The Nerv Toolkit User Manual

NOTE: This readme is obsolete and will be rearranged, for further information, please check http://nerv-sjtu.github.io/nerv/

This user manual will information about how to use Nerv and Nerv's interface.

How to make and start using

First make sure you have lua and CUDA installed on your computer.
Nerv is currently developed via github.You can download and make Nerv by doing the following:

cd ~
git clone https://github.com/Nerv-SJTU/nerv.git
cd nerv
git submodule init && git submodule update
make
#To include some new CUDA feature(e.x. atomicCAS), use "make CFLAGS=-D__NERV_FUTURE_CUDA_7"

#further, if you want the speech modules
git clone https://github.com/Nerv-SJTU/nerv-speech.git speech 
make speech

The git submodule command is for the luajit repository inside Nerv.
Now, you can try to run some example scripts.

./install/bin/nerv examples/cumatrix_example.lua

To get an example of DNN(for ASR) training, run(this requires the speech modules)
You need to be at or (copy files from) /slfs1(SJTU speechlab cluster) to get this running.

./install/bin/nerv nerv/examples/asr_trainer.lua nerv/examples/swb_baseline.lua

How to contribute

Fork the original repository, then use the pull&merge function in github to contribute.
The pull&merge request can be found on your dashboard in github. See this sync-help to sync with the original repository.

Nerv Packages

• luaT
  Nerv uses luaT(a Torch library) to define lua class in C.
• The Nerv OOP
  Enables object-oriented programming in Nerv.
• The Nerv utility functions
  Inlcudes some utility functions from luaT to implement Nerv.Class.
• The Nerv Matrix Package
  The matrix package is a basic package in Nerv that is used to store and manipulate matrices.
• The Nerv IO Package
  The IO package is used to read and write parameters to file.
• The Nerv Parameter Package
  The parameter package is used to store, read model parameters from file.
• The Nerv Layer Package
  The layer package is used to define propagation and backpropagation of different type of layers.
• The Nerv NN Package
  The nn package is for organizing a neural network, it contains nerv.LayerRepo, nerv.ParamRepo, and nerv.DAGLayer.