深度学习框架Caffe在Mac上的安装和测试

cluster 2016-06-22

深度学习框架介绍

先概括一下深度学习的几大流行的框架:Pylearn2, Theano, Caffe, Torch, Cuda-covnet,Deeplarning4j等。

  • Theano是一个Python库,也是一个强大的数学表达式编译器。Pylearn2是在Theano基础上建立的机器学习库。用户可以用数学表达式写Pylearn2的插件(新的model, algorithm等), Theano将这些表达式进行优化和稳定化,然后进行编译。
  • Caffe是由Berkely Vision and Learning Center的贾杨清博士(毕业后在谷歌工作)主导开发的基于ConvNets和C++的深度学习库。Caffe允许网络模型和优化方法都定义在配置文件中而不需要写代码,可以很方便地在CPU和GPU之间切换。
  • Torch更偏向企业级应用,是用Lua写的,Facebook AI实验室和Google DeepMind团队等都使用Torch。可以为机器学习算法提供类似于Matlab的环境。Lua可以轻易地与C结合,任何C或者C++库都可以成为Lua库。OverFeat是用Torch7在ImageNet上训练得到的特征提取工具。
  • Cuda-convnet或者CuDNN是NVIDIA提供的基于GPU加速的深度学习工具,对主流的软件包括Caffe,Torch和Theano都提供支持。
  • Deeplarning4j面向商业应用,是基于Java的机器学习框架。更多介绍可阅读各自的网站或者阅读这篇文章

Caffe的安装

    Caffe的网站上提供了安装说明。由于其依赖的库比较多,通常安装过程会出现许多问题,在不同的机器和操作系统上可能遇到不同的问题。安装时可以根据网站上提供的说明步骤进行,遇到有问题时用Google搜索一下基本都能找到。本文记录了笔者在Mac上安装遇到的问题和解决办法。系统版本:OS X 10.9.5。

    1,安装Caffe的依赖库

    1.1 安装CUDA。推荐7.0以上版本,6.*版本也可以。我安装的是最新版CUDA 7.5。

    1.2 安装BLAS。这里我使用了OpenBLAS。推荐使用brew安装:brew install openblas

    1.3 安装Boost。

          通过brew install boost默认安装版本为1.60。但建议使用1.59。因为1.60编译后可能会出现问题。 

$ brew search boost
boost                                    homebrew/versions/boost-python159 ✔    
boost-bcp                                homebrew/versions/boost149             
boost-build                              homebrew/versions/boost150             
boost-python                             homebrew/versions/boost155             
homebrew/science/boost-compute           homebrew/versions/boost159 ✔           
Caskroom/cask/iboostup                   Caskroom/cask/turbo-boost-switcher     
Caskroom/cask/pivotalbooster
$ brew install –build-from-source homebrew/versions/boost159

 

 

            安装好后可以后在/usr/local/opt/boost159下看到该库。Caffe中把某些依赖库所在的文件夹名字限定为boost,可以将/usr/local/opt/boost159复制粘贴产生备份,将备份改名为/usr/local/opt/boost。

 

    1.4 安装CuDNN。下载cuDNN v5.0版本。解压后将include和bin文件夹中的内容分别复制到/usr/lcoal 下面的/include和/bin中。

    1.5 使用brew install 分别安装 protobuf, glog, gflags, hdf5, snappy, leveldb, szip, lmdb等。

          如果使用python, protobuf安装命令为

$ brew install --build-from-source --with-python -vd protobuf<code> </code><code></code>

 

    1.6  (可选)OpenCV, 我使用2.4.6版本。

    1.7  (可选)Python 版本:2.7。

              需要安装numpy。推荐使用Anaconda,里面包含了一个python版本2.7.11并且包含了大多数所需要的库,包括hdf5、numpy等。Anaconda默认安装在$(HOME)/anaconda目录下。

              还需要安装python-boost。与boost类似的方法,推荐1.59版本。

    1.8  (可选)Matlab 版本 2015a

    2,安装Caffe

    2.1 下载Caffe后在caffe-master文件夹下,以Makefile.config.example为模板根据第一步中的安装情况,建立配置文件Makefile.config,内容如下:

## Refer to http://caffe.berkeleyvision.org/installation.html
# Contributions simplifying and improving our build system are welcome!

# cuDNN acceleration switch (uncomment to build with cuDNN).
USE_CUDNN := 1

# CPU-only switch (uncomment to build without GPU support).
# CPU_ONLY := 1

# uncomment to disable IO dependencies and corresponding data layers
# USE_OPENCV := 0
# USE_LEVELDB := 0
# USE_LMDB := 0

# uncomment to allow MDB_NOLOCK when reading LMDB files (only if necessary)
#    You should not set this flag if you will be reading LMDBs with any
#    possibility of simultaneous read and write
ALLOW_LMDB_NOLOCK := 1

# Uncomment if you're using OpenCV 3
OPENCV_VERSION := 2.4

# To customize your choice of compiler, uncomment and set the following.
# N.B. the default for Linux is g++ and the default for OSX is clang++
CUSTOM_CXX := clang++

# CUDA directory contains bin/ and lib/ directories that we need.
CUDA_DIR := /usr/local/cuda
# On Ubuntu 14.04, if cuda tools are installed via
# "sudo apt-get install nvidia-cuda-toolkit" then use this instead:
# CUDA_DIR := /usr

# CUDA architecture setting: going with all of them.
# For CUDA < 6.0, comment the *_50 lines for compatibility.
CUDA_ARCH := -gencode arch=compute_20,code=sm_20 \
        -gencode arch=compute_20,code=sm_21 \
        -gencode arch=compute_30,code=sm_30 \
        -gencode arch=compute_35,code=sm_35 \
        -gencode arch=compute_50,code=sm_50 \
        -gencode arch=compute_50,code=compute_50

# BLAS choice:
# atlas for ATLAS (default)
# mkl for MKL
# open for OpenBlas
BLAS := open
# Custom (MKL/ATLAS/OpenBLAS) include and lib directories.
# Leave commented to accept the defaults for your choice of BLAS
# (which should work)!
# BLAS_INCLUDE := /usr/local/opt/openblas/include
# BLAS_LIB := /usr/local/opt/openblas/lib

# Homebrew puts openblas in a directory that is not on the standard search path
BLAS_INCLUDE := $(shell brew --prefix openblas)/include
BLAS_LIB := $(shell brew --prefix openblas)/lib

# This is required only if you will compile the matlab interface.
# MATLAB directory should contain the mex binary in /bin.
# MATLAB_DIR := /usr/local
MATLAB_DIR := /Applications/MATLAB_R2015a.app

# NOTE: this is required only if you will compile the python interface.
# We need to be able to find Python.h and numpy/arrayobject.h.
#PYTHON_INCLUDE := /usr/include/python2.7 \
        /Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/numpy/core/include
# Anaconda Python distribution is quite popular. Include path:
# Verify anaconda location, sometimes it's in root.
ANACONDA_HOME := $(HOME)/anaconda
PYTHON_INCLUDE := $(ANACONDA_HOME)/include \
          $(ANACONDA_HOME)/include/python2.7 \
          $(ANACONDA_HOME)/lib/python2.7/site-packages/numpy/core/include \

# Uncomment to use Python 3 (default is Python 2)
# PYTHON_LIBRARIES := boost_python3 python3.5m
# PYTHON_INCLUDE := /usr/include/python3.5m \
#                 /usr/lib/python3.5/dist-packages/numpy/core/include

# We need to be able to find libpythonX.X.so or .dylib.
#PYTHON_LIB := /usr/lib
PYTHON_LIB := $(ANACONDA_HOME)/lib

# Homebrew installs numpy in a non standard path (keg only)
# PYTHON_INCLUDE += $(dir $(shell python -c 'import numpy.core; #print(numpy.core.__file__)'))/include
# PYTHON_LIB += $(shell brew --prefix numpy)/lib
#PYTHON_LIB +=/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/numpy/core/lib
PYTHON_LIB +=$(ANACONDA_HOME)/lib/python2.7/site-packages/numpy/core/lib

# Uncomment to support layers written in Python (will link against Python libs)
WITH_PYTHON_LAYER := 1

# Whatever else you find you need goes here.
INCLUDE_DIRS := $(PYTHON_INCLUDE) /usr/local/Cellar/boost159/1.59.0/include
LIBRARY_DIRS := $(PYTHON_LIB) /usr/local/cuda/lib /usr/local/Cellar/boost159/1.59.0/lib /usr/local/opt/boost-python159/lib

# If Homebrew is installed at a non standard location (for example your home directory) and you use it for general dependencies
INCLUDE_DIRS += $(shell brew --prefix)/include
LIBRARY_DIRS += $(shell brew --prefix)/lib

# Uncomment to use `pkg-config` to specify OpenCV library paths.
# (Usually not necessary -- OpenCV libraries are normally installed in one of the above $LIBRARY_DIRS.)
# USE_PKG_CONFIG := 1

BUILD_DIR := build
DISTRIBUTE_DIR := distribute

# Uncomment for debugging. Does not work on OSX due to https://github.com/BVLC/caffe/issues/171
# DEBUG := 1

# The ID of the GPU that 'make runtest' will use to run unit tests.
TEST_GPUID := 0

# enable pretty build (comment to see full commands)
Q ?= @

 

 


    2.2,命令行进入caffe-master文件夹下,运行:

 

 

$ make all

    出现问题:

PROTOC src/caffe/proto/caffe.proto
make: protoc: No such file or directory

    解决办法:  需要用brew建立protobuf的链接。为此,运行

$ brew link protobuf

    如果运行上述命令又出问题,比如brew的权限问题:permission denied for /usr/local。需要设置一下权限,更新一下brew, 为此,运行

$ sudo chown -R $USER:admin /usr/local
$ cd /usr/local
$ git reset --hard origin/master
$ brew update

  上述问题可以得到解决。

 

  2.3 上一步通过后,运行

$ make test

  这一步没问题。将build_release/lib下的所有文件复制到/usr/local/lib

$ cp -a .build_release/lib/. /usr/local/lib/

  再运行

$ make runtest

 

 报错:

.build_release/tools/caffe
dyld: Library not loaded: @rpath/libcudart.7.5.dylib
Referenced from: /Developer/caffe/.build_release/tools/caffe
Reason: image not found

  为此需要设置一下环境变量DYLD_FALLBACK_LIBRARY_PATH

$ export DYLD_FALLBACK_LIBRARY_PATH=/usr/local/cuda/lib:/usr/local/lib:$(HOME)/<span style="color:black;">anaconda/lib</span>

  再运行make runtest,一切顺利。

 

    2.4 如果使用python,再运行

<pre name="code" class="html">$ make pycaffe
$ make pytest

 

 

 

  3, 运行mnist的例子。

  详细步骤见:http://caffe.berkeleyvision.org/gathered/examples/mnist.html

    3.1,下载mnist数据。在caffe-master目录下运行

$ ./data/mnist/get_mnist.sh

 

    3.2,建立训练数据和测试数据,运行

 

$ ./examples/mnist/create_mnist.sh

    出现以下错误,说convert_mnist_data.bin找不到:

Creating lmdb...
./examples/mnist/create_mnist.sh: line 16: build/examples/mnist/convert_mnist_data.bin: No such file or directory
./examples/mnist/create_mnist.sh: line 18: build/examples/mnist/convert_mnist_data.bin: No such file or directory
Done.

    解决办法:搜索convert_mnist_data.bin发现该文件位于./distribute/bin目录下,因此在在./examples/mnist/create_mnist.sh文件中将BUILDdistribute/bin即可

 

    3.3,训练和测试,运行:

$ ./examples/mnist/create_mnist.sh

    如果出现和上面类似的错误,说caffe找不到 (caffe.bin位于./distribute/bin目录下或者build/tools下),检查create_mnist.sh的内容,保证caffe.bin的路径正确

./distribute/bin/caffe.bin train--solver=examples/mnist/lenet_solver.prototxt

 

    然后就能看到运行结果了。

      4, 在python中使用caffe的例子

    详见:http://nbviewer.jupyter.org/github/BVLC/caffe/blob/master/examples/00-classification.ipynb

    该例子用Caffe中已经训练好的模型(基于Alexnet的结构)对图像进行分类。并且可以显示不同层中训练得到的特征。

相关推荐