dive-into-llms

Why we included this project

Built from two Shanghai Jiao Tong University courses, one on NLP and one on AI security, this free tutorial series is written in Chinese and aimed at people getting their first real hands-on experience with large language models. Each chapter pairs lecture slides with runnable Jupyter notebooks, so you can work through fine-tuning and deploying a pretrained model, prompting and chain-of-thought reasoning, knowledge editing, and distilling a small math-reasoning model. The later chapters go further into research territory, covering text watermarking, jailbreak attacks, steganography, multimodal models, GUI agents, and RLHF safety alignment, so students planning coursework or academic projects will find material beyond the usual tutorials. A companion full-pipeline course built with the Huawei Ascend community extends the material to model migration and tuning on domestic hardware. If you learn best by running code rather than reading theory, the notebooks are a good place to start.