Datasets:
Tasks:
Text Generation
Modalities:
Text
Formats:
parquet
Languages:
English
Size:
10M - 100M
License:
Update README.md
Browse files
README.md
CHANGED
|
@@ -46,15 +46,6 @@ Documents scoring ≥5 are kept.
|
|
| 46 |
| DCLM | 13,144 | ~229 GB | ~18.8M |
|
| 47 |
| FineWeb | 8,800 | ~669 GB | ~47.5M |
|
| 48 |
|
| 49 |
-
Combined ~66M rows before dedup.
|
| 50 |
-
|
| 51 |
-
## Schema
|
| 52 |
-
|
| 53 |
-
```
|
| 54 |
-
text: string
|
| 55 |
-
term_score: int32
|
| 56 |
-
```
|
| 57 |
-
|
| 58 |
## Use case
|
| 59 |
|
| 60 |
-
|
|
|
|
| 46 |
| DCLM | 13,144 | ~229 GB | ~18.8M |
|
| 47 |
| FineWeb | 8,800 | ~669 GB | ~47.5M |
|
| 48 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 49 |
## Use case
|
| 50 |
|
| 51 |
+
Mostly for upsampling agentic-adjacent data during pretraining.
|