[PyTorch小试牛刀]实战四·CNN实现逻辑回归对FashionMNIST数据集进行分类（使用GPU）

import torch as t
import torchvision as tv
import numpy as np
import time


# 超参数
EPOCH = 5
BATCH_SIZE = 100
DOWNLOAD_MNIST = True   # 下过数据的话, 就可以设置成 False
N_TEST_IMG = 10          # 到时候显示 5张图片看效果, 如上图一



class DNN(t.nn.Module):
    def __init__(self):
        super(DNN, self).__init__()

        train_data = tv.datasets.FashionMNIST(
        root="./fashionmnist/",
        train=True,
        transform=tv.transforms.ToTensor(),
        download=DOWNLOAD_MNIST
        )

        test_data = tv.datasets.FashionMNIST(
        root="./fashionmnist/",
        train=False,
        transform=tv.transforms.ToTensor(),
        download=DOWNLOAD_MNIST
        )

        print(test_data)


        # Data Loader for easy mini-batch return in training, the image batch shape will be (50, 1, 28, 28)
        self.train_loader = t.utils.data.DataLoader(
            dataset=train_data, 
            batch_size=BATCH_SIZE,
            shuffle=True)

        self.test_loader = t.utils.data.DataLoader(
            dataset=test_data, 
            batch_size=1000,
            shuffle=True)

        self.cnn = t.nn.Sequential(
            t.nn.Conv2d(
                in_channels=1,      # input height
                out_channels=32,    # n_filters
                kernel_size=5,      # filter size
                stride=1,           # filter movement/step
                padding=2,      # 如果想要 con2d 出来的图片长宽没有变化, padding=(kernel_size-1)/2 当 stride=1
            ),                  # output shape (16, 28, 28)
            t.nn.ELU(),    # activation
            t.nn.MaxPool2d(kernel_size=2), 

            t.nn.Conv2d(
                in_channels=32,      # input height
                out_channels=64,    # n_filters
                kernel_size=3,      # filter size
                stride=1,           # filter movement/step
                padding=1,      # 如果想要 con2d 出来的图片长宽没有变化, padding=(kernel_size-1)/2 当 stride=1
            ),                  # output shape (64, 14, 14)
            t.nn.ELU(),    # activation
            t.nn.MaxPool2d(kernel_size=2)  # output shape (64, 7, 7)
        )

        self.dnn = t.nn.Sequential(
            t.nn.Linear(7*7*64,256),
            t.nn.Dropout(0.5),
            t.nn.ELU(),
            t.nn.Linear(256,10),
        )

        self.lr = 0.001
        self.loss = t.nn.CrossEntropyLoss()
        self.opt = t.optim.Adam(self.parameters(), lr = self.lr)

    def forward(self,x):
        cnn1 = self.cnn(x)
        #print(cnn1.shape)
        cnn1 = cnn1.view(-1,7*7*64)
        #print(cnn1.shape)
        out = self.dnn(cnn1)
        #print(out.shape)
        return(out)

def train():
    use_gpu =  t.cuda.is_available()
    model = DNN()
    if(use_gpu):
        model.cuda()
    print(model)
    loss = model.loss
    opt = model.opt
    dataloader = model.train_loader
    testloader = model.test_loader

    
    for e in range(EPOCH):
        step = 0
        ts = time.time()
        for (x, y) in (dataloader):
            

            model.train()# train model dropout used
            step += 1
            b_x = x.view(-1,1,28,28)   # batch x, shape (batch, 28*28)
            #print(b_x.shape)
            b_y = y
            if(use_gpu):
                b_x = b_x.cuda()
                b_y = b_y.cuda()
            out = model(b_x)
            losses = loss(out,b_y)
            opt.zero_grad()
            losses.backward()
            opt.step()
            if(step%100 == 0):
                if(use_gpu):
                    print(e,step,losses.data.cpu().numpy())
                else:
                    print(e,step,losses.data.numpy())
                
                model.eval() # train model dropout not use
                for (tx,ty) in testloader:
                    t_x = tx.view(-1,1, 28,28)   # batch x, shape (batch, 28*28)
                    t_y = ty
                    if(use_gpu):
                        t_x = t_x.cuda()
                        t_y = t_y.cuda()
                    t_out = model(t_x)
                    if(use_gpu):
                        acc = (np.argmax(t_out.data.cpu().numpy(),axis=1) == t_y.data.cpu().numpy())
                    else:
                        acc = (np.argmax(t_out.data.numpy(),axis=1) == t_y.data.numpy())

                    print(time.time() - ts ,np.sum(acc)/1000)
                    ts = time.time()
                    break#只测试前1000个
            


    t.save(model, './model.pkl')  # 保存整个网络
    t.save(model.state_dict(), './model_params.pkl')   # 只保存网络中的参数 (速度快, 占内存少)
    #加载参数的方式
    """net = DNN()
    net.load_state_dict(t.load('./model_params.pkl'))
    net.eval()"""
    #加载整个模型的方式
    net = t.load('./model.pkl')
    net.cpu()
    net.eval()
    for (tx,ty) in testloader:
        t_x = tx.view(-1, 1,28,28)   # batch x, shape (batch, 28*28)
        t_y = ty

        t_out = net(t_x)
        #acc = (np.argmax(t_out.data.CPU().numpy(),axis=1) == t_y.data.CPU().numpy())
        acc = (np.argmax(t_out.data.numpy(),axis=1) == t_y.data.numpy())

        print(np.sum(acc)/1000)

if __name__ == "__main__":
    train()

DNN(
  (cnn): Sequential(
    (0): Conv2d(1, 32, kernel_size=(5, 5), stride=(1, 1), padding=(2,
2))
    (1): ELU(alpha=1.0)
    (2): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    (3): Conv2d(32, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (4): ELU(alpha=1.0)
    (5): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  )
  (dnn): Sequential(
    (0): Linear(in_features=3136, out_features=256, bias=True)
    (1): Dropout(p=0.5)
    (2): ELU(alpha=1.0)
    (3): Linear(in_features=256, out_features=10, bias=True)
  )
  (loss): CrossEntropyLoss()
)
0 100 0.5667072
2.720407485961914 0.801
0 200 0.39575616
1.7416255474090576 0.843
0 300 0.37888268
1.7285969257354736 0.862
0 400 0.40818048
1.773937702178955 0.869
0 500 0.47720864
1.7295997142791748 0.862
0 600 0.30158585
1.7265923023223877 0.867
1 100 0.27911857
1.7963228225708008 0.885
1 200 0.2902728
1.7476909160614014 0.89
1 300 0.25626943
1.8007855415344238 0.884
1 400 0.3532468
1.7679908275604248 0.871
1 500 0.27845666
1.7266316413879395 0.909
1 600 0.3446595
1.7566702365875244 0.895
0.884
0.89
0.885
0.892
0.899
0.895
0.892
0.869
0.898
0.871