2015.08.17 Ubuntu 14.04+cuda 7.5+caffe安装配置

2016.06.10 update cuda 7.5 and cudnn v5

2015.10.23更新：修改了一些地方，身边很多人按这个流程安装，完全可以安装

折腾了两个星期的caffe，windows和ubuntu下都安装成功了。其中windows的安装配置参考官网推荐的那个blog，后来发现那个版本的caffe太老，和现在的不兼容，一些关键字都不一样，果断回到Linux下。这里记录一下我的安装配置流程。

电脑配置：

ubuntu 14.0464bit

8G 内存

GTX650显卡

软件版本：

CUDA 7.0

caffe 当天从github下载的版本

安装ubuntu的过程省略，建议安装后关闭自动更新，上一次安装caffe后用的很好，结果有一天晚上没关电脑，自己半夜更新了显卡驱动，然后...

caffe的安装流程主要参考这个blog，稍有改动：Caffe + Ubuntu 14.04 64bit + CUDA 6.5 配置说明

Caffe 安装配置步骤：

1， 安装开发所需的依赖包

[plain] view plain copy print ?

1. sudoapt-getinstallbuild-essential#basicrequirement
2. sudoapt-getinstalllibprotobuf-devlibleveldb-devlibsnappy-devlibopencv-devlibboost-all-devlibhdf5-serial-devlibgflags-devlibgoogle-glog-devliblmdb-devprotobuf-compiler#requiredbycaffe

Before install CUDA 7.5,you need update gcc 4.8+ to gcc 4.9+

reference:update gcc/g++

2，安装CUDA 7.5

验证过程省略，按照官方文档自己操作吧（遇到问题首先要看官方文档啊，血泪教训）

安装CUDA有两种方法，

离线.run安装：从官网下载对应版本的.run安装包安装，安装过程挺复杂，尝试过几次没成功，遂放弃。

在离线.deb安装：deb安装分离线和在线，我都尝试过都安装成功了，官网下载地址

安装之前请先进行md5校验，确保下载的安装包完整

切换到下载的deb所在目录，执行下边的命令

[plain] view plain copy print ?

1. sudodpkg-icuda-repo-<distro>_<version>_<architecture>.deb
2. sudoapt-getupdate
3. sudoapt-getinstallcuda

然后重启电脑：sudo reboot

NOTE:装不成功卸了多来几遍，总会成的

3，安装cuDNN

下载cudnn-7.5-linux-x64-v5.0-ga.tgz，官网申请不到，网上自己找的，就不给地址了。

[plain] view plain copy print ?

1. tar-zxvfcudnn-7.5-linux-x64-v5.0-ga.tgz
2. cdcuda
3. sudocplib/lib*/usr/local/cuda/lib64/
4. sudocpinclude/cudnn.h/usr/local/cuda/include/

更新软连接

cd /usr/local/cuda/lib64/
sudo chmod +r libcudnn.so.5.0.5
sudo ln -sf libcudnn.so.5.0.5 libcudnn.so.5
sudo ln -sf libcudnn.so.5 libcudnn.so
sudo ldconfig

4，设置环境变量

在/etc/profile中添加CUDA环境变量

sudo gedit /etc/profile

[plain] view plain copy print ?

1. PATH=/usr/local/cuda/bin:$PATH
2. exportPATH

保存后,执行下列命令,使环境变量立即生效

[plain] view plain copy print ?

1. source/etc/profile

同时需要添加lib库路径： 在 /etc/ld.so.conf.d/加入文件 cuda.conf,内容如下

[plain] view plain copy print ?

1. /usr/local/cuda/lib64

保存后，执行下列命令使之立刻生效

[plain] view plain copy print ?

1. sudoldconfig

5，安装CUDA SAMPLE

进入/usr/local/cuda/samples,执行下列命令来build samples

[plain] view plain copy print ?

1. sudomakeall-j4

整个过程大概10分钟左右,全部编译完成后， 进入 samples/bin/x86_64/linux/release,运行deviceQuery

[plain] view plain copy print ?

1. ./deviceQuery

如果出现显卡信息， 则驱动及显卡安装成功：

[plain] view plain copy print ?

1. ./deviceQueryStarting...
3. CUDADeviceQuery(RuntimeAPI)version(CUDARTstaticlinking)
5. Detected1CUDACapabledevice(s)
7. Device0:"GeForceGTX670"
8. CUDADriverVersion/RuntimeVersion6.5/6.5
9. CUDACapabilityMajor/Minorversionnumber:3.0
10. Totalamountofglobalmemory:4095MBytes(4294246400bytes)
11. (7)Multiprocessors,(192)CUDACores/MP:1344CUDACores
12. GPUClockrate:1098MHz(1.10GHz)
13. MemoryClockrate:3105Mhz
14. MemoryBusWidth:256-bit
15. L2CacheSize:524288bytes
16. MaximumTextureDimensionSize(x,y,z)1D=(65536),2D=(65536,65536),3D=(4096,4096,4096)
17. Maximumlayered1DTextureSize,(num)layers1D=(16384),2048layers
18. Maximumlayered2DTextureSize,(num)layers2D=(16384,16384),2048layers
19. Totalamountofconstantmemory:65536bytes
20. Totalamountofsharedmemoryperblock:49152bytes
21. Totalnumberofregistersavailableperblock:65536
22. Warpsize:32
23. Maximumnumberofthreadspermultiprocessor:2048
24. Maximumnumberofthreadsperblock:1024
25. Maxdimensionsizeofathreadblock(x,z):(1024,1024,64)
26. Maxdimensionsizeofagridsize(x,z):(2147483647,65535,65535)
27. Maximummemorypitch:2147483647bytes
28. Texturealignment:512bytes
29. Concurrentcopyandkernelexecution:Yeswith1copyengine(s)
30. Runtimelimitonkernels:Yes
31. IntegratedGPUsharingHostMemory:No
32. Supporthostpage-lockedmemorymapping:Yes
33. AlignmentrequirementforSurfaces:Yes
34. DevicehasECCsupport:disabled
35. DevicesupportsUnifiedAddressing(UVA):Yes
36. DevicePCIBusID/PCIlocationID:1/0
37. ComputeMode:
38. <Default(multiplehostthreadscanuse::cudaSetDevice()withdevicesimultaneously)>
40. deviceQuery,CUDADriver=CUDART,CUDADriverVersion=6.5,CUDARuntimeVersion=6.5,NumDevs=1,Device0=GeForceGTX670
41. Result=PASS

NOTE:上边的显卡信息是从别的地方拷过来的，我的GTX650显卡不是这些信息，如果没有这些信息，那肯定是安装不成功，找原因吧！

6，安装Intel MKL 或Atlas

我没有MKL，装的Atlas

安装命令：

[plain] view plain copy print ?

1. sudoapt-getinstalllibatlas-base-dev

7，安装OpenCV

我安装的是2.4.10

1）下载 安装脚本

2）进入目录 Install-OpenCV/Ubuntu/2.4

3）执行脚本

[plain] view plain copy print ?

1. sudosh./opencv2_4_10.sh

8，安装Caffe所需要的Python环境

按caffe官网的推荐使用Anaconda

去Anaconda官网下载安装包
切换到文件所在目录，执行

[plain] view plain copy print ?

1. bashAnaconda-2.3.0-Linux-x86_64.s<em>h</em>

NOTE:后边的文件名按自己下的版本号更改，整个安装过程请选择默认

8.1，添加Anaconda Library Path

在/etc/ld.so.conf最后加入以下路径，并没有出现重启不能进入界面的问题( NOTE:下边的username要替换)

[plain] view plain copy print ?

1. /home/username/anaconda/lib

在~/.bashrc最后添加下边路径

[plain] view plain copy print ?

1. exportLD_LIBRARY_PATH="/home/username/anaconda/lib:$LD_LIBRARY_PATH"

9，安装python依赖库
去caffe的github下载caffe源码包
进入caffe-master下的python目录

执行如下命令

[plain] view plain copy print ?

1. forreqin$(catrequirements.txt);dopipinstall$req;done

10，编译Caffe

终于来到这里了

进入caffe-master目录，复制一份Makefile.config.examples

[plain] view plain copy print ?

1. cpMakefile.config.exampleMakefile.config

修改其中的一些路径，如果前边和我说的一致，都选默认路径的话，那么配置文件应该张这个样子

[plain] view plain copy print ?

1. ##Refertohttp://caffe.berkeleyvision.org/installation.html
2. #Contributionssimplifyingandimprovingourbuildsystemarewelcome!
4. #cuDNNaccelerationswitch(uncommenttobuildwithcuDNN).
5. USE_CUDNN:=1
7. #cpu-onlyswitch(uncommenttobuildwithoutGPUsupport).
8. #cpu_ONLY:=1
10. #Tocustomizeyourchoiceofcompiler,uncommentandsetthefollowing.
11. #N.B.thedefaultforLinuxisg++andthedefaultforOSXisclang++
12. #CUSTOM_CXX:=g++
14. #CUDAdirectorycontainsbin/andlib/directoriesthatweneed.
15. CUDA_DIR:=/usr/local/cuda
16. #OnUbuntu14.04,ifcudatoolsareinstalledvia
17. #"sudoapt-getinstallnvidia-cuda-toolkit"thenusethisinstead:
18. #CUDA_DIR:=/usr
20. #CUDAarchitecturesetting:goingwithallofthem.
21. #ForCUDA<6.0,commentthe*_50linesforcompatibility.
22. CUDA_ARCH:=-gencodearch=compute_20,code=sm_20\
23. -gencodearch=compute_20,code=sm_21\
24. -gencodearch=compute_30,code=sm_30\
25. -gencodearch=compute_35,code=sm_35\
26. -gencodearch=compute_50,code=sm_50\
27. -gencodearch=compute_50,code=compute_50
29. #BLASchoice:
30. #atlasforATLAS(default)
31. #mklforMKL
32. #openforOpenBlas
33. BLAS:=atlas
34. #Custom(MKL/ATLAS/OpenBLAS)includeandlibdirectories.
35. #LeavecommentedtoacceptthedefaultsforyourchoiceofBLAS
36. #(whichshouldwork)!
37. #BLAS_INCLUDE:=/path/to/your/blas
38. #BLAS_LIB:=/path/to/your/blas
40. #Homebrewputsopenblasinadirectorythatisnotonthestandardsearchpath
41. #BLAS_INCLUDE:=$(shellbrew--prefixopenblas)/include
42. #BLAS_LIB:=$(shellbrew--prefixopenblas)/lib
44. #Thisisrequiredonlyifyouwillcompilethematlabinterface.
45. #MATLABdirectoryshouldcontainthemexbinaryin/bin.
46. #MATLAB_DIR:=/usr/local
47. #MATLAB_DIR:=/Applications/MATLAB_R2012b.app
49. #NOTE:thisisrequiredonlyifyouwillcompilethepythoninterface.
50. #WeneedtobeabletofindPython.handnumpy/arrayobject.h.
51. #PYTHON_INCLUDE:=/usr/include/python2.7\
52. /usr/lib/python2.7/dist-packages/numpy/core/include
53. #AnacondaPythondistributionisquitepopular.Includepath:
54. #Verifyanacondalocation,sometimesit'sinroot.
55. ANACONDA_HOME:=$(HOME)/anaconda
56. PYTHON_INCLUDE:=$(ANACONDA_HOME)/include\
57. $(ANACONDA_HOME)/include/python2.7\
58. $(ANACONDA_HOME)/lib/python2.7/site-packages/numpy/core/include\
60. #WeneedtobeabletofindlibpythonX.X.soor.dylib.
61. #PYTHON_LIB:=/usr/lib
62. PYTHON_LIB:=$(ANACONDA_HOME)/lib
64. #Homebrewinstallsnumpyinanonstandardpath(kegonly)
65. #PYTHON_INCLUDE+=$(dir$(ShellPython-c'importnumpy.core;print(numpy.core.__file__)'))/include
66. #PYTHON_LIB+=$(shellbrew--prefixnumpy)/lib
68. #UncommenttosupportlayerswritteninPython(willlinkagainstPythonlibs)
69. #WITH_PYTHON_LAYER:=1
71. #Whateverelseyoufindyouneedgoeshere.
72. INCLUDE_Dirs:=$(PYTHON_INCLUDE)/usr/local/include
73. LIBRARY_Dirs:=$(PYTHON_LIB)/usr/local/lib/usr/lib
75. #IfHomebrewisinstalledatanonstandardlocation(forexampleyourhomedirectory)andyouuseitforgeneraldependencies
76. #INCLUDE_Dirs+=$(shellbrew--prefix)/include
77. #LIBRARY_Dirs+=$(shellbrew--prefix)/lib
79. #Uncommenttouse`pkg-config`tospecifyOpenCVlibrarypaths.
80. #(Usuallynotnecessary--OpenCVlibrariesarenormallyinstalledinoneoftheabove$LIBRARY_Dirs.)
81. #USE_PKG_CONfig:=1
83. BUILD_DIR:=build
84. distribute_DIR:=distribute
86. #Uncommentfordebugging.DoesnotworkonOSXduetohttps://github.com/BVLC/caffe/issues/171
87. #DEBUG:=1
89. #TheIDoftheGPUthat'makeruntest'willusetorununittests.
90. TEST_GPUID:=0
92. #enableprettybuild(commenttoseefullcommands)
93. Q?=@

保存退出

编译

[plain] view plain copy print ?

1. makeall-j4
2. maketest
3. makeruntest

11，编译Python wrapper

[plain] view plain copy print ?

1. makepycaffe

到这里就基本结束了，跑个自带的例子测试一下吧！ NOTE:以上是我在自己PC上的安装步骤，因软件版本不同，硬件环境不同，按照以上方式可能出现错误，请耐心查找错误，欢迎留言