Add 100M-token wol_Latn slice from fineweb-2
Browse files- README.md +10 -0
- data/train-00000-of-00001.parquet +3 -0
README.md
ADDED
|
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# FineWeb2-wol_Latn-100M
|
| 2 |
+
|
| 3 |
+
100M-token slice of HuggingFaceFW/fineweb-2 (wol_Latn) for LAS tokenizer training under the MultilingualUnigramLM org.
|
| 4 |
+
|
| 5 |
+
- Source: HuggingFaceFW/fineweb-2
|
| 6 |
+
- Config: wol_Latn
|
| 7 |
+
- Documents: 6,621
|
| 8 |
+
- Whitespace-tokens: 3,286,701
|
| 9 |
+
- Chars: 17,776,324
|
| 10 |
+
- Format: single `data/train-00000-of-00001.parquet`, schema `(text: string)`. Mirrors the `LangMap-TheStack-*-100M` layout.
|
data/train-00000-of-00001.parquet
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:de36238e722d0363309eff1865466708f27edadd9063b24f83c6f4eb260142df
|
| 3 |
+
size 8722427
|