Chainer 3.2.0 と imagenet その３

前回の V100 や GeForce GTX 1080 Ti などと比べると極めて遅いです（当然でしょうが）。。。

◯CPU Intel(R) Xeon(R) CPU E5-2670 v3 @ 2.30GHz x 2個 + NVIDIA Quadro K420 1個
$ time python ../imagenet/train_imagenet.py -a alex -g 0 -E 50 train.txt test.txt
epoch iteration main/loss validation/main/loss main/accuracy validation/main/accuracy lr
4 1000 3.40144 0.277312 0.01
9 2000 1.99988 0.510063 0.01
13 3000 1.33209 0.647812 0.01
18 4000 0.883646 0.757375 0.01
23 5000 0.599879 0.827563 0.01
27 6000 0.43296 0.875969 0.01
32 7000 0.329996 0.906187 0.01
37 8000 0.262041 0.924563 0.01
41 9000 0.222094 0.938438 0.01
46 10000 0.197076 0.944594 0.01

real 470m9.974s
user 227m48.250s
sys 284m33.631s

◯CPU Intel(R) Xeon(R) CPU E5-2670 v3 @ 2.30GHz x 2個
# time python ../imagenet/train_imagenet.py -a alex -g -1 -E 50 train.txt test.txt

epoch iteration main/loss validation/main/loss main/accuracy validation/main/accuracy lr
4 1000 3.36402 0.285094 0.01
9 2000 1.95647 0.519844 0.01
13 3000 1.29434 0.657719 0.01
18 4000 0.844572 0.766125 0.01
23 5000 0.561211 0.842094 0.01
27 6000 0.413539 0.8815 0.01
32 7000 0.312991 0.910937 0.01
37 8000 0.254715 0.927937 0.01
41 9000 0.19088 0.945312 0.01
46 10000 0.181867 0.950031 0.01

real 1032m4.749s
user 3634m40.588s
sys 6542m35.312s