egnn_pytorch_geometric.py seem some error #47
Description
Hi,
In egnn_pytorch_geometric.py line 426 - 429 seem some error
layer[0] is EGNN_Sparse and layer[-1] is GlobalLinearAttention,
so the true code should be
x_attn = layer[-1](x[:, self.pos_dim:], global_tokens) x = torch.cat( (x[:, :self.pos_dim], x_attn), dim=-1) x = layer[0](x, edge_index, edge_attr, batch=batch, size=bsize)