1.1 Install PyTorch

Install PyTorch related libraries and verify whether the installation is successful!

Created Date: 2025-04-26

The main task is to install and verify PyTorch related libraries. For beginners, you can skip the GPU installation and use the CPU version of PyTorch for learning (GPU will need to be installed later). The installation instructions of CPU are as follows:

pip3 install torch torchvision torchaudio

1.1.1 Install IDE

We should have an IDE (Integrated Development Environment) for writing, compiling, and debugging code. You can choose what you like. I recommend using VS Code here because that’s what I use.

1.1.2 Install Python

Go to the Python official website to download Python. It is recommended to use the latest version.

1.1.3 Install CUDA

The NVIDIA^® CUDA^® Toolkit provides a development environment for creating high-performance, GPU-accelerated applications. With it, you can develop, optimize, and deploy your applications on GPU-accelerated embedded systems, desktop workstations, enterprise data centers, cloud-based platforms, and supercomputers. The toolkit includes GPU-accelerated libraries, debugging and optimization tools, a C/C++ compiler, and a runtime library.

After installed the CUDA package, we can confirm whether the installation was successful by typing nvidia-smi or nvcc --version :

1.1.4 Install PyTorch

Go to the PyTorch official website, select preferences and run the command to install PyTorch locally:

pip3 install torch torchvision torchaudio

If CUDA installed, we can use the following command:

pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu128

1.1.5 Verify

Download the tutorial repository, open the verify.py file, import the torch library, and print its version:

import torch

print(torch.__version__)

2.6.0+cu124

🎉 Congratulations, you have completed the PyTorch installation task!

We can use torch to create a Tensor, which will introduction in next section:

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

tensor([[0.9032, 0.6900, 0.7404],
        [0.2300, 0.5118, 0.5050],
        [0.9662, 0.1400, 0.8662],
        [0.8696, 0.9460, 0.6607],
        [0.9800, 0.5627, 0.3977]])

torch.cuda.is_available() to verify whether PyTorch can use CUDA acceleration, and torch.mps.is_available() to detect macOS acceleration settings:

print(torch.cuda.is_available())
print(torch.mps.is_available())