Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,27 @@
|
|
1 |
---
|
2 |
license: apache-2.0
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: apache-2.0
|
3 |
---
|
4 |
+
# ColorShadow-7B
|
5 |
+
|
6 |
+
This is a Gradient-SLERP merge between diffnamehard/Mistral-CatMacaroni-slerp-7B and diffnamehard/Mistral-CatMacaroni-slerp-7B performed using mergekit.
|
7 |
+
|
8 |
+
Here is the config file used:
|
9 |
+
|
10 |
+
```
|
11 |
+
slices:
|
12 |
+
- sources:
|
13 |
+
- model: diffnamehard/Mistral-CatMacaroni-slerp-7B
|
14 |
+
layer_range: [0, 32]
|
15 |
+
- model: cookinai/Valkyrie-V1
|
16 |
+
layer_range: [0, 32]
|
17 |
+
merge_method: slerp
|
18 |
+
base_model: diffnamehard/Mistral-CatMacaroni-slerp-7B
|
19 |
+
parameters:
|
20 |
+
t:
|
21 |
+
- filter: self_attn
|
22 |
+
value: [0, 0.5, 0.3, 0.7, 1]
|
23 |
+
- filter: mlp
|
24 |
+
value: [1, 0.5, 0.7, 0.3, 0]
|
25 |
+
- value: 0.5 # fallback for rest of tensors
|
26 |
+
dtype: float16
|
27 |
+
```
|