Update README.md
Browse files
README.md
CHANGED
|
@@ -3,10 +3,11 @@ language:
|
|
| 3 |
- en
|
| 4 |
---
|
| 5 |
A tiny 55 million parameter trained on 1.3 billion tokens using a custom dataset mixture.
|
|
|
|
| 6 |
| Dataset | Weight |
|
| 7 |
|---|---|
|
| 8 |
| `HuggingFaceFW/fineweb-edu` | 50% |
|
| 9 |
| `epfml/FineWeb-HQ` | 30% |
|
| 10 |
| `HuggingFaceTB/cosmopedia` (stories split) | 20% |
|
| 11 |
-
|
| 12 |
The tokenizer is a basic bpe tokenizer that was trained on a smaller subset of 80_000 samples of this same data mixture with a vocab size of 8000.
|
|
|
|
| 3 |
- en
|
| 4 |
---
|
| 5 |
A tiny 55 million parameter trained on 1.3 billion tokens using a custom dataset mixture.
|
| 6 |
+
|
| 7 |
| Dataset | Weight |
|
| 8 |
|---|---|
|
| 9 |
| `HuggingFaceFW/fineweb-edu` | 50% |
|
| 10 |
| `epfml/FineWeb-HQ` | 30% |
|
| 11 |
| `HuggingFaceTB/cosmopedia` (stories split) | 20% |
|
| 12 |
+
|
| 13 |
The tokenizer is a basic bpe tokenizer that was trained on a smaller subset of 80_000 samples of this same data mixture with a vocab size of 8000.
|