NaNs detected in the static data

[atulkakrana]

Description

Using timegan and timevae for timeseries / sequential data leads to an error:

File ~/miniconda3/envs/syncity/lib/python3.10/site-packages/synthcity/plugins/core/models/ts_model.py:251, in TimeSeriesModel.forward(self, static_data, temporal_data, observation_times)
    240 @validate_arguments(config=dict(arbitrary_types_allowed=True))
    241 def forward(
    242     self,
   (...)
    247     # x shape (batch, time_step, input_size)
    248     # r_out shape (batch, time_step, output_size)
    250     if torch.isnan(static_data).sum() != 0:
--> 251         raise ValueError("NaNs detected in the static data")
    252     if torch.isnan(temporal_data).sum() != 0:
    253         raise ValueError("NaNs detected in the temporal data")

ValueError: NaNs detected in the static data

How to Reproduce

model_type = "timevae" 
trn_n_iter = 20
trn_batch_size  = 12
trn_n_hidden_units = 100

ts_syn_model = Plugins().get(model_type, 
    n_iter=trn_n_iter,             # Number of training iterations
    batch_size=trn_batch_size,     # Batch size
    generator_n_units_hidden = trn_n_hidden_units,     # Number of units in the hidden layers of the generator; default: 500
    discriminator_n_units_hidden = trn_n_hidden_units, # Number of units in the hidden layers of the discriminator; default: 500,


ts_syn_model.fit(data)
    )

System Information

• OS: ubuntu-noble-24.04-arm64-server
• OS Version: 24.04
• Language Version: Python 3.10

Additional Context

Add any other context about the problem here.