Deep Learning Frameworks
深層学習フレームワーク
https://developer.nvidia.com/deep-learning-frameworks
|
|
Deep learning framework by the BVLC (Berkeley Vision and Lerning Center)
http://caffe.berkeleyvision.org/
NVIDIA® Caffe
https://developer.nvidia.com/taxonomy/term/690
http://caffe.berkeleyvision.org/installation.html
Caffe is a deep learning framework made with expression, speed, and modularity in mind. Caffe is developed by the Berkeley Vision and Learning Center (BVLC), as well as community contributors and is popular for computer vision.
Caffe supports cuDNN v5 for GPU acceleration.
（まだ、CuDNN v6 は、正式サポートされていません）
Supported interfaces: C, C++, Python（インストール時に Version 2 / 3 を要選択）, MATLAB, Command line interface
Learning Resources
•Deep learning course: Getting Started with the Caffe Framework
•Blog: Deep Learning for Computer Vision with Caffe and cuDNN
http://caffe.berkeleyvision.org/installation.html
■ Ubuntu 12.04 の場合
Download ZIP
https://github.com/BVLC/caffe/
Install git
$ cd
$ sudo apt-get install git
Install caffe
$ git clone https://github.com/BVLC/caffe.git
Compilation with Make
$ cd caffe
$ cp Makefile.config.example Makefile.config
cnDNN を使用して高速化をする場合は、Makefile.config の Line 5 の 冒頭の#を取ってください。
# USE_CUDNN := 1
↓
USE_CUDNN := 1
●Compilation
Make コマンドにより、コンパイルしてください。
$ make all
6-Core/12-Thread 対応のハイエンド・デスクトップ用の Intel@ Core™ i7 processor （i7-6980K、i7-6985K等）の場合、-j12 と指定しますと、プロセッサの全ての Thread を使用して最大の能力を使って、高速にコンパイルができます。
$ make -j12 all
$ make -j12 test
$ make -j12 runtest
以上のテストが全てOKとなれば、 Caffe の環境構築は終了となります。
Ubunt 16.04 の場合は、NVIDIA DIGITS 4.0 / 5.0 を ソースからビルドするときに、こちらのページの指示に従って、Caffe をビルドしておく必要があります。
Build Caffe from source using Git
最新版の ubuntu 16.04.3 日本語版では、何とか コンパイルは通過しても、リンカー ld が異常終了して、Core Dump してしまいます。（BLVC、NVIDIA版の両方とも）
どちらかというと、NVIDIA版の方が、警告の表示が少なく、お勧めです。
どうしても、日本語の環境で ld が異常終了してしまう場合、英語版を試してみてください。
|
|
Caffe2 is a deep learning framework enabling simple and flexible deep learning.
Built on the original Caffe, Caffe2 is designed with expression, speed, and modularity in mind, allowing for a more flexible way to organize computation.
Caffe2 supports cuDNN v5.1 for GPU acceleration.
Supported interfaces: C++, Python
Learning Resources
Download Caffe2 Download cuDNN
|
|
https://docs.microsoft.com/en-us/cognitive-toolkit/Setup-CNTK-on-your-machine
Microsoft® The Computational Network Toolkit
unified deep-learning toolkit from Microsoft Research that makes it easy to train and combine popular model types across multiple GPUs and servers.
The Microsoft Cognitive Toolkit —previously known as CNTK— is a unified deep-learning toolkit from Microsoft Research that makes it easy to train and combine popular model types across multiple GPUs and servers. Microsoft Cognitive Toolkit implements highly efficient CNN and RNN training for speech, image and text data.
Microsoft Cognitive Toolkit supports cuDNN v5.1 for GPU acceleration.
Supported interfaces: Python, C++, C# and Command line interface
Download CNTK Download cuDNN
|
|
Googleが開発した データフローグラフを用いた DeepLearning、人工知能、多層ニューラルネットワーク、数値計算のためのソフトウェアライブラリ。2015年11月、オープンソース公開されました。
https://www.tensorflow.org/install/#download-and-setup
Software library for numerical computation using data flow graphs, developed by Google’s Machine Intelligence research organization.
TensorFlow is a software library for numerical computation using data flow graphs, developed by Google’s Machine Intelligence research organization.
TensorFlow supports cuDNN v5.1 for GPU acceleration.
Supported interfaces: C++, Python
Download TensorFlow Download cuDNN
|
|
http://deeplearning.net/software/theano/
Theano is a math expression compiler that efficiently defines, optimizes, and evaluates mathematical expressions involving multi-dimensional arrays.
Theano supports cuDNN v5 for GPU acceleration.
Supported interfaces: Python
Learning resources
•Deep learning course: Getting Started with the Theano Framework
https://pypi.python.org/pypi/Theano
Download Theano-0.8.2.tar.gz
Download Theano-0.9.0.tar.gz
Optimizing compiler for evaluating mathematical expressions on CPUs and GPUs.
Theano is a Python library that allows you to define, optimize, and efficiently evaluate mathematical expressions involving multi-dimensional arrays.
It is built on top of NumPy.
|
|
http://torch.ch/
Torch is a scientific computing framework that offers wide support for machine learning algorithms.
Torch supports cuDNN v5 for GPU acceleration.
Supported interfaces: C, C++, Lua
Learning resources
•Deep learning course: Getting Started with the Torch Framework
•Blog: Understanding Natural Language with Deep Neural Networks Using Torch
https://github.com/soumith/cudnn.torch
■ Ubuntu 12.04 の場合
Download ZIP
https://github.com/torch/torch7
torch7-nv は、0.9.92 から 0.9.98-1+cuda7.5 にアップデート
Ubunt 16.04 の場合は、NVIDIA DIGITS 4.0 / 5.0 を ソースからビルドするときに、こちらのページの指示に従って、torch7 をビルドしておくおくことが、推奨されています。（インストールしませんと、DIGITS がエラーとなることが多々ありますので、基本的には必ずインストール）
Build torch from source using Git
|
|
http://mxnet.readthedocs.io/en/latest/how_to/build.html
Deep learning framework designed for both efficiency and flexibility that allows you to mix the flavors of symbolic programming and imperative programming to maximize efficiency and productivity.
MXnet is a deep learning framework designed for both efficiency and flexibility that allows you to mix the flavors of symbolic programming and imperative programming to maximize efficiency and productivity.
MXnet supports cuDNN v5 for GPU acceleration.
Supported Interfaces: Python, R, C++, Julia
Download MXnet Download cuDNN
|
|
Preferred Networks／Preferred Infrastructure が開発した日本製の深層学習フレームワーク。
、ニューラルネットワークを実装するためのライブラリ
define-by-runの原則の上に設計されており、ランタイムでのネットワーク修正が可能で、任意の制御フローステートメントを使えるようになります
http://chainer.org/
A Powerful, Flexible, and Intuitive Framework of Neural Networks
Pythonで動作し、実行にはCUDAが必須となります。
https://github.com/pfnet/chainer
Deep learning framework that’s designed on the principle of define-by-run.
$ sudo pip install chainer
Chainer is a deep learning framework that’s designed on the principle of define-by-run. Unlike frameworks that use the define-and-run approach, Chainer lets you modify networks during runtime, allowing you to use arbitrary control flow statements.
Chainer supports cuDNN v5.1 for GPU acceleration.
Supported Interfaces: Python
Download Chainer Download cuDNN
|
BIDMach
|
Welcome to the BID Data Project! Here you will find resources for the fastest Big Data tools on the Web. See our Benchmarks on github. BIDMach running on a single GPU-equipped host holds the records for many common machine learning problems, on single nodes or clusters.
Try It! BIDMach is an interactive environment designed to make it extremely easy to build and use machine learning models. BIDMach runs on Linux, Windows 7&8, and Mac OS X, and we have a pre-loaded Amazon EC2 instance. See the instructions in the Download Section.
Develop with it. BIDMach includes core classes that take care of managing data sources, optimization and distributing data over CPUs or GPUs. It’s very easy to write your own models by generalizing from the models already included in the Toolkit.
Download
http://bid2.berkeley.edu/bid-data-project/download/
Download Source
https://github.com/BIDData/BIDMach
The BID Data Suite is freely available under a BSD-style license. Complete bundles for Linux, Windows and Mac are available here. BIDMat is a self-contained matrix toolkit. BIDMach uses BIDMat as a library but includes it in the bundle, so you dont need to download BIDMat as well.
The basic bundles require Java 1.7, and for GPU use, CUDA 7.0. If you want to use the IScala notebooks, you need a recent version of IPython installed. The current release is Version 1.0.3.
•BIDMach 1.0.3 for 64-bit Linux
•BIDMach 1.0.3 for 64-bit windows
•BIDMach 1.0.3 for 64-bit Mac OSX
•BIDMat 1.0.3 for 64-bit Linux
•BIDMat 1.0.3 for 64-bit windows
•BIDMat 1.0.3 for 64-bit Mac OSX
Full bundles which include the CUDA 7.0 runtime.
•BIDMach 1.0.3 for 64-bit Linux
•BIDMat 1.0.3 for 64-bit Linux
CUDA 7.5、8.0 を使用する場合は、Github よりソースを入手し、コンパするする必要あり。
Source code is available on Github here:
https://github.com/BIDData/BIDMach
•BIDMat
•BIDMach
|
Keras
|
Keras は，Python で書かれた、TensorFlow または Theano 上で実行可能な高水準のニューラルネットワークライブラリです。
Keras は，迅速な実験を可能にすることに重点を置いて開発されました。
Keras は 次のような深層学習ライブラリが必要な場合に適しています。:
・簡単で早くプロトタイプ作成が可能 (全体的なモジュール性，ミニマリズム，および拡張性による)
・CNNとRNNの両方，およびこれらの2つの組み合わせをサポート
・任意の接続方式 (複数入力および複数出力の学習を含む) をサポート
・CPUとGPUでシームレスな実行
Keras は Python 2.7-3.5 と互換性があります．
https://keras.io/ （English）
https://keras.io/ja/ （日本語）
GitHub fcholet Keras
https://github.com/fchollet/keras
Download Source from GitHub
https://github.com/fchollet/keras/archive/master.zip
|
