arrow-left

All pages
gitbookPowered by GitBook
1 of 10

Loading...

Loading...

Loading...

Loading...

Loading...

Loading...

Loading...

Loading...

Loading...

Loading...

Week 5 - Advanced Retrieval Augmented Generation

hashtag
Sessions schedule

  • Quiz

  • Short Recap

  • Impulses on Sparse Priming Representations (SPRs) and HyDE (Hypothetical Document Embeddings)

  • Project Prototype Presentations

  • Homework for next week

hashtag
Resources

hashtag
Homework for next week

  • Watch the course “”

  • Install Gradio and explore the examples from the Gradio playground locally ()

  • Think about the elements your application user interface should contain, create a sketch or wireframe of your app and implement it with Gradio

Week 8 - Open-Source LLMs

hashtag
today's schedule

  • special guest: Jan Monika

(If possible connect your prototype to the user interface)

file-pdf
1MB
231128_Advanced Retrieval_Augmented_Generation.pptx.pdf
PDF
arrow-up-right-from-squareOpen
Building Generative AI Applications with Gradioarrow-up-right
https://www.gradio.app/playgroundarrow-up-right

Tips 'n Tricks on Open Source LLMs

  • Questions

  • The Template for Code Submission

  • Guidelines for the Final Presentation

  • hashtag
    Slides

    file-pdf
    406KB
    231219_OpenSource_LLMs_and_Guest.pptx.pdf
    PDF
    arrow-up-right-from-squareOpen

    Week 1- General Introduction

    hashtag
    Session content

    • Getting to know each other

    • What can you expect

    • Requirements for credits

    • Course outline

    • Demos

    • Breakout session

    • Content and homework

    hashtag
    Todo until next session

    Watch and read the course content

    • Register on

    • Watch "" (20 min)

    • Watch the short course "" and work through the Jupyter notebooks (~1h 30min)

    Tasks

    1. Watch the provided videos and review the text to identify various tactics and techniques used in Prompt Engineering. List these tactics and techniques with clear examples.

    2. Handling Ambiguity

      The word “Java” has at least three meanings (programming language, island in Indonesia, or coffee).

      Task:

    hashtag
    Additional resources

    • Slides of Session 1:

    Readings on simple prompting techniques

    Week 3 - Introduction to LangChain

    hashtag
    content

    • Quiz

    • Homework presentation

    • short recab on LangChain

    • overview of project ideas & pitches

    • building project groups

    hashtag

    hashtag
    this sessions presentation

    hashtag
    homework

    • Watch and work through the course “” on Deeplearning.AI

    • Work through the tutorial , understand the code and do the tasks

    file-pdf
    2MB
    141123_Introduction to LangChain.pdf
    PDF
    arrow-up-right-from-squareOpen
    LangChain: Chat with your Dataarrow-up-right
    HomeworkRAGarrow-up-right
    file-download
    18KB
    HomeworkRetrievalAugmentedGeneration.ipynb
    arrow-up-right-from-squareOpen

    Watch "" (10 min)

  • Read "" and ""

  • Try out the following prompt:

    Can you tell me about Java?

    Task:

    Suppose you are developing an app and do not know beforehand what your app's user means. Can you improve the prompt to help the model to clarify what the user represents and generate an answer?

  • Generate three potential course project ideas that you could work on. Be prepared to discuss these ideas in the next session. Feel free to consider alternative projects if needed.

  • Select one of the potential project ideas from Task 3 and create prompts for the language model to solve the problem. Use different prompting tactics and techniques to see which one produces the best results in addressing the chosen project idea.

  • To be continued...
    learn.deeplearning.aiarrow-up-right
    Language Models, the Chat Format and Tokensarrow-up-right
    ChatGPT Prompt Engineering for Developersarrow-up-right
    file-pdf
    785KB
    241023_GeneralIntroduction_PEwLLMs.pptx.pdf
    PDF
    arrow-up-right-from-squareOpen
    Instruction promptingarrow-up-right
    Role promptingarrow-up-right
    Priming promptarrow-up-right
    Chain-of-Thought Reasoningarrow-up-right
    Contructing promptsarrow-up-right
    Prompt Engineering for Generative AIarrow-up-right

    Week 6 - Building User Interfaces with Gradio

    hashtag
    today's topics

    • Project Prototype Presentations

    • Gradio

    • Self-hosted models with LM Studio

    • Breakout Session ‘Concrete Planning’

    • Homework for next week

    hashtag
    resources

    hashtag
    homework

    • Watch the videos “” and “” from the course “”

    • Watch the video “” from the course “”

    Watch the video from Sam Witteveen

    Week 7 - Evaluation of LLM outputs and structured outputs

    hashtag
    today's schedule

    • Quiz

    • Short Recap

    • Short presentation of current status

    • Breakout Session ‘Next Steps’

    • Homework for next week

    hashtag
    resources

    hashtag
    homework

    • Take a look at this article:

    • Take a look at some open-source / open-access LLM frameworks:

    file-pdf
    258KB
    231205_Building User Interfaces with Gradio.pptx.pdf
    PDF
    arrow-up-right-from-squareOpen
    Evaluation Part Iarrow-up-right
    Evaluation Part IIarrow-up-right
    Building Systems with the ChatGPT APIarrow-up-right
    Evaluationarrow-up-right
    LangChain for LLM Application Developmentarrow-up-right
    “Using LangChain Output Parsers to get what you want out of LLMs”arrow-up-right

    Ollama (Mac, Linux):

  • LM Studio (Windows, Mac, Linux): - --

  • GPT4All

  • Test some open-source/open-access LLMs either locally downloaded or in HuggingFace spaces or any other test environment (e.g. HuggingChat, H2O.ai, etc.)

    • Do you see significant differences in the output of proprietary and open-source/open-access LLMs?

  • file-pdf
    687KB
    231212_Evaluation of LLM outputs and structured output.pptx (1).pdf
    PDF
    arrow-up-right-from-squareOpen
    https://blog.n8n.io/open-source-llm/arrow-up-right
    https://www.youtube.com/watch?v=Ox8hhpgrUi0arrow-up-right
    https://www.youtube.com/watch?v=k_1pOF1mj8karrow-up-right
    https://medium.com/@genebernardin/running-llms-locally-using-lm-studio-38070f286413arrow-up-right

    Practical Engineering with LLMs

    Week 4 - Introduction to Retrieval Augmented Generation

    hashtag
    This sessions schedule

    • Quiz

    • Homework presentation

    • Short Recap

    • Projects

    • Homework for next week

    hashtag
    Ressources

    hashtag
    Homework

    1. Watch the course “”

    2. Prepare a short presentation (max 5 min per group) with 2 parts:

    • Present your project outline using the template (or improvise)

    • Present a rough prototype (or mockup) using LangFlow

    Use one of the available no-code tools (or, if you prefer, LangChain) to test the feasibility of your idea by creating a first small prototype.

    LangFlow:

    Flowise:

    GPTs:

    file-pdf
    1MB
    231121_Introduction_to_Retrieval_Augmented_Generation_final.pptx.pdf
    PDF
    arrow-up-right-from-squareOpen
    Large Language Models with Semantic Searcharrow-up-right
    https://github.com/logspace-ai/langflowarrow-up-right
    https://github.com/FlowiseAI/Flowisearrow-up-right
    https://openai.com/blog/introducing-gptsarrow-up-right

    Week 9 - Project Presentations

    Week 2 - Prompt Engineering

    hashtag
    Session Content

    • Quiz

    • Homework presentation

    • Short recap on intro to LLMs & prompt engineering

    • Breakout session on prompt hubs

    • Anatomy of an app

    • Breakout session on project ideas

    • Presentation of project ideas and default projects

    • Homework for next week

    hashtag
    Session Slides

    hashtag
    Todo until next session

    • Watch the short course (~1h 40min)

    • If the videos and are too confusing for you right now, as they introduce a lot of concepts in a short amount of time, you can skip these videos as this content will also be part of later sessions.

    • Work through the Jupyter notebook and complete the tasks at the bottom of the notebook.

    (Copy the notebook to your Google Drive before working on it.)

    file-pdf
    1MB
    231107_PromptEngineering.pdf
    PDF
    arrow-up-right-from-squareOpen
    "LangChain for LLM Application Development"arrow-up-right
    "Question and Answer"arrow-up-right
    "Evaluation"arrow-up-right
    "HomeworkLangChain"arrow-up-right
    file-download
    14KB
    HomeworkLangChain.ipynb
    arrow-up-right-from-squareOpen
    "Colab Link HomeworkLangChain"arrow-up-right