Pytorch_geometric: Undefined symbol: _ZN5torch3jit17parseSchemaOrNameERKSs

Created on 30 Mar 2020 · 14Comments · Source: rusty1s/pytorch_geometric

📚 Installation

Environment

OS:18.4
Python version: 3.6
PyTorch version:1.4
CUDA/cuDNN version:10.0
GCC version:7.4
How did you try to install PyTorch Geometric and its extensions (pip, source): pip wheel
Any other relevant information:

_libgcc_mutex 0.1 main
_pytorch_select 0.2 gpu_0
blas 1.0 mkl
ca-certificates 2020.1.1 0
certifi 2019.11.28 py36_1
cffi 1.14.0 py36h2e261b9_0
chardet 3.0.4 pypi_0 pypi
cudatoolkit 10.0.130 0
cudnn 7.6.5 cuda10.0_0
cycler 0.10.0 pypi_0 pypi
decorator 4.4.2 pypi_0 pypi
freetype 2.9.1 h8a8886c_1
googledrivedownloader 0.4 pypi_0 pypi
h5py 2.10.0 pypi_0 pypi
idna 2.9 pypi_0 pypi
imageio 2.8.0 pypi_0 pypi
intel-openmp 2020.0 166
isodate 0.6.0 pypi_0 pypi
joblib 0.14.1 pypi_0 pypi
jpeg 9b h024ee3a_2
kiwisolver 1.1.0 pypi_0 pypi
ld_impl_linux-64 2.33.1 h53a641e_7
libedit 3.1.20181209 hc058e9b_0
libffi 3.2.1 hd88cf55_4
libgcc-ng 9.1.0 hdf63c60_0
libgfortran-ng 7.3.0 hdf63c60_0
libpng 1.6.37 hbc83047_0
libstdcxx-ng 9.1.0 hdf63c60_0
libtiff 4.1.0 h2733197_0
llvmlite 0.31.0 pypi_0 pypi
matplotlib 3.2.1 pypi_0 pypi
mkl 2020.0 166
mkl-service 2.3.0 py36he904b0f_0
mkl_fft 1.0.15 py36ha843d7b_0
mkl_random 1.1.0 py36hd6b4f25_0
ncurses 6.2 he6710b0_0
networkx 2.4 pypi_0 pypi
ninja 1.9.0 py36hfd86e86_0
numba 0.48.0 pypi_0 pypi
numpy 1.18.1 py36h4f9e942_0
numpy-base 1.18.1 py36hde5b4d6_1
olefile 0.46 py36_0
openssl 1.1.1e h7b6447c_0
pandas 1.0.3 pypi_0 pypi
pillow 7.0.0 py36hb39fc2d_0
pip 20.0.2 py36_1
plyfile 0.7.2 pypi_0 pypi
pycparser 2.20 py_0
pyparsing 2.4.6 pypi_0 pypi
python 3.6.10 hcf32534_1
python-dateutil 2.8.1 pypi_0 pypi
pytorch 1.4.0 cuda100py36h75a84eb_0
pytz 2019.3 pypi_0 pypi
pywavelets 1.1.1 pypi_0 pypi
rdflib 4.2.2 pypi_0 pypi
readline 8.0 h7b6447c_0
requests 2.23.0 pypi_0 pypi
scikit-image 0.16.2 pypi_0 pypi
scikit-learn 0.22.2.post1 pypi_0 pypi
scipy 1.4.1 pypi_0 pypi
setuptools 46.1.1 py36_0
six 1.14.0 py36_0
sqlite 3.31.1 h7b6447c_0
tk 8.6.8 hbc83047_0
torch-cluster 1.5.3 pypi_0 pypi
torch-geometric 1.4.3 pypi_0 pypi
torch-scatter 2.0.4 pypi_0 pypi
torch-sparse 0.6.1 pypi_0 pypi
torch-spline-conv 1.2.0 pypi_0 pypi
torchvision 0.2.1 py36_0
urllib3 1.25.8 pypi_0 pypi
wheel 0.34.2 py36_0
xz 5.2.4 h14c3975_4
zlib 1.2.11 h7b6447c_3
zstd 1.3.7 h0b5b093_0

Checklist

[ ] I followed the installation guide.
[ ] I cannot find my error message in the FAQ.
[ 1 ] I set up CUDA correctly and can compile CUDA code via nvcc.
[ ] I have cloned the repository and tried a manual installation from source.
[ ] I do have multiple CUDA versions on my machine.
[ ] I checked if the official extension example runs on my machine.
[ ] The offical extension example runs on my machine.

Additional context

my problem like https://github.com/rusty1s/pytorch_geometric/issues/999

File "main.py", line 2, in
from torch_geometric.datasets import TUDataset
File "/home/xiejunlin/anaconda3/envs/graph/lib/python3.6/site-packages/torch_geometric/__init__.py", line 2, in
import torch_geometric.nn
File "/home/xiejunlin/anaconda3/envs/graph/lib/python3.6/site-packages/torch_geometric/nn/__init__.py", line 2, in
from .data_parallel import DataParallel
File "/home/xiejunlin/anaconda3/envs/graph/lib/python3.6/site-packages/torch_geometric/nn/data_parallel.py", line 5, in
from torch_geometric.data import Batch
File "/home/xiejunlin/anaconda3/envs/graph/lib/python3.6/site-packages/torch_geometric/data/__init__.py", line 1, in
from .data import Data
File "/home/xiejunlin/anaconda3/envs/graph/lib/python3.6/site-packages/torch_geometric/data/data.py", line 7, in
from torch_sparse import coalesce
File "/home/xiejunlin/anaconda3/envs/graph/lib/python3.6/site-packages/torch_sparse/__init__.py", line 23, in
raise OSError(e)
OSError: /home/xiejunlin/anaconda3/envs/graph/lib/python3.6/site-packages/torch_sparse/_version.so: undefined symbol: _ZN5torch3jit17parseSchemaOrNameERKSs

can you have any idea?

Source

jun0wanan

Most helpful comment

I encounter this problem too. The installation of "pytorch_geometric" is a terrible, awful, and horrible experience ever!!!

dongweidongwei on 30 Mar 2020

👍17

All 14 comments

I have install it for 4 days ... 😭😭 who can help me ?

jun0wanan on 30 Mar 2020

I encounter this problem too. The installation of "pytorch_geometric" is a terrible, awful, and horrible experience ever!!!

dongweidongwei on 30 Mar 2020

👍17

I encounter this problem too. The installation of "pytorch_geometric" is a terrible, awful, and horrible experience ever!!!

Maybe because we are newer .... Do you know if we only can use pytorch==1.4 >< ? I want to use pytorch=1.2......

jun0wanan on 30 Mar 2020

If you want to use PyTorch 1.2, you can install older versions of the extensions, e.g.:

torch-scatter==1.4.0
torch-sparse==0.4.3
torch-spline-conv==1.1.1
torch-cluster==1.4.5

The latest version require torch==1.4.0 though.

The error undefined symbol: _ZN5torch3jit17parseSchemaOrNameERKSs usually indicates that you have multiple PyTorch versions installed.

For a better debugging experience, you can install the extensions from source, and post any errors in the installation log.

rusty1s on 30 Mar 2020

I encounter this problem too. The installation of "pytorch_geometric" is a terrible, awful, and horrible experience ever!!!

Maybe because we are newer .... Do you know if we only can use pytorch==1.4 >< ? I want to use pytorch=1.2......

I think that an excellent project should have a wonderful usability. Because we focus on our work efficiently via an easy-to-use software. However, compared with "DGL", the usability of "pytorch_geometric" is very terrible. According to the author, multiple versions of CUDA or pytorch coexisting in your OS may result in this problem, but I have only CUDA 10,1 and pytorch 1.4. I have no idea about this problem!

dongweidongwei on 30 Mar 2020

Well, I have tested all wheels on multiple platforms and they all ran fine, so I do not know exactly what the problem here is. You are invited to dig into this issue and find out why PyTorch cannot load the _version.so on your OS.

Furthermore, note that you do not need to use the wheels, you can always install from source.

rusty1s on 30 Mar 2020

Well, I have tested all wheels on multiple platforms and they all ran fine, so I do not know exactly what the problem here is. You are invited to dig into this issue and find out why PyTorch cannot load the _version.so on your OS.

Furthermore, note that you do not need to use the wheels, you can always install from source.

Thank your reply~~ This is a problem . Because I also have no other pytorch,and I only install one. Maybe can you put a virtual environment to give others .. I think it will be fine to many newers

jun0wanan on 31 Mar 2020

I encounter this problem too. The installation of "pytorch_geometric" is a terrible, awful, and horrible experience ever!!!

Maybe because we are newer .... Do you know if we only can use pytorch==1.4 >< ? I want to use pytorch=1.2......

I think that an excellent project should have a wonderful usability. Because we focus on our work efficiently via an easy-to-use software. However, compared with "DGL", the usability of "pytorch_geometric" is very terrible. According to the author, multiple versions of CUDA or pytorch coexisting in your OS may result in this problem, but I have only CUDA 10,1 and pytorch 1.4. I have no idea about this problem!

hello? You can download the environment to install it in cuda10.1...
https://pan.baidu.com/s/1TtMvo9haPegmtesECAmSCg : vijp

then use conda create -f **.yaml then you can use it
ref: https://blog.csdn.net/ft_sunshine/article/details/92215164

jun0wanan on 31 Mar 2020

I have solved this problem. The reason is that: do not install pytorch via anaconda-navigator ! It will install both cpu and gpu versions, even though you only select the pytorch. You can download the specific version of the pytorch from:

https://download.pytorch.org/whl/torch_stable.html,

and this problem will be solved. All dependent packages are listed:
cu101/torch-1.4.0-cp38-cp38-linux_x86_64.whl
cu101/torchvision-0.5.0-cp38-cp38-linux_x86_64.whl
torch_cluster-1.5.3+cu101-cp38-cp38-linux_x86_64.whl
torch_scatter-2.0.4+cu101-cp38-cp38-linux_x86_64.whl
torch_sparse-0.6.1+cu101-cp38-cp38-linux_x86_64.whl
torch_spline_conv-1.2.0+cu101-cp38-cp38-linux_x86_64.whl

The author is right! Multi-versions of pytorch (cpu and gpu) result in this problem! I am sorry about my comments in yesterday. Pytorch geometric is an excellent project !!!

dongweidongwei on 31 Mar 2020

👍2

I encounter this problem too. The installation of "pytorch_geometric" is a terrible, awful, and horrible experience ever!!!

Maybe because we are newer .... Do you know if we only can use pytorch==1.4 >< ? I want to use pytorch=1.2......

I think that an excellent project should have a wonderful usability. Because we focus on our work efficiently via an easy-to-use software. However, compared with "DGL", the usability of "pytorch_geometric" is very terrible. According to the author, multiple versions of CUDA or pytorch coexisting in your OS may result in this problem, but I have only CUDA 10,1 and pytorch 1.4. I have no idea about this problem!

hello? You can download the environment to install it in cuda10.1...
https://pan.baidu.com/s/1TtMvo9haPegmtesECAmSCg : vijp

then use conda create -f **.yaml then you can use it
ref: https://blog.csdn.net/ft_sunshine/article/details/92215164

https://download.pytorch.org/whl/torch_stable.html,

The author is right! Multi-versions of pytorch (cpu and gpu) result in this problem! I am sorry about my comments in yesterday. Pytorch geometric is an excellent project !!!

dongweidongwei on 31 Mar 2020

Well, I have tested all wheels on multiple platforms and they all ran fine, so I do not know exactly what the problem here is. You are invited to dig into this issue and find out why PyTorch cannot load the _version.so on your OS.

Furthermore, note that you do not need to use the wheels, you can always install from source.

https://download.pytorch.org/whl/torch_stable.html,

You are right! Multi-versions of pytorch (cpu and gpu) result in this problem! I am sorry about my comments in yesterday. Pytorch geometric is an excellent project !!!

dongweidongwei on 31 Mar 2020

好吧，我已经在多个平台上测试了所有车轮，并且它们都运行良好，所以我不知道到底是什么问题。邀请您深入研究此问题，并找出为什么PyTorch无法_version.so在您的OS上加载。
此外，请注意，您无需使用轮子，可以始终从源代码安装。

我已经解决了这个问题。原因是：不要通过anaconda-navigator安装pytorch！即使您仅选择pytorch，它也会安装cpu和gpu版本。您可以从以下位置下载特定版本的pytorch：

https://download.pytorch.org/whl/torch_stable.html，

并且这个问题将得到解决。列出了所有相关软件包：
cu101 / torch-1.4.0-cp38-cp38-linux_x86_64.whl
cu101 / torchvision-0.5.0-cp38-cp38-linux_x86_64.whl
torch_cluster-1.5.3 + cu101-cp38-cp38-cp38-linux_x86_64。 whl
torch_scatter-2.0.4 + cu101-cp38-cp38-linux_x86_64.whl
torch_sparse-0.6.1 + cu101-cp38-cp38-linux_x86_64.whl
torch_spline_conv-1.2.0 + cu101-cp38-cp38-linux_x86_64.whl

你是对的！pytorch的多版本（cpu和gpu）导致此问题！我对昨天的评论感到抱歉。Pytorch Geometric是一个出色的项目！

So we are newer ~~~ But I think if you have a friend who have installed it , you can ask him to copy the environment to you.Because it is faster .... Anyway ,thanks~~~~~~🙂

jun0wanan on 31 Mar 2020

好吧，我已经在多个平台上测试了所有车轮，并且它们都运行良好，所以我不知道到底是什么问题。邀请您深入研究此问题，并找出为什么PyTorch无法_version.so在您的OS上加载。
此外，请注意，您无需使用轮子，可以始终从源代码安装。

我已经解决了这个问题。原因是：不要通过anaconda-navigator安装pytorch！即使您仅选择pytorch，它也会安装cpu和gpu版本。您可以从以下位置下载特定版本的pytorch：
https://download.pytorch.org/whl/torch_stable.html，
并且这个问题将得到解决。列出了所有相关软件包：
cu101 / torch-1.4.0-cp38-cp38-linux_x86_64.whl
cu101 / torchvision-0.5.0-cp38-cp38-linux_x86_64.whl
torch_cluster-1.5.3 + cu101-cp38-cp38-cp38-linux_x86_64。 whl
torch_scatter-2.0.4 + cu101-cp38-cp38-linux_x86_64.whl
torch_sparse-0.6.1 + cu101-cp38-cp38-linux_x86_64.whl
torch_spline_conv-1.2.0 + cu101-cp38-cp38-linux_x86_64.whl
你是对的！pytorch的多版本（cpu和gpu）导致此问题！我对昨天的评论感到抱歉。Pytorch Geometric是一个出色的项目！

So we are newer ~~~ But I think if you have a friend who have installed it , you can ask him to copy the environment to you.Because it is faster .... Anyway ,thanks~~~~~~🙂

Do you mean that a "newer" is a green hand? I am not a green hand, otherwise, I will not solve this problem independently. I just focus on my work attentively, but this problem makes me waste a lot of time.

dongweidongwei on 31 Mar 2020

on windows no gpu

import torch_geometric
Traceback (most recent call last):
File "G:\anaconda\envs\python3.6\lib\site-packages\torch_sparse__init__.py", line 15, in
library, [osp.dirname(__file__)]).origin)
File "G:\anaconda\envs\python3.6\lib\site-packages\torch_ops.py", line 106, in load_library
ctypes.CDLL(path)
File "G:\anaconda\envs\python3.6\lib\ctypes__init__.py", line 348, in __init__
self._handle = _dlopen(self._name, mode)
OSError: [WinError 126] 找不到指定的模块。

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "", line 1, in
File "G:\anaconda\envs\python3.6\lib\site-packages\torch_geometric__init__.py", line 2, in
import torch_geometric.nn
File "G:\anaconda\envs\python3.6\lib\site-packages\torch_geometric\nn__init__.py", line 2, in
from .data_parallel import DataParallel
File "G:\anaconda\envs\python3.6\lib\site-packages\torch_geometric\nn\data_parallel.py", line 5, in
from torch_geometric.data import Batch
File "G:\anaconda\envs\python3.6\lib\site-packages\torch_geometric\data__init__.py", line 1, in
from .data import Data
File "G:\anaconda\envs\python3.6\lib\site-packages\torch_geometric\data\data.py", line 7, in
from torch_sparse import coalesce
File "G:\anaconda\envs\python3.6\lib\site-packages\torch_sparse__init__.py", line 23, in
raise OSError(e)
OSError: [WinError 126] 找不到指定的模块。