Data certification API

Le Wagon Data Science certification exam starter pack for the predictive API test.

💡  This challenge is completely independent of other challenges. It is not required to complete any other challenge in order to work on this challenge.

Setup

Duplicate the repository for the API challenge

📝  Let's duplicate the repository of the API challenge.

Go to https://github.com/lewagon/data-certification-api:

• Click on Use this template
• Enter the repository name data-certification-api
• Set it as Public
• Click on Create repository from template
• Click on Code
• Select SSH
• Copy the SSH URL of the repository (the format is git@github.com:YOUR_GITHUB_NICKNAME/data-certification-api.git)

Clone the repository for the API challenge

📝  Now we will clone your new repository.

Open your terminal and run the following commands:

👉  replace YOUR_GITHUB_NICKNAME with your github nickname and PASTE_REPOSITORY_URL_HERE with the SSH URL you just copied:

cd ~/code/YOUR_GITHUB_NICKNAME
git clone PASTE_REPOSITORY_URL_HERE
cd data-certification-api

Look around

💡  The content of the challenge should look like this:

tree

.
├── Dockerfile
├── MANIFEST.in
├── Makefile
├── README.md
├── api
│   ├── __init__.py
│   └── app.py
├── exampack
│   ├── __init__.py
│   ├── data
│   ├── models
│   ├── predictor.py
│   ├── tests
│   │   └── __init__.py
│   └── utils.py
├── notebooks
├── requirements.txt
├── scripts
│   └── exampack-run
└── setup.py

Open your favourite text editor and proceed with the challenge.

API challenge

📝  In this challenge, you are provided with a trained model saved as model.joblib. The goal is to create an API that will predict the popularity of a song based on its other features.

👉  You will only need to edit the code of the API in api/app.py 🚨

👉  The package versions listed in requirements.txt should work out of the box with the pipelined model saved in model.joblib

Install the required packages

The requirements.txt file lists the exact version of the packages required in order to be able to load the pipelined model that we provide.

pip install -r requirements.txt

👉  If you encounter a version conflict while installing the packages 👈

 

In this case you will need to create a new virtual environment in order to be able to load the pipeline.

👉  Only execute this commands if you encounter an issue while installing the packages 🚨

pyenv install 3.8.6
pyenv virtualenv 3.8.6 certif
pyenv local certif
pip install -r requirements.txt

Run a uvicorn server

📝  Start a uvicorn server in order to make sure that the setup works correctly.

Run the server:

uvicorn api.app:app --reload

Open your browser at http://localhost:8000/

👉  You should see the response { "ok": true }

You will now be able to work on the content of the API while uvicorn automatically reloads your code as it changes.

API specification

Predict the popularity of a Spotify song

GET /predict

Parameter	Type	Description
acousticness	float	whether the track is acoustic
danceability	float	describes how suitable a track is for dancing
duration_ms	int	duration of the track in milliseconds
energy	float	represents a perceptual measure of intensity and activity
explicit	int	whether the track has explicit lyrics
id	string	id for the track
instrumentalness	float	predicts whether a track contains no vocals
key	int	the key the track is in
liveness	float	detects the presence of an audience in the recording
loudness	float	the overall loudness of a track in decibels
mode	int	modality of a track
name	string	name of the track
release_date	string	release date
speechiness	float	detects the presence of spoken words in a track
tempo	float	overall estimated tempo of a track in beats per minute
valence	float	describes the musical positiveness conveyed by a track
artist	string	artist who performed the track

Returns a dictionary with the artist, the name of the song and predicted popularity as an integer.

Example request:

/predict?acousticness=0.654&danceability=0.499&duration_ms=219827&energy=0.19&explicit=0&id=0B6BeEUd6UwFlbsHMQKjob&instrumentalness=0.00409&key=7&liveness=0.0898&loudness=-16.435&mode=1&name=Back%20in%20the%20Goodle%20Days&release_date=1971&speechiness=0.0454&tempo=149.46&valence=0.43&artist=John%20Hartford

Example response:

{
  "artist": "John Hartford",
  "name": "Back in the Goodle Days",
  "popularity": 22
}

👉 It is your turn, code the endpoint in api/app.py. If you want to verify what data types the pipeline expects, have a look at the docstring of the create_pipeline method in exampack/trainer.py.

API in production

📝  Push your API to production on the hosting service of your choice.

👉  If you opt for Google Cloud Platform 👈

 

Once you have changed your GCP_PROJECT_ID in the Makefile, run the directives of the Makefile to build and deploy your containerized API to Container Registry and finally Cloud Run.