birgermoell commited on
Commit
4792cd8
1 Parent(s): a65787c

Upload folder using huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +12 -19
README.md CHANGED
@@ -3,38 +3,31 @@ tags:
3
  - merge
4
  - mergekit
5
  - lazymergekit
6
- - timpal0l/Mistral-7B-v0.1-flashback-v2
7
  - mlabonne/NeuralBeagle14-7B
8
  base_model:
9
- - timpal0l/Mistral-7B-v0.1-flashback-v2
10
  - mlabonne/NeuralBeagle14-7B
11
  ---
12
 
13
  # NeuralBeagle-Flashback
14
 
15
  NeuralBeagle-Flashback is a merge of the following models using [LazyMergekit](https://colab.research.google.com/drive/1obulZ1ROXHjYLn6PPZJwRR6GzgQogxxb?usp=sharing):
16
- * [timpal0l/Mistral-7B-v0.1-flashback-v2](https://huggingface.co/timpal0l/Mistral-7B-v0.1-flashback-v2)
17
  * [mlabonne/NeuralBeagle14-7B](https://huggingface.co/mlabonne/NeuralBeagle14-7B)
18
 
19
  ## 🧩 Configuration
20
 
21
  ```yaml
22
- slices:
23
- - sources:
24
- - model: timpal0l/Mistral-7B-v0.1-flashback-v2
25
- layer_range: [0, 32]
26
- - model: mlabonne/NeuralBeagle14-7B
27
- layer_range: [0, 32]
28
- merge_method: slerp
29
- base_model: timpal0l/Mistral-7B-v0.1-flashback-v2
30
- parameters:
31
- t:
32
- - filter: self_attn
33
- value: [0, 0.5, 0.3, 0.7, 1]
34
- - filter: mlp
35
- value: [1, 0.5, 0.7, 0.3, 0]
36
- - value: 0.5
37
- dtype: bfloat16
38
  ```
39
 
40
  ## 💻 Usage
 
3
  - merge
4
  - mergekit
5
  - lazymergekit
 
6
  - mlabonne/NeuralBeagle14-7B
7
  base_model:
 
8
  - mlabonne/NeuralBeagle14-7B
9
  ---
10
 
11
  # NeuralBeagle-Flashback
12
 
13
  NeuralBeagle-Flashback is a merge of the following models using [LazyMergekit](https://colab.research.google.com/drive/1obulZ1ROXHjYLn6PPZJwRR6GzgQogxxb?usp=sharing):
 
14
  * [mlabonne/NeuralBeagle14-7B](https://huggingface.co/mlabonne/NeuralBeagle14-7B)
15
 
16
  ## 🧩 Configuration
17
 
18
  ```yaml
19
+ models:
20
+ - model: timpal0l/Mistral-7B-v0.1-flashback-v2
21
+ # No parameters necessary for base model
22
+ - model: mlabonne/NeuralBeagle14-7B
23
+ parameters:
24
+ density: 0.53
25
+ weight: 0.6
26
+ merge_method: dare_ties
27
+ base_model: timpal0l/Mistral-7B-v0.1-flashback-v2
28
+ parameters:
29
+ int8_mask: true
30
+ dtype: bfloat16
 
 
 
 
31
  ```
32
 
33
  ## 💻 Usage