Question about how to test the learned policy

[xwbxxx]

Thank you for open-sourcing your work.
Causal discovery with interventions is a new field for me, so I don't quite understand the test procedure. Here I have 2 questions:

1. During test time, the true structure is unknown. So how will the observation vectors update variable values after doing interventions? Does the training process shown below still hold in the test time?
   

2. I would be very appreciated if you could point me to the code file and the specific lines where I can see test procedure.