Error when loading datasets from the torch_geometric.dataset package

[TNodeCode]

🐛 Describe the bug

Hello community,

when I try to load a dataset from the torch_geometric.datasetspackage I get an error.

This is my code:

from torch_geometric.datasets import Planetoid

dataset = Planetoid(root="./data", name='Cora')

I am using PyTorch version 2.0.0+cu117 and torch_geometric version 2.3.0

And this is the error I get:

Downloading https://github.com/kimiyoung/planetoid/raw/master/data/ind.cora.x
Downloading https://github.com/kimiyoung/planetoid/raw/master/data/ind.cora.tx
Downloading https://github.com/kimiyoung/planetoid/raw/master/data/ind.cora.allx
Downloading https://github.com/kimiyoung/planetoid/raw/master/data/ind.cora.y
Downloading https://github.com/kimiyoung/planetoid/raw/master/data/ind.cora.ty
Downloading https://github.com/kimiyoung/planetoid/raw/master/data/ind.cora.ally
Downloading https://github.com/kimiyoung/planetoid/raw/master/data/ind.cora.graph
Downloading https://github.com/kimiyoung/planetoid/raw/master/data/ind.cora.test.index
Processing...
----------------------------------------------------------------
ValueError                     Traceback (most recent call last)
Input In [11], in <cell line: 2>()
      1 # Download the 'Cora' dataset and cache it in the local './data' directory
----> 2 dataset = Planetoid(root="./data", name='Cora')

File ~\AppData\Local\Programs\Python\Python310\lib\site-packages\torch_geometric\datasets\planetoid.py:92, in Planetoid.__init__(self, root, name, split, num_train_per_class, num_val, num_test, transform, pre_transform)
     89 self.split = split.lower()
     90 assert self.split in ['public', 'full', 'geom-gcn', 'random']
---> 92 super().__init__(root, transform, pre_transform)
     93 self.data, self.slices = torch.load(self.processed_paths[0])
     95 if split == 'full':

File ~\AppData\Local\Programs\Python\Python310\lib\site-packages\torch_geometric\data\in_memory_dataset.py:57, in InMemoryDataset.__init__(self, root, transform, pre_transform, pre_filter, log)
     49 def __init__(
     50     self,
     51     root: Optional[str] = None,
   (...)
     55     log: bool = True,
     56 ):
---> 57     super().__init__(root, transform, pre_transform, pre_filter, log)
     58     self._data = None
     59     self.slices = None

File ~\AppData\Local\Programs\Python\Python310\lib\site-packages\torch_geometric\data\dataset.py:97, in Dataset.__init__(self, root, transform, pre_transform, pre_filter, log)
     94     self._download()
     96 if self.has_process:
---> 97     self._process()

File ~\AppData\Local\Programs\Python\Python310\lib\site-packages\torch_geometric\data\dataset.py:230, in Dataset._process(self)
    227     print('Processing...', file=sys.stderr)
    229 makedirs(self.processed_dir)
--> 230 self.process()
    232 path = osp.join(self.processed_dir, 'pre_transform.pt')
    233 torch.save(_repr(self.pre_transform), path)

File ~\AppData\Local\Programs\Python\Python310\lib\site-packages\torch_geometric\datasets\planetoid.py:150, in Planetoid.process(self)
    149 def process(self):
--> 150     data = read_planetoid_data(self.raw_dir, self.name)
    152     if self.split == 'geom-gcn':
    153         train_masks, val_masks, test_masks = [], [], []

File ~\AppData\Local\Programs\Python\Python310\lib\site-packages\torch_geometric\io\planetoid.py:21, in read_planetoid_data(folder, prefix)
     19 def read_planetoid_data(folder, prefix):
     20     names = ['x', 'tx', 'allx', 'y', 'ty', 'ally', 'graph', 'test.index']
---> 21     items = [read_file(folder, prefix, name) for name in names]
     22     x, tx, allx, y, ty, ally, graph, test_index = items
     23     train_index = torch.arange(y.size(0), dtype=torch.long)

File ~\AppData\Local\Programs\Python\Python310\lib\site-packages\torch_geometric\io\planetoid.py:21, in <listcomp>(.0)
     19 def read_planetoid_data(folder, prefix):
     20     names = ['x', 'tx', 'allx', 'y', 'ty', 'ally', 'graph', 'test.index']
---> 21     items = [read_file(folder, prefix, name) for name in names]
     22     x, tx, allx, y, ty, ally, graph, test_index = items
     23     train_index = torch.arange(y.size(0), dtype=torch.long)

File ~\AppData\Local\Programs\Python\Python310\lib\site-packages\torch_geometric\io\planetoid.py:106, in read_file(folder, prefix, name)
    103     return out
    105 out = out.todense() if hasattr(out, 'todense') else out
--> 106 out = torch.Tensor(out)
    107 return out

ValueError: too many dimensions 'matrix'

Environment

• PyG version: 2.3.0
• PyTorch version: 2.0.0+cu117
• OS: Windows 11
• Python version: 3.10.6
• CUDA/cuDNN version: 11.4
• How you installed PyTorch and PyG (conda, pip, source): pip
• Any other relevant information (e.g., version of torch-scatter):
  torch_scatter-2.1.1+pt20cpu
  torch_sparse-0.6.17+pt20cpu
  torch_spline_conv-1.2.2+pt20cpu

[EdisonLeeeee]

Thanks for reporting. We should use torch.tensor rather than torch.Tensor here.

[rusty1s]

I fixed this in #7141. Can you please confirm that this works for you?