search

tijiang13 / InstantAvatar

InstantAvatar: Learning Avatars from Monocular Video in 60 Seconds (CVPR 2023)
376 stars 32 forks source link

How much memory of gpu is enough to process custom video. #70

Closed bayunCC closed 7 months ago

bayunCC commented 7 months ago

Thank for your great job. I want to use custom video to run this project.But it appears error cudaSuccess(2 vs. 0) out of memory when I run openpose to estimate images.I have a gpu of 6gb, it's not enough.So I want to know how much memory of gpu is enough to process custom video.

bayunCC commented 7 months ago

The version of cuda and cudnn can't match in my ubuntu system, so I need more than 12GB memory of GPU to train video without acceleration of cudnn. So I try to use Windows system to install openpose and process my video to get correct keypoints.npy file.I prepare Windows environment:CUDA10.2 CUDNN7.6.5 OpenPosev1.7.0 visual studio2017 pycharm and cmake3.16.2.