# UserWarning: Mixed memory format inputs

**URL:** <https://discuss.pytorch.org/t/userwarning-mixed-memory-format-inputs/97571>\
**Category:** Uncategorized\
**Created:** [September 26, 2020, 4:52pm UTC](https://discuss.pytorch.org/t/userwarning-mixed-memory-format-inputs/97571 "2020-09-26T16:52:18Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![NearIt](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/nearit/32/12260_2.png) [@NearIt](https://discuss.pytorch.org/u/NearIt)\
**Post date:** [September 26, 2020, 4:52pm UTC](https://discuss.pytorch.org/t/userwarning-mixed-memory-format-inputs/97571/1 "2020-09-26T16:52:18Z")

</div>

I have implemented Unet with custom loss function : Dice loss.

while training, I receive the following message :  
`/usr/local/lib/python3.6/dist-packages/ipykernel_launcher.py:32: UserWarning: Mixed memory format inputs detected while calling the operator. The operator will output contiguous tensor even if some of the inputs are in channels_last format. (Triggered internally at /pytorch/aten/src/ATen/native/TensorIterator.cpp:918.)`

this warning appear only when calling the Dice loss function. this does not happens when I am calling pytorch function torch.nn.CrossEntropyLoss().

this is my Dice implementation (copied from some webpage):

```auto
def calculate_dice_loss(logits,true, eps = 1e-7):

    """Computes the Sørensen–Dice loss.
    Note that PyTorch optimizers minimize a loss. In this
    case, we would like to maximize the dice loss so we
    return the negated dice loss.
    Args:
        true: a tensor of shape [Batch size x 512 x 512].
        logits: a tensor of shape [Batch size x numLabels x 512 x 512]. Corresponds to
            the raw output or logits of the model.
        eps: added to the denominator for numerical stability.
    Returns:
        dice_loss: the Sørensen–Dice loss.
    """
    num_classes = logits.shape[1]
    true = true.unsqueeze(1) # now true: a tensor of shape [Batch size x 1 x 512 x 512].
    true_1_hot = torch.eye(num_classes)[true.squeeze(1)]
    true_1_hot = true_1_hot.permute(0, 3, 1, 2).float()
    probas = F.softmax(logits, dim = 1)
    true_1_hot = true_1_hot.type(logits.type())
    dims = (0,) + tuple(range(2, true.ndimension()))
    intersection = torch.sum(probas * true_1_hot, dims)
    cardinality = torch.sum(probas + true_1_hot, dims)
    dice_loss = (2. * intersection / (cardinality + eps)).mean()
    return (1 - dice_loss)

```

and this is how I call the loss function while training:

```auto
# loss values
    CrossEntropy_Loss = CrossEntropy_criterion(logits, true_labels)
    dice_loss = calculate_dice_loss(logits,true_labels)

    loss = dice_loss 

    # Back propagation

    unet.zero_grad()
    loss.backward()
    optimizer.step()
    loss_list.append(loss.item())

```

---

<div class="post-metadata">

**Author:** ![ptrblck](https://discuss.pytorch.org/user_avatar/discuss.pytorch.org/ptrblck/32/1823_2.png) [@ptrblck](https://discuss.pytorch.org/u/ptrblck)\
**Post date:** [September 27, 2020, 1:51am UTC](https://discuss.pytorch.org/t/userwarning-mixed-memory-format-inputs/97571/2 "2020-09-27T01:51:31Z")

</div>

Are you using the `channels_last` memory format in your model?  
If not, could you print the shapes of `logits` and `true_labels` before passing them into the loss function?