Hello,
If I want to replicate the experiment of the following one,

Do I need to change the default train setting in read.me? I try the default setting (super_rate=0.5) with meddg dataset, however, the final result is less than yours (e.g., EG=37.90).