Similar settings were successful for an ollama container and this open-webui-container can also make use of nvidia-smi. But it always throws an error during initialization (pls. see Logs & Screenshots)
Actual Behavior
Whatever I do in open-webui the tool nvtop does not show any activity. So I'm afraid CUDA is really not used.
Steps to Reproduce
Run open-webui with that image and docker-compose.yml; have a look at the log.
Logs & Screenshots
REPOSITORY TAG IMAGE ID CREATED SIZE
ghcr.io/open-webui/open-webui cuda dbc45818aeb9 19 hours ago 8.57GB
open-webui | CUDA is enabled, appending LD_LIBRARY_PATH to include torch/cudnn & cublas libraries.
open-webui | /app/backend/open_webui
open-webui | /app/backend
open-webui | /app
open-webui | INFO:open_webui.env:GLOBAL_LOG_LEVEL: INFO
open-webui | ERROR:open_webui.env:Error when testing CUDA but USE_CUDA_DOCKER is true. Resetting USE_CUDA_DOCKER to false: CUDA not available
open-webui | NoneType: None
root@open-webui:/app/backend# nvidia-smi
Sun Mar 9 16:20:04 2025
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 570.124.04 Driver Version: 570.124.04 CUDA Version: 12.8 |
|-----------------------------------------+------------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+========================+======================|
| 0 NVIDIA T400 4GB On | 00000000:01:00.0 Off | N/A |
| 38% 33C P8 N/A / 31W | 4MiB / 4096MiB | 0% Default |
| | | N/A |
+-----------------------------------------+------------------------+----------------------+
+-----------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=========================================================================================|
| No running processes found |
+-----------------------------------------------------------------------------------------+
Additional Information
Do I have false expectations or is there still something to configure? What does CUDA not available mean in that context?
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Originally created by @SchwarzerA on GitHub (Mar 9, 2025).
Check Existing Issues
Installation Method
Git Clone
Open WebUI Version
v0.5.20
Ollama Version (if applicable)
0.5.13
Operating System
openSUSE Tumbleweed | 20250307-0 | x86_64
Browser (if applicable)
any (Edge, FireFox, Chromium)
Confirmation
README.md.Expected Behavior
I expected that open-webui:cuda would use the CUDA functionality to improve performance with the help of this docker-compose.yml:
Similar settings were successful for an ollama container and this open-webui-container can also make use of nvidia-smi. But it always throws an error during initialization (pls. see Logs & Screenshots)
Actual Behavior
Whatever I do in open-webui the tool nvtop does not show any activity. So I'm afraid CUDA is really not used.
Steps to Reproduce
Run open-webui with that image and docker-compose.yml; have a look at the log.
Logs & Screenshots
Additional Information
Do I have false expectations or is there still something to configure? What does CUDA not available mean in that context?
Installed software in running os: