Now that you're connected to your Jupyter notebook, let's prepare a data set for processing. We're going to tackle the customer churn prediction problem found in this SageMaker example, except we're going to use a neural network in TensorFlow rather than XGBoost.
We need an S3 bucket to store our training data. Making sure that you're working in the same region as your EKS cluster, run:
aws s3 mb s3://<BUCKET>
Choose an appropriate name in place of <BUCKET>.
We need to give our EKS nodes permission to use the S3 bucket we just made. We do this by associating an IAM policy to the IAM role that eksctl created for the nodes.
First, create the following IAM policy in the AWS IAM console. Name it smoperators-s3. Be sure to replace the value BUCKET with the name of the S3 bucket you just created.
{
"Version": "2012-10-17",
"Statement": [
{
"Sid": "VisualEditor0",
"Effect": "Allow",
"Action": [
"s3:PutObject",
"s3:GetObject",
"s3:ListBucket",
"s3:PutObjectTagging"
],
"Resource": [
"arn:aws:s3:::BUCKET/*",
"arn:aws:s3:::BUCKET"
]
},
{
"Sid": "VisualEditor1",
"Effect": "Allow",
"Action": "s3:HeadBucket",
"Resource": "*"
}
]
}
Now find the IAM role whose name starts with eksctl-eks-kubeflow-nodegroup. Add the new policy to this role.
Before running a notebook, we'll install the boto3 module so that we can work with S3 through the Python SDK. In your notebook, click New -> Terminal. In the new terminal window, enter:
pip install --user boto3 pandas numpy sklearn
Next, download the example notebook. In your notebook, click Upload and select this notebook. Click Upload again to complete the upload.
Now click the imported notebook to open it. Read through and execute each cell in the notebook, making sure to edit the bucket and prefix variables in the first code cell.
After you finish executing this notebook, you'll have train, validation, and test data sets stored in S3.