criterion = torch.nn.CrossEntropyLoss()

transform = transforms.Compose([ transforms.ToTensor(), transforms.Normalize((0.1307, ), (0.3081, )) ])

transforms.Normalize((0.1307, ), (0.3081, ))

import torch
from torchvision import transforms
from torchvision import datasets
from torch.utils.data import DataLoader
import torch.nn.functional as F
import torch.optim as optim
 
# prepare dataset
 
batch_size = 64
 
transform = transforms.Compose([
    transforms.ToTensor(), #先將圖像變換成一個(gè)張量tensor。
    transforms.Normalize((0.1307,), (0.3081,))
    #其中的0.1307是MNIST數(shù)據(jù)集的均值，0.3081是MNIST數(shù)據(jù)集的標(biāo)準(zhǔn)差。
])  # 歸一化,均值和方差
 
train_dataset = datasets.MNIST(root='../dataset/mnist/', train=True,
                               download=True, transform=transform)
train_loader = DataLoader(train_dataset, shuffle=True, batch_size=batch_size)
 
test_dataset = datasets.MNIST(root='../dataset/mnist/', train=False,
                               download=True, transform=transform)
test_loader = DataLoader(test_dataset, shuffle=False, batch_size=batch_size)
 
# design model using class
class Net(torch.nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.l1 = torch.nn.Linear(784, 512)
        self.l2 = torch.nn.Linear(512, 256)
        self.l3 = torch.nn.Linear(256, 128)
        self.l4 = torch.nn.Linear(128, 64)
        self.l5 = torch.nn.Linear(64, 10)
 
    def forward(self, x):
        # 28 * 28 = 784
        # 784 = 28 * 28,即將N *1*28*28轉(zhuǎn)化成 N *1*784
        x = x.view(-1, 784)  # -1其實(shí)就是自動(dòng)獲取mini_batch
        x = F.relu(self.l1(x))
        x = F.relu(self.l2(x))
        x = F.relu(self.l3(x))
        x = F.relu(self.l4(x))
        return self.l5(x)  # 最后一層不做激活，不進(jìn)行非線性變換
 
model = Net()
 
#CrossEntropyLoss <==> LogSoftmax + NLLLoss。
#也就是說(shuō)使用CrossEntropyLoss最后一層(線性層)是不需要做其他變化的；
#使用NLLLoss之前，需要對(duì)最后一層(線性層)先進(jìn)行SoftMax處理，再進(jìn)行l(wèi)og操作。
 
# construct loss and optimizer
criterion = torch.nn.CrossEntropyLoss()
#momentum 是帶有優(yōu)化的一個(gè)訓(xùn)練過(guò)程參數(shù)
optimizer = optim.SGD(model.parameters(), lr=0.01, momentum=0.5)
 
# training cycle forward, backward, update
 
def train(epoch):
    running_loss = 0.0
    #enumerate()函數(shù)用于將一個(gè)可遍歷的數(shù)據(jù)對(duì)象(如列表、元組或字符串)組合為一個(gè)索引序列，
    #同時(shí)列出數(shù)據(jù)和數(shù)據(jù)下標(biāo)，一般用在 for 循環(huán)當(dāng)中。
    #enumerate(sequence, [start=0])
    for batch_idx, data in enumerate(train_loader, 0):
        # 獲得一個(gè)批次的數(shù)據(jù)和標(biāo)簽
        inputs, target = data
        optimizer.zero_grad()
 
        #forward + backward + update
        # 獲得模型預(yù)測(cè)結(jié)果(64, 10)
        outputs = model(inputs)
        # 交叉熵代價(jià)函數(shù)outputs(64,10),target（64）
        loss = criterion(outputs, target)
        loss.backward()
        optimizer.step()
 
        running_loss += loss.item()
        if batch_idx % 300 == 299:
            print('[%d, %5d] loss: %.3f' % (epoch + 1, batch_idx + 1, running_loss / 300))
            running_loss = 0.0
 
def test():
    correct = 0
    total = 0
    with torch.no_grad():#不需要計(jì)算梯度。
        for data in test_loader:
            images, labels = data
            outputs = model(images)
            #orch.max的返回值有兩個(gè)，第一個(gè)是每一行的最大值是多少，第二個(gè)是每一行最大值的下標(biāo)(索引)是多少。
            _, predicted = torch.max(outputs.data, dim=1)  # dim = 1 列是第0個(gè)維度，行是第1個(gè)維度
            total += labels.size(0)
            correct += (predicted == labels).sum().item()  # 張量之間的比較運(yùn)算
    print('accuracy on test set: %d %% ' % (100 * correct / total))
 
if __name__ == '__main__':
    for epoch in range(10):
        train(epoch)
        test()