import geometric fail

[JHZhou123]

Hi, I can not import torch_geometric.nn, it says

  raceback (most recent call last):

  File "C:\Users\81568\Anaconda3\lib\site-packages\torch_cluster\__init__.py", line 15, in <module>
    library, [osp.dirname(__file__)]).origin)

  File "C:\Users\81568\Anaconda3\lib\site-packages\torch\_ops.py", line 104, in load_library
    ctypes.CDLL(path)

  File "C:\Users\81568\Anaconda3\lib\ctypes\__init__.py", line 364, in __init__
    self._handle = _dlopen(self._name, mode)

OSError: The specified procedure could not be found


During handling of the above exception, another exception occurred:

Traceback (most recent call last):

  File "<ipython-input-180-8d54e42149ae>", line 1, in <module>
    from torch_geometric.nn import GCNConv

  File "C:\Users\81568\Anaconda3\lib\site-packages\torch_geometric\nn\__init__.py", line 5, in <module>
    from .to_hetero_transformer import to_hetero

  File "C:\Users\81568\Anaconda3\lib\site-packages\torch_geometric\nn\to_hetero_transformer.py", line 11, in <module>
    from torch_geometric.nn.fx import Transformer

  File "C:\Users\81568\Anaconda3\lib\site-packages\torch_geometric\nn\fx.py", line 8, in <module>
    from torch_geometric.nn.conv import MessagePassing

  File "C:\Users\81568\Anaconda3\lib\site-packages\torch_geometric\nn\conv\__init__.py", line 6, in <module>
    from .gravnet_conv import GravNetConv

  File "C:\Users\81568\Anaconda3\lib\site-packages\torch_geometric\nn\conv\gravnet_conv.py", line 11, in <module>
    from torch_cluster import knn

  File "C:\Users\81568\Anaconda3\lib\site-packages\torch_cluster\__init__.py", line 21, in <module>
    f'Expected PyTorch version {t_major}.{t_minor} but found '

RuntimeError: Expected PyTorch version 1.4 but found version 1.9.

I have uninstall geometric and reinstall it for 1.9 version.
Seems like the pytorch version issue, however, other part like from torch_geometric.data import Data,DataLoader can run successfully.

Could you helper me with it?

[QuanticDisaster]

Maybe it is a discrepancy between the version of torch-cluster (1.4) and the version of PyG (1.9) ? How did you install it ?

[JHZhou123]

I reinstalled cluster via pip and it's sill not working.

[QuanticDisaster]

If you really want to do it with pip, i think you have to uninstall torch geometric pip uninstall torch-geometric (run the command multiple times) before reinstalling with pip and pip install torch-cluster -f https://data.pyg.org/whl/torch-${TORCH}+${CUDA}.html

But I see you are using anaconda so I would advise to install thorugh the command : conda install pyg -c pyg -c conda-forge which I think will handle it better

[JHZhou123]

I have unintalled torch-geometric and reinstall torch-cluster but the error is still there, any other suggestion?

[rusty1s]

Can you show us the commands that you used to install PyTorch and torch-cluster + any installation logs? Thanks!

[rusty1s]

That's weird :) Your versions also match with my versions. Mhh. What happens if you increase the batch_size to 2? Does it run?

[JHZhou123]

There is still the same error.

  File "<ipython-input-222-15d640de48e7>", line 45, in <module>
    x=gcn(d.x,d.edge_index,d.batch)

  File "C:\Users\81568\Anaconda3\lib\site-packages\torch\nn\modules\module.py", line 1051, in _call_impl
    return forward_call(*input, **kwargs)

  File "<ipython-input-222-15d640de48e7>", line 22, in forward
    x = global_mean_pool(x, batch)  # [batch_size, hidden_channels]

  File "C:\Users\81568\Anaconda3\lib\site-packages\torch_geometric\nn\glob\glob.py", line 48, in global_mean_pool
    size = int(batch.max().item() + 1) if size is None else size

AttributeError: 'NoneType' object has no attribute 'max'

[rusty1s]

Can you do me a favor and try to install PyG from source?

pip uninstall torch-geometric
pip uninstall torch-geometric
pip install git+https://github.com/pyg-team/pytorch_geometric.git

Let me know if this resolves your issue.

[JHZhou123]

There are still some bugs with torch-geometric.

[autoreload of torch_geometric.data failed: Traceback (most recent call last):
  File "C:\Users\81568\Anaconda3\lib\site-packages\IPython\extensions\autoreload.py", line 245, in check
    superreload(m, reload, self.old_objects)
  File "C:\Users\81568\Anaconda3\lib\site-packages\IPython\extensions\autoreload.py", line 394, in superreload
    module = reload(module)
  File "C:\Users\81568\Anaconda3\lib\imp.py", line 314, in reload
    return importlib.reload(module)
  File "C:\Users\81568\Anaconda3\lib\importlib\__init__.py", line 169, in reload
    _bootstrap._exec(spec, module)
  File "<frozen importlib._bootstrap>", line 630, in _exec
  File "<frozen importlib._bootstrap_external>", line 728, in exec_module
  File "<frozen importlib._bootstrap>", line 219, in _call_with_frames_removed
  File "C:\Users\81568\Anaconda3\lib\site-packages\torch_geometric\data\__init__.py", line 7, in <module>
    from .lightning_datamodule import LightningDataset, LightningNodeData
  File "C:\Users\81568\Anaconda3\lib\site-packages\torch_geometric\data\lightning_datamodule.py", line 7, in <module>
    from torch_geometric.typing import InputNodes
ImportError: cannot import name 'InputNodes' from 'torch_geometric.typing' (C:\Users\81568\Anaconda3\lib\site-packages\torch_geometric\typing.py)
]
[autoreload of torch_geometric.transforms.to_dense failed: Traceback (most recent call last):
  File "C:\Users\81568\Anaconda3\lib\site-packages\IPython\extensions\autoreload.py", line 245, in check
    superreload(m, reload, self.old_objects)
  File "C:\Users\81568\Anaconda3\lib\site-packages\IPython\extensions\autoreload.py", line 410, in superreload
    update_generic(old_obj, new_obj)
  File "C:\Users\81568\Anaconda3\lib\site-packages\IPython\extensions\autoreload.py", line 347, in update_generic
    update(a, b)
  File "C:\Users\81568\Anaconda3\lib\site-packages\IPython\extensions\autoreload.py", line 302, in update_class
    if update_generic(old_obj, new_obj): continue
  File "C:\Users\81568\Anaconda3\lib\site-packages\IPython\extensions\autoreload.py", line 347, in update_generic
    update(a, b)
  File "C:\Users\81568\Anaconda3\lib\site-packages\IPython\extensions\autoreload.py", line 266, in update_function
    setattr(old, name, getattr(new, name))
ValueError: __repr__() requires a code object with 0 free vars, not 1
]
Traceback (most recent call last):

  File "<ipython-input-223-15d640de48e7>", line 45, in <module>
    x=gcn(d.x,d.edge_index,d.batch)

  File "C:\Users\81568\Anaconda3\lib\site-packages\torch\nn\modules\module.py", line 1051, in _call_impl
    return forward_call(*input, **kwargs)

  File "<ipython-input-223-15d640de48e7>", line 22, in forward
    x = global_mean_pool(x, batch)  # [batch_size, hidden_channels]

  File "C:\Users\81568\Anaconda3\lib\site-packages\torch_geometric\nn\glob\glob.py", line 48, in global_mean_pool
    size = int(batch.max().item() + 1) if size is None else size

AttributeError: 'NoneType' object has no attribute 'max'

[rusty1s]

I'm clueless TBH. Can you debug why this code block does not seem to get called for you?