Overview

Sagemaker Utilities

CLI

There are some command line utilities to help manage sagemaker endpoints.

sagemaker-endpoint-management

Overview

Argument	Description
action	The action to take. Valid values are create or delete
endpoint-name	The name of the endpoint to create or delete
endpoint-config	The name of the endpoint config to use

Installation

brew tap jmoney/aws
brew install sagemaker-endpoint-management

Run locally

go run cmd/cli/endpoint/main.go --action create --endpoint-name <endpoint-name> --endpoint-config <endpoint-config>

Sagemaker Lambda Proxy

A proxy that runs as a Lambda to sagemaker to testing certain sagemaker models. This is a proof of concept and not meant for production use.

Prerequisites

brew bundle --file Brewfile
tfenv use

Deploy

make build
terraform init
AWS_PROFILE=<profile> terraform plan -out=plan.out -var endpoint_name=<endpoint-name>
AWS_PROFILE=<profile> terraform apply plan.out

Destroy

AWS_PROFILE=<profile> terraform destroy -var endpoint_name=<endpoint-name>

Testing

curl --silent --http1.1 --request POST --header "Content-Type: application/json" --header "X-NONCE: $(terraform output -json | jq -r .nonce.value)" "$(terraform output -json | jq -r .url.value)" -d @data.json | jq -r '.[].generation.content'

There is a data.json file at the root of this repo with a question you can use.

Docker

There is a dockerfile built to be used with lambda for the container image package type. It has not been tested for local runs but in theory it should work. Lambda does not allow images pulled from ghcr and only AWS ECR registries so the docker image would need to be mirrored from ghcr to ecr in some way.