Changing number of topics

Nobody has claimed this yet.

### Have you searched existing issues? 🔎

- [x] I have searched and found no existing issues

### Desribe the bug

I have been running BERTopic on a dataset for the Chinese language. I see every time I run the model, I get either two topics or around 315 topics. I'm not sure what is causing this drastic change and which result is acceptable.

### Reproduction

```python
from bertopic import BERTopic

device = 'cuda' if torch.cuda.is_available() else 'cpu'
sentence_model = SentenceTransformer('paraphrase-multilingual-MiniLM-L12-v2', device=device)

umap_model = umap.UMAP(n_neighbors=15, n_components=5, metric='cosine')
topic_model = BERTopic(embedding_model=sentence_model, umap_model=umap_model, language="multilingual")
topics, _ = topic_model.fit_transform(texts)

topic_model.get_topic_info()
# either two or around 315 topics
```

### BERTopic Version

0.16.4

Start with the BERTopic fit_transform call and the configured UMAP model in the reproduction, then inspect get_topic_info() across repeated runs. Reproduce the two-topic and approximately 315-topic outcomes with the same inputs and determine which configuration or behavior causes the variation. Done means identifying the source of the discrepancy and documenting which result is expected.

Written by the indexing model from the issue text.