In class CompGCNConv and CompGCNConvBasis, it seems that w_loop, w_in and w_out are not used in the project. The 'out' (obtained by in_res, loop_res and out_res), and the bias do not match as the dimension of in_res, loop_res, out_res is num_entity × in_channels, while the dimension of bias is out_channel.
In class CompGCNConv and CompGCNConvBasis, it seems that w_loop, w_in and w_out are not used in the project. The 'out' (obtained by in_res, loop_res and out_res), and the bias do not match as the dimension of in_res, loop_res, out_res is num_entity × in_channels, while the dimension of bias is out_channel.