The Web UI does nothing

[SODDINGIT]

The Web Interface does nothing. I can switch tabs in the menu but no button does anything. I cannot start training or open/select files or paths. The console stays like this until I close it. I have tried 3 different browsers and disabled all ad/script blocking. I have made the server public and accessed it with my phone. I guess I can rule out a browser issue. Something is wrong with the interface.

Editing this to reflect the fact that the issue actually manifsted on both Windows and Linux. The issue seems to have been with Gradio v.3.33.1

←[1;33m=============================================================
Modules installed outside the virtual environment were found.
This can cause issues. Please review the installed modules.

You can uninstall all local modules with:

←[1;34mdeactivate
pip freeze > uninstall.txt
pip uninstall -y -r uninstall.txt
←[1;33m=============================================================←[0m

23:03:59-368512 INFO Version: v21.7.16
23:03:59-374512 INFO nVidia toolkit detected
23:04:01-364454 INFO Torch 1.12.1+cu116
23:04:01-435115 INFO Torch backend: nVidia CUDA 11.6 cuDNN 8700
23:04:01-438119 INFO Torch detected GPU: NVIDIA GeForce RTX 2070 VRAM 8192 Arch (7, 5) Cores 36
23:04:01-440116 INFO Verifying modules instalation status from requirements_windows_torch1.txt...
23:04:01-443118 INFO Verifying modules instalation status from requirements.txt...
23:04:04-707490 INFO headless: False
23:04:04-715490 INFO Load CSS...
Running on local URL: http://127.0.0.1:7860

To create a public link, set share=True in launch().

[toyssamurai]

I am on Ubuntu (on paperspace), and it does nothing, too. In fact, even typing into some of the fields would throw an "error" in the Web UI (such as, the configuration file field's location.)

[sebas1111111]

I'm getting the exact same issue. I've installed following OlivioSarikas' instructions only 2 hours ago and didn't make any changes to the base installation other than installing CUDNN (I'm on an RTX 3080).

[kaan36875]

Same issue, I didn't have an issue when I download it at morning. I was just trying to fresh install now, I saw the old one still works but new one is not.

[sebas1111111]

Same issue, I didn't have an issue when I download it at morning. I was just trying to fresh install now, I saw the old one still works but new one is not.

What version of Torch did you install? I installed Torch 2.

[kaan36875]

I installed Torch 2, too.

[Thebogurtlen]

Same issue. I'm on windows 11. I just installed it an hour ago.

[martinobettucci]

Up, same here: it also broke our previous installation when rebuilt so it seams to me something coming from an upstream dependency @bmaltais

[kaan36875]

I managed to fix it. They will probably fix it soon, but here's a quick fix:

On main folder, you should open cmd and write those commands in order:

.\venv\Scripts\activate
pip uninstall fastapi
pip uninstall pydantic
pip install fastapi==0.99.1
pip install pydantic==1.10.11

[bux]

Still doesn't work with @kaan36875 's fix

[bmaltais]

Is there a k own issue with those two modules? I don't train on Linux so I never experience those issues there... But everything work fine for me on windows. Been training sdxl with the new code in the sdxl branch for a while with no issues. Maybe you could give that updated version a try? Major code rework in there.

[jimtalksdata]

Yes, Pydantic 2 incorporates some breaking changes compared to Pydantic 1. See:

https://docs.pydantic.dev/2.0/migration/
https://github.com/tiangolo/fastapi/releases/tag/0.100.0

[ashleykleynhans]

If this is related to starting the UI with --share, then its a gradio issue and is resolved by bumping gradio to the latest (3.36.1). Seems they broke backwards compatibility in their tunnel or something.

[Pashahlis]

installing gradio v. 3.36.1 fixed the issues for me

python3 -m pip install gradio==3.36.1

[SODDINGIT]

installing gradio v. 3.36.1 fixed the issues for me

python3 -m pip install gradio==3.36.1

i have done this and it did not work at first. Venv would re-install the old Gradio.

20:41:54-941483 WARNING Package wrong version: gradio 3.36.1 required 3.33.1
20:41:54-942106 INFO Installing package: gradio==3.33.1

You have to also change the version in the requirements.txt in the project's root folder. this will keep gradio at 3.36.1

[CodR-2]

I managed to fix it. They will probably fix it soon, but here's a quick fix:

On main folder, you should open cmd and write those commands in order:

.\venv\Scripts\activate pip uninstall fastapi pip uninstall pydantic pip install fastapi==0.99.1 pip install pydantic==1.10.11

Thanks for the fix, it worked for me.

[tkittich]

I managed to fix it. They will probably fix it soon, but here's a quick fix:

On main folder, you should open cmd and write those commands in order:

.\venv\Scripts\activate pip uninstall fastapi pip uninstall pydantic pip install fastapi==0.99.1 pip install pydantic==1.10.11

This fixed the problem for me too. Thanks.

[timnetworks]

I am also seeing the same symptoms, [HTTP/1.1 422 Unprocessable Entity 2ms] when POSTing to http://127.0.0.1:7861/run/predict on latest Ubuntu/Mint. Not able to initiate training at all, and also getting the ERROR popups when entering anything into the DreamBooth LoRA folder paths. Thank you,

[studiomav]

I'm having this same issue, none of the fixes mentioned in this thread work for me. (M1 Mac)