Update README.md
Browse files
README.md
CHANGED
|
@@ -1,25 +1,27 @@
|
|
| 1 |
-
---
|
| 2 |
-
pretty_name: test strict shuffled
|
| 3 |
-
task_categories:
|
| 4 |
-
- text-classification
|
| 5 |
-
tags:
|
| 6 |
-
- parquet
|
| 7 |
-
- shuffled
|
| 8 |
-
- exact-shuffle
|
| 9 |
-
---
|
| 10 |
-
|
| 11 |
-
# Strict shuffled copy of `KantaHayashiAI/test`
|
| 12 |
-
|
| 13 |
-
This dataset was produced by assigning each source row a deterministic pseudo-random sort key
|
| 14 |
-
derived from:
|
| 15 |
-
|
| 16 |
-
- source parquet path
|
| 17 |
-
- row index inside that parquet
|
| 18 |
-
- shuffle seed `2026-04-01-strict-shuffle-v1`
|
| 19 |
-
|
| 20 |
-
Rows were first partitioned into `512` buckets by the high bits of the key,
|
| 21 |
-
then each bucket was fully sorted by `(__key_hi, __key_lo, __file_id, __row_idx)`.
|
| 22 |
-
This yields a deterministic global shuffled order without requiring the full dataset
|
| 23 |
-
to be materialized twice on local disk.
|
| 24 |
-
|
| 25 |
-
Expected train shard count: `512`.
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
pretty_name: test strict shuffled
|
| 3 |
+
task_categories:
|
| 4 |
+
- text-classification
|
| 5 |
+
tags:
|
| 6 |
+
- parquet
|
| 7 |
+
- shuffled
|
| 8 |
+
- exact-shuffle
|
| 9 |
+
---
|
| 10 |
+
|
| 11 |
+
# Strict shuffled copy of `KantaHayashiAI/test`
|
| 12 |
+
|
| 13 |
+
This dataset was produced by assigning each source row a deterministic pseudo-random sort key
|
| 14 |
+
derived from:
|
| 15 |
+
|
| 16 |
+
- source parquet path
|
| 17 |
+
- row index inside that parquet
|
| 18 |
+
- shuffle seed `2026-04-01-strict-shuffle-v1`
|
| 19 |
+
|
| 20 |
+
Rows were first partitioned into `512` buckets by the high bits of the key,
|
| 21 |
+
then each bucket was fully sorted by `(__key_hi, __key_lo, __file_id, __row_idx)`.
|
| 22 |
+
This yields a deterministic global shuffled order without requiring the full dataset
|
| 23 |
+
to be materialized twice on local disk.
|
| 24 |
+
|
| 25 |
+
Expected train shard count: `512`.
|
| 26 |
+
|
| 27 |
+
This dataset was created using the Supermicro ARS-111GL-DNHR-LCC and FUJITSU Server PRIMERGY CX2550 M7 (Miyabi) at the Joint Center for Advanced High Performance Computing (JCAHPC).
|