Datasets:
Update README.md
Browse files
README.md
CHANGED
@@ -30,4 +30,26 @@ configs:
|
|
30 |
data_files:
|
31 |
- split: train
|
32 |
path: data/train-*
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
33 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
30 |
data_files:
|
31 |
- split: train
|
32 |
path: data/train-*
|
33 |
+
license: odc-by
|
34 |
+
task_categories:
|
35 |
+
- text-generation
|
36 |
+
language:
|
37 |
+
- en
|
38 |
+
size_categories:
|
39 |
+
- 10K<n<100K
|
40 |
---
|
41 |
+
|
42 |
+
# FineWeb 10MT
|
43 |
+
Roughly 10 million tokens worth of English FineWeb data. Obtained by taking the [10BT Sample](https://huggingface.co/datasets/HuggingFaceFW/fineweb/viewer/sample-10BT) of [FineWeb](https://huggingface.co/datasets/HuggingFaceFW/fineweb), shuffling it, sharding into 1000 shards, and selecting the first shard.
|
44 |
+
|
45 |
+
## Code
|
46 |
+
To reproduce FineWeb 10MT, simply use the following:
|
47 |
+
|
48 |
+
```python
|
49 |
+
from datasets import load_dataset
|
50 |
+
|
51 |
+
fineweb = load_dataset("HuggingFaceFW/fineweb", "sample-10BT", split="train").shuffle().shard(num_shards=1000, index=0)
|
52 |
+
|
53 |
+
token = "token"
|
54 |
+
fineweb.push_to_hub("OxxoCodes/fineweb-10MT", token=token)
|
55 |
+
```
|