2d64684041ddaa57c8692d7ddc114a408fec3982,torch_geometric/nn/conv/sage_conv.py,SAGEConv,forward,#SAGEConv#Any#Any#,49

Before Change

        edge_index, _ = add_self_loops(edge_index, num_nodes=x.size(0))

        x = x.unsqueeze(-1) if x.dim() == 1 else x
        row, col = edge_index

        x = torch.matmul(x, self.weight)
        out = scatter_mean(x[col], row, dim=0, dim_size=x.size(0))

        if self.bias is not None:
            out = out + self.bias

        if self.normalize:
            out = F.normalize(out, p=2, dim=-1)

After Change

        x = x.unsqueeze(-1) if x.dim() == 1 else x
        x = torch.matmul(x, self.weight)

        return self.propagate(edge_index, x=x)

    def message(self, x_j):
        return x_j

Instances