Where to Find Pip Dependencies for AI for Beginners: Complete Setup Guide

The AI for Beginners repository maintains two separate requirements.txt files at the root level and in the binder/ folder, plus an optional environment.yml for Conda users.

Setting up the Microsoft AI for Beginners curriculum requires installing the correct Python packages. The repository provides version-pinned dependency files that ensure compatibility across different deployment scenarios. This guide maps each file to its specific use case and shows you how to install them.

---

Root requirements.txt: Core Curriculum Dependencies

The primary pip dependencies for AI for Beginners live in requirements.txt at the repository root.

This file captures the core libraries needed for the majority of lessons:

• TensorFlow 2.17
• PyTorch
• Keras 3

Installing the Root Dependencies

Use this file for local development, virtual environments, or when running notebooks outside of Binder:

pip install -r https://raw.githubusercontent.com/microsoft/AI-For-Beginners/main/requirements.txt

---

binder/requirements.txt: Binder-Specific Packages

The repository maintains a separate dependency file at binder/requirements.txt for cloud notebook deployment.

This variant pins different versions to ensure compatibility with the Binder infrastructure:

• TensorFlow 2.13 (older stable version for Binder)
• tensorflow-text
• transformers

Installing the Binder Dependencies

Use this version only when specifically configuring a Binder-compatible environment:

pip install -r https://raw.githubusercontent.com/microsoft/AI-For-Beginners/main/binder/requirements.txt

---

Alternative: Conda Environment Setup

For users preferring Conda, the repository provides environment.yml at the root. This file includes the same packages plus required conda channels in a single specification:

conda env create -f https://raw.githubusercontent.com/microsoft/AI-For-Beginners/main/environment.yml

---

File Reference Summary

File Path	Purpose	Best For
requirements.txt	Core pip dependencies	Local development, standard Python environments
binder/requirements.txt	Binder-tuned pip dependencies	MyBinder.org cloud deployment
environment.yml	Conda environment specification	Conda users needing channel management

---

Summary

• Two requirements.txt files serve different deployment targets in the AI for Beginners repository.
• Root requirements.txt contains the latest stable versions (TensorFlow 2.17, Keras 3) for local use.
• binder/requirements.txt pins older versions (TensorFlow 2.13) plus tensorflow-text and transformers for cloud compatibility.
• environment.yml provides a Conda-native alternative with identical package coverage.
• All files use strict package==version syntax for reproducible installations.

---

Frequently Asked Questions

What is the difference between the two requirements.txt files?

The root requirements.txt targets modern local environments with TensorFlow 2.17 and Keras 3, while binder/requirements.txt pins TensorFlow 2.13 and adds tensorflow-text and transformers specifically for MyBinder infrastructure compatibility.

Can I use Conda instead of pip for AI for Beginners?

Yes. The repository includes environment.yml at the root level, which specifies the same packages through Conda channels. Run conda env create -f environment.yml to build a complete environment without manually managing pip dependencies.

Why does Binder use different dependency versions?

Binder's container infrastructure has specific constraints around CUDA drivers and pre-installed system libraries. The binder/requirements.txt pins TensorFlow 2.13 because that version has verified compatibility with the Binder build pipeline, whereas 2.17 may introduce breaking changes in that environment.

How do I verify my dependencies installed correctly?

After installation, import the key libraries in Python to confirm version alignment:

import tensorflow as tf
import torch
import keras

print(f"TensorFlow: {tf.__version__}")
print(f"PyTorch: {torch.__version__}")
print(f"Keras: {keras.__version__}")

Compare the output against the versions specified in your chosen requirements.txt file.

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