antiven0m commited on
Commit ·
608a64c
1
Parent(s): ea211e5
Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,31 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
---
|
|
|
|
| 1 |
+
## Model Composition
|
| 2 |
+
|
| 3 |
+
- `NurtureAI/neural-chat-7b-v3-16k`: Weight - 30%
|
| 4 |
+
- `xDAN-AI/xDAN-L1-Chat-RL-v1`: Weight - 30%
|
| 5 |
+
- `rwitz/go-bruins-v2`: Weight - 30%
|
| 6 |
+
- `segmed/MedMistral-7B-v0.1`: Weight - 10%
|
| 7 |
+
|
| 8 |
+
### Code Snippet for Model Merging
|
| 9 |
+
The following Python code demonstrates how to create this mixed model using the LM-Cocktail approach:
|
| 10 |
+
|
| 11 |
+
```python
|
| 12 |
+
from LM_Cocktail import mix_models_by_layers
|
| 13 |
+
|
| 14 |
+
model = mix_models_by_layers(
|
| 15 |
+
model_names_or_paths=[
|
| 16 |
+
"NurtureAI/neural-chat-7b-v3-16k",
|
| 17 |
+
"xDAN-AI/xDAN-L1-Chat-RL-v1",
|
| 18 |
+
"rwitz/go-bruins-v2",
|
| 19 |
+
"segmed/MedMistral-7B-v0.1"
|
| 20 |
+
],
|
| 21 |
+
model_type='decoder',
|
| 22 |
+
weights=[0.3, 0.3, 0.3, 0.1],
|
| 23 |
+
output_path='./mixed_llm'
|
| 24 |
+
)
|
| 25 |
+
```
|
| 26 |
+
|
| 27 |
+
```Disclaimer: This project is highly experimental. The chatbot's capabilities and the effectiveness of the model integration are subjects of ongoing research and development.```
|
| 28 |
+
|
| 29 |
---
|
| 30 |
license: apache-2.0
|
| 31 |
---
|