Segmentation fault (core dumped) (CUDA9.0，C++)

Questions
#1

I compile the mxnet models to TVM.

  1. target = ‘llvm’
    -the inference was correct both with python and C++
  2. target = ‘cuda’
    -the inference was correct with python when I set target=tvm.gpu()
    -with C++, I set device_type = kDLGPU; and I get the executable program. When I ran the program, Segmentation fault (core dumped) occurred. Have anyone encountered the same problem ???