Using Intel Arc A750 for PyTorch

Ask Fedora
,
1

I want to use my Intel Arc A750 with PyTorch, so I was following these instructions for setup:

  1. PyTorch instructions for Intel GPUs. It points to the following driver installation page.
  2. Driver installation, which points to the following
  3. Drivers for Ubuntu

I’m hoping someone can help me figure out the equivalent Fedora packages. If a package is unavailable, I can also consider building them myself (I’m familiar with building packages on Copr).

So far I managed to setup the oneAPI YUM repo mentioned in (2). And I have intel-openapi-runtime\* and ntel-deep-learning-essentials-2025.0. But I’m not sure what are the equivalent Fedora packages from (3).

At the moment, after installing PyTorch I get this:

>>> import torch
>>> torch.xpu.is_available()
/home/user/.virtualenvs/xpu/lib64/python3.13/site-packages/torch/xpu/__init__.py:60: UserWarning: XPU device count is zero! (Triggered internally at /pytorch/c10/xpu/XPUFunctions.cpp:60.)
  return torch._C._xpu_getDeviceCount()
False