FedRAG
An open-source Python framework for fine-tuning retrieval-augmented generation systems across both centralized and federated architectures.
Visit website (opens in a new tab)vectorinstitute.github.io
FedRAG is a framework for fine-tuning retrieval-augmented generation (RAG) systems, supporting both centralized training and federated learning across multiple parties. It provides RAG fine-tuning methods that can be federated with relatively little extra work.
It integrates with popular frameworks including HuggingFace, LlamaIndex and LangChain, and offers lightweight abstractions intended to simplify fine-tuning while keeping full control over the process.
It is installable from PyPI or conda-forge, with optional extras such as a HuggingFace integration, and is licensed under Apache-2.0.