search

invoke-ai / InvokeAI

Invoke is a leading creative engine for Stable Diffusion models, empowering professionals, artists, and enthusiasts to generate and create visual media using the latest AI-driven technologies. The solution offers an industry leading WebUI, and serves as the foundation for multiple commercial products.
https://invoke-ai.github.io/InvokeAI/
Apache License 2.0
23.57k stars 2.43k forks source link

[bug]: Long UI loading #6310

Open VeyDlin opened 6 months ago

VeyDlin commented 6 months ago

Is there an existing issue for this problem?

Operating system

Windows

GPU vendor

Nvidia (CUDA)

GPU model

RTX 2080 TI

GPU VRAM

11GB

Version number

v4.2.0a4

Browser

firefox developer edition v126.0b9

Python dependencies

No response

What happened

I have a lot of LoRA installed (1000+) and when I first launch the UI, the browser freezes for a few minutes

I installed LoRA by simply copying it to a folder, perhaps this is important Restarting does not solve the problem

What you expected to happen

It shouldn't be that long

How to reproduce the problem

No response

Additional context

No response

Discord username

veydlin

psychedelicious commented 6 months ago

Do all of the LoRAs show up in the app after the long wait?

VeyDlin commented 6 months ago

Yes

Also, after loading, there will be a hang every time when switching to the model manager, but it will only be a few seconds

psychedelicious commented 6 months ago

Ok, thanks. We will have to try to reproduce the issue. That many models is an unusually high number - I'm not sure when we'll have time to address this.

VeyDlin commented 5 months ago

I test a lot of models when I train a new LoRA and they accumulate easily It seems to me that this may not be a rare problem

psychedelicious commented 5 months ago

Pretty sure this is the first report, not a common problem.

Should be easy to reproduce and diagnose but the solution may be involved.

x-CK-x commented 1 month ago

Ok, thanks. We will have to try to reproduce the issue. That many models is an unusually high number - I'm not sure when we'll have time to address this.

I wonder what the approximate threshold is for the number of LoRAs & full models scanned into the program is for this problem to start appearing.

I've also noticed this issue with a large amount of models scanned into the app.