lxtGH/SFSegNets

Train on a single GPU

nur-917Hub opened this issue · 1 comments

nur-917Hub commented

Hi,
I saw earlier close issue #59 but my case is a bit different.

I'd like to train SFNet but I only have 1 GPU.
Where I should tweek the code If possible using only single GPU?
-Thanks-

lxtGH commented

you may modify this line.

SFSegNets/train.py

Line 202 in 872c277

net = network.wrap_network_in_dataparallel(net, args.apex)

by not using the dataparallel.