Failed to detect number of GPUs: {e}

[Corey Hahn]

After an error free installation I am getting the following error.

$ python
Python 3.10.8 (main, Nov 24 2022, 14:13:03) [GCC 11.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import cupy as cp
>>> cp.show_config()
/home/cah/anaconda3/envs/cupy/lib/python3.10/site-packages/cupyx/_runtime.py:279: UserWarning: Failed to detect number of GPUs: cudaErrorUnknown: unknown error
  warnings.warn(f'Failed to detect number of GPUs: {e}')
OS                        : Linux-5.15.0-56-generic-x86_64-with-glibc2.31
Python Version            : 3.10.8
CuPy Version              : 11.4.0
CuPy Platform             : NVIDIA CUDA
NumPy Version             : 1.23.5
SciPy Version             : 1.9.3
Cython Build Version      : 0.29.32
Cython Runtime Version    : None
CUDA Root                 : /usr/local/cuda/bin
nvcc PATH                 : None
CUDA Build Version        : 11080
CUDA Driver Version       : 11060
CUDA Runtime Version      : CUDARuntimeError('cudaErrorUnknown: unknown error')
cuBLAS Version            : (available)
cuFFT Version             : 10600
cuRAND Version            : 10209
cuSOLVER Version          : (11, 3, 2)
cuSPARSE Version          : (available)
NVRTC Version             : (11, 6)
Thrust Version            : 101501
CUB Build Version         : 101501
Jitify Build Version      : 4a37de0
cuDNN Build Version       : 8600
cuDNN Version             : 8101
NCCL Build Version        : 21505
NCCL Runtime Version      : 2804
cuTENSOR Version          : None
cuSPARSELt Build Version  : None

Linux xxx 5.15.0-56-generic #62~20.04.1-Ubuntu SMP Tue Nov 22 21:24:20 UTC 2022 x86_64 x86_64 x86_64 GNU/Linux

Nvidia Driver 510.47.03
CUDA Version: 11.6  

(I install with 11.6 via conda as well).

 new conda environment python 3.10
conda install -c conda-forge cupy cudatoolkit=11.6

no errors on installation

[Corey Hahn]

NVIDIA GeForce RTX 3090