Can't allocate memory error on running train_test_split_edges.py

[adg-git]

I am getting the following error
RuntimeError: [enforce fail at CPUAllocator.cpp:65] . DefaultCPUAllocator: can't allocate memory: you tried to allocate 141957139984 bytes. Error code 12 (Cannot allocate memory)

for line
https://github.com/rusty1s/pytorch_geometric/blob/master/torch_geometric/utils/train_test_split_edges.py#L73

I have 380K nodes in my graph.

Has anybody experienced similar issues? Any suggestions on how to resolve it would be greatly appreciated.

[rusty1s]

The current train_test_split_edges method does not scale well to large graphs, since it will create a dense BoolTensor for negative sampling. An alternative is to create negative samples via torch.utils.negative_sampling(method='sparse').