Why My Generations are So Slow? #9631
-
I have a RTX4090 and a generation with 60 steps 512x512 is taking 8 seconds in euler a, wich is the same or I think slower time I used to take on my RTX 3080, is something wrong? |
Beta Was this translation helpful? Give feedback.
Replies: 4 comments 5 replies
-
If you're using torch 1.13 (the default), download this and put the contents of the This is necessary for 40xx cards with torch < 2.0. |
Beta Was this translation helpful? Give feedback.
-
The other fork you said has been upgraded to PyTorch 2 and the main project has not yet, try to replace the code here, then delete the venv and reinstall.
|
Beta Was this translation helpful? Give feedback.
-
The guy who's fork you downloaded coded an extension for A1111 that shows all your system info (all versions etc). It'll also run a benchmark with your current setup which you can use to compare with other people and see what versions (CUDA/Pytorch/etc/etc) they're using to get better speeds. Extension's repo: https://github.com/vladmandic/sd-extension-system-info Site to just look at benchmarks/info other people have shared (can sort by speed/gpu/etc): https://vladmandic.github.io/sd-extension-system-info/pages/benchmark.html |
Beta Was this translation helpful? Give feedback.
-
Beta Was this translation helpful? Give feedback.
If you're using torch 1.13 (the default), download this and put the contents of the
bin
folder instable-diffusion-webui\venv\Lib\site-packages\torch\lib
This is necessary for 40xx cards with torch < 2.0.