def build_output_layer(self):
print('Building an OUTPUT layer of {}x{}'.format(
self.hidden_dimensions[-2], self.hidden_dimensions[-1]))
return RelGraphConv(self.hidden_dimensions[-2],
self.hidden_dimensions[-1],
self.num_rels,
regularizer='basis',
dropout=self.feat_drop,
num_bases=self.num_bases,
activation=torch.tanh)
def build_input_layer(self):
print('Building an INPUT layer of {}x{}'.format(
self.in_dim, self.hidden_dimensions[0]))
return RelGraphConv(self.in_dim,
self.hidden_dimensions[0],
self.num_rels,
regularizer='basis',
dropout=self.feat_drop,
num_bases=self.num_bases,
activation=F.leaky_relu)
def build_hidden_layer(self, i):
print('Building an HIDDEN layer of {}x{}'.format(
self.hidden_dimensions[i], self.hidden_dimensions[i + 1]))
return RelGraphConv(self.hidden_dimensions[i],
self.hidden_dimensions[i + 1],
self.num_rels,
regularizer='basis',
dropout=self.feat_drop,
num_bases=self.num_bases,
activation=F.leaky_relu)
def build_gnn_output_layer(self):
print(
f'Building an OUTPUT layer for RGCN DeulingDQN of {self.gnn_hidden_dimensions[-2]}x{self.gnn_hidden_dimensions[-1]} (activation:{self.gnn_activation[-1]})'
)
return RelGraphConv(self.gnn_hidden_dimensions[-2],
self.gnn_hidden_dimensions[-1],
self.num_rels,
regularizer='basis',
dropout=self.feat_drop,
num_bases=self.num_bases,
activation=self.gnn_activation[-1])