Forecasted Values with a lag of two days

Question

I am using a CNN-LSTM autoencoder + CNN-LSTM forecasting model where the autoencoder is used for compressing my 10 features which are the macroeconomic indicators into a 4 dimen