bertopic version 0.16.0 - probs are empty when executing with zero_shot

Nobody has claimed this yet.

when executing zero shot. the following probs is empty:
topics, probs = topic_model.fit_transform(docs, embeddings)

following configuration:
from umap import UMAP

umap_model = UMAP(n_neighbors=15, n_components=5, min_dist=0.0, metric='cosine', random_state=42)
from hdbscan import HDBSCAN

hdbscan_model = HDBSCAN(min_cluster_size=200, metric='euclidean', cluster_selection_method='eom', prediction_data=True)
from sklearn.feature_extraction.text import CountVectorizer
vectorizer_model = CountVectorizer(stop_words="english")
zero_shot_topics_list=["vitiligo"]

embedding_model_name="thenlper/gte-base"
embedding_model = SentenceTransformer(embedding_model_name)
embeddings = embedding_model.encode(docs, show_progress_bar=True)
topic_model = BERTopic(

# Pipeline models
embedding_model=embedding_model,
umap_model=umap_model,
hdbscan_model=hdbscan_model,
vectorizer_model = CountVectorizer(stop_words="english")
zeroshot_topic_list=zero_shot_topics_list,
zeroshot_min_similarity=.8,
calculate_probabilities=True,
# representation_model=representation_model,
# Hyperparameters
top_n_words=10,
verbose=True
)

topics, probs = topic_model.fit_transform(docs, embeddings)

Start by reproducing the reported BERTopic 0.16.0 configuration with zero-shot topics, calculate_probabilities=True, and the provided fit_transform call. Trace how zero-shot assignment produces the probs result; done means the call returns the expected probability values instead of an empty result, with regression coverage for this configuration.

Written by the indexing model from the issue text.