- Installing PyTorch on Windows OS should not be a difficult process for programming techies.
- You have to get the Anaconda auxiliary package, then you install PyTorch from the official site.
- After the installation, take a look at our Development Tools Hub for more articles on development & IT-related software.
- If you want to read more tutorials, we also have a dedicated How-To page as well.
PyTorch is a popular machine learning library, easier to learn than other similar deep learning tools, according to specialists.
Before installing PyTorch, you must make sure you meet some prerequisites. Namely, to run Windows 7 or greater, Windows Server 2008 r2 or greater.
It is also recommended to have already installed the Anaconda auxiliary package for PyTorch 3.x (the only version compatible with Windows).
In short, installing PyTorch involves two steps: getting the auxiliary packages (included in the Anaconda Python package manager) and then run the install command that the developer offers on their official page, depending on your settings.
How do I install PyTorch?
- Go to the official Anaconda site and find the right installer.
- Click on it to download and then hit Run.
- You can accept default settings and follow the prompts to finish the installation.
- Use the official PyTorch local installation instructions.
- Choose your configuration. Note that there is no need to install CUDA separately, but you have to select a version of CUDA if your system supports Nvidia GPU.
- Run the recommended command to complete the installation.
The process doesn’t take long, but if it fails, you’ll get a specific error message. Most of the time, it’s a version compatibility issue.
3. Verify the installation
Additionally, you can check if your GPU driver and CUDA are enabled, with the following command:
import torch torch.cuda.is_available()
That’s about it! We hope you found this article useful. If you have anything that you might want to add for this installation process, please use the comments section below.
Frequently Asked Questions
Yes, PyTorch can be installed on Windows since 2018, starting version PyTorch 0.4.0. The library can be installed on Windows OS via Conda or Pip command line.
You have to download the official PyTorch pip package and copy it to the offline computer. Then run pip install * command on the offline computer, in the copied folder.
Yes, PyTorch installs a CUDA toolkit. But you have to check if you have a CUDA-capable device before installing PyTorch, and which version is better.