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.

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

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

• Click on Use this template
• Enter the repository name data-certification-api-movies
• 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 [email protected]:YOUR_GITHUB_NICKNAME/data-certification-api-movies.git)

📝  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-movies

💡  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
│   ├── tests
│   │   └── __init__.py
│   ├── trainer.py
│   └── utils.py
├── model.joblib
├── notebooks
├── requirements.txt
├── scripts
│   └── exampack-run
└── setup.py

Open your favourite text editor and proceed with the 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 movie 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

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

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

📝  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.

Predict the popularity of a Spotify song

GET /predict

Returns a dictionary with the title of the movie, and predicted popularity as a float.

Example request:

/predict?title=Harry%20Potter&original_title=Harry%20Potter&release_date=2010-06-09&duration_min=150&description=Harry%20is%20a%20wizard%20that%20tries%20to%20save%20the%20world%20from%20crazy%20guys&budget=1000000&original_language=en&status=Released&number_of_awards_won=80&number_of_nominations=120&has_collection=1&all_genres=Fantasy,%20Family,%20Adventure&top_countries=United%20States%20of%20America,,%20United%20Kindgom&number_of_top_productions=3&available_in_english=True

Example response:

{
  "title": "Harry Potter",
  "popularity": 15
}

👉 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.

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