Summary

This repository offers open-source educational slides and code for AI algorithms, focusing on deep learning and reinforcement learning. It targets engineers, researchers, and students, bridging theoretical math principles with practical code implementation for a deep understanding of AI mechanics.

How It Works

AI concepts are presented via detailed mathematical derivations and visual explanations, linked to Bilibili video lectures. Each topic includes downloadable slide decks (PPTX, PDF) and code snippets, covering Transformer architectures (RoPE, KV Cache), parameter-efficient tuning (LoRA), and RL algorithms (Q-Learning, DQN, PPO). The approach emphasizes understanding AI mechanics through rigorous mathematical grounding.

Quick Start & Requirements

This is an educational resource, not a runnable project. No installation is required. Users can access and download slide decks (PPTX, PDF) and view associated video lectures via Bilibili.

Highlighted Details

• Covers LLM topics: Transformer components (RoPE, KV Cache), DeepSeek sparse attention (MLA, NSA), and LoRA for efficient tuning.
• Explains RL algorithms from Q-Learning to PPO, including math derivations and code, with RLHF modules.
• Provides downloadable, editable PPTX slide decks and PDF versions.
• Links to Bilibili video lectures for each topic.

Maintenance & Community

Maintained by Bilibili user "东川路第一可爱猫猫虫" (@东川路第一可爱猫猫虫), with community engagement primarily via their Bilibili profile. Content is continuously updated.

Licensing & Compatibility

Licensed under CC BY-NC 4.0. Allows sharing and adaptation for non-commercial purposes with attribution. Commercial use is restricted.

Limitations & Caveats

This is a collection of educational notes and slides, not a deployable AI model or framework. Users seeking to run AI models directly will need separate implementations. Interaction is via slides and videos, with code as supplementary material.