下载tensorflow-gpu版本的源

Posted qwangxiao

tags:

篇首语:本文由小常识网(cha138.com)小编为大家整理,主要介绍了下载tensorflow-gpu版本的源相关的知识,希望对你有一定的参考价值。

  每次换了个地方就要重新配置自己的开发环境那是特别蛋疼的,尤其是要弄到服务器跑的时候,不小心把环境弄崩了是非常惨的。
  
  下载tensorflow-gpu版本的源
  
  docker pull daocloud.io/daocloud/tensorflow:0.11.0-gpu
  
  1
  
  编辑方便的脚本文件启动docker
  
  先查下你有几个GPU设备
  
  [[email protected] ~]# ls -la /dev | grep nvidia
  
  crw-rw-rw-. 1 root root 195, 0 Sep 16 13:49 nvidia0
  
  crw-rw-rw-. 1 root root 195, 255 Sep 16 13:49 nvidiactl
  
  crw-rw-rw-. 1 root root 247, 0 Sep 16 13:54 nvidia-uvm
  
  1
  
  2
  
  3
  
  4
  
  然后再查你的docker镜像
  
  [email protected]:~$ sudo docker images
  
  [sudo] password for y:
  
  REPOSITORY TAG IMAGE ID CREATED SIZE
  
  daocloud.io/daocloud/tensorflow 0.11.0-gpu dd645f420f1d 8 weeks ago 2.713 GB
  
  daocloud.io/daocloud/tensorflow 0.10.0-devel-gpu fa886c09638d 3 months ago 5.014 GB
  
  hello-world
  
  1
  
  2
  
  3
  
  4
  
  5
  
  6
  
  然后就可以启动咯
  
  sudo docker run -ti -v /home/:/mnt/home --privileged=true --device /dev/nvidia0:/dev/nvidia0 --device /dev/nvidiactl:/dev/nvidiactl --device /dev/nvidia-uvm:/dev/nvidia-uvm daocloud.io/daocloud/tensorflow:0.11.0-gpu /bin/bash
  
  1
  
  上面这句有点长把它写到docker.sh文件,然后
  
  sh docker.sh
  
  1
  
  完成。 上面的意思是把本地的/home映射到docker的/mnt目录
  
  以及各种显卡设备也映射进去
  
  进去之后
  
  别急着用tensorflow,可能会报错,因为我发现LD_LIBLABRARY_PATH环境变量设置的不对。但是又没有vim。于是先更新软件源。把软件源文件放到本机的/home,再去docker的/mnt/home里面复制到
  
  /etc/apt/sources.list
  
  1
  
  deb http://mirrors.aliyun.com/ubuntu/ trusty main multiverse restricted universe
  
  deb http://mirrors.aliyun.com/ubuntu/ trusty-backports main multiverse restricted universe
  
  deb http://mirrors.aliyun.com/ubuntu/ trusty-proposed main multiverse restricted universe
  
  deb http://mirrors.aliyun.com/ubuntu/ trusty-security main multiverse restricted universe
  
  deb http://mirrors.aliyun.com/ubuntu/ trusty-updates main multiverse restricted universe
  
  deb-src http://mirrors.aliyun.com/ubuntu/ trusty main multiverse restricted universe
  
  deb-src http://www.jyz521.com/ /ubuntu/ trusty-backports main multiverse restricted universe
  
  deb-src http://mirrors.aliyun.com/ubuntu/ trusty-proposed main multiverse restricted universe
  
  deb-src http://www.chushiyl.cn /ubuntu/ trusty-security main multiverse restricted universe
  
  deb-src http://www.huachengjpt.com /ubuntu/ trusty-updates main multiverse restricted universe
  
  然后
  
  apt-get update
  
  apt-get install 你要安装的东西
  
  1
  
  2
  
  安装常用软件
  
  在~/.bashrc里面最后加上
  
  export LD_LIBLABRARY_PATH = /usr/local/cuda-8.0/lib64:$LD_LIBLABRARY_PATH
  
  1
  
  2
  
  退出终端
  
  主机保存镜像为新版本
  
  sudo docker ps -l
  
  [email protected]:~$ sudo docker ps -l
  
  CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
  
  a1f2ac36a2c9 daocloud.io/daocloud/tensorflow:0.11.0-gpu "/bin/bash" 10 minutes ago Up 10 minutes 6006/tcp, 8888/tcp
  
  1
  
  2
  
  3
  
  4
  
  把a1f2ac36a2c9这个名字记住
  
  然后
  
  docker commit a1f2ac36a2c9 新名字
  
  1
  
  OK了
  
  把镜像存到移动硬盘里
  
  sudo docker save -o "要存的地址" daocloud.io/daocloud/tensorflow:0.11.0-gpu
  
  1
  
  加载本地的
  
  sudo docker load --input "本地地址"
  
  1
  
  删除镜像
  
  docker rmi "镜像IDa1f2ac36a2c9 "
  
  1
  
  使用tensorflow
  
  在docker里面发现import tensorflow报错
  
  I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcublas.so locally
  
  I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcudnn.so locally
  
  I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcufft.so locally
  
  I tensorflow/stream_executor/dso_loader.cc:105] Couldn‘t open CUDA library libcuda.so.1. LD_LIBRARY_PATH: /usr/local/cuda-8.0/lib64:/usr/local/nvidia/lib:/usr/local/nvidia/lib64:
  
  I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:160] hostname: ad8b0d82bec1
  
  I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:185] libcuda reported version is: Not found: was unable to find libcuda.so DSO loaded into this program
  
  I tensorflow/stream_executor/cuda/cuda_diagnostics.cc:356] driver version file contents: """NVRM version: NVIDIA UNIX x86_64 Kernel Module 367.44 Wed Aug 17 22:24:07 PDT 2016
  
  1
  
  2
  
  3
  
  4
  
  5
  
  6
  
  7
  
  估计是cuda.so这个没有集成进来,干脆直接把本机的放进docker来好了。
  
  -v /usr/lib/x86_64-linux-gnu/:/usr/lib/x86_64-linux-gnu/
  
  1
  
  这一句重新加入docker run里面去。重启docker OK
  
  [email protected]:/notebooks# python\
  
  >
  
  Python 2.7.6 (default, Jun 22 2015, 17:58:13)
  
  [GCC 4.8.2] on linux2
  
  Type "help", "copyright", "credits" or "license" for more information.
  
  >>> import tensorflow
  
  I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcublas.so locally
  
  I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcudnn.so locally
  
  I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcufft.so locally
  
  I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcuda.so.1 locally
  
  I tensorflow/stream_executor/dso_loader.cc:111] successfully opened CUDA library libcurand.so locally
  
  >>>
  
  测试一下效果
  
  ...
  
  >>> import tensorflow as tf
  
  >>> hello = tf.constant(‘Hello, TensorFlow!‘)
  
  >>> sess = tf.Session()
  
  >>> print(sess.run(hello))
  
  Hello, TensorFlow!
  
  >>> a = tf.constant(10)
  
  >>> b = tf.constant(32)
  
  >>> print(sess.run(a + b))
  
  42

以上是关于下载tensorflow-gpu版本的源的主要内容,如果未能解决你的问题,请参考以下文章

tensorflow-gpu安装的一些注意

ubuntu14.04_64位安装tensorflow-gpu

win10安装CUDA CUDNN tensorflow-gpu

记录 window10安装TensorFlow-gpu2.4.0

记录 window10安装TensorFlow-gpu2.4.0

tensorflow-gpu安装脚本