Add data/ar_clean_00000-00.parquet

README.md

@@ -0,0 +1,10 @@
+---
+configs:
+- config_name: default
+  data_files: 'data/*.parquet'
+---
+# Beetle-Data/ar-28B
+
+Pretokenized chunks (`input_ids` = 513-token packed sequences,
+no cross-document bleeding). Sharded incrementally; the marker
+`data/_finalized.json` is committed once all parts are uploaded.

data/ar_clean_00000-00.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:255791fb84e01ef3ed83bb303e119359f4ff57649e271eb96677442e822a4ab6
+size 72753054