Skip to content

cudnn: Library of deep learning/neural network primitives for GPUs

Overview of package

Package: cudnn
Description: Library of deep learning/neural network primitives for GPUs
For more information: https://developer.nvidia.com/cuda-zone
Categories: SoftwareDevelopment
MachineLearning
License: Free2Use (NVIDIA Software License)

General Usage

cuDNN is a library of primitives supporting deep learning and neural networks using NVIDIA GPUs.

The following environmental variables have been defined:

  • $CUDNN_ROOT has been set to the root of the cudnn installation
  • $CUDNN_LIBDIR points to the directory containing the libraries
  • $CUDNN_INCDIR points to the directory containing the header files

You will probably wish to use these by adding the following flags to your compilation command (e.g. to CFLAGS in your Makefile):

  • -I$CUDNN_INCDIR

and the following flags to your link command (e.g. LDFLAGS in your Makefile):

  • -L$CUDNN_LIBDIR -Wl,-rpath,$CUDNN_LIBDIR

Available versions by cluster

This section lists the available versions of the package cudnnon the UMD HPC clusters.

Versions available on the Zaratan cluster

Version Module tags CPU(s) optimized for GPU ready?
7.6.5.32-10.2 7.6.5.32-10.2 zen Yes
8.4.0.27-11.6 8.4.0.27-11.6 zen2 Yes
8.9.7.29-12 8.9.7.29-12 icelake, zen2 Yes