Skip to content

Having trouble while generating images on ComfyUI by AMD with Zluda #379

Description

@Knowes

Custom Node Testing

Your question

Hi, I'm a first time Comfy enjoyer and finally got it to open without problems on my RX 6800xt and Windows. I used this repo https://github.com/patientx/ComfyUI-Zluda

Turns out I'm kinda lost with the workflow and nodes. I was having problems using the base custom workflow for sdxl provided by the platform itself, having to disable custom nodes.
Now I'm using a random workflow from OpenArt.
I tried both w/o refiner, but I don't even know whats the refiner node for, and when (if) should I use. Anyways, everytime I tried generating images, I got a black image, I'm not sure how to fix that.

My cmd was:

HIP_PATH = C:\Program Files\AMD\ROCm\6.2\

HIP_PATH_57 = C:\Program Files\AMD\ROCm\5.7\

HIP_PATH_62 = C:\Program Files\AMD\ROCm\6.2\

[INFO] Detected Visual Studio Build Tools version: 18.0.2

[INFO] Detected Git version: 2.47.1.windows.1

[INFO] ComfyUI-Zluda current path: E:\ComfyUI-Zluda\

[INFO] ComfyUI-Zluda current version: 2025-11-25 14:05:26 hash: de43f7b branch: master

[INFO] Checking and updating to a new version if possible...

[INFO] Already up to date.

[INFO] AMD Software version: 25.11.1

[INFO] ZLUDA version: 3.9.5 [nightly build]

[INFO] Launching application via ZLUDA...

Checkpoint files will always be loaded safely.

%%% [backported] imported triton knobs

:: Checking package versions...

Found pydantic: 2.12.4, pydantic-settings: 2.12.0

:: Pydantic packages are compatible, skipping reinstall

Installed version of comfyui-frontend-package: 1.30.6

Installed version of comfyui-workflow-templates: 0.7.9

Installed version of av: 16.0.1

Installed version of comfyui-embedded-docs: 0.3.1

:: Package version check complete.

:: ------------------------ ZLUDA ----------------------- ::

:: Auto-detecting AMD GPU architecture for Triton...

:: Detected GPU via Windows registry: AMD Radeon RX 6800 XT

:: Set TRITON_OVERRIDE_ARCH=gfx1030

:: Triton core imported successfully

:: Detected Triton version: 3.4.0

:: Enabled cuDNN

:: Running Triton kernel test...

%%% [info] triton/runtime/build/platform_key: AMD64,Windows,64bit,WindowsPE

:: Triton kernel test passed successfully

:: Triton initialized successfully

:: Patching ONNX Runtime for ZLUDA — disabling CUDA EP.

:: Using ZLUDA with device: AMD Radeon RX 6800 XT [ZLUDA]

:: Applying core ZLUDA patches...

:: Initializing Triton optimizations

:: Configuring Triton device properties...

:: Triton device properties configured

:: Flash attention components found

:: AMD flash attention enabled successfully

:: Configuring PyTorch backends...

:: Disabled CUDA flash attention

:: Enabled math attention fallback

:: ZLUDA initialization complete

:: ------------------------ ZLUDA ----------------------- ::

Total VRAM 16368 MB, total RAM 32678 MB

pytorch version: 2.7.0+cu118

Set vram state to: NORMAL_VRAM

Device: cuda:0 AMD Radeon RX 6800 XT [ZLUDA] : native

Enabled pinned memory 14705.0

Using sub quadratic optimization for attention, if you have memory or speed issues try using: --use-split-cross-attention

Python version: 3.11.9 (tags/v3.11.9:de54cf5, Apr 2 2024, 10:12:12) [MSC v.1938 64 bit (AMD64)]

ComfyUI version: 0.3.71

ComfyUI frontend version: 1.30.6

[Prompt Server] web root: E:\ComfyUI-Zluda\venv\Lib\site-packages\comfyui_frontend_package\static

Total VRAM 16368 MB, total RAM 32678 MB

pytorch version: 2.7.0+cu118

Set vram state to: NORMAL_VRAM

Device: cuda:0 AMD Radeon RX 6800 XT [ZLUDA] : native

Enabled pinned memory 14705.0

Skipping loading of custom nodes

Context impl SQLiteImpl.

Will assume non-transactional DDL.

No target revision found.

Starting server

To see the GUI go to: http://127.0.0.1:8188

got prompt

model weight dtype torch.float16, manual cast: None

model_type EPS

Using split attention in VAE

Using split attention in VAE

VAE load device: cuda:0, offload device: cpu, dtype: torch.float32

Requested to load SDXLClipModel

loaded completely; 95367431640625005117571072.00 MB usable, 1560.80 MB loaded, full load: True

CLIP/text encoder model load device: cuda:0, offload device: cpu, current: cuda:0, dtype: torch.float16

Requested to load SDXL

loaded completely; 11503.12 MB usable, 4897.05 MB loaded, full load: True

100%|██████████████████████████████████████████████████████████████████████████████████| 25/25 [00:33<00:00, 1.34s/it]

Requested to load AutoencoderKL

loaded partially: 418.18 MB loaded, lowvram patches: 0

loaded completely; 361.69 MB usable, 319.11 MB loaded, full load: True

E:\ComfyUI-Zluda\nodes.py:1594: RuntimeWarning: invalid value encountered in cast

img = Image.fromarray(np.clip(i, 0, 255).astype(np.uint8))

Prompt executed in 48.95 seconds

got prompt

100%|██████████████████████████████████████████████████████████████████████████████████| 25/25 [01:49<00:00, 4.38s/it]

Prompt executed in 111.20 seconds

got prompt

Prompt executed in 0.00 seconds

got prompt

Prompt executed in 0.00 seconds

got prompt

loaded completely; 8346.05 MB usable, 1560.80 MB loaded, full load: True

100%|██████████████████████████████████████████████████████████████████████████████████| 25/25 [00:29<00:00, 1.17s/it]

loaded partially: 443.20 MB loaded, lowvram patches: 0

Prompt executed in 54.01 seconds

got prompt

100%|██████████████████████████████████████████████████████████████████████████████████| 25/25 [01:15<00:00, 3.04s/it]

Prompt executed in 77.60 seconds

got prompt

Using split attention in VAE

Using split attention in VAE

VAE load device: cuda:0, offload device: cpu, dtype: torch.float32

loaded completely; 8346.05 MB usable, 1560.80 MB loaded, full load: True

Compilation is in progress. Please wait...

0%| | 0/25 [00:00<?, ?it/s]E:\ComfyUI-Zluda\venv\Lib\site-packages\torchsde_brownian\brownian_interval.py:608: UserWarning: Should have tb<=t1 but got tb=14.614640235900879 and t1=14.61464.

warnings.warn(f"Should have {tb_name}<=t1 but got {tb_name}={tb} and t1={self._end}.")

100%|██████████████████████████████████████████████████████████████████████████████████| 25/25 [00:29<00:00, 1.20s/it]

0%| | 0/5 [00:00<?, ?it/s]E:\ComfyUI-Zluda\venv\Lib\site-packages\torchsde_brownian\brownian_interval.py:608: UserWarning: Should have tb<=t1 but got tb=0.10288411378860474 and t1=0.102884.

warnings.warn(f"Should have {tb_name}<=t1 but got {tb_name}={tb} and t1={self._end}.")

100%|████████████████████████████████████████████████████████████████████████████████████| 5/5 [00:05<00:00, 1.19s/it]

Requested to load AutoencoderKL

loaded partially: 418.18 MB loaded, lowvram patches: 0

loaded completely; 348.98 MB usable, 319.11 MB loaded, full load: True

Prompt executed in 125.47 seconds


I'm using these:

set "COMMANDLINE_ARGS=--auto-launch --use-quad-cross-attention --reserve-vram 0.9 --disable-all-custom-nodes --fp32-vae"

Had to use --disable-all-custom-nodes cause apparently you can get errors while running with it enabled (and after I disabled it, some TRITON recognition errors got fixed);
I'm using --fp32-vae cause I got some VAE errors and searched in some AI chatbots and got this resolution

Since I'cant upload the whole workflow, here is the link: https://openart.ai/workflows/openart/-/Q6n55PcgIumjDRoPmPHX

I'm trying to generate some 2d anime style images while using waiNSFWIllustrious_v120.safetensors model
After oploading the worflow, I just changed the model on both nodes and the prompt

Here are some info about the current state of it:

Image

Can someone help me?

Logs

Other

Since I'm new to all of it, and tried to get info from some other sources and from AI, I guess I need practical experience knowledge. Where can I get the most basic but functional SDXL worflow (trying to use with Illustrious) so I can test it out to see if there are any other causes to it?

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions