Number of Layers in DGL example (GAT)

[data-weirdo]

The code below is the GAT example from homepage.

Isn’t the first layer counted as layer?
For example, if I have 4 layer, according to the implementation, I get 5 layers instead.
Can anybody explain?

class GAT(nn.Module):
    def __init__(self,
                 g,
                 num_layers,
                 in_dim,
                 num_hidden,
                 num_classes,
                 heads,
                 activation,
                 feat_drop,
                 attn_drop,
                 negative_slope,
                 residual):
        super(GAT, self).__init__()
        self.g = g
        self.num_layers = num_layers
        self.gat_layers = nn.ModuleList()
        self.activation = activation
        # input projection (no residual)
        self.gat_layers.append(GATConv(
            in_dim, num_hidden, heads[0],
            feat_drop, attn_drop, negative_slope, False, self.activation))
        # hidden layers
        for l in range(1, num_layers):
            # due to multi-head, the in_dim = num_hidden * num_heads
            self.gat_layers.append(GATConv(
                num_hidden * heads[l-1], num_hidden, heads[l],
                feat_drop, attn_drop, negative_slope, residual, self.activation))
        # output projection
        self.gat_layers.append(GATConv(
            num_hidden * heads[-2], num_classes, heads[-1],
            feat_drop, attn_drop, negative_slope, residual, None))

    def forward(self, inputs):
        h = inputs
        for l in range(self.num_layers):
            h = self.gat_layers[l](self.g, h).flatten(1)
        # output projection
        logits = self.gat_layers[-1](self.g, h).mean(1)
        return logits

[minjie]

This is indeed a bit confusing. Sorry for that! You are right that if you set num_layers=4, it will contain one input->hidden, three hidden->hidden and one hidden->output layer (a total of 5). We will change the implementation so that num_layers means the actual number of layers in them model. Thanks for the report.