AWS Setup Tutorial

AWS or Amazon Web Services are a bunch of Cloud services that are provided by Amazon.

There a bunch of options to choose from for Deep learning, I prefer AWS the most because of the huge number of options that it offers.

• First you will need to set up your account by hopping over to the aws website Here
• After setting all up, login Here

 

• Next, we need to request for increase in the service limits, since by default all of the access to compute power is set to 0 and we need to request Amazon for access.

• Click on Support from the Top Right Corner
• Select Support Centre in the Drop Down Menu
• Select Create a New Case from the Page

 

• Select Service limit increase
• Limit Type: EC2
• Region: Select the region closest to you. AWS hosts its server in varying places, for the lowest lag, select one that is closest to you.
• Select Instance type as p2.xlarge
• Select 1 or more as the limit
• Description: In my case, I’ll be following FAST AI MOOC.

Now you might have to wait for a while untill you hear from Amazon, once your limit increase has been approved

• From the Side menu (left side), select Instances under Instances

• Click on Launch instance

• Next up, we have to select an AMI- Select one from the Community AMI: Anything that has the libraries needed should work fine for you.

AMI is an ‘Image’ of the OS with everything installed. Consider a snapshot of the states of OS after installing all the software. It can save you the pain of setting it all up, you can dive right into working.

• Next Select your Instance type, Select GPU Compute
• Tick P2.xlarge
• Proceed by clicking on Review and Launch
• Next we need to create a Key-Pair to login to the Instance
• I’m using a pre-created .pem. You may create a new one by selecting that from the Drop Down menu.

Key Pair: We use Secure Shell Login or SSH to login to our instances, the security is ensured by A Key-Pair which gives you access to the instance

• Download the Key pair ( A xyz.pem) file
• SAVE THIS FILE, DELETING IT WILL MAKE YOU LOSE ACCESS TO THE INSTANCE.
• Next you will be redirected to the EC2 Dashboard
• First, you need to enable permissions for the .pem file
• Open a terminal
• Cd to the directory where the .pem file is
• chmod 400 xyz.pem
• This will set the permissions to use the file

• Next, click on the Instance and click Connect from the Menu
• Just copy the “ssh -i xyz.pem….” command and run it inside your terminal
• Type yes to save the Auth files
• Voila! You’re in.
• Remember to Stop the Instance once you’re done, AWS bills you for the number of hours the instance has been running.
• A little charges might still be incurred on monthly basis-based on the Usage of space by the instance, Terminate the Instance to completely remove it, if you don’t need to work with it.