TangXing / mxnet

An efficient and flexible distributed framework for deep learning

Home Page:http://mxnet.rtfd.org

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

MXNet

Build Status Documentation Status GitHub Stats Hex.pm

This is a project that combines lessons and ideas we learnt from cxxnet, minerva and purine2.

  • The interface is designed in collaboration by authors of three projects.
  • Nothing is yet working

Guidelines

  • Use google c style
  • Put module header in include
  • Depend on dmlc-core
  • Doxygen comment every function, class and variable for the module headers
  • Minimize dependency, if possible only depend on dmlc-core
  • Macro Guard CXX11 code by
    • Try to make interface compile when c++11 was not avaialable(but with some functionalities pieces missing)
#include <dmlc/base.h>
#if DMLC_USE_CXX11
  // c++11 code here
#endif
  • Update the dependencies by
git submodule foreach --recursive git pull origin master
  • For heterogenous hardware support (CPU/GPU). Hope the GPU-specific component could be isolated easily. That is too say if we use USE_CUDA macro to wrap gpu-related code, the macro should not be everywhere in the project.

About

An efficient and flexible distributed framework for deep learning

http://mxnet.rtfd.org

License:Other


Languages

Language:C++ 72.7%Language:Python 20.2%Language:C 4.5%Language:Makefile 1.2%Language:Cuda 1.0%Language:Shell 0.4%