Discover a simpler way to install PyTorch tailored to your hardware: Wheel variants for Release 2.8 are now available for testing on the PyTorch test channel

release/packaging
1

In collaboration with Astral, NVIDIA, and Quansight PyTorch team have produced an experimental set of Wheels for Release 2.8.

What are Wheel Variants ?

  • Wheel variants are a mechanism for publishing platform-dependent Python wheels and selecting the most suitable package variant for a given platform.

  • This approach helps to remove the need for local identifier experience in PyTorch packaging and enhance user experience installing PyTorch

  • Please see this RFC for more information.

What is currently supported ?

The following variants will be supported for PyTorch 2.8.0:

  • CUDA 12.6: Linux X86 and Windows X86

  • CUDA 12.8: Linux X86 and Windows X86

  • CUDA 12.9: Linux X86, Linux aarch64 CPU and GPU, Windows X86

  • CPU only: Linux, Linux aarch64, MacOS M1 and Windows CPU Only

How to test it ?

​​You can start using it today by installing Final RC for PyTorch 2.8.0 with the variant-enabled uv, which will automatically choose the best PyTorch build for your machine’s GPU.

Using Linux/MacOS:

curl -LsSf https://astral.sh/uv/install.sh | INSTALLER_DOWNLOAD_URL=https://wheelnext.astral.sh sh
uv pip install torch torchvision

Using Windows:

powershell -ExecutionPolicy Bypass -c “$env:INSTALLER_DOWNLOAD_URL=‘https://wheelnext.astral.sh’; irm https://astral.sh/uv/install.ps1 | iex”
uv pip install torch torchvision

If there are any questions, please feel free to reach out.

Cheers,

Team PyTorch

2 Likes