stable-diffusion-webui icon indicating copy to clipboard operation
stable-diffusion-webui copied to clipboard

2GB Nvidia Geforce GT 1030 (2GB varient) not working

Open Seagate-1TB opened this issue 2 years ago • 6 comments

Is there an existing issue for this?

  • [X] I have searched the existing issues and checked the recent builds/commits

What happened?

I use Windows 10 x64, and I use an chrome varient for my Browser, and I can't generate art, here's the error, so how do I fix it? error

Steps to reproduce the problem

  1. Use any 2GB GPU from nvidia, or their partners
  2. run SD
  3. try to generate art
  4. error pops up

What should have happened?

I think torch needs to be fixed or something

Commit where the problem happens

https://github.com/AUTOMATIC1111/stable-diffusion-webui/commit/22bcc7be428c94e9408f589966c2040187245d81

What Python version are you running on ?

Python 3.10.x

What platforms do you use to access the UI ?

Windows

What device are you running WebUI on?

Other GPUs

What browsers do you use to access the UI ?

Google Chrome

Command Line Arguments

@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS= --lowvram --always-batch-cond-uncond --precision full --no-half

call webui.bat

List of extensions

No

Console logs

The console log wasn't copied, but there is an screenshot

Additional information

This is a torch issue, I think

Seagate-1TB avatar Jun 15 '23 20:06 Seagate-1TB

I decided to switch from--precision full, to --precision autocast

Seagate-1TB avatar Jun 15 '23 20:06 Seagate-1TB

What model are you using? I think you need at least 4 GB of VRAM or you can try:

set COMMANDLINE_ARGS=--use-cpu all

chrme avatar Jun 16 '23 12:06 chrme

Consider using Stable Diffusion on a Google Colab or some other online resource. A GT 1030 is just plain insufficient. Even if you do get the program to start, you'll be waiting forever to get your output.

CUDA out of memory means exactly what it says - it doesn't have enough memory available for everything it's trying to do.

steamrick avatar Jun 16 '23 12:06 steamrick

This is a torch issue, I think

No, this is an insufficient hardware issue.

Munichschntz avatar Jun 19 '23 03:06 Munichschntz

Can you people stop contaminating and bloating the repo's Issues section with this garbage? There are actual significant bugs and interesting feature requests that the devs would rather focus on. The Issues page has been a complete mess lately, one guy's trying to run on a 1030, another's trying to run without Git, just do some basic research, FFS.

highnrgappalachian avatar Jun 19 '23 11:06 highnrgappalachian

The Issues page has been a complete mess lately, one guy's trying to run on a 1030, another's trying to run without Git, just do some basic research, FFS. But everyone NEEDs a real AI running on their machine so they can have it trained enough to do your dishes and math homework by fall and be able to play CS:GO while it's cranking out the artificial intelligences. :P

Wait for the next exciting report where the card has been migrated to a 6GB model that was somehow located, but now "files start loading and computer locks up for a very long time then resets" where we'll find out about the 4GB of system RAM. "But that's why I'm using 64-bit Windows, so it can have more". Fix your code.

NeedsMoar avatar Jun 22 '23 17:06 NeedsMoar

Yeah, pretty sure you can't run SD properly on 2 GB, lowvram hacks or not. Sorry.

akx avatar Jun 30 '23 07:06 akx