Deep Learning Notebook AMI Instructions


1. You can access the AMI either from the AWS marketplace url or directly from the EC2 Console in the AWS Marketplace category.

On the marketplace AMI page, click on "Continue to Subscribe" and Accept to enable the AMI for your account. Note that subscribing does not incur charges on your account. It simply allows you access to the AMI.



2. You will get a confirmation and after a short period the product will be enabled for your account



3. Select the region, vpc and your key pair settings. The configured security group has ports 8888 (jupyter notebook) and 22 (ssh) open by default.

When selecting the EC2 instance type, if you select a p2.* or p3.* instance, those come with GPUs, and the AMI will automatically configure the required drivers and packages to enable Tensorflow and Pytorch to take advantage of the GPU acceleration.

  (Optional) You can set user data when configuring the instance to enable additional options:

  • If you wish to use S3 as storage for your notebooks, add this line to the user-data field when configuring the instance: S3_BUCKET=your-s3-bucket-name
    The instance will mount the bucket as the notebook directory.
    You also need to assign to the instance the right IAM role with S3 access to the bucket
  • You can use your own SSL certificates for the instance.
    Add SSL_CERT=/home/ec2-user/s3/path-to-cert.crt and SSL_KEY=/home/ec2-user/s3/path-to-cert.key to the user-data field.
    This will let the instance load the certificate and private key.
    This can be useful if you don't want to use a self-signed certificate for the web server.



4. After launching the instance, open the ec2 console to find out the IP address and the instance id (the instance id is the password in the jupyter login screen)



5. Go to https://<instance-ip>:8888 and you should see a warning screen about SSL. This warning just means the instance is using self signed certificates. Click advanced and accept the security exception to continue




6. You should see the Jupyter log in screen. Use the instance id as the password.


7. You should now be in the Jupyter environment. You can create a new notebook and start coding.