Remove old shard-named files

Rename validation shard to validation.parquet

Rename train shard to train.parquet

Remove old 16-shard Parquet files

Use streaming ParquetWriter for single-file output

Single shard per split with page index for random access

Add page index and content-defined chunking to reshard script

Add page index and content-defined chunking to conversion scripts

Re-shard with page index and smaller row groups for dataset viewer compatibility

Remove old 307-shard Parquet files

Add reshard script

Add conversion scripts

Re-shard to 16 larger Parquet files (1M rows each)

Add LICENSE from original dataset

Add dataset card

Convert JSONL to sharded Parquet for dataset viewer compatibility

initial commit