Problem when applying convolutions on model.py

[giulliabraga]

Hello Yan! First of all, thank you for your paper and your code, they are both very thorough and neatly done, congratulations on your work! I am a student in a Computer Science Masters program, and as part of a coursework assignment I have been trying to reproduce your results. However, I have encoutered a few obstacles, for example the fact that I was not able to install the packages on the versions you specified. I am trying to adapt the code to comply with the versions I am using, but there is one specific problem I haven't been able to fix:

I believe there is a bug caused by model.py, on the function forward, during the convolutional layers step:

def forward(self, x_dict, edge_index_dict):
        
        x_dict["example"]  = self.post_transform(self.pretransform_exp(x_dict["example"]))
        x_dict['window'] = self.post_transform(self.pretransform_win(x_dict['window']))
        x_dict["example_y"] = self.pretransform_ey(x_dict["example"][:,-250:])
        
        for conv in self.convs:
            x_dict = conv(x_dict, edge_index_dict)
            x_dict = {key: self.leaklyrelu(x) for key, x in x_dict.items()}
        return self.lin(self.pool(x_dict, edge_index_dict))

From what I have tested, during the convolution process the key 'example' disappears from x_dict, and on my understanding this induces an error in csra.py, this one specifically:

ERROR
On csra.py, line 19, in forward:

src = x_dict['example']
KeyError: 'example'

I am pretty sure the key 'example' only disappears from x_dict after these convolutions, and then it induces this error. Do you have any insights that could help me fix this error? Thank you in advance!

[LiuXintongPKU]

Hi there! I'm encountering the same issue in my environment. Have you found a way to solve it? Any insights would be greatly appreciated!