Transformer
Complete graph with learned attention weights
GNN
Sparse graph with radius cutoff
Cutoff Radius
2.8Å