CPUd的训练速度的确不快啊

当我们进行训练服装数据集合的时候，服装数据集有60000个训练 10000个测试

每一个的大小为28*28的

那么相关的操作时候，用cpu的确比较慢

训练和测试的数据集可以进行研究测试

Train Epoch: 9 [57600/60000 (96%)]      Loss: 0.319133
Train Epoch: 9 [58240/60000 (97%)]      Loss: 0.222361
Train Epoch: 9 [58880/60000 (98%)]      Loss: 0.129223
Train Epoch: 9 [59520/60000 (99%)]      Loss: 0.273222

Test set: Average loss: 0.2939, Accuracy: 8935/10000 (89%)

Train Epoch: 10 [0/60000 (0%)]  Loss: 0.321073
Train Epoch: 10 [640/60000 (1%)]        Loss: 0.143354
Train Epoch: 10 [1280/60000 (2%)]       Loss: 0.221344
Train Epoch: 10 [1920/60000 (3%)]       Loss: 0.237351
Train Epoch: 10 [2560/60000 (4%)]       Loss: 0.328981
Train Epoch: 10 [3200/60000 (5%)]       Loss: 0.261786
Train Epoch: 10 [3840/60000 (6%)]       Loss: 0.279668
Train Epoch: 10 [4480/60000 (7%)]       Loss: 0.269053
Train Epoch: 10 [5120/60000 (9%)]       Loss: 0.460862
Train Epoch: 10 [5760/60000 (10%)]      Loss: 0.158513
Train Epoch: 10 [6400/60000 (11%)]      Loss: 0.184277
Train Epoch: 10 [7040/60000 (12%)]      Loss: 0.150012
Train Epoch: 10 [7680/60000 (13%)]      Loss: 0.233210
Train Epoch: 10 [8320/60000 (14%)]      Loss: 0.217571
Train Epoch: 10 [8960/60000 (15%)]      Loss: 0.259001
Train Epoch: 10 [9600/60000 (16%)]      Loss: 0.330698
Train Epoch: 10 [10240/60000 (17%)]     Loss: 0.201906
Train Epoch: 10 [10880/60000 (18%)]     Loss: 0.306483
Train Epoch: 10 [11520/60000 (19%)]     Loss: 0.237035
Train Epoch: 10 [12160/60000 (20%)]     Loss: 0.144198