docs: correct verified parameter count
Browse files
docs/MODEL_SPECIFICATION.md
CHANGED
|
@@ -3,7 +3,7 @@
|
|
| 3 |
Neo is an English-only System One decision model.
|
| 4 |
|
| 5 |
- Bidirectional Transformer encoder.
|
| 6 |
-
-
|
| 7 |
- 8 layers, hidden size 512, 8 attention heads.
|
| 8 |
- Intermediate size 2048.
|
| 9 |
- Maximum context 1,024 tokens.
|
|
|
|
| 3 |
Neo is an English-only System One decision model.
|
| 4 |
|
| 5 |
- Bidirectional Transformer encoder.
|
| 6 |
+
- 41,391,654 trainable parameters (approximately 41.4M).
|
| 7 |
- 8 layers, hidden size 512, 8 attention heads.
|
| 8 |
- Intermediate size 2048.
|
| 9 |
- Maximum context 1,024 tokens.
|