Unable to load cudnn library

Could not load library libcudnn_cnn_infer.so.8. Error: libnvrtc.so: cannot open shared object file: No such file or directory

How can I install CuDNN and carry out the remaining training steps? Please provide the steps in detail.

Thanks in advance.

Hi, the cudnn libraries should be installed in the images used when launching with jobman with the latest-cuda tag. For example: jobman submit -i ubuntu-python-pytorch:latest-cuda -r small-gpu -- ... jobman submit -i ubuntu-python-tensorflow:latest-cuda -r small-gpu -- ...

With the option "-i " we indicate to use one of the GPU-ready images (with libraries and tools for it), and with the option "-r " we indicate in which resource we want the job to be launched.

It's possible that you're trying to run on the desktop itself (and not as a job) or with a normal image that does not include these libraries.

If, knowing this, the problem persists, please send us the execution line (with the image used) to our email chaimeleon@quibim.com so that we can reproduce the problem and/or check the image. Hope we could help!

FYI. I did use jobman as intended and was able to get a GPU. The error occured when I try to train the model (model.fit(...))

Hi, we still need to know what image you use, if you put the latest-cuda tag at the end.

Hi,

I've emailed chaimeleon@quibim.com with an explanation of the details you mentioned. I hope you look into this and get back to me soon.