# datawhalechina/happy-llm

📚 从零开始构建大模型

Repository: https://github.com/datawhalechina/happy-llm
Canonical: https://ross.abutalabs.com/products/happy-llm
Homepage: https://datawhalechina.github.io/happy-llm/
Language: Jupyter Notebook
License: NOASSERTION
License Family: other
Topics: agent, llm, rag
Last push: 2026-08-08T17:18:48+00:00

## Health v2 (maintenance only)
Score: 70/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 96, release rhythm 44, longevity 59
- inputs: {"age_days": 827, "days_push": 25, "days_rel": 216, "gap_med": 116.0, "n_releases_24m": 3}
- flags: no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 33272, forks 3155 (observed 2026-08-28T04:11:58.198412+00:00)

## What it is
Happy-LLM is a free, open-source Chinese-language tutorial by Datawhale that teaches large language model principles from scratch, covering Transformer architecture, pretraining, and fine-tuning. Learners implement a full LLaMA2 model and practice RAG, Agent, and agentic-RL techniques through hands-on Jupyter Notebook exercises.

## Use cases
- learn how large language models work from scratch
- implement a LLaMA2 model by hand
- understand the Transformer architecture and attention mechanism
- practice pretraining and fine-tuning an LLM including LoRA and QLoRA
- learn how to build RAG and agent applications
- study agentic reinforcement learning like GRPO
- find a structured free LLM course in Chinese

## When to choose
- you want a systematic, hands-on curriculum for understanding and building LLMs from first principles
- you prefer learning by implementing models in code rather than just reading theory
- you want free, well-maintained educational material covering pretraining through RAG and agents

## When to avoid
- you need a production-ready LLM training framework or library rather than a tutorial
- you need English-only learning materials
- you want a quick reference guide for using existing LLM APIs instead of building models yourself

## Facets
- artifact type: learning-resource
- maturity: active
- function: llm-training, machine-learning, rag, agent-framework, prompt-engineering
- domain: large-language-models, deep-learning, tutorials, education
- platform: python, cross-platform
- tags: llm-from-scratch, transformer, llama2, fine-tuning, lora, chinese, datawhale, jupyter-notebook, pretraining, agentic-rl, natural-language-processing

## Member repositories
- datawhalechina/happy-llm (main) score 70

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:11:58.198412+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T16:28:29.067510+00:00, confidence not recorded.
  - readme: https://github.com/datawhalechina/happy-llm (fetched 2026-08-28T04:11:58.198412+00:00, sha fed0c90c3e65)
  - homepage: https://datawhalechina.github.io/happy-llm/ (fetched 2026-08-29T07:48:44.173231+00:00, sha 325d6b446972)
- Data as of 2026-08-30T08:39:29.467469+00:00.