[Bug]: Error: The 'onnxruntime-gpu' distribution was not found and is required by the application

[SunGreen777]

Checklist

• The issue exists after disabling all extensions
• The issue exists on a clean installation of webui
• The issue is caused by an extension, but I believe it is caused by a bug in the webui
• The issue exists in the current version of the webui
• The issue has not been reported before recently
• The issue has been reported before but has not been fixed yet

What happened?

I have an error, does anyone know what it is?
Error: The 'onnxruntime-gpu' distribution was not found and is required by the application

Steps to reproduce the problem

What should have happened?

What browsers do you use to access the UI ?

No response

Sysinfo

Console logs

-

Additional information

No response

[waylaa]

Did you follow the steps from here?

After that, if you want ONNX to use DirectML, follow this

[lshqqytiger]

Add --skip-ort if you don't want ONNX.

[SunGreen777]

This is a little confusing, I don’t understand at all whether it is needed or not, and if it is better, faster, then why is it not enabled by default? Can explain please )

[lshqqytiger]

ONNX is slightly fast, Olive optimized ONNX model is much faster. But the overhead follows. (model conversion and optimization)

[SunGreen777]

ONNX is slightly fast, Olive optimized ONNX model is much faster. But the overhead follows. (model conversion and optimization)

I'm understood, thank you :)

[vutym]

open CMD as administrator or Consider using the --user option or check the permissions.

pip install onnxruntime-gpu --extra-index-url https://aiinfra.pkgs.visualstudio.com/PublicPackages/_packaging/onnxruntime-cuda-12/pypi/simple/

[e2heintz]

I had the same error. Partially managed to fix it, but only partially.

I tried using Forge yesterday and it wasn't working properly. Came back to Directml this morning and had lots of errors, ending up with this 'no onnx runtime' thing. Now webui bypassed both my GPU-s and generated on the CPU.

What I did:

1. Deleted venv fom Directml folder : no help
2. pip install torch-directml and onnxruntime-gpu : no help
3. Completely deleted both the Forge folder and the Directml folder, then git pull, then added torch-directml to requirements, added cmd args --use-directml --skip-ort : somewhat working, but…

I cannot use --device-id. Getting ‘invalid string’. Which is very bad. I really really really need that other GPU with its clean vram untouched by the OS and all the apps.

[e2heintz]

Update:
Uninstalled Python and Git. Deleted webui folder comlepetely. Restarted. Reinstalled Python and Git. Git pulled webui folder.
SD is working with --skip-torch-cuda-test --use-directml, but it is using the wrong GPU. I don't want it to use my display-out GPU, I want it to use my secondary GPU.
When adding --device-id, I get this error "'onnxruntime-gpu' distribution was not found' and "invalid device string".
--skip-ort does not help.

[luckyzsd1975]

已经安装了pip install onnxruntime，但在运行open-webui serve时还是报上面的错误，who knows the reason?

[Chr-Wol]

get the same issue, have you got help to solve it?