Issue with data formatting for TimeSeriesDataLoader ("outcome" input)

[codingkimchi]

Description

Hello,
I've been trying out the TimeSeriesDataLoader with my custom dataset and I found that when my "outcome" input is set to None, the model.fit(loader) process would fail, and throw the following error:

[/usr/local/lib/python3.11/dist-packages/synthcity/plugins/core/models/ts_model.py](https://localhost:8080/#) in dataloader(self, static_data, temporal_data, observation_times, outcome)
    426         stratify = None
    427         _, out_counts = torch.unique(outcome, return_counts=True)
--> 428         if out_counts.min() > 1:
    429             stratify = outcome.cpu()
    430 

RuntimeError: min(): Expected reduction dim to be specified for input.numel() == 0. Specify the reduction dim with the 'dim' argument.

To reproduce this, I tried to follow the recommended tutorial notebook, tutorial6_time_series_data_preparation.ipynb

Tutorial 6 Linked here: https://github.com/vanderschaarlab/synthcity/blob/main/tutorials/tutorial6_time_series_data_preparation.ipynb

So, using the same tutorial notebook, I set "outcome" and "static_data" to None, after doing the data preparation, to test this out. The same error occured.

I only managed to try this for the 'timevae' model so far, I'm not sure if the error is present for the other time series models as well.

I read the synthcity documentation on the inputs to the TimeSeriesDataLoader module, and it states that the "outcome" and "static_data" inputs are optional, and default to None.

Synthcity documentation linked here (please refer to TimeSeriesDataLoader specifically):
https://synthcity.readthedocs.io/en/latest/generated/synthcity.plugins.core.dataloader.html

How to Reproduce

Please refer to the github gist here which demonstrates how I produced the error.
https://gist.github.com/codingkimchi/fbac549c8eafc0cf977b52df10d9e5fe

Expected Behavior

I expected 'timevae' to train even with no specified 'outcome' input.

Notes

Apologies in advance if this issue was addressed before. I saw some advice on a separate Github Issue that all inputs to TimeSeriesDataLoader must be the same length, but this issue was not addressed, to the best of my knowledge!