LLMs-Zero-to-Hero is a Jupyter Notebook project focused on training and fine-tuning large language models from scratch, with chapters and notebooks guiding pre-training, SFT, DPO, and RLHF workflows. It includes code structure for chapters, src/hero, and video-focused notebooks, plus deployment and tooling references.
Collecting history — the radar snapshots this repo daily. The trend line appears after 3 days of data (1 so far).
What it is
LLMs-Zero-to-Hero appears to be a Jupyter Notebook–based educational repository that documents from-scratch large language model training and fine-tuning workflows. It includes sections for full-stack topics such as pre-training and supervised fine-tuning, with references to related tools and a code structure that places final implementations under src/hero and chapter-wise notebooks under chapter directories. The project also links to accompanying videos, blogs, and external tools (ApeRouter, Bloome).
How it works
The repository organizes content into chapters and notebooks. The code layout shows a src/hero directory intended for final self-developed models, along with chapter01, chapter02 containing training-related scripts (e.g., train.py in chapter02) and video-oriented code under src/video. It references pipelines for large-model basics, MT and RL workflows, and deployment topics such as inference optimization and quantization. It also references external tooling and resources for learning and experimentation.
Getting started
The repository README implies a structured learning path rather than a single install command. There is no explicit installation or setup command in the provided excerpt. It lists code areas like src/hero and chapter directories and points to external resources for setup and tutorials. Users may need to clone the repo and explore notebooks in the chapter and src directories, plus follow linked external tools for environment setup.
Recent releases
RELEASES (latest 0):
- none
Traction
Stars: 2269 Forks: 153 Open issues: 4
Behind the repo
Not provided in the excerpt.
Caveats
License: Apache-2.0 Language: Jupyter Notebook Created: 2025-01-05 Last push: 2026-07-19