resnet 的运行结果
step 0, training cost 1.37196e+11
step 20, training cost 117.753
step 40, training cost 36.3039
step 60, training cost 8.12384
step 80, training cost 3.16211
step 100, training cost 1.91433
step 120, training cost 1.95805
step 140, training cost 1.8385
step 160, training cost 1.81032
step 180, training cost 1.7807
step 200, training cost 1.79874
step 220, training cost 1.73078
step 240, training cost 1.7901
step 260, training cost 1.76452
step 280, training cost 1.77152
step 300, training cost 1.6973
step 320, training cost 1.6824
step 340, training cost 1.6871
step 360, training cost 1.55129
step 380, training cost 1.36596
step 400, training cost 1.34946
step 420, training cost 0.874308
step 440, training cost 1.02222
step 460, training cost 1.14761
step 480, training cost 0.347405
step 500, training cost 0.254293
step 520, training cost 0.281947
step 540, training cost 0.290539
step 560, training cost 0.578437
step 580, training cost 0.530058
step 600, training cost 0.133229
step 620, training cost 0.335239
step 640, training cost 0.609641
step 660, training cost 0.297039
step 680, training cost 0.558695
step 700, training cost 0.631659
step 720, training cost 0.387947
step 740, training cost 0.136434
step 760, training cost 0.208913
step 780, training cost 0.106486
step 800, training cost 0.0879618
step 820, training cost 0.140663
step 840, training cost 0.155591
step 860, training cost 0.0321576
step 880, training cost 0.0482099
step 900, training cost 0.153577
step 920, training cost 0.304375
step 940, training cost 0.138238
step 960, training cost 0.328786
step 980, training cost 0.0123621
training data accuracy 0.986111
test accuracy 0.833333