對LLM進行了深入的介紹。然後，我們將探索各種主流架構框架，包括專有模型（GPT-4o、o1/Gemini 2/ Claude）和開源模型，並分析了它們的獨特優勢和差異。我們將重點放在基於 Python 的輕量級框架 aisuite、LangChain 與 介面Gradio\Streamlit 。我們引導完成創建智慧agent的過程，這些agent能夠從非結構化資料中檢索資訊，並使用 LLM 和強大的工具處理結構化資料。此外，還涉足了LLM\SLM領域，除了語言建模，涵蓋一些AI任務和模式，如視覺和音訊. A detailed introduction to LLM was made. We will then explore various mainstream architecture frameworks, including dedicated models (GPT-4o, o1/Gemini 2/Claude) and open source models, and analyze their unique advantages and differences. We will focus on lightweight frameworks based on Python aisuite, LangChain and interface Gradio\Streamlit. We lead the process of creating smart agents that can retrieve information from non-structured data and process structured data using LLM and powerful tools. In addition, it has also entered the LLM\SLM field, which in addition to language modeling, covers some AI tasks and models, such as vision and audio.
LangChain https://python.langchain.com/docs/tutorials/
Llama https://ai.meta.com/llama/get-started/
Streamlit https://docs.streamlit.io/
Gradio https://huggingface.co/blog/gradio-5
LangChain https://python.langchain.com/docs/tutorials/
Llama https://ai.meta.com/llama/get-started/
Streamlit https://docs.streamlit.io/
Gradio https://huggingface.co/blog/gradio-5
| 評分項目 Grading Method | 配分比例 Grading percentage | 說明 Description |
|---|---|---|
| Google Colab 程式作業Google Colab 程式作業 Google Colab Programming |
20 | aisuite 與 LangChain操作應用 |
| Python 程式作業Python 程式作業 Python Programming |
30 | VScode開發Streamlit 實作應用 |
| 期末報告期末報告 Final report |
30 | 生成式AI應用實作 |
| 出席出席 Attend |
20 |
