GINEConv assumes nn is subscriptable

[oiao]

🐛 Describe the bug

There is an issue with the GINEConv class when providing the edge_dim parameter.
Line

pytorch_geometric/torch_geometric/nn/conv/gin_conv.py

Line 144 in e805f0c

if hasattr(self.nn[0], 'in_features'):

assumes that the passed nn is subscriptable, which is not the case when it is directly derived from nn.Module:

from torch import nn
from torch_geometric import nn as gnn

class Net(nn.Module):
    def __init__(self):
        super().__init__()
        self.lin = nn.Linear(1, 2)
    def forward(self, x):
        return self.lin(x)

gnn.GINEConv(Net(), edge_dim=3)

File torch_geometric/nn/conv/gin_conv.py:144, in GINEConv.__init__(self, nn, eps, train_eps, edge_dim, **kwargs)
    [142]     self.register_buffer('eps', torch.Tensor([eps]))
    [143] if edge_dim is not None:
--> [144]     if hasattr(self.nn[0], 'in_features'):
    [145]         in_channels = self.nn[0].in_features
    [146]     else:

TypeError: 'Net' object is not subscriptable

Environment

• PyG version: 2.0.4
• PyTorch version: 1.12
• OS: Ubuntu
• Python version: 3.8
• CUDA/cuDNN version: 11.3
• How you installed PyTorch and PyG (conda, pip, source): pip

[rusty1s]

Thanks for reporting. Will look into it.

[Padarn]

I had some spare time, so raised a PR to address this.