PyTorch is an open-source deep learning framework widely adopted by researchers, Data Scientists, and AI developers for creating and training machine learning models. It is preferred over other frameworks as it offers flexibility, dynamic computation graphs and is easy to use. But before jumping into deep learning projects, the first task is to install PyTorch properly.

For a new user, setting up PyTorch may appear to pose challenges, but a little training for users does wonders in making this setup a breeze. In this guide, we will cover every step of the installation so you can have everything set up correctly and get started with PyTorch without any problems. Let’s get started.

What Is PyTorch?

PyTorch is an open-source deep learning framework that combines the Torch machine learning library with a high-level API written in Python. Initially created by Facebook AI Research (now known as Meta) and later governed by the PyTorch Foundation, it is one of the dominant tools for designing and training neural networks. PyTorch is widely used in research and AI applications as it supports a variety of architectures, including linear regression and deep learning, CNNs, and transformer architectures.

Flexibility and ease of use are one of the main strengths of PyTorch. This includes dynamic computation graphs that enable developers to change and test different parts of their code in real time without needing to wait for the complete execution. As such, PyTorch is incredibly useful for fast prototyping, debugging, and iterative research. Adding to this, its tight integration with Python, combined with a wide range of ready-to-use and pre-trained models, makes it a go-to framework for data scientists and AI researchers alike across the globe.

1. Installing PyTorch on Windows

To install PyTorch on a Windows operating system, follow the steps below.

Prerequisites

Before you install, make sure that you have the following requirements:

• Python: Python should be available on the system.
• Package Managers: You should have pip or conda installed for the installation process.

A. Installing PyTorch Using Pip

Step 1: Check if Python is installed

First, you have to make sure that Python is installed on your system. You can check it by running the following command in the command prompt:

python -V

If Python is installed, you will see an output similar to:

Output:
Python 3.10.2 

Step 2: Verify Pip Installation

Next, check if pip,  the Python package manager, is installed by running:

pip -V

If pip is installed, you will see an output like:

Pip 21.2.4

If pip is not installed, you may need to install or update it before proceeding.

Step 3: Install PyTorch for CPU

To install PyTorch for CPU usage, run the following command:

pip3 install torch torchvision torchaudio

Once the installation is complete, PyTorch will be ready for use.

Step 4: Verify Installation

To confirm that PyTorch was installed successfully, check its version using:

pip3 show torch

Pro Tip: In case you want to uninstall PyTorch, you can use the following command:

pip uninstall torch

B.  Installing PyTorch Using Conda

In case you want to use the conda package manager, follow the instructions below to install PyTorch in a Windows environment.

Prerequisites

Check the next steps before the installation:

• Anaconda/Miniconda Installed: If you do not already have Conda installed, download and install Anaconda or Miniconda.
• Ensure Conda is Installed: Run the following command in Anaconda Command Prompt to make sure Conda is installed:

conda -V

If Conda is installed, you will see an output similar to:

conda 4.14.0

Now that we have ensured that Conda is installed, let’s look at the step-by-step procedure to install PyTorch Using Conda: 

Step 1: Install PyTorch for CPU

To install PyTorch for CPU-only usage, run:

conda install pytorch torchvision torchaudio cpuonly -c pytorch

This command will download and install the required PyTorch libraries.

Step 2: Verify Installation

conda list -f pytorch

If the installation was successful, you will see an output like:

# packages in environment at C:\Users\YourUsername\anaconda3:
# Name                      Version
pytorch                    1.12.1

C.  Installing PyTorch with CUDA support

If you have a compatible NVIDIA GPU and CUDA available, you can install a version of PyTorch with GPU acceleration by just changing the installation command:

For CUDA 11.3:

conda install pytorch torchvision torchaudio cudatoolkit=11.3 -c pytorch

For CUDA 11.6:

conda install pytorch torchvision torchaudio cudatoolkit=11.6 -c pytorch -c conda-forge

These commands install a version of PyTorch that is optimized for GPU computation, giving significant performance improvements for deep learning tasks.

After installation, you can check the installation using:

conda list -f pytorch

You have successfully installed PyTorch on your Windows system using Conda.

2.  Installing PyTorch on macOS

Installing PyTorch on macOS requires very little configuration. PyTorch does not support CUDA natively under macOS, which means unless you install from source, it’ll fall back to using the CPU.

A. Installing PyTorch Using PIP

Prerequisites

Before installation, ensure:

• Python is installed: You can check the installed version using:

python3 --version

Example output:
Python 3.9.7

pip (Python package manager) is installed: Check its version using:

pip3 --version

Example output:
pip 22.2.2

Now that we have ensured that PIP is installed, let’s look at the step-by-step procedure to install PyTorch using PIP: 

Step 1: Install PyTorch

To install PyTorch for CPU, run the following command:

pip3 install torch torchvision torchaudio

This will install PyTorch and its related libraries.

Step 2: Verify Installation

After installation, verify it using:

pip3 show torch

Example output:
Name: torch  
Version: 1.12.1  

If you see the installed version, PyTorch is successfully installed.

Pro Tip: In case you want to uninstall PyTorch, simply run:

pip uninstall torch

Once completed, PyTorch will be removed from your system.

B. Installing PyTorch on macOS Using Conda

Another method for installing PyTorch on macOS is through the use of the Conda package manager. This approach is another option for those who are working inside of virtual environments that are maintained by Anaconda.

Step 1: Activate Anaconda Prompt

Before installing PyTorch, activate the Anaconda prompt (if it’s not already active) using:

conda activate

Step 2: Check Conda Version

Ensure that Conda is installed by checking its version:

conda-- version

If Conda is not installed, you will need to install Anaconda or Miniconda before proceeding.

Step 3: Install PyTorch

Now, install PyTorch using the following command:

conda install pytorch torchvision torchaudio -c pytorch

This will install PyTorch along with its necessary dependencies.

Step 4: Verify Installation

Once the installation is complete, confirm the installed version of PyTorch using:

conda list -f pytorch

Example output:

# packages in environment

pytorch   1.12.1   py3.9_0   pytorch

If you see PyTorch listed, the installation was successful.

3. Installing PyTorch for Linux

A. Install Pytorch on Linux Using PIP

We can install PyTorch on Linux using pip, which is the default package manager for Python. Make sure to meet the prerequisites before continuing.

Prerequisites

• Glibc Version: Ensure your system has Glibc v2.17 or higher.
  
• Python Installed: PyTorch requires Python 3.6 or later.
  
• Pip Installed: The package manager pip must be installed

Now that PIP is installed, let’s explore the step-by-step procedure to install PyTorch using PIP.

Step 1: Check Python Version

First, verify that Python is installed by running:

python -V

Example output:
Python 3.8.10

Step 2: Check Pip Version

Ensure that pip is installed by checking its version:

pip -V

Example output:

pip 21.2.4 from /usr/lib/python3/dist-packages/pip (python 3.8)

Step 3: Install PyTorch (CPU Version)

To install PyTorch for CPU-only support, use the following command:

pip3 install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cpu

Step 4: Verify Installation

Once the installation is complete, confirm the installed version of PyTorch:

pip3 show torch

Example output:

Name: torch
Version: 1.12.1

Summary: Tensors and dynamic neural networks in Python with strong GPU acceleration

If you see the correct version listed, PyTorch has been successfully installed.

B. Install Pytorch using CUDA Support

For GPU acceleration using CUDA, install the corresponding CUDA version.

For CUDA 11.3, use:

pip3 install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu113

For CUDA 11.6, use:

pip3 install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu116

Pro Tip: If you need to remove PyTorch, simply run:

pip uninstall torch

This will uninstall the currently installed version of PyTorch from your system.

Getting Started with PyTorch on Google Colab

What is Google Colab?

Google Colaboratory (Google Colab), a tool designed for machine learning and deep learning research, offers a free online Jupyter Notebook environment. It provides an interactive coding environment that enables users to write and run Python code in the cloud without any local setup requirements.

Key Features:

• No installation is required – it works directly in your browser.
• Free GPU/TPU access – Ideal for ML & AI tasks.
•  Cloud storage integration – Supports Google Drive.
•  Collaborative – Share notebooks easily with others.

Getting Started with Google Colab

Follow these steps to start using Google Colab:

1. Open Google Colab:
   
   • Visit Google Colaboratory
     
2. Sign in:
   
   • Log in using your Google Account.
     
3. Create a new Jupyter Notebook:
   
   • Click File → New notebook (Python 3).
     

Now, you are all set to use Google Colab! 

A. Setting Up GPU in Google Colab

By default, Colab runs on a CPU. To use a GPU, follow these steps:

1. Go to “Runtime” in the menu.
   
2. Click on “Change runtime type”.
   
3. Under “Hardware Accelerator”, select “GPU”.
   
4. Click “Save”.
   

 Note:

• Google Colab provides 12 hours of continuous runtime.
  
• If inactive for more than 1 hour, your session disconnects.
  
• After 12 hours, the system will reset all RAM, disk storage, and data.

B. Installing PyTorch On Google Collab

Let’s look at the step-by-step process to install PyTorch in Google Colab

Colab comes with PyTorch pre-installed, but if needed, install/update it using:

!pip install torch torchvision torchaudio

To check if PyTorch is installed correctly, run:

import torch
print(torch.__version__)

If you are using GPU, verify it with:

print(torch.cuda.is_available())  # Should return True if GPU is active

That’s it! You’re now ready to use PyTorch on Google Colab! 

Conclusion

PyTorch offers cross-platform support, with installation instructions available for Windows, macOS, and Linux. Available for installation through both pip and conda, meaning you can integrate it wherever you need it in your flow. Their PyTorch deep learning library is a popular framework that allows developers to easily create complex neural networks and train them on either local or cloud-based infrastructure.

With this guide to installation steps, you are ready to go with PyTorch with a properly configured CPU or CUDA with your configuration for Machine Learning. With PyTorch installed, you can now begin exploring the powerful features and capabilities of this deep learning framework for training models and AI research.