Cross_entropy_loss(): argument 'input' (position 1) must be Tensor, not collections.OrderedDict

vision
#1

hi friends
I was trying to execute a simple code and encountered this error while calculating the loss . Can anyone tell me the code problem? Thanks

from torchvision.datasets import Cityscapes

transform = transforms.Compose([
    transforms.Resize([256,256]),
    transforms.ToTensor()
])

trainset =  Cityscapes(root='/cityscape', split='train', mode='fine', target_type='semantic', transform=transform, target_transform=transform)
valset =  Cityscapes(root='/cityscape', split='train', mode='fine', target_type='semantic', transform=transform, target_transform=transform)
trainloader = torch.utils.data.DataLoader( trainset, batch_size=2,  shuffle=True, num_workers=0)
valloader = torch.utils.data.DataLoader( valset, batch_size=2,  shuffle=True, num_workers=0)

from torchvision import models
model = models.segmentation.deeplabv3_resnet101(pretrained=False,progress=True,num_classes=19)
model = model.to(device)

optimizer = torch.optim.SGD(model.parameters(), lr=0.001, momentum=0.9)
criterion = nn.CrossEntropyLoss()

def train_model(model, dataloaders, criterion, optimizer, num_epochs):

    since = time.time()
    val_acc_history = []

    best_model_wts = copy.deepcopy(model.state_dict())
    best_acc = 0.0

    for epoch in range(num_epochs):

        # Each epoch has a training and validation phase
        for phase in ['train', 'val']:
            if phase == 'train':
                model.train()  # Set model to training mode
            else:
                model.eval()   # Set model to evaluate mode
            running_loss = 0.0
            running_corrects = 0
            for inputs, labels in dataloaders[phase]:

                inputs = inputs.to(device)
                labels = labels.to(device)

                # zero the parameter gradients
                optimizer.zero_grad()

                # track history if only in train
                with torch.set_grad_enabled(phase == 'train'):

                    outputs = model(inputs)
                    loss = criterion(outputs, labels)
                    _, preds = torch.max(outputs, 1)

'''
/usr/local/lib/python3.7/dist-packages/torch/nn/functional.py in cross_entropy(input, target, weight, size_average, ignore_index, reduce, reduction, label_smoothing)
   2844     if size_average is not None or reduce is not None:
   2845         reduction = _Reduction.legacy_get_string(size_average, reduce)
-> 2846     return torch._C._nn.cross_entropy_loss(input, target, weight, _Reduction.get_enum(reduction), ignore_index, label_smoothing)
   2847 
   2848 

TypeError: cross_entropy_loss(): argument 'input' (position 1) must be Tensor, not collections.OrderedDict
'''
#2

your model must be returning a dictionary that you need to use part of it maybe, I dont know the model so I can not speculate about the return of it, could you print output.keys() and send it here?

#3

Thank you for your attention. Iā€™m not sure you mean this or not. But when I apply this :

outputs = model(inputs)
print(outputs.keys())

I get this result :

/usr/local/lib/python3.7/dist-packages/torchsummary/torchsummary.py in hook(module, input, output)
     24                 ]
     25             else:
---> 26                 summary[m_key]["output_shape"] = list(output.size())
     27                 summary[m_key]["output_shape"][0] = batch_size
     28 

AttributeError: 'collections.OrderedDict' object has no attribute 'size'
#4

change the line:

outputs = model(inputs)

for:

outputs = model(inputs)['out']

should be working.

1 Like