Declare num_bytes/download_size/dataset_size (HF-computed values) so the
dataset viewer reports coherent size info.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Update corpus to upstream Text-Fabric release v0.2.7: new tablet KTU 2.33
plus revisions to 124 existing tablets (279 tablets, 7,616 lines).
Restructure as a single typed Parquet artifact (data/cuc.parquet) with
tablet/column/line columns and a globally unique row_id; drop the always-empty
entities/relations fields and the 278 per-tablet root JSONL files. Add an
explicit configs + dataset_info block and language tag (uga).
build_parquet.py renders data/cuc.parquet directly from the CUC Text-Fabric
data (no intermediate JSONL), vendoring the line-rendering logic from the
ugaritic-nb exporter.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
