mandarjoshi90 / coref

BERT for Coreference Resolution

Apache License 2.0

441 stars 92 forks source link

setup_all check gcc version #29

Closed Naman-ntc closed 4 years ago

Naman-ntc commented 4 years ago

Hi, Thanks for releasing the code. While running predict.py I was getting

tensorflow.python.framework.errors_impl.NotFoundError: ./coref_kernels.so: undefined symbol: _ZN10tensorflow12OpDefBuilder4AttrESs

I searched and found this link. It would be nice to either check the GCC version or keep this as a warning in the script to help anyone else stuck here!!

jkkummerfeld commented 4 years ago

Thanks, I ran into this problem and the same thing fixed it for me.

mandarjoshi90 commented 4 years ago

Sorry, I missed this earlier. I'll add a warning to the README. Just so I understand the solution, removing the flag -D_GLIBCXX_USE_CXX11_ABI=0 from setup_all.sh did the trick?

jkkummerfeld commented 4 years ago

That's right.

rainyrainyguo commented 3 years ago

I removed the flag from setup_all.sh but still, get the error:

tensorflow.python.framework.errors_impl.NotFoundError: ./coref_kernels.so: undefined symbol: _ZN10tensorflow8internal21CheckOpMessageBuilder9NewStringEv

mariaBio commented 3 years ago

Thank you very much for everyone who contributed to the code release.

I am running the software on the SLURM cluster and facing exactly that same issue: lib_handle = py_tf.TF_LoadLibrary(library_filename) tensorflow.python.framework.errors_impl.NotFoundError: ./coref_kernels.so: undefined symbol: _ZN10tensorflow12OpDefBuilderC1ENSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEEE

The setup_all.sh has the following line:

Linux (pip)

g++ -std=c++11 -shared coref_kernels.cc -o coref_kernels.so -fPIC ${TF_CFLAGS[@]} ${TF_LFLAGS[@]} -O2 (so the flag that was suggested to be removed is not there)