Stage 4: Deep Learning and Neural Networks

A large model programming practice tutorial developed by Shanghai Jiao Tong University, covering a complete set of learning resources on 11 core topics including fine-tuning deployment, prompt engineering, model editing, and security techniques.

Start Learning AI →
LLMDeepLearningMLGitHubTextFree中文

"Hands-on Large Models" Course Detailed Introduction

Course Overview

"Hands-on Large Models" is a practical programming tutorial for large models developed by Shanghai Jiao Tong University, originating from the Spring 2024 course "AI Security Technologies" (NIS3353). Led by Instructor Zhang Zhuosheng, this tutorial aims to provide learners with an introductory programming reference related to large models. It is a public welfare, completely free, open-source educational resource.

Tutorial Features

  • Practice-Oriented: Helps learners quickly get started with large models through simple practical exercises.
  • Completely Free: Public welfare in nature, without any cost.
  • Industry-Academia Integration: An extension of course lectures from a top university.
  • Continuously Updated: Significant updates were made in June 2025, adding new topics and localized content.

Core Teaching Content

Main Chapter Structure

Chapter 1: Fine-tuning and Deployment

  • Key Focus: Guide to pre-trained model fine-tuning and deployment.
  • Learning Objectives: Master how to select appropriate pre-trained models, fine-tune them for specific tasks, and deploy the fine-tuned models as usable demos.
  • Resources Provided:
    • Courseware (PDF format)
    • Detailed tutorial documentation
    • Practice scripts (Jupyter Notebook)

Chapter 2: Prompt Engineering

  • Learning Resources:
    • dive-into-prompting.pdf Courseware
    • README.md Tutorial documentation
    • dive-prompting.ipynb Practice scripts

Chapter 3: Model Editing

  • Learning Resources:
    • dive_edit_0410.pdf Courseware
    • Complete tutorial documentation
    • dive_edit.ipynb Practice scripts

Chapter 4: Mathematical Reasoning

  • Learning Resources:
    • math.pdf Courseware
    • Tutorial documentation
    • sft_math.ipynb Mathematical reasoning practice scripts

Chapter 5: Watermarking Techniques

  • Learning Resources:
    • watermark.pdf Courseware
    • Tutorial documentation
    • watermark.ipynb Practice scripts

Chapter 6: Jailbreak Attacks

  • Learning Resources:
    • dive-Jailbreak.pdf Courseware
    • Tutorial documentation
    • dive-jailbreak.ipynb Practice scripts

Chapter 7: Steganography

  • Learning Resources:
    • stega.pdf Courseware
    • Tutorial documentation
    • llm_stega.ipynb Practice scripts

Chapter 8: Multimodal Large Models

  • Learning Resources:
    • mllms.pdf Courseware
    • Tutorial documentation
    • mllms.ipynb Practice scripts

Chapter 9: GUI Agent

  • Learning Resources:
    • GUIagent.pdf Courseware
    • Tutorial documentation
    • GUIagent.ipynb Practice scripts

Chapter 10: AI Safety

  • Learning Resources:
    • dive-into-safety.pdf Courseware
    • Tutorial documentation
    • agent.ipynb Practice scripts

Chapter 11: Large Model Alignment (RLHF)

  • Learning Resources:
    • RLHF.pdf Courseware
    • Tutorial documentation
    • RLHF.ipynb Practice scripts

Special Highlights

Localized "Full Workflow for Large Model Development" Series

Updated in June 2025, this tutorial, in collaboration with the Huawei Ascend Community, launched a localized version:

  • Technical Support: Developed based on Huawei Ascend foundational software and hardware.
  • Tutorial Format: Includes PPTs, experiment manuals, videos, and other diverse formats.
  • Difficulty Levels: Divided into beginner, intermediate, and advanced series.
  • Learning Path: Caters to different large model practical needs.
  • Practice-Oriented: Presents cutting-edge technologies through code practice.

Learning Resource Structure

Each chapter provides three core types of resources:

  1. Courseware (PDF): Theoretical knowledge and concept explanations.
  2. Tutorial Documentation (README.md): Detailed step-by-step guidance.
  3. Practice Scripts (Jupyter Notebook): Executable code examples.

Target Audience

  • Beginners hoping to get started with large models.
  • Students needing to complete course projects.
  • Researchers engaged in academic research.
  • Developers interested in large model practices.
  • Professionals looking to understand AI safety technologies.

Technical Features

  • Strong Practicality: Each chapter has corresponding executable code.
  • Good Systematization: Progresses systematically from basic concepts to advanced applications.
  • High Cutting-Edge Relevance: Covers the latest large model technologies and security issues.
  • Easy to Understand: Derived from university courses, with a complete teaching system.

Access and Usage

  • Open Source and Free: The project is completely open source, and anyone can access it for free.
  • Continuous Updates: The project team continuously maintains and updates the content.
  • Community Support: Welcome to submit Issues and Pull Requests.
  • Academic Background: Developed based on an official course at Shanghai Jiao Tong University.

Update History

  • June 2025: Major update, adding localized content and new topics.
  • Ongoing Maintenance: The project team regularly updates and refines the content.

Note: All content in this tutorial is based on the contributors' personal experience, internet data, and research work accumulation, and is provided for reference and learning purposes only.