Closed cygerts closed 5 years ago
Ok, resolved it. It turned out that the that compilation error was a result of previous error: error: constexpr function return is non-constant
Solution is to add -DNDEBUG flag to nvcc compilationm -> https://github.com/tensorflow/tensorflow/issues/22766
While running make.sh this error occurs:
1 error detected in the compilation of "/tmp/tmpxft_00000ff5_00000000-6_psroi_pooling_op_gpu.cu.cpp1.ii". g++: error: psroi_pooling_op.cu.o: No such file or directory
All paths (PATH, LD_LIBRARY_PATH) set up correctly, Cuda 9.0, Cudnn 7.0, gcc 5.5, Ubuntu 18.04, Python 3.6, TF - 1.12.0
Related to this issue: https://github.com/CharlesShang/TFFRCNN/issues/34 But those solutions do not work here