def forward(self, data):
x = forward_op(data.node_feature, self.dropouts1)
x = forward_op(x, self.relus1)
x = self.conv1(x, data.edge_index)
x = forward_op(x, self.dropouts2)
x = forward_op(x, self.relus2)
x = self.conv2(x, data.edge_index)
return x
def forward(self, data):
x = forward_op(data.node_feature, F.dropout, p=self.dropout, training=self.training)
x = forward_op(x, F.relu)
x = self.conv1(x, data.edge_index)
x = forward_op(x, F.dropout, p=self.dropout, training=self.training)
x = forward_op(x, F.relu)
x = self.conv2(x, data.edge_index)
return x
def forward(self, node_feature, edge_index):
x = node_feature
x = self.convs1(x, edge_index)
x = forward_op(x, self.bns1)
x = forward_op(x, self.relus1)
x = self.convs2(x, edge_index)
x = forward_op(x, self.bns2)
x = forward_op(x, self.relus2)
x = forward_op(x, self.post_mps)
return x
def forward(self, data):
x = forward_op(data.node_feature, self.dropout1)
x = forward_op(x, self.relu1)
x = self.conv1(x, data.edge_index)
x = forward_op(x, self.dropout2)
x = forward_op(x, self.relu2)
x = self.conv2(x, data.edge_index)
pred = {}
for message_type in data.edge_label_index:
nodes_first = torch.index_select(
x['n1'], 0, data.edge_label_index[message_type][0, :].long())
nodes_second = torch.index_select(
x['n1'], 0, data.edge_label_index[message_type][1, :].long())
pred[message_type] = torch.sum(nodes_first * nodes_second, dim=-1)
return pred
def forward(self, data):
x = data.node_feature
edge_index = data.edge_index
x = self.convs1(x, edge_index)
x = forward_op(x, self.bns1)
x = forward_op(x, self.relus1)
x = self.convs2(x, edge_index)
x = forward_op(x, self.bns2)
pred = {}
for message_type in data.edge_label_index:
nodes_first = torch.index_select(
x['n1'], 0, data.edge_label_index[message_type][0, :].long())
nodes_second = torch.index_select(
x['n1'], 0, data.edge_label_index[message_type][1, :].long())
pred[message_type] = torch.sum(nodes_first * nodes_second, dim=-1)
return pred
def test_hetero_gnn_forward(self):
xs = {}
layers = nn.ModuleDict()
emb_dim = 5
feat_dim = 10
num_samples = 8
keys = ['a', 'b', 'c']
for key in keys:
layers[key] = nn.Linear(feat_dim, emb_dim)
xs[key] = torch.ones(num_samples, feat_dim)
ys = forward_op(xs, layers)
for key in keys:
self.assertEqual(ys[key].shape[0], num_samples)
self.assertEqual(ys[key].shape[1], emb_dim)
