New extension: System Info

[vladmandic]

Creates a top-level System Info tab in Automatic WebUI

URL: https://github.com/vladmandic/sd-extension-system-info

System Info

Note:

• State & memory info are auto-updated every second if tab is visible (no updates are performed when tab is not visible)
• All other information is updated once upon WebUI load and can be force refreshed if required

Current information:

• Server start time
• Version
• Current Model & VAE
• Current State
• Current Memory statistics
  System data:
• Platform details
• Torch, CUDA and GPU details
• Active CMD flags such as low-vram or med-vram
• Versions of critical libraries as xformers, transformers, etc.
• Versions of dependent repositories such as k-diffusion, etc.
  Models
  Hypernetworks
  Embeddings (including info on number of vectors per embedding)

Benchmarks

Run standardized benchmark and (optionally) publish results online
Online viewer allows for sort/search/filter of all submitted results for comparison

• Extension URL
• Details
• Online Results

[Nacurutu]

AWESOME... back again.... thanks...

[cgessai]

I'm not seeing how to install this and I don't see it available on the Extensions tab to install through the UI. (?)

[Nacurutu]

I'm not seeing how to install this and I don't see it available on the Extensions tab to install through the UI. (?)

Just paste the link of the repo in the "Install from URL" tab on Extensions.

[mclsugi]

IMHO, this should be a built-in extension, and perhaps could generate some sort of files that is mandatory to be included in every issue/bug reports to avoid reporter saying 'last version' or 'since I updates yesterday' or giving a wrong/blank system info, or think they can run three cross attention together, etc, etc

[vladmandic]

New feature published couple days ago

Run standardized benchmark and (optionally) publish results online

Online viewer allows for sort/search/filter of all submitted results for comparison

• Extension URL
• Details
• Online Results

[vladmandic]

one month in and collected nearly 1,200 unique benchmark samples, ranging from 0.08 it/s to 51+ it/s - thanks all!

[Nacurutu]

thank you for the benchmark tool..!

[quasiblob]

Thanks! Sent my results now to add one more :D

BTW some suggestions;

• Don't populate username field with Windows login name automatically. I wouldn't want my login name accidentally sent
• Suggest some standard testing setup in UI? Not sure if model matters that much, but now folks have used whatever model they have in use

[vladmandic]

Don't populate username field with Windows login name automatically. I wouldn't want my login name accidentally sent

As most users don't populate it themselves, I want to have a default and currently logged in user is only thing that comes to mind.
You can always clear or edit it before running.

Suggest some standard testing setup in UI? Not sure if model matters that much, but now folks have used whatever model they have in use

I haven't seen any major impact on model used, so I left it. Switching model can be time consuming, so this is just path of least resistence.

[quasiblob]

Well it could be simply "user" or "anonymous", but I get your point.