idea-iitd / graphgen Goto Github PK

Hi! I'm checking the code now, and I found a small problem :

if 'inactive' in args.graph_type:
        args.current_dataset_path = base_path + 'inactive_graphs/'
        **include_file = base_path + 'actives.txt'**
        produce_complementary = True
        args.min_dfscode_path = base_path + 'inactive_min_dfscodes/'
        min_dfscode_tensor_path = base_path + 'inactive_min_dfscode_tensors/'

    elif 'active' in args.graph_type:
        args.current_dataset_path = base_path + 'active_graphs/'
        **include_file = base_path + 'actives.txt'**
        produce_complementary = False
        args.min_dfscode_path = base_path + 'active_min_dfscodes/'
        min_dfscode_tensor_path = base_path + 'active_min_dfscode_tensors/'

Schouldn't in if 'inactive' in args.graph_type: block be include_file = base_path + **'in_actives.txt'** instead of **include_file = base_path + 'actives.txt'** ?

Thank you so much for this git! However, I'm experiencing a lot of errors. Is there someone who I can get in contact with that has run/edited this library on/for windows? My hunch is that this is made in Linux.

Cheers, Oma.

Hi thanks for the nice work - would you mind adding a license such as Apache or MIT for this project? Thanks

Hi! I've found some issues with requirements.txt file. So, first of all I had a problem with tqdm - the library is not in requirements, so for me, I had this error:

sh-4.2$ python3 main.py
Traceback (most recent call last):
  File "main.py", line 8, in <module>
    from datasets.process_dataset import create_graphs
  File "/home/ec2-user/SageMaker/graphgen/datasets/process_dataset.py", line 11, in <module>
    from tqdm.auto import tqdm
ModuleNotFoundError: No module named 'tqdm'

Solved it by adding to requitements.

Another error, sklearn=1.1.0 seems to be not found - I had this error:

ERROR: Could not find a version that satisfies the requirement sklearn==1.1.0 (from -r requirements.txt (line 3)) (from versions: 0.0)
ERROR: No matching distribution found for sklearn==1.1.0 (from -r requirements.txt (line 3))

Solved by just deleting it from requirements.

Hi! I'm trying to make the code up and running, but unfortunately I have this kind of error while ./build.sh execution :

metrics/isomorph.cpp: In function ‘bool is_subgraph_isomorphic(Graph, Graph)’:
metrics/isomorph.cpp:54:16: error: ‘vf2_subgraph_mono’ is not a member of ‘boost’
     bool res = boost::vf2_subgraph_mono(smallGraph, bigGraph, callback, boost::vertex_order_by_mult(smallGraph),
                ^
metrics/unique.cpp: In function ‘bool is_subgraph_isomorphic(Graph, Graph)’:
metrics/unique.cpp:52:16: error: ‘vf2_subgraph_mono’ is not a member of ‘boost’
     bool res = boost::vf2_subgraph_mono(smallGraph, bigGraph, callback, boost::vertex_order_by_mult(smallGraph),
                ^

Do you have any ideas why it could happen? Maybe a specific version of boost is needed to use?

Hi! I have a problem running the training with GPU. I've installed cuda 10.0, which is compatible with my hardware. Anyways to troubleshoot?
I'll appreciate your help.

Great work!
Thank you for uploading the code.
I replicated your fantastic code, and everything works fine except for the DGMG part.
Somehow it works fine when we use CPU mode.
But in GPU mode, we get the following errors.

~/path/to/graphgen/baselines/dgmg/model.py in forward(self, g_list)
78 bg = dgl.batch(g_list)
79 print("debug:", bg, "ndata:",bg.ndata)
---> 80 bhv = bg.ndata['hv']
81 bg.ndata['hg'] = self.node_gating(bhv) * self.node_to_graph(bhv)
82
~/anaconda3/lib/python3.7/site-packages/dgl/view.py in getitem(self, key)
65 return ret
66 else:
---> 67 return self._graph._get_n_repr(self._ntid, self._nodes)[key]
68
69 def setitem(self, key, val):
~/anaconda3/lib/python3.7/site-packages/dgl/frame.py in getitem(self, name)
493 Column data.
494 """
--> 495 return self._columns[name].data
496
497 def setitem(self, name, data):
KeyError: 'hv'

I have been debugging this for ages, but in vain.
Do you have any suggestions on what might be the problem here?
Thank you in advance for your reply.

Hi, I am planning to move all the real datasets from graphRNN to your implementation, and there is an option (degree/cc) I am a little confused, can you point me out what it is?