can't run with this error: undefined symbol: _ZN5torch3jit17parseSchemaOrNameERKSs
See original GitHub issue😵 Installation
Hi!
I installed the PyG according to the documentation(https://pytorch-geometric.readthedocs.io/en/latest/notes/installation.html). tried to run some sample code to test if I installed successfully, then I encountered this issue.
Traceback (most recent call last): File “test_dataloader.py”, line 54, in <module> from torch_geometric.data import Data File “/usr/local/lib/python3.8/dist-packages/torch_geometric/init.py”, line 4, in <module> import torch_geometric.data File “/usr/local/lib/python3.8/dist-packages/torch_geometric/data/init.py”, line 1, in <module> from .data import Data File “/usr/local/lib/python3.8/dist-packages/torch_geometric/data/data.py”, line 3, in <module> from torch_geometric.typing import OptTensor, NodeType, EdgeType File “/usr/local/lib/python3.8/dist-packages/torch_geometric/typing.py”, line 4, in <module> from torch_sparse import SparseTensor File “/usr/local/lib/python3.8/dist-packages/torch_sparse/init.py”, line 15, in <module> torch.ops.load_library(importlib.machinery.PathFinder().find_spec( File “/usr/lib/python3/dist-packages/torch/_ops.py”, line 104, in load_library ctypes.CDLL(path) File “/usr/lib/python3.8/ctypes/init.py”, line 373, in init self._handle = _dlopen(self._name, mode)
OSError: /usr/local/lib/python3.8/dist-packages/torch_sparse/_version_cuda.so: undefined symbol: _ZN5torch3jit17parseSchemaOrNameERKSs
I tried to search previous issues, there is someplace mentioned about that system cuda and PyTorch cuda mismatch. I check mine, they both at 11.1, so I think it may not be the root cause of my issue.
Can anyone help me with this? Is there any other possibility? I have been stuck here for 4 days and have no clue about next move. Thank you so much for any advice!
Environment
- PyG version (
torch_geometric.__version__
): 2.0.3 - PyTorch version: (
torch.__version__
): 1.9 - OS (e.g., Linux): Linux
- Python version (e.g.,
3.9
): 3.8 - CUDA/cuDNN version: cuda 11.1 (nvcc version also 11.1)
- How you installed PyTorch and PyG (
conda
,pip
, source): pip - Any other relevant information (e.g., version of
torch-scatter
):
Additional context
Issue Analytics
- State:
- Created 2 years ago
- Comments:16 (8 by maintainers)
Top GitHub Comments
@rusty1s I followed version corresbondences on [https://pytorch-geometric.com/whl/], thus
torch-1.6.0
andtorch-scatter-2.0.5
should match. And your doubt about the 1.6 version of Pytorch is correct to some extent, cause that error does not occur when I use 1.4 . (chose 1.4 because my cuda version is 10.0, kind of old.) I will try Pytorch 1.10 as you suggested for testing and give an update later. Thank you.Thanks for confirming. I hope you are good to go now.