PyTorch NotImplementedError 转发

Posted 2023-03-12

【中文标题】PyTorch NotImplementedError 转发

【英文标题】：PyTorch NotImplementedError in forward

【发布时间】：2019-02-13 22:29:36

【问题描述】：

import torch
import torch.nn as nn

device = torch.device('cuda' if torch.cuda.is_available() else 
'cpu')

class Model(nn.Module):
    def __init__(self):
        super(Model, self).__init__()
        self.layer = nn.Sequential(
            nn.Conv2d(1, 16, kernel_size=3, stride=1, padding=1),
            nn.BatchNorm2d(16),
            nn.ReLU(),
            nn.MaxPool2d(kernel_size=2, stride=2), # 16x16x650
            nn.Conv2d(16, 32, kernel_size=3, stride=1, padding=1), # 32x16x650
            nn.ReLU(),
            nn.Dropout2d(0.5),
            nn.Conv2d(32, 64, kernel_size=3, stride=1, padding=1), # 64x16x650
            nn.ReLU(),
            nn.MaxPool2d(kernel_size=2, stride=2), # 64x8x325
            nn.Conv2d(64, 128, kernel_size=3, stride=1, padding=1),
            nn.ReLU()) # 64x8x325

        self.fc = nn.Sequential(
            nn.Linear(64*8*325, 128),
            nn.ReLU(),
            nn.Linear(128, 256),
            nn.ReLU(),
            nn.Linear(256, 1),
        )

        def forward(self, x):
            out = self.layer1(x)
            out = self.layer2(out)
            out = out.reshape(out.size(0), -1)
            out = self.fc(out)
            return out

# HYPERPARAMETER
learning_rate = 0.0001 
num_epochs = 15

import data

def main():
    model = Model().to(device)

    criterion = nn.CrossEntropyLoss()
    optimizer = torch.optim.Adam(model.parameters(), 
lr=learning_rate)

    total_step = len(data.train_loader)
    for epoch in range(num_epochs):
        for i, (images, labels) in enumerate(data.train_loader):
            images = images.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 + 1, num_epochs, i + 1, total_step, loss.item()))

    model.eval()
    with torch.no_grad():
        correct = 0
        total = 0
        for images, labels in data.test_loader:
            images = images.to(device)
            labels = labels.to(device)
            outputs = model(images)
            _, predicted = torch.max(outputs.data, 1)
            total += labels.size(0)
            correct += (predicted == labels).sum().item()

        print('Test Accuracy of the model on the 10000 test images:  %'.format(100 * correct / total))

if __name__ == '__main__':
    main()

错误：

File "/home/rladhkstn8/Desktop/SWID/tmp/pycharm_project_853/model.py", line 82, in <module>
    main()
  File "/home/rladhkstn8/Desktop/SWID/tmp/pycharm_project_853/model.py", line 56, in main
    outputs = model(images)
  File "/home/rladhkstn8/anaconda3/lib/python3.6/site-packages/torch/nn/modules/module.py", line 477, in __call__
    result = self.forward(*input, **kwargs)
  File "/home/rladhkstn8/anaconda3/lib/python3.6/site-packages/torch/nn/modules/module.py", line 83, in forward
    raise NotImplementedError
NotImplementedError

我不知道问题出在哪里。我知道应该实现NotImplementedError，但是当有未实现的代码时就会发生。

【问题讨论】：

如果您尝试通过ModuleList 而不是Sequential 呼叫前转，也会发生这种情况

【参考方案1】：

只需在 Model 类中取消缩进你的 forward 方法。

像这样：

class Model(nn.Module):
    def __init__(self):
        super(Model, self).__init__()
        self.layer = nn.Sequential(
            nn.Conv2d(1, 16, kernel_size=3, stride=1, padding=1),
            nn.BatchNorm2d(16),
            nn.ReLU(),
            nn.MaxPool2d(kernel_size=2, stride=2), # 16x16x650
            nn.Conv2d(16, 32, kernel_size=3, stride=1, padding=1), # 32x16x650
            nn.ReLU(),
            nn.Dropout2d(0.5),
            nn.Conv2d(32, 64, kernel_size=3, stride=1, padding=1), # 64x16x650
            nn.ReLU(),
            nn.MaxPool2d(kernel_size=2, stride=2), # 64x8x325
            nn.Conv2d(64, 128, kernel_size=3, stride=1, padding=1),
            nn.ReLU()) # 64x8x325

        self.fc = nn.Sequential(
            nn.Linear(64*8*325, 128),
            nn.ReLU(),
            nn.Linear(128, 256),
            nn.ReLU(),
            nn.Linear(256, 1),
        )

    def forward(self, x):
        out = self.layer1(x)
        out = self.layer2(out)
        out = out.reshape(out.size(0), -1)
        out = self.fc(out)
        return out

【参考方案2】：

请仔细查看您的__init__ 函数的indentation：您的forward 是__init__ 的一部分，而不是您的模块的一部分。

【讨论】：

我不明白你的意思。一个代码示例会很有帮助。

@samisnotinsane 如果您要从定义__init__ 的位置垂直握住标尺并让它垂直向下运行您的代码，则应该在标尺碰到它的线的位置定义forward。相反，您的从标尺缩进一个制表符，即标尺和forward 之间有一个制表符的空格。您使用两个制表符缩进了def forward，而不是像def __init__ 这样的一个制表符。这意味着您在__init__ 中定义了forward，当它是它自己的方法时，独立于__init__。

【参考方案3】：

当您没有从超类中实现所需的方法时，会发生此错误，在我的情况下，我在函数名称前有拼写错误。我建议你检查你的代码缩进。