NovaSearch-Team/RAG-Retrieval
An open-source library for unified fine-tuning and inference of RAG retrieval models including embedding, ColBERT, and reranker architectures.
Velocity · 7d
+1.4
★ / day
Trend
→steady
star history
RAG-Retrieval provides code for training, inference, and distillation of RAG retrieval systems. It supports fine-tuning any open-source RAG retrieval model including embedding models (BERT-based and LLM-based), late interaction models like ColBERT, and reranker models. The project includes a Python library that offers a unified API to call different ranking models for inference. It also supports knowledge distillation from larger models to smaller ones.