Update README.md
Browse files
README.md
CHANGED
|
@@ -4,7 +4,7 @@ license: mit
|
|
| 4 |
|
| 5 |
### Model Description
|
| 6 |
|
| 7 |
-
This model is a mixture of experts merge consisting of 3 mistral based models
|
| 8 |
|
| 9 |
base model, **openchat/openchat-3.5-0106**
|
| 10 |
|
|
@@ -12,6 +12,42 @@ code expert, **beowolx/CodeNinja-1.0-OpenChat-7B**
|
|
| 12 |
|
| 13 |
math expert, **meta-math/MetaMath-Mistral-7B**
|
| 14 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 15 |
### Usage
|
| 16 |
|
| 17 |
```python
|
|
|
|
| 4 |
|
| 5 |
### Model Description
|
| 6 |
|
| 7 |
+
This model is a mixture of experts merge consisting of 3 mistral based models:
|
| 8 |
|
| 9 |
base model, **openchat/openchat-3.5-0106**
|
| 10 |
|
|
|
|
| 12 |
|
| 13 |
math expert, **meta-math/MetaMath-Mistral-7B**
|
| 14 |
|
| 15 |
+
This is the config used in the merging process:
|
| 16 |
+
``` yaml
|
| 17 |
+
base_model: openchat/openchat-3.5-0106
|
| 18 |
+
experts:
|
| 19 |
+
- source_model: openchat/openchat-3.5-0106
|
| 20 |
+
positive_prompts:
|
| 21 |
+
- "chat"
|
| 22 |
+
- "assistant"
|
| 23 |
+
- "tell me"
|
| 24 |
+
- "explain"
|
| 25 |
+
- "I want"
|
| 26 |
+
- source_model: beowolx/CodeNinja-1.0-OpenChat-7B
|
| 27 |
+
positive_prompts:
|
| 28 |
+
- "code"
|
| 29 |
+
- "python"
|
| 30 |
+
- "javascript"
|
| 31 |
+
- "programming"
|
| 32 |
+
- "algorithm"
|
| 33 |
+
- "C#"
|
| 34 |
+
- "C++"
|
| 35 |
+
- "debug"
|
| 36 |
+
- "runtime"
|
| 37 |
+
- "html"
|
| 38 |
+
- "command"
|
| 39 |
+
- "nodejs"
|
| 40 |
+
- source_model: meta-math/MetaMath-Mistral-7B
|
| 41 |
+
positive_prompts:
|
| 42 |
+
- "reason"
|
| 43 |
+
- "math"
|
| 44 |
+
- "mathematics"
|
| 45 |
+
- "solve"
|
| 46 |
+
- "count"
|
| 47 |
+
- "calculate"
|
| 48 |
+
- "arithmetic"
|
| 49 |
+
- "algebra"
|
| 50 |
+
```
|
| 51 |
### Usage
|
| 52 |
|
| 53 |
```python
|