ImportError: /home/rd4/.conda/envs/jmlr/lib/python3.7/site-packages/topologylayer/functional/persistence.cpython-37m-x86_64-linux-gnu.so: undefined symbol: _ZN3c105ErrorC1ENS_14SourceLocationERKSs

[changshuowang]

Hello, after I installed it, I encountered a problem in the process of running the program:

(jmlr) [rd4@localhost levelset]$ python noisy_circle.py Traceback (most recent call last): File "noisy_circle.py", line 3, in from topologylayer.nn import LevelSetLayer2D, SumBarcodeLengths, PartialSumBarcodeLengths File "/home/rd4/.conda/envs/jmlr/lib/python3.7/site-packages/topologylayer/init.py", line 1, in import topologylayer.nn File "/home/rd4/.conda/envs/jmlr/lib/python3.7/site-packages/topologylayer/nn/init.py", line 3, in from .levelset import LevelSetLayer, LevelSetLayer1D, LevelSetLayer2D File "/home/rd4/.conda/envs/jmlr/lib/python3.7/site-packages/topologylayer/nn/levelset.py", line 1, in from ..functional.sublevel import SubLevelSetDiagram File "/home/rd4/.conda/envs/jmlr/lib/python3.7/site-packages/topologylayer/functional/sublevel.py", line 6, in from .persistence import SimplicialComplex, persistenceForwardCohom, persistenceBackward, persistenceForwardHom ImportError: /home/rd4/.conda/envs/jmlr/lib/python3.7/site-packages/topologylayer/functional/persistence.cpython-37m-x86_64-linux-gnu.so: undefined symbol: _ZN3c105ErrorC1ENS_14SourceLocationERKSs ----------------------------------------------------------------------------------------------My environment is GCC 7.3.0 cudatoolkit 9.2
cudnn 7.6.5
numpy 1.18.1
python 3.7.7 pytorch 1.2.0 torchvision 0.4.0 scipy 1.4.1

Can you help me solve it？ Thanks!

[bruel-gabrielsson]

Hi! Thanks for sharing. Are you using MacOS?

[yvonna-li]

I am also encountering the same error with the following specifications: numpy==1.16.4 scipy==1.2.1 torch==1.1.0 python3.6.8 centos 4.18.0-193.28.1.el8_2.x86_64 gcc-c++-8.3.1-5.el8.0.2.x86_64 I built from source using python setup.py install --record files.txt Could you please help?

[heitere]

I also had the error reported by @changshuowang on Ubuntu 21.10. Here are two setups that show the error and one that works. After installing the requirements with conda, I installed topologylayer with pip.

Error setup (1), requirements.yml: channels:

• pytorch
• defaults
• conda-forge

dependencies:

• cudatoolkit=11.3.1
• matplotlib=3.4.3
• numpy=1.21.2
• python=3.9.7
• pytorch=1.9.0
• scipy=1.7.1

resulting in the following package versions:

• _libgcc_mutex=0.1=main
• _openmp_mutex=4.5=1_gnu
• cpuonly=1.0=0
• cudatoolkit=11.3.1=h2bc3f7f_2
• libgcc-ng=9.3.0=h5101ec6_17
• libgomp=9.3.0=h5101ec6_17
• libstdcxx-ng=9.3.0=hd4cf53a_17
• pytorch=1.9.0=py3.9_cpu_0

Error: >> import topologylayer ImportError: /path/to/environment/lib/python3.9/site-packages/ topologylayer/functional/persistence.cpython-39-x86_64-linux-gnu.so: undefined symbol: _ZNSt15__exception_ptr13exception_ptr10_M_releaseEv

Error setup (2), requirements.yml (with versions as specified in the requirements file of this repository): channels:

• pytorch
• defaults
• conda-forge

dependencies:

• matplotlib
• numpy=1.16.4
• pytorch=1.1.0
• scipy=1.2.1
• torchvision

resulting in the following package versions:

• _libgcc_mutex=0.1=main
• _openmp_mutex=4.5=1_gnu
• cudatoolkit=10.0.130=0
• libgcc-ng=9.3.0=h5101ec6_17
• libgomp=9.3.0=h5101ec6_17
• libstdcxx-ng=9.3.0=hd4cf53a_17
• python=3.7.11=h12debd9_0
• pytorch=1.1.0=py3.7_cuda10.0.130_cudnn7.5.1_0

Error: >> import topologylayer ImportError: /path/to/environment/lib/python3.7/site-packages/ topologylayer/functional/persistence.cpython-37m-x86_64-linux-gnu.so: undefined symbol: _ZNSt15__exception_ptr13exception_ptr10_M_releaseEv

Working setup, requirements.yml: channels:

• pytorch
• defaults
• conda-forge

dependencies:

• cudatoolkit=11.1.1
• matplotlib=3.4.3
• numpy=1.21.2
• python=3.9.7
• pytorch=1.9.0
• scipy=1.7.1

resulting in the following package versions:

• _libgcc_mutex=0.1=conda_forge
• _openmp_mutex=4.5=1_llvm
• cudatoolkit=11.1.1=h6406543_9
• libgcc-ng=11.2.0=h1d223b6_11
• libstdcxx-ng=11.2.0=he4da1e4_11
• llvm-openmp=12.0.1=h4bd325d_1
• pytorch=1.9.0=py3.9_cuda11.1_cudnn8.0.5_0