junosan / fractal Goto Github PK

===========================================================================
         Fractal: Flexible Recurrent ArChitecture TrAining Library
===========================================================================


   Copyright 2015 Kyuyeon Hwang

   Licensed under the Apache License, Version 2.0 (the "License");
   you may not use this file except in compliance with the License.
   You may obtain a copy of the License at

       http://www.apache.org/licenses/LICENSE-2.0

   Unless required by applicable law or agreed to in writing, software
   distributed under the License is distributed on an "AS IS" BASIS,
   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
   See the License for the specific language governing permissions and
   limitations under the License.



1. Prerequisites

  - g++ 4.7 or higher
  - CUDA Toolkit 7.0 or higher
  - GNU Autotools (only if you need to re-configure libfractal)
  - NVIDIA GPU with Fermi architecture or higher



2. Quick start guide

  2.1. Install libfractal (local)

    - Generate libfractal library files in build/lib/
    - Generate libfractal header files in build/include/fractal/

    $ ./build-local.sh

    TROUBLESHOOTING: aclocal version mismatch
      (Required: GNU Autotools)
      $ cd libfractal
      $ ./autoreconf


  2.2. Compile and run "movingAverage" example

    - Compute the moving average of recent 10 binary input values

    $ cd example/movingAverage/src
    $ make
    $ cd ../run
    $ ./run.sh


  2.3. Compile and run "languageModel" example

    - Learn plain text for character-level language modeling
    - Compute P(x_{t+1} | x_{1:t})
    - Can be run as a generative model

    $ cd example/languageModel/src
    $ make
    $ cd ../run
    $ ./train_wiki.sh

    After or during the training, we can test the generative model by

    $ ./textGen.sh



3. Install libfractal (global)

  - Install libfractal in the global zone (e.g. /usr/local/lib/, /usr/local/include/)

  $ cd libfractal
  $ mkdir build
  $ cd build
  $ ../configure
  $ make
  $ sudo make install



4. How to use the library

  - Include the Fractal header file in your C++ source code:
    #include <fractal/fractal.h>

  - Currently, you need to specify where the CUDA include directory is:
    $ g++ -o foo.o -c foo.cpp -I/usr/local/cuda/include

  - Link with -lfractal:
    $ g++ -o foo foo.o -lfractal