Undefined symbol in roi_pooling.so

[yakovenkodenis]

I'm trying to run the demo following the instructions in readme, however, when I run the command

python ./tools/demo.py --model ./lib/pretrained/VGGnet_fast_rcnn_iter_70000.ckpt

I get the error below:

➜ Faster-RCNN_TF git:(master) ✗ python ./tools/demo.py --model ./lib/pretrained/VGGnet_fast_rcnn_iter_70000.ckpt
I tensorflow/stream_executor/dso_loader.cc:128] successfully opened CUDA library libcublas.so locally
I tensorflow/stream_executor/dso_loader.cc:128] successfully opened CUDA library libcudnn.so locally
I tensorflow/stream_executor/dso_loader.cc:128] successfully opened CUDA library libcufft.so locally
I tensorflow/stream_executor/dso_loader.cc:128] successfully opened CUDA library libcuda.so.1 locally
I tensorflow/stream_executor/dso_loader.cc:128] successfully opened CUDA library libcurand.so locally
Traceback (most recent call last):
File "./tools/demo.py", line 11, in
from networks.factory import get_network
File "/home/denis/WEB/DeepLearning/Faster-RCNN_TF/tools/../lib/networks/init.py", line 8, in
from .VGGnet_train import VGGnet_train
File "/home/denis/WEB/DeepLearning/Faster-RCNN_TF/tools/../lib/networks/VGGnet_train.py", line 2, in
from networks.network import Network
File "/home/denis/WEB/DeepLearning/Faster-RCNN_TF/tools/../lib/networks/network.py", line 3, in
import roi_pooling_layer.roi_pooling_op as roi_pool_op
File "/home/denis/WEB/DeepLearning/Faster-RCNN_TF/tools/../lib/roi_pooling_layer/roi_pooling_op.py", line 5, in
_roi_pooling_module = tf.load_op_library(filename)
File "/usr/local/lib/python2.7/dist-packages/tensorflow/python/framework/load_library.py", line 64, in load_op_library
None, None, error_msg, error_code)
tensorflow.python.framework.errors_impl.NotFoundError: /home/denis/WEB/DeepLearning/Faster-RCNN_TF/tools/../lib/roi_pooling_layer/roi_pooling.so: undefined symbol: _ZN10tensorflow7strings6StrCatB5cxx11ERKNS0_8AlphaNumE

---

CUDA: 8.0
CuDNN: 5
Python: 2.7.12
Tensorflow: 0.12.0-rc1
GPU: NVidia GeForce 750M (sm_30 architecture)

Due to my setup above, I modified CUDA_PATH in make.sh file to be like this:

CUDA_PATH=/usr/local/cuda-8.0/

and the nvcc instruction to be like this:

	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 \
		-arch=sm_30

---

Am I doing something wrong?

Could you please help me with running the demo properly?

[civilman628]

update g++ line as below: (add -D_GLIBCXX_USE_CXX11_ABI=0 )

g++ -std=c++11 -shared -o roi_pooling.so roi_pooling_op.cc \
	roi_pooling_op.cu.o -I $TF_INC  -D GOOGLE_CUDA=1 -fPIC $CXXFLAGS  -D_GLIBCXX_USE_CXX11_ABI=0 \
	-lcudart -L $CUDA_PATH/lib64

[civilman628]

and re-run make

[yakovenkodenis]

@civilman628 Thank you, that worked perfectly!

[mathetes87]

@civilman628 in what file should I add that argument?

[mathetes87]

Found it: make.sh file in lib folder

[mathetes87]

But now I get this error:

g++: error: roi_pooling_op.cc: No such file or directory
g++: error: roi_pooling_op.cu.o: No such file or directory

Running make without @civilman628 's code it finishes without issues

[mathetes87]

Nevermind, I was mixing roi_pooling_op section with psroi_pooling_op...

[showaichuan]

@mathetes87 Hi， I got the same problem ,so how did you solve this issue at last?

[mathetes87]

@showaichuan If you look in the file make.sh, there are 2 sections, one for roi_pooling_op and one for psroi_pooling_op. My error was that I was editing the wrong one. Try looking very carefully so as to edit only the one corresponding to roi_pooling_op

[tp227]

@civilman628 thank you very much! I run the demo.py successfully!

[WLpub]

@civilman628 thank you very much!

[zdm123]

@civilman628 Hi, I added "-D_GLIBCXX_USE_CXX11_ABI=0" as you said. But still have the same problem. How can fix it?

[jakeoung]

Me too. Same problem..

[tpatel0409]

Yeah, Even I faced the same problem, adding the extra argument in g++ is not helping.
I am using intel core i7 4th gen CPU and I don't have graphic card installed.

I tried adding up the same flag in g++ in the else branch in make.sh but still does not help.
Thanks

[aiqu]

tensorflow/tensorflow#13607

It might help

[danaodai]

but my make.sh file has no psroi_pooling_op section。so how to sovle the problem?

[UpCoder]

@civilman628 Thanks, It works for me!
Tensorflow: 1.3.0
TITAN X -arch=sm_52
and add -D_GLIBCXX_USE_CXX11_ABI=0

[JiteshPshah]

no matter what tensorflow you have, what gcc you this will work :
follow this link: http://windstop.tistory.com/155