What does the argument parameter "args.target" sets for multivariate time series forecasting?

[rohaan2614]

Thank you for sharing the code, really helpful. However, I noticed that for multi-feature forecasting, we still need to provide a target feature, which is by default "OT". My question is does the network only why is that required? Will not the network be trained to forecast all the features?

Also, how are the multiple features passed into the network? So is it done by concatenating the multiple features from the conv layer and use that as input to the network, or something else?