GFPGAN icon indicating copy to clipboard operation
GFPGAN copied to clipboard

How to improve GPU-Util in NVIDIA GPU

Open kenZhangCn opened this issue 2 years ago • 2 comments

When i use GFPGAN to process same images, i find the GPU-Util (NVIDIA 3090) always under 50% (while GPU Memory Usage around 3G/24G). I want to speed up the whole process so i think maybe improve the GPU-Util is a way(or maybe there are better way please let me know)

ENV: Ubuntu 20.04.6 LTS CUDA Version: 11.4 python 3.7.17 torch===1.13.1 opencv-python==4.1.0.25 GFP paras: -v 1.4 -s 2 --only_center_face --bg_upsampler None

kenZhangCn avatar Jul 13 '23 04:07 kenZhangCn

change the for loop in inference_gfpgan.py tomultiprocessing exp : https://github.com/FXmonkey/GFPGAN-Speed

FXmonkey avatar Nov 01 '23 07:11 FXmonkey

change the for loop in inference_gfpgan.py tomultiprocessing exp : https://github.com/FXmonkey/GFPGAN-Speed

This works for me. It speeds up a lot!

Crestina2001 avatar Dec 24 '23 11:12 Crestina2001