cuDNN Libraries

Hi Everybody,

I see many options for cuDNN as below. What are their differences? I am using Ubuntu 16.04 so that should I install the last three? or is the first one is enough instead of other three? Thanks in advance.

cuDNN v7.1.3 Library for Linux

cuDNN v7.1.3 Runtime Library for Ubuntu16.04 (Deb)

cuDNN v7.1.3 Developer Library for Ubuntu16.04 (Deb)

cuDNN v7.1.3 Code Samples and User Guide for Ubuntu16.04 (Deb)

Best regards, Ender.

https://stackoverflow.com/questions/48784645/which-nvidia-cudnn-release-type-for-tensorflow-on-ubuntu-16-04/48788965#48788965

Thank you.

Dear txbob,

I have just installed cuDNN according to the Nvidia Guide. (for .tgz). To verify the installation, I checked for the address of

/usr/src/

to see cudnn_samples_v9.1, however this directory is empty. May you please enlight me about this issue?

King regards, Ender.

There isn’t going to be any cudnn_samples_v9.1

However just run this command to locate where the install put it:

sudo ls -R / |grep cudnn_samples

output of the command that you provided is:

ls: cannot open directory ‘/run/user/1000/gvfs’: Permission denied

ok try this then:

sudo ls -R / | grep cudnn

output became:

cudnn-9.1-linux-x64-v7.1.tgz
cudnn.h
libcudnn.so
libcudnn.so.7
libcudnn.so.7.1.3
libcudnn_static.a
ls: cannot open directory ‘/run/user/1000/gvfs’: Permission denied
cudnn.h
libcudnn.so
libcudnn.so.7
libcudnn.so.7.1.3
libcudnn_static.a

try:

sudo find / -name cudnn.h

output:

find: ‘/run/user/1000/gvfs’: Permission denied
/usr/local/cuda-9.1/targets/x86_64-linux/include/cudnn.h
/home/deep/Downloads/cuda/include/cudnn.h

Note: during the installation, I copied extracted files into cuda-9.1 instead of cuda in the local file. Can that be the problem?

Your install looks OK. However it appears that the .tgz file doesn’t contain the sample codes for this version of CUDNN. So if you are looking for the samples, you will need to grab the deb installer that contains the samples.

cuDNN v7.1.3 Code Samples and User Guide for Ubuntu16.04 (Deb)

Sorry for the misinformation, things seem to have changed.

If it is installed successfully as you checked. It is alright for me. Thank you for the support :)

We created a new “Deep Learning Training and Inference” section in Devtalk to improve the experience for deep learning and accelerated computing, and HPC users:
https://devtalk.nvidia.com/default/board/301/deep-learning-training-and-inference-/

We are moving active deep learning threads to the new section.

URLs for topics will not change with the re-categorization. So your bookmarks and links will continue to work as earlier.

-Siddharth