[BUG] Error when training Tide with more than 1 unknown reals

[mrgreen3325]

Describe the bug

[561](https://vscode-remote+ssh-002dremote-002b192-002e168-002e8-002e123.vscode-resource.vscode-cdn.net/opt/conda/envs/py39_pytorch222/lib/python3.9/site-packages/lightning/pytorch/trainer/trainer.py:561) call._call_and_handle_interrupt(

[562](https://vscode-remote+ssh-002dremote-002b192-002e168-002e8-002e123.vscode-resource.vscode-cdn.net/opt/conda/envs/py39_pytorch222/lib/python3.9/site-packages/lightning/pytorch/trainer/trainer.py:562)     self, self._fit_impl, model, train_dataloaders, val_dataloaders, datamodule, ckpt_path
[563](https://vscode-remote+ssh-002dremote-002b192-002e168-002e8-002e123.vscode-resource.vscode-cdn.net/opt/conda/envs/py39_pytorch222/lib/python3.9/site-packages/lightning/pytorch/trainer/trainer.py:563) )

File /opt/conda/envs/py39_pytorch222/lib/python3.9/site-packages/lightning/pytorch/trainer/call.py:48, in _call_and_handle_interrupt(trainer, trainer_fn, *args, **kwargs)
46 if trainer.strategy.launcher is not None:
47 return trainer.strategy.launcher.launch(trainer_fn, *args, trainer=trainer, **kwargs)
---> 48 return trainer_fn(*args, **kwargs)
50 except _TunerExitException:
51 _call_teardown_hook(trainer)
...
267 x_dynamic_future_covariates = self.future_cov_projection(
268 x_dynamic_future_covariates
269 )

RuntimeError: Sizes of tensors must match except in dimension 1. Expected size 4 but got size 2 for tensor number 1 in the list.
-->

To Reproduce

tsdataset = TimeSeriesDataSet(
    training_data,
    time_idx="time_idx",
    target="y",  # Single target variable
    group_ids=["group"],
    min_encoder_length=max_encoder_length,
    max_encoder_length=max_encoder_length,
    min_prediction_length=max_prediction_length,
    max_prediction_length=max_prediction_length,
    static_categoricals=[],
    static_reals=[],
    time_varying_known_reals=["time_idx"]+["day"],
    time_varying_unknown_reals= ["y","temperature","humidity"],  # Include both inputs and target
    add_relative_time_idx=False,
    add_target_scales=False,
)

model = TiDEModel.from_dataset(
    tsdataset,
    num_encoder_layers = 32,
    num_decoder_layers = 32,
    learning_rate=0.001,
    hidden_size=32,
    dropout=0.1,
    # loss=SMAPE(),
    loss=RMSE(),
    optimizer="adam"
)
batch_size = 32
train_dataloader = tsdataset.to_dataloader(train=True, batch_size=batch_size, num_workers=8)

Expected behavior

Additional context

Versions
pytorch_forecasting=1.3.0
lightning = 2.5.1
pytorch=2.4.0
python=3.9