阿葛廷

人一定要靠自己!

1 minute read

Tags: , , ,

一個風和日麗即將過年的星期一早晨　:sunny::sunny::sunny:

想到上禮拜的快樂 AI 分享，如果要快速建置一個 ml‵ 環境，用 docker 是很完美的解決方案。

耶～

tensorflow 官網，有很多詳細介紹使用。

我用 jupyter notebbok 的版本。

docker run -it -p 8888:8888 tensorflow/tensorflow:nightly-py3-jupyter

大致上是這樣

  • 抓官方的 image
    • run image
      • docker container run --name yuting-tf -p 8888:8888 tensorflow/tensorflow:nightly-py3-jupyter
    • exec 進 container
      • docker container exec -it yuting-tf bash
    • 安裝 pandas, seaborn, sklearn (官方沒有)
      • pip install pandas seaborn skleran
    • 讓此狀態 (官方tensorflow image + pip install pandas, seaborn, skleran) 下的 container 變成一image
      • docker commit
        • docker commit <container-id/name>
  • build 新的 tag
    • 查看剛剛　docker commit
      • docker image ls 會看到一個剛新建只有 IMAGE ID 的 
        • PS E:\> docker image ls
REPOSITORY                 TAG                   IMAGE ID            CREATED             SIZE
<none>                     <none>                9d651abad0b1        6 seconds ago       2.97GB
...
    • docker image tag <source-id/name> <target-image-name>
      • 把剛剛 commit image id 當成 source 來源 新build 一個 image
        • docker iamge tag 9d651abad0b1 yuting-tensorflow
    • 看現在 image list 情況 
      • PS E:\> docker image ls
REPOSITORY                 TAG                   IMAGE ID            CREATED             SIZE
yuting-tensorflow          latest                9d651abad0b1        2 minutes ago       2.97GB

  • 上傳至 Docker Hub

    • 把剛剛已經 build 好的 image 照 Docker Hub 規則 <usr>/<repo>:<tag> 新 build 一個 image

      • docker image tag yuting-tensorflow tim23656/tensorflow:yuting-v1

      • 可以清楚看到 IMAGE ID 是一模一樣的唷～

        •  PS E:\> docker image tag yuting-tensorflow tim23656/tensorflow:yuting-v1
 PS E:\> docker image ls
 REPOSITORY                 TAG                   IMAGE ID            CREATED             SIZE
 yuting-tensorflow          latest                9d651abad0b1        4 minutes ago       2.97GB
 tim23656/tensorflow        yuting-v1             9d651abad0b1        4 minutes ago       2.97GB

    • push 上 Docker Hub

        PS E:\> docker image push tim23656/tensorflow:yuting-v1
  The push refers to repository [docker.io/tim23656/tensorflow]
  39cf1631e553: Pushed                                                                                                    
  96ae1e489cb8: Mounted from tensorflow/tensorflow                                                                        
  dd29d83dc0e3: Mounted from tensorflow/tensorflow                                                                        
  33fa6610ad9f: Mounted from tensorflow/tensorflow                                                                        
  aee8200659e9: Mounted from tensorflow/tensorflow                                                                        
  30e6f3b975fd: Mounted from tensorflow/tensorflow                                                                        
  fdd7eaf8fa52: Mounted from tensorflow/tensorflow                                                                        
  a3390fe65a10: Mounted from tensorflow/tensorflow                                                                        
  cbb1460d1298: Mounted from tensorflow/tensorflow                                                                        
  9fd8ecc50338: Mounted from tensorflow/tensorflow                                                                        
  f6f58c106474: Mounted from tensorflow/tensorflow                                                                        
  8c797d8ac223: Mounted from tensorflow/tensorflow                                                                        
  334ca2803159: Mounted from tensorflow/tensorflow                                                                        
  e443f5e673a6: Mounted from tensorflow/tensorflow                                                                        
  b97be3bb8a66: Mounted from tensorflow/tensorflow                                                                        
  f7753afc08bb: Mounted from tensorflow/tensorflow                                                                        
  2141677d99ae: Mounted from tensorflow/tensorflow                                                                        
  204d46de2d69: Mounted from tensorflow/tensorflow                                                                        
  5e10c362b98c: Mounted from tensorflow/tensorflow                                                                        
  1f21658a9230: Mounted from tensorflow/tensorflow                                                                        
  d89d90ba791e: Mounted from tensorflow/tensorflow                                                                        
  caf9b8f602b4: Mounted from tensorflow/tensorflow                                                                        
  f851662e7833: Mounted from tensorflow/tensorflow                                                                        
  918efb8f161b: Mounted from tensorflow/tensorflow                                                                        
  27dd43ea46a8: Mounted from tensorflow/tensorflow                                                                        
  9f3bfcc4a1a8: Mounted from tensorflow/tensorflow                                                                        
  2dc9f76fb25b: Mounted from tensorflow/tensorflow                                                                        
  yuting-v1: digest: sha256:87c2d98f22efc603ef5db2b65c7eb29bde27b9d1a55dfe40d721fea565134e49 size: 5964

  • 我的 Docker Hub

其他

  • 如果想要讓 container ml 的環境 (PYTHON: jupyter notebbok) 吃到本機的資料夾？

    • docker container run -v $PWD:/tf -w /tf -p 8888:8888 tim23656/tensorflow:yuting-v1
      • -v: volume bind mounting
      • -w: -w, –workdir string Working directory inside the container

You May Also Enjoy