Chainer 3.2.0 と imagenet

やっぱり Tesla V100 や GeForce GTX 1080 Ti は速い。。。

◯ Tesla V100 PCIe
$ time python ../imagenet/train_imagenet.py -g 0 -E 18 train.txt test.txt
epoch iteration main/loss validation/main/loss main/accuracy validation/main/accuracy lr
4 1000 3.43079 0.259062 0.01
9 2000 2.33857 0.428969 0.01
13 3000 1.69085 0.558969 0.01

real 2m52.302s
user 16m11.994s
sys 0m49.653s

◯ GeForce GTX 1080 Ti
$ time python ../imagenet/train_imagenet.py -g 1 -E 18 train.txt test.txt
epoch iteration main/loss validation/main/loss main/accuracy validation/main/accuracy lr
4 1000 3.46599 0.254469 0.01
9 2000 2.37635 0.42025 0.01
13 3000 1.74475 0.546969 0.01

real 2m31.105s
user 13m37.571s
sys 0m35.873s

◯ Tesla K40m
$ time python ../imagenet/train_imagenet.py -g 1 -E 18 train.txt test.txt
epoch iteration main/loss validation/main/loss main/accuracy validation/main/accuracy lr
4 1000 3.49999 0.246375 0.01
9 2000 2.36371 0.418094 0.01
13 3000 1.76144 0.541906 0.01

real 11m7.945s
user 22m9.749s
sys 4m7.984s