xinyun/ComfyUI
1
0
Fork 0
mirror of https://git.datalinker.icu/comfyanonymous/ComfyUI synced 2026-11-12 22:57:03 +08:00
Code Issues Packages Projects Releases Wiki Activity

feat(README.md): add zluda instructions

Browse Source
This commit is contained in:
RaINi_ 2024-02-18 19:44:30 +01:00 committed by GitHub
parent dccca1daa5
commit c37513b113
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 24 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

24
README.md
View File

@ -95,6 +95,30 @@ Put your SD checkpoints (the huge ckpt/safetensors files) in: models/checkpoints
Put your VAE in: models/vae Put your VAE in: models/vae
### AMD GPUs (ZLUDA)
Thanks to the efforts of @vosen and @lshqqytiger AMD users are now able to use applications utilizing cuda.
Check if your gpu is compatible with the hip sdk at <https://rocm.docs.amd.com/projects/install-on-windows/en/develop/reference/system-requirements.html>
Download and install the hip sdk from <https://www.amd.com/en/developer/resources/rocm-hub/hip-sdk.html>.
Download the latest release of zluda from <https://github.com/lshqqytiger/ZLUDA/releases>, extract all files.
Add `%HIP_PATH%bin` and the zluda folder with the files you extracted into to your windows PATH enviroment variable
Replace the installed torch and torchvision package with a zluda compatible one by running:
```
pip uninstall torch torchvision torch-directml -y
pip install torch==2.2.0 torchvision --index-url https://download.pytorch.org/whl/cu118
```
Now in your zluda folder make a copy of and rename `cublas.dll` to `cublas64_11.dll` and `cusparse.dll` to `cusparse64_11.dll`.
Lastly copy `cublas64_11.dll` and `cusparse64_11.dll` into your `site-packages\torch\lib` folder replacing the already existing ones.
You should now be able to run comfy normally and enjoy stable diffusion with your AMD gpu.
Keep in mind that zluda is still very experimental and some things may not work properly at the moment.
### AMD GPUs (Linux only) ### AMD GPUs (Linux only)
AMD users can install rocm and pytorch with pip if you don't have it already installed, this is the command to install the stable version: AMD users can install rocm and pytorch with pip if you don't have it already installed, this is the command to install the stable version: