运行时错误：给定组=1，权重大小为[64,64,3,3]，预期输入[64,3,32,32]有64个通道，但得到了3个通道

Question

我遇到一个运行时错误，指出在使用以 Conv2d(3, 64, kernel_size=(3, 3)) 层开头的自定义模型进行分类期间，期望 64 个通道的层收到了具有 3 个通道的输入。即使输入形状看起来正确 ([64, 3, 32, 32]) 并且符合模型的初始层期望，也会发生错误。

*第一批输入的形状：torch.Size([64, 3, 32, 32]) *

RuntimeError                              Traceback (most recent call last)
\<ipython-input-25-b3283edbd61f\> in \<cell line: 22\>()
30         optimizer.zero_grad()
31
\---\> 32         outputs = model2(inputs)
33         loss = criterion(outputs, labels)
34         loss.backward()

16 frames
/usr/local/lib/python3.10/dist-packages/torch/nn/modules/conv.py in \_conv_forward(self, input, weight, bias)
454                             weight, bias, self.stride,
455                             \_pair(0), self.dilation, self.groups)
\--\> 456         return F.conv2d(input, weight, bias, self.stride,
457                         self.padding, self.dilation, self.groups)
458

RuntimeError: Given groups=1, weight of size \[64, 64, 3, 3\], expected input\[64, 3, 32, 32\] to have 64 channels, but got 3 channels instead

**我的模型： **

[model2_code](https://i.stack.imgur.com/X3ZFm.png)

\*\*Model2 OUTPUT: \*\*

[OUTPUT_Model2](https://i.stack.imgur.com/vH5Sc.png)


\*\*Lines that produce the error: \*\*


model2.to(device)
optimizer = optim.Adam(model2.parameters(), lr=0.001)
criterion = nn.CrossEntropyLoss()

losses = \[\]
def update_plot(epoch, loss):
losses.append(loss)
plt.plot(losses, '-x')
plt.xlabel('Epoch')
plt.ylabel('Loss')
plt.title('Training Loss')
plt.pause(0.001)

num_epochs = 10

for epoch in range(1, num_epochs + 1):
start_time = time.time()
running_loss = 0.0
total_batches = 0

    for i, (inputs, labels) in enumerate(train_loader, 0):
        inputs, labels = inputs.to(device), labels.to(device)
    
        optimizer.zero_grad()
    
        outputs = model2(inputs)
        loss = criterion(outputs, labels)
        loss.backward()
        optimizer.step()
    
        running_loss += loss.item()
        total_batches += 1
    
    avg_l_loss / total_batches
    update_plot(epoch, avg_loss) 
    elapsed_time = time.time() - start_time
    
    if epoch % 10 == 0 or epoch == 1:
        print(f'Epoch {epoch}/{num_epochs} - Loss: {avg_loss:.4f} - Time: {elapsed_time:.2f}s')

plt.show()

Answer 1

再次检查你的前传，确保你没有犯错。我认为您将原始输入（形状

[64, 3, 32, 32]

）作为

block_2

的输入，而不是

block_1

（形状

[64, 64, 32, 32]

）的输出。