bamec66557
commited on
Update config.yaml
Browse files- config.yaml +26 -26
config.yaml
CHANGED
@@ -1,27 +1,27 @@
|
|
1 |
-
models:
|
2 |
-
- model:
|
3 |
-
#no parameters necessary for base model
|
4 |
-
- model:
|
5 |
-
parameters:
|
6 |
-
density: 1
|
7 |
-
weight: 1
|
8 |
-
merge_method: ties
|
9 |
-
base_model:
|
10 |
-
tokenizer_source: base
|
11 |
-
chat_template: "auto"
|
12 |
-
parameters:
|
13 |
-
normalize: false
|
14 |
-
int8_mask: true
|
15 |
-
dtype: bfloat16
|
16 |
-
regularization:
|
17 |
-
- method: weight_clipping
|
18 |
-
clip_range: [-0.1, 0.1]
|
19 |
-
- method: l2_norm
|
20 |
-
scale: 0.002
|
21 |
-
# Postprocessing to refine the merged model
|
22 |
-
postprocessing:
|
23 |
-
- operation: dynamic_scaling
|
24 |
-
scale_range: [0.9, 1.1]
|
25 |
-
- operation: gaussian_smoothing
|
26 |
-
sigma: 1.0
|
27 |
- operation: normalize
|
|
|
1 |
+
models:
|
2 |
+
- model: Mistral-Nemo-Instruct-2407
|
3 |
+
#no parameters necessary for base model
|
4 |
+
- model: VICIOUS_MESH-12B-BETA
|
5 |
+
parameters:
|
6 |
+
density: 1
|
7 |
+
weight: 1
|
8 |
+
merge_method: ties
|
9 |
+
base_model: Mistral-Nemo-Instruct-2407
|
10 |
+
tokenizer_source: base
|
11 |
+
chat_template: "auto"
|
12 |
+
parameters:
|
13 |
+
normalize: false
|
14 |
+
int8_mask: true
|
15 |
+
dtype: bfloat16
|
16 |
+
regularization:
|
17 |
+
- method: weight_clipping
|
18 |
+
clip_range: [-0.1, 0.1]
|
19 |
+
- method: l2_norm
|
20 |
+
scale: 0.002
|
21 |
+
# Postprocessing to refine the merged model
|
22 |
+
postprocessing:
|
23 |
+
- operation: dynamic_scaling
|
24 |
+
scale_range: [0.9, 1.1]
|
25 |
+
- operation: gaussian_smoothing
|
26 |
+
sigma: 1.0
|
27 |
- operation: normalize
|