An academic project that includes some of the concepts covered during the "DevOps for Data Science" course DSIA-5201A at ESIEE Paris.

Keywords: Machine learning, DevOps, Flask, Python, Docker

Authors:

• Vincent Barbosa Vaz
• Dhanashri Rajput
• Clea Ngouansavanh
• Elias Bouillanne

Teacher: Pouya Yousefi

Course unit: DSIA-5201A (ESIEE Paris)

• Know what to expect in enterprise

  • Understand tools, processes, and techniques for deploying data science software into a production environment as you would in a corporate role.
  • Look at a dataset and / or business problem and provide solutions
  • Be able to talk and understand data science software dev and infrastructure
  • Be able to speak Agile & Scrum

• Know your worth

  • Developer
  • Data Scientist: Machine Learning Engineer
  • DevOps: Site Reliability Engineer, Infrastructure Engineer, Cloud Engineer

Checkout the app: https://simple-iris-webapp.herokuapp.com/

Note: The website may also be down from time to time (free plan).

Heroku: Build apps for free on Heroku

1. cd into DSIA5201A_app
2. Create the environment [1]
3. Activate the environment and install libraries [2]
4. Export the environment variable and run the app [3]

Exit the project:

5. Quit the application with CTRL+C
6. Deactivate the environment [4]

[1] Create the environment

python3 -m venv venv/

[2] Activate the environment and install libraries

. venv/bin/activate
pip install -r requirements.txt

Ensure that you are using the right pip:

which pip

Output should be something like : path_to/DSIA5201A/DSIA5201A_app/venv/bin/pip

If not:

alias pip=path_to/DSIA5201A/DSIA5201A_app/venv/bin/pip

Alternatively you can specify the path to the pip binary you want to use before any pip command:

path_to/venv/bin/pip yourcommand

[3] Export the environment variable and run the app

export FLASK_APP=DSIA5201A_app
pip install -e .
python3 -m flask run

[4] Deactivate the environment

deactivate

Misc

Output installed packages in requirements format (alternative pip list):

pip freeze

Generate a requirements file:

pip freeze > requirements.txt

Turn on the development features

export FLASK_ENV=development

simple-iris-webapp/DSIA5201A_app/: the app

simple-iris-webapp/DSIA5201A_app/DSIA5201A_app/static/ai/: python code to build the classifier

If you want to rebuild manually the docker image

cd into DSIA5201A_app folder:

cd path_to/DSIA5201A_app

Create the docker containers (depending on your computer and your connexion this operation can take a while)

docker build -t app_flask:latest .

If you don't want to rebuild manually the docker image

Fetch the image from docker hub:

docker pull eliasbouillanne/dsia5201_group4:latest

Launch the container

docker run -p 5000:5000 app_flask

Open: http:https://localhost:5000/