I am using the evolveGCN as a reference to build a similar model on HeteroGeneous graphs. When I put the parameter weight=False, the GraphConv class doesn’t output the correct feature size or dimensions. When I change the weight=True it outputs the correct feature size. Please if anyone can help with what is going on in my code.
class EvolveGCNH(nn.Module):
def __init__(self, in_feats=166, n_hidden=76, num_layers=2, n_classes=2, classifier_hidden=510):
# default parameters follow the official config
super(EvolveGCNH, self).__init__()
self.num_layers = num_layers
self.pooling_layers = nn.ModuleList()
self.recurrent_layers = nn.ModuleList()
self.gnn_convs = nn.ModuleList()
self.gcn_weights_list = nn.ParameterList()
self.pooling_layers.append(TopK(in_feats, n_hidden))
# similar to EvolveGCNO
self.recurrent_layers.append(MatGRUCell(in_feats=in_feats, out_feats=n_hidden))
self.gcn_weights_list.append(Parameter(torch.Tensor(in_feats, n_hidden)))
# weight=False doesn't work and doesn't output n_hidden feature
self.gnn_convs.append(
HeteroGraphConv(
{
"follower": GraphConv(
in_feats=in_feats,
out_feats=n_hidden,
bias=True, activation=nn.RReLU(), weight=True,
),
"following": GraphConv(
in_feats=in_feats,
out_feats=n_hidden,
bias=True, activation=nn.RReLU(), weight=True,
),
},
aggregate="sum",
)
)
for _ in range(num_layers - 1):
self.pooling_layers.append(TopK(n_hidden, n_hidden))
self.recurrent_layers.append(MatGRUCell(in_feats=n_hidden, out_feats=n_hidden))
self.gcn_weights_list.append(Parameter(torch.Tensor(n_hidden, n_hidden)))
self.gnn_convs.append(
HeteroGraphConv(
{
"follower": GraphConv(
in_feats=n_hidden,
out_feats=n_hidden,
bias=False, activation=nn.RReLU(), weight=False,
),
"following": GraphConv(
in_feats=n_hidden,
out_feats=n_hidden,
bias=False, activation=nn.RReLU(), weight=False,
),
},
aggregate="sum",
)
)
self.mlp = nn.Sequential(nn.Linear(n_hidden, classifier_hidden),
nn.ReLU(),
nn.Linear(classifier_hidden, n_classes))
self.reset_parameters()
def reset_parameters(self):
for gcn_weight in self.gcn_weights_list:
init.xavier_uniform_(gcn_weight)
def forward(self, g_list):
feature_list = []
for g in g_list:
feature_list.append(g.ndata['feat'])
for i in range(self.num_layers):
W = self.gcn_weights_list[i]
for j, g in enumerate(g_list):
X_tilde = self.pooling_layers[i](feature_list[j])
#print(W.shape)
W = self.recurrent_layers[i](W, X_tilde)
#print(W.shape)
#print(self.gnn_convs[i])
result = self.gnn_convs[i](g,{'user':feature_list[j]}, {'weight': W})['user']
#print(result.shape)
feature_list[j] = result
return self.mlp(feature_list[-1])