Recommender-system

Set up

Download rating.csv & movie.csv files from kaggle link

Download also the trained model pickle files at this google drive link Please unzip these files and put it under app/resources/ folder.

Environment variables to set

export ROOT_FOLDER=""
export PYTHONPATH="$PWD"

Install requirements

pip install -r requirements.txt

Setup Git hooks

pre-commit install #set up git hooks

Keep requirements clean and updated

Please make sure pip-tools is properly installed in your virtual environment with

pip install pip-tools

This should be done by adding the required package in the requirements.in file. Then using pip-tools the requirements.txt file can be reproduced.

pip-compile --output-file requirements.txt --quiet requirements.in
pip-sync requirements.txt

Or using the makefile

make requirements

Build docker image and run container

docker build -t registry.heroku.com/pagerank-webapp/web .
docker run --rm --name pagerank.webapp.container -e PORT=8080 -p 8080:8080 registry.heroku.com/pagerank-webapp/web:latest

Or using makefile

make dockerbuild
make dockerrun

To mount the datafiles in local, cd into project directory, uncomment the first 2 lines in .dockerignore and run:

make dockerrun-local-mount

Deploy on Heroku

Install heroku

brew tap heroku/brew && brew install heroku

Login

heroku login #opens web browser
heroku container:login #to tell heroku we use container login for deployment

Choose a name for webapp on Heroku website after login: pagerank-webapp

Push docker image & deploy

registry.heroku.com/pagerank-webapp/web
heroku container:release -a pagerank-webapp web

Remarks The name of the docker image should be in this format:registry.heroku.com/<app-name>/web

References

Pagerank modeling

1. https://nlp.stanford.edu/IR-book/html/htmledition/topic-specific-pagerank-1.html

Deploy using Heroku

1. https://testdriven.io/blog/fastapi-machine-learning/