Terry Heo

For Bazel build, you need to provide `--macos_cpus=arm64` So the following command might work. ``` $ CUSTOM_BAZEL_FLAGS=--macos_cpus=arm64 build_pip_package_with_bazel.sh ``` For CMake build, you need to use it as following. ```...

If you're using `tensorflow/lite/tools/pip_package/build_pip_package_with_cmake.sh` you'd better check if `${PYTHON_INCLUDE}` and `${PYBIND11_INCLUDE}` are defined properly. https://github.com/tensorflow/tensorflow/blob/master/tensorflow/lite/tools/pip_package/build_pip_package_with_cmake.sh#L30 https://github.com/tensorflow/tensorflow/blob/master/tensorflow/lite/tools/pip_package/build_pip_package_with_cmake.sh#L114

1. Could you update C++ style? https://www.tensorflow.org/community/contribute/code_style#c_coding_style 2. Please resolve the conflict on tensorflow/lite/CMakeLists.txt

The patch looks a bit hacky to me. @robert-kalmar could you share what kind of error do you have in which compiler?

Thanks for the contribution. I've merged a similar patch. Plz let me know if you still have an issue.

I'm seeing the following build error. ``` [100%] Built target pthreadpool -- Downloading FXdiv to /tmpfs/tensor_flow/tflite_build/FXdiv-source (define FXDIV_SOURCE_DIR to avoid it) -- Configuring done -- Generating done -- Build files...

This PR still has an error. ``` CMake Error at [tools/cmake/modules/OverridableFetchContent.cmake:566](https://cs.corp.google.com/piper///depot/google3/tools/cmake/modules/OverridableFetchContent.cmake?l=566&cl=507511477) (message): Required license file not found for opengl_headers Call Stack (most recent call first): [tools/cmake/modules/opengl_headers.cmake:36](https://cs.corp.google.com/piper///depot/google3/tools/cmake/modules/opengl_headers.cmake?l=36&cl=507511477) (OverridableFetchContent_Populate) [tools/cmake/modules/Findopengl_headers.cmake:16](https://cs.corp.google.com/piper///depot/google3/tools/cmake/modules/Findopengl_headers.cmake?l=16&cl=507511477) (include) CMakeLists.txt:262...

The error happens with GPU delegate enabled. $ cmake ../tensorflow/lite -DTFLITE_ENABLE_GPU=ON -DTFLITE_KERNEL_TEST=ON

@arvinjhou thanks for your contribution. Could you share if this PR is still needed with ToT TFLite? If yes, could you elaborate your issue?

Updating build toolchain makes a regression. I've reverted the change and will revisit this.