Update dataset card for 30B export layout
Browse files
README.md
CHANGED
|
@@ -1,19 +1,20 @@
|
|
| 1 |
---
|
| 2 |
license: odc-by
|
| 3 |
-
pretty_name: Parameter Golf FineWeb
|
| 4 |
source_datasets:
|
| 5 |
- HuggingFaceFW/fineweb
|
| 6 |
---
|
| 7 |
|
| 8 |
-
# Parameter Golf FineWeb
|
| 9 |
|
| 10 |
-
This repository hosts tokenizer-matched export artifacts derived from [`HuggingFaceFW/fineweb`](https://huggingface.co/datasets/HuggingFaceFW/fineweb), specifically from the
|
| 11 |
|
| 12 |
The repository contains:
|
| 13 |
|
| 14 |
-
- pretokenized training and validation shards under `
|
| 15 |
-
- tokenizer artifacts under `
|
| 16 |
-
- the export manifest at `
|
|
|
|
| 17 |
|
| 18 |
## License
|
| 19 |
|
|
|
|
| 1 |
---
|
| 2 |
license: odc-by
|
| 3 |
+
pretty_name: Parameter Golf FineWeb Export
|
| 4 |
source_datasets:
|
| 5 |
- HuggingFaceFW/fineweb
|
| 6 |
---
|
| 7 |
|
| 8 |
+
# Parameter Golf FineWeb Export
|
| 9 |
|
| 10 |
+
This repository hosts tokenizer-matched export artifacts derived from [`HuggingFaceFW/fineweb`](https://huggingface.co/datasets/HuggingFaceFW/fineweb), specifically a 30B subset pulled from the 100B FineWeb cut used for parameter-golf experiments.
|
| 11 |
|
| 12 |
The repository contains:
|
| 13 |
|
| 14 |
+
- pretokenized training and validation shards under `datasets/datasets/`
|
| 15 |
+
- tokenizer artifacts under `datasets/tokenizers/`
|
| 16 |
+
- the export manifest at `datasets/manifest.json`
|
| 17 |
+
- selected-document metadata at `datasets/docs_selected.jsonl`
|
| 18 |
|
| 19 |
## License
|
| 20 |
|