Convs layer supports both edge_attr/edge_weights and negative edges

[dxyzx0]

GCNconv only supports positive edges and SignedConv only supports edge_index without edge_weight. Is there a conv layer that supports both edge_attr/edge_weights and negative edges? Because my Adjacency matrix contains negative values.

[rusty1s]

Not that I am aware of, but you can quite easily write something like this by yourself:

class MyConv(MessagePassing):
    def __init__(self, in_channels, out_channels, edge_dim):
        super().__init__()
        self.lin1 = Linear(in_channels + edge_dim, out_channels, bias=False)
        self.lin2 = Linear(in_channels, out_channels)
    
    def forward(self, x, edge_index, edge_weight, edge_attr):
        out = self.propagate(edge_index, x=x=, edge_weight=edge_weight, edge_attr=edge_attr)
        return self.lin2(x) + out        

    def message(self, x_j, edge_weight, edge_attr):
        out = self.lin1(torch.cat([x_j, edge_attr], dim=-1))
        return edge_weight.view(-1, 1) * out