Azure-Samples/azureai-foundry-finetuning-raft
A recipe that will walk you through using either Meta Llama 3.1 405B or OpenAI GPT-4o deployed on Azure AI to generate a synthetic dataset using UC Berkeley's Gorilla project RAFT method.
This project helps improve the accuracy of Retrieval Augmented Generation (RAG) systems by teaching language models to better use retrieved information. It takes an existing RAG setup and uses powerful "teacher" models like GPT-4o or Llama 3.1 405B to create specialized training data. This data then fine-tunes a smaller "student" model, leading to more precise and relevant answers. This is for AI practitioners and RAG system developers looking to enhance their model's performance on specific knowledge domains.
No commits in the last 6 months.
Use this if you have a RAG system and want to make your language model more precise and reliable when answering questions based on retrieved documents or data.
Not ideal if you are looking for a general model distillation technique that doesn't focus on improving RAG system precision.
Stars
77
Forks
27
Language
Jupyter Notebook
License
MIT
Category
Last pushed
Jul 17, 2025
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/rag/Azure-Samples/azureai-foundry-finetuning-raft"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
llmware-ai/llmware
Unified framework for building enterprise RAG pipelines with small, specialized models
Sinapsis-AI/sinapsis-chatbots
Monorepo for sinapsis templates supporting LLM based Agents
aimclub/ProtoLLM
Framework for prototyping of LLM-based applications
xi029/Qwen3-VL-MoeLORA
在千问最新的多模态image-text模型Qwen3-VL-4B-Instruct 进行多种lora微调对比效果，通过langchain+RAG+多智能体(Multi-Agent)进行部署
pkargupta/taxoadapt
Dynamically constructs and adapts an LLM-generated taxonomy to a given corpus across multiple dimensions.