vrr/kvcache-ai-ktransformers
2
0
Fork 0
mirror of https://github.com/kvcache-ai/ktransformers.git synced 2025-09-06 04:30:03 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 2

Merge pull request #1306 from aubreyli/xpu-doc

Browse source 
xpu.md: add device discovery tips
This commit is contained in:
aubreyli 2025-05-15 14:18:47 +08:00 • committed by GitHub
commit af9472b518
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
1 changed files with 18 additions and 1 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

17
doc/en/xpu.md
View file

@ -11,6 +11,15 @@ We are excited to introduce **Intel GPU support** in KTransformers (Beta release
Begin by installing the GPU drivers for your Intel GPU: Begin by installing the GPU drivers for your Intel GPU:
- [Official GPU Installation Guide for Intel GPUs](https://dgpu-docs.intel.com/driver/overview.html) - [Official GPU Installation Guide for Intel GPUs](https://dgpu-docs.intel.com/driver/overview.html)
To verify that the kernel and compute drivers are installed and functional:
```bash
clinfo --list | grep Device
`-- Device #0: 13th Gen Intel(R) Core(TM) i9-13900K
`-- Device #0: Intel(R) Arc(TM) A770 Graphics
`-- Device #0: Intel(R) UHD Graphics 770
```
> [!Important] > [!Important]
> Ensure that **Resizable BAR** is enabled in your system's BIOS before proceeding. This is essential for optimal GPU performance and to avoid potential issues such as `Bus error (core dumped)`. For detailed steps, please refer to the official guidance [here](https://www.intel.com/content/www/us/en/support/articles/000090831/graphics.html). > Ensure that **Resizable BAR** is enabled in your system's BIOS before proceeding. This is essential for optimal GPU performance and to avoid potential issues such as `Bus error (core dumped)`. For detailed steps, please refer to the official guidance [here](https://www.intel.com/content/www/us/en/support/articles/000090831/graphics.html).
@ -114,3 +123,11 @@ pip uninstall triton pytorch-triton-xpu
# Reinstall correct version of pytorch-triton-xpu # Reinstall correct version of pytorch-triton-xpu
pip install pytorch-triton-xpu==3.3.0 --index-url https://download.pytorch.org/whl/xpu pip install pytorch-triton-xpu==3.3.0 --index-url https://download.pytorch.org/whl/xpu
``` ```
4. `ValueError: Unsupported backend: CUDA_HOME ROCM_HOME MUSA_HOME are not set and XPU is not available.`
Ensure you have permissions to access /dev/dri/renderD*. This typically requires your user to be in the render group:
```bash
sudo gpasswd -a ${USER} render
newgrp render
```