From e7969a8ab0f42e1a7051933ff2069eb02bf14b77 Mon Sep 17 00:00:00 2001 From: Dmitry Rogozhkin Date: Tue, 2 Jun 2026 15:16:20 -0700 Subject: [PATCH 1/2] xpu: add XPU to additional pages XPU is a PyTorch device backend designed to support hardware acceleration on Intel GPUs. Signed-off-by: Dmitry Rogozhkin --- _additional_platforms/xpu.json | 12 +++++ _get_started/additional_platforms/xpu.md | 56 ++++++++++++++++++++++++ 2 files changed, 68 insertions(+) create mode 100644 _additional_platforms/xpu.json create mode 100644 _get_started/additional_platforms/xpu.md diff --git a/_additional_platforms/xpu.json b/_additional_platforms/xpu.json new file mode 100644 index 000000000000..bf61174064a7 --- /dev/null +++ b/_additional_platforms/xpu.json @@ -0,0 +1,12 @@ +{ + "name": "XPU", + "support_channel": "https://github.com/pytorch/pytorch", + "stable": { + "linux": "pip3 install torch torchvision --index-url https://download.pytorch.org/whl/xpu", + "windows": "pip3 install torch torchvision --index-url https://download.pytorch.org/whl/xpu" + }, + "preview": { + "linux": "pip3 install torch torchvision --pre --index-url https://download.pytorch.org/whl/nightly/xpu", + "windows": "pip3 install torch torchvision --pre --index-url https://download.pytorch.org/whl/nightly/xpu" + } +} diff --git a/_get_started/additional_platforms/xpu.md b/_get_started/additional_platforms/xpu.md new file mode 100644 index 000000000000..725b08ae29c2 --- /dev/null +++ b/_get_started/additional_platforms/xpu.md @@ -0,0 +1,56 @@ +# Installing on Intel GPU (XPU) Platform + +XPU is a PyTorch device backend designed to support hardware acceleration on Intel GPUs. Key technical features: + +* Native support for FP32, BF16, FP16, and Automatic Mixed Precision (AMP) +* Extensions of operator set through custom SYCL kernels +* Graph compilation +* Distributed training (through `XCCL`) + +## Prerequisites + +### Hardware Requirements + +* Intel Client GPU: + + * Intel® Arc A-Series Graphics (CodeName: Alchemist) + * Intel® Arc B-Series Graphics (CodeName: Battlemage) + * Intel® Core™ Ultra Processors with Intel® Arc™ Graphics (CodeName: Meteor Lake-H) + * Intel® Core™ Ultra Processors (Series 2) with Intel® Arc™ Graphics (CodeName: Arrow Lake-H) + * Intel® Core™ Ultra Mobile Processors (Series 2) with Intel® Arc™ Graphics (CodeName: Lunar Lake) + * Intel® Core™ Ultra Mobile Processors (Series 3) with Intel® Arc™ Graphics (CodeName: Panther Lake) + +* Intel Data Center GPU: + + * Intel® Data Center GPU Max Series (CodeName: Ponte Vecchio) + +### Software Requirements + +* [Intel GPU Driver](https://www.intel.com/content/www/us/en/developer/articles/tool/pytorch-prerequisites-for-intel-gpu.html) +* Python 3.10 or later + +## Installation + +### pip + +Use the pip package manager to install PyTorch with XPU support. Select your preferred options in the selector above to get the installation command. + +## Verification + +To ensure that PyTorch was installed correctly with XPU support, run the following code: + +```python +import torch +print(torch.__version__) + +# Check XPU availability +if torch.xpu.is_available(): + print("XPU is available!") + print(f"XPU devices: {torch.xpu.device_count()}") +else: + print("XPU is not available.") +``` + +## Documentation + +For more information, please visit the [Getting Started on Intel GPU](https://docs.pytorch.org/docs/stable/notes/get_start_xpu.html). From 58b7a76e7ee61f53a2078465a7eca0a8fb505ddb Mon Sep 17 00:00:00 2001 From: Dmitry Rogozhkin Date: Mon, 15 Jun 2026 15:38:12 -0700 Subject: [PATCH 2/2] Update and simplify xpu.md Signed-off-by: Dmitry Rogozhkin --- _get_started/additional_platforms/xpu.md | 34 +++++++----------------- 1 file changed, 9 insertions(+), 25 deletions(-) diff --git a/_get_started/additional_platforms/xpu.md b/_get_started/additional_platforms/xpu.md index 725b08ae29c2..e558d89ba99c 100644 --- a/_get_started/additional_platforms/xpu.md +++ b/_get_started/additional_platforms/xpu.md @@ -1,33 +1,17 @@ # Installing on Intel GPU (XPU) Platform -XPU is a PyTorch device backend designed to support hardware acceleration on Intel GPUs. Key technical features: +XPU device backend brings native Intel GPU support to PyTorch, enabling performant training and inference on both Linux and Windows: -* Native support for FP32, BF16, FP16, and Automatic Mixed Precision (AMP) -* Extensions of operator set through custom SYCL kernels -* Graph compilation -* Distributed training (through `XCCL`) +* Supports both eager and graph execution +* Built-in support for FP32, BF16, FP16, FP8 and AMP +* Broad operator coverage and model readiness +* Supports PyTorch CPP Extension API through SYCL-based custom kernels +* Enables training and inference workflows +* Scales across devices with distributed training via the `XCCL` backend ## Prerequisites -### Hardware Requirements - -* Intel Client GPU: - - * Intel® Arc A-Series Graphics (CodeName: Alchemist) - * Intel® Arc B-Series Graphics (CodeName: Battlemage) - * Intel® Core™ Ultra Processors with Intel® Arc™ Graphics (CodeName: Meteor Lake-H) - * Intel® Core™ Ultra Processors (Series 2) with Intel® Arc™ Graphics (CodeName: Arrow Lake-H) - * Intel® Core™ Ultra Mobile Processors (Series 2) with Intel® Arc™ Graphics (CodeName: Lunar Lake) - * Intel® Core™ Ultra Mobile Processors (Series 3) with Intel® Arc™ Graphics (CodeName: Panther Lake) - -* Intel Data Center GPU: - - * Intel® Data Center GPU Max Series (CodeName: Ponte Vecchio) - -### Software Requirements - -* [Intel GPU Driver](https://www.intel.com/content/www/us/en/developer/articles/tool/pytorch-prerequisites-for-intel-gpu.html) -* Python 3.10 or later +The system with configured Intel GPU card is required. For detailed list of supported devices and driver install instructions refer to [Getting Started on Intel GPU](https://docs.pytorch.org/docs/stable/notes/get_start_xpu.html). ## Installation @@ -53,4 +37,4 @@ else: ## Documentation -For more information, please visit the [Getting Started on Intel GPU](https://docs.pytorch.org/docs/stable/notes/get_start_xpu.html). +For more information, please visit the [torch.xpu](https://docs.pytorch.org/docs/stable/xpu.html).