RECCE icon indicating copy to clipboard operation
RECCE copied to clipboard
verified publisher icon VISION-SJTU

How to run on multiple GPUs?

Open CharlesTran opened this issue 1 year ago • 1 comments

CharlesTran avatar Aug 06 '24 07:08 CharlesTran

Hi, it has been a long time since the code deployment, and I currently do not have a similar environment for this project. However, I think you may try the following command first. Please also refer to PyTorch Distributed Overview for multi-GPU training.

CUDA_VISIBLE_DEVICES=0,1 python -m torch.distributed.launch --nproc_per_node=2 --master_port 12345 train.py --config path/to/config.yaml

XJay18 avatar Aug 07 '24 13:08 XJay18

Close due to inactivation. Please feel free to reopen this issue if you still have related problems.

XJay18 avatar Oct 16 '24 14:10 XJay18