# Stability AI, 세계 최초의 오픈소스 RLHF LLM 챗봇 StableVicuna 공개

> Clean Markdown view of GeekNews topic #9079. Use the original source for factual precision when an external source URL is present.

## Metadata

- GeekNews HTML: [https://news.hada.io/topic?id=9079](https://news.hada.io/topic?id=9079)
- GeekNews Markdown: [https://news.hada.io/topic/9079.md](https://news.hada.io/topic/9079.md)
- Type: news
- Author: [xguru](https://news.hada.io/@xguru)
- Published: 2023-05-01T10:32:02+09:00
- Updated: 2023-05-01T10:32:02+09:00
- Original source: [stability.ai](https://stability.ai/blog/stablevicuna-open-source-rlhf-chatbot)
- Points: 16
- Comments: 0

## Topic Body

- Vicuna v0 13B를 명령어 파인튜닝 및 RLHF 훈련  
- 3개의 데이터셋을 이용하여 SFT(Supervised Fine-Tuning)  
  - OpenAssistant Conversations Dataset(OASST1)  
  - GPT4All Prompt Generations   
  - Alpaca   
- HuggingFace에서 LLaMA모델에 대한 가중치 델타로 다운로드 가능(오리지널 LLaMA 데이터 필요)

## Comments



_No public comments on this page._