search

graphdeco-inria / gaussian-splatting

Original reference implementation of "3D Gaussian Splatting for Real-Time Radiance Field Rendering"
https://repo-sam.inria.fr/fungraph/3d-gaussian-splatting/
Other
13.52k stars 1.73k forks source link

how to choose GPU #894

Open Moreor1ess opened 1 month ago

Moreor1ess commented 1 month ago

Do you know to choose GPU? The default is to run on GPU 0, but I want it to run on GPU 1

jaco001 commented 1 month ago

for convert.py try colmap parameters in the code --SiftExtraction.use_gpu 0 --SiftMatching.use_gpu 0

for trainig.py you can try https://github.com/graphdeco-inria/gaussian-splatting?tab=readme-ov-file#--data_device --data_device cpu

Moreor1ess commented 1 month ago

Thank you for your answer

Kkun12345 commented 1 month ago

Thank you for your answer

I tried to change --data_device to cuda:1, but it still doesn't work. Is it my misunderstanding? May I ask how did you modify the GPU?

Moreor1ess commented 1 month ago

I haven't tried it specifically, but my understanding is that data_device only supports switching between CPU and GPU, but does not support selecting specific GPU serial numbers

Kkun12345 commented 1 month ago

I haven't tried it specifically, but my understanding is that data_device only supports switching between CPU and GPU, but does not support selecting specific GPU serial numbers Ok, Thank you for your answer

Kkun12345 commented 1 month ago

for convert.py try colmap parameters in the code --SiftExtraction.use_gpu 0 --SiftMatching.use_gpu 0

for trainig.py you can try https://github.com/graphdeco-inria/gaussian-splatting?tab=readme-ov-file#--data_device --data_device cpu

sorry, I tried to change --data_device to cuda:1 for train.py, but it still defaults to cuda:0. May I ask how can i modify the GPU?