Failed call to cuDevicePrimaryCtxRetain: CUDA_ERROR_UNKNOWN

i installed tensorflow-gpu==1.12.0, NVIDIA-Linux-x86_64-390.75-grid.run and cuda_9.0.176_384.81_linux.run(not include the nvidia driver),but i run python with “from tensorflow.python.client import device_lib;print(device_lib.list_local_devices())”, it reported the error “failed initializing StreamExecutor for CUDA device ordinal 0: Internal: failed call to cuDevicePrimaryCtxRetain: CUDA_ERROR_UNKNOWN”

i can’t install nvidia driver on the host which is nvidia virtual GPU, so i install nvidia grid driver, I’m not sure that whether it works.

GRID M60-1Q doesn’t support CUDA.

https://docs.nvidia.com/cuda/vGPU/index.html

https://docs.nvidia.com/grid/latest/grid-vgpu-user-guide/index.html#cuda-open-cl-support-vgpu

I recommend asking vgpu questions on the vgpu forums.