2

I'm trying to set up stable diffusion on a server so that users can access it via RDP and generate what they need. The server in place I hosted on-premise and doesn't have any internet connection. I installed all needed libs via whl files. I was able to set up everything fine for my user, meaning when I run the webui-user.bat everything works fine.

As soon as a different user connects to the server and runs the webui-user.bat that startup runs till a certain point and stops with ConnectionTimeoutError. Here is the console output from the part where it fails.

Launching Web UI with arguments: --no-gradio-queue -disable-nan-check --skip-install
No module 'xformers'. Procedding without it.
Loading weights [aba96b389d] from C:\AutomaticStableDiff\stable-diffusion-webui\models\Stable-diffusion\mdjrny-v4.safetensors
Creating model from config: C:\AutomaticStableDiff\stable-diffusion-webui\configs\v1-inference.yaml
LatentDiffusion: Running in eps-prediction mode
DiffusionWrapper has 859.52 M params.
'HTTPSConnectionPool(host='huggingface.co',port=443): Max retries exceeded with url: /openai/clip-vit-large-patch14/resolve/main/vocab.json (Caused by ConnectTimeoutError(urllib3.connection.HTTPSConnection object at 0x0000021B2E0C9DB0>, 'Connection to huggingface.co timed out. (connect timeout=10)'))' thrown while requesting HEAD https://huggingface.co/openai/clip-vit-large-patch14/resolve/main/vocab.json

I ensured Python is installed and set up globaly on the server not only for my user. I don't get why the problem seems to be user-specific.

What I've tried so far is:

jonrsharpe
  • 115,751
  • 26
  • 228
  • 437
Sacul0815
  • 21
  • 2

1 Answers1

0

Welcome to SO. I would maybe advise against using RDP to share Automatic1111's Stable Diffusion GUI in that way, it could be possible that once you're trying to boot multiple instances of the webui (which I have found that it does not like) simultaneously instead of connecting directly to the instance. I would recommend instead in webui-user.bat use COMMANDLINE ARGS=--share, once it's loaded you'll get a sharable grad.io link in the console. If you want the LAN version, you could use --listen instead of --share, that might give you the solution you're looking for.

Brock Brown
  • 956
  • 5
  • 11