Hugging Face's logo

Shitao
/
OmniGen-v1

Text-to-Image
Diffusers
Safetensors
phi3
image-to-image
Model card Files Community
16

Hope someone will reduce the hardware requirements

#5
by loojlm - opened
Discussion
loojlm

Costs more than 22G vram when generating 1024x1024 image. and I have no idea how to bring it down

Shitao
 Owner

Hi, @loojlm , we update the inference code to reduce memory cost. You can refer to https://github.com/VectorSpaceLab/OmniGen/blob/main/docs/inference.md#requiremented-resources for required resources

fullsoftwares

@loojlm

If you want to run on low VRAM

try this
https://www.youtube.com/watch?v=9ZXmXA2AJZ4

loojlm

Thanks a lot !

loojlm changed discussion status to closed

Sign up or log in to comment