I am trying to build a seq2seq model for time series forecasting where the input is a sequence with shape [batch_size,time_steps,no_of_features] and the output is a sequence of