Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

VAEdecode and AITemplate dont get along, slower latent decoding #907

Open
2blackbar opened this issue Aug 19, 2023 · 3 comments
Open

VAEdecode and AITemplate dont get along, slower latent decoding #907

2blackbar opened this issue Aug 19, 2023 · 3 comments

Comments

@2blackbar
Copy link

2blackbar commented Aug 19, 2023

Ok i think the issue is when i use AITemplate node, it breaks the vaedecoder node and it starts to choke deconding takes few seconds and i pretty much have to restart comfy to get it working instantly again.
When i dont use AITemplate nodes then VAEdecode decodes latent to image instantly, when i use even one AITemplate node then it all goes to crap and it takes from 3 to 7 seconds to vaedecode latent to image

Even if i use AITemplate node at the end of the chain (after VAEdecode) then it will slow down vaedecode after awhile so theres definintely some issue going on with it.

I tried to use AITemplae vae decoder but it does not work with face detailer node, so im stuck

@aakhundov
Copy link
Contributor

@2blackbar Thanks for reporting the issue. As currently described, it's very hard to reason about the problem. Would you like to share a minimal repro of the issue, i.e. when VAE decoding with AITemplate takes longer than you'd expect? Thank you.

@2blackbar
Copy link
Author

Its easy, use non AITemplate VAE node first, generate like 3 pics, then swap this node to AITemplate and enjoy the slowness

@2blackbar
Copy link
Author

This repo AIT also breaks other workflows in comfyui even if the node is not added to the workflow, somethin is ver very wrong here

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants