Issue with RandCropByPosNegLabeld

[tararelan]

Issue with RandCropByPosNegLabeld

[KumoLiu]

Hi @tararelan, could you please share the whole error message?
At first glance, I can only wonder if you ensured that the image and label sizes entered are consistent.

[tararelan]

---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
[/usr/local/lib/python3.10/dist-packages/monai/transforms/transform.py](https://localhost:8080/#) in apply_transform(transform, data, map_items, unpack_items, log_stats, lazy, overrides)
    140             return [_apply_transform(transform, item, unpack_items, lazy, overrides, log_stats) for item in data]
--> 141         return _apply_transform(transform, data, unpack_items, lazy, overrides, log_stats)
    142     except Exception as e:

15 frames
[/usr/local/lib/python3.10/dist-packages/monai/transforms/transform.py](https://localhost:8080/#) in _apply_transform(transform, data, unpack_parameters, lazy, overrides, logger_name)
     97 
---> 98     return transform(data, lazy=lazy) if isinstance(transform, LazyTrait) else transform(data)
     99 

[/usr/local/lib/python3.10/dist-packages/monai/transforms/croppad/dictionary.py](https://localhost:8080/#) in __call__(self, data, lazy)
    999 
-> 1000         self.randomize(d.get(self.label_key), fg_indices, bg_indices, d.get(self.image_key))
   1001 

[/usr/local/lib/python3.10/dist-packages/monai/transforms/croppad/dictionary.py](https://localhost:8080/#) in randomize(self, label, fg_indices, bg_indices, image)
    981     ) -> None:
--> 982         self.cropper.randomize(label=label, fg_indices=fg_indices, bg_indices=bg_indices, image=image)
    983 

[/usr/local/lib/python3.10/dist-packages/monai/transforms/croppad/array.py](https://localhost:8080/#) in randomize(self, label, fg_indices, bg_indices, image)
   1151                 raise ValueError("label must be provided.")
-> 1152             fg_indices_, bg_indices_ = map_binary_to_indices(label, image, self.image_threshold)
   1153         _shape = None

[/usr/local/lib/python3.10/dist-packages/monai/transforms/utils.py](https://localhost:8080/#) in map_binary_to_indices(label, image, image_threshold)
    359         img_flat, *_ = convert_to_dst_type(img_flat, label, dtype=bool)
--> 360         bg_indices = nonzero(img_flat & ~label_flat)
    361     else:

[/usr/local/lib/python3.10/dist-packages/monai/data/meta_tensor.py](https://localhost:8080/#) in __torch_function__(cls, func, types, args, kwargs)
    281             kwargs = {}
--> 282         ret = super().__torch_function__(func, types, args, kwargs)
    283         # if `out` has been used as argument, metadata is not copied, nothing to do.

[/usr/local/lib/python3.10/dist-packages/torch/_tensor.py](https://localhost:8080/#) in __torch_function__(cls, func, types, args, kwargs)
   1417         with _C.DisableTorchFunctionSubclass():
-> 1418             ret = func(*args, **kwargs)
   1419             if func in get_default_nowrap_functions():

RuntimeError: The size of tensor a (7663256) must match the size of tensor b (4919040) at non-singleton dimension 0

The above exception was the direct cause of the following exception:

RuntimeError                              Traceback (most recent call last)
[<ipython-input-9-ca3a090a9802>](https://localhost:8080/#) in <cell line: 2>()
      1 check_iter = iter(train_loader)
----> 2 check_data = next(check_iter)
      3 
      4 image, label = check_data["image"].squeeze().numpy(), check_data["label"].squeeze().numpy()
      5 print(f"image shape: {image.shape}, label shape: {label.shape}")

[/usr/local/lib/python3.10/dist-packages/torch/utils/data/dataloader.py](https://localhost:8080/#) in __next__(self)
    629                 # TODO(https://github.com/pytorch/pytorch/issues/76750)
    630                 self._reset()  # type: ignore[call-arg]
--> 631             data = self._next_data()
    632             self._num_yielded += 1
    633             if self._dataset_kind == _DatasetKind.Iterable and \

[/usr/local/lib/python3.10/dist-packages/torch/utils/data/dataloader.py](https://localhost:8080/#) in _next_data(self)
    673     def _next_data(self):
    674         index = self._next_index()  # may raise StopIteration
--> 675         data = self._dataset_fetcher.fetch(index)  # may raise StopIteration
    676         if self._pin_memory:
    677             data = _utils.pin_memory.pin_memory(data, self._pin_memory_device)

[/usr/local/lib/python3.10/dist-packages/torch/utils/data/_utils/fetch.py](https://localhost:8080/#) in fetch(self, possibly_batched_index)
     49                 data = self.dataset.__getitems__(possibly_batched_index)
     50             else:
---> 51                 data = [self.dataset[idx] for idx in possibly_batched_index]
     52         else:
     53             data = self.dataset[possibly_batched_index]

[/usr/local/lib/python3.10/dist-packages/torch/utils/data/_utils/fetch.py](https://localhost:8080/#) in <listcomp>(.0)
     49                 data = self.dataset.__getitems__(possibly_batched_index)
     50             else:
---> 51                 data = [self.dataset[idx] for idx in possibly_batched_index]
     52         else:
     53             data = self.dataset[possibly_batched_index]

[/content/drive/MyDrive/Colab Notebooks/COMP4211/Project/loader.py](https://localhost:8080/#) in __getitem__(self, idx)
     14         data_dict = self.data[idx]
     15         if self.transform:
---> 16             data_dict = self.transform(data_dict)
     17         return data_dict
     18 

[/usr/local/lib/python3.10/dist-packages/monai/transforms/compose.py](https://localhost:8080/#) in __call__(self, input_, start, end, threading, lazy)
    333     def __call__(self, input_, start=0, end=None, threading=False, lazy: bool | None = None):
    334         _lazy = self._lazy if lazy is None else lazy
--> 335         result = execute_compose(
    336             input_,
    337             transforms=self.transforms,

[/usr/local/lib/python3.10/dist-packages/monai/transforms/compose.py](https://localhost:8080/#) in execute_compose(data, transforms, map_items, unpack_items, start, end, lazy, overrides, threading, log_stats)
    109         if threading:
    110             _transform = deepcopy(_transform) if isinstance(_transform, ThreadUnsafe) else _transform
--> 111         data = apply_transform(
    112             _transform, data, map_items, unpack_items, lazy=lazy, overrides=overrides, log_stats=log_stats
    113         )

[/usr/local/lib/python3.10/dist-packages/monai/transforms/transform.py](https://localhost:8080/#) in apply_transform(transform, data, map_items, unpack_items, log_stats, lazy, overrides)
    169             else:
    170                 _log_stats(data=data)
--> 171         raise RuntimeError(f"applying transform {transform}") from e
    172 
    173 

RuntimeError: applying transform <monai.transforms.croppad.dictionary.RandCropByPosNegLabeld object at 0x7f32347b8c10>

[KumoLiu]

Hi @tararelan, could you please try Dataset instead of your customized one? Didn't see special reason you write your own dataset.

apply_transform has some logic when your data is a list.
https://github.com/Project-MONAI/MONAI/blob/dev/monai/transforms/transform.py#L101

[tararelan]

Hi, I used Dataset instead of the customised one and it works. However, when I try to train my model using:

train_ds = Dataset(train_data_dicts, train_transforms)
val_ds = Dataset(val_data_dicts, val_transforms)

train_loader = DataLoader(train_ds, batch_size=8, shuffle=True, collate_fn=pad_list_data_collate)
val_loader = DataLoader(val_ds, batch_size=8, shuffle=False, collate_fn=pad_list_data_collate)

import torch.nn as nn
from torchsummary import summary
from monai.networks.nets import UNet
from monai.networks.layers import Norm

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

model = UNet(
    spatial_dims=3,
    in_channels=1,
    out_channels=2,
    channels=(16, 32, 64, 128, 256),
    strides=(2, 2, 2, 2),
    num_res_units=2,
    norm=Norm.BATCH,
).to(device)

model = model.to(device)
summary(model, input_size=(1, 16, 128, 128))

from monai.metrics import DiceMetric, SurfaceDistanceMetric, HausdorffDistanceMetric, MeanIoU
from monai.losses import DiceLoss

optimizer = torch.optim.Adam(model.parameters(), 1e-4)

loss_function = DiceLoss(to_onehot_y=True, softmax=True, jaccard=False)
dice_metric = DiceMetric(include_background=False, reduction="mean")
asd_metric = SurfaceDistanceMetric(include_background=False, distance_metric="euclidean")
hausdorff_metric = HausdorffDistanceMetric(include_background=False,percentile=95)
jaccard_metric = MeanIoU(include_background=False)


max_epochs = 2
val_interval = 2

best_metric = -1
best_metric_epoch = -1

epoch_loss_values = []
dice_values = []
asd_values = []
hausdorff_values = []
jaccard_values = []

post_pred = Compose([AsDiscrete(argmax=True, to_onehot=2)])
post_label = Compose([AsDiscrete(to_onehot=2)])


from monai.inferers import sliding_window_inference
from monai.data import decollate_batch
from monai.transforms import SpatialCrop

wandb.init(project="Project", name="Trial1")

for epoch in range(max_epochs):
    print("-" * 10)
    model.train()
    epoch_loss = 0
    step = 0
    for batch_data in train_loader:
        step += 1
        inputs, labels = (
            batch_data["image"].to(device),
            batch_data["label"].to(device),
        )
        optimizer.zero_grad()
        outputs = model(inputs)
        loss = loss_function(outputs, labels)
        loss.backward()
        optimizer.step()
        epoch_loss += loss.item()

    epoch_loss /= step
    epoch_loss_values.append(epoch_loss)
    print(f"epoch {epoch + 1} average loss: {epoch_loss:.4f}")
    wandb.log({'train_loss': epoch_loss}, step= epoch)

    if (epoch + 1) % val_interval == 0:
        model.eval()
        with torch.no_grad():
            for val_data in val_loader:
                val_inputs, val_labels = (
                    val_data["image"].to(device),
                    val_data["label"].to(device),
                )

                roi_size = (96, 96, 96)
                sw_batch_size = 4

                val_outputs = sliding_window_inference(val_inputs, roi_size, sw_batch_size, model)
                val_outputs = [post_pred(i) for i in decollate_batch(val_outputs)]
                val_labels = [post_label(i) for i in decollate_batch(val_labels)]

                # val_outputs = torch.stack(val_outputs)
                # val_labels = torch.stack(val_labels)

                print(val_outputs.shape, val_labels.shape)
                dice_metric(val_outputs, val_labels)
                asd_metric(y_pred=val_outputs, y=val_labels)
                hausdorff_metric(y_pred=val_outputs, y=val_labels)
                jaccard_metric(y_pred=val_outputs, y=val_labels)

            dice = dice_metric.aggregate().item()
            asd = asd_metric.aggregate().item()
            hausdorff = hausdorff_metric.aggregate().item()
            jaccard = jaccard_metric.aggregate().item()

            dice_metric.reset()
            asd_metric.reset()
            hausdorff_metric.reset()
            jaccard_metric.reset()

            dice_values.append(dice)
            asd_values.append(asd)
            hausdorff_values.append(hausdorff)
            jaccard_values.append(jaccard)
            wandb.log({'val_dice': dice}, step= epoch)
            wandb.log({'val_asd': asd}, step= epoch)
            wandb.log({'val_hausdorff': hausdorff}, step= epoch)
            wandb.log({'val_jaccard': jaccard}, step= epoch)

            if dice > best_metric:
                best_metric = dice
                best_metric_epoch = epoch + 1
                torch.save(model.state_dict(), "./best_metric_model.pth")
                print("saved new best metric model")
            print(
                f"current epoch: {epoch + 1} current mean dice: {dice:.4f}"
                f"\nbest mean dice: {best_metric:.4f} "
                f"at epoch: {best_metric_epoch}"
            )

I get the following error:

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
[/usr/local/lib/python3.10/dist-packages/monai/transforms/transform.py](https://localhost:8080/#) in apply_transform(transform, data, map_items, unpack_items, log_stats, lazy, overrides)
    139         if isinstance(data, (list, tuple)) and map_items:
--> 140             return [_apply_transform(transform, item, unpack_items, lazy, overrides, log_stats) for item in data]
    141         return _apply_transform(transform, data, unpack_items, lazy, overrides, log_stats)

16 frames
[/usr/local/lib/python3.10/dist-packages/monai/transforms/transform.py](https://localhost:8080/#) in <listcomp>(.0)
    139         if isinstance(data, (list, tuple)) and map_items:
--> 140             return [_apply_transform(transform, item, unpack_items, lazy, overrides, log_stats) for item in data]
    141         return _apply_transform(transform, data, unpack_items, lazy, overrides, log_stats)

[/usr/local/lib/python3.10/dist-packages/monai/transforms/transform.py](https://localhost:8080/#) in _apply_transform(transform, data, unpack_parameters, lazy, overrides, logger_name)
     97 
---> 98     return transform(data, lazy=lazy) if isinstance(transform, LazyTrait) else transform(data)
     99 

[/usr/local/lib/python3.10/dist-packages/monai/transforms/croppad/array.py](https://localhost:8080/#) in __call__(self, img, mode, lazy, **pad_kwargs)
   1470         lazy_ = self.lazy if lazy is None else lazy
-> 1471         ret = self.padder(self.cropper(img, lazy_), mode=mode, lazy=lazy_, **pad_kwargs)
   1472         # remove the individual info and combine

[/usr/local/lib/python3.10/dist-packages/monai/transforms/croppad/array.py](https://localhost:8080/#) in __call__(self, img, lazy)
    534             img=img,
--> 535             slices=self.compute_slices(img.peek_pending_shape() if isinstance(img, MetaTensor) else img.shape[1:]),
    536             lazy=lazy_,

AttributeError: 'dict' object has no attribute 'shape'

The above exception was the direct cause of the following exception:

RuntimeError                              Traceback (most recent call last)
[/usr/local/lib/python3.10/dist-packages/monai/transforms/transform.py](https://localhost:8080/#) in apply_transform(transform, data, map_items, unpack_items, log_stats, lazy, overrides)
    140             return [_apply_transform(transform, item, unpack_items, lazy, overrides, log_stats) for item in data]
--> 141         return _apply_transform(transform, data, unpack_items, lazy, overrides, log_stats)
    142     except Exception as e:

[/usr/local/lib/python3.10/dist-packages/monai/transforms/transform.py](https://localhost:8080/#) in _apply_transform(transform, data, unpack_parameters, lazy, overrides, logger_name)
     97 
---> 98     return transform(data, lazy=lazy) if isinstance(transform, LazyTrait) else transform(data)
     99 

[/usr/local/lib/python3.10/dist-packages/monai/transforms/compose.py](https://localhost:8080/#) in __call__(self, input_, start, end, threading, lazy)
    334         _lazy = self._lazy if lazy is None else lazy
--> 335         result = execute_compose(
    336             input_,

[/usr/local/lib/python3.10/dist-packages/monai/transforms/compose.py](https://localhost:8080/#) in execute_compose(data, transforms, map_items, unpack_items, start, end, lazy, overrides, threading, log_stats)
    110             _transform = deepcopy(_transform) if isinstance(_transform, ThreadUnsafe) else _transform
--> 111         data = apply_transform(
    112             _transform, data, map_items, unpack_items, lazy=lazy, overrides=overrides, log_stats=log_stats

[/usr/local/lib/python3.10/dist-packages/monai/transforms/transform.py](https://localhost:8080/#) in apply_transform(transform, data, map_items, unpack_items, log_stats, lazy, overrides)
    170                 _log_stats(data=data)
--> 171         raise RuntimeError(f"applying transform {transform}") from e
    172 

RuntimeError: applying transform <monai.transforms.croppad.array.ResizeWithPadOrCrop object at 0x7f32347b8b80>

The above exception was the direct cause of the following exception:

RuntimeError                              Traceback (most recent call last)
[<ipython-input-21-727b3f1caebb>](https://localhost:8080/#) in <cell line: 7>()
     10     epoch_loss = 0
     11     step = 0
---> 12     for batch_data in train_loader:
     13         step += 1
     14         inputs, labels = (

[/usr/local/lib/python3.10/dist-packages/torch/utils/data/dataloader.py](https://localhost:8080/#) in __next__(self)
    629                 # TODO(https://github.com/pytorch/pytorch/issues/76750)
    630                 self._reset()  # type: ignore[call-arg]
--> 631             data = self._next_data()
    632             self._num_yielded += 1
    633             if self._dataset_kind == _DatasetKind.Iterable and \

[/usr/local/lib/python3.10/dist-packages/torch/utils/data/dataloader.py](https://localhost:8080/#) in _next_data(self)
    673     def _next_data(self):
    674         index = self._next_index()  # may raise StopIteration
--> 675         data = self._dataset_fetcher.fetch(index)  # may raise StopIteration
    676         if self._pin_memory:
    677             data = _utils.pin_memory.pin_memory(data, self._pin_memory_device)

[/usr/local/lib/python3.10/dist-packages/torch/utils/data/_utils/fetch.py](https://localhost:8080/#) in fetch(self, possibly_batched_index)
     49                 data = self.dataset.__getitems__(possibly_batched_index)
     50             else:
---> 51                 data = [self.dataset[idx] for idx in possibly_batched_index]
     52         else:
     53             data = self.dataset[possibly_batched_index]

[/usr/local/lib/python3.10/dist-packages/torch/utils/data/_utils/fetch.py](https://localhost:8080/#) in <listcomp>(.0)
     49                 data = self.dataset.__getitems__(possibly_batched_index)
     50             else:
---> 51                 data = [self.dataset[idx] for idx in possibly_batched_index]
     52         else:
     53             data = self.dataset[possibly_batched_index]

[/usr/local/lib/python3.10/dist-packages/monai/data/dataset.py](https://localhost:8080/#) in __getitem__(self, index)
    110             # dataset[[1, 3, 4]]
    111             return Subset(dataset=self, indices=index)
--> 112         return self._transform(index)
    113 
    114 

[/usr/local/lib/python3.10/dist-packages/monai/data/dataset.py](https://localhost:8080/#) in _transform(self, index)
     96         """
     97         data_i = self.data[index]
---> 98         return apply_transform(self.transform, data_i) if self.transform is not None else data_i
     99 
    100     def __getitem__(self, index: int | slice | Sequence[int]):

[/usr/local/lib/python3.10/dist-packages/monai/transforms/transform.py](https://localhost:8080/#) in apply_transform(transform, data, map_items, unpack_items, log_stats, lazy, overrides)
    169             else:
    170                 _log_stats(data=data)
--> 171         raise RuntimeError(f"applying transform {transform}") from e
    172 
    173 

RuntimeError: applying transform <monai.transforms.compose.Compose object at 0x7f315d5ca4a0>

[KumoLiu]

Hi @tararelan, could you please share a reproducible code then I can take a deep look at it? Please check your data after transforms and run the code step by step to see where the size mismatch is. Thanks.

(Move to discussion, more like a usage issue)