Installing and Managing Packages (pip / conda)

Python's real power comes from its vast ecosystem of third-party libraries, tools like NumPy, Pandas, Scikit-learn, and TensorFlow that make AI development possible.

To use these libraries, you need a package manager, a tool that downloads, installs, updates, and removes packages for you.

Python has two primary package managers: pip and conda. Knowing how to use them confidently is an essential practical skill for every Python and AI developer.

What is a Package?

A package is a collection of pre-written Python code that extends what Python can do. Instead of writing complex mathematical or data processing functions from scratch, you install a package and use it directly.

1. pip — Python's default package manager, works with the Python Package Index (PyPI).

2. conda — Anaconda's package manager, handles both Python packages and system-level dependencies.

pip — Python's Default Package Manager

pip comes pre-installed with Python. It connects to PyPI (pypi.org) — a repository of over 400,000 packages, and installs them with a single command.

Core pip Commands

bash

# Install a package

pip install numpy

# Install a specific version

pip install numpy==1.24.0

# Install multiple packages at once

pip install numpy pandas matplotlib

# Upgrade a package to the latest version

pip install --upgrade numpy

# Uninstall a package

pip uninstall numpy

# List all installed packages

pip list

# Show details about a specific package

pip show numpy

# Check for outdated packages

pip list --outdated

Installing from a Requirements File

In professional projects, all required packages are listed in a requirements.txt file. This allows anyone to recreate the same environment instantly.

requirements.txt:

Install everything at once:

Generate a requirements file from your current environment:

conda — Anaconda's Package Manager

conda is the package manager that comes with Anaconda and Miniconda. It is more powerful than pip for AI/ML work because it also manages non-Python dependencies and handles complex library compatibility automatically.

Core conda Commands

bash

# Install a package
conda install numpy

# Install from a specific channel
conda install -c conda-forge scikit-learn

# Install a specific version
conda install numpy=1.24.0

# Update a package
conda update numpy

# Update all packages
conda update --all

# Uninstall a package
conda remove numpy

# List installed packages
conda list

# Search for a package
conda search pandas

Managing Environments

An environment is an isolated workspace with its own Python version and packages. This prevents conflicts between projects that need different versions of the same library.

Why Environments Matter: Imagine Project A needs TensorFlow 2.10 and Project B needs TensorFlow 2.15. Installing both globally causes conflicts. With environments, each project gets its own isolated setup.

conda Environment Commands

bash

# Create a new environment
conda create --name ai_project python=3.10

# Activate the environment
conda activate ai_project

# Deactivate the environment
conda deactivate

# List all environments
conda env list

# Remove an environment
conda env remove --name ai_project

# Export environment to a file
conda env export > environment.yml

# Recreate environment from file
conda env create -f environment.yml

pip Virtual Environments (venv)

If you are using pip without Anaconda, use Python's built-in venv module:

pip vs conda — Key Differences

Installing Common AI Packages

Troubleshooting Common Issues

Best Practices

Following these habits keeps your Python environment clean and your projects reproducible:

1. Always use a virtual environment — never install packages directly into the system Python.

2. Pin package versions in requirements.txt or environment.yml for reproducibility.

3. Use conda for AI/ML projects — it handles complex dependencies better than pip alone.

4. Keep environments project-specific — one environment per project avoids conflicts.

5. Regularly update packages — especially security-sensitive ones.