chrislevn / Deploying-GPT-Example

Deploying GPT Miniature on Flask, Docker, and Kubernetes

Deploying Miniature GPT on Flask, Docker, Kubernetes

What it does

Deploy a Machine Learning model (Miniature GPT with movie reviews dataset) with Docker, Kubernetes, Flask API.

How to run:

1. Create a hub on Docker - https://hub.docker.com/ (As of current version, Kubernetes will take public image. )
2. Login to docker docker login
3. Build Docker image (current port is 5002).
   docker build -t tagname
docker push image:tagname
4. In deployment.yaml, rename image (line 32) with the image name from your Docker Hub.
   Optional: To have Docker image exits and removes itself, run docker run -rm <IMAGE_NAME>
5. Run kubernetes kubectl create -f deployment.yaml
kubectl apply -f deployment.yaml (this will trigger deployment and load balancer)
6. Run minikube minikube start
minikube service list (get the list of current service. get the service running your name)
   minikube service <SERVICE_NAME>
7. Test the API with Postman

Future developments

• Connect Kubernetes with private Docker image
• Add CI/CD
• Add monitoring