InvokeAI: WebUI performance issues (memory leak?)
Describe your environment
- GPU: mps
- VRAM: 64 GB
- CPU arch: ARM
- OS: macOS
- Python: anaconda
- Branch: Development
- Commit: 3170c83d8de7189fae3aa6df555bd40abc2d0328
Describe the bug
pulled latest dev branch this morning and webUI seems very slow to respond/very laggy. Taking a look at memory usage, I am seeing python consuming 12GB of RAM at idle. When running just dream.py without --web, idle memory usage is 4.5 GB of RAM. when starting the web server, it starts at 4.5, after generating one image in the UI, that goes up to 7.2, it appears that usage may be climbing with each subsequent image generation
To Reproduce
launch dream.py with --web and --host flags, generate a few images. general usage of the webUI on Mac
Additional context not sure if this is Mac specific or how to isolate further. I did reset the webUI as well as clearing the output directory incase this was trying to load up all of the image previews in gallery and issue persisted
About this issue
- Original URL
- State: closed
- Created 2 years ago
- Comments: 20 (12 by maintainers)
Please try out #1009 and see if it resolves this issue.
@psychedelicious Seems to have isolated the bug. It’s some stupid CSS related issue. He’ll deploy a fix soon.
As I posted somewhere else, problem the release-canidate-2 thread, it seems to kick in as images show up in the gallery, no images, gives decent response even with the gallery visible. It may even be more images in the gallery the slower it gets but my crappy 8Gb is not fast enough for me to generate the images in any sensible time to test that properly.
It definitely true for the first 6 or so, 2 images it barely noticeable but by 6 images it’s obvious.
thought it may be safari specific. @blessedcoolant and @Any-Winter-4079, let me know if there’s any additional information or logs I can provide if more debug info is needed