You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Running the consistency decoder takes several seconds and most of this time is spent in a stalled state and reducing the number of diffusion steps leads to no meaningful speed increase. The default SD1.5 decoder is ~100x faster running the code example in the readme.
I'm on Pytorch 2.0.1 on Linux kernel 6.1 with an RTX 3060
The text was updated successfully, but these errors were encountered:
Running the consistency decoder takes several seconds and most of this time is spent in a stalled state and reducing the number of diffusion steps leads to no meaningful speed increase. The default SD1.5 decoder is ~100x faster running the code example in the readme.
I'm on Pytorch 2.0.1 on Linux kernel 6.1 with an RTX 3060
The text was updated successfully, but these errors were encountered: