Skip to content

A CUDNN minimal deep learning training code sample using LeNet.

Notifications You must be signed in to change notification settings

ljlabs/cudnn-training

 
 

Repository files navigation

A CUDNN minimal deep learning training code sample using LeNet.

Prerequisites

Set the CUDNN_PATH environment variable to where CUDNN is installed.

Compilation

The project can either be compiled with CMake (cross-platform) or Visual Studio.

To enable gflags support, uncomment the line in CMakeLists.txt. In the Visual Studio project, define the macro USE_GFLAGS.

Running

Extract the MNIST training and test set files (*-ubyte) to a directory (if gflags are not used, the default is the current path).

You can also use the pre-trained weights published along with CUDNN, using the "pretrained" flag.

About

A CUDNN minimal deep learning training code sample using LeNet.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Cuda 89.4%
  • C++ 10.6%