cuDNN Linux Verification Error (Ubuntu 22.04)

I have installed cuda11.7 and cuDNN v8.7.0. Following:

I am using 2 x NVIDIA RTX-3090 GPUs and Ubuntu-22.04 OS.

I finished the installation process without any errors, but when I tried to Verify the Install on Linux, I got the following error:

Testing single precision
Could not load library Error: cannot open shared object file: No such file or directory
Aborted (core dumped)

I have verified CUDA; it is running correctly. What can I do to interface cuDNN with CUDA?

Edit-1: From online research, I found that cudnn libraries should be looking for ‘’, and my library is searching for ‘’; does this have anything to do with the error?

Hi @harish2sista ,
Would you mind checking the support matrix to validate the compatibility.

1 Like

I resolved the issue.
I had the latest preinstalled NVIDIA drivers on my system before installing CUDA. The CUDA drivers conflicted with the preinstalled version of NVIDIA drivers, so I reinstalled the OS and let CUDA installation take of the NVIDIA drivers, and the error has been resolved.