Skip to content
This repository has been archived by the owner on Aug 10, 2023. It is now read-only.

High Vram usage and UI issues #58

Closed
Vendetta-S opened this issue Dec 14, 2022 · 7 comments
Closed

High Vram usage and UI issues #58

Vendetta-S opened this issue Dec 14, 2022 · 7 comments

Comments

@Vendetta-S
Copy link

The High Vram usage I have been experiencing never really happen and the UI issues didn't bother me, until recently. I did the same thing and yesterday it worked but now something is off.
The unusual high vram usage I am seeing when it starts to generate and while it's generating too.
image
(The orange peak happened when it was done generating and decoded the images)
I know it didn't peak like this, before I screenshotted the image I did my normal AI generating and it peaked on the RED (which means it used almost all of the Vram).
And the UI is broken too, it stops halfway when I was generating it and it always happens if I used the High res fix.
image
The High Vram usage might be the model or the server or even colab but the UI one is really annoying.

@acheong08
Copy link
Owner

Regarding the high vram usage, you can enable --medvram in the vram options.

Gradio is unstable for an unknown reason. It will take some time to figure out

@Enferlain
Copy link

Yeah the ui thing is really weird, some days I get 0 issues, today it gets stuck on every normal generation.

@Vendetta-S
Copy link
Author

Vendetta-S commented Dec 15, 2022

Apologies if I didn't respond quickly, but I noticed this isn't a isolated incident. I tested another colab and it had the same issues as this one, assuming that other colabs are the same that means this is something from either AUTOMATIC1111's github or gradio itself.
And Yes, I did use the --medvram option but it still takes majority of the Vram. But, I still don't know why it happened. I'll close this issue if the UI issues and Vram issues still persist.

@acheong08
Copy link
Owner

I have tested this myself and verified that it is a universal issue. Gradio is currently experiencing extremely high demand which causes some network issues. Automatic's repo doesn't handle these errors well which is why it gets stuck.

And Yes, I did use the --medvram option but it still takes majority of the Vram. But, I still don't know why it happened. I'll close this issue if the UI issues and Vram issues still persist.

This seems to be caused new code in Automatic's repo. Difficult to fix from my end but it'll check it out

@Vendetta-S
Copy link
Author

Vendetta-S commented Dec 15, 2022

Decided to run it and this happened:
image
Well crap. Somehow GFPGAN decided to not work anymore.
I run it again and the error didn't happen. And, another error happened and CLIP didn't want to work.
image
And it decided to work again.

@Vendetta-S
Copy link
Author

And another issue:
image
Since multiple issues have been reported, I'll split these two issues into a separate issue page for later fixing.

@Vendetta-S
Copy link
Author

The high Vram usage isn't nearly problematic now but it still peaks at the orange peaks. For now, it seems like the only issue here is the high Vram one.

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

No branches or pull requests

3 participants