missing code implementation of "accessing to future predicted values of the exogenous variables"

[chadh0829]

I was intrigued by the claim that the model achieves the best performance when it has access to future predicted values of the exogenous variables in addition to historical information of exogenous variables. (this is explanation is in section H.1 and figure 11 (f) )
However, upon reviewing the code provided on your GitHub repository, I could not locate the implementation where future exogenous variables are incorporated into the data loader.

As I am attempting to replicate your results and further understand the impact of including future exogenous variables, I would greatly appreciate it if you could provide the relevant code or clarify the implementation details regarding this aspect.

Thank you