6GB aesthetic-gradients RuntimeError: CUDA out of memory.

[baphilia]

Is there an existing issue for this?

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

What happened?

tried using aesthetic-gradients extension and get RuntimeError: CUDA out of memory.

It might just require more vram than I have (6GB) and I'm out of luck, but I thought it was possible that it might just be bypassing the memory optimizations used elsewhere.

Steps to reproduce the problem

1. Have 6GB vram and a pt prepared for an aesthetic gradient
2. Go to Txt2Img
3. Press the controls to select an aesthetic gradient
4. Press generate.

What should have happened?

It generates an image.

Commit where the problem happens

737eb28faca8be2bb996ee0930ec77d1f7ebd939

What platforms do you use to access UI ?

Windows

What browsers do you use to access the UI ?

Brave

Command Line Arguments

no

Additional information, context and logs

RuntimeError: CUDA out of memory. Tried to allocate 146.00 MiB (GPU 0; 6.00 GiB total capacity; 5.17 GiB already allocated; 0 bytes free; 5.31 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

[ClashSAN]

You will need 8 gb to train/make this I believe, it is a similar requirement to textual inversion. I don't know how it plays out if you're using smaller images, however.

[ShadowPower]

https://github.com/MalumaDev/stable-diffusion-webui-aesthetic It seems that older versions work fine with 6GB vram I have used on this commit: e4f8b5f00dd33b7547cc6b76fbed26bb83b37a64

[baphilia]

https://github.com/MalumaDev/stable-diffusion-webui-aesthetic It seems that older versions work fine with 6GB vram I have used on this commit: e4f8b5f

can confirm it does indeed work with 6gb of vram. thank you. for anyone else who has an issue, I couldn't get it to run immediately after training, I had to restart the webui, and not just through the ui itself. The embeds would get a nonetype error unless I exited the webui completely and relaunched it.