There is a problem with the code in gnn_parse.py

[zj12346666]

attention = torch.sum(torch.softmax(energy, dim=-1), dim=-1)，The result is always 1