训练cifar100的临界点到来！（59.94分到来！）

郁闷！又遇到这种事情！要趟过去，还得下功夫！

架构改来改去，又折腾回来！

layers.emplace_back(std::make_shared<Conv2D>(cudnn, batch, 5, 32, 32, 32, 3, 1, 1));

layers.emplace_back(std::make_shared<residualExt22>(cudnn, batch, 32, 32, 32));

layers.emplace_back(std::make_shared<Conv2D>(cudnn, batch, 32, 64, 32, 32, 3, 1, 1));

layers.emplace_back(std::make_shared<residualExt22>(cudnn, batch, 64, 32, 32));

layers.emplace_back(std::make_shared<Conv2D>(cudnn, batch, 64, 128, 32, 32, 3, 1, 1));

layers.emplace_back(std::make_shared<residualExt22>(cudnn, batch, 128, 32, 32));

layers.emplace_back(std::make_shared<MaxPool2D>(cudnn, batch, 128, 32, 32, 2, 2, 0, 2));

layers.emplace_back(std::make_shared<Conv2D>(cudnn, batch, 128, 256, 16, 16, 3, 1, 1));

layers.emplace_back(std::make_shared<residualExt22>(cudnn, batch, 256, 16, 16));

layers.emplace_back(std::make_shared<MaxPool2D>(cudnn, batch, 256, 16, 16, 2, 2, 0, 2));

layers.emplace_back(std::make_shared<Conv2D>(cudnn, batch, 256, 512, 8, 8, 3, 1, 1));

layers.emplace_back(std::make_shared<residualExt22>(cudnn, batch, 512, 8, 8));//se

layers.emplace_back(std::make_shared<averPool2D>(cudnn, batch, 512, 8, 8, 2, 2, 0, 2));

/* layers.emplace_back(std::make_shared<Conv2D>(cudnn, batch, 512, 2048, 4, 4, 4));

layers.emplace_back(std::make_shared<BN>(cudnn, batch, 2048, 1, 1));

layers.emplace_back(std::make_shared<LeakyRL>(cudnn, batch, 2048, 1, 1));*/

layers.emplace_back(std::make_shared<Linear>(cublas, batch, 512*4*4, 2048));

layers.emplace_back(std::make_shared<BN>(cudnn, batch, 2048, 1, 1));

layers.emplace_back(std::make_shared<LeakyRL>(cudnn, batch, 2048, 1, 1));

layers.emplace_back(std::make_shared<Dropout>(cudnn, batch, 2048, 1, 1,0.5f));

layers.emplace_back(std::make_shared<Linear>(cublas, batch, 2048, 100));

有时候很奇怪！糟糕的运行完！改回来！什么也没变！成绩就会变好一些！你说奇怪不奇怪!遇到很多次了！就像这种临界点！上一次是cifar10的85分临界点，一直过不去！这一次又是cifar100的60分临界点！

不过，得感谢老东家富士康，他说，方法总比困难多！继续努力！

轮次：0

learn rate：0.01

时间: 67135.625000 ms

train Classification result: 0.98% ok (used 49984 images)

时间: 5302.966797 ms

Test Classification result: 1.00% ok (used 9984 images)

轮次：1

learn rate：0.01

时间: 72225.867188 ms

train Classification result: 0.95% ok (used 49984 images)

时间: 5054.175781 ms

Test Classification result: 1.00% ok (used 9984 images)

轮次：2

learn rate：0.01

时间: 73110.882813 ms

train Classification result: 4.49% ok (used 49984 images)

时间: 5095.684082 ms

Test Classification result: 1.00% ok (used 9984 images)

轮次：3

learn rate：0.01

时间: 73329.195313 ms

train Classification result: 18.48% ok (used 49984 images)

时间: 5073.086914 ms

Test Classification result: 1.00% ok (used 9984 images)

轮次：4

learn rate：0.01

时间: 73330.343750 ms

train Classification result: 30.99% ok (used 49984 images)

时间: 5118.119141 ms

Test Classification result: 1.00% ok (used 9984 images)

轮次：5

learn rate：0.001

时间: 73460.710938 ms

train Classification result: 46.02% ok (used 49984 images)

时间: 5134.735840 ms

Test Classification result: 0.99% ok (used 9984 images)

轮次：6

learn rate：0.001

时间: 73362.953125 ms

train Classification result: 50.64% ok (used 49984 images)

时间: 5149.645996 ms

Test Classification result: 1.00% ok (used 9984 images)

轮次：7

learn rate：0.001

时间: 73708.781250 ms

train Classification result: 54.03% ok (used 49984 images)

时间: 5166.069824 ms

Test Classification result: 0.99% ok (used 9984 images)

轮次：8

learn rate：0.001

时间: 73720.414063 ms

train Classification result: 57.24% ok (used 49984 images)

时间: 5168.429199 ms

Test Classification result: 1.00% ok (used 9984 images)

轮次：9

learn rate：0.001

时间: 74132.976563 ms

train Classification result: 60.24% ok (used 49984 images)

时间: 5147.817871 ms

Test Classification result: 0.99% ok (used 9984 images)

轮次：10

learn rate：0.001

时间: 73930.906250 ms

train Classification result: 63.46% ok (used 49984 images)

时间: 5145.777832 ms

Test Classification result: 1.00% ok (used 9984 images)

轮次：11

learn rate：0.001

时间: 73768.304688 ms

train Classification result: 66.78% ok (used 49984 images)

时间: 5122.628906 ms

Test Classification result: 1.00% ok (used 9984 images)

轮次：12

learn rate：0.001

时间: 73637.054688 ms

train Classification result: 70.58% ok (used 49984 images)

时间: 5121.869141 ms

Test Classification result: 1.00% ok (used 9984 images)

轮次：13

learn rate：0.001

时间: 73705.875000 ms

train Classification result: 74.35% ok (used 49984 images)

时间: 5142.958008 ms

Test Classification result: 1.00% ok (used 9984 images)

轮次：14

learn rate：0.001

时间: 73906.039063 ms

train Classification result: 78.62% ok (used 49984 images)

时间: 5146.979004 ms

Test Classification result: 1.00% ok (used 9984 images)

轮次：15

learn rate：0.001

时间: 73742.007813 ms

train Classification result: 82.90% ok (used 49984 images)

时间: 5133.905762 ms

Test Classification result: 1.00% ok (used 9984 images)

轮次：16

learn rate：0.001

时间: 73956.820313 ms

train Classification result: 86.78% ok (used 49984 images)

时间: 5128.584961 ms

Test Classification result: 1.00% ok (used 9984 images)

轮次：17

learn rate：0.001

时间: 73924.054688 ms

train Classification result: 89.96% ok (used 49984 images)

时间: 5134.266113 ms

Test Classification result: 1.00% ok (used 9984 images)

轮次：18

learn rate：0.0001

时间: 73813.054688 ms

train Classification result: 94.39% ok (used 49984 images)

时间: 5166.653809 ms

Test Classification result: 58.34% ok (used 9984 images)

轮次：19

learn rate：0.0001

时间: 73980.843750 ms

train Classification result: 96.28% ok (used 49984 images)

时间: 5176.334961 ms

Test Classification result: 57.37% ok (used 9984 images)

轮次：20

learn rate：1e-05

时间: 73717.640625 ms

train Classification result: 96.88% ok (used 49984 images)

时间: 5185.070801 ms

Test Classification result: 59.84% ok (used 9984 images)

轮次：21

learn rate：1e-05

时间: 73662.054688 ms

train Classification result: 97.10% ok (used 49984 images)

时间: 5188.983887 ms

Test Classification result: 59.92% ok (used 9984 images)

轮次：22

learn rate：1e-06

时间: 73708.179688 ms

train Classification result: 97.07% ok (used 49984 images)

时间: 5166.612793 ms

Test Classification result: 59.88% ok (used 9984 images)

轮次：23

learn rate：1e-06

时间: 73865.960938 ms

train Classification result: 96.97% ok (used 49984 images)

时间: 5240.098145 ms

Test Classification result: 59.89% ok (used 9984 images)

轮次：24

learn rate：1e-07

时间: 73640.687500 ms

train Classification result: 97.17% ok (used 49984 images)

时间: 5167.064941 ms

Test Classification result: 59.94% ok (used 9984 images)

轮次：25

learn rate：1e-07

时间: 73854.000000 ms

train Classification result: 97.40% ok (used 49984 images)

时间: 5224.902832 ms

Test Classification result: 59.88% ok (used 9984 images)

请按任意键继续. . .