今天嘗試自己建立模型，並且用昨天創建好的資料來訓練，首先引入需要用到的模組：

from torch import nn
import torch

辨識要使用的裝置：

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

建立CNN模型：

class CustomConvNeuralNetwork(nn.Module):
    def __init__(self):
        super().__init__()
        self.flatten = nn.Flatten()
        self.cnn_module = nn.Sequential(
            nn.Conv2d(in_channels=3, out_channels=6, kernel_size=3, stride=(1, 1), padding=1),
            nn.ReLU(),
            nn.MaxPool2d(2, 2),
            nn.Conv2d(in_channels=6, out_channels=16, kernel_size=3, stride=(1, 1), padding=1),
            nn.MaxPool2d(2, 2),
            nn.ReLU(),
        )
        self.fc_modeul = nn.Sequential(
            nn.Linear(16 * 15 * 15, 120),
            nn.ReLU(),
            nn.Linear(120, 84),
            nn.ReLU(),
            nn.Linear(84, 7)
        )


    def forward(self, x):
        x = self.cnn_module(x)
        x = self.flatten(x)
        x = self.fc_modeul(x)
        
        return x

cnn_model = CustomConvNeuralNetwork().to(device)
print(cnn_model)

設定超參數：

# 超參數
learning_rate = 1e-3
batch_size = 64
epochs = 10
# 初始化loss function
loss_fn = nn.CrossEntropyLoss()
optimizer = torch.optim.SGD(cnn_model.parameters(), lr=learning_rate)

設定訓練和測試的迴圈：

def train_loop(dataloader, model, loss_fn, optimizer):
    size = len(dataloader.dataset)
    model.train()
    for batch, (X, y) in enumerate(dataloader):
        try:
            y = torch.tensor(y, dtype=torch.long)
            # 將資料讀取到GPU中
            X, y = X.to(device), y.to(device)
            # 運算出結果並計算loss
            pred = model(X)
            loss = loss_fn(pred, y)

            # 反向傳播
            loss.backward()
            optimizer.step()
            optimizer.zero_grad()
        except:
            print("image error")
        if batch % 100 == 0:
            loss, current = loss.item(), (batch + 1) * len(X)
            print(f"loss: {loss:>7f}  [{current:>5d}/{size:>5d}]")

def test_loop(dataloader, model, loss_fn):
    model.eval()
    size = len(dataloader.dataset)
    num_batches = len(dataloader)
    test_loss, correct = 0, 0
    # 驗證或測試時記得加入 torch.no_grad() 讓神經網路不要更新
    with torch.no_grad():
        for X, y in dataloader:
            y = torch.tensor(y, dtype=torch.long)
            X, y = X.to(device), y.to(device)
            pred = model(X)
            test_loss += loss_fn(pred, y).item()
            correct += (pred.argmax(1) == y).type(torch.float).sum().item()

    test_loss /= num_batches
    correct /= size
    print(f"Test Error: \n Accuracy: {(100*correct):>0.1f}%, Avg loss: {test_loss:>8f} \n")

接下來就是訓練的流程：

for t in range(epochs):
    print(f"Epoch {t+1}\n-------------------------------")
    train_loop(train_dataloader, cnn_model, loss_fn, optimizer)
    test_loop(test_loader, cnn_model, loss_fn)
print("Done!")

But，寫程式最容易遇到的 But，訓練時報了錯誤，查了一下Google發現是我的 label 格式錯誤，因此做了修改。

y = torch.tensor(y, dtype=torch.long) 我在資料及新增了這行，順利運行了幾百筆資料又出錯了，這次的錯誤是：

個人猜測有部分資料出問題，可能是圖片損毀之類的，但是今天有點晚了，明天再來嘗試解開這個問題吧~

圖片來源