Please make sure libcudnn_ops_infer.so.8 is in your library path?

[softcatala]

Please make sure libcudnn_ops_infer.so.8 is in your library path?

[mayeaux]

This is what I use, I think it's for the same issue you're having:

echo 'export LD_LIBRARY_PATH=/opt/ctranslate2/lib:$LD_LIBRARY_PATH' >> ~/.bashrc

[silvacarl2]

$ whisper-ctranslate2 12694263-Joan-Svarczkopf.mp3 --compute_type int8_float16
Detecting language using up to the first 30 seconds. Use --language to specify the language
100%|████████████████████████████████████████| 922M/922M [00:03<00:00, 273MiB/s]
100%|█████████████████████████████████████| 2.29k/2.29k [00:00<00:00, 13.3MiB/s]
100%|███████████████████████████████████████| 449k/449k [00:00<00:00, 2.69MiB/s]
100%|█████████████████████████████████████| 2.10M/2.10M [00:00<00:00, 9.16MiB/s]
Estimating duration from bitrate, this may be inaccurate
Could not load library libcudnn_ops_infer.so.8. Error: libcudnn_ops_infer.so.8: cannot open shared object file: No such file or directory
Please make sure libcudnn_ops_infer.so.8 is in your library path!
Aborted (core dumped)

$ sudo find / -name "libcudnn_ops_infer.so.8"
/home/silvacarl/.local/lib/python3.8/site-packages/torch/lib/libcudnn_ops_infer.so.8

$ echo $LD_LIBRARY_PATH
/usr/local/cuda-11.7/targets/x86_64-linux/lib/:/usr/local/cuda-11.7/lib64:/usr/lib/x86_64-linux-gnu:/home/silvacarl/.local/lib/python3.8/site-packages/torch/lib/

$ ls /opt/ctranslate2/lib
ls: cannot access '/opt/ctranslate2/lib': No such file or directory

[Botoni]

The same happened to me using an Anaconda virtual environment, this is the solution I found:

mkdir -p $CONDA_PREFIX/etc/conda/activate.d

echo 'CUDNN_PATH=$(dirname $(python3 -c "import nvidia.cudnn;print(nvidia.cudnn.__file__)"))' >> $CONDA_PREFIX/etc/conda/activate.d/env_vars.sh

echo 'export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$CONDA_PREFIX/lib/:$CUDNN_PATH/lib' >> $CONDA_PREFIX/etc/conda/activate.d/env_vars.sh

source $CONDA_PREFIX/etc/conda/activate.d/env_vars.sh

[kanazei]

I had this error in runpod.io. The following helped:
!apt-get install libcudnn8
!apt-get install libcublas-12-1 -y
Or a specific version of cuda so you don't have to upgrade libcublas:
!apt-get install libcudnn8=8.9.2.26-1+cuda11.8 -y

Upd.
In the end, it all came down to copying.
To do this, we are looking for where our libcudnn is located:
!find / -name "libcudnn_ops_infer.so.8"
Then we look where it should be:
!apt-get install libcudnn8=8.9.2.26-1+cuda11.8 -y
!find / -name "libcudnn_ops_infer.so.8"
And in the following sessions, we simply copy all libcudnn to the desired directory.
In my case it is:
!cp /usr/local/lib/python3.10/dist-packages/torch/lib/libcudnn* /usr/lib/x86_64-linux-gnu/

[silvacarl2]

this nailed it for me:

export LD_LIBRARY_PATH=$(echo "$LD_LIBRARY_PATH:/usr/local/cuda-11.7/targets/x86_64-linux/lib/:/home/silvacarl/.local/lib/python3.8/site-packages/nvidia/cudnn/lib/")

[YusufCandraArif]

@silvacarl2 you save my day, thank you

[silvacarl2]

any time, ctranslate2 is awesome