Is there a way to use a lstm model to make multistep forecast for multiple windows?
For example, I train the model to output 5 steps, is there a way that, if I want to predict 10 I could easily do it? Or is this intrinsic to the model architecture I there is no way to do it so that I can have more flexibility in predictions?
I searched solutions like seq2seq encoder decoder but could not find an answer.
Thank you
I tried simple lstm models but could only have a fixed output...