How to Run a Specific Lesson's Code from the Command Line in AI Engineering From Scratch

Navigate to the lesson directory under phases/<phase-slug>/<lesson-slug>/ and execute the run command documented in code/README.md, such as python3 code/main.py for Python implementations or cargo run for Rust.

The AI Engineering From Scratch repository by rohitg00 organizes hands-on curriculum content into structured phase directories. Each lesson includes a self-contained implementation in the code/ subdirectory that you can execute directly from the terminal to validate concepts and experiment with modifications.

Locate the Lesson Directory

Every lesson resides at phases/<phase-slug>/<lesson-slug>/ within the repository root. For example, the "Gradient Checkpointing" lesson lives under phases/10-llms-from-scratch/34-gradient-checkpointing/.

Navigate to the lesson root before executing any commands:

cd phases/10-llms-from-scratch/34-gradient-checkpointing

Inside this directory, the code/ folder contains the implementation files. The code/README.md file documents the exact command required to run that specific lesson.

Install Required Dependencies

Most lessons are self-contained, but some require third-party packages. The code/README.md specifies the exact installation command for the lesson's programming language.

Python lessons typically use:

pip install -r requirements.txt

TypeScript lessons require:

npm install

Rust lessons handle dependencies automatically through Cargo.toml when you run cargo build, but ensure you have the Rust toolchain installed.

Execute the Lesson Code

The code/README.md in each lesson specifies the precise command to launch the implementation. The convention varies by language:

• Python: python3 code/main.py
• TypeScript: npm start or npx ts-node code/main.ts
• Rust: cargo run --manifest-path code/Cargo.toml
• Julia: julia code/main.jl

For example, to run the Gradient Checkpointing lesson according to the source code:

cd phases/10-llms-from-scratch/34-gradient-checkpointing
python3 code/main.py

This prints the equivalence check and cost table, then exits with status 0 on success.

For the Terminal-Native Coding Agent TypeScript lesson:

cd phases/19-capstone-projects/01-terminal-native-coding-agent
npm install
npm start

To launch the interactive REPL mode instead:

npm start -- --repl

For Rust implementations like the Pre-training Mini-GPT lesson:

cd phases/10-llms-from-scratch/04-pre-training-mini-gpt
cargo run --manifest-path code/Cargo.toml

Run the Test Suite

Each lesson includes a code/tests/ directory containing unit tests. Execute the language-specific test runner to verify the implementation's correctness.

Python:

python3 -m unittest discover code/tests -v

TypeScript:

npm test

Rust:

cargo test --manifest-path code/Cargo.toml

Julia:

julia --project=code -e 'using Pkg; Pkg.test()'

All lessons exit with status 0 when tests pass. A non-zero exit status indicates a failure in the lesson's self-test suite.

Summary

• Each lesson in AI Engineering From Scratch follows the path phases/<phase>/<lesson>/ with implementation files in the code/ subdirectory.
• The code/README.md file contains the exact terminal command required to run that lesson's code.
• Python lessons execute via python3 code/main.py, TypeScript via npm start, Rust via cargo run --manifest-path code/Cargo.toml, and Julia via julia code/main.jl.
• Install dependencies using pip install -r requirements.txt or npm install as documented in the lesson's README before running.
• Run the test suite using language-specific commands like python3 -m unittest discover code/tests or cargo test to verify correctness.

Frequently Asked Questions

How do I find the correct command to run a specific lesson?

Check the code/README.md file inside the lesson directory. This file explicitly documents the run command, such as python3 code/main.py for Python lessons or npm start for Node.js projects, along with any required dependencies and flags.

What should I do if a lesson requires external packages?

Install the dependencies listed in the lesson's code/README.md. Python lessons typically provide a requirements.txt file for pip install, while TypeScript lessons require running npm install to fetch packages like hono or zod before executing npm start.

How do I verify that a lesson executed correctly?

All lessons exit with status 0 on successful completion. The output will display the lesson's results, such as benchmark tables or equivalence checks. For additional verification, run the unit tests in code/tests/ using the appropriate language test runner.

Can I run lessons interactively or modify the code?

Yes. After navigating to the lesson directory, you can edit code/main.<ext> directly. For interactive sessions, some lessons support specific flags; for example, the Terminal-Native Coding Agent supports npm start -- --repl to launch an interactive REPL instead of the scripted demo.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}