Omost patch 1 - for 8GB VRAM GPU - 1920x1080

[xhoxye]

The inspiration came from here https://github.com/lllyasviel/Omost/pull/27, but I couldn't run it, so I modified a version myself https://github.com/lllyasviel/Omost/issues/63

1. Seed randomization via -1, max int32 2 ** 31 - 1
2. Use the drop-down menu to select LLM Model and Image Diffusion Model, and the Image Diffusion Model list reads the Omost\models\checkpoints path
3. Save the image to the outputs directory and format the image name
4. Print the loaded model name, seed value, dialogue and image rendering time in cmd, and other information
5. Load the .safetensors file using the fp16 type

My pc, windows 10, 2070s-8GB vram, 64G ram, CPU 5600X

[xhoxye]

如果你在国内，也可以使用我的中文版本，https://github.com/xhoxye/Omost_CN

[ginto-sakata]

Can you explain what your patch doing? Will it work on 8Gb cards?

[xhoxye]

@ginto-sakata Yes, it works fine on my 2070S-8GB vram gpu fp16 model_pipeline = StableDiffusionXLImg2ImgPipeline.from_single_file(model_path, torch_dtype=torch.float16, variant="fp16")

[xhoxye]

1920x1080 no problem

[qwyan]

你好，怎么联系呢？

[xhoxye]

@qwyan 有什么问题在这里问 有其他事情可以私信，或者到我的分支上问

[ping3839]

如果你在国内，也可以使用我的中文版本，https://github.com/xhoxye/Omost_CN

我在国外也一样用