rlabuguen / cxxnet

move forward to https://github.com/dmlc/mxnet

We move forward to MXNet !

---

Dear users,

Thanks for using and supporting cxxnet. Today, we finally make a hard but exciting decision: we decide to deprecate cxxnet and fully move forward to next generation toolkit MXNet.

Please check the feature highlights, speed/memory comparation and examples in MXNet.

cxxnet developers,

28th, Sep, 2015

---

Note: We provide a very simple converter to MXNet. Check guide to see whether your model is able to be converted.

---

#cxxnet

CXXNET is a fast, concise, distributed deep learning framework.

Contributors: https://github.com/antinucleon/cxxnet/graphs/contributors

• Documentation
• Learning to use cxxnet by examples
• Note on Code
• User Group(TODO)

###Feature Highlights

• Lightweight: small but sharp knife
  • cxxnet contains concise implementation of state-of-art deep learning models
  • The project maintains a minimum dependency that makes it portable and easy to build
• Scale beyond single GPU and single machine
  • The library works on multiple GPUs, with nearly linear speedup
  • THe library works distributedly backed by disrtibuted parameter server
• Easy extensibility with no requirement on GPU programming
  • cxxnet is build on mshadow
  • developer can write numpy-style template expressions to extend the library only once
  • mshadow will generate high performance CUDA and CPU code for users
  • It brings concise and readable code, with performance matching hand crafted kernels
• Convenient interface for other languages
  • Python interface for training from numpy array, and prediction/extraction to numpy array
  • Matlab interface

News

• 24-May, 2015: Pretrained Inception model with 89.9% Top-5 Correctness is ready to use.
• 09-Apr, 2015: Matlab Interface is ready to use

Backbone Library

CXXNET is built on MShadow: Lightweight CPU/GPU Tensor Template Library

• MShadow is an efficient, device invariant and simple tensor library
  • MShadow allows user to write expressions for machine learning while still provides
  • This means developer do not need to have knowledge on CUDA kernels to extend cxxnet.
• MShadow also provides a parameter interface for Multi-GPU and distributed deep learning
  • Improvements to cxxnet can naturally run on Multiple GPUs and being distributed

###Build

• Copy make/config.mk to root foler of the project
• Modify the config to adjust your enviroment settings
• Type ./build.sh to build cxxnet