本文为Kaggle上5 天 Gen AI 强化课程（Google Learn Guide）

Welcome to our 5-Day Gen AI Intensive Course with Google! This was a live event from November 11-15, 2024, now made available as a self-paced learning guide for anyone interested in learning more about the fundamental technologies and techniques behind Generative AI.

What's covered:

• Day 1: Foundational Models & Prompt Engineering - Explore the evolution of LLMs, from transformers to techniques like fine-tuning and inference acceleration. Get trained with the art of prompt engineering for optimal LLM interaction.
• Day 2: Embeddings and Vector Stores/Databases - Learn about the conceptual underpinning of embeddings and vector databases, including embedding methods, vector search algorithms, and real-world applications with LLMs, as well as their tradeoffs.
• Day 3: Generative AI Agents - Learn to build sophisticated AI agents by understanding their core components and the iterative development process.
• Day 4: Domain-Specific LLMs - Delve into the creation and application of specialized LLMs like SecLM and Med-PaLM, with insights from the researchers who built them.
• Day 5: MLOps for Generative AI - Discover how to adapt MLOps practices for Generative AI and leverage Vertex AI's tools for foundation models and generative AI applications.

Best of Luck!
- brought to you by Anant Nawalgaria, Mark McDonald, Paige Bailey, and many other contributors from Google.

Set Up

Follow the following steps to get set up before diving into the daily assignments:

Please note that if you would like to post on other channels on the Kaggle discord you will need to link your Kaggle account to discord here: https://kaggle.com/discord/confirmation.

Day 1 (Foundational Large Language Models & Prompt Engineering)

Welcome to Day 1.

Today you’ll explore the evolution of LLMs, from transformers to techniques like fine-tuning and inference acceleration. You’ll also get trained in the art of prompt engineering for optimal LLM interaction.

The code lab will walk you through getting started with the Gemini API and cover several prompt techniques and how different parameters impact the prompts.

Day 1 Assignments:

1. Complete the Intro Unit: “Foundational Large Language Models”, which is:

• [Optional] Listen to the summary podcast episode for this unit (created by NotebookLM).
• Read the “Foundational Large Language Models & Text Generation” whitepaper.

2. Complete Unit 1 – “Prompt Engineering”, which is:

• [Optional] Listen to the summary podcast episode for this unit (created by NotebookLM).
• Read the “Prompt Engineering” whitepaper.
• Complete this code lab on Kaggle where you’ll learn prompting fundamentals. Make sure you phone verify your account before starting, it's necessary for the code labs.
• [Optional] Read this case study to learn how a leading bank leveraged advanced prompt engineering and other contents discussed in assignments of day 1 to automate their financial advisory workflows, achieving significant productivity gains.

3. Watch the YouTube livestream recording. Paige Bailey will be joined by expert speakers from Google - Mohammadamin Barekatain, Lee Boonstra, Logan Kilpatrick, Daniel Mankowitz, Majd Merey Al, Anant Nawalgaria, Aliaksei Severyn and Chuck Sugnet to discuss today's readings and code labs.

Day 2 (Embeddings and Vector Stores/Databases)

Welcome to Day 2.

Today you will learn about the conceptual underpinning of embeddings and vector databases and how they can be used to bring live or specialist data into your LLM application. You’ll also explore their geometrical powers for classifying and comparing textual data.

Day 2 Assignments:

1. Complete Unit 2: “Embeddings and Vector Stores/Databases”, which is:

• [Optional] Listen to the summary podcast episode for this unit (created by NotebookLM).
• Read the “Embeddings and Vector Stores/Databases” whitepaper.
• Complete these code labs on Kaggle:
  • 1. Build a RAG question-answering system over custom documents
  • 2. Explore text similarity with embeddings
  • 3. Build a neural classification network with Keras using embeddings

2. Watch the YouTube livestream recording. Paige Bailey will be joined by expert speakers from Google - Omid Fatemieh, Jinhyuk Lee, Alan Li, Iftekhar Naim, Anant Nawalgaria, Yan Qiao, and Xiaoqi Ren to discuss embeddings and vector stores/databases.

Day 3 (Generative Agents)

Welcome to Day 3.

Learn to build sophisticated AI agents by understanding their core components and the iterative development process.
The code labs cover how to connect LLMs to existing systems and to the real world. Learn about function calling by giving SQL tools to a chatbot, and learn how to build a LangGraph agent that takes orders in a café.

Day 3 Assignments:

1. Complete Unit 3: “Generative Agents”, which is:

•  [Optional] Listen to the summary podcast episode for this unit (created by NotebookLM).
•  Read the “Generative AI Agents” whitepaper.
•  [Optional] Read a case study which talks about how a leading technology regulatory reporting solutions provider used an agentic generative AI system to automate ticket-to-code creation in software development, achieving a 2.5x productivity boost.
• Complete these code labs on Kaggle:
  • 1. Talk to a database with function calling
  • 2. Build an agentic ordering system in LangGraph

Day 4 (Domain-Specific LLMs)

Welcome to Day 4.

In today’s reading, you’ll delve into the creation and application of specialized LLMs like SecLM and MedLM/Med-PaLM, with insights from the researchers who built them.

In the code labs you will learn how to add real world data to a model beyond its knowledge cut-off by grounding with Google Search. You will also learn how to fine-tune a custom Gemini model using your own labeled data to solve custom tasks.

Day 4 Assignments:

1. Complete Unit 4: “Domain-Specific LLMs”, which is:

•  [Optional] Listen to the summary podcast episode for this unit (created by NotebookLM).
• Read the “Solving Domain-Specific Problems Using LLMs” whitepaper".
• Complete these code labs on Kaggle:
  • 1. Use Google Search data in generation
  • 2.Tune a Gemini model for a custom task

Day 5 (MLOps for Generative AI)

Welcome to Day 5.

Discover how to adapt MLOps practices for Generative AI and leverage Vertex AI's tools for foundation models and generative AI applications.

Day 5 Assignments:

1. Complete Unit 5: “MLOps for Generative AI”, which is:

•  [Optional] Listen to the summary podcast episode for this unit (created by NotebookLM).
• Read the “MLOps for Generative AI” whitepaper.
• No code lab for today! We will do a code walkthrough and live demo of goo.gle/e2e-gen-ai-app-starter-pack, a resource created for making MLOps for Gen AI easier and accelerating the path to production. Please go through the repository in advance.

2. Watch the YouTube livestream recording. Paige Bailey will be joined by expert speakers from Google - Advait Bopardikar, Sokratis Kartakis, Gabriela Hernandez Larios, Veer Muchandi, Anant Nawalgaria, Elia Secchi, and Olivia Wiles to discuss MLOps practices for generative AI.

Bonus Assignment

There's more!

This bonus notebook walks you through a few more things you can do with the Gemini API that weren't covered during the course. This material doesn't pair with the whitepapers or podcast, but covers some extra capabilities you might find useful when building Gemini API powered apps.

If you want to learn more about future live courses, please fill out our outreach form .

最后附上相关附件：