fastEASE 0.3.0

An easy-to-use Python library for building EASE recommendation systems with CUDA

fast EASE for CUDA

EASY is a unique and decisive approach to recommendation for a limited number of users and items. One challenge is that the matrix inversion process becomes computationally intensive, requiring significant processing time on the central processing unit (CPU). This issue is addressed in the current project by leveraging CUDA, a powerful technology specifically designed for parallel processing. The key distinction is that this solution is intended not for research purposes but rather for deployment in production environments.

Framework

Structure

• .github: GitHub Actions workflows.
• src: Library's source code.
• tests: Unit tests.
• .editorconfig: Editor settings for consistent coding style.
• .gitignore: Excludes files generated by Python and Poetry.
• LICENSE: License file.
• Makefile: Manage tasks like testing, linting, and formatting.
• pyproject.toml: Poetry's configuration file.

Getting Started

Prerequisites

• Python >= 3.10
• Poetry (should work with uv as well)
• GNU Make

Tested on Ubuntu 24.04 LTS and Debian 12. But the template should work on other operating systems as well.

Setting Things Up

1. Clone the repository:

   git clone https://github.com/fkrasnov2/fastEASE
   cd fastEASE
   

2. Install dependencies:

   make install
   

Development Workflow Management

# Run the unit tests
make test

# Lint the code
make lint