ml_project_template

Tutorial for deployment website in AWS instance can be found here

Tutorial

This is tutorial for local development on local machine

Install

sudo -s
apt update
apt install -y npm
apt install -y python3.6 python3-pip
apt install -y nginx
apt update && apt install -y libsm6 libxext6
apt install libxrender1
apt install -y redis-server

Clone the repo

apt install git
git clone https://github.com/thanhhau097/ml_project_template.git 

Install requirements

cd ml_project_template
alias python=python3
alias pip=pip3

Requirements for Flask app

cd api
pip install -r requirements.txt
npm install

Requirements for web app

cd ../web
npm install

Change your custom path

api/app.py

You need to config your AWS account in local machine: (optional, it is useful when you want to save uploaded data to AWS s3)

apt install awscli
aws configure

Change your bucket in AWS S3 (it is optional, when you need to upload user data to s3 bucket)

async_data = {
    'data': {'image': image_utils.encode(image), 'result': result},
    'bucket': 'your-bucket',
    'object_name': 'file-path-in-bucket/{}.pkl'.format(file_name)
}

nginx/nginx.conf

Comment out the SSL config in this file because you don't need domain name for local development (line 4-6, 29.41). If you want to deploy into production, please see this link.

Write your code

There are 3 tasks that you need to do for your project:

1. Write prediction for you model in model/predictor.py and update your weights in model/weights folder
2. Write your API in api/app.py using Flask framework (you can use the template that was written for image)
3. Write your web app using ReactJS (you can use the demo template that I wrote in web/)

How to run the service

Open multiple terminal windows, each process should be handle in one window.

1. Web

cd web/
npm run build
npm run start

2. Flask

export LC_ALL=C.UTF-8
export LANG=C.UTF-8
export PYTHONPATH=$PYTHONPATH:./ 
python api/app.py 

3. Redis

Change redis conf in /etc/redis/redis.conf

bind 0.0.0.0

systemctl stop redis
systemctl start redis

4. Celery

Change 'redis://redis:6379/0' to 'redis://localhost:6379/0' because you are running redis in local machine

celery worker -A api.app.celery_app --loglevel=info

5. Nginx

Change the app and web in nginx/nginx.conf file to 0.0.0.0

sudo -s
rm /etc/nginx/sites-enabled/default
cp nginx/nginx.conf /etc/nginx/sites-enabled/
systemctl reload nginx

Now you can go to your browser and see what is happening: 0.0.0.0

Tools are used in this template

1. Flask
2. Redis
3. ReactJS
4. Nginx
5. Certbot (optional, when you have a domain name)
6. Celery
7. Docker
8. Jenkins (optional)