1 Introduction
AWS Tutorial
CS224D Spring 2016 April 17, 2016
This tutorial explains how to set up your EC2 instance using our provided AMI which has TensorFlow installed. Our AMI is cs224d tensorflow (ami-d8433cb8). We’ve installed on it:
• CUDA 7.0
• cuDNN 4.0
• TensorFlow 0.7
2 Create an AWS account and apply for AWS Educate Program
2.1 Create an AWS account
Go to AWS homepage http://www.aws.amazon.com. Click the Sign In to Con- sole or Create an AWS account button on the top right corner. This will bring you to the sign in/sign up page. Create your account there with your email and password.
2.2 Apply for AWS Educate Program
AWS Educate Program gives $35 AWS credits per student. You can apply for it here: https://aws.amazon.com/education/awseducate/.
1
Figure 1: AWS homepage
3 Launch an EC2 instance using existing AMI
Note: You may need to wait for 2 hours after sign-up before you can launch instances.
Go to AWS homepage and sign in to your console. After you sign in, you should be able to see a page like this
Figure 2: Console Home
You can see your region on the top right. Make sure it is N. California.
On the left you can find EC2 under Compute category. Click EC2, which will bring to the EC2 dashboard:
2
Figure 3: EC2 dashboard
We can find a blue Launch Instance button. Click it and we will see lists of AMIs that we can use. An AMI (Amazon Machine Image) is a pack of data which provides the information required to launch an instance. We can modify existing AMIs, such as installing softwares on it, and then save it as our own AMI. Here we will use the AMI created for CS224D which has TensorFlow installed.
Figure 4: Lists of available AMIs
Click Community AMIs on the sidebar. Search for “cs224d”. You will find an AMI named cs224d tensorflow (ami-d8433cb8).
3
Figure 5: Select an AMI
Select the cs224d tensorflow AMI. It then asks you to select an instance type. We need to select a GPU instance. Scroll down and find g2.2xlarge and click Next:
Figure 6: Choose instance type
After that, it asks you to configure instance details. We don’t need to modify things here, simply click Next. Then we reach step 4: Add Storage. Make sure the size of Root is at least 16GB.
4
Figure 7: Add Storage
Click next and in Step 5: Tag Instance, we also click Next.
Now we are at Step 6: Configure Security Group. If you hadn’t created a security group before, choose Create a new security group. Otherwise, choose Select an existing security group and then choose the security group that you had created. Make sure SSH is included in the type column. After you’ve done this, click Review and Launch.
Figure 8: Configure Security Group
In Step 7: Review Instance Launch, click Launch and we reach the final step: Select an existing key pair or create a new key pair. This key pair would be needed when you SSH into your instance. If you hadn’t created a key pair before, select create a new key pair and give the key pair a name, such as cs224d. Then click Download Key Pair and save it as a .pem file. Make sure to store it at somewhere you can find. We will need it later.
5
Figure 9: Select an existing key pair or create a new key pair
The last step is to click Launch Instance. Congratulations! You now have a running EC2 instance! (Note: Be sure to read section 4 to learn how to close an instance. Amazon would charge you for every running instance. ) You can view status of instances in the EC2 dashboard when you click instances. You will see the Instance State goes from pending to running. You can also see the Public IP of your instance, as shown in figure 10.
Figure 10: Inspect the status of your EC2 instance
After the instance is running, we can now ssh into our instance to do our programming assignments and projects. But first, we need to change the per- mission of the previous .pem that you’ve downloaded. Open terminal and type the command:
$ chmod 400 path-to-pem-file Now we ssh into our instance use:
$ ssh -i path-to-pem-file ubuntu@ip-address 6
We can check that TensorFlow is working by following the steps here:
https://www.tensorflow.org/versions/r0.8/get_started/os_setup.html# test-the-tensorflow-installation
4 Close a running instance
Amazon charges you for each running instance, so make sure to close them when
you’ve finished. In the page where you inspect your instance, right click your
instance and in Instance State, select Terminate or Stop. For information on
their differences, you can look at http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/ec2-instance-lifecycle. html#lifecycle-differences.
Figure 11: Close a running instance
5 Other topics
Here are some topics and related links that you may need for doing your assign- ments/pro ject:
1. Amazon Machine Image:
http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/AMIs.html
2. AWS Identity and Access Management (IAM):
http://docs.aws.amazon.com/IAM/latest/UserGuide/introduction. html
This is very useful when you do project in group.
3. Exchange file between EC2 and S3:
http://docs.aws.amazon.com/AWSEC2/latest/UserGuide/AmazonS3.html
7