How To Install PyTorch on Rocky Linux 9

PyTorch has emerged as a leading open-source machine-learning framework, widely adopted by researchers and developers for its flexibility and robustness. It offers tools and libraries for various tasks, including natural language processing and computer vision. Rocky Linux 9, known for its stability and enterprise-grade reliability, provides an excellent platform for deploying and experimenting with PyTorch. This guide provides a detailed walkthrough on how to install PyTorch on Rocky Linux 9. Whether you are setting up a deep-learning workstation or configuring a cloud server, this article covers the essential steps to get you started. We will explore two primary installation methods: using Anaconda and using pip, each with its own advantages.

Prerequisites

Before diving into the installation process, ensure your system meets the necessary prerequisites. This preparation will help streamline the installation and avoid common issues. Let’s get started.

System Requirements

• Hardware Requirements: A modern CPU with multiple cores and sufficient RAM (at least 8GB, but 16GB or more is recommended for more intensive tasks). Adequate storage space for the installation and datasets.
• Operating System: Rocky Linux 9 should already be installed and running smoothly on your system.
• Internet Connection: A stable internet connection to download the necessary packages and dependencies.

Software Requirements

• Python: PyTorch requires Python 3.8 or later. Older versions might cause compatibility issues.
• pip: The pip package manager is essential for installing Python packages. It usually comes pre-installed with Python, but it’s good to verify its presence.
• Anaconda (Optional): Anaconda is a distribution of Python that simplifies package management and environment isolation. While optional, it is recommended for beginners due to its ease of use.

Checking Existing Installation

First, you should check if Python is already installed on your Rocky Linux 9 system. Open a terminal and run the following command:

python3 --version

or

python --version

If Python is installed, the command will display the version number. If not, you’ll need to install Python. Next, verify that pip is installed. Use this command:

pip3 --version

or

pip --version

If pip is installed, the command will show its version. If not, you’ll need to install pip separately.

Choosing an Installation Method

When it comes to installing PyTorch, you have two main options: Anaconda and pip. Each method has its pros and cons, making them suitable for different users and scenarios. Let’s explore these methods to help you decide which one fits your needs best.

• Anaconda: Anaconda is a comprehensive distribution of Python that includes a package manager, an environment manager, and a collection of pre-built packages.
• pip: pip is a package installer for Python. It allows you to install and manage packages from the Python Package Index (PyPI).

Anaconda is generally recommended for beginners because it simplifies dependency management and provides a more controlled environment. However, it comes with a larger installation size due to the inclusion of numerous packages. Pip, on the other hand, is lightweight and installs only the packages you need. This makes it ideal for advanced users who prefer manual dependency resolution and a minimal installation footprint.

Installing PyTorch with Anaconda

Anaconda simplifies the installation process with its package and environment management capabilities. Here’s how to install PyTorch using Anaconda.

Installing Anaconda

1. Download the Anaconda Installer: Visit the official Anaconda website and download the Linux installer. Choose the version that corresponds to Python 3.x.
2. Verify the Integrity of the Installer: Before running the installer, it’s a good practice to verify its integrity using the checksum provided on the Anaconda website. This ensures that the downloaded file is complete and has not been tampered with.
3. Run the Installer: Open a terminal, navigate to the directory where you downloaded the installer, and run it using the following command:

bash Anaconda3-xxxxxxxx-Linux-x86_64.sh

Replace Anaconda3-xxxxxxxx-Linux-x86_64.sh with the actual name of the installer file.

4. Follow the On-Screen Instructions: The installer will guide you through the installation process. Accept the license agreement and choose the installation directory.
5. Add Anaconda to the System PATH: The installer will ask if you want to add Anaconda to your system PATH. It is recommended to choose “yes” to make Anaconda commands available from any terminal.
6. Initialize Anaconda: After the installation is complete, initialize Anaconda with the following command:

conda init

This command configures your shell to use Anaconda. Close and reopen your terminal for the changes to take effect.

Creating a Conda Environment

It’s best to create a separate environment for your PyTorch installation to avoid conflicts with other Python packages. Here’s how:

1. Create a New Environment: Use the following command to create a new Conda environment:

conda create -n pytorch_env python=3.x

Replace 3.x with your desired Python version (e.g., 3.8, 3.9, 3.10). This command creates an environment named pytorch_env with the specified Python version.

2. Activate the Environment: Activate the newly created environment using the following command:

conda activate pytorch_env

Once activated, your terminal prompt will change to indicate that you are working within the pytorch_env environment.

Installing PyTorch

With the Conda environment activated, you can now install PyTorch. Use the conda install command to install PyTorch, torchvision, and torchaudio. Torchvision and torchaudio are essential packages for computer vision and audio processing, respectively.

If you have a CUDA-enabled GPU, specify the appropriate CUDA toolkit version to leverage GPU acceleration. If you do not have a GPU or want to use the CPU version, omit the CUDA toolkit specification.

conda install pytorch torchvision torchaudio cudatoolkit=[version] -c pytorch

Replace [version] with the CUDA toolkit version compatible with your NVIDIA drivers (e.g., 11.3, 11.6, 11.8). If you want to install the CPU-only version, use the following command:

conda install pytorch torchvision torchaudio -c pytorch

Verifying the Installation

After the installation is complete, verify that PyTorch has been installed correctly. Follow these steps:

1. Open a Python Interpreter: Open a Python interpreter within the Conda environment by typing python in the terminal.
2. Import the torch Library: Import the torch library using the following command:

import torch

3. Check the PyTorch Version: Check the PyTorch version using the following command:

torch.__version__

This will display the version of PyTorch that you have installed. Finally, run a simple test to ensure that PyTorch is functioning correctly. Create a random tensor and print it:

x = torch.rand(5, 3)
 print(x)

If the installation is successful, you will see a tensor of random numbers printed in the terminal.

Installing PyTorch with pip

If you prefer using pip, the Python package manager, follow these steps to install PyTorch on Rocky Linux 9. This method is suitable for users who want a lightweight installation and are comfortable managing dependencies manually.

Installing pip

First, ensure that you have the latest version of pip installed. Update pip to the latest version using the following command:

python -m pip install --upgrade pip

This command updates pip to the newest version available on PyPI.

Installing PyTorch

Use pip to install PyTorch, torchvision, and torchaudio. For CPU-only support, use the following command:

pip install torch torchvision torchaudio

If you have a CUDA-enabled GPU and want to leverage it, install the CUDA-enabled version of PyTorch. Specify the appropriate CUDA version using the --index-url option:

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu[version]

Replace [version] with the CUDA version that matches your NVIDIA drivers (e.g., cu113, cu116, cu117, cu118).

Resolving Dependencies

When using pip, you may need to manually install any missing dependencies. If you encounter errors during the installation, read the error message to identify the missing packages. Install them using pip:

pip install [package_name]

Replace [package_name] with the name of the missing package.

Verifying the Installation

After installing PyTorch with pip, verify that it has been installed correctly. Open a Python interpreter by typing python in the terminal. Then, import the torch library:

import torch

Check the PyTorch version:

torch.__version__

Run a simple test:

x = torch.rand(5, 3)
 print(x)

If the installation is successful, you will see the PyTorch version and a tensor of random numbers printed in the terminal.

GPU Support (Optional)

To leverage the power of GPUs for accelerated computing, follow these steps to set up GPU support for PyTorch on Rocky Linux 9.

Checking for CUDA-Enabled GPU

First, verify that you have an NVIDIA GPU available on your system. Use the nvidia-smi command to check the GPU status and information. If the command is not found, you may need to install the NVIDIA drivers.

Next, check the CUDA and cuDNN versions installed on your system. PyTorch requires specific versions of CUDA and cuDNN to function correctly. You can check the installed versions using the following commands:

nvcc --version
cat /usr/local/cuda/version.txt

Installing NVIDIA Drivers

1. Add the NVIDIA Repository: Add the NVIDIA repository to your system to get the latest drivers. Create a new file /etc/yum.repos.d/nvidia.repo with the following content:

[nvidia]
 name=NVIDIA - x86_64
 baseurl=https://developer.download.nvidia.com/compute/cuda/repos/rocky9/x86_64
 enabled=1
 gpgcheck=1
 gpgkey=https://developer.download.nvidia.com/compute/cuda/repos/rocky9/x86_64/gpgkey

2. Install the NVIDIA Drivers: Install the latest NVIDIA drivers using the following command:

dnf install nvidia-driver

3. Verify the Driver Installation: Verify that the drivers have been installed correctly using the nvidia-smi command. This command should now display information about your NVIDIA GPU.

Installing CUDA Toolkit

Download the CUDA Toolkit from NVIDIA’s website, ensuring you select the version compatible with your GPU and drivers. Follow the installation instructions provided by NVIDIA. After installing CUDA, set the CUDA environment variables:

export PATH=/usr/local/cuda/bin:$PATH
 export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH

Add these lines to your ~/.bashrc or ~/.bash_profile to make them permanent.

Configuring PyTorch for GPU

When installing PyTorch, make sure to install the CUDA-enabled version. Use either Anaconda or pip, as described earlier, and specify the appropriate CUDA toolkit version. Verify that PyTorch is using the GPU by running the following commands in a Python interpreter:

import torch
 print(torch.cuda.is_available())
 print(torch.cuda.get_device_name(0))

If torch.cuda.is_available() returns True and torch.cuda.get_device_name(0) displays the name of your NVIDIA GPU, then PyTorch is properly configured to use the GPU.

Troubleshooting Common Issues

During the installation process, you may encounter some common issues. Here are some troubleshooting tips to help you resolve them.

• Dependency Conflicts: Dependency conflicts can arise when different packages require conflicting versions of the same library. To avoid this, use virtual environments to isolate dependencies. Anaconda and venv can help to maintain separate environments and prevent version conflicts.
• Import Errors: If you encounter import errors when trying to import the torch library, verify that PyTorch is installed in the correct environment and that the environment is activated. Check the PYTHONPATH environment variable to ensure it includes the path to the PyTorch installation.
• CUDA Errors: CUDA errors often indicate issues with the NVIDIA drivers or the CUDA toolkit installation. Ensure that you have installed the correct CUDA version compatible with your GPU and drivers. Verify that the NVIDIA drivers are properly installed and that the CUDA environment variables are set correctly.
• Reinstalling PyTorch: If you continue to experience issues, try completely removing PyTorch and its dependencies before reinstalling. Use conda remove or pip uninstall to remove the packages, and then reinstall them following the steps outlined in this guide.

Performance Optimization

To maximize the performance of PyTorch on Rocky Linux 9, consider the following optimization techniques.

• Use multiple CPU cores for training by adjusting the number of worker threads used for data loading.
• Leverage GPU acceleration for faster computations by moving tensors and models to the GPU.
• Optimize data loading and preprocessing to reduce bottlenecks during training.
• Monitor resource usage during training to identify and address performance issues.

Congratulations! You have successfully installed PyTorch. Thanks for using this tutorial for installing the PyTorch on Rocky Linux 9 system. For additional help or useful information, we recommend you check the official PyTorch website.