---
license: apache-2.0
---

A cleaned version of [CCI3](https://huggingface.co/datasets/BAAI/CCI3-Data) (first 700 files), where:
- All files with encoding errors (can't be read with pandas.read_json) are removed.
- Samples with fewer than 100 characters are removed.
- The jsonl files are converted to parquet for disk space efficiency.