search

dmlc / dgl

Python package built to ease deep learning on graph, on top of existing DL frameworks.
http://dgl.ai
Apache License 2.0
13.46k stars 3.01k forks source link

Bug in dgl.data.CiteseerGraphDataset #3467

Closed tmargary closed 2 years ago

tmargary commented 2 years ago

I am trying to use CiteseerGraphDataset and other datasets and it shows the error as below:

Environment

Code

import dgl
from dgl.data import CiteseerGraphDataset

dataset = CiteseerGraphDataset()
g = dataset[0]

Error

Finished data loading and preprocessing.
  NumNodes: 3327
  NumEdges: 9228
  NumFeats: 3703
  NumClasses: 6
  NumTrainingSamples: 120
  NumValidationSamples: 500
  NumTestSamples: 1000

---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
/tmp/ipykernel_34530/3377010863.py in <module>
----> 1 dataset = CiteseerGraphDataset()
      2 g = dataset[0]
      3 g

~/anaconda3/envs/rapids-21.10/lib/python3.8/site-packages/dgl/data/citation_graph.py in __init__(self, raw_dir, force_reload, verbose, reverse_edge)
    547         name = 'citeseer'
    548 
--> 549         super(CiteseerGraphDataset, self).__init__(name, raw_dir, force_reload, verbose, reverse_edge)
    550 
    551     def __getitem__(self, idx):

~/anaconda3/envs/rapids-21.10/lib/python3.8/site-packages/dgl/data/citation_graph.py in __init__(self, name, raw_dir, force_reload, verbose, reverse_edge)
     66         self._reverse_edge = reverse_edge
     67 
---> 68         super(CitationGraphDataset, self).__init__(name,
     69                                                    url=url,
     70                                                    raw_dir=raw_dir,

~/anaconda3/envs/rapids-21.10/lib/python3.8/site-packages/dgl/data/dgl_dataset.py in __init__(self, name, url, raw_dir, hash_key, force_reload, verbose)
    280     """
    281     def __init__(self, name, url, raw_dir=None, hash_key=(), force_reload=False, verbose=False):
--> 282         super(DGLBuiltinDataset, self).__init__(name,
    283                                                 url=url,
    284                                                 raw_dir=raw_dir,

~/anaconda3/envs/rapids-21.10/lib/python3.8/site-packages/dgl/data/dgl_dataset.py in __init__(self, name, url, raw_dir, save_dir, hash_key, force_reload, verbose)
     91             self._save_dir = save_dir
     92 
---> 93         self._load()
     94 
     95     def download(self):

~/anaconda3/envs/rapids-21.10/lib/python3.8/site-packages/dgl/data/dgl_dataset.py in _load(self)
    177             self._download()
    178             self.process()
--> 179             self.save()
    180             if self.verbose:
    181                 print('Done saving data into cached files.')

~/anaconda3/envs/rapids-21.10/lib/python3.8/site-packages/dgl/data/citation_graph.py in save(self)
    171         info_path = os.path.join(self.save_path,
    172                                  self.save_name + '.pkl')
--> 173         save_graphs(str(graph_path), self._g)
    174         save_info(str(info_path), {'num_classes': self.num_classes})
    175 

~/anaconda3/envs/rapids-21.10/lib/python3.8/site-packages/dgl/data/graph_serialize.py in save_graphs(filename, g_list, labels)
    128     g_sample = g_list[0] if isinstance(g_list, list) else g_list
    129     if type(g_sample) == DGLHeteroGraph:  # Doesn't support DGLHeteroGraph's derived class
--> 130         save_heterographs(filename, g_list, labels)
    131     else:
    132         raise DGLError(

~/anaconda3/envs/rapids-21.10/lib/python3.8/site-packages/dgl/data/heterograph_serialize.py in save_heterographs(filename, g_list, labels)
     26         g_list = [g_list]
     27     assert all([type(g) == DGLHeteroGraph for g in g_list]), "Invalid DGLHeteroGraph in g_list argument"
---> 28     gdata_list = [HeteroGraphData.create(g) for g in g_list]
     29     _CAPI_SaveHeteroGraphData(filename, gdata_list, tensor_dict_to_ndarray_dict(labels))
     30 

~/anaconda3/envs/rapids-21.10/lib/python3.8/site-packages/dgl/data/heterograph_serialize.py in <listcomp>(.0)
     26         g_list = [g_list]
     27     assert all([type(g) == DGLHeteroGraph for g in g_list]), "Invalid DGLHeteroGraph in g_list argument"
---> 28     gdata_list = [HeteroGraphData.create(g) for g in g_list]
     29     _CAPI_SaveHeteroGraphData(filename, gdata_list, tensor_dict_to_ndarray_dict(labels))
     30 

~/anaconda3/envs/rapids-21.10/lib/python3.8/site-packages/dgl/data/heterograph_serialize.py in create(g)
     40             edata_list.append(tensor_dict_to_ndarray_dict(g.edges[etype].data))
     41         for ntype in g.ntypes:
---> 42             ndata_list.append(tensor_dict_to_ndarray_dict(g.nodes[ntype].data))
     43         return _CAPI_MakeHeteroGraphData(g._graph, ndata_list, edata_list, g.ntypes, g.etypes)
     44 

~/anaconda3/envs/rapids-21.10/lib/python3.8/site-packages/dgl/data/heterograph_serialize.py in tensor_dict_to_ndarray_dict(tensor_dict)
     15     ndarray_dict = {}
     16     for key, value in tensor_dict.items():
---> 17         ndarray_dict[key] = F.zerocopy_to_dgl_ndarray(value)
     18     return convert_to_strmap(ndarray_dict)
     19 

~/anaconda3/envs/rapids-21.10/lib/python3.8/site-packages/dgl/backend/pytorch/tensor.py in zerocopy_to_dgl_ndarray(data)
    332 
    333 if LooseVersion(th.__version__) >= LooseVersion("1.10.0"):
--> 334     def zerocopy_to_dgl_ndarray(data):
    335         if data.dtype == th.bool:
    336             data = data.byte()

RuntimeError: Bool type is not supported by dlpack
VoVAllen commented 2 years ago

Could you try uninstall dgl and reinstall the nightly version again? It should be fixed now.

tmargary commented 2 years ago 
conda create --name env_temp python==3.9
conda activate env_temp
pip install --pre dgl-cu111 -f https://data.dgl.ai/wheels-test/repo.html
pip install torch==1.10
python

Python 3.9.0 (default, Nov 15 2020, 14:28:56) 
[GCC 7.3.0] :: Anaconda, Inc. on linux
Type "help", "copyright", "credits" or "license" for more information.

>>> import dgl
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/tigran/anaconda3/envs/env_temp/lib/python3.9/site-packages/dgl/__init__.py", line 13, in <module>
    from .backend import load_backend, backend_name
  File "/home/tigran/anaconda3/envs/env_temp/lib/python3.9/site-packages/dgl/backend/__init__.py", line 107, in <module>
    load_backend(get_preferred_backend())
  File "/home/tigran/anaconda3/envs/env_temp/lib/python3.9/site-packages/dgl/backend/__init__.py", line 41, in load_backend
    from .._ffi.base import load_tensor_adapter # imports DGL C library
  File "/home/tigran/anaconda3/envs/env_temp/lib/python3.9/site-packages/dgl/_ffi/base.py", line 44, in <module>
    _LIB, _LIB_NAME, _DIR_NAME = _load_lib()
  File "/home/tigran/anaconda3/envs/env_temp/lib/python3.9/site-packages/dgl/_ffi/base.py", line 34, in _load_lib
    lib = ctypes.CDLL(lib_path[0])
  File "/home/tigran/anaconda3/envs/env_temp/lib/python3.9/ctypes/__init__.py", line 382, in __init__
    self._handle = _dlopen(self._name, mode)
OSError: libcudart.so.11.0: cannot open shared object file: No such file or directory
VoVAllen commented 2 years ago

Hi,

Could you try conda list | grep dgl to see is there any legacy dgl version?

VoVAllen commented 2 years ago

Hi, we've just released 0.7.2. Could you try conda install -c dglteam dgl-cuda10.2 ?

tmargary commented 2 years ago

Hi. Thank you. I have just created a new conda env and here is the results: 

---------------------------------------------------------------------------
OSError                                   Traceback (most recent call last)
/tmp/ipykernel_42765/704036641.py in <module>
----> 1 import dgl
      2 from dgl.data import CiteseerGraphDataset
      3 
      4 dataset = CiteseerGraphDataset()
      5 g = dataset[0]

~/miniconda3/envs/gnn_env/lib/python3.8/site-packages/dgl/__init__.py in <module>
     11 
     12 # Should import backend before importing anything else
---> 13 from .backend import load_backend, backend_name
     14 
     15 from . import function

~/miniconda3/envs/gnn_env/lib/python3.8/site-packages/dgl/backend/__init__.py in <module>
    105 
    106 
--> 107 load_backend(get_preferred_backend())
    108 
    109 

~/miniconda3/envs/gnn_env/lib/python3.8/site-packages/dgl/backend/__init__.py in load_backend(mod_name)
     39         raise NotImplementedError('Unsupported backend: %s' % mod_name)
     40 
---> 41     from .._ffi.base import load_tensor_adapter # imports DGL C library
     42     version = mod.__version__
     43     load_tensor_adapter(mod_name, version)

~/miniconda3/envs/gnn_env/lib/python3.8/site-packages/dgl/_ffi/base.py in <module>
     42 __version__ = libinfo.__version__
     43 # library instance of nnvm
---> 44 _LIB, _LIB_NAME, _DIR_NAME = _load_lib()
     45 
     46 # The FFI mode of DGL

~/miniconda3/envs/gnn_env/lib/python3.8/site-packages/dgl/_ffi/base.py in _load_lib()
     32     """Load libary by searching possible path."""
     33     lib_path = libinfo.find_lib_path()
---> 34     lib = ctypes.CDLL(lib_path[0])
     35     dirname = os.path.dirname(lib_path[0])
     36     basename = os.path.basename(lib_path[0])

~/miniconda3/envs/gnn_env/lib/python3.8/ctypes/__init__.py in __init__(self, name, mode, handle, use_errno, use_last_error, winmode)
    367 
    368         if handle is None:
--> 369             self._handle = _dlopen(self._name, mode)
    370         else:
    371             self._handle = handle

OSError: libcudart.so.10.2: cannot open shared object file: No such file or directory

PyTorch: 1.10.0+cu113 How you installed DGL: conda install -c dglteam dgl-cuda10.2 Python version: 3.8

VoVAllen commented 2 years ago

Please try conda install -c dglteam dgl-cuda11.3 if you are using cuda 11.3

BarclayII commented 2 years ago

Please feel free to reopen if the solution above did not work for you.