search

ubicomplab / rPPG-Toolbox

rPPG-Toolbox: Deep Remote PPG Toolbox (NeurIPS 2023)
https://arxiv.org/abs/2210.00716
Other
503 stars 127 forks source link

DEVICE is set to cuda: 5, but the result still runs on GPU0 #264

Closed 408550969 closed 7 months ago

408550969 commented 7 months ago

Hi,I modified the DEVICE: to cuda: 5 under UBFC-rPPG_UBFC-rPPG-PURE EFFICIENTPHYS.yaml, but from the results, it still runs on GPU0. Why is this?

yahskapar commented 7 months ago

Hi @408550969,

Can you try using CUDA_VISIBLE_DEVICES while executing your python commands, for example in your case to use just the GPU with the ID 5:

CUDA_VISIBLE_DEVICES=5 python main.py --config_file ./configs/train_configs/UBFC-rPPG_UBFC-rPPG_PURE_EFFICIENTPHYS.yaml

You can also specify multiple devices (e.g., 0,1,2,3). In the future we can try to make the config option more robust and usable the way you've tried to use it, but this may require some additional refactoring of the repo given how early that device binding needs to take place. For the time being doing what I suggested above should solve your issue.

408550969 commented 7 months ago

Thanks, it worked!