27_pytorch全连接层，使用MNIST的分类案例（学习笔记）

# -*- coding: UTF-8 -*-

import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
from torchvision import datasets, transforms

batch_size=200
learning_rate=0.01
epochs=10

# torchvision.transforms.Compose()类。这个类的主要作用是串联多个图片变换的操作。
# (1) transforms.Compose就是将transforms组合在一起
# (2) transforms.Normalize使用如下公式进行归一化
# (3) torchvision.transforms.ToTensor() 起到的作用是把PIL.Image或者numpy.narray数据类型转变为torch.FloatTensor类型，shape是C*H*W，数值范围缩小为[0.0, 1.0]
# (4) 如果想把数值范围调整为[-1.0, 1.0],则可加torchvision.transforms.Normalize([mean_channel1,mean_channel2,mean_channel3],[std_channel1,std_channel2,std_channel3])
train_loader = torch.utils.data.DataLoader(
    datasets.MNIST('../data', train=True, download=True,
                   transform=transforms.Compose([
                       transforms.ToTensor(),
                       transforms.Normalize((0.1307,), (0.3081,))
                   ])),
    batch_size=batch_size, shuffle=True
)

test_loader = torch.utils.data.DataLoader(
    datasets.MNIST('../data', train=False,
                   transform=transforms.Compose([
                       transforms.ToTensor(),
                       transforms.Normalize((0.1307,), (0.3081,))
                   ])),
    batch_size=batch_size, shuffle=True
)


# 继承自nn.Module
class MLP(nn.Module):
    def __init__(self):
        # 初始化层在 __init__中
        super(MLP,self).__init__()

        self.model = nn.Sequential(
            nn.Linear(784, 200),
            # 使用ReLU激活函数，关于22个激活函数的，可以参考：https://blog.csdn.net/tototuzuoquan/article/details/113791252?spm=1001.2014.3001.5501
            nn.ReLU(inplace=True),
            nn.Linear(200, 200),
            nn.ReLU(inplace=True),
            nn.Linear(200, 10),
            nn.ReLU(inplace=True),
        )

    # 要实现forward()
    def forward(self, x):
        x = self.model(x)

        return x


net = MLP()
# PyTorch的十个优化器：https://blog.csdn.net/tototuzuoquan/article/details/113779970?spm=1001.2014.3001.5501
# 另外一篇关于优化器的文章（这里面有优化器的特征、数学公式等）：https://blog.csdn.net/tototuzuoquan/article/details/112724028?spm=1001.2014.3001.5501
optimizer = optim.SGD(net.parameters(), lr=learning_rate)
# 关于PyTorch的十九个损失函数的文章有：https://blog.csdn.net/tototuzuoquan/article/details/113777788?spm=1001.2014.3001.5501
criteon = nn.CrossEntropyLoss()


# enumerate() 函数用于将一个可遍历的数据对象(如列表、元组或字符串)组合为一个索引序列，同时列出数据和数据下标，一般用在 for 循环当中。
# enumerate() 方法的语法:
# 参数：
# sequence -- 一个序列、迭代器或其他支持迭代对象
# start -- 下标起始位置


for epoch in range(epochs):
    for batch_idx, (data, target) in enumerate(train_loader):
        data = data.view(-1, 28*28)

        logits = net(data)
        loss = criteon(logits, target)

        optimizer.zero_grad()
        loss.backward()
        # print(w1.grad.norm(), w2.grad.norm())
        optimizer.step()

        if batch_idx % 100 == 0:
            print('Train Epoch: {} [{}/{} ({:.0f}%)]\tLoss: {:.6f}'.format(
                epoch, batch_idx * len(data), len(train_loader.dataset),
                100. * batch_idx / len(train_loader), loss.item()
            ))

    test_loss = 0
    correct = 0
    for data, target in test_loader:
        data = data.view(-1, 28 * 28)
        logits = net(data)
        test_loss += criteon(logits, target).item()

        pred = logits.data.max(1)[1]
        correct += pred.eq(target.data).sum()

    test_loss /= len(test_loader.dataset)
    print('\nTest set: Average loss: {:.4f}, Accuracy: {}/{} ({:.0f}%)\n'.format(
        test_loss, correct, len(test_loader.dataset),
        100. * correct / len(test_loader.dataset)))

输出结果：

runfile('E:/workspace/pytorch-learn/27_MLP网络层/main.py', wdir='E:/workspace/pytorch-learn/27_MLP网络层')
Train Epoch: 0 [0/60000 (0%)]	Loss: 2.311366
Train Epoch: 0 [20000/60000 (33%)]	Loss: 2.019119
Train Epoch: 0 [40000/60000 (67%)]	Loss: 1.321077
Test set: Average loss: 0.0040, Accuracy: 8394/10000 (84%)
Train Epoch: 1 [0/60000 (0%)]	Loss: 0.845444
Train Epoch: 1 [20000/60000 (33%)]	Loss: 0.593395
Train Epoch: 1 [40000/60000 (67%)]	Loss: 0.391533
Test set: Average loss: 0.0020, Accuracy: 8918/10000 (89%)
Train Epoch: 2 [0/60000 (0%)]	Loss: 0.489720
Train Epoch: 2 [20000/60000 (33%)]	Loss: 0.439235
Train Epoch: 2 [40000/60000 (67%)]	Loss: 0.424977
Test set: Average loss: 0.0017, Accuracy: 9050/10000 (90%)
Train Epoch: 3 [0/60000 (0%)]	Loss: 0.354985
Train Epoch: 3 [20000/60000 (33%)]	Loss: 0.367654
Train Epoch: 3 [40000/60000 (67%)]	Loss: 0.279793
Test set: Average loss: 0.0015, Accuracy: 9133/10000 (91%)
Train Epoch: 4 [0/60000 (0%)]	Loss: 0.241317
Train Epoch: 4 [20000/60000 (33%)]	Loss: 0.336772
Train Epoch: 4 [40000/60000 (67%)]	Loss: 0.314246
Test set: Average loss: 0.0014, Accuracy: 9195/10000 (92%)
Train Epoch: 5 [0/60000 (0%)]	Loss: 0.380214
Train Epoch: 5 [20000/60000 (33%)]	Loss: 0.184812
Train Epoch: 5 [40000/60000 (67%)]	Loss: 0.316628
Test set: Average loss: 0.0013, Accuracy: 9237/10000 (92%)
Train Epoch: 6 [0/60000 (0%)]	Loss: 0.331256
Train Epoch: 6 [20000/60000 (33%)]	Loss: 0.288215
Train Epoch: 6 [40000/60000 (67%)]	Loss: 0.228315
Test set: Average loss: 0.0012, Accuracy: 9277/10000 (93%)
Train Epoch: 7 [0/60000 (0%)]	Loss: 0.312616
Train Epoch: 7 [20000/60000 (33%)]	Loss: 0.219588
Train Epoch: 7 [40000/60000 (67%)]	Loss: 0.294207
Test set: Average loss: 0.0012, Accuracy: 9298/10000 (93%)
Train Epoch: 8 [0/60000 (0%)]	Loss: 0.302778
Train Epoch: 8 [20000/60000 (33%)]	Loss: 0.227025
Train Epoch: 8 [40000/60000 (67%)]	Loss: 0.212254
Test set: Average loss: 0.0011, Accuracy: 9333/10000 (93%)
Train Epoch: 9 [0/60000 (0%)]	Loss: 0.178860
Train Epoch: 9 [20000/60000 (33%)]	Loss: 0.232927
Train Epoch: 9 [40000/60000 (67%)]	Loss: 0.224159
Test set: Average loss: 0.0011, Accuracy: 9384/10000 (94%)