56.9K ·GitHub
What is this?
It's a free Chinese-language tutorial series called 'Dive into LLMs' (《动手学大模型》) that teaches large language model concepts through hands-on coding exercises, built from Shanghai Jiao Tong University's NLP and AI-security course materials.
It's a free Chinese-language tutorial series called 'Dive into LLMs' (《动手学大模型》) that teaches large language model concepts through hands-on coding exercises, built from Shanghai Jiao Tong University's NLP and AI-security course materials.Is this one course or a collection of separate topic tutorials?It's a collection of 11 independent topic modules (fine-tuning/deployment, prompt learning, knowledge editing, math reasoning, watermarking, jailbreak attacks, LLM steganography, multimodal models, GUI agents, agent security, RLHF alignment), each with slides, a written tutorial, and runnable Jupyter notebook scripts.
It's a collection of 11 independent topic modules (fine-tuning/deployment, prompt learning, knowledge editing, math reasoning, watermarking, jailbreak attacks, LLM steganography, multimodal models, GUI agents, agent security, RLHF alignment), each with slides, a written tutorial, and runnable Jupyter notebook scripts.How do I start using these tutorials?Pick a topic from the README's tutorial table and open its slides (课件), written guide (教程), and Jupyter notebook script (脚本); each module is self-contained.
Pick a topic from the README's tutorial table and open its slides (课件), written guide (教程), and Jupyter notebook script (脚本); each module is self-contained.Do I need prior Python or deep-learning experience?Not explicitly stated, but since the material covers fine-tuning, RLHF/PPO, and model editing with runnable notebooks, basic Python and machine-learning familiarity is expected to follow along.
Not explicitly stated, but since the material covers fine-tuning, RLHF/PPO, and model editing with runnable notebooks, basic Python and machine-learning familiarity is expected to follow along.Is the content guaranteed to be accurate and complete?No — the authors state it's an ongoing project drawn from personal experience and internet data, with no guarantee everything is fully correct, and issues/PRs are welcomed to fix gaps.
No — the authors state it's an ongoing project drawn from personal experience and internet data, with no guarantee everything is fully correct, and issues/PRs are welcomed to fix gaps.Who is this tutorial series for?It's aimed at students and researchers wanting a quick, practical entry point into LLM topics for coursework or academic research, not production engineering.
It's aimed at students and researchers wanting a quick, practical entry point into LLM topics for coursework or academic research, not production engineering.Is there anything beyond the original tutorial series?Yes — a separate, Huawei Ascend (昇腾)-based series called 大模型开发全流程 (Full LLM Development Pipeline) was added, with beginner/intermediate/advanced tracks including slides, lab manuals, and videos.
Yes — a separate, Huawei Ascend (昇腾)-based series called 大模型开发全流程 (Full LLM Development Pipeline) was added, with beginner/intermediate/advanced tracks including slides, lab manuals, and videos.What makes the LLM steganography module interesting?It shows how to make a model's generated text covertly carry a hidden message readable only by those who know the scheme, while the output still reads naturally.
It shows how to make a model's generated text covertly carry a hidden message readable only by those who know the scheme, while the output still reads naturally.What does the jailbreak-attack module teach?It walks through how adversarial prompting techniques can bypass a model's safety guardrails, framed as learning attacks to better understand defenses.
It walks through how adversarial prompting techniques can bypass a model's safety guardrails, framed as learning attacks to better understand defenses.What are similar projects?nlp_course, LLM101n, ai-engineering-hub, llm_interview_note, llm-course, self-llm, ai-agents-the-definitive-guide, Ai-jailbreak
nlp_course, LLM101n, ai-engineering-hub, llm_interview_note, llm-course, self-llm, ai-agents-the-definitive-guide, Ai-jailbreakWhich topics is it in?rlhf, jailbreak, ai-security, reasoning, curriculum, course, jupyter notebook, fine-tuning, multimodal, academic-research, large-language-models, nlp
rlhf, jailbreak, ai-security, reasoning, curriculum, course, jupyter notebook, fine-tuning, multimodal, academic-research, large-language-models, nlp