Need help with multivariate multi-step timeseries prediction using RNN

My network is as follows:

  1. LSTM Encoder (Take a sequence of n lags and output a hidden end vector)
  2. Fully Connected Dense Layer (To create a context vector)
  3. Decoder network (LSTM layer that process context vector for m lags)

I am having trouble with the decoder part. Basically, I need to repeat my context vector and then apply LSTM recursively. Any suggestion on how to do that would be helpful? For multi-step output is there a better method?