How to import numpy using layer in lambda aws?

Question

I am trying to add the numpy package to my lambda function, but I can't import it.

I have followed several tutorials, but they all have the same problem. On my last attempt, I executed the following step by step:

• I created a lambda function in AWS and tested it
• Installed the numpy package on my local machine and zipped it
• Created lambda layer and loaded the numpy package
• Fixed the layer to lambda and tested

When I run my lambda function without importing numpy, it works perfectly, however when I import it I get this error:

{
  "errorMessage": "Unable to import module 'lambda_function': No module named 'numpy'",
  "errorType": "Runtime.ImportModuleError",
  "requestId": "0153834a-6b28-44d1-889f-3e2e3ead9c4a",
  "stackTrace": []
}

A very common error on the forum, however everything is fine with lambda_function, because as I said it works fine if I'm not importing any module.

• lambda_function.py

import json
import numpy 

def lambda_handler(event, context):
    # TODO implement
    return {
        'statusCode': 200,
        'body': json.dumps('Hello from Lambda!')
    }

I would like to learn how to use Layers inside Lambda.

Answer 1

There are at least two ways to add NumPy to your AWS Lambda as layer:

1. Add an AWS (Managed) Layer:
   From your Lambda console, choose AWS Layers and AWSSDKPandas-Python39 (Choose a layer accordingly for your Python version). Since Pandas is built on top of NumPy, you should be able to use NumPy once you add this Pandas layer.
2. Add a custom layer:
   Create a virtual environment on your local machine, install NumPy, create a zip file, and upload to AWS as a layer. Ensure the version of Python in your local Virtual env is the same as the Lambda. Step by step instruction here. This is a general approach which you can use to add any external Python libraries as Lambda layers.