InvokeAI icon indicating copy to clipboard operation
InvokeAI copied to clipboard

Published 20 hours ago • verified publisher icon invoke-ai

[enhancement]: Performance optimization through requirements changes

Open brucethemoose opened this issue 2 years ago • 0 comments

Is there an existing issue for this?

  • [X] I have searched the existing issues

Contact Details

brucethemoose in discord

What should this feature add?

  • Bump pytorch from CUDA 11.6 to 11.7 or 11.8. 11.8 provides slightly better performance on my 2060, but there are reports of much more significant speedups on RTX 4000 hardware.
  • Add pyston-lite-autoload to the requirements for a general boost to python. I've been running system python with pyston-lite-autoload for some time, and have noticed no issues: https://blog.pyston.org/2022/06/08/announcing-pyston-lite-our-python-jit-as-an-extension-module/

brucethemoose avatar Dec 26 '22 07:12 brucethemoose