How to use add/mean aggregation in GENConv

[riskiem]

I am trying to use GENConv and I am not getting good results with the default softmax. I want to try out add or mean. When I change my code it throws error.

class DeeperGCN(torch.nn.Module):
    def __init__(self, hidden_channels, num_layers):
        super(DeeperGCN, self).__init__()

        self.node_encoder = Linear(data.x.size(-1), hidden_channels)
        self.edge_encoder = Linear(data.edge_attr.size(-1), hidden_channels)

        self.layers = torch.nn.ModuleList()
        for i in range(1, num_layers + 1):
         
            conv = GENConv(hidden_channels, hidden_channels, aggr='add',
                            num_layers=2, norm='layer')
            norm = LayerNorm(hidden_channels, elementwise_affine=True)
            act = ReLU(inplace=True)

            layer = DeepGCNLayer(conv, norm, act, block='res+', dropout=0.5,
                                 ckpt_grad=i % 3)
            self.layers.append(layer)

        self.lin = Linear(hidden_channels, 2)

    def forward(self, x, edge_index, edge_attr):
        x = self.node_encoder(x)
        edge_attr = self.edge_encoder(edge_attr)

        x = self.layers[0].conv(x, edge_index, edge_attr)

        for layer in self.layers[1:]:
            x = layer(x, edge_index, edge_attr)

        x = self.layers[0].act(self.layers[0].norm(x))
        

        return self.lin(x)

I get the following error:

---

AssertionError Traceback (most recent call last)
in ()
----> 1 model = DeeperGCN(hidden_channels=64, num_layers=10)
2 #optimizer = torch.optim.Adam(model.parameters(), lr=0.1)
3 criterion = torch.nn.CrossEntropyLoss()
4 #optimizer = torch.optim.SGD(model.parameters(), lr=0.1, momentum=0.9)
5 optimizer = torch.optim.Adam(model.parameters(), lr=0.1)

1 frames
/usr/local/lib/python3.7/dist-packages/torch_geometric/nn/conv/gen_conv.py in init(self, in_channels, out_channels, aggr, t, learn_t, p, learn_p, msg_norm, learn_msg_scale, norm, num_layers, eps, **kwargs)
101 self.eps = eps
102
--> 103 assert aggr in ['softmax', 'softmax_sg', 'power']
104
105 channels = [in_channels]

AssertionError:

Can anyone help me fix this?

[rusty1s]

That's indeed a mistake. I fixed it in master.

[riskiem]

Thank you! I downloaded the latest version. I still can't run this code without error.

[rusty1s]

This shouldn't be the case. Note that you need to install PyG from master:

pip install git+https://github.com/rusty1s/pytorch_geometric.git