rife-ncnn-vulkan icon indicating copy to clipboard operation
rife-ncnn-vulkan copied to clipboard

ncnn-vulkan time performance

Open liye0713 opened this issue 4 years ago • 1 comments

您好。我用了这个ncnn 的推理框架,在 nvidia-T4上测试发现,它相比于 pytorch 实现的rife, 速度慢了很了。推理时间几乎达到了pytorch 版本的 3-5倍左右。请问这个原因是什么呀?是不是在云上ncnn 的框架相比于pytorch 没有速度优势呢?

liye0713 avatar Jul 09 '21 02:07 liye0713

I noticed this too! The python version i used was almost twice as fast as this version. For reference, the python version im using is here: https://github.com/hzwer/arXiv2020-RIFE

KCGD avatar Jan 11 '22 18:01 KCGD