Multi-target not supported at ..\aten\src\THNN/generic/ClassNLLCriterion.c:20

1

Hello People,

I am newbie, working with adult dataset.

I would please appreciate any help here on how to fix this.

iter = 0
for epoch in range(num_epochs):
for i, (inputs, labels) in enumerate(train_loader):

    # Load inputs as Variable
    inputs = Variable(X_trainTensor)
    labels = Variable(y_trainTensor).type(torch.LongTensor)
    
    #y_tensor = torch.tensor(y_train, dtype=torch.long, device=device)
    # Clear gradients w.r.t. parameters
    optimizer.zero_grad()
    
    # Forward pass to get output/logits
    outputs = model(inputs.float())
    
    # Calclate Loss: softmax --> cross entropy loss
    loss = criterion(outputs, labels)
    
    # Getting gradients w.r.t. parameters
    loss.backward()
    
    # Updating parameters
    optimizer.step()

Error:

RuntimeError Traceback (most recent call last)
in
72
73 # Calclate Loss: softmax --> cross entropy loss
—> 74 loss = criterion(outputs, labels)
75
76 # Getting gradients w.r.t. parameters
.
.
.
.
RuntimeError: multi-target not supported at …\aten\src\THNN/generic/ClassNLLCriterion.c:20

2

What criterion are you using?
Can you check if the labels tensor is of size (N,) and outputs tensor is of size (N, C)?
N=batch_size, C=number of classes

3

Thank you very much Raghul.

I have the following:

]:

​
print(inputs.shape)
print(labels.shape)
print(outputs.shape)

torch.Size([100, 4])
torch.Size([100, 1])
torch.Size([100, 1])

Please any suggestion on how to for about this?

input_dim = 4
hidden_dim = 10
output_dim = 1

model = FeedforwardNeuralNetModel(input_dim, hidden_dim, output_dim)

criterion = nn.CrossEntropyLoss()

learning_rate = 0.01

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

#return model, loss_fnc, optimizer

#load_model(0.01)

for epoch in range(num_epochs):
for i, (inputs, labels) in enumerate(train_loader):
# Load inputs as Variable
inputs = Variable(torch.Tensor(inputs.view(-1, 4).float()))
labels = Variable(torch.Tensor(labels.float()))#.type(torch.LongTensor)

    # Clear gradients w.r.t. parameters
    optimizer.zero_grad()
    
    # Forward pass to get output/logits
    outputs = model(inputs.float())
    
    # Calclate Loss: softmax --> cross entropy loss
    
    loss = criterion(outputs.view(-1, 1), labels.type(torch.LongTensor))
    
    # Getting gradients w.r.t. parameters
    loss.backward()
    #loss.backward()
    
    # Updating parameters
    optimizer.step()

RuntimeError Traceback (most recent call last)
in
97 # Calclate Loss: softmax → cross entropy loss
98
—> 99 loss = criterion(outputs.view(-1, 1), labels.type(torch.LongTensor))
100
101 # Getting gradients w.r.t. parameters

~\AppData\Local\Continuum\anaconda3\lib\site-packages\torch\nn\modules\module.py in call(self, *input, **kwargs)
491 result = self._slow_forward(*input, **kwargs)
492 else:
→ 493 result = self.forward(*input, **kwargs)
494 for hook in self._forward_hooks.values():
495 hook_result = hook(self, input, result)

~\AppData\Local\Continuum\anaconda3\lib\site-packages\torch\nn\modules\loss.py in forward(self, input, target)
940 def forward(self, input, target):
941 return F.cross_entropy(input, target, weight=self.weight,
→ 942 ignore_index=self.ignore_index, reduction=self.reduction)
943
944

~\AppData\Local\Continuum\anaconda3\lib\site-packages\torch\nn\functional.py in cross_entropy(input, target, weight, size_average, ignore_index, reduce, reduction)
2054 if size_average is not None or reduce is not None:
2055 reduction = _Reduction.legacy_get_string(size_average, reduce)
→ 2056 return nll_loss(log_softmax(input, 1), target, weight, None, ignore_index, None, reduction)
2057
2058

~\AppData\Local\Continuum\anaconda3\lib\site-packages\torch\nn\functional.py in nll_loss(input, target, weight, size_average, ignore_index, reduce, reduction)
1869 .format(input.size(0), target.size(0)))
1870 if dim == 2:
→ 1871 ret = torch._C._nn.nll_loss(input, target, weight, _Reduction.get_enum(reduction), ignore_index)
1872 elif dim == 4:
1873 ret = torch._C._nn.nll_loss2d(input, target, weight, _Reduction.get_enum(reduction), ignore_index)

RuntimeError: multi-target not supported at …\aten\src\THNN/generic/ClassNLLCriterion.c:20

4

Hi,

The labels should be of size (N,) not (N, 1). You can use labels.squeeze_(-1) to remove the extra dimension of size 1.

5

Hi AlbanD,

Thank you very much for the suggestion.

But this:

RuntimeError: Assertion `cur_target >= 0 && cur_target < n_classes’ failed. at …\aten\src\THNN/generic/ClassNLLCriterion.c:92

Any idea please?

6

@Samuel_Adu I think since the size of outputs is (N, 1) where n_classes or C = 1 here, whereas the labels tensor contains values like for eg. [1, 0, 1, 0, 0] i.e labels for class 0 and 1, this error is thrown.
The variable output_dim should be 2 and size of outputs tensor should be (N, 2).
Just to make sure, is this a binary classification task?

7

This is a different error. This one says that the values contained in your target tensor are larger than the number of classes (or negative).

8

Hi Mailcorahul,

Thank you very much for your kind response.Yes the labels tensor is like this [1, 0, 1, 0, 0] but the outputs tensor size is (N, 2) and Yes, this is a binary classification task…please below…

print(inputs.shape)
print(labels.shape)
print(outputs.shape)
print(labels)
print(outputs)

torch.Size([100, 4])
torch.Size([100])
torch.Size([100, 1])
tensor([1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0,
0, 1, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0,
0, 1, 0, 0, 0, 1, 1, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0, 0, 1, 1, 1, 0, 0, 0,
0, 0, 1, 0])
tensor([[ -0.6877],
[ -79.4583],
[ -0.6877],
[ -0.6877],
[ -0.6877],
[-231.3748],

9

Hi AlbanD,

Thank you very much for your kind response.

Yes it is. It came up after I used labels.squeeze_(-1)…still trying to figure out the solution.

I’d appreciate any further suggestion.

10

Hi,

So the problem is that your output is of size (100, 1), meaning that C = 1 for your loss: you have a single class. I think this should be 2 in your case no?

11

Wow…Thank you very much Raghul

This solves it.

1 Like
12

Wow…Thank you very much AlbanD for your suggestions

This is resolved.