[Bug]: Segmentation faul when

### What happened?

在运行时
collection.add(
documents=docs,
metadatas=metadatas,
ids=ids
)
这一句的时候报错Segmentation fault，没有其他提示，尝试不使用embedding是可以的，另外内存没有溢出，减小batch也不行

### Versions

python 10.0
torch 2.4.0
tqdm 4.66.4
transformers 4.43.3
chromadb 1.0.16

### Relevant log output

```shell
2025-08-14 17:02:49,169 - INFO - PyTorch version 2.4.0 available.
2025-08-14 17:02:49,893 - INFO - Load pretrained SentenceTransformer: BAAI/bge-m3
2025-08-14 17:02:58,529 - INFO - Anonymized telemetry enabled. See
https://docs.trychroma.com/telemetry for more information.
2025-08-14 17:02:58,688 - INFO - Loaded existing collection: concepts_only_name
2025-08-14 17:02:58,689 - INFO - Loading data from CSV
Processing batches: 0%| | 0/1 [00:00

Start with the reported collection.add(documents=docs, metadatas=metadatas, ids=ids) call and reproduce it using the listed Python, PyTorch, transformers, and Chroma versions. Compare the embedding-enabled path with the no-embedding path and the logged batch; done means identifying a reproducible failure cause or a confirmed fix with a regression test or minimal reproduction.

Written by the indexing model from the issue text.