Build and deploy your first machine learning web app to Linode

A beginner’s guide to train and deploy machine learning pipelines in Python using PyCaret

Inspired by these articles and Github repos:

• How to Move Machine Learning Model to Production Linode
• https://github.com/abalarin/Flask-on-Linode
• Build and Deploy Your First Machine Learning Model
• https://github.com/pycaret/deployment-heroku

Application Breakdown

This is the breakdown of a fairly simple Flask Application serving a PyCaret pickled model.

Flask-on-Linode
│   README.md
│   FlaskDeployment.md
│   .gitignore
└───flask_app
│   │   app.py
│   │   deploy_28082021.pkl
│   │   requirements.txt
│   └───static
│   │   │   style.css
│   └───templates
│   │   │   home.html

Deploying Locally

Lets walk through setting up your development environment and deploying this application on your local machine

1. Install Python, pip, and virtualenv

• Python
• pip
• Virtualenv

2. Clone this repo and CD into the projects directory

git clone https://github.com/getcontrol/Flask-model-Linode  
cd flask_app  

3. Create and activate a virtualenv

virtualenv venv
source venv/bin/activate

4. Install packages

pip install -r flask_app/requirements.txt

5. Create Flask environment variables

export FLASK_APP=flask_app/app.py
export FLASK_ENV=development

6. Run it

flask run

Move your App to your Linode

Retrieving your Application from source control

1. SSH into your Linode

ssh user@<Linode-IP>

2. Navigate to your Home directory

cd /home

3. Install your essential Linux packages to get started (enter command one by one):

sudo apt update
sudo apt upgrade
sudo apt install git

3. Pull from source control (Replace my repo with yours)

git clone https://github.com/getcontrol/Flask-model-Linode && cd Flask-model-Linode

Configure NGINX

NGINX is a free, open-source, high-performance HTTP server and reverse proxy, as well as an IMAP/POP3 proxy server.

1. Install NGINX

sudo apt install nginx

2. Create an NGINX Configuration file

sudo nano /etc/nginx/sites-enabled/flaskapp

server {
	listen 80;
	server_name <Your Linodes IP>;

	location / {
		proxy_pass http://127.0.0.1:8000;
		proxy_set_header Host $host;
		proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
	}
}

3. Unlink the NGINX default config

unlink /etc/nginx/sites-enabled/default

4. Reload your NGINX server

sudo nginx -s reload

If you try navigating to your Linode's IP in a web browser you should get the following or a similar error.

Install Python and Packages

You should now be in your applications root directory on your Linode.

1. Install Python 3

sudo apt install python3

2. Install pip and make sure its the latest version

sudo apt install python3-pip
python3 -m pip install --upgrade --force-reinstall pip

3. Install Flask Packages/libraries.

cd flask_app && pip install -r requirements.txt

Deploy your Application

1. Start the app

python3 app.py

2. Your app is running successfully if you see this!

Transformation Pipeline and Model Sucessfully Loaded
 * Serving Flask app 'model' (lazy loading)
 * Environment: production
   WARNING: This is a development server. Do not use it in a production deployment.
   Use a production WSGI server instead.
 * Debug mode: on
 * Running on http://xx.xx.xx.xx:5000/ (Press CTRL+C to quit)
 * Restarting with stat
Transformation Pipeline and Model Sucessfully Loaded
 * Debugger is active!
 * Debugger PIN: 673-488-585

Custom Model Training

You can train your own Pycret model using the Notebook:

1. Train the model and save it with a unique name "mymodel.pkl"

2. Change line 9 (remove file extension) :

model = load_model('mymodel')

3. Start the app

python3 app.py