GH Repository · datawhalechina
tiny-universe
《大模型白盒子构建指南》：一个全手搓的Tiny-Universe
- stars
- 5,060
- 30-day movement
- +52/day
- Related entries
- 60
- Connections
- 1
Jupyter Notebooktransformersevaluation-metricsllamadiffusionagentqwenrag
An educational repository from Datawhale titled 《大模型白盒子构建指南》(A Guide to Building Large Models from a White-Box Perspective), described as a fully hand-built Tiny-Universe. Based on its topics and Jupyter Notebook format, it appears to be tutorial content covering building RAG, agents, diffusion models, and evaluation around LLMs such as Llama and Qwen.
Reach for it when you want to learn how LLM-adjacent components are built from scratch rather than calling them through a framework.
Use it to
- Work through hands-on notebooks on RAG and agents
- Study evaluation metrics for language models
- Learn diffusion model basics alongside LLMs
- Explore Llama and Qwen with transformers
For Chinese-speaking learners building LLM components from scratch
- Role
- rag
- Language
- Jupyter Notebook
- Forks
- 477
- Open issues
- 11
- Last push
- 2026-02-12
topicsragagentdiffusionevaluation-metricsllamaqwen