PyTorch Implement for Path Attention Graph Network

error:

self._PropertyMap__map[k] = v Boost.Python.ArgumentError: Python argument types in EdgePropertyMap<double>.__setitem__(EdgePropertyMap<double>, tuple, numpy.float32) did not match C++ signature: __setitem__(graph_tool::PythonPropertyMap<boost::checked_vector_property_map<double, boost::adj_edge_index_property_map<unsigned long> > > {lvalue}, graph_tool::PythonEdge<boost::filt_graph<boost::undirected_adaptor<boost::adj_list<unsigned long> >, graph_tool::detail::MaskFilter<boost::unchecked_vector_property_map<unsigned char, boost::adj_edge_index_property_map<unsigned long> > >, graph_tool::detail::MaskFilter<boost::unchecked_vector_property_map<unsigned char, boost::typed_identity_property_map<unsigned long> > > > const>, double) During handling of the above exception, another exception occurred:

self._PropertyMap__map[k] = self._PropertyMap__convert(v) Boost.Python.ArgumentError: Python argument types in EdgePropertyMap<double>.__setitem__(EdgePropertyMap<double>, tuple, float) did not match C++ signature: __setitem__(graph_tool::PythonPropertyMap<boost::checked_vector_property_map<double, boost::adj_edge_index_property_map<unsigned long> > > {lvalue}, graph_tool::PythonEdge<boost::filt_graph<boost::undirected_adaptor<boost::adj_list<unsigned long> >, graph_tool::detail::MaskFilter<boost::unchecked_vector_property_map<unsigned char, boost::adj_edge_index_property_map<unsigned long> > >, graph_tool::detail::MaskFilter<boost::unchecked_vector_property_map<unsigned char, boost::typed_identity_property_map<unsigned long> > > > const>, double)

@ihollywhy Could you help me with this? thank you so much! I'm not so sure if it's a 'version' problem again but I do believe if I can have some information about the exact packages' version you used would be helpful Thanks again

Originally posted by @scarletocean in https://github.com/ihollywhy/SPAGAN/issues/4#issuecomment-754764757

As I understand, if the SPAGAN has two layers and the max value of c is three for the first layer and two for the last layer, the size of receptive field of this SPAGAN is 5=3+2. While, according to your paper Section 5.3, 3-layer GAT has the same size of receptive field with SPAGAN. It seems the size of receptive field of SPAGAN is 3.

So, how to calculate the size of receptive field? why it is 3 rather than 5?

Thanks for reading. GAT stacks layer for larger receptive field. But SPAGAN obviously does not need to do that. But why does the SPAGAN still stack layers. I would appreciate it if someone could reply to me.

Traceback (most recent call last): e_rowsum = spmm(edge, edge_e, N, torch.ones(size=(N,1)).cuda()) # e_rowsum: N x 1 TypeError: spmm() missing 1 required positional argument: 'matrix'

Hi, @ihollywhy ,

I see you mentioned in the paper that on Citeseer dataset we needed to adjust the learning rate and the value of weight attenuation, which had reached a relatively better result. However, according to your parameter setting, the accuracy rate was only 18%,……，Could you tell me how to solve it?