深度学习-torch-手写数字识别-线性降维-Sequential

深度学习-torch-手写数字识别-线性降维-Sequential

Sequential：是torch.nn自带的功能函数，实乃相见恨晚，它相当于把神经网络的各层装进一个容器，层与层之间通道数的变化也更加清晰。以基于线性降维的全连接对手写数字识别为例：
代码：

import torch
import torch.nn as nn
from torchvision import datasets
from torchvision import transforms
from torch.utils.data import DataLoader
import matplotlib.pyplot as plt
import sys
# 雷同的代码相应注释可查 “从梦到西洲” 早期博客

# 模型加载和进行数据运算可以选择GPU
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
# ---------------准备数据集--------------------
batch_size = 100
transform = transforms.Compose([transforms.ToTensor(), transforms.Normalize((0.1307,), (0.3081,))])
train_dataset = datasets.MNIST(root="../Hello/dataset/mnist", train=True, download=True, transform=transform)
test_dataset = datasets.MNIST(root="../Hello/dataset/mnist", train=False, download=True, transform=transform)
print(len(train_dataset))
print(len(test_dataset))
train_loader = DataLoader(train_dataset, shuffle=True, batch_size=batch_size)
test_loader = DataLoader(test_dataset, shuffle=False, batch_size=batch_size)
# -----------

# -----------struct model--------------
# Net是torch.nn.Module的子类
class Net(nn.Module):
    # __init__初始类中数据成员
    def __init__(self, *args):
        # super()函数的常见的用途是通过构建子类来扩展先前构造的类的功能，从而节省重写子类的这些方法的时间
        super(Net, self).__init__()
        # Sequential可以将多层整合到一起被调用
        self.model = nn.Sequential(
            nn.Linear(784, 512),
            nn.ReLU(),
            nn.Linear(512, 256),
            nn.ReLU(),
            nn.Linear(256, 128),
            nn.ReLU(),
            nn.Linear(128, 64),
            nn.ReLU(),
            nn.Linear(64, 10)
        )
    # forward是定义在每次调用时执行的计算
    def forward(self, x):
        x = x.view(-1, 784)
        x = self.model(x)
        return x


model = Net()
model.to(device)
# ------------------------------------

# -----------loss optimizer--------------
criterion = nn.CrossEntropyLoss()
optimizer = torch.optim.SGD(model.parameters(), lr=0.01)
# ------------------------------------

# -----------train test--------------
train_loss = []
def train(epoch):
    running_loss = 0.0
    for batch_idx, data in enumerate(train_loader):
        # inputs,target=data
        inputs, target = data[0].to(device), data[1].to(device)
        optimizer.zero_grad()  # 梯度清零

        outputs = model(inputs)  # 加载模型
        loss = criterion(outputs, target)  # 计算损失
        loss.backward()  # 反向传播
        optimizer.step()  # 梯度更新
        running_loss += loss  # 损失叠加

        if batch_idx % 600 == 599:
            train_loss.append((running_loss / 600).item())  # 计算每一轮的损失

            print("%d  loss: %.3f" % (epoch + 1, running_loss/ 600))

            running_loss = 0.0


acc = []
def test(epoch):
    correct = 0
    total = 0
    with torch.no_grad():
        for data in test_loader:
            # images,labels=data
            images, labels = data[0].to(device), data[1].to(device)
            ouputs = model(images)
            _, predicted = torch.max(ouputs.data, dim=1)
            correct += (predicted == labels).sum().item()
            total += labels.size(0)
    print("accurancy: %.3f %%" % (correct / total * 100))
    acc.append((correct / total) * 100)


if __name__ == '__main__':
    for epoch in range(150):
        train(epoch)
        test(epoch)
    # sys.exit(0)
    epochs = []
    for i in range(0, 150):
        epochs.append(i)

    # 刻画测试数据集准确率和轮数的关系
    plt.xlabel('epochs')
    plt.xlabel('acc')
    plt.plot(epochs, acc)
    plt.show()

    # 刻画训练损失和轮数的关系
    plt.xlabel('epochs')
    plt.xlabel('train_loss')
    plt.plot(epochs, train_loss)
    plt.show()
# ------------------------------------

刻画训练损失和轮数的关系

刻画测试数据集准确率和轮数的关系