undefined symbol: _ZN10tensorflow8internal21CheckOpMessageBuilder9NewStringEv

[abadcd]

When I solved the ZTIN10tensorflow8OpKernelE problem, I encountered a new problem _ZN10tensorflow8internal21CheckOpMessageBuilder9NewStringEv, if you know please tell me, thank you very much

This is my mistake： Traceback (most recent call last): File "demo.py", line 3, in from pixel2mesh.models import GCN File "build/bdist.linux-x86_64/egg/pixel2mesh/models.py", line 3, in from pixel2mesh.losses import * File "build/bdist.linux-x86_64/egg/pixel2mesh/losses.py", line 2, in from pixel2mesh.cd_dist import * File "build/bdist.linux-x86_64/egg/pixel2mesh/cd_dist.py", line 3, in nn_distance_module=tf.load_op_library('./external/tf_nndistance_so.so') File "/home/software/anaconda/envs/tensorflow-gpu/lib/python2.7/site-packages/tensorflow/python/framework/load_library.py", line 56, in load_op_library lib_handle = py_tf.TF_LoadLibrary(library_filename) tensorflow.python.framework.errors_impl.NotFoundError: ./external/tf_nndistance_so.so: undefined symbol: _ZN10tensorflow8internal21CheckOpMessageBuilder9NewStringEv

[abadcd]

I solved it If you have the same problem, this is probably related to the gcc version. Check out the gcc version, gcc=5.x to use conda install tensorflow, gcc=4.x to use pip install tensorflow, because the tensorflow installed by pip may be Gcc4 compiled. If it is wrong, please indicate.

[bodhicheng]

The problem is caused by the difference between the local and compiled gcc versions of tensorflow. Conda's tensorflow is compiled by gcc7. My local version is 4.8. When I installed with pip, the problem was solved.

[anguyen216]

@abadcd How did you solve the ZTIN10tensorflow8OpKernelE problem? I'm encountering the same problem when trying to compile the functions on Google Colab (cuda 10.1, python 3.7, tensorflow 1.15.2)