fix: Conversion of tabular dataset to tensors

src/safeds/data/labeled/containers/_image_dataset.py

@@ -294,7 +294,10 @@ def __init__(self, table: Table) -> None:
         _init_default_device()
 
         self._column_names = table.column_names
-        self._tensor = torch.Tensor(table._data_frame.to_torch()).to(_get_device())
+        if table.number_of_rows == 0:
+            self._tensor = torch.empty((0, table.number_of_columns), dtype=torch.float32).to(_get_device())
+        else:
+            self._tensor = table._data_frame.to_torch().to(_get_device())
 
         if not torch.all(self._tensor.sum(dim=1) == torch.ones(self._tensor.size(dim=0))):
             raise ValueError(

src/safeds/data/labeled/containers/_time_series_dataset.py

@@ -322,8 +322,8 @@ def _create_dataset(features: torch.Tensor, target: torch.Tensor) -> Dataset:
 
     class _CustomDataset(Dataset):
         def __init__(self, features_dataset: torch.Tensor, target_dataset: torch.Tensor):
-            self.X = features_dataset
-            self.Y = target_dataset.unsqueeze(-1)
+            self.X = features_dataset.float()
+            self.Y = target_dataset.unsqueeze(-1).float()
             self.len = self.X.shape[0]
 
         def __getitem__(self, item: int) -> tuple[torch.Tensor, torch.Tensor]:
@@ -341,8 +341,8 @@ def _create_dataset_predict(features: torch.Tensor) -> Dataset:
     _init_default_device()
 
     class _CustomDataset(Dataset):
-        def __init__(self, features: torch.Tensor):
-            self.X = features
+        def __init__(self, datas: torch.Tensor):
+            self.X = datas.float()
             self.len = self.X.shape[0]
 
         def __getitem__(self, item: int) -> torch.Tensor:

src/safeds/ml/nn/_output_conversion_time_series.py

@@ -78,7 +78,7 @@ def _data_conversion(self, input_data: TimeSeriesDataset, output_data: Tensor, *
         window_size: int = kwargs["window_size"]
         forecast_horizon: int = kwargs["forecast_horizon"]
         input_data_table = input_data.to_table()
-        input_data_table = input_data_table.slice_rows(window_size + forecast_horizon)
+        input_data_table = input_data_table.slice_rows(start=window_size + forecast_horizon)
 
         return input_data_table.add_columns(
             [Column(self._prediction_name, output_data.tolist())],

tests/safeds/data/labeled/containers/_time_series_dataset/test_into_dataloader.py

@@ -175,7 +175,7 @@ def test_should_create_dataloader_invalid(
             1,
             0,
             OutOfBoundsError,
-            r"forecast_horizon \(=0\) is not inside \[1, \u221e\).",
+            None,
         ),
         (
             Table(
@@ -189,7 +189,7 @@ def test_should_create_dataloader_invalid(
             0,
             1,
             OutOfBoundsError,
-            r"window_size \(=0\) is not inside \[1, \u221e\).",
+            None,
         ),
     ],
     ids=[
@@ -204,7 +204,7 @@ def test_should_create_dataloader_predict_invalid(
     window_size: int,
     forecast_horizon: int,
     error_type: type[ValueError],
-    error_msg: str,
+    error_msg: str | None,
     device: Device,
 ) -> None:
     configure_test_with_device(device)

tests/safeds/ml/nn/test_forward_workflow.py

@@ -23,12 +23,8 @@ def test_forward_model(device: Device) -> None:
         path=resolve_resource_path(_inflation_path),
     )
     table_1 = table_1.remove_columns(["date"])
-    table_2 = table_1.slice_rows(length=table_1.number_of_rows - 14)
-    table_2 = table_2.add_columns(
-        [
-            table_1.slice_rows(start=14).get_column("value").rename("target"),
-        ]
-    )
+    table_2 = table_1.slice_rows(start=0, length=table_1.number_of_rows - 14)
+    table_2 = table_2.add_columns([(table_1.slice_rows(start=14)).get_column("value").rename("target")])
     train_table, test_table = table_2.split_rows(0.8)
 
     ss = StandardScaler()