Master LLM Fine-Tuning & Inference — From White Belt to Black Belt
Free, open-source curriculum for mastering large language models. 83 hands-on Google Colab notebooks across 7 stages — from transformer basics to custom CUDA kernels and production inference.
| Stage | Belt | Topics | Notebooks |
|---|---|---|---|
| 0 | 🤍 White | Foundations, Transformers, Tokenization, Data | 7 |
| 1 | 💛 Yellow | Full Fine-Tuning, Custom Loss, AMP, Evaluation | 10 |
| 2 | 💚 Green | LoRA, QLoRA, Adapters, Prompt Tuning, DoRA | 10 |
| 3 | 💙 Blue | FlashAttention, DeepSpeed, FSDP, Instruction Tuning | 10 |
| 4 | 🤎 Brown | RLHF, DPO, Constitutional AI, MoE, Safety | 10 |
| 5 | 🖤 Black | CUDA Kernels, Triton, Quantization, vLLM, TRT-LLM | 10 |
| Inf | ❤️ Red | Inference Optimization, KV Cache, GPTQ, AWQ, GGUF | 23 |
npm install && npm run dev- Replace
YOUR_GITHUB_USERNAMEeverywhere with your GitHub username - Update
SITE_CONFIG.urlinsrc/data/curriculum.ts - Replace
ca-pub-XXXXXXXXXXXXXXXXwith your AdSense publisher ID insrc/app/layout.tsxandsrc/components/AdBanner.tsx - Replace ad slot placeholder IDs with real ad unit IDs
npx vercel --prod- Per-page metadata, OG tags, Twitter cards
- JSON-LD: WebSite, Course, LearningResource, ItemList
- Auto-generated sitemap.xml (92 URLs)
- robots.txt
- All 92 pages statically pre-rendered
MIT