How to predict multi time-steps forecasting using lstm layer and "predict"?

The chickenpox example of lstm forecasting accepts only a single input array. It is suggested to use the function "predict" for the multi-input data. However, by using "predict", how can one loop over to forecast multi time-step forecasting? "predict" is used only to predict a one-time step ahead. I see the function "closeloop" was used for this purpose but not with "predict". How can I use "closeloop" to forecast multi time-steps in an example with "predict". My neural network is lstm.

Thanks.

[Gordon]

I too am wondering how to do this.