Version number :) (#1)
Browse files- Version number :) (b6cfc11ba37bc858fbcf69320f445a20bd6462d4)
Co-authored-by: algorithm <algorithm@users.noreply.huggingface.co>
README.md
CHANGED
@@ -11,7 +11,7 @@ license: apache-2.0
|
|
11 |
**Creator** [Nicolas Mejia-Petit](https://twitter.com/mejia_petit)
|
12 |
|
13 |
### Overview
|
14 |
-
Just one day after the release of **Mixtral-8x-22b**, we are excited to introduce our handcrafted experimental model, **Mistral-22b-V.
|
15 |
|
16 |
### Capabilities
|
17 |
- **Math Proficiency**: The model exhibits strong mathematical abilities. Dispite not being trained on math.
|
|
|
11 |
**Creator** [Nicolas Mejia-Petit](https://twitter.com/mejia_petit)
|
12 |
|
13 |
### Overview
|
14 |
+
Just one day after the release of **Mixtral-8x-22b**, we are excited to introduce our handcrafted experimental model, **Mistral-22b-V.02**. This model is a culmination of equal knowledge distilled from all experts into a single, dense 22b model. This model is not a single trained expert, rather its a compressed MOE model, turning it into a dense 22b mode. This is the first working MOE to Dense model conversion.
|
15 |
|
16 |
### Capabilities
|
17 |
- **Math Proficiency**: The model exhibits strong mathematical abilities. Dispite not being trained on math.
|