Skip to content
This repository has been archived by the owner on Nov 23, 2021. It is now read-only.
/ magmadnn Public archive
forked from MagmaDNN/magmadnn

basic NN framework in c++

License

Notifications You must be signed in to change notification settings

ckf42/magmadnn

 
 

Repository files navigation

MagmaDNN GCN

This repo is forked from the C++ neural network library MagmaDNN (commit hash 8e80c066ee52aa07b0f73eb0e9e55e22fcb857a2).

In this repo, a graph convolution layer (or to be more specific the one developed by Kipf & Welling) is implemented. The GPU implementation relies on CUDA CuBlas library.

At the moment the implementation uses dense matrix multiplications in batches and hence supports only graphs in dense format. Sparse graphs may be supported if Magma/CuSparse/other libraries have the corresponding routines or native routines are implemented.

Other graph convolution layers and pooling layers may be added in the future.

As this is forked from the dev branch of MagmaDNN, bugs may exist and features in this repo may be removed in upcoming MagmaDNN versions.

author: Kam Fai Chan

original author: Daniel Nichols

original co-author: Sedrick Keh

About

basic NN framework in c++

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • C++ 87.4%
  • Cuda 8.9%
  • Makefile 1.8%
  • Shell 1.3%
  • C 0.6%