Graph convolutional network with various sizes of input

Hi there,

I’m wondering whether it is possible to forward various size of input to the graph convolutional networks (GCNs) or not.
For example, I have 3 different inputs as below, (please ignore the values of tensors(0))

in_a = torch.zeros(50, 100)
in_b = torch.zeros(80, 100)
in_c = torch.zeros(150, 100)

Will GCNs which have 100x100 nodes be trained well with those inputs?
Please reply any comments whenever you need further explanations.