MCFreddie777/2019

[DP] Session based recommendation system

Prerequisites

Getting Started

• Make sure you have installed the NVIDIA driver installed on the host system
• Docker
• Download dataset into /data/input directory.

Spawning and running the container

docker compose up

Usage

Go to http://127.0.0.1:8888 in your browser or click the link in the console.

Installing packages

This image contains conda + pip, therefore install the packages as followed:

docker compose exec jupyterlab bash -ic "conda install <package>"

or

docker compose exec jupyterlab bash -ic "pip install <package>"

To persist the packages when rebuilding the image, packages need to be freezed into environment file.

docker compose exec jupyterlab bash -ic "conda env export --no-build > environment.yml"

Scripts

Project also contains helper scripts to generate or process data directly from console. No need to run jupyter notebooks.

Usage

1. Clone .env.example file and rename it to .env file. Change the env variables if you wish or keep the defaults.

2. From root directory of project run

docker compose exec jupyterlab bash -ic "python src/<script_name>.py"

e.g.

docker compose exec jupyterlab bash -ic "python src/preprocess.py"