Update README.md
Browse files
README.md
CHANGED
|
@@ -23,7 +23,7 @@ tags:
|
|
| 23 |
|
| 24 |
# Tiny-Word
|
| 25 |
|
| 26 |
-
Tiny-Word is an extremely tiny Mistral-like model, approximately ~
|
| 27 |
|
| 28 |
## Architecture
|
| 29 |
|
|
@@ -78,7 +78,7 @@ Tiny-Word was trained on Google Colaboratory, with 1 Nvidia Tesla T4 GPU, 15 GB
|
|
| 78 |
| 12000 | 3.5139 | 3.5161 | ~33.6 | ~33.7 |
|
| 79 |
| 15000 | 3.4784 | 3.4861 | ~32.4 | ~32.6 |
|
| 80 |
|
| 81 |
-
Tiny-Word shows promising results, even at its tiny size (~
|
| 82 |
|
| 83 |
## Generation Examples
|
| 84 |
|
|
|
|
| 23 |
|
| 24 |
# Tiny-Word
|
| 25 |
|
| 26 |
+
Tiny-Word is an extremely tiny Mistral-like model, approximately ~134k parameters. It generates English or Spanish words or word-like sequences.
|
| 27 |
|
| 28 |
## Architecture
|
| 29 |
|
|
|
|
| 78 |
| 12000 | 3.5139 | 3.5161 | ~33.6 | ~33.7 |
|
| 79 |
| 15000 | 3.4784 | 3.4861 | ~32.4 | ~32.6 |
|
| 80 |
|
| 81 |
+
Tiny-Word shows promising results, even at its tiny size (~134k parameters). Given the relatively easy task (predicting subwords inside single words), this is expected.
|
| 82 |
|
| 83 |
## Generation Examples
|
| 84 |
|