CuDNN library

From MetaCentrum
Jump to navigation Jump to search


The NVIDIA CUDA® Deep Neural Network library (cuDNN) is a GPU-accelerated library of primitives for deep neural networks. cuDNN provides highly tuned implementations for standard routines such as forward and backward convolution, pooling, normalization, and activation layers. cuDNN is part of the NVIDIA Deep Learning SDK.


Development tools and environments


Versions and modules:

cudnn-7.4.2-cuda10         (for CUDA 10.0)
cudnn-7.1.4-cuda90         (for CUDA 9.0)
cudnn-7.0                  (for CUDA 8.0 and later)
cudnn-6.0                  (for CUDA 7.5 and later)
cudnn-5.1                  (for CUDA 7.5 and later)
cudnn-5.0                  (for CUDA 7.5 and later)
cudnn-4.0                  (for CUDA 7.0 and later)

Notice: This is a licenced software. If you want to use it, you must confirm the licence form. But first you have to accept a licence at NVIDIA's site.
Notice 2: These are the standalone modules. Usually you need to use it with some of CUDA modules.

Supporting GPU clusters

CuDNN only works on GPUs with high enough computing capabilities. In this table, you can see information about individual GPU clusters and if their GPUs support CuDNN library:

GPU clusters in MetaCentrum
Cluster Nodes GPUs per node Compute Capability CuDNN gpu_cap= - 2x nVidia Tesla T4 16GB 7.5 YES cuda20,cuda35,cuda61,cuda75 - 2x nVidia Tesla K20 5GB (aka Kepler) 3.5 YES cuda20,cuda35 - 4x GPU nVidia GeForce GTX 1080 Ti 6.1 YES cuda20,cuda35,cuda61 - 2x nVidia Tesla K20Xm 6GB (aka Kepler) 3.5 YES cuda20,cuda35 - nVidia 1080Ti GPU 6.1 YES cuda20,cuda35,cuda61 nVidia 1080Ti GPU 6.1 YES cuda20,cuda35,cuda61 nVidia TITAN V GPU 7.0 YES cuda20,cuda35,cuda61,cuda70 nVidia Tesla K40 3.5 YES cuda20,cuda35 nVidia Tesla P100 6.0 YES cuda20,cuda35, cuda60


You have to be registered in NVIDIA Accelerated Computing Developer Program and agree with their licence. Then confirm the licence form


module load cudnn-7.0

To plan your job on clusters with certain Compute Capability, use qsub command like this:

qsub -q gpu -l select=1:ncpus=1:ngpus=X:gpu_cap=cuda35 <job batch file>



Program manager