undefined symbol error: _ZNK10tensorflow4Node11type_stringEv

[pacifastacus]

I installed ngraph-tensorflow-bridge pip package in a virtualenv by following the readme.

virtualenv --system-site-packages -p python3 ngraph-venv
source ./ngraph-venv/bin/activate
pip install -U tensorflow==1.14.0
pip install -U ngraph-tensorflow-bridge

After that when i want to import the ngraph_bridge module i get an error: NotFoundError: /home/***/ngraph-venv/lib/python3.7/site-packages/ngraph_bridge/libngraph_bridge.so: undefined symbol: _ZNK10tensorflow4Node11type_stringEv

My system: Processor : Intel(R) Core(TM) i5-2520M CPU @ 2.50GHz Memory : 10114MB (2114MB used) Operating System : Debian GNU/Linux 10 (buster)

[arne0x7]

I'm getting the same error.

Intel® Core™ i5-6500 CPU @ 3.20GHz × 4 Ubuntu 19.04 Python 3.7.3

[SleepProgger]

Same here in an anaconda environment with python 3.6.7. Happens as soon as ngraph_bridge is imported.

# python -c "import ngraph_bridge"
Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "/home/nope/.conda/envs/test/lib/python3.6/site-packages/ngraph_bridge/__init__.py", line 97, in <module>
    _ = load_library.load_op_library(full_lib_path)
  File "/home/nope/.conda/envs/test/lib/python3.6/site-packages/tensorflow/python/framework/load_library.py", line 61, in load_op_library
    lib_handle = py_tf.TF_LoadLibrary(library_filename)
tensorflow.python.framework.errors_impl.NotFoundError: /home/nope/.conda/envs/test/lib/python3.6/site-packages/ngraph_bridge/libngraph_bridge.so: undefined symbol: _ZNK10tensorflow4Node11type_stringEv

System: CPU: AMD Ryzen 5 3600 6-Core Processor(Altho the same thing occurred about a month ago with my old CPU) OS: Manjaro stable with 5.2.8-1 kernel.

Edit: I acidentally had tensorflow installed via conda in that environment. After uninstalling it and reinstalling tensorflow via pip that error is gone.

[AlexMV12]

I have the same issue, both on an Ubuntu (19.04) machine and a ArchLinux machine. EDIT After setting up a VM with Ubuntu 18.04 I noticed that it worked on such OS: so, looking at the differences, I realized that the problem seems to be with Python 3.7. In fact, with Python 3.6 this problem doesn't happen.

[drwebb]

I also had this problem when I was using a tensorflow package install in a conda environment using conda install. Using pip install -U tensorflow==1.14.0 fixed the issue inside the conda environment.