README.md

metadata

title: Stable Diffusion Webui on Cpu (GPU optional)
emoji: 🎪🎪
colorFrom: blue
colorTo: gray
sdk: gradio
sdk_version: 3.41.2
app_file: app.py
pinned: false
python_version: 3.10.6
duplicated_from: DreamSunny/stable-diffusion-webui-cpu

This space is duplicated from DreamSunny's space, except that I've thrown in some quality of life improvements, disabled some extensions (especially the ones that required a GPU, but yet works on CPU, but awfully slow or don't work at all), different selection of checkpoint models compared to the original space (including optional ones to install), bug fixes in terms of how the original author of the space coded (Loras not showing up on Lora tab for example), as well as a few nice extensions that could be useful in some way. The most important one (but optional) is that there is now an option to change startup code geared towards running it on GPU, but it's not tested a lot sadly, more on that little later. This can be viewed as a better version of the original space, of which this space was originally duplicated from. Credits to the original author of that space that made this possible that made this space exist in the first space.

By default, this space is running on CPU, which might be slow and certain features may or may not work at all. If you want to run it on GPU, you must duplicate this space first, after that, upgrade your newly duplicated space to run it on a GPU.

Using Stable Diffusion XL-based checkpoint models when a space is running CPU-powered is unfortunately not supported and may actually crash the whole space (it might get a connection time out I think) if one tries to do so (be it edited by its source code or downloaded via the use of extensions installed by default), so basically it's probably clear that the only actual way (as far as I am aware of) to use SDXL-based checkpoints is to in fact upgrade it to GPU in order to use them without any kinds of issues.

Do keep in mind that its default startup settings are geared towards running it on CPU (and CPU upgrade), so check up with the source code (its the app dot py file) that runs this space to read instructions on how to change startup settings that are geared towards running it on an upgraded GPU.

Another thing to point out is that unfortunately, the option to change startup settings to run it on GPU is not actually tested at all, so it'd be of great help if you can help test out that option by at least duplicating this space, change the source code by reading the instructions, and upgrading a newly duplicated space to run it on a GPU to test if it's working well or not. I could provide a few startup options, if necessary, in the source code if I get some kind of feedback depending on the hardware upgrade options in question they used.

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference