Welcome!
Let’s …• Set up an Amazon Web Services (AWS) Virtual Machine (VM)
• Connect to the VM using a PuTTY (Put TeleTYpe) Secure SHell (SSH) client
• Configure the Kaggle Application Programming Interface (API)
• [optional] Edit inbound rules, to allow the node package manager (npm) to serve a custom version of the Tensorflow Playground
• Bonus: Thinking about buying a Graphics Processing Unit (GPU)?
Amazon Web Services: Elastic Compute Cloud [AWS EC2]• Navigate to https://aws.amazon.com/education/awseducate/ and apply for
the AWS Educate Promotional Credits• Use your “uw.edu” email address to register, as University of Washington students
can receive a $200 credit [UW is an "AWS Educate" Institution]
• After you have received the AWS Promotional Credit Code (via email to your uw.edu email address)• Redeem this code at https://console.aws.amazon.com/billing/home#/credits• Visit https://aws.amazon.com/console/ to setup a p2.xlarge Virtual Machine
• Don’t forget to terminate the virtual machine when you’re done with your homework assignment: the p2.xlarge Virtual Machine (VM) costs $0.90 per hour [April 7, 2020]
• Note: When you first try to create a p2.xlarge VM, you will probably need to request that they increase your limit [this took 6 hours (for me)]
[US_WEST_2]: EC2 Instances / Instance Limit (p2.xlarge), New Limit = 1
Selecting an Amazon Machine Image (AMI)
• On the next screen, I picked the Ubuntu Server 18.04 Long Term Support (LTS) Hardware Virtual Machine (HVM), Solid State Drive (SSD) Volume Type [the first choice on the next slide]• This choice requires installing the Common Unified Device Architecture
(CUDA) driver, toolkit, and Deep Neural Network (CUDNN) library, as well as tensorflow-gpu and keras: it mimics what you might have to do at home
• Alternatively, you could pick the Deep Learning Amazon Machine Image (AMI) Version 27.0 [the third choice on the next slide]• This choice means you can start with installing the Kaggle API
PuTTY: a Secure Shell (SSH) Client
• Install from here: https://www.putty.org/Download Putty > MSI (Microsoft Software Installer) > 64-bit
• Save your private key• Run "C:\Program Files\PuTTY\puttygen.exe“• Click the “Load” button• Browse to the “.pem” file (new key pair) that you just downloaded from AWS• Click the “Save private key” button
• Run "C:\Program Files\PuTTY\putty.exe“• Use the “IPv4 Public IP” for “Host Name (or IP address)”• On the left, expand “Connection” > “SSH” > “Auth”, and “Browse” to the
“.ppk” putty private key file that you created with puttygen
Software Installation
Commands for installing software can be found here:
http://cross-entropy.net/ml410/aws-script.txt
Console output from my session can be found here:
http://cross-entropy.net/ml410/aws-console.txt
The entire session (both software installation and homework execution) took around 15 minutes
Termination
• After you’re done with the Virtual Machine, do not forget to pull down the “Actions” menu and “Terminate” the instance
• This should stop any further charges
Generate Kaggle API Token
• Navigate to kaggle.com• Create an account with your uw.edu email address
• Use your favorite city name for your display name
• Click on your account icon, in the upper, right-hand corner• Select “My Account”
• Select “Create New API Token”
C:\Users\dadebarr> type %USERPROFILE%\Downloads\kaggle.json{"username":"mlearn310","key":“Secret Globally Unique IDentifier (GUID) Goes Here"}C:\Users\dadebarr>
Thinking About Buying a GPU Card?• Turing is the name of Nvidia’s latest architecture
• A list of their latest consumer cards appears below
• If you’re thinking about buying a card, make sure you understand whether your system has …• a Peripheral Component Interconnect express (PCIe) 3.0 x16 card socket [this
card should probably be the only installed graphics card]
• enough power (and appropriate power connectors) to support the card [requires more than just the card socket]
• https://en.wikipedia.org/wiki/GeForce_20_series