Skip to content

Latest commit

 

History

History
28 lines (22 loc) · 662 Bytes

README.md

File metadata and controls

28 lines (22 loc) · 662 Bytes

trying_graph_rag

  • CPU-only GraphRag + Default RAG for Multi-Hop QA
  • Uses qunatized (Q4) Gemma2 2B through Ollama

Setup

curl -fsSL https://ollama.com/install.sh | sh
ollama serve &
ollama pull gemma2:2b
huggingface-cli login  # for gemma 2 tokenizer access

Indexing

poetry run python scripts/batch_index.py dataset/sampled_gold_data.json

Querying

poetry run python scripts/batch_query.py dataset/sampled_gold_data.json . dataset/graph_rag_prediction.csv

Evaluation

poetry run python scripts/hotpot_evaluate_v1.py dataset/graph_rag_prediction_formatted.json dataset/sampled_gold_data.json