I am a TensorFlow beginner. The environment I use is ubuntu18.04, python3.7, TensorFlow-GPU1.14, CUDA10.0.

I want to build a custom OP (using GPU), I saw from the official website tutorial that you need to include #include "tensorflow/core/util/cuda_kernel_helper.h" in the source code.
But when I was "make", I got an error: "cuda_kernel_helper.h" was not found.

In addition, I used the python environment built by anaconda. I found gpu_kernel_helper.h in the folder in the environment. I guess this is a replacement file for cuda_kernel_helper.h. When I replace the file name in my source code, another one appears. Error: "third_party/gpus/cuda/include/cuda_fp16.h" file not found.
After that, I only found the "eigen3" folder in the “third_party” folder and did not find the "gpus" folder.
I installed TensorFlow-gpu using pip. Is there a problem with my installation method?

I hope to get help to solve this problem. thank you all.