Open rspezialetti opened 6 years ago
facing a similar error. It has probably got to do something with version changes in the internal dependencies of TF 1.3.
Its like all the headerfiles aren't being included before the shared library.
here's an example:
/usr/local/lib/libtensorflow_all.so: undefined reference to
tensorflow::strings::FloatToBuffer(float, char*)'
`
all the functions that are causing errors are defined in headers that we copy to the system, I confirmed that. Maybe a link or a directory change is not accounted for in the bazel config?
^plausible. Another reason: https://github.com/tensorflow/tensorflow/issues/13607
I am facing the same problem here. Any suggestions please?
Same for me !
Been stuck in this problem for two days. Using the following command works for me. Then follow the rest of the instructions as per stated.
bazel build --config=opt --config=cuda --config=monolithic tensorflow:libtensorflow_all.so
I was referring to the following websites to arrive at this solution: creating-tensorflow-c-headers-and-libraries Installing Tensorflow from Sources Issue #35
I was also facing the 'Protobuf undefined reference to LogMessage ...'. For me, it is solved by capitalizing the Protobuf in the FindProtobuf.cmake file and CMakeLists.txt file, i.e.
Protobuf_LIBRARIES
to be rewritten as PROTOBUF_LIBRARIES
Protobuf_INCLUDE_DIRS
to be rewritten as PROTOBUF_INCLUDE_DIRS
Again, I was refering to this website.
I am using Tensorflow in ROS, hence the usage of CMakeLists in the provided solution.
When i try to use libtensorflow_all.so in my project i got 1048 error, like this: undefined reference to `tensorflow::TensorShapeBase::TensorShapeBase(tensorflow::gtl::ArraySlice)'
Tesorflow version is 1.3