Is Inference speed margin reasonable?

Open Goddard-Wen opened this issue 8 months ago • 0 comments

I test model on AMD R5 7500F and Nvidia 4070TiS. I run run_demo.py and added code to record inference speed, I get frames per second(fps) only has 5 to 8, it much less than author said in paper(~32Hz).
However, the running device is different, but this big margin is that reasonable?

Apr 27 '25 07:04 Goddard-Wen