Add verbosity TVAE (progress bar + save the loss values)

[npatki]

Problem Description

Similar to #298, we should also add a verbosity option for TVAE that shows the user the fitting progress per epoch. After fitting, it should be possible to see the loss values.

Expected behavior

1. Add a verbose parameter to TVAE (just like CTGAN). By default it can be False
2. When set to True, increment a progress bar per (epoch x data iteration). Also show/refresh the loss value, similar to Add progress bar for CTGAN fitting (+ save the loss values) #298 (except there is only 1 loss value)
3. Always save the loss values that correspond to the last fitting iteration. The columns can now be Epoch, Batch and Loss.

>>> model = TVAE(verbose=True, epochs=100)
>>> model.fit()
Fitting TVAE ...
Loss 0.024 |████████████████████████████         | 750/1000 [00:33<00:10, 229.00it/s]

# after modeling
>>> model.loss_values
   Epoch     Batch    Loss 
0      1     1        1.7863
1      1     2        1.5484
2      1     3        1.3633

[TkinterinShanghai]

Did you remove this in TVAESynthesizer? This is probably quite a useful feature

[npatki]

Hi @TkinterinShanghai nothing should be removed in TVAESynthesizer. If you are having any problems, would you be able to file a new issue describing the code you're trying, errors you are seeing, etc.

[TkinterinShanghai]

According to the docs there is no verbose option. It doesnt seem to be present in the code either, based on what I saw in the docstring

[npatki]

Hi @TkinterinShanghai, I see what's going on.

We are currently discussing an issue in the CTGAN library, which implements some raw ML algorithms for synthetic data (such as CTGAN and TVAE). You are referring to API and documentation in the SDV library, which wraps around these algorithms (TVAESynthesizer)

If you use the code within this current repo, the verbose option is available. However, the TVAESynthesizer is a wrapper in a different library, where we haven't yet exposed the verbose parameter. I will add a feature request for it in SDV.

[TkinterinShanghai]

Awesome, thanks!