Running on CPU instead of GPU in an AMD system. How to fix?

[morpheuskibbe]

Hello. I have an AMD system. i got it to run by using these commands

set COMMANDLINE_ARGS=--skip-torch-cuda-test --no-half

Initially because it refused to start with no cuda. hence the skip cuda option and then because it had a runtime error mentioning "LayerNormKernelImpl" not implemented for 'Half'

It's running now but according to task manager it's running on CPU only... the GPU is not being hit in any way and predictably the performance is terrible.

what commands should I be using to get expected behavior?

[morpheuskibbe]

OH MY GOD the instructions were WRONG

The automatic installation section says
git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui.git

It should say
git clone https://github.com/lshqqytiger/stable-diffusion-webui-directml.git

[robonxt]

Actually, if you go to the AMD installation guide, it says to use the ishqqytiger fork: https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Install-and-Run-on-AMD-GPUs

[morpheuskibbe]

yes. hence my comment about incorrect instructions. if you goto the ishqqtiger fork: https://github.com/lshqqytiger/stable-diffusion-webui-directml and scroll down to the instructions it has a copyable git clone command that points at the original, not the fork. Hence my original issue.

[Enferlain]

I cloned the directml one and it still runs on cpu 4shrug

[Aev69]

Same here. Both the original and fork just run on CPU, really defeats the purpose of using the directml one

[Ming2k8-Coder]

What your parameter in webui-user.bat Vào 09:09, T.2, 8 Th1, 2024 Aev69 ***@***.***> đã viết:

…

Same here. Both the original and fork just run on CPU, really defeats the purpose of using the directml one — Reply to this email directly, view it on GitHub <#124 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/A7VS7K5LNNGMUCNG5RBKZXTYNNINDAVCNFSM6AAAAAAYGUMC2GVHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM4DANBTGM2DE> . You are receiving this because you are subscribed to this thread.Message ID: <lshqqytiger/stable-diffusion-webui-directml/repo-discussions/124/comments/8043342 @github.com>

[diekf]

try add --use-directml

[dylanh724]

Still an issue. Adding --use-directml doesn't work, either; heck it's not even detected as a command. Confirming I'm on this fork.

[lshqqytiger]

Make sure that you are on the latest version 1.9.3.

[tonkgai]

May I ask. version 1.9.3 for what? I have the same problem! I followed this one "git clone https://github.com/lshqqytiger/stable-diffusion-webui-directml && cd stable-diffusion-webui-directml && git submodule init && git submodule update"

[shazam1234raja]

I have been having the same issue
I am using SD.next as you see in the picture
NO GPU showing anywhere
1 generation takes with Width: 320 Height 320 takes more then 30mins

Running
windows 10 pro
GPU - AMD Radeon RX 6600 XT
CPU - AMD Ryzen 5 5600 6-Core Processor
VRAM - 8176 MB GDDR6 2000 MHz
RAM - 16 GB

Starting SD.Next
17:26:35-998149 INFO Logger: file="E:\Stable Diffusion\ZLUDA STABLE diffusion\automatic\sdnext.log" level=DEBUG
size=65 mode=create
17:26:36-000125 INFO Python: version=3.10.6 platform=Windows bin="E:\Stable Diffusion\ZLUDA STABLE
diffusion\automatic\venv\Scripts\python.exe" venv="E:\Stable Diffusion\ZLUDA STABLE
diffusion\automatic\venv"
17:26:36-431118 INFO Version: app=sd.next updated=2024-12-09 hash=a2f5a39a branch=master
url=https://github.com/vladmandic/automatic/tree/master ui=main
17:26:37-020432 INFO Platform: arch=AMD64 cpu=AMD64 Family 25 Model 33 Stepping 2, AuthenticAMD system=Windows
release=Windows-10-10.0.19045-SP0 python=3.10.6 docker=False
17:26:37-024286 DEBUG Packages: venv=venv site=['venv', 'venv\lib\site-packages']
17:26:37-025287 INFO Args: ['--use-zluda', '--debug', '--autolaunch']
17:26:37-027287 DEBUG Setting environment tuning
17:26:37-028288 DEBUG Torch allocator: "garbage_collection_threshold:0.80,max_split_size_mb:512"
17:26:37-034289 DEBUG Torch overrides: cuda=False rocm=False ipex=False directml=False openvino=False zluda=True
17:26:37-045292 INFO Python: version=3.10.6 platform=Windows bin="E:\Stable Diffusion\ZLUDA STABLE
diffusion\automatic\venv\Scripts\python.exe" venv="E:\Stable Diffusion\ZLUDA STABLE
diffusion\automatic\venv"
17:26:37-050315 INFO ROCm: AMD toolkit detected
17:26:37-499652 INFO ROCm: agents=['gfx1032']
17:26:37-501652 INFO ROCm: version=6.2, using agent gfx1032
17:26:37-537660 WARNING Failed to load ZLUDA: Could not find module 'E:\Stable Diffusion\ZLUDA STABLE
diffusion\automatic.zluda\nvrtc64_112_0.dll' (or one of its dependencies). Try using the full
path with constructor syntax.
17:26:37-539702 INFO Using CPU-only torch
17:26:37-726465 INFO Install: verifying requirements
17:26:37-730508 INFO Verifying packages
17:26:37-817570 DEBUG Timestamp repository update time: Mon Dec 9 13:57:14 2024
17:26:37-819571 INFO Startup: standard
17:26:37-820571 INFO Verifying submodules
17:26:43-252439 DEBUG Git submodule: extensions-builtin/sd-extension-chainner / main
17:26:43-520436 DEBUG Git detached head detected: folder="extensions-builtin/sd-extension-system-info" reattach=main
17:26:43-522436 DEBUG Git submodule: extensions-builtin/sd-extension-system-info / main
17:26:43-710877 DEBUG Git submodule: extensions-builtin/sd-webui-agent-scheduler / main
17:26:43-984535 DEBUG Git detached head detected: folder="extensions-builtin/sdnext-modernui" reattach=main
17:26:43-985536 DEBUG Git submodule: extensions-builtin/sdnext-modernui / main
17:26:44-176897 DEBUG Git submodule: extensions-builtin/stable-diffusion-webui-rembg / master
17:26:44-366913 DEBUG Git submodule: modules/k-diffusion / master
17:26:44-638891 DEBUG Git detached head detected: folder="wiki" reattach=master
17:26:44-640893 DEBUG Git submodule: wiki / master
17:26:44-758919 DEBUG Register paths
17:26:44-813931 DEBUG Installed packages: 182
17:26:44-814931 DEBUG Extensions all: ['Lora', 'sd-extension-chainner', 'sd-extension-system-info',
'sd-webui-agent-scheduler', 'sdnext-modernui', 'stable-diffusion-webui-rembg']
17:26:44-970965 DEBUG Extension installer: E:\Stable Diffusion\ZLUDA STABLE
diffusion\automatic\extensions-builtin\sd-webui-agent-scheduler\install.py
17:26:46-277078 DEBUG Extension installer: E:\Stable Diffusion\ZLUDA STABLE
diffusion\automatic\extensions-builtin\stable-diffusion-webui-rembg\install.py
17:26:55-136993 DEBUG ONNX: version=1.20.1 provider=CPUExecutionProvider, available=['AzureExecutionProvider',
'CPUExecutionProvider']
17:26:55-179003 INFO Device:

[lshqqytiger]

main branch does not support HIP SDK 6.2. Try dev branch with new venv.

[shazam1234raja]

can you please simplify the answer

benchmark problem