RecaLLMate: python framework for flexible and reproducible study of large language models based recommender systems
The framework is aimed at developing recommender systems that utilize large language models. It includes the following key modules:
- working with user and item features (data)
- functionality for loading and fine-tuning LLMs (llm)
- solving recommendation problems with LLMs (tasks)
- interface for wrapping and using classical RecSys modules
- functionality for using agent-based approaches, generating instruments, memory and parsing responses for agents (agents)
- evaluation of RecSys results (evaluate)
First of all, you need to install the necessary dependencies:
pip install -r requirements/requirements.txt