// the find
meta-llama/llama-cookbook
Welcome to the Llama Cookbook! This is your go to guide for Building with Llama: Getting started with Inference, Fine-Tuning, RAG. We also show you how to solve end to end problems using Llama model family and using them on various provider services
Meta's official collection of Jupyter notebooks for building with Llama models — inference, fine-tuning, RAG, and end-to-end use cases. It's the reference point if you're starting with Llama and want working examples rather than reading docs cold. Primarily for ML engineers and researchers who want to run something before committing to an architecture.
The 3p-integrations directory is genuinely useful — working examples for AWS Bedrock, GCP Vertex, vLLM, Together AI, and Groq that you'd otherwise have to piece together from scattered blog posts. The fine-tuning section covers FSDP and LoRA with actual training scripts, not toy demos. Llama 4's 5M token context window gets a dedicated notebook with concrete usage patterns. The repo tracks new model releases quickly — Llama 4 Scout and Maverick recipes appeared alongside the model launch.
It's a notebook dump, not a library — nothing is importable or testable, and you can't build on top of it programmatically. The refactor left broken links and missing folders that the FAQ acknowledges but hasn't fully fixed months later. Quality is uneven: some notebooks are polished walkthroughs, others are clearly vendor-contributed marketing content (the Groq section is essentially Groq's own cookbook copy-pasted in). There are no pinned dependency versions in most notebooks, so running a six-month-old notebook is a gamble on whether transformers and torch still agree.