Skip to content

Comment on Stable Diffusion forming images from text: image snapshots at each stepparent

Comments

Interesting. I suppose GPUs could calculate things differently. I just checked, I can rerun both 40/80 step runs and the final images are bit-identical to the first runs. So at least in my scenario the same parameters are deterministic, but changing the number of ddim sampling steps changes the result.

Maybe it's doing something fancy with the total number of steps, beyond just stopping after the count is reached.

It will most likely render differently on different hardware since gpu float math is nondeterministic across different environments

How are cloud GPU providers handling this then? Do the fancy A1000 chips solve this?

aaaand there goes people trying to turn this into a 99% compression system

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.