import torch
import torch.nn as nn
from torch.nn import functional as F

class CnnNet(nn.Module):
    def __init__(self):
        super(CnnNet, self).__init__()
        # images input channle = 1
        # output channel = 6
        # conv kernel 5x5
        self.conv1 = nn.Conv2d(1, 6, 5)
        self.conv2 = nn.Conv2d(6, 16, 5)
        # affine operator for: y = Wx + b
        self.fc1 = nn.Linear(16 * 5 * 5, 120)
        self.fc2 = nn.Linear(120, 84)
        self.fc3 = nn.Linear(84, 10)

    def forward(self, input):
        # max_pooling window size: 2x2
        # 经过relu激活函数，再经过max_pooling池化
        input = self.conv1(input)
        input = F.relu(input)
        input = F.max_pool2d(input, (2, 2))

        # If the size is a square you can only specify a single number for pool window size
        input = self.conv2(input)
        input = F.relu(input)
        input = F.max_pool2d(input, 2)

        # 铺平
        input = input.view(-1, self.flat_feature(input))

        input = self.fc1(input)
        input = F.relu(input)

        input = self.fc2(input)
        input = F.relu(input)
        input = self.self(input)
        return input


    def flat_feature(self, input):
        sizes = input.size[1:]   # all dimensions except the batch dimension
        num_feat = 1
        for size in sizes:
            num_feat *= size
        return num_feat

params = list(cnn_net.parameters())
print(len(params))  # expected 5: conv1, conv2, fc1, fc2, fc3
print(params[0].size())  # conv1's .weight

input = torch.randn(1, 1, 32, 32)
out = cnn_net(input)

# clear grad
cnn_net.zero_grad()

# 调用反向传播，括号里面的是贡献系数
out.backward(torch.randn(1, 10))

output = cnn_net(input)
target = torch.randn(10)    # dumy target, actual target need to be obtained from the dataset
# make it the same shape as output
target = target.view(1, -1)

loss = nn.MSELoss(output, target)
print(loss)

# 在反向传播之前清空梯度
cnn_net.zero_grad()

print('conv1.bias.grad before backward')
print(net.conv1.bias.grad)

loss.backward()

print('conv1.bias.grad after backward')
print(net.conv1.bias.grad)

import torch.optim as optim
optimizer = optim.SGD(cnn_net.parameters(), lr = 0.01)

# 清空梯度
optimizer.zero_grad()
output = cnn_net(input)
loss = nn.MSELoss(output, target)
loss.backward()

# 更新权重
optimizer.step()

from torchvision import datasets, transforms
import torch

transform = transforms.Compose([
    transforms.ToTensor(),  # 转为张量
    transforms.Normalize((0.5,), (0.5,))  # 归一化到 [-1, 1]
])

# 加载 MNIST 数据集
train_dataset = datasets.MNIST(root='./data', train=True, transform=transform, download=True)
test_dataset = datasets.MNIST(root='./data', train=False, transform=transform, download=True)

train_loader = torch.utils.data.DataLoader(dataset=train_dataset, batch_size=64, shuffle=True)
test_loader = torch.utils.data.DataLoader(dataset=test_dataset, batch_size=64, shuffle=False)

from torchvision import datasets, transforms
import torch

transform = transforms.Compose([
    transforms.ToTensor(),  # 转为张量
    transforms.Normalize((0.5,), (0.5,))  # 归一化到 [-1, 1]
])

# 加载 MNIST 数据集
train_dataset = datasets.MNIST(root='./data', train=True, transform=transform, download=False)
test_dataset = datasets.MNIST(root='./data', train=False, transform=transform, download=False)

train_loader = torch.utils.data.DataLoader(dataset=train_dataset, batch_size=64, shuffle=True)
test_loader = torch.utils.data.DataLoader(dataset=test_dataset, batch_size=64, shuffle=False)

class SimpleCNN(nn.Module):
    def __init__(self, in_channels = 1, num_classes = 10):
        super(SimpleCNN, self).__init__()

        # 定义卷积层1：输入1通道，输出32通道，卷积核大小3x3
        self.conv1 = nn.Conv2d(in_channels, 32, kernel_size=3, stride=1, padding=1)

        # 定义卷积层2：输入32通道，输出64通道
        self.conv2 = nn.Conv2d(32, 64, kernel_size=3, stride=1, padding=1)

        # 定义全连接层
        self.fc1 = nn.Linear(64 * 7 * 7, 128)  # 输入大小 = 特征图大小 * 通道数
        self.fc2 = nn.Linear(128, num_classes)  # 10 个类别

    def forward(self, x):
        # 第一层卷积 + ReLU
        x = F.relu(self.conv1(x))
        # 最大池化
        x = F.max_pool2d(x, 2)

        # 第二层卷积 + ReLU
        x = F.relu(self.conv2(x))
        # 最大池化
        x = F.max_pool2d(x, 2)

        # 展平操作
        x = x.view(-1, 64 * 7 * 7)

        # 全连接层 + ReLU
        x = F.relu(self.fc1(x))
        # 全连接层输出
        x = self.fc2(x)
        return x

# 创建模型实例
# model = SimpleCNN(in_channels = 1, num_classes = 10).to(device)

# 多分类交叉熵损失
criterion = nn.CrossEntropyLoss()

# 学习率和动量
optimizer = optim.SGD(model.parameters(), lr=0.01, momentum=0.9)

num_epochs = 25
# 设为训练模式
model.train()

for epoch in range(num_epochs):
    total_loss = 0
    for batch_index, (data, targets) in enumerate(tqdm(train_loader)):
        data = data.to(device)
        targets = targets.to(device)

        # 前向传播
        outputs = model(data)
        loss = criterion(outputs, targets)

        # 反向传播
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()

        total_loss += loss.item()
    if (epoch + 1) % 5 == 0:
        print(f"Epoch [{epoch+1}/{num_epochs}], Loss: {total_loss / len(train_loader):.4f}")

# 设置为评估模式
model.eval()
correct = 0
total = 0

# 评估时不需要计算梯度
with torch.no_grad():
    for data, targets in test_loader:
        data.to(device)
        targets.to(device)
        outputs = model(data)
        _, predicted = torch.max(outputs, 1)
        total += targets.size(0)
        correct += (predicted == targets).sum().item()
accuracy = 100 * correct / total
print(f"Test Accuracy: {accuracy:.2f}%")

dataiter = iter(test_loader)
images, labels = next(dataiter)
outputs = model(images)
_, predictions = torch.max(outputs, 1)

fig, axes = plt.subplots(1, 6, figsize=(12, 4))
for i in range(6):
    axes[i].imshow(images[i][0], cmap='gray')
    axes[i].set_title(f"Label: {labels[i]}\nPred: {predictions[i]}")
    axes[i].axis('off')
plt.savefig('./predict_result.png')
# plt.show()