import hangs on LinkPropPredDataset

[moschles]

import torch
try:
    print('ogb')
    import ogb    
except Exception :
    print("import fail ogb")
else:
    print("ogb. okay")
print( dir(ogb) )

try:
    print('linkproppred')
    from ogb.linkproppred.dataset import LinkPropPredDataset as LPPD
except Exception :
    print("import fail linkproppred")
else:
    print("linkproppred. okay")

.

ogb
ogb. okay
['__builtins__', '__cached__', '__doc__', '__file__', '__loader__', '__name__', '__package__', '__path__', '__spec__', '__version__', 'version']
linkproppred

No exception is captured. Python just goes to infinite hang here.

Python 3.8.13
ogb-1.3.4-pyhd8ed1ab_0
pytorch-1.6.0-py3.8_cuda102_cudnn7_0
torch-geometric-2.1.0.post1-pypi_0
numpy-1.21.5-py38h7a0a035_3
pandas-1.4.4-py38hd77b12b_0
urllib3-1.26.11-pyhd8ed1ab_0
scikit-learn-1.1.1-py38h9235413_0
outdated-0.2.1-pyhd8ed1ab_0


Device 0: "NVIDIA GeForce GTX 1070"
  CUDA Driver Version / Runtime Version          11.7 / 10.2
  CUDA Capability Major/Minor version number:    6.1
CUDA Driver = CUDART, CUDA Driver Version = 11.7, CUDA Runtime Version = 10.2, NumDevs = 1 

[moschles]

This code also hangs.
.

try:
    print('linkproppred')
    from ogb.linkproppred import PygLinkPropPredDataset as PLPPD
except Exception :
    print("import fail linkproppred")
else:
    print("linkproppred. okay")

[weihua916]

Please update to 1.3.5

[weihua916]

Does importing torch and torch_geometrichang?

[rusty1s]

Should be fixed in master. Can we release a new version?

[weihua916]

Ah got it, thanks! Just updated to 1.3.5

[moschles]

@weihua916

import torch
print( "torch ",  torch.__version__ ) 
import torch_geometric
print( "torch_geometric ", torch_geometric.__version__)
import ogb
print( "ogb ", ogb.__version__)

--
torch 1.12.0
torch_geometric 2.1.0
ogb 1.3.5

[moschles]

@weihua916

I have solved this problem by copying files dataset_pyg.py and master.csv to my local machine, and then performing,

from dataset_pyg import PygLinkPropPredDataset as PLPPD

Everything else is working. Presumably this is a problem related to how python is accessing subfolders and whether the import refers to a class name or a file name.