cuDNN

Overview

• cuDNN: 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.

  • documentation - release notes:

Using

Use the module name cudnn to discover versions available and to load the application.

Note: This library is tied in directly with a version of the cudatoolkit (and subsequently the compile the compile the toolkit was built with). Call ml from the command line to see what has been loaded within your environment.

The library does expect to be run against a GPU - make sure to request one within your job submission.