Hi,
I am trying to build TVM from source, but I am facing some problems when enabling the tflite support. Without tflite I am able to complete the building process (libtvm.so).
Any insight would be greatly appreciated.
Thanks
Expected behavior
Building TVM from source with tflite enable
Actual behavior
When “make” try to link libtensorflow-lite.a, the process fails to do to some undefined references, the first being
/usr/bin/ld: /home/user/TVM/tflite_build/libtensorflow-lite.a(rfft2d.cc.o): in function
tflite::ops::builtin::rfft2d::Rfft2dImpl(int, int, double**, int*, double*)': rfft2d.cc:(.text+0x916): undefined reference to
rdft2d’
Here is the full cmake log: TVM cmake log - Pastebin.com
Environment
- Ubuntu 20.04 kernel 5.13.0-41
- Python 3.8.10
- GCC,G++ 11.1
- LLVM 10
- CUDA 11.7
- cuDNN 8.4
- ninja 1.11.0.git
- NNPack from source (compiled as https ://tvm.apache.org/docs/install/nnpack.html)
- Tensorflow lite from source (compiled as https ://www.tensorflow.org/lite/guide/build_cmake#create_a_cmake_project_which_uses_tensorflow_lite)
Steps to reproduce
I am following the steps from TVM docs as explained here:
https ://tvm.apache.org/docs/install/from_source.html
This is my config.cmake file: config.cmake - Pastebin.com