HyGRAG

The code implementation of HyGRAG, accepted by WWW' 26. "A Unified Framework for Context-Aware and Relation-Aware Graph Retrieval-Augmented Generation"

Quick Start 🚀

Dependencies

Ensure you have the required dependencies installed:

conda env create -f experiment.yml

Example Configuration (config.yaml):

llm:
  api_type: "open_llm"  # Options: "openai" or "open_llm" (For Ollama and LlamaFactory) 
  model: "YOUR_LOCAL_MODEL_NAME"
  base_url: "YOUR_LOCAL_URL"  # Change this for local models
  api_key: "YOUR_API_KEY"  # Not required for local models

For LlamaFactory or Ollama or vllm, ensure the model is correctly installed and running in your local environment.

You can refer to the Readme of LlamaFactory

llm:
  api_type: "open_llm"  # Options: "openai" or "open_llm" (For Ollama and LlamaFactory) 
  model: "YOUR_LOCAL_MODEL_NAME"
  base_url: "YOUR_LOCAL_URL"  # Change this for local models
  api_key: "ANY_THING_IS_OKAY"  # Not required for local models

Run Methods

1. start HyGRAG

python main.py -opt Option/Data/multihop-rag.yaml -dataset_name multihop-rag

2. incremental test

python main_incremental.py -opt Option/Ours/HKGraphTreeDynamic.yaml -dataset_name multihop-rag -mode incremental -incremental_ratio 0.2

---

Cite me.