Hi DropEdge Team,

I am running experiments on 8-layer GCN (using DropEdge) in the semi-supervised setting. I used the default hyper-parameters as 2-layer GCN and changed --nbaseblocklayer 0 to --nbaseblocklayer 6 in script/semi-supervised/cora_GCN.sh. On Cora dataset, the 2-layer GCN performance is 82.8% while 8-lay GCN is only 16%, not 75.80%. Could you please tell me how I can reproduce the results shown in readme?

Thank you.