Create new file
Browse files
README.md
ADDED
|
@@ -0,0 +1,18 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language:
|
| 3 |
+
- de
|
| 4 |
+
task_categories:
|
| 5 |
+
- text-generation
|
| 6 |
+
size_categories:
|
| 7 |
+
- 100k<n<1M
|
| 8 |
+
task_ids:
|
| 9 |
+
- language-modeling
|
| 10 |
+
- masked-language-modeling
|
| 11 |
+
---
|
| 12 |
+
German validation dataset from WECHSEL () to evaluate LLM perplexity.
|
| 13 |
+
|
| 14 |
+
JSON-line files (on JSON object per line):
|
| 15 |
+
|
| 16 |
+
- `valid.json.gz`: Gzipped validation set as generated by the paper (163,698 docs)
|
| 17 |
+
- `valid.random_1636.json.gz`: Random 1% (1636 docs) of the validation set
|
| 18 |
+
|