search

wenbowen123 / iros20-6d-pose-tracking

[IROS 2020] se(3)-TrackNet: Data-driven 6D Pose Tracking by Calibrating Image Residuals in Synthetic Domains
Other
388 stars 67 forks source link

about multiply GPU training #7

Closed baojun701 closed 3 years ago

baojun701 commented 3 years ago

I modify the CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7, in the train.sh, but it always training by only one GPU, how can I train by multiply GPU?

wenbowen123 commented 3 years ago

The current code does not support multi-gpu. It needs to be modified for that purpose. However, multi-gpu is not recommended here since currently the batchsize is already around 200. If you want to train faster you can reduce the epochs, e.g. change to 150 or 200. The performance should not hurt much.