Troubleshooting
- The program needs 16gb of regular RAM to run smoothly. If you have 8gb RAM, consider making an 8gb page file/swap file, or use the --lowram option (if you have more gpu vram than ram).
- The installer creates a python virtual environment, so none of the installed modules will affect existing system installations of python.
- To use the system's python rather than creating a virtual environment, use custom parameter replacing
set VENV_DIR=-. - To reinstall from scratch, delete directories:
venv,repositories. - When starting the program for the first time, the path to python interpreter is displayed. If this is not the python you installed, you can specify full path in the
webui-userscript; see Running with custom parameters. - If the desired version of Python is not in PATH, modify the line
set PYTHON=pythoninwebui-user.batwith the full path to the python executable.- Example:
set PYTHON=B:\soft\Python310\python.exe
- Example:
- Installer requirements from
requirements_versions.txt, which lists versions for modules specifically compatible with Python 3.10.6. If this doesn't work with other versions of Python, setting the custom parameterset REQS_FILE=requirements.txtmay help.
When running on video cards with a low amount of VRAM (<=4GB), out of memory errors may arise. Various optimizations may be enabled through command line arguments, sacrificing some/a lot of speed in favor of using less VRAM:
- If you have 4GB VRAM and want to make 512x512 (or maybe up to 640x640) images, use
--medvram. - If you have 4GB VRAM and want to make 512x512 images, but you get an out of memory error with
--medvram, use--lowvram --always-batch-cond-uncondinstead. - If you have 4GB VRAM and want to make images larger than you can with
--medvram, use--lowvram.
This is one of the most frequently mentioned problems, but it's usually not a WebUI fault, there are many reasons for it.
- WebUI uses GPU by default, so if you don't have suitable hardware, you need to add
--use-cpu. - Make sure you configure the WebUI correctly, refer to the corresponding installation tutorial in the wiki.
- If you encounter this issue after some component updates, try undoing the most recent actions.
If you are one of the above, you should delete the venv folder.
If you still can't solve the problem, you need to submit some additional information when reporting.
- Open the console under
venv\Scripts - Run
python -m torch.utils.collect_env - Copy all the output of the console and post it
Video cards
Certain GPU video cards don't support half precision: a green or black screen may appear instead of the generated pictures. Use --upcast-sampling. This should stack with --xformers if you are using.
If still not fixed, use command line arguments --precision full --no-half at a significant increase in VRAM usage, which may require --medvram.
Your installed xformers is incompatible with your GPU. If you use Python 3.10, have a Pascal or higher card and run on Windows, add --reinstall-xformers --xformers to your COMMANDLINE_ARGS to upgrade to a working version. Remove --reinstall-xformers after upgrading.
If you use Windows, this means your Python is too old. Use 3.10
If Linux, you'll have to build xformers yourself or just avoid using xformers.
Windows defender/antiviruses sometimes blocks Gradio's ability to create a public URL.
- Go to your antivirus
- Check the protection history:
- Add it as an exclusion
Related issues:
This issue has been noted, 3 times. It is apparently something users in china may experience. #8537
Solution:
This problem is caused by errors in the CSS file type information in my computer registry, which leads to errors in CSS parsing and application. Solution:
According to the above image to locate, and modify the last Content Type and PerceivedType. Finally, reboot the machine, delete the browser cache, and force refresh the web page (shift+f5). Thanks to https://www.bilibili.com/read/cv19519519