The cora dataset cannot be loaded

[J-zin]

Hi,
I try to run the code with

python refine.py --dataset cora \
    --solve-proxy-epochs 200 \
    --solve-proxy-eval-every 20 \
    --solve-proxy-node-lr 1e-2 \
    --solve-proxy-edge-lr 2e-3 \
    --refine-node-lr 1e-3 \
    --refine-edge-lr 2e-4 \
    SAGE

but some errors are thrown
Traceback (most recent call last):
File "refine.py", line 126, in
refine(args)
File "refine.py", line 38, in refine
train_loader, val_loader, test_loader = prepare_dataloaders(args)
File "/apdcephfs/share_1364275/jzinou/github-repo/SPN/dataset.py", line 126, in prepare_dataloaders
train_loader, val_loader, test_loader = map(precompute_edge_label_and_reverse, (train_loader, val_loader, test_loader))
File "/apdcephfs/share_1364275/jzinou/github-repo/SPN/dataset.py", line 17, in precompute_edge_label_and_reverse
for data in dataset:
File "/root/anaconda3/envs/shapley/lib/python3.7/site-packages/torch_geometric/data/dataset.py", line 197, in getitem
data = self.get(self.indices()[idx])
File "/root/anaconda3/envs/shapley/lib/python3.7/site-packages/torch_geometric/data/in_memory_dataset.py", line 83, in get
return copy.copy(self.data)
File "/root/anaconda3/envs/shapley/lib/python3.7/copy.py", line 106, in copy
return _reconstruct(x, None, *rv)
File "/root/anaconda3/envs/shapley/lib/python3.7/copy.py", line 275, in _reconstruct
y = func(*args)
TypeError: call() missing 1 required positional argument: 'base_cls'

It seems that the provided cora dataset has some mistakes. Similar errors are found when I simply replace cora with citeseer, pubmed, but the dblp works well.

[mnqu]

Hi Zijing,

Thanks for the question. Our codes are not compatible with the latest torch-geometric package (2.0.X). You may consider using an older version (1.X.X) instead.

[J-zin]

Thanks for your super quick reply, torch-geometric==1.7.2 works well for me 🤗