We will create a RAG system that searches the AI White Paper of the Catalan AI Strategy
These notebooks shows how to create a basic Retrieval Augmented Generation (RAG) system for Catalan using a QA-optimized models based on the Flor foundational models created for the AINA project. FlorQARAG and Flor1.3RAG
These demos should run with the Tesla T4 instance provided free of charge in Google Colab, but the 1.3 model can also run on a local machine CPU with a few adjustments. You can create a copy of these notebooks (RAGDemo_LlamaIndex_Flor6.3_EN.ipynb and RAGDemo_LlamaIndex_Flor1.3_EN.ipynb) in your drive and use another pdf documents to interrogate, if you wish.
For a more in-depth description of RAG, go to this blog from IBM research: "What is retrieval-augmented generation?"
Demo with 6.3b model
Demo with 1.3b model