Tensorflow Lambda Layer using AWS SAM

Tensorflow version 1.4.1

Layer unzipped size: 185.6 MB chi

Pre-requisite

1. Docker

2. Install AWS SAM CLI. You must have AWS IAM access to create role, aws lambda and upload to S3.

3. Edit Makefile and specify value of STACK_NAME and your S3_LAMBDA_BUCKET

Step 1: Build docker image

make image

This command will run docker build with a base image from amazon.

The dockerfile will run ./build.sh that will install and package all the dependencies to be used in amazon lambda.

Step 2: Package dependencies for lambda

make install

This command copies all dependencies in the local build folder

Step 3: Deploy the lambda layer

make deploy

This command will package the build/layer with AWS SAM CLI and upload to S3.

Edit template.yml to configure name and compatible runtimes if necessary

Step 4: Copy the Layer ARN

After a successful deployment, copy the Layer ARN output and paste it in your own lambda app template.yml Layers value

Reference

Dependencies

bleach==1.5.0
html5lib==0.9999999
importlib-metadata==1.7.0; python_version < '3.8'
markdown==3.2.2; python_version >= '3.5'
numpy==1.19.2; python_version >= '3.6'
protobuf==3.13.0
six==1.15.0; python_version >= '2.7' and python_version not in '3.0, 3.1, 3.2'
tensorflow-tensorboard==0.4.0
tensorflow==1.4.1
werkzeug==1.0.1; python_version >= '2.7' and python_version not in '3.0, 3.1, 3.2, 3.3, 3.4'
wheel==0.35.1; python_version >= '3'
zipp==3.1.0; python_version >= '3.6'

Lambda Layers

https://docs.aws.amazon.com/lambda/latest/dg/configuration-layers.html