undefined symbol: _ZN10tensorflow8internal21CheckOpMessageBuilder9NewStringEv 问题

修改make.sh文件为以下内容

########################

TF_INC=$(python -c 'import tensorflow as tf; print(tf.sysconfig.get_include())')
TF_LIB=$(python -c 'import tensorflow as tf; print(tf.sysconfig.get_lib())')

CUDA_PATH=/usr/local/cuda/
CXXFLAGS=''

if [[ "$OSTYPE" =~ ^darwin ]]; then
        CXXFLAGS+='-undefined dynamic_lookup'
fi

cd roi_pooling_layer

if [ -d "$CUDA_PATH" ]; then
        nvcc -std=c++11 -c -o roi_pooling_op.cu.o roi_pooling_op_gpu.cu.cc \
                -I $TF_INC -D GOOGLE_CUDA=1 -x cu -Xcompiler -fPIC $CXXFLAGS --expt-relaxed-constexpr\
                -arch=sm_37

        g++ -std=c++11 -shared -o roi_pooling.so roi_pooling_op.cc  -D_GLIBCXX_USE_CXX11_ABI=1  \
                roi_pooling_op.cu.o -I $TF_INC -I $TF_INC/external/nsync/public  -L $TF_LIB -D GOOGLE_CUDA=1  -ltensorflow_framework -fPIC $CXXFLAGS \
                -lcudart -L $CUDA_PATH/lib64
else
        g++ -std=c++11 -shared -o roi_pooling.so roi_pooling_op.cc \
                -I $TF_INC -fPIC $CXXFLAGS
fi

cd ..

#cd feature_extrapolating_layer

#nvcc -std=c++11 -c -o feature_extrapolating_op.cu.o feature_extrapolating_op_gpu.cu.cc \
#       -I $TF_INC -D GOOGLE_CUDA=1 -x cu -Xcompiler -fPIC -arch=sm_50

#g++ -std=c++11 -shared -o feature_extrapolating.so feature_extrapolating_op.cc \
#       feature_extrapolating_op.cu.o -I $TF_INC -fPIC -lcudart -L $CUDA_PATH/lib64
#cd ..

#################