| # Genesis Parquet Pack |
| **Developed by: Within Us AI** |
| Generated: 2026-01-01 |
|
|
| This pack includes: |
| - A converter script to turn JSONL splits into **Parquet shards** suitable for fast scanning/streaming. |
| - Recommended sharding conventions (train-00000.parquet, etc.). |
|
|
| Why Parquet: |
| - Faster scans |
| - Better compatibility with dataset viewers and downstream pipelines |
| - Easy sharding and parallel reads |
|
|
| ## Usage |
| ```bash |
| pip install datasets pyarrow pandas |
| python jsonl_to_parquet_shards.py --jsonl train.jsonl --out_dir data --rows_per_shard 5000 |
| ``` |
|
|
