Replies: 2 comments 1 reply
-
I agree that implementing Teacache for flux workflows would speed up generation for many users 2x without loss in quality. Not sure about wavespeed. |
Beta Was this translation helpful? Give feedback.
0 replies
-
It can be enabled in Performance settings (called "Dynamic Caching") since v1.31.0 |
Beta Was this translation helpful? Give feedback.
1 reply
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Hi, Acly, could you please consider adding teacache or wavespeed nodes to Krita AI? My graphics card is an RTX 2060 Super 8G. When using Krita AI to generate a 1024x1024 image with 20 steps using flux-dev-fp8, it takes an average of 103 seconds. after using the wavespeed node, it only takes 67 seconds, improving the speed by 1.5 to 1.6 times, and there doesn't seem to be any noticeable difference in image quality.
I look forward to seeing this node added to Krita AI, as it would be a great help for generating images using the FLUX model. Thank you!
WaveSpeed: https://github.com/chengzeyi/Comfy-WaveSpeed
Teacache: https://github.com/ali-vilab/TeaCache
Beta Was this translation helpful? Give feedback.
All reactions