how to predict the next n steps not one？

[guoyang007]

the tutorial's output is the next one value,but how to use the same time series data to predict the next n steps?
how to design the model to fit it?i have know the 'TimeDistributed()'can make the 3D output,but the output shape is the same as input.shape ,i can only change the output_dim ,so the problem is how to change the time_step params in the output?