rllm-team/rllm

Pytorch Library for Relational Table Learning with LLMs.

What it solves

rLLM addresses the complexity of Relational Table Learning (RTL). It provides a standardized way to integrate and co-train different types of models—specifically Graph Neural Networks (GNNs), Large Language Models (LLMs), and Tabular Neural Networks (TNNs)—to better analyze data stored across multiple linked tables.

How it works

It is a PyTorch library that breaks down state-of-the-art GNNs, LLMs, and TNNs into standardized modules. This allows developers to build new models by combining, aligning, and co-training these modules. It treats various graph structures (such as e-commerce or social networks) as multiple tables connected by foreign keys, making them compatible with this relational approach.

Who it’s for

This tool is designed for researchers and developers working on tabular data governance, relational table learning, and AI applications that require integrating structured relational data with LLMs.

Highlights

  • Modular Interface: Designed for LLM-oriented applications and integrates with Hugging Face transformers and LangChain.
  • Extensive Model Library: Includes over 15 state-of-the-art GNN and TNN models, such as InRTL, OGC, and ExcelFormer.
  • Versatile Data Handling: Capable of processing diverse graph types by treating them as linked tables.
  • New Benchmarks: Introduces three new relational table datasets for model design and testing.

Related

  • Project
    stanford-star/relbench

    RelBench is an open benchmark for predictive machine learning on relational databases, providing a standardized way to load and evaluate models on multi-table data.

  • Project
    pytorch-tabular/pytorch_tabular

    A high-level PyTorch framework that provides a unified interface for applying various deep learning architectures to tabular data.

  • Project
    pyg-team/pytorch-frame

    A modular deep learning extension for PyTorch designed to build neural network models on heterogeneous tabular data containing mixed column types like text, images, and numerical values.

  • Project
    pykt-team/pykt-toolkit

    A PyTorch-based library for training and benchmarking deep learning knowledge tracing models with standardized data preprocessing and multiple model implementations.

  • Project
    TorchSpatiotemporal/tsl

    A PyTorch-based library for neural spatiotemporal forecasting that provides a unified framework for data preprocessing, model prototyping, and Graph Neural Network implementation.