An error message

[zixiaoshuixin]

Hi,
I got an error message when I reproduced your project named "GNNStealing" and could you please tell how to solve it. Thank you so much!

when I executed "python train_target_model.py --dataset citeseer_full --target-model gat --num-hidden 256 --gpu 2", the error occurred. It seems that the "dgl.data.utils.split_dataset(..)" can't accept a parameter whose type is "DGLHeteroGraph" which has no length, so the error message is "TypeError: object of type 'DGLHeteroGraph' has no len()"

The error message in console is as follows：

[20:17:18] /opt/dgl/src/runtime/tensordispatch.cc:43: TensorDispatcher: dlopen failed: /etc/anaconda3/envs/gnn_model_stealing/lib/python3.9/site-packages/dgl/tensoradapter/pytorch/libtensoradapter_pytorch_1.10.2.so: cannot open shared object file: No such file or directory

Traceback (most recent call last):

File "/home/gpu/sunqin/GNNStealing-master/GNNStealing-master/code/train_target_model.py", line 55, in

train_g, val_g, test_g = split_graph(g, args.dataset, frac_list=[0.6, 0.2, 0.2])

File "/home/gpu/sunqin/GNNStealing-master/GNNStealing-master/code/src/utils.py", line 323, in split_graph

train_subset, val_subset, test_subset = dgl.data.utils.split_dataset(

File "/etc/anaconda3/envs/sunqin/lib/python3.9/site-packages/dgl/data/utils.py", line 76, in split_dataset

num_data = len(dataset)

TypeError: object of type 'DGLHeteroGraph' has no len()

[xinleihe]

This might be a hidden tiny bug of DGL=0.7

You may also test this example (which may reproduce the same error):

# https://docs.dgl.ai/generated/dgl.graph.html?highlight=dgl%20graph#dgl.graph
import dgl
import torch


# Source nodes for edges (2, 1), (3, 2), (4, 3)
src_ids = torch.tensor([2, 3, 4])
# Destination nodes for edges (2, 1), (3, 2), (4, 3)
dst_ids = torch.tensor([1, 2, 3])
g = dgl.graph((src_ids, dst_ids))

# https://docs.dgl.ai/generated/dgl.data.utils.split_dataset.html?highlight=split_dataset
a,b,c = dgl.data.utils.split_dataset(g, frac_list=[0.6, 0.2, 0.2])

To solve this problem, you may consider degrading DGL to a lower version like 0.6x or consider the tiny change:

# See also https://github.com/dmlc/dgl/blob/master/python/dgl/data/utils.py#L78
num_data = len(dataset) => num_data = len(dataset.nodes())