Motorbike-Classification - MLOPS

Table of Contents

1. Architecture
2. Repo Structure
3. API Example
4. I. How-to Guide

• 4.1. Running model locally
• 4.2. Model-serving with Azure Kubernetes Services (AKS)

Architecture

Repo structure

├── app: 
│   ├── data
│   ├── utils
│   └── main.py
├── helm-charts:
│   ├── jaeger-tracing
│   ├── model-serving
│   ├── monitoring
│   └── nginx-ingress
├── notebook: 
│   ├── resnet-50.ipynb
│   └── visualize.ipynb
├── terraform
│   ├── main.tf
│   ├── outputs.tf
│   ├── provider.tf
│   ├── README.md
│   ├── ssh.tf
│   └── variables.tf
├── docker-compose.yaml
├── dockerfile
└── requirements.txt

API Example

Motorbike Image to classify:

Returned Result :

I. How-to Guide

1. Running model locally

1.1 Creating & activating conda environment with python 3.9

conda create -n my_env python==3.9
conda activate my_env

1.2 Install prerequisites

pip install -r requirements.txt

1.3 Navigate to app directory

cd ./app/

1.4 Connect service app with Uvicorn

uvicorn main:app --host 0.0.0.0 --port 30000

1.5 Another way to deploy model locally with Docker Compose

*** If you already have Docker Engine in your local machine, just execute 1 command:

docker compose -f docker-compose.yaml up -d

---

*** When model loading , use command to check

docker image ls
docker logs img_id

2. Model-serving with Azure Kubernetes Services (AKS)

Install Azure CLI Packages

• Install Azure CLI

sudo curl -sL https://aka.ms/InstallAzureCLIDeb | sudo bash 

• Setup your Azure project:

• Initialize Azure account:

  az login

2.1 Create AKS cluster with Terraform (infrastruture as code)

cd terraform # Navigate to terraform folder
terraform plan # Preview cluster plan
terraform apply # Create cluster

2.2 Connect to AKS cluster with Suggest command

2.3 Switch to your AKS cluster environment (using kubectl)

kubectx your_aks_cluster

2.4 Create cluster namespace

kubectl create ns nginx-ingress # Nginx Ingress Controller
kubectl create ns model-serving

2.5 Deploy Nginx Ingress Controller with Helm-chart

helm upgrade --install nginx-ingress helm-charts/nginx-ingress -n nginx-ingress           

2.6 Deploy Cat Breed Classification Application

helm upgrade --install nginx-ingress helm-charts/model-serving -n model-serving           

2.7 Config Domain Name to nginx-ingress's External IP

kubectl get svc --all-namespaces # Listing all services & finding your nginx External IP

sudo vim /etc/hosts # Editting hosts_file with Vim
your_nginx_externalIP 

your_nginx-ingress_host cbp.com # Updated content

2.8 Access application at above address http://cbp.com/docs

Because I'm using a free Student Subscription, I don't have enough memory to pull the image size. Instead of running it on the cloud, I am running it on minikube.

Fail because not enough memory :

Result when use minikube :