Update README.md
Browse files
README.md
CHANGED
|
@@ -43,13 +43,17 @@ The Neurograf v1 models provide a controlled comparison of language-model scalin
|
|
| 43 |
|
| 44 |
### Training Diagnostics
|
| 45 |
|
| 46 |
-
|
|
|
|
|
|
|
| 47 |
|
| 48 |
The training results show consistent improvements in validation bits per byte (BPB) as model size increases.
|
| 49 |
|
| 50 |
### Fixed-Data Scaling
|
| 51 |
|
| 52 |
-
|
|
|
|
|
|
|
| 53 |
|
| 54 |
A power-law fit to the five measured model sizes achieves R² = 0.99999, illustrating the relationship between model capacity and validation performance under a fixed Romanian-language training budget.
|
| 55 |
|
|
|
|
| 43 |
|
| 44 |
### Training Diagnostics
|
| 45 |
|
| 46 |
+
<p align="center">
|
| 47 |
+
<img src="https://cdn-uploads.huggingface.co/production/uploads/67acfe9a42ea14a2b9aeaeff/Pea2tYO4r6ttf63WvAG1S.png" width="60%">
|
| 48 |
+
</p>
|
| 49 |
|
| 50 |
The training results show consistent improvements in validation bits per byte (BPB) as model size increases.
|
| 51 |
|
| 52 |
### Fixed-Data Scaling
|
| 53 |
|
| 54 |
+
<p align="center">
|
| 55 |
+
<img src="https://cdn-uploads.huggingface.co/production/uploads/67acfe9a42ea14a2b9aeaeff/4PBYrApGaoMcKQLCBmNWZ.png" width="60%">
|
| 56 |
+
</p>
|
| 57 |
|
| 58 |
A power-law fit to the five measured model sizes achieves R² = 0.99999, illustrating the relationship between model capacity and validation performance under a fixed Romanian-language training budget.
|
| 59 |
|