AWS演習（2）- DLAMI・Jupyter Notebook・TensorFlow

授業の演習で使用した資料です．一コマの演習，また2年生を対象とした演習で，1年生では専門教育を行わず，2年生から学部・学科に配属されるため，この辺りに水準を合わせています．

Transcript

1. AWS2 DLAMI
   JUPYTER NOTEBOOK
   TENSORFLOW axi-sugiki
2. +)02 ◦ Amazon EC2 ◦ Amazon '(!  ◦ Deep Learning AMI5DLAMI6 ◦ /%$21 '4"  ◦ Jupyter Notebook ◦ .-$23,1
      ◦ TensorFlow ◦ &*#
   .-$2 2
3. DLAMI4Deep Learning AMI5 ◦ Amazon
   ("+%$0-  ◦ Python /, 4Conda5 ◦ +%$0  4MXNet, TensorFlow, PyTorch, Chainer, CNTK, Keras, Caffe…5 ◦ GPU-4CUDA5 ◦ 3.#)- 4Jupyter Notebook5 3 +%$0&2*1!'("
4. DLAMI 1/3 ◦ EC2 
   Deep Learning AMI  ◦ UbuntuBase AMI    4
5. DLAMI >+@2/3A ◦ && ◦
   t2.medium@2/4GiBA?2 ◦ *.CPUB !$0:  ◦ "$%' ◦
   Web Security Group)6@SSH 0:A ◦ ' ◦ 4'(1 ◦  ◦
   Name
   dlami-@#'-A ;/ ◦
   Owner
   @#'-A ;/ 5 <7  *. 58=3,9
6. DLAMI&(3/3) ◦ !Running
   *SSH#%  "
   
   6 ! DNS IP   ($'%) AMI
7. SSH Port Forwarding&Tunneling' 7 ◦ Local%  Remote%    SSH !" # SSH"&TCP/22' 8888 localhost 8888 Local Forwarding&-L' Remote Forwarding&-R'
   Local Remote   Local%   : %    %   $ " Remote%  $
8. Tera TermSSH-2 57 ◦ (1) SSH-2 ◦ -2$EC2
    IP
    DNS ◦  (
   ec2-user ◦  1. ◦
   RSA/DSA9"/ 8+:  ,); ◦ (2) SSH 574) ◦ SSH-2*<
   4)→
   SSH57 8+ ◦
   6' 8+<!3 %& ◦
     
   88880 ◦
    #
   127.0.0.1
   88880 8
9. Jupyter Notebook$ ◦ Tera TermSSH"  $ jupyter notebook ◦ ! URL Web  # 
    9 URL Web # 
    
10. Jupyter Notebook ◦ '&-  0%) 3  ◦ Python14
    !
    9R14). ◦ Numpy7#2+65:/(+)8 ◦ Pandas7  0%:$, )8 ◦ Matplotlib7 "*8 10
11. Notebook #%! ◦ 
    New notebook! ◦ Notebook ◦ * &$'( 11 ) Python3 +" New+"  )
    Python3 +"
12. Notebook"1/5# ◦  $! 12 (1)     (2)  $Shift + Enter (3) 
    
13. Notebook%*2/5+ ◦  % 13 Notebook(Close and Halt '& #! Notebook $*)"+.ipynb   $
    +
14. Notebook ( 43/55 ◦ 2 (  14 + -/ + 3/ 4
      .25 4 ( 5 Shift+Enter&0 ( Enter Esc vi $, %) Esc1' Enter/Esc !#" *
15. Notebook /#>4/5? ◦
     $3/# ◦ ↑  k@!  5) ◦ ↓  j@"  5) ◦ a@! 08#. ◦ b@" 08#. ◦ dd@'=>d2*%(? ◦ h@/# 974 ◦ m@Markdown-,>;1:<3? &2 ◦ y@-,>+63? &2 15  %(  >Esc &2?
16. Notebook '55/56 ◦ Markdown&$ 03 ◦ #7##7###8/ 51,2,3,…6 ◦ -8-+)5,# 1.6 ◦ ***%2****8%2 16  Markdown/Code!*
    ".  41(03 
17. TensorFlow/C ◦ *
    GoogleLE >=3H#&!$& ◦ DLAMITensorFlowB2.< ◦ :@1NConda/COQ ◦ SSH%&" 'I+0NJupyter)?P,4J1O $ source activate tensorflow_p36 ◦ :@2NJupyter NotebookGDOQ ◦ (7; EnvironmentNconda_tensorflow_p36OK8 17 %% 5 %&" P 9-F4;M A6
18. 49 ◦ DLAMI TensorFlow16%/ ◦ https://www.tensorflow.org/tutorials/ ◦
    MNIST'+)$25;16 7& < ◦ :,-
    Keras16 !. ◦ #  =0( " 8 %/ ◦ =Markdown 3*16 18
19. MNIST
     1/2 ◦ Matplotlib 
     ◦  
     19
20. MNIST 2/2 ◦   
      20
21.  59< ◦ "4 Notebook(2  ◦ A. Geron, “Hands-On Machine Learning with Scikit-Learn & TensorFlow”, O’Reilly. ◦ https://github.com/ageron/handson-ml ◦ )+8%6,” & :.*,$&1”,#/ ◦ https://github.com/YutaroOgawa/Deep-Reinforcement- Learning-Book 21
    clone or download 30URL-'; "4  SSH  ! (2 $ git clone {URL} CloneNotebook Jupyter Notebook 7
22. #17Amazon SageMaker ◦ AmazonJupyter Notebook,+&0     ◦ -3 '65/6&06)4(*! ◦ AWS,+&0.%
    ◦ "  $2 22
23. ") ' ◦ #% & (
    ' &  ◦
    !  ' (
    $  23