Sören Wacker

That would be amazing. I implemented something like that, using a Flask server for the auth, before, for a scientific application, but looks quite bumpy, and I found some security...

I tried replacing Input("pkl-table", "data"), with Input("pkl-table", "dataSorted"), but that does not change anything.

Same problem. Even when serving at / I just get a blank page.

https://discuss.streamlit.io/t/streamlit-0-72-0-behind-nginx-connecting-forever/7913

I have something similar with pycaret 2.3.10. It seems pycaret fails to remove the feature used as fold_groups from the features list and uses it as a categorical feature. The...

Compare this to the case when providing the groups directly:  However, this also does not work:  Notice that using `GroupKFold` forces the user to use `imputation_type='iterative'`. Maybe I...

Apparently, the problem here was the use of `data_split_stratify=True`. Setting it to `False` finally generated models.

Actually this happens also with binary classification sometimes. It looks like it is not working for 'lightgbm' but working for 'catboost'.

It would be nice to get an error message when save failed. I had lost some edits due to this because I falsely assumed the file was saved.

Thanks. Such a converter would be a great add on. #featurerequest