#Cmake install install#

How can I enable hexagon support with the CMake Installation of the TfLite ? All hexagon libraries are on board. Will moving the android_rc19_ndk to a directory on board the RB5 and setting ANDROID_NDK_HOME successfully compile the NNAPI delegate ? Will this enable NNAPI correctly? Because now it does not believe it skips that compilation because it cannot find the correct headers (not installed yet)Д«asically I have the following questions: I have downloaded the Hexagon SDK and will copy the android_ndk _rc19 workspace to somewhere on board and set the path. There is no documentation using cmake to enable hexagon support in the TensorFlow lite library and I see no flags in the CMakeLists.txt.

#Cmake install install#

The RB5 has hexagon libraries already installed and I will also install the libhexagon_nn_skel.so libraries. The GPU delegate causes the program to fail when specifying it (maybe not linking correct adreno library to the inferencing code).could use some help there)Д«ut mostly I am concerned about building NNAPI and HEXAGON delegate support. However the only delegate that works is the XNN_PACK delegate. I have the following flags enabled -DTFLITE_ENBLE_XNN_PACK -DTFLITE_-DTFLITE_ENABLE_GPU -DTFLITE_ENABLE_NNAPIДЄnd I am able to compile a working(ish) library and do inferencing. I have been using CMAKE / GCC installation natively on board. Hello, I am having trouble understanding the tutorials and getting delegate support for Tensorflow lite c++/c library on the Qualcomm RB5. GCC/Compiler version (if compiling from source): gcc 7.5.Bazel version (if compiling from source):.Installed using virtualenv? pip? conda?: no.

As per our GitHub Policy, we only address code/doc bugs, performance issues, feature requests and build/installation issues on GitHub. Please make sure that this is a build/installation issue.