Closed alessandroperilli closed 10 months ago
It doesn't decrease iteration time, it decreases the number of steps needed. So to make generations faster, you need to decrease the steps (4 to 8 range works well)
hey i tried again and for an instant painting app workflow and seems to be having no issues and generating at super cool fast speeds. May be i will give a link to a video using same. https://www.youtube.com/watch?v=ZF1GJasG7kc check out this workflow here and use case.
OK. Now I better understand the point. And, after tests, I can see the LCM-LoRA works as intended in the Efficiency Loader node. Thank you both for the help.
I've seen videos of people loading the new LCM-LoRAs greatly reducing the image generation time.
However, loading one of them via the
Efficiency Loader
node seems to have no effect on my system:The image generation speed remains identical.
It might well be that, despite what I read, these new LoRAs don't support Apple MPS, but I wanted to flag it just in case it's a problem with how the LoRAs are loaded by the Efficient Loader node.
Thank you.