error:tf_sampling_so.so: undefined symbol: _ZTIN10tensorflow8OpKernelE

[han1366]

when I run command_train_v2.sh ,I get this mistake "tf_sampling_so.so: undefined symbol: _ZTIN10tensorflow8OpKernelE".I have already modified the file "tf_sampling_comling.sh".By the way ,I run these files in anaconda virtual environment python2.7,tensorflow-gpu1.13.0.

[witignite]

@han1366 , Could you tell me if you compiled the custom TensorFlow ops with -D_GLIBCXX_USE_CXX11_ABI=0?

I had exactly the same error message, but that was when I removed -D_GLIBCXX_USE_CXX11_ABI=0 from the tf_x_compile.sh script. With this flag in the script (as is in the author code), I didn't have any problem. My OS was Ubuntu 16.04, tensorflow-gpu-1.12.0 and CUDA 9.0.

If you solved the problem, can you also let me know how did you solve it?

[amardeepjaiman]

I am getting the same error when i am compiling on windows environment. What could be the reason ?