model_name: APPNP 
k_hop: 10
alpha: 0.1
num_layer2: 1
learning_rate: 0.01
dropout: 0.5
hidden_size: 64
weight_decay: 0.0005