top of page
Asaf Shimshoviz

Part 1: Tutorial, Developing Deep Learning on AWS

Updated: May 16

Connecting to an Instance


Introduction


In recent years, we have seen huge developments in computer vision engineering when it comes to deep learning. Development for these projects usually requires large computational resources on powerful GPU processors. When we work on these AI projects, our team at Vision Elements will offload compute power to cloud resources, for example Amazon cloud computing (AWS).


Cloud computing can give us the flexibility to use large computational resources whenever we need it, however, it may be tricky to make it work efficiently without losing time and money. In this series of tutorials, our computer vision services team will take you through the steps we have adopted to develop deep learning on AWS, starting from launching the instance, loading and handling the data, creating a convenient environment for development, and more to make your journey as smooth as possible.

 

Part 1: Launching and connecting to an AWS instance.

 

Step 1: Under the EC2 dashboard choose the AMI catalog.



Step 2: Type deep learning in the search bar. Choose an instance that suits your needs. Note that on some of the AMIs, there is a list of the supported instances. It's important to make sure that the AMI supports the type of instance that you need.

 


Step 3: Name the instance.



Step 4: Choose an instance type that will work for what you are trying to achieve. In our computer vision services team example we will choose an instance of p3.2xlarge which has 1 GPU with 16GB of memory, which works for moderate deep learning projects (https://aws.amazon.com/ec2/instance-types/p3/).


 

Step 5: Create a key pair.  In order to prevent anyone other than yourself from accessing the instance, you should create a key pair which is a pem file. You need to use this file each time you connect to the instance. Select the key pair name, choose pem, and push create key pair. A key pair will be downloaded to your machine. We will use it to connect the instance.

 

 


Step 6: Push the launch instance. Under instances you will be able to see your new instance. Wait for the instance state to change to "running."


 

Step 7: Connection to AWS using SSH. Open the command prompt. Type: ssh -i "path to your key pair" ubuntu@instance-adress. Replace the instance address with the Public IPv4 address. We are now connected to our AWS machine and can start using it.


In Part 2 of this blog series, our computer vision services team will walk through how to connect to the AWS instance using a remote desktop as well as other useful tips that will allow you to leverage AWS for your vision engineering projects.

Comments


bottom of page