Issues with Cuda when trying to configure tensorflow

Hey guys,

I’m trying to get tensorflow installed on my Windows computer. I’ve got a NVIDIA Quadro M2200 and have installed CUDA 9.2 and CUDNN 7.3.1.

However, a few steps in to python ./configure.py, when i set the compute capability to default[3.5,7.0], i get:

Traceback (most recent call last):
File “./configure.py”, line 1693, in
main()
File “./configure.py”, line 1655, in main
raise UserInputError('SYCL / CUDA / ROCm are mututally exclusive. ’
main.UserInputError: SYCL / CUDA / ROCm are mututally exclusive. At most 1 GPU platform can be configured.

Im fairly vanilla when it comes to the command prompt and what not, so treat me as a dummy.

Best regards

Hello
I met same problems.

I find out the code of tensorflow and find out the solution:
“if environ_cp.get(‘TF_NEED_OPENCL_SYCL’) == ‘1’:
gpu_platform_count += 1
if environ_cp.get(‘TF_NEED_ROCM’) == ‘1’:
gpu_platform_count += 1
if environ_cp.get(‘TF_NEED_CUDA’) == ‘1’:
gpu_platform_count += 1
if gpu_platform_count >= 2:
raise UserInputError('SYCL / CUDA / ROCm are mututally exclusive. ’
‘At most 1 GPU platform can be configured.’)”

I choose all of SYCL\ROC\CUDA so things went wrong.

Next time, I only choose CUDA and correct this problem.

3 Likes