def forward(self, g, x):
g = g.local_var()
g.nodes['n'].data['h'] = x
for i in g.etypes:
one_hot = (th.ones((g.number_of_edges(etype=i)), dtype=th.long) *
self.rel_dict[i])
if self.use_cuda:
g.edges[i].data['type'] = one_hot.cuda()
else:
g.edges[i].data['type'] = one_hot
if self.self_loop:
loop_message = utils.matmul_maybe_select(x, self.loop_weight)
# message passing
for i in g.etypes:
g.update_all(self.message_func,
fn.sum(msg='msg', out='h'),
etype=i)
# apply bias and activation
node_repr = g.nodes['n'].data['h']
if self.bias:
node_repr = node_repr + self.h_bias
if self.self_loop:
node_repr = node_repr + loop_message
if self.activation:
node_repr = self.activation(node_repr)
node_repr = self.dropout(node_repr)
return node_repr
def forward(self, g, x, etypes, norm=None):
g = g.local_var()
g.ndata['h'] = x
g.edata['type'] = etypes
if norm is not None:
g.edata['norm'] = norm
if self.self_loop:
loop_message = utils.matmul_maybe_select(x, self.loop_weight)
# message passing
g.update_all(self.message_func, self.reduce_func(msg='msg', out='h'))
# apply bias and activation
node_repr = g.ndata['h']
if self.bias:
node_repr = node_repr + self.h_bias
if self.self_loop:
node_repr = node_repr + loop_message
if self.activation:
node_repr = self.activation(node_repr)
node_repr = self.dropout(node_repr)
return node_repr
def forward(self, g):
assert g.is_homograph(), \
"not a homograph; convert it with to_homo and pass in the edge type as argument"
with g.local_scope():
if self.self_loop:
loop_message = dgl_utils.matmul_maybe_select(
g.ndata['h'], self.loop_weight)
# message passing
g.update_all(self.message_func, fn.sum(msg='msg', out='h'))
# apply bias and activation
node_repr = g.ndata['h']
if self.bias:
node_repr = node_repr + self.h_bias
if self.self_loop:
node_repr = node_repr + loop_message
if self.activation:
node_repr = self.activation(node_repr)
node_repr = self.dropout(node_repr)
return node_repr