[Bug]: text2img error

[flyonch]

Is there an existing issue for this?

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

What happened?

when evaluating text2img in more than 1000x1000 type errors rtx 3070 ti.

while my friend renders quietly at 1500x1500 on rtx 3060

Steps to reproduce the problem

1. Go to ....
2. Press ....
3. ...

What should have happened?

Commit where the problem happens

What platforms do you use to access the UI ?

No response

What browsers do you use to access the UI ?

No response

Command Line Arguments

-

List of extensions

Console logs

torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 3.75 GiB (GPU 0; 8.00 GiB total capacity; 5.89 GiB already allocated; 0 bytes free; 6.00 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation.  See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF

Additional information

No response

[vidiotgameboss]

same issue as you since a couple days ago #8571 used to be able to generate 1080p on the same settings then suddenly getting OOM errors

[Xyem]

Could you try using --opt-split-attention-v1 to see if that mitigates the issue?

[nhari999]

Maybe this error depend on the VRAM . Check the size and the usage of your VRAM and to optimize the image size and avoid memory errors use "--medvram" : This option allocates memory in a way that optimizes the usage of your VRAM.

But if you get an out of memory error with" --medvram" ,use" --lowvram --always-batch-cond-uncond" instead: This suggests that you are getting out of memory errors when using the --medvram option ,and this option further reduces memory usage and ensures that the batch is always processed.

Or if you want to generate images larger than you can with "--medvram" relative to the size of your VRAM , you should switch to the "--lowvram " option to further reduce memory usage.

Try those one of this 3 points to see if that mitigates the issue ?

[arshtepe]

I started to get this error when I try to teach dreambooth, but I have rtx 3090 with 24 gb VRAM and it worked before

[lozn00]

@Xyem @nhari999 how to config param ,i just know use webui.bat

[nhari999]

@Xyem @nhari999 how to config param ,i just know use webui.bat

To config param and put them into your webUi user bat : go to webUI-user.bat file , right click and then click on edit ; the appearance will be like this in the text editor

  @echo off

   set PYTHON="C:\Users\MSI pulse\AppData\Local\Programs\Python\Python310\python.exe"
   set GIT=
   set VENV_DIR=
   set COMMANDLINE_ARGS=
  git pull
  call webui.bat  

Then put the command behind the "set COMMANDLINE_ARGS=" like this "set COMMANDLINE_ARGS= --medvram" and then save and restart the stable diffusion again

[lozn00]

@nhari999 thank you

[flyonch]

torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 3.64 GiB (GPU 0; 8.00 GiB total capacity; 5.34 GiB already allocated; 571.00 MiB free; 5.39 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF

torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 19.02 GiB (GPU 0; 8.00 GiB total capacity; 1.81 GiB already allocated; 4.08 GiB free; 1.87 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF

return _VF.einsum(equation, operands)  # type: ignore[attr-defined]

torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 7.28 GiB (GPU 0; 8.00 GiB total capacity; 189.80 MiB already allocated; 5.73 GiB free; 224.00 MiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF

1000х1000