FastAPI-Model-Deployement

Template of simple API for machine learning models deployment

How to use it

You can run the backend service directly on your machine, use docker-compose to deploy it on a VPS or build it as an image and use that as a backbone of AWS Lambda

Simple run

docker-compose.yaml should look like that:

version: '3.7'
services:
  model_api:
    build: .
#    image: <user>/<docker-image-name>:latest
    ports:
      - "8011:80"

and just run

docker-compose up

The service will be avalible at: 127.0.0.1:8011/docs

On VPS, EC2 or similar services

We use automatic docker build and push using Github Actions. So eveyryting you need to do is to commit you changes, update values in docker.yml and docker-compose.yaml and that is basicly it, you are good to go

1. Update docker._yml

*disclaimer: update this values directly on gihtub, don't push it from the repo (it is possible but requires additional actions)

Rename it to docker.yml

Set DOCKER_USERNAME and DOCKER_PASSWORD in your repository secrets

Replace line 34 repository: <user>/<docker-image-name>:latest with the proper repository name (the one to which you want to push it)

Save it

2. Modify docker-compose.yaml

Replace <user>/<docker-image-name> in docker-compose.yaml, comment build and commit the change

version: '3.7'
services:
  model_api:
#    build .
    image: <user>/<docker-image-name>:latest
    ports:
      - "81:80"

3. Clone your repo on a remote machine

git clone <link-to-your-repo>

4. Run docker-compose up on remote machine

docker-compose up

With AWS Lambda or similar services

Authors

• Piotr Mazurek