pycaffe - Making sure if Caffe is using multiple-GPUs -
i new using training neural networks. have access gpu cluster , fine-tuning version of alex-net scene classification.
i have access 2 gpus right , want use both of them training. nvidia-smi command gives me id of gpus (which 0 , 1).
this how training use both of gpus:
caffe.set_mode_gpu()
caffe.set_device([0,1])
is right way use it?
python allows choose single gpu using set_device()
. multi-gpu supported on c++ interface. --gpu
flag used purpose discussed here. gpus used training can set --gpu
flag on command line caffe
tool. example,
build/tools/caffe train --solver=models/bvlc_alexnet/solver.prototxt --gpu=0,1
will train on gpus 0 , 1.
Comments
Post a Comment