3

我想使用 pretrain inceptionV3 训练一个 100 类的模型，但是在我的训练代码中，我遇到了一个奇怪的错误，显示如下错误：

我的培训代码是：

step = -1  
print_inter=50
val_inter=400
train_size = ceil(len(data_set['train']) / dataloader['train'].batch_size)
for epoch in range(50):
    # train phase
    exp_lr_scheduler.step(epoch)
    inception_v3.train(True)
    for batch_cnt, data in enumerate(dataloader['train']):
        step += 1
        inception_v3.train(True)

        inputs, labels = data

        inputs = torch.autograd.Variable(inputs.cuda())
        labels = torch.autograd.Variable(torch.from_numpy(np.array(labels)).long().cuda())
        outputs = inception_v3(inputs)
        # zero the parameter gradients
        optimizer.zero_grad()

        outputs = inception_v3(inputs)
        print(inputs.shape)
        print(outputs[0].shape)
        print(outputs[1].shape)
        print(labels.shape)

        loss = criterion(outputs[0], labels)
        loss += criterion(outputs[1], labels)
        outputs = (outputs[0] + outputs[1]) / 2

        _, preds = torch.max(outputs, 1)
        loss.backward()
        optimizer.step()
        # batch loss

        inception_v3.train(False)  # Set model to evaluate mode

        for batch_cnt_val, data_val in enumerate(dataloader['val']):
            # print data
            inputs, labels = data_val

            inputs = Variable(inputs.cuda())
            labels = Variable(torch.from_numpy(np.array(labels)).long().cuda())

            # forward
            outputs = inception_v3(inputs)
            print(inputs.shape)
            print(outputs[0].shape)
            print(outputs[1].shape)
            print(labels.shape)
            loss = criterion(outputs[0], labels)
            loss += criterion(outputs[1], labels)
            outputs = (outputs[0] + outputs[1]) / 2

运行此代码后，得到以下结果：

torch.Size([8, 3, 299, 299])
torch.Size([8, 100])
torch.Size([8, 100])
torch.Size([8])
torch.Size([8, 3, 299, 299])
torch.Size([100])
torch.Size([100])
torch.Size([8])
        ---------------------------------------------------------------------------
    RuntimeError                              Traceback (most recent call last)
    <ipython-input-21-0cfbe57aae73> in <module>()
         49             print(outputs[1].shape)
         50             print(labels.shape)
    ---> 51             loss = criterion(outputs[0], labels)
         52             loss += criterion(outputs[1], labels)
         53             outputs = (outputs[0] + outputs[1]) / 2

    ~/anaconda3/lib/python3.6/site-packages/torch/nn/modules/module.py in __call__(self, *input, **kwargs)
        489             result = self._slow_forward(*input, **kwargs)
        490         else:
    --> 491             result = self.forward(*input, **kwargs)
        492         for hook in self._forward_hooks.values():
        493             hook_result = hook(self, input, result)

    ~/anaconda3/lib/python3.6/site-packages/torch/nn/modules/loss.py in forward(self, input, target)
        757         _assert_no_grad(target)
        758         return F.cross_entropy(input, target, self.weight, self.size_average,
    --> 759                                self.ignore_index, self.reduce)
        760 
        761 

    ~/anaconda3/lib/python3.6/site-packages/torch/nn/functional.py in cross_entropy(input, target, weight, size_average, ignore_index, reduce)
       1440         >>> loss.backward()
       1441     """
    -> 1442     return nll_loss(log_softmax(input, 1), target, weight, size_average, ignore_index, reduce)
       1443 
       1444 

    ~/anaconda3/lib/python3.6/site-packages/torch/nn/functional.py in log_softmax(input, dim, _stacklevel)
        942     if dim is None:
        943         dim = _get_softmax_dim('log_softmax', input.dim(), _stacklevel)
    --> 944     return torch._C._nn.log_softmax(input, dim)
        945 
        946 

    RuntimeError: dimension out of range (expected to be in range of [-1, 0], but got 1)

您可以看到相同的模型获得相同的输入大小，但我的第二个输出错过了批处理维度。原谅我蹩脚的英文

4

0 回答 0