1.9 - Model Deployment
After successfully developing and testing our employee attrition prediction model locally, let's explore how to deploy.
Docker - Containerise the Application
To containerise the application make sure all the required files are available.
With the above application files add the following Dockerfile, .dockerignore and requirements.txt file.
a. Dockerfile
Create a Dockerfile and copy the below content into it.
FROM python:3.11.6-slim
WORKDIR /app
COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt
COPY . /app
RUN python3 model.py
EXPOSE 5000
ENV FLASK_APP=app.py
CMD ["flask", "run", "--host=0.0.0.0"]
Here the Dockerfile uses a minimal Python 3.11.6 as the base image. Set the working directory inside the container to /app and copies the requirements.txt file into it.
Installs the Python packages listed in requirements.txt using pip. Then it copies all files and folders from your local directory into the container's /app directory.
Then executes the Python script model.py during the image build process and expose the container on port 5000. The CMD instruction specifies the command to run when the container starts.
b. requirements.txt
Create a file requirements.txt and copy the below contents into it.
scikit-learn
pandas
flask_cors
c. .dockerignore
This file ignores the files mentioned during the Docker build process. Create a file .dockerignore and copy the below contents.
// .dockerignore file
Dockerfile
In our project we use it to ignore copying Dockerfile into the Docker image while building the application image.
And the final directory structure will look like this:
.
├── Dockerfile
├── app.py
├── employee_attrition_test.csv
├── employee_attrition_train.csv
├── model.py
├── requirements.txt
├── .dockerignore
├── static
└── templates
Once every files are ready run the following command to build the Docker image.
docker build -t employee-attrition-app:1.0 .
To test the image run it locally using the following command.
docker run -p 5000:5000 employee-attrition-app:1.0
If it runs without any issue, tag it and push it to a Docker registry, so that we can deploy it on Kubernetes.
docker tag employee-attrition-app:1.0 techiescamp/employee-attrition-app:1.0
docker push techiescamp/employee-attrition-app:1.0
Deploy it on Kubernetes
Create a manifest file deploy.yaml and copy the below content.
apiVersion: apps/v1
kind: Deployment
metadata:
name: employee-attrition-app
labels:
app: employee-attrition
spec:
replicas: 1
selector:
matchLabels:
app: employee-attrition
template:
metadata:
labels:
app: employee-attrition
spec:
containers:
- name: employee-attrition
image: techiescamp/employee-attrition-app:1.0
ports:
- containerPort: 5000
To expose the application as a service, create a file service.yaml and copy the below contents into it.
apiVersion: v1
kind: Service
metadata:
name: employee-attrition-service
spec:
selector:
app: employee-attrition
ports:
- protocol: TCP
port: 80
targetPort: 5000
type: ClusterIP
Apply the manifest file by running the following command.
kubectl apply -f deploy.yaml
kubectl apply -f service.yaml
To check if the pods are up and running use the following command.
kubectl get po
Thus Employee Attrition Prediciton project is up and available to all.