Incomprehensible behavior of AutoRegressiveBaseModelWithCovariates in Pytorch

Question

When I try to create an AutoRegressiveBaseModelWithCovariates from the Pytorch Module with the Method .from_dataset with an TimeseriesDataset(in this case called "train_dataset") i always get the following error although i dont have the keyword static_categoricals in my TimeseriesDataset

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Cell In[10], line 1
----> 1 model = AutoRegressiveBaseModelWithCovariates.from_dataset(dataset=train_dataset, 
      2                                             #    hidden_size=HIDDEN,
      3                                             #    lstm_layers=LSTMLAYERS,
      4                                             #    dropout=DROPOUT,
      5                                             #    output_size=OUTPUT_SIZE,
      6                                             #    loss=QuantileLoss(quantiles=QUANTILES),
      7                                             #    attention_head_size=ATTHEADS,
      8                                             #    max_encoder_length=ENCODER_LEN_MAX,
      9                                             #    hidden_continuous_size=HIDDEN_CONTINIOUS,
     10                                                learning_rate=LEARNING_RATE,
     11                                                log_interval=LOG_INTERVAL,
     12                                                reduce_on_plateau_patience = PLATEU_PATIENCE,
     13                                                optimizer='adam',# weigth_decay
     14                                                )
     16 #todo Autoregressionmodel while handing over **kwargs gets error in BaseModel from_dataset function although it is working on TemporalFusionTransformer??
     18 print(f"Number of parameters in TFT network: {model.size()/1e3:.1f}k")

File c:\Users\juo7ho\.conda\envs\hyperparam_pytorch\lib\site-packages\pytorch_forecasting\models\base_model.py:1679, in BaseModelWithCovariates.from_dataset(cls, dataset, allowed_encoder_known_variable_names, **kwargs)
   1675 new_kwargs.update(kwargs)
   1676 # print("basewithcovariates_from")
   1677 # print(new_kwargs)
   1678 #super is BaseModel
...
-> 1242 net = cls(**kwargs)
   1243 net.dataset_parameters = dataset.get_parameters()
   1244 if dataset.multi_target:

TypeError: BaseModel.__init__() got an unexpected keyword argument 'static_categoricals'

This my Dataset:

train_dataset = TimeSeriesDataSet(data=SG.df.loc[SG.df['Usecase'] == 'Train'],                          # dataframe with sequence data
                                  time_idx='Sequence_idx',                                              # integer column denoting the time index
                                  target='a',                                                      # column/list denoting the target                                                  
                                  group_ids=['Sequence_ID'],                                            # column names that identifying a time series
                                  max_encoder_length=ENCODER_LEN_MAX,                                   # maximum length to encode
                                  min_encoder_length=ENCODER_LEN_MIN,                                   # minimum allowed length to encode
                                  max_prediction_length=PREDICTION_LEN_MAX,                             # maximum prediction/decoder length
                                  min_prediction_length=PREDICTION_LEN_MIN,                             # minimum prediction/decoder length
                                  time_varying_known_reals=["size"],          # continuous variables that change over time and are known in the future
                                  time_varying_unknown_reals=['a',                                 # continuous variables that change over time and are NOT known in the future
                                                              'b', 
                                                              'c',
                                                              'd',
                                                              ],
                                  allow_missing_timesteps=False,                                          # allow missing timesteps that are automatically filled up
                                  add_relative_time_idx=False,                                            # add a relative time index as feature
                                  add_target_scales=False,                                               # add scales for target to static real features
                                  add_encoder_length=False,                                            # add decoder length to list of static real variables
                                  target_normalizer=TorchNormalizer(),           
                                  predict_mode=False                                                     # this will take choose for each time series identified by group_ids the last max_prediction_length samples of each time series as prediction samples and everthing previous up to max_encoder_length samples as encoder samples
                                  )

I tried using a different Class from Pytorch named TemporalFusionTransformer and with that it worked flawlessly. I looked at the Code from Pytorch and I noticed that AutoRegressiveBaseModelWithCovariates inherets from BaseModelWithCovariates which inherets from BaseModel. At the end the BaseModel __init__ Method is called with the **kwargs where static cathegorials are in and throws an error mentiont above because the BaseModel does not have static cathegorials. Am I doing something wrong or is this a bug in the Pytorch Libary?

Thanks in advance.