labs/docs/00_getting_started/README.md

Lab 00: Getting Started (The Simplest Starting Point)

Overview

Welcome! Before spinning up massive cloud GPUs or diving into complex AI loops, this is the absolute best place to start. This lab provides a completely local, interactive introduction to the foundational concepts of AI training and the Modular ecosystem.

Setup

No cloud GPU is required. This lab runs entirely on your local CPU.

If you don't have Jupyter installed, you can launch it using uv:

uv tool run --from jupyterlab jupyter-lab

Usage

Open 00_getting_started.ipynb in the Jupyter interface. The notebook is an interactive journey that walks you through:

1. The Data: Inspecting the raw TinyStories dataset.
2. The Tokenizer: Seeing how English sentences are translated into numerical tokens using our custom BPE tokenizer.
3. The Mojo Compiler: Compiling and executing our first Mojo code locally using the pixi package manager.

Goals

• Understand the raw ingredients of an AI model (Data + Tokens).
• Verify your local Python environment can read parquet data shards.
• Verify your local pixi environment can successfully compile Mojo (.mojo) files.
• Establish the baseline knowledge needed to start writing Neural Network architectures in Lab 03.