# Advancing RAG with Fine-tuned Gemma - 廖聖傑(JimmyLiao)

{%hackmd @HWDC/BJOE4qInR %}

>#### 》[議程介紹](https://hwdc.ithome.com.tw/2024/lab-page/3234)
>#### 》[填寫議程滿意度問卷｜回饋建言給辛苦的講者](https://forms.gle/M5dLgWuLjx4yUsMq7)

> Slide: https://docs.google.com/presentation/d/1e7wKsOAp3LcKUW4iKXkwyWBnKn6E90YIm0kVpcXi1Ng/mobilepresent?slide=id.g2fdf3371a50_0_524

## Part0 - RAG

特定知識的問答系統，跟 ChatGPT 不同

ref:
https://www.google.com/url?q=https://www.linkedin.com/posts/mr-deepak-bhardwaj_ai-machinelearning-datascience-activity-7238890112581787648-ImO7&sa=D&source=editors&ust=1726023675281826&usg=AOvVaw02y6N4rFgH2HRiuPjciQvf


## Part1 - Dify

Cloud: https://cloud.dify.ai/apps
Source: https://github.com/langgenius/dify

Template: Knowledge Retreival + Chatbot

Low Code 快速 PoC 使用

## Part2

## Part3 - Fine-Tune

開源模型 => 初始化參數(Adapter) + Fundation Model

微調完可以 Match 某程度的特定能力（如公司相關的領域知識）,但也會伴隨 Loss 某些其他能力

https://github.com/jimmyliao/workshop-gemma