Help Required Building Tensorflow with CMAKE

TensorFlow General Discussion
, ,
1

Hi all,
I hope somebody can help me with my frustration before I finally give up on tensorflow and go back to opencv.
I have followed the guide ‘Build with cmake’, step by step, and get absolutely no where quickly.
I can get the source, make the build directory, then use the build command ```
cmake …/tensorflow_src/tensorflow/lite -DTFLITE_ENABLE_INSTALL=ON

And it fails, on everything I think. 
- Configuring done (1.1s)
CMake Error in tools/cmake/modules/ml_dtypes/CMakeLists.txt:
  Target "ml_dtypes" INTERFACE_INCLUDE_DIRECTORIES property contains path:

    "/home/steff/tfbuild/ml_dtypes"

  which is prefixed in the build directory.


CMake Error in tools/cmake/modules/ml_dtypes/CMakeLists.txt:
  Target "ml_dtypes" INTERFACE_INCLUDE_DIRECTORIES property contains path:

    "/home/steff/tfbuild/ml_dtypes/ml_dtypes"

  which is prefixed in the build directory.


CMake Error: install(EXPORT "tensorflow-liteTargets" ...) includes target "tensorflow-lite" which requires target "absl_flags" that is not in any export set.
CMake Error: install(EXPORT "tensorflow-liteTargets" ...) includes target "tensorflow-lite" which requires target "absl_hash" that is not in any export set.
CMake Error: install(EXPORT "tensorflow-liteTargets" ...) includes target "tensorflow-lite" which requires target "absl_status" that is not in any export set.
CMake Error: install(EXPORT "tensorflow-liteTargets" ...) includes target "tensorflow-lite" which requires target "absl_strings" that is not in any export set.
CMake Error: install(EXPORT "tensorflow-liteTargets" ...) includes target "tensorflow-lite" which requires target "absl_synchronization" that is not in any export set.
CMake Error: install(EXPORT "tensorflow-liteTargets" ...) includes target "tensorflow-lite" which requires target "absl_variant" that is not in any export set.
CMake Error: install(EXPORT "tensorflow-liteTargets" ...) includes target "tensorflow-lite" which requires target "ruy" that is not in any export set.
CMake Error: install(EXPORT "tensorflow-liteTargets" ...) includes target "tensorflow-lite" which requires target "pthreadpool" that is not in any export set.
CMake Error: install(EXPORT "tensorflow-liteTargets" ...) includes target "tensorflow-lite" which requires target "XNNPACK" that is not in any export set.
-- Generating done (0.8s)
CMake Generate step failed.  Build files cannot be regenerated correctly.



Can anybody help, is there a propper set of instruction maybe somewhere Im missing. Im guessing that this is due to a whole load of missed pre-reqa, but the documentation fails to mention these.....

Help Please...?

Thanks Steff
3

OK so bit more info, I have reverted through the branches / tags a bit, and found this error occurs in all branches after 2.10, which does build, but fails to install the package to any usr/lib or usr/local.
I can build to .a, and I assume if I give the correct toolchain I can cross compile too. Its annoying though that the install package doesnt seem to work, as it would have been a cleaner solution.

5

OK so cross compiling works, but the c++ project still cannot find the package with find_package(tensorflow-lite CONFIG), any help would be really really appreciated.
I guess my alternative to this frustration is to simply convert the model to load with opencv dnn module.

7

Hi @Steffany_Raphael, Could please follow the steps mention in this document for building tensorflow lite C++ using cmake. Thank You.

9

I met exactly the same problem when building tensorflow-lite. Is any solution available? Thanks !

11

I had the exact same problem when trying to build with TensorFlow lite. Directories, which I just set, were not allowing me to add a sub-directory and throwing errors. I simple set the path will adding the sub-directory and it worked. Not the best option, but it worked for me.

# Set the main Directory and add TensorFlow Lite as a subdirectory with its own unique binary directory
set(TFLM_MICRO_BUILD_DIR (add_subdirectory("${TFLM_DIR}/lite" ${CMAKE_BINARY_DIR})))