How to use multi-GPU to train the model?

RQuispeC / pytorch-ACSCP

Unofficial implementation of "Crowd Counting via Adversarial Cross-Scale Consistency Pursuit" with pytorch - CVPR 2018

MIT License

18 stars 4 forks source link

How to use multi-GPU to train the model? #2

Closed fengwhu closed 5 years ago

fengwhu commented 5 years ago

Hi, I have tried to train the ACSCP model using single GPU by default, but it seemed too slow. Thus, I wonder how to use multi-GPU to train the model?

RQuispeC commented 5 years ago

hi @fengwhu , sorry for answering after so long. I have not tried using multi-GPU but this link may help you https://pytorch.org/tutorials/beginner/former_torchies/parallelism_tutorial.html

To make training faster you may want to increase --train-batch , I used 24 images per batch due to GPU capability (around 11 gb)

fengwhu commented 5 years ago

Thank you.