Not working on Docker Desktop (Windows)

[tudorms]

It gets stuck after loading the VAE, I let it sit there for a long time and there are no additional logs or progress. This is Docker Desktop on Windows 11 (WSL2), and a Nvidia 5090. The same model and workflow loads and runs fine with the ComfyUI desktop application. Tried both cu128 and cu130 with the same result (ComfyUI Desktop is using cu130)

Logs:

docker run -it --rm --runtime=nvidia --gpus all -p 8188:8188 -v .\storage:/root -v .\storage-models\models:/root/ComfyUI/models -v .\storage-models\hf-hub:/root/.cache/huggingface/hub -v .\storage-models\torch-hub:/root/.cache/torch/hub -v .\storage-user\input:/root/ComfyUI/input -v .\storage-user\output:/root/ComfyUI/output -v .\storage-user\workflows:/root/ComfyUI/user/default/workflows -e CLI_ARGS="--disable-xformers" yanwk/comfyui-boot:cu130-slim
########################################
[INFO] Running set-proxy script...
[INFO] Continue without proxy.
[INFO] Using existing ComfyUI in user storage...
[INFO] Running pre-start script...
[INFO] Continue without pre-start script.
[INFO] Starting ComfyUI...
########################################
[START] Security scan
[DONE] Security scan
## ComfyUI-Manager: installing dependencies done.
** ComfyUI startup time: 2026-02-04 03:09:51.668
** Platform: Linux
** Python version: 3.13.11 (main, Dec 19 2025, 18:38:03) [GCC]
** Python executable: /usr/bin/python3
** ComfyUI Path: /root/ComfyUI
** ComfyUI Base Folder Path: /root/ComfyUI
** User directory: /root/ComfyUI/user
** ComfyUI-Manager config path: /root/ComfyUI/user/__manager/config.ini
** Log path: /root/ComfyUI/user/comfyui.log
[ComfyUI-Manager] PyTorch is not installed
[ComfyUI-Manager] 'comfyui-frontend-package' dependency were fixed

Prestartup times for custom nodes:
  12.3 seconds: /root/ComfyUI/custom_nodes/ComfyUI-Manager

Checkpoint files will always be loaded safely.
Total VRAM 32607 MB, total RAM 46972 MB
pytorch version: 2.10.0+cu130
Set vram state to: NORMAL_VRAM
Device: cuda:0 NVIDIA GeForce RTX 5090 : cudaMallocAsync
Using async weight offloading with 2 streams
Enabled pinned memory 44623.0
Found comfy_kitchen backend eager: {'available': True, 'disabled': False, 'unavailable_reason': None, 'capabilities': ['apply_rope', 'apply_rope1', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8', 'scaled_mm_nvfp4']}
Found comfy_kitchen backend triton: {'available': True, 'disabled': True, 'unavailable_reason': None, 'capabilities': ['apply_rope', 'apply_rope1', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8']}
Found comfy_kitchen backend cuda: {'available': True, 'disabled': False, 'unavailable_reason': None, 'capabilities': ['apply_rope', 'apply_rope1', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8', 'scaled_mm_nvfp4']}
Using pytorch attention
Python version: 3.13.11 (main, Dec 19 2025, 18:38:03) [GCC]
ComfyUI version: 0.11.1
ComfyUI frontend version: 1.37.11
[Prompt Server] web root: /usr/local/lib/python3.13/site-packages/comfyui_frontend_package/static
2026-02-04 03:10:17.906558440 [W:onnxruntime:Default, device_discovery.cc:164 DiscoverDevicesForPlatform] GPU device discovery failed: device_discovery.cc:89 ReadFileContents Failed to open file: "/sys/class/drm/card0/device/vendor"
### Loading: ComfyUI-Manager (V3.39.2)
[ComfyUI-Manager] network_mode: public
[ComfyUI-Manager] ComfyUI per-queue preview override detected (PR #11261). Manager's preview method feature is disabled. Use ComfyUI's --preview-method CLI option or 'Settings > Execution > Live preview method'.
### ComfyUI Version: v0.11.1 | Released on '2026-01-29'

Import times for custom nodes:
   0.0 seconds: /root/ComfyUI/custom_nodes/websocket_image_save.py
   0.5 seconds: /root/ComfyUI/custom_nodes/ComfyUI-Manager

[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/model-list.json
Context impl SQLiteImpl.
Will assume non-transactional DDL.
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/alter-list.json
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/github-stats.json
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/extension-node-map.json
Assets scan(roots=['models']) completed in 0.138s (created=0, skipped_existing=26, total_seen=26)
[ComfyUI-Manager] default cache updated: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/custom-node-list.json
Starting server

To see the GUI go to: http://0.0.0.0:8188
To see the GUI go to: http://[::]:8188
FETCH ComfyRegistry Data: 5/123
FETCH ComfyRegistry Data: 10/123
FETCH ComfyRegistry Data: 15/123
FETCH ComfyRegistry Data: 20/123
FETCH ComfyRegistry Data: 25/123
FETCH ComfyRegistry Data: 30/123
FETCH ComfyRegistry Data: 35/123
FETCH ComfyRegistry Data: 40/123
FETCH ComfyRegistry Data: 45/123
FETCH ComfyRegistry Data: 50/123
FETCH ComfyRegistry Data: 55/123
FETCH ComfyRegistry Data: 60/123
FETCH ComfyRegistry Data: 65/123
FETCH ComfyRegistry Data: 70/123
FETCH ComfyRegistry Data: 75/123
FETCH ComfyRegistry Data: 80/123
FETCH ComfyRegistry Data: 85/123
FETCH ComfyRegistry Data: 90/123
FETCH ComfyRegistry Data: 95/123
FETCH ComfyRegistry Data: 100/123
FETCH ComfyRegistry Data: 105/123
FETCH ComfyRegistry Data: 110/123
FETCH ComfyRegistry Data: 115/123
FETCH ComfyRegistry Data: 120/123
FETCH ComfyRegistry Data [DONE]
[ComfyUI-Manager] default cache updated: https://api.comfy.org/nodes
FETCH DATA from: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/custom-node-list.json [DONE]
[ComfyUI-Manager] All startup tasks have been completed.
[DEPRECATION WARNING] Detected import of deprecated legacy API: /scripts/ui.js. This is likely caused by a custom node extension using outdated APIs. Please update your extensions or contact the extension author for an updated version.
[DEPRECATION WARNING] Detected import of deprecated legacy API: /extensions/core/groupNode.js. This is likely caused by a custom node extension using outdated APIs. Please update your extensions or contact the extension author for an updated version.
[DEPRECATION WARNING] Detected import of deprecated legacy API: /scripts/ui/components/buttonGroup.js. This is likely caused by a custom node extension using outdated APIs. Please update your extensions or contact the extension author for an updated version.
[DEPRECATION WARNING] Detected import of deprecated legacy API: /scripts/ui/components/button.js. This is likely caused by a custom node extension using outdated APIs. Please update your extensions or contact the extension author for an updated version.
got prompt
Using pytorch attention in VAE
Using pytorch attention in VAE
VAE load device: cuda:0, offload device: cpu, dtype: torch.bfloat16
^C
Stopped server

[YanWenKun]

Looks like the KSampler is not running at all. Can SD 1.5 models work?

[tudorms]

I tried SD 3.5 Simple from the built-in Templates. It got a little bit further in the logs but it still gets stuck here:

FETCH ComfyRegistry Data [DONE]
[ComfyUI-Manager] default cache updated: https://api.comfy.org/nodes
FETCH DATA from: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/custom-node-list.json [DONE]
[ComfyUI-Manager] All startup tasks have been completed.
got prompt
Found quantization metadata version 1
Detected mixed precision quantization
Using mixed precision operations
model weight dtype torch.float16, manual cast: torch.float16
model_type FLOW
Using pytorch attention in VAE
Using pytorch attention in VAE
VAE load device: cuda:0, offload device: cpu, dtype: torch.bfloat16
Found quantization metadata version 1
Using MixedPrecisionOps for text encoder
^C
Stopped server

[YanWenKun]

1. You can check the disk I/O in Task Manager while running the workflow, if it reads at 100 MB/s then it's a WSL2 bottleneck. Transferring files between VM and host is very slow (via network protocols - SMB and 9P). In that case you can put the model files inside the VM (e.g. using Docker volume).

2. If it's not about I/O, it could be a compatibility issue. We need more information, try the basic SD 1.5 model with verbose logging:

docker run -it --rm \
  --name comfyui-cu130 \
  --runtime nvidia \
  --gpus all \
  -p 8188:8188 \
  -v temp-models:/root/ComfyUI/models \
  -e CLI_ARGS="--verbose=DEBUG" \
  yanwk/comfyui-boot:cu130-slim \
    bash -c 'cd /root/ComfyUI/models/checkpoints/ \
      && curl -LO https://huggingface.co/Comfy-Org/stable-diffusion-v1-5-archive/resolve/main/v1-5-pruned-emaonly-fp16.safetensors \
      && bash /runner-scripts/entrypoint.sh'

Then in the browser's ComfyUI page, close the blank workflow tab, so the default workflow will show up. Try running it to see if any useful message pop up. If it hangs, before pressing Ctrl+C, you can try clicking "Stop" button in ComfyUI to see if it can properly stop the job.

[tudorms]

The SD 1.5 model was successful, it generated an image pretty quickly.

I then tried Z-Image Turbo. After 29 minutes the load was complete and the original prompt ran. Subsequent runs with the same model came back very quickly so you were right this is some kind of bottleneck with the mapped mounts from the Windows file system.

Do you know if there any workarounds besides keeping the files on the WSL (Linux) side or inside the container itself? I was hoping to be able to share the model folder with ComfyUI Desktop (and StabilityMatrix) to avoid duplicates.

Thank you!

[ReinerBforartists]

Just curious, is there a reason to run ComfyUI in a docker at Windows through the WSL bottleneck? Why not simply download the Portable? It should work just fine besides the desktop version.