Pytorch深度学习之简单前馈神经网络

这里还是使用MNIST数据集进行处理，直接上代码（不懂看注释~）

#导入相应包
import torch
import torch.nn as nn
import torchvision
import torchvision.transforms as transform
#定义cuda加速
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
#定义输入维度、隐藏层维度、输出维度、训练次数、batch大小、学习率
input_size = 28 * 28
hidden_size = 500
num_classes = 10
num_epochs = 5
batch_size = 100
learning_rate = 0.001
#定义训练集测试集并转为DataLoader格式
train_dataset = torchvision.datasets.MNIST(root='data/',
                                           train=True,
                                           transform=transform.ToTensor(),
                                           download=True)
test_dataset = torchvision.datasets.MNIST(root='data/',
                                          train=False,
                                          transform=transform.ToTensor())

train_loader = torch.utils.data.DataLoader(dataset=train_dataset,
                                           batch_size=batch_size,
                                           shuffle=True)
test_loader = torch.utils.data.DataLoader(dataset=test_dataset,
                                          batch_size=batch_size,
                                          shuffle=False)
#定义模型
class NeuralNet(nn.Module):
	#定义初始化以及内部计算函数
    def __init__(self, input_size, hidden_size, num_classes):
        super(NeuralNet, self).__init__()
        self.fc1 = nn.Linear(input_size, hidden_size)
        self.relu = nn.ReLU()
        self.fc2 = nn.Linear(hidden_size, num_classes)
    #使用初始化函数中进行前馈
    def forward(self, x):
        out = self.fc1(x)
        out = self.relu(out)
        out = self.fc2(out)
        return out
#实例化模型对象
model = NeuralNet(input_size, hidden_size, num_classes).to(device)
#定义损失函数和优化器
criterion = nn.CrossEntropyLoss()
optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate)

total_step = len(train_loader)
for epoch in range(num_epochs):
	#遍历数据集进行训练
    for i, (images, labels) in enumerate(train_loader):
        images = images.reshape(-1, 28 * 28).to(device)
        labels = labels.to(device)
        outputs = model(images)
        loss = criterion(outputs, labels)
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()
        if((i+1) % 100 == 0):
            print('Epoch [{}/{}], Step [{}/{}], Loss: {:.4f}'.format(epoch, num_epochs, i+1, total_step, loss.item()))

#设置无梯度模式进行测试查看精度
with torch.no_grad():
    correct = 0
    total = 0
    for images, labels in test_loader:
        images = images.reshape(-1, 28 * 28).to(device)
        labels = labels.to(device)
        outputs = model(images)
        _, predict = torch.max(outputs.data, 1)
        total += labels.shape[0]
        correct += (predict == labels).sum().item()
    print('Accuracy of the network on the 10000 test images: {} %'.format(100 * correct / total))

torch.save(model.state_dict(), 'model.ckpt')

输出如下：

Epoch [0/5], Step [100/600], Loss: 0.1658
Epoch [0/5], Step [200/600], Loss: 0.3905
Epoch [0/5], Step [300/600], Loss: 0.1464
Epoch [0/5], Step [400/600], Loss: 0.2937
Epoch [0/5], Step [500/600], Loss: 0.2443
Epoch [0/5], Step [600/600], Loss: 0.0820
Epoch [1/5], Step [100/600], Loss: 0.1177
Epoch [1/5], Step [200/600], Loss: 0.1299
Epoch [1/5], Step [300/600], Loss: 0.1143
Epoch [1/5], Step [400/600], Loss: 0.0830
Epoch [1/5], Step [500/600], Loss: 0.1383
Epoch [1/5], Step [600/600], Loss: 0.0745
Epoch [2/5], Step [100/600], Loss: 0.1163
Epoch [2/5], Step [200/600], Loss: 0.0630
Epoch [2/5], Step [300/600], Loss: 0.0738
Epoch [2/5], Step [400/600], Loss: 0.0474
Epoch [2/5], Step [500/600], Loss: 0.0874
Epoch [2/5], Step [600/600], Loss: 0.1105
Epoch [3/5], Step [100/600], Loss: 0.0503
Epoch [3/5], Step [200/600], Loss: 0.0636
Epoch [3/5], Step [300/600], Loss: 0.0599
Epoch [3/5], Step [400/600], Loss: 0.0385
Epoch [3/5], Step [500/600], Loss: 0.0644
Epoch [3/5], Step [600/600], Loss: 0.0296
Epoch [4/5], Step [100/600], Loss: 0.0091
Epoch [4/5], Step [200/600], Loss: 0.0263
Epoch [4/5], Step [300/600], Loss: 0.0255
Epoch [4/5], Step [400/600], Loss: 0.0307
Epoch [4/5], Step [500/600], Loss: 0.0537
Epoch [4/5], Step [600/600], Loss: 0.0589
Accuracy of the network on the 10000 test images: 98.03 %

可以看到精度达到98%，而且我们还是比较简单的初级模型，并未对参数进行寻优，且为对数据进行处理。