import torch
import torchvision
import torchaudio
from torch import nn
from torch.utils.data import DataLoader
from torchvision import datasets
from torchvision.transforms import ToTensor
from matplotlib import pyplot as plt

print(torch.__version__)
print(torchaudio.__version__)
print(torchvision.__version__)

training_data = datasets.MNIST(
    root="data",
    train=True,
    download=True,
    transform=ToTensor(),
)

test_data = datasets.MNIST(
    root="data",
    train=False,
    download=True,
    transform=ToTensor(),
)

print(len(training_data))

figure = plt.figure()
for i in range(9):
    img, label = training_data[i + 59000]
    figure.add_subplot(3, 3, i + 1)
    plt.title(label)
    plt.axis("off")
    plt.imshow(img.squeeze(), cmap="gray")
plt.show()

# 增加批次大小
train_dataloader = DataLoader(training_data, batch_size=128)  # 增大batch size
test_dataloader = DataLoader(test_data, batch_size=128)

for X, y in test_dataloader:
    print(f"Shape of X[N,C,H,W]:{X.shape}")
    print(f"Shape of y:{y.shape} {y.dtype}")
    break

device = "cuda" if torch.cuda.is_available() else "mps" if torch.backends.mps.is_available() else "cpu"
print(f"Using {device} device")

class NeuralNetwork(nn.Module):
    def __init__(self):
        super().__init__()
        self.a = 10
        self.flatten = nn.Flatten()
        原始架構(gòu)
        self.hidden1 = nn.Linear(28 * 28, 128)
        self.hidden2 = nn.Linear(128, 256)
        self.out = nn.Linear(256, 10)
        
    
    def forward(self, x):
        # 原始前向傳播
        x = self.flatten(x)
        x = self.hidden1(x)
        x = torch.sigmoid(x)
        x = self.hidden2(x)
        x = torch.sigmoid(x)
        return x

model = NeuralNetwork().to(device)
print(model)

def train(dataloader, model, loss_fn, optimizer):
    model.train()
    batch_size_num = 1
    for X, y in dataloader:
        X, y = X.to(device), y.to(device)
        pred = model.forward(X)
        loss = loss_fn(pred, y)
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()

        loss_value = loss.item()
        if batch_size_num % 100 == 0:
            print(f"loss: {loss_value:>7f} [number:{batch_size_num}]")
        batch_size_num += 1

def test(dataloader, model, loss_fn):
    size = len(dataloader.dataset)
    num_batches = len(dataloader)
    model.eval()
    test_loss, correct = 0, 0
    with torch.no_grad():
        for X, y in dataloader:
            X, y = X.to(device), y.to(device)
            pred = model.forward(X)
            test_loss = loss_fn(pred, y)
            correct += (pred.argmax(1) == y).type(torch.float).sum().item()
            a = (pred.argmax(1) == y)
            b = (pred.argmax(1) == y).type(torch.float)
    test_loss /= num_batches
    correct /= size

    print(f"Test result:\n Accuracy:{(100 * correct):.2f}%, Avg loss: {test_loss}")

loss_fn = nn.CrossEntropyLoss()

# 原始優(yōu)化器
optimizer = torch.optim.SGD(model.parameters(), lr=0.01)

train(train_dataloader, model, loss_fn, optimizer)
test(train_dataloader, model, loss_fn)

epochs = 10
for t in range(epochs):
    print(f"Epoch {t+1}\n----------------------")
    train(train_dataloader, model, loss_fn, optimizer)
print("Done!")
test(test_dataloader, model, loss_fn)

        # 改進(jìn)架構(gòu)
        self.hidden1 = nn.Linear(28 * 28, 512)  # 增加神經(jīng)元
        self.dropout1 = nn.Dropout(0.2)  # 添加Dropout
        self.hidden2 = nn.Linear(512, 256)
        self.dropout2 = nn.Dropout(0.2)  # 添加Dropout
        self.hidden3 = nn.Linear(256, 128)  # 增加一層
        self.out = nn.Linear(128, 10)

        # 改進(jìn)的前向傳播
        x = self.flatten(x)
        x = self.hidden1(x)
        x = torch.relu(x)  # 使用ReLU替代sigmoid
        x = self.dropout1(x)  # 訓(xùn)練時隨機(jī)丟棄
        x = self.hidden2(x)
        x = torch.relu(x)  # 使用ReLU替代sigmoid
        x = self.dropout2(x)  # 訓(xùn)練時隨機(jī)丟棄
        x = self.hidden3(x)
        x = torch.relu(x)
        x = self.out(x)

# 改進(jìn)優(yōu)化器
optimizer = torch.optim.Adam(model.parameters(), lr=0.001)  # 降低學(xué)習(xí)率