97d7f88df32ae58875d7534c7c517905a3eb1341,deepchem/models/torch_models/gat.py,GAT,forward,#GAT#Any#,113

Before Change

    node_feat = self.embedding(node_feat)

    // convolutional layer
    for conv in self.conv_layers:
      node_feat = conv(node_feat, edge_index)

    // pooling
    graph_feat = self.pooling(node_feat, data.batch)
    graph_feat = F.leaky_relu(self.fc(graph_feat))
    out = self.out(graph_feat)

    if self.mode == "regression":

After Change

            This is only returned when self.mode = "classification", the output consists of the
            logits for classes before softmax.
        
    node_feats = g.ndata[self.nfeat_name]
    out = self.model(g, node_feats)

    if self.mode == "classification":
      if self.n_tasks == 1:
        logits = out.view(-1, self.n_classes)
        softmax_dim = 1
      else:
        logits = out.view(-1, self.n_tasks, self.n_classes)
        softmax_dim = 2

Instances