Supervised Topic Modelling unable to produce same output using fit_transform() and transform()

Nobody has claimed this yet.

Hi,
I referenced the code from https://maartengr.github.io/BERTopic/getting_started/supervised/supervised.html for BERTopic Supervised modelling. Although I was able to obtain the correct output using fit_transform(docs, y=y), I was unable to obtain the same output using transform(docs) even though the inputted doc is the same. Most of the topics generated were labelled -1.

May I know what has gone wrong?

For reference, this is my model code:
empty_dimensionality_model = BaseDimensionalityReduction()
clf = LogisticRegression()
ctfidf_model = ClassTfidfTransformer(reduce_frequent_words=True)

topic_model = BERTopic(
umap_model=empty_dimensionality_model,
hdbscan_model=clf,
ctfidf_model=ctfidf_model,
n_gram_range=(1,3))

Thank you! Any help would be truly appreciated.

Start with the supervised BERTopic example linked in the issue and compare the shown fit_transform(docs, y=y) and transform(docs) calls using the provided model configuration. Investigate why transform assigns many topics as -1; done means the behavior is explained and the same input produces consistent topic assignments, with a regression test if the repository provides one.

Written by the indexing model from the issue text.