XAVIER);
        listBuilder.layer(i, hiddenLayerBuilder.build());
    }

    listBuilder.layer(HIDDEN_LAYER_CONT, new DenseLayer.Builder()
            .nIn(HIDDEN_LAYER_WIDTH)
            .nOut(DENSE_LAYER_WIDTH)
            .activation(Activation.RELU)
            .build());
    listBuilder.layer(HIDDEN_LAYER_CONT+1, new RnnOutputLayer.Builder()
            .nIn(DENSE_LAYER_WIDTH)
            .nOut(experimentDto.getNumOutputs())
            .activation(Activation.IDENTITY)
            .lossFunction(LossFunctions.LossFunction.MSE)
            .build());

    return listBuilder.build();
}

```

Please, where can I find example for encoder/decoder system?

Thanks,  
Vladimir