lab2

Lab 2: Create a private interface

In this lab you will use AWS Lambda to create a private interfaace for your Amazon SageMaker endpoint

Task 1: Create a Lambda function

Create an AWS Lambda function that uses Amazon SageMaker to classify an aircraft in an image

1. Browse to the AWS Lambda console to create a new function: https://console.aws.amazon.com/lambda/home#/create
2. Select the Author from Scratch option:

3. Fill out the following information for the Lambda function:

• Name: mlclassify
• Runtime: Python 3.6
• Role: Create a custom role

4. Specify the following information for the IAM role, then click the Allow button to continue:

• IAM Role: Create a new IAM Role
• Role Name: mlclassify

5. Click the Create function' button to finish:

Task 2: Update the IAM role

Update the IAM role to allow invocation of the SageMaker InvokeEndpoint API

1. Browse to the AWS IAM console to edit the mlclassify IAM role: https://console.aws.amazon.com/iam/home#/roles/mlclassify
2. Click on the Add inline policy button

3. Click on the JSON tab and replace the existing policy with the following. Replace <SageMaker Endpoint ARN> weith the ARN captured previously.

{
    "Version": "2012-10-17",
    "Statement": [
        {
            "Sid": "mlclassify",
            "Effect": "Allow",
            "Action": "sagemaker:InvokeEndpoint",
            "Resource": "<SageMaker Endpoint ARN>"
        }
    ]
}

4. Click on the Review policy button to continue

5. Name the policy mlclassify, then click on the Create policy button to finish

Task 3: Update the Lambda function

Update the Lambda Function classify an aircraft in an image using Amazon SageMaker InvokeEndpoint API

1. Browse to the AWS Lambda console to edit the mlclassify Lamda function: https://console.aws.amazon.com/lambda/home#/functions/mlclassify
2. Replace the lambda_function.py template code with the following (mlclassify-lambda.py). Replace <SageMaker Endpoint Name> with the name of your SageMaker endpoint.

import json
import boto3

from botocore.vendored import requests
 
sage = boto3.Session().client(service_name='runtime.sagemaker') 
names = ['airbus-a320','boeing-747','dornier-328']

def lambda_handler(event, context):
   
    url = event["url"]

    # download image bytes
    bytes = requests.get(url).content
    
    # classify aircraft in the image
    response = sage.invoke_endpoint(EndpointName='<SageMaker Endpoint Name>', 
                                   ContentType='application/x-image', 
                                   Body=bytes)
    scores = response['Body'].read()
    scores = json.loads(scores)

    aircraft = ""
    if max(scores) > 0.90:
        aircraft = names[scores.index(max(scores))]
    
    return {
        "statusCode": 200,
        "body": aircraft
    }

3. Click the Save button to finish

Task 4: Test the Lambda function

Create a test event and test your Lambda function

1. Browse to the AWS Lambda console to edit the mlclassify Lamda function: https://console.aws.amazon.com/lambda/home#/functions/mlclassify
2. Click on the Select a test event.. drop down and select Configure test events

3. Specify the following onformation for the test event:

• Event template: Hello World
• Event name: mlclassify
• Code:

{
  "url": "https://s3-us-west-2.amazonaws.com/awsgeek-devweek-austin/boeing-747.jpg"
}

4. Click on the Create button to continue

5. Click click on the Test button, then verify the output of the test matches the example output below

This is the end of the lab